See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/382323454

Function Spaces in Quaternionic and Clifford Analysis

Preprint · July 2024

CITATIONS READS

0 142

1 author:

Mircea Martin
Baker University
101 PUBLICATIONS 621 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mircea Martin on 17 July 2024.

The user has requested enhancement of the downloaded file.

Function Spaces in Quaternionic and Clif-
ford Analysis
Mircea Martin

Abstract. The first part of the article is a brief report on the history
of quaternionic and Clifford analysis, two research fields that general-
ize single variable complex analysis to higher dimension. Appropriate
counterparts of the Cauchy–Riemann operator and the Cauchy kernel
defined in the frameworks of Hamilton quaternions and Clifford algebras
yield spaces of regular and monogenic functions that extend the concept
of holomorphic functions. Part two introduces Cauchy–Pompeiu and
Bochner–Martinelli–Koppelman integral representation formulas with
remainders for operator–kernel couples consisting of a first order differ-
ential operator on a Euclidean space with coefficients in a unital Banach
algebra, and a smooth homogeneous algebra valued kernel. The general
formulas underscore the role played by Dirac, Cauchy–Riemann, and
Laplace operators in Clifford analysis. The third part of the article is
concerned with sharp estimates of fractional integral transforms in a
Banach algebra setting, and quantitative Hartogs–Rosenthal theorems
on uniform approximation of continuos functions on compact sets of
Euclidean spaces by solutions of operator–kernel couples which, in par-
ticular include regular and monogenic functions.
Keywords: Quaternionic analysis, Clifford analysis, Cauchy–Pompeiu
formulas, Bochner–Martinelli–Koppelman formulas, Quantitave Hartogs–
Rosenthal theorems
2020 Mathematics Subject Classification: 41A30, 44A15, 30E20, 32W05

1. Introduction
The function spaces studied in quaternionic and Clifford analysis are exten-
sions to higher dimension of spaces of complex valued holomorphic functions
in single variable complex analysis. They are introduced by replacing the field
C of complex numbers with the skew field H of Hamilton quaternions, or with
Euclidean Clifford algebras Cn (R), n ≥ 2. The critical part in developing an
appropriate several variable function theory relies on the discovery of special
2 Mircea Martin

classes of functions defined on open subsets of H ≡ R4 , or Rn , n ≥ 2, called

regular and monogenic, respectively.
The definitions of quaternions and of differentiable quaternion valued func-
tions of a quaternion variable were given by W. R. Hamilton in Elements of
Quaternions, 1866, Hamilton [30]. His groundbreaking discoveries did not re-
ceive a wide recognition, and the study of quaternions and of their potential
uses as part of multivariable calculus was minimal for several decades due to
the development of vector calculus.
The spaces of regular quaternion valued functions of a quaternion variable
were defined and investigated by R. Fueter in a series of articles, Fueter
[21–25], 1932–1937. Fueter introduced the classes of left and right regular
functions as left or right solutions of a quaternion extension of the Cauchy–
Riemann operator. As highlights of his work we should mention the general-
izations of Cauchy, Morera, and Liouville theorems. In retrospect, it is hard to
believe that for almost three decades there were only a few new developments
of this noteworthy line of research.
The vindication of Fueter and early developments of Clifford analysis are due
to R. Delange, Delange [16–18], 1970–1972. After 1972, Clifford analysis and
the study of monogenic functions underwent a remarkable growth. The mono-
graph Clifford Analysis, Brackx, Delange, Sommen [20], 1982, is arguably the
best reference for an introduction to Clifford analysis that includes requisites
and a great deal of relevant issues.
The subjects addressed in this article have been selected using objective and
subjective criteria. Section 2 has a historical interest. It provides a review of
the concepts alluded to above of function spaces that generalize the concept
of holomorphic functions to higher dimension. We recall the definitions of
differential operators associated with algebra H, or Euclidean Clifford alge-
bras Cn (R), n ≥ 2, that generalize the Cauchy–Riemann operator, and of the
corresponding several variable substitues for the Cauchy kernel.
Section 3 is concerned with integral representation formulas with remainders
that extend Cauchy–Pompeiu and Bochner–Martinelli–Koppelman formulas.
The generalized formulas are proved for operator–kernel couples (D, Φ) con-
sisting of a first order homogeneous differential operator on Euclidean space
Rm , m ≥ 2, with coefficients in a unital Banach algebra, and a smooth alge-
bra valued kernel homogeneous of degree 1 − m. The operators which yield
the simplest forms of such general formulas are the Euclidean Dirac and
Cauchy–Riemann operators in Clifford analysis, and the standard operators
of exterior differentiation in several variable real or complex analysis.
As a generalization of uniform approximation of continuous functions on com-
pact subsets of C by rational functions, Section 4 addresses the problem of
approximation of continuous functions on compact subsets of Rn , n ≥ 2, by
solutions of operator–kernel couples (D, Φ). The main result is a quantitative
Hartogs–Rosenthal theorem, which has qualitative consequences.
 Function Spaces in Quaternionic and Clifford Analysis 3

Section 5 provides additional information on research projects developed over

the years in the framework of quaternionic and Clifford analysis, and their
applications. The point of view, we deliberately decided to promote in our
article, is to identify relevant defining requirements, single them out, and
accordingly introduce extensions of particular basic concepts. This approach
connects quaternionic and Clifford analysis with other well established areas,
such as harmonic analysis, spin geometry, and multivariable operator theory.
At the same time, close inspections of proofs that rely on a minimal set of
assumptions might be potential sources of further developments.

2. The Search of Regular and Monogenic Functions

Suppose H is a finite dimensional real vector space of dimension n ≥ 2. A
linear basis for H identifies H with Rn , an identification that yields a topo-
logical structure on H. We might also transfer the Euclidean inner product
⟨ ·, · ⟩ and norm | · | from Rn to H. Assume next that X ⊆ H is an open set,
and E is a second finite dimensional real vector space. Recall that a function
U : X → E is continuously differentiable provided there exists a continuous
mapping DU : X → HomR (H, E), where HomR (H, E) is the space of linear
operators from H into E, such that
U (x + t ξ) − U (x)
lim = DU (x)(ξ), x ∈ X, ξ ∈ H.
t→0 t
Mapping DU is the first order derivative of U . If DU is continuously dif-
ferentiable, we take its derivative, DDU = D2 U , referred to as the second
order derivative of U , and U will be called twice continuously differentiable.
A continuation of this process yields the space C ∞ (X, E) of differentiable, or
smooth functions U : X → E that have continuous derivatives of any order.

2.1. Quatenions and Differentiable Functions of a Quaterion Variable

In the particular case when H = E = H, the definition of C ∞ (X, E) reduces
to the concept introduced by Hamilton [30] of differentiable quaternion valued
functions of a quaternion variable. For convenience, we review some basic
requisites. Each quaternion x ∈ H is given by
x = x0 + x1 i + x2 j + x3 k, (x0 , x1 , x2 , x3 ) ∈ R4 , (2.1)
where the quaternion units i, j, k are subject to the Hamiton requirements,
i2 = j2 = k2 = ijk = −1, ij = k = −ji, jk = i = −kj, ki = j = −ik. (2.2)

The conjugate x̄ ∈ H of quaternion x ∈ H as in (2.1) is defined by

x̄ = x0 − x1 i − x2 j − x3 k,
and the structure equations (2.2) imply
xx̄ = x̄x = x20 + x21 + x22 + x23 = |x|2 .
4 Mircea Martin

Component x0 = Re (x) is called the real part of x, and

1
x0 = (x − i x i − j x j − k x k), x ∈ H, (2.3)
4
an observation made by F. Hausdorff [32] as early as 1900. Since, in addition,
x1 = Re (−x i), x2 = Re (−x j), x3 = Re (−x k), we get that all real coordi-
nates x0 , x1 , x2 , x3 of x ∈ H are expressed in terms of x, without using the
conjugate quaternion x̄.
Each function U ∈ C ∞ (X, H), with X ⊆ H an open set, is given by
U = u0 + u1 i + u2 j + u3 k, u0 , u1 , u2 , u3 ∈ C ∞ (X, R). (2.4)
2.2. Regular Functions of a Quaterion Variable
Function spaces C ∞ (X, H) are not substitutes in the quaternion setting for
spaces of holomorphic functions. The next quaternion counterparts of three
defining properties of holomorphic functions are natural attempts to finding
subspaces of C ∞ (X, H) that might turn out to be appropriate.
Definition – H–differentiable functions. Suppose U ∈ C ∞ (X, H), with X ⊆ H
an open set. Function U is called left or right H–differentiable provided the
following limits in H of left or right difference quotients,
DL U (x) = lim ζ −1 [U (x + ζ) − U (x)], x ∈ X, ζ ∈ H \ {0},
ζ→0

DR U (x) = lim [U (x + ζ) − U (x)] ζ −1 , x ∈ X, ζ ∈ H \ {0},

ζ→0
exist for any x ∈ X, respectively.
With regard to this definition, we note that the resulting class of functions
is rather scarce. Actually, G. Scheffers, Scheffers [62], 1893, discovered that
the existence of limits DL U , DR U of left or right difference quotients for the
squaring function U (x) = x2 , x ∈ A, where A is a finite dimensional division
algebra, implies that A is commutative. More to the point, N. M. Krylov,
Krylov [36], 1947, and A. Z. Mejlihzon, Mejlihzon [51], 1948, proved that the
only left or right H–differentiable functions are of the form U (x) = a0 + a1 x,
or U (x) = a0 + x a1 , with a0 , a1 ∈ H, respectively. If both requirements are
imposed, then a1 ∈ R ⊆ H. A different proof that relied on complex analysis
was given by A. Sudbery, Sudbery [67], 1979.
Definittion – H–analytic functions. Function U ∈ C ∞ (X, H) is H–analytic
provided for each a ∈ X there is a ball Bn [a, ε] = {x ∈ H : |x − a| ≤ ε} ⊆ X
of radius ε > 0, such that the restriction of U to Bn [a, ε] is the sum of an
absolutely convergent quaternion power series,
U (x) = c00 +c10 (x−a)c11 +c20 (x−a)c21 (x−a)c22 +c30 (x−a)c31 (x−a)c32 (x−a)c33 + · · · ,
where x ∈ Bn (a, ε) and c00 , c10 , c11 , c20 , c21 , c22 , c30 , c31 , c32 , c33 , · · · , are in H.
Before describing functions that satisfy this definition, we note that quater-
nion coefficients are inserted in the homogeneous terms of quaternion power
series because H is a skew field. This is not a problem, the terms of quater-
nion power series only include the quaternion variable x, hence such series
 Function Spaces in Quaternionic and Clifford Analysis 5

are similar to power series representations of holomorphic functions. How-

ever, Hausdorff identity (2.3) and its three consequences show that the real
coordinates x0 , x1 , x2 , x3 of x ∈ H are all expressible in terms of x. Therefore,
any U ∈ C ∞ (X, H) locally represented by quaternion power series is also lo-
cally represented by power series in four real variables, i.e., U is an H–valued
real analytic function of four real variables. This space is inappropriate, too.
In order to state the third definition, based on equation (2.1) we subsequently
identify quaternions with points x = (x, x1 , x2 , x3 ) ∈ R4 . We next follow
Fueter and introduce the first order differential operator DH with coefficients
in H, called the Cauchy–Riemann–Fueter operator, defined by
DH = ∂0 + ∂1 i + ∂2 j + ∂3 k,
where ∂0 = ∂/∂x0 , ∂1 = ∂/∂x1 , ∂2 = ∂/∂x2 , ∂3 = ∂/∂x3 .
Definition – Regular functions. Function U ∈ C ∞ (X, H) is called left or right
regular provided U is a left or right solution of operator DH , i.e., respectively,
DH U (x) = ∂0 U (x) + ∂1 U (x) i + ∂2 U (x) j + ∂3 U (x) k = 0, x ∈ X, (2.5)

U DH (x) = ∂0 U (x) + i ∂1 U (x) + j ∂2 U (x) + k ∂3 U (x) = 0, x ∈ X. (2.6)

Assuming that function U ∈ C ∞ (X, H) is represented as in (2.4), equations

(2.5), (2.6) reduce to linear systems of differential equations for the real valued
components u0 , u1 , u2 , u3 ∈ C ∞ (X, R).
It turns out that regular functions yield appropriate function spaces for ex-
tending complex analysis in the quaternion setting. The articles by H. Haefli,
Haefli [29], 1947, V. Iftimie, Iftimie [34], 1965, C. A. Deavours, Deavours
[15], 1973, and the already mentioned article by A. Sudbery, Sudbery [67],
1979, contributed over the years to the dissemination of Fueter’s work and
the development of quaternionic analysis.
The quaternion conjugate of differential operator DH is given by
D̄H = ∂0 − ∂1 i − ∂2 j − ∂3 k,
and direct calculations show that DH D̄H = D̄H DH = ∆4 , where
∆4 = ∂02 + ∂12 + ∂22 + ∂32
is the Laplace operator on R4 . Therefore, left or right solutions of DH are
solutions of ∆4 , i.e., regular functions are harmonic. Moreover, operator DH
is elliptic with fundamental solution
1 ξ¯
ΦH (ξ) = 3 · , ξ ∈ H \ {0} ≡ R4 \ {0}, (2.7)
|S | |ξ|4
derived by applying operator D̄H to the fundamental solution of ∆4 , where
|S3 | is the surface area of the unit sphere S3 ⊆ R4 , and |ξ| denotes the norm
of ξ ∈ H ≡ R4 .
6 Mircea Martin

2.3. Euclidean Clifford Algebras, Dirac Operators, and Monogenic Functions

Euclidean Clifford algebra Cn (R), n ≥ 1, is the real unital associative algebra
defined by assuming that R ⊕ Rn ⊆ Cn (R), with unit 1Cn (R) = 1 ∈ R, and
such that the standard orthonormal basis {e1 , e2 , . . . , en } for Rn is a complete
set of generators satisfying the Clifford relations
e2i = −1, 1 ≤ i ≤ n, ei ej + ej ei = 0, 1 ≤ i, j ≤ n, i ̸= j. (2.8)
In particular, C1 (R) = C and C2 (R) = H. Assuming that 0 ≤ p ≤ n, let Ipn
be the collection of p–element subsets I ⊆ {1, 2, . . . , n}. If p = 0, I0n = {∅}.
Each I ∈ Ipn , p ≥ 1, is expressed as an ordered p–tuple
I = (i1 , i2 , . . . , ip ), 1 ≤ i1 < i2 < · · · < ip ≤ n.
Sn
We next assign to I ∈ In = p=0 Ipn the element eI ∈ Cn (R) given by
e∅ = 1, eI = ei1 ei2 · · · eip , I = (i1 , i2 , . . . , ip ), 1 ≤ p ≤ n. (2.9)
The set {eI : I ∈ In } is a linear basis for Cn (R), and Cn (R) has an inner
product ⟨· , ·⟩ with corresponding norm |·| defined by assuming that this basis
is orthonormal.
We identify each a ∈ Cn (R) with the linear operator La : Cn (R) → Cn (R) on
the Hilbert space [ Cn (R), ⟨· , ·⟩ ] defined as left multiplication by a. This one–
to–one representation makes Cn (R) an operator C ∗ –algebra. Therefore, Cn (R)
has an operator norm ∥·∥, too, different from the inner product norm |·|. The
two norms coincide on R ⊕ Rn ⊆ Cn (R). As a C ∗ –algebra, Cn (R) also has an
involution operation ∗, called Clifford conjugation. Since e∗i = −ei , 1 ≤ i ≤ n,
from (2.8), (2.9) we get that each eI ∈ Cn (R), I ∈ In , is a unitary element.
Let σn : Rn → Cn (R) denote the embedding of Rn into Cn (R), i.e.,
σn (ξ) = ξ = ξ1 e1 + ξ2 e2 + · · · + ξn en , ξ = (ξ1 , . . . , ξn ) ∈ Rn , (2.10)
and let
Dn = e1 ∂1 + e2 ∂2 + · · · + en ∂n , ∂i = ∂/∂xi , 1 ≤ i ≤ n, (2.11)
be the first order differential operator with symbol mapping s(Dn ) = σn ,
where x1 , x2 , . . . , xn denote the coordinate functions on Rn , referred to as
the Euclidean Dirac operator on Rn . The domains of Dirac operator Dn are
spaces C ∞ (X, E), with X ⊆ Rn an open set and E a Banach Cn (R)–module.
The actions of Dn on U ∈ C ∞ (X, E) are defined by
Dn U (x) = e1 ∂1 U (x) + e2 ∂2 U (x) + · · · + en ∂n U (x), x ∈ X,

U Dn (x) = ∂1 U (x) e1 + ∂2 U (x) e2 + · · · + ∂n U (x) en , x ∈ X,

according as E is a left or right Cn (R)–module, respectively. If E is a bimodule,
for instance when E = Cn (R), n ≥ 2, Dn U and U Dn might be different.
Definition – Monogenic functions. Function U ∈ C ∞ (X, E) is called left or
right monogenic provided U is a left or right solution of Dirac operator Dn ,
i.e., Dn U (x) = 0, x ∈ X, or U Dn (x) = 0, x ∈ X.
 Function Spaces in Quaternionic and Clifford Analysis 7

The Clifford conjugate D̄n of Dirac operator Dn is D̄n = − Dn , with symbol

mapping σ̄n = − σn . The identity σn (ξ) σ̄n (ξ) = σ̄n (ξ) σn (ξ) = |ξ|2 , ξ ∈ Rn ,
relates Dirac operator Dn with the Laplace operator ∆n on Rn given by
∆n = ∂12 + ∂22 + · · · + ∂n2 . (2.12)
2 n
Since s(∆n )(ξ) = |ξ| , ξ ∈ R , we get
Dn D̄n = D̄n Dn = ∆n . (2.13)
Therefore, monogenic functions are harmonic, and operator Dn , n ≥ 2, is
elliptic. The fundamental solution of Dn , n ≥ 2, called the Euclidean Cauchy
kernel on Rn , is the function Φn : Rn \ {0} → Cn (R) derived by applying D̄n
to the fundamental solution of ∆n , and is given by
1 σ̄n (ξ)
Φn (ξ) = n−1 · , ξ ∈ Rn0 = Rn \ {0}, (2.14)
|S | |ξ|n
where |Sn−1 | is the surface area of the unit sphere Sn−1 ⊆ Rn and |ξ| stands
for the Euclidean norm of ξ ∈ Rn ⊆ Cn (R).
2.4. Euclidean Cauchy–Riemann Operators
Though C1 R) = C ≡ R ⊕ R, Dirac operator D1 is not the Cauchy–Riemann
operator. To get Cauchy–Riemann operators in Clifford analysis we use the
augmented embedding σn,+ : R ⊕ Rn ≡ Rn+1 ⊆ Cn (R), n ≥ 1, i.e.,
σn,+ (ζ) = ζ0 + ζ1 e1 + ζ2 e2 · · · + ζn en , ζ = (ζ0 , ζ1 , . . . , ζn ) ∈ Rn+1 , (2.15)
and introduce the Euclidean Cauchy–Riemann operator Dn,+ with symbol
mapping s(Dn,+ ) = σn,+ defined as
Dn,+ = ∂0 + e1 ∂1 + e2 ∂2 + · · · + en ∂n , ∂i = ∂/∂zi , 0 ≤ i ≤ n, (2.16)
n+1
where z0 , z1 , . . . , zn are the coordinate functions on R . Next, we take the
Clifford conjugates of σn,+ and Dn,+ given by
σ̄n,+ (ζ) = ζ0 − (ζ1 e1 + ζ2 e2 · · · + ζn en ), (2.17)
D̄n,+ = ∂0 − (e1 ∂1 + e2 ∂2 + · · · + en ∂n ). (2.18)
∞
The domains of Dn,+ and D̄n,+ consist of functions U ∈ C (Z, E), where
Z ⊆ Rn+1 is an open set and, as in Subsection 2.3, E is a left or right
Banach Cn (R)–module. Accordingly, we have to distinguish between left or
right actions of operators Dn,+ and D̄n,+ on spaces C ∞ (Z, E).
Since σn,+ (ζ) σ̄n,+ (ζ) = σ̄n,+ (ζ) σn,+ (ζ) = |ζ|2 , ζ ∈ Rn+1 , we get that
Dn,+ D̄n,+ = D̄n,+ Dn,+ = ∆n+1 , (2.19)
where ∆n+1 = ∂02 +∂12 +∂22 +· · ·+∂n2
is the Laplace operator on R . There- n+1

fore, the left and right solutions of Dn,+ or D̄n,+ are harmonic functions, too,
and both operators are elliptic. The fundamental solutions of Dn,+ and D̄n,+
are the Cn (R)–valued functions
1 σ̄n,+ (ζ)
Φn,+ (ζ) = n · , ζ ∈ Rn+1
0 = Rn+1 \ {0}, (2.20)
|S | |ζ|n+1
8 Mircea Martin

1 σn,+ (ζ)
Φ̄n,+ (ζ) = n · , ζ ∈ Rn+1
0 = Rn+1 \ {0}, (2.21)
|S | |ζ|n+1
respectively, with |Sn | and |ζ| the surface area of the unit sphere Sn ⊆ Rn+1
and the norm of ζ ∈ Rn+1 .
We note that quaternionic and Clifford analysis use H = C2 (R) in different
ways to define DH and D2,+ as several variable counterparts of the Cauchy–
Riemann operator. To make another point, let D denote one of the operators
DH , Dn , n ≥ 2, or Dn,+ , n ≥ 1. Using quaternion or Clifford conjugates, we
assigned to D a conjugate operator D̄. At the same time, each such D has a
formal adjoint operator D∗ . It turns out that D∗ = −D̄, hence, in particular,
all Dirac operators Dn , n ≥ 2, are formally self–adjoint.

3. Representation Formulas for Operator–Kernel Couples

Motivated by Section 2, we will next introduce and analyze operator–kernel
couples (D, Φ), which generalize the couples (DH , ΦH ), (Dn , Φn ), n ≥ 2, and
(Dn,+ , Φn,+ ), n ≥ 1, in a Banach algebra setting. The goal is to set up
Cauchy–Pompeiu and Bochner–Martinelli–Koppelman integral representa-
tion formulas with remainders for arbitrary couples (D, Φ), and characterize
the couples that yield remainder–free formulas. Parts of this section are based
on two articles, Deconstructing Dirac Operators II. Integral Representation
Formulas, Martin [47], 2011, and Deconstructing Dirac Operators III. Dirac
and Semi–Dirac Pairs of Differential Operators, Martin [46], 2009.
3.1. The Setting and Main Results
Suppose A is a Banach algebra with unit 1A and let E be a Banach A–module.
Let D : C ∞ (Rm , E) → C ∞ (Rm , E), m ≥ 2, be a first order homogeneous
differential operator with constant coefficients in algebra A given by
D = a1 ∂1 + a2 ∂2 + · · · + am ∂m , ∂i = ∂/∂xi , 1 ≤ i ≤ m. (3.1)
According as E is a left or right A–module, we set
D U = a1 · ∂1 u + a2 · ∂2 u + . . . + am · ∂n u, U ∈ C ∞ (Rm , E),
U D = ∂1 u · a1 + ∂2 u · a2 + . . . + ∂m u · am , U ∈ C ∞ (Rm , E).
We denote the coefficient m–tuple of D by a = (a1 , a2 , . . . , am ) ∈ Am , and
introduce the symbol mapping σ = s(D) of D, σ : Rm → A, given by
σ(ξ) = ξ1 a1 + ξ2 a2 + · · · + ξm am , ξ = (ξ1 , ξ2 , . . . , ξm ) ∈ Rm . (3.2)
Operator D, m–tuple a, and function σ uniquely determine each other.
Next, assume that Φ : Rm
0 → A is a smooth A–valued kernel defined by
1 φ(ξ)
Φ(ξ) = · , ξ ∈ Rm m
0 = R \ {0}, (3.3)
|Sm−1 | |ξ|m
where φ : Rm → A is a smooth function homogeneous of degree 1, i.e.,
φ(t ξ) = t φ(ξ), t ∈ (0, ∞), ξ ∈ Rm .
 Function Spaces in Quaternionic and Clifford Analysis 9

Therefore, kernel Φ is homogeneous of degree 1 − m,

Φ(t ξ) = t1−m Φ(ξ), t ∈ (0, ∞), ξ ∈ Rm . (3.4)
We will refer to (D, Φ) as an operator–kernel couple on Rm with A–valued
parameters (σ, φ). From (3.1) and (3.3), elementary calculations show that
1 1
Φ D(ξ) = · · [ |ξ|2 φ D(ξ) − m φ(ξ) · σ(ξ) ], ξ ∈ Rm
0 , (3.5)
|Sm−1 | |ξ|m+2
1 1
D Φ(ξ) = · · [ |ξ|2 D φ(ξ) − m σ(ξ) · φ(ξ) ], ξ ∈ Rm
0 . (3.6)
|Sm−1 | |ξ|m+2
As additional links between σ, φ, or D, Φ, we define µR , µL ∈ A by
Z Z
1
µR = m−1 φ(ξ) · σ(ξ) darea(ξ) = Φ(ξ) · σ(ξ) darea(ξ), (3.7)
|S | Sm−1 Sm−1
Z Z
1
µL = m−1 σ(ξ) · φ(ξ) darea(ξ) = σ(ξ) · Φ(ξ) darea(ξ), (3.8)
|S | Sm−1 Sm−1
where darea is the surface area measure on the unit sphere Sm−1 . We refer
to µR , µL as the right and left spherical means of φ with weight σ, or the
right and left spherical means of operator–kernel couple (D, Φ).
Let Λ ⊆ Rm be a bounded open set with a smooth and oriented boundary Σ.
Assuming that E is a left A–module, associated with D, Φ, Λ, Σ we introduce
four integral operators,
PΛ , RR, Λ , RL, Λ , CΣ : C ∞ (Rm , E) → C ∞ (Rm \ Σ, E),
defined by Z
PΛ U (x) = p.v. Φ(ξ − x) · U (ξ) dvol(ξ), (3.9)
Λ
Z
RR, Λ U (x) = p.v. Φ D(ξ − x) · U (ξ) dvol(ξ), (3.10)
Λ
Z
RL, Λ U (x) = p.v. D Φ(ξ − x) · U (ξ) dvol(ξ), (3.11)
Λ
Z
CΣ U (x) = Φ(ξ − x) σ(ν(ξ)) · U (ξ) darea(ξ), (3.12)
Σ
for U ∈ C ∞ (Rm , E) and x ∈ Rm \ Σ, where dvol is the volume measure on Λ,
p.v. stands for principal value, darea is the surface area measure on Σ, and
ν(ξ) = (ν1 (ξ), ν2 (ξ), . . . , νm (ξ)) ∈ Rm (3.13)
denotes the unit outer normal vector to Σ at points ξ ∈ Σ. We decided to
employ skewed convolution operators due to custom, a slight adjustment that
yields familiar forms of integral representation formulas.
The fact that operators PΛ , RR, Λ , RL, Λ transform smooth functions on Rm
into smooth functions on Rm \ Σ is a consequence of Calderon–Zygmund
theory. For details, we refer to Stein [64, 65] and Tarkhanov [68].
10 Mircea Martin

Subsequently, we will also use operator TΛ : C ∞ (Rm , E) → C ∞ (Rm \ Σ, E),

the truncation operator associated with Λ, defined by
TΛ U (x) = χΛ (x) U (x), U ∈ C ∞ (Rm , E), x ∈ Rm \ Σ, (3.14)
where χΛ is the characteristic function of set Λ.
We continue assuming that (D, Φ) is an operator–kernel couple and Λ ⊆ Rm
is an arbitrary bounded open set with a smooth oriented boundary Σ.
Theorem A — Integral Representation Formulas with Remainders.
(i) If U ∈ C ∞ (Rm , E) and x ∈ Rm \ Σ, then
µR · TΛ U (x) = CΣ U (x) − PΛ DU (x) − RR,Λ U (x).
∞
(ii) If U ∈ C (Rm , E) and x ∈ Rm \ Σ, then
(µR + µL ) · TΛ U (x) = CΣ U (x) − PΛ DU (x) − DPΛ U (x) − (RR,Λ + RL,Λ )U (x).

Theorem A has two consequences, which characterize couples (D, Φ) with

standard Cauchy–Pompeiu or Bochner–Martinelli–Koppelman formulas.
Corollary 1. The following statements are equivalent.
(i) If (Λ, Σ) is arbitrary, U ∈ C ∞ (Rm , E), and x ∈ Rn \ Σ, then
TΛ U (x) = CΣ U (x) − PΛ DU (x).
(ii) The parameter couple (σ, φ) has the next two properties,
Z
φ(ξ) · σ(ξ) darea(ξ) = |Sm−1 | 1A ,
Sm−1
2
|ξ| φD(ξ) − m φ(ξ) · σ(ξ) = 0, ξ ∈ Rm .
Corollary 2. The following statements are equivalent.
(i) If (Λ, Σ) is arbitrary, U ∈ C ∞ (Rm , E), and x ∈ Rm \ Σ, then
TΛ U (x) = CΣ U (x) − PΛ DU (x) − DPΛ U (x).
(ii) The parameter couple (σ, φ) has the next two properties,
Z
[φ(ξ) · σ(ξ) + σ(ξ) · φ(ξ)] darea(ξ) = |Sm−1 | 1A ,
Sm−1
2
|ξ| [φD(ξ) + Dφ(ξ)] − m [φ(ξ) · σ(ξ) + σ(ξ) · φ(ξ)] = 0, ξ ∈ Rm .
The proofs of Corollaries 1, 2 are straightforward. By equations (3.7) and
(3.5), part (ii) in Corollary 1 is equivalent to µR = 1A and Φ D = 0, i.e.,
RR,Λ = 0 for any (Λ, Σ), hence, by statement (i) in Theorem A, part (ii)
implies part (i). In a similar way, by equations (3.7), (3.8) and (3.5), (3.6), we
note that part (ii) in Corollary 2 reduces to µR +µL = 1A and Φ D+D Φ = 0,
i.e., RR,Λ + RL,Λ = 0 for any (Λ, Σ), which by statement (ii) in Theorem A
lead to part (i). The implications from part (i) to part (ii) in both corollaries
are elementary.
Definition – Cauchy–Pompeiu and Bochner–Martinelli–Koppelman couples.
Couple (D, Φ) is called a Cauchy–Pompeiu or Bochner–Martinelli–Koppelman
 Function Spaces in Quaternionic and Clifford Analysis 11

operator–kernel couple provided it has remainder–free integral representation

formulas as stated in the previous two corollaries, respectively.

3.2. Proof of Theorem A

Let a = (a1 , a2 , · · · , am ) ∈ Am be the coefficient m–tuple of operator D as in
equation (3.1). Let ⋆ dξi , 1 ≤ i ≤ m, be the (m − 1)–form on Rm defined by
⋆ dξi = (−1)i−1 dξ1 ∧ · · · ∧ dξi−1 ∧ dξi+1 ∧ · · · ∧ dξm ,
where ξi , 1 ≤ i ≤ m, are the coordinate functions on Rm , and introduce the
following A–valued form ω = ω(a) on Rm ,
ω = a1 ⋆ dξ1 + a2 ⋆ dξ2 + · · · + an ⋆ dξm . (3.15)
Next, assume that Ω ⊆ Rm is a compact oriented smooth submanifold of
dimension m with boundary ∂Ω. Suppose Θ ∈ C ∞ (Ω, A) and U ∈ C ∞ (Rm , E)
are arbitrary, and observe that the exterior derivative of the E–valued smooth
(m − 1)–form Θ · ω · U on Ω is given by
d(Θ · ω · U ) = (Θ · DU + ΘD · U ) dξ,
where dξ = dξ1 ∧ dξ2 ∧ · · · ∧ dξm . Accordingly, Stokes Theorem implies
Z Z
Θ · ω · U = (Θ · DU + ΘD · U ) dξ. (3.16)
∂Ω Ω
The two integrals of E–valued forms reduce to integrals relative to area
measure darea on ∂Ω and volume measure dvol on Ω, respectively. We let
ν(ξ) = (ν1 (ξ), ν2 (ξ), · · · , νm (ξ)) ∈ Rm be the unit outer normal vector to ∂Ω
at ξ ∈ ∂Ω, and since ⋆ dξi | ∂Ω ≡ νi (·) darea, 1 ≤ i ≤ m, we get
Z Z
Θ·ω·U = Θ(ξ) · σ(ν(ξ)) · U (ξ) darea(ξ). (3.17)
∂Ω ∂Ω
For the integral on the right hand side of (3.16) we obviously have
Z Z
(Θ · Du + ΘD · U ) dξ = [ Θ(ξ) · Du(ξ) + ΘD(ξ) · U (ξ) ] dvol(ξ). (3.18)
Ω Ω
Proof of Statement (i). Under the assumptions in statement (i), first suppose
that x ∈ Rn \ Λ. Introduce the compact oriented manifold Ω = Λ ∪ Σ and
define Θ ∈ C ∞ (Ω, A) as
Θ(ξ) = Φ(ξ − x), ξ ∈ Ω. (3.19)
Since ∂Ω = Σ, by (3.17) and (3.12) we observe that
Z
Θ · ω · U = CΣ U (x).
∂Ω
At the same time, from (3.18), (3.9), (3.10) we have
Z
(Θ · Du + ΘD · U ) dξ = PΛ Du(x) + RR,Λ U (x).
Ω
Therefore, by (3.16) we get that CΣ U (x) = PΛ Du(x) + RR,Λ U (x), an equa-
tion that proves (3.7) because TΛ U (x) = 0.
12 Mircea Martin

Assume next that x ∈ Λ and choose ε > 0 such that Bm [x, ε] ⊆ Λ, where
Bm [x, ε] ⊆ Rm is the closed ball of center x and radius ε. Let Ω ⊆ Rm be
the closure of Λ \ Bm [x, ε]. Its oriented boundary ∂Ω consists of Σ with the
positive orientation, and sphere Sm−1 [x, ε] ⊆ Rm of center x and radius ε
with the negative orientation. We define Θ ∈ C ∞ (Ω, A) as in (3.19), and
from (3.16), (3.17), (3.18) we get
Z Z
Φ(ξ−x)·σ(ν(ξ))·U (ξ) darea(ξ)− Φ(ξ−x)·σ(ν(ξ))·U (ξ) darea(ξ)
Σ Sm−1 [x,ε]

Z Z
= Φ(ξ − x) · DU (ξ) dvol(ξ) + ΦD(ξ − x) · U (ξ) dvol(ξ).
Λ\Bm [x,ε] Λ\Bn [x,ε]

Since kernel Φ is homogeneous of degree 1 − m, using the change of variable

ξ = x + ε η, η ∈ Sm−1 , we have
Z Z
Φ(ξ−x)·σ(ν(ξ))·U (ξ) darea(ξ) = Φ(η)·σ(η)·U (x+ε η) darea(η),
Sm−1 [x,ε] Sm−1

and by (3.7) we get

Z
lim Φ(ξ − x) · σ(ν(ξ)) · U (ξ) darea(ξ) = µR · U (x). (3.20)
ε↓0 Sm−1 (x,ε)

Since obviously
Z
lim Φ(ξ − x) · DU (ξ) dvol(ξ) = PΛ Du(x),
ε↓0 Λ\Bm (x,ε)

Z
lim ΦD(ξ − x) · U (ξ) dvol(ξ) = RR,Λ U (x),
ε↓0 Λ\Bm (x,ε)

the proof of statement (i) is complete.

Proof of Statement (ii). We only need to show that
µL · TΛ U (x) = −DPΛ U (x) − RL,Λ U (x), (3.21)
for U ∈ C ∞ (Rm , E) and x ∈ Rm \ Σ. We rely on a general result proved in
Tarkhanov [68], which in our setting yields the equations
Z
Di PΛ U (x) = −p.v. Di Φ(ξ − x) · U (ξ) dvol(ξ) − µi · TΛ U (x), (3.22)
Λ

where
Z Z
1
µi = ξi φ(ξ) darea(ξ) = ξi Φ(ξ) darea(ξ), (3.23)
|Sm−1 | Sm−1 Sm−1

for any 1 ≤ i ≤ m. To get (3.21), we multiply (3.22), (3.23) by ai , 1 ≤ i ≤ m,

add the resulting equations, and just use the definitions of D, RL,Λ , and µL .
The proof of Theorem A is complete. □
 Function Spaces in Quaternionic and Clifford Analysis 13

3.3. Dirac and Semi–Dirac pairs of Differential Operators

In addition to operator D and its symbol mapping σ : Rn → A defined by
equations (3.1), (3.2), let σ † : Rn → A be a second linear mapping with
coefficient m–tuple a† = (a†1 , a†2 , . . . , a†m ) ∈ Am ,
σ † (ξ) = a†1 ξ1 + a†2 ξ2 + · · · + a†m ξm , ξ = (ξ1 , ξ2 , . . . , ξm ) ∈ Rm ,
and let D† be the associated differential operator with symbol s(D† ) = σ † ,
D† = a†1 ∂1 + a†2 ∂2 + · · · + a†m ∂m .
We next introduce the homogeneous kernels Φ, Φ† : Rm
0 → A by setting
1 σ † (ξ) 1 σ(ξ)
Φ(ξ) = · , Φ† (ξ) = · , ξ ∈ Rm
0 .
|Sm−1 | |ξ|m |Sm−1 | |ξ|m
Specifically, the operator–kernel couples (D, Φ) and (D† , Φ† ) are associated
with parameteres (σ, σ † ) and (σ † , σ), respectively. Corollaries 1, 2 in Subsec-
tion 3.2 have the following counterparts.
Corollary 3. Let (D, Φ) and (D† , Φ† ) be the two previously defined couples.
(i) (D, Φ) and (D† , Φ† ) are Cauchy–Pompeiu couples only if σ, σ † satisfy the
symbol equations
σ(ξ) · σ † (ξ) = σ † (ξ) · σ(ξ) = |ξ|2 1A , ξ ∈ Rm . (3.24)

(ii) (D, Φ) and (D† , Φ† ) are Bochner–Martinelli–Koppelman couples only if

σ, σ † satisfy the symbol equation
σ(ξ) · σ † (ξ)] + σ † (ξ) · σ(ξ) = |ξ|2 1A , ξ ∈ Rm . (3.25)

Direct calculations convert (3.24), (3.25) into systems of equations for the
coefficients (a1 , a2 , · · · , am ) and (a†1 , a†2 , . . . , a†m ) of σ and σ † . The equations
also relate operator pairs (D, D† ) with the Laplace operator ∆m on Rm .
Regarding D, D† , ∆m as operators on C ∞ (Rm , E), where E is an A–module,
since the symbol mapping s(∆m ) : Rm → A of ∆m is s(∆m )(ξ) = |ξ|2 1A , ξ ∈
Rm , Corollary 3 leads to the next result.
Corollary 4. Let (D, Φ) and (D† , Φ† ) be the two previously defined couples.
(i) (D, Φ) and (D† , Φ† ) are Cauchy–Pompeiu couples only if D, D† satisfy
the operator equations D D† = D† D = ∆m .
(ii) (D, Φ) and (D† , Φ† ) are Bochner–Martinelli–Koppelman couples only if
D, D† satisfy the operator equation D D† + D† D = ∆m .
Definition – Dirac and semi–Dirac pairs. Pairs (D, D† ) with symbol mappings
σ, σ † satisfying the symbol or operator equations in part (i) or part (ii) of
Corollaries 3, 4 are called Dirac and semi–Dirac pairs, respectively.
Relying on earlier remarks, we note that quaternionic and Clifford analysis
provide three distinct families of Dirac pairs (D, D† ). Specifically, suppose D
is either DH , Dn , n ≥ 2, or Dn,+ , n ≥ 1, and let D† = −D∗ , where D∗ is the
formal adjoint of D. Section 2 justifies our claim about (D, D† ).
14 Mircea Martin

Semi–Dirac pairs, related to Dirac operators, are provided by operators of

exterior differentiation acting on spaces of smooth differential forms in several
⋆
real or complex variables. Consider first the exterior graded algebra Em (R)
m ⋆ ⋆
of differential forms on R , m ≥ 2, and let dm : Em (R) → Em (R) be the
operator of exterior differentiation, with formal adjoint d∗m : Em
⋆
(R) → Em⋆
(R).
Standard calculations yield the next equation,
dm d∗m + d∗m dm = −∆m , (3.26)
which implies that ( dm , −d∗m ) is a semi–Dirac pair. In this framework, we
also get a representation of Dirac operator Dm , namely, Dm ∼ = dm + d∗m .
⋆⋆
Consider next the bigraded complex algebra Em (C) of complex valued dif-
m
ferential forms on C , m ≥ 1, and let ∂m : Em¯ ⋆⋆
(C) → Em⋆⋆
(C) be the
operator of exterior anti–holomorphic differentiation, with formal adjoint
∂¯m
∗ ⋆⋆
: Em ⋆⋆
(C) → Em (C). Similar to equation (3.26), we have
∂¯m ∂¯m
∗
+ ∂¯m
∗ ¯
∂m = −□m , (3.27)
where □m stands for the complex Laplace operator on Cm defined by
□m = 2 ∂ 2 /∂ζ1 ∂ ζ̄1 + ∂ 2 /∂ζ2 ∂ ζ̄2 + · · · + ∂ 2 /∂ζm ∂ ζ̄m , m ≥ 1,


√ √
and ζi = ξ2i−1 + −1 ξ2i , ζ̄i = ξ2i−1 − −1 ξ2i , 1 ≤ i ≤ m, are the complex
and complex conjugate cooordinate functions on Cm ≡ R2m . Switching from
complex to real coordinates we get
∆2m = 2 □m , m ≥ 1. (3.28)
√ √ ∗
By (3.27), (3.28) we conclude ¯ ¯
that ( 2 ∂m , − 2 ∂m ) is a semi–Dirac pair
√
and, in addition, D2m ∼
= 2 ( ∂¯m + ∂¯m
∗
).
Previous definitions imply the existence of Bochner–Martinelli–Koppelman
√ √ ∗
formulas for each component of the semi–Dirac pair ( 2 ∂¯m , − 2 ∂¯m ). The
specific one corresponding to ∂¯m , m ≥ 2, derived from Theorem A, is just
a non–standard form of the classical formula in several variable complex
analysis. Our approach circumvented the rather intricate traditional setting,
⋆⋆
without using the bigraded complex algebra Em (C), or relying on the funda-
mental solution of the complex Laplace operator □m on Cm . Nevertheless,
such concepts are significant parts of the background, and the monograph by
Aizenberg, Dautov [2] illustrates their relevance and uses.

4. Hartogs–Rosenthal Theorems for Operator–Kernel Couples

Suppose that Ω ⊆ C is a compact set, let C(Ω) be the uniform algebra of
complex valued continuous functions on Ω, and let R(Ω) denote the closure
in C(Ω) of the subalgebra of rational functions on Ω, i.e., restrictions to Ω
of rational functions with poles in C \ Ω. The following classical result on
rational approximation was discovered by Hartogs, Rosenthal [31], 1931.
Hartogs–Rosenthal Theorem. If area (Ω) = 0, then R(Ω) = C(Ω).
 Function Spaces in Quaternionic and Clifford Analysis 15

Several applications of Hartogs–Rosenthal theorem related to the theory of

uniform algebras are presented in Gamelin [25], 1969. This section refines
an approach developed in the articles Higher–Dimensional Ahlfors–Beurling
Inequalities in Clifford Analysis, Martin [40], 1998, and Deconstructing Dirac
Operators I. Quantitative Hartogs–Rosenthal Theorems, Martin [45], 2009.
4.1. Three Inequalities in Single Variable Complex Analysis
The next three inequalities provide quantitative estimates in the same frame-
work as Hartogs–Rosenthal theorem, and motivate in part the main themes
addressed in this section.
Ahlfors–Beurling Inequality. Suppose Ω ⊆ C is a compact set. Then,
Z  1/2
1 darea(ζ) 1
≤ · area (Ω) , z ∈ C. (4.1)
π Ω ζ −z π
Alexander Inequality. The distance in C(Ω) from the complex conjugate co-
ordinate function ζ̄ on C to R(Ω) satisfies the upper estimate
 1/2
1
distC(Ω) [ ζ̄, R(Ω) ] ≤ · area (Ω) . (4.2)
π

Putinar Inequality. Suppose u : C → [0, ∞] and u ∈ L1 (C) ∩ L∞ (C). Then,

Z  1/2
1 u(ζ) darea(ζ) 1
≤ · ∥u∥∞ · ∥u∥1 , z ∈ C. (4.3)
π C ζ −z π
Each inequality makes it possible to derive Hartogs–Rosenthal theorem as a
qualitative consequence. For instance, inequality (4.2) shows that ζ̄ ∈ R(Ω)
when area (Ω) = 0, whence we get that R(Ω) = C(Ω) by relying on Stone–
Weierstrass theorem. For proofs, we refer to Ahlfors, Beurling [1], 1950,
Alexander [3, 4], 1973, and Putinar [53], 1988. Perhaps the reader should
note the rather long time intervals that separate Hartogs–Rosenthal theorem
and the three previous inequalities, from 1931 to 1950, 1973, 1988.
4.2. L∞ –Estimates for Convolution Operators with Homogeneous Kernels
Our next goal is to generalize Ahlfors–Beurling, Putinar, and Alexander in-
equalities. The objects of interest are extended counterparts of Riesz potential
operators in Euclidean harmonic analysis. As in Section 3, suppose A is a uni-
tal Banach algebra, E a Banach A–module, and let C0 (Rn , E), n ≥ 1, be the
A–module of continuous compactly supported functions U : Rn → E. The
norms on A and E are denoted by ∥ · ∥. On C0 (Rn , E) we use the norms
Z
∥U ∥∞ = sup { ∥U (x)∥ : x ∈ Rn }, ∥U ∥1 = ∥U (x)∥ dvol(x),
Rn
∞ n 1 n
and let L (R , E) and L (R , E) be the spaces obtained by completing
C0 (Rn , E) relative to these norms, respectively. Assume next that Φ : Rn0 → A
is a continuous function homogeneous of degree α − n, 0 < α < n, i.e.,
Φ(t η) = tα−n Φ(η), t ∈ (0, ∞), η ∈ Rn0 = Rn \ {0}. (4.4)
16 Mircea Martin

The skewed convolution operator P = P(Φ) with kernel Φ transforms func-

tions U ∈ C0 (Rn , E) into continous functions from Rn to E defined by
Z
P U (x) = Φ(ξ − x) · U (ξ) dvol(ξ), x ∈ Rn , (4.5)
Rn
and has an extension from L1 (Rn , E) ∩ L∞ (Rn , E) to L∞ (Rn , E). Associated
with Φ we introduce the compact set BnΦ = {η ∈ Rn0 : ∥Φ(η)∥ ≥ 1} ∪ {0} and
the constant
Γ(Φ) = n/α · vol ( BnΦ ) 1−α/n . (4.6)
⋆ ⋆
In addition, assuming that A is the dual space of A and θ ∈ A , we define
the set BnΦ,θ = {η ∈ Rn0 : θ ◦ Φ(η) ≥ 1} ∪ {0} and a second constant,
Γ+ (Φ) = n/α · sup { vol ( BnΦ,θ ) 1−α/n : θ ∈ A⋆ , ∥θ∥ = 1}. (4.7)
We are now in a position to state the main result in this subsection.
Theorem B — L∞ – Estimates for Operator P = P(Φ).
(i) If U ∈ L1 (Rn , E) ∩ L∞ (Rn , E), then
α/n
∥P U ∥∞ ≤ Γ(Φ) · ∥U ∥1−α/n
∞ · ∥U ∥1 . (4.8)

(ii) Suppose U : Rn → E is defined as U ( · ) = u( · ) e, where e ∈ E, ∥e∥ = 1,

and u : Rn → [0, ∞), u ∈ L1 (Rn ) ∩ L∞ (Rn ). Then
α/n
∥P u( · ) e∥∞ ≤ Γ+ (Φ) · ∥u∥∞
1−α/n
· ∥u∥1 . (4.9)
In particular, if e = 1A , the unit of A, and u = χΩ is the characteristic
function of a compact set Ω ⊆ Rn , then
∥P χΩ ( · ) 1A ∥∞ ≤ Γ+ (Φ) · vol (Ω)α/n . (4.10)
Statement (ii) provides generalizations of Ahlfors–Beurling and Putinar in-
equalities in a Banach algebra setting for a rather large class of kernels. Even-
tually, Theorem B will be used to generalize Alexander inequality, too, by
assuming that Φ is the kernel component of an operator–kernel couple (D, Φ)
subject to some requirements. For instance, we may require D : C ∞ (Rn , E) →
C ∞ (Rn , E), n ≥ 2, to be an elliptic constant coefficient differential operator
homogeneous of order α, and let Φ : Rn0 → A be its fundamental solution.
4.3. Proof of Theorem B
For convenience, we divide the proof into two parts.
Part I — We first prove a scalar form of Theorem B when A = E = R. We
assume that the kernel is a continuos non–negative function ρ : Rn0 → [0, ∞),
homogeneous of degree α − n, and define the associated fractional integral
operator P(ρ) : L1 (Rn ) ∩ L∞ (Rn ) → L∞ (Rn ) as in (4.5), i.e.,
Z
P(ρ) u(x) = ρ(ξ − x) u(ξ) dvol(ξ), x ∈ Rn , u ∈ L1 (Rn ) ∩ L∞ (Rn ).
Rn
Next, we introduce the compact sets
Bnρ [x, t] = {ξ ∈ Rn : ξ ̸= x, ρ(ξ − x) ≥ t} ∪ {x}, x ∈ Rn , t ∈ (0, ∞),
 Function Spaces in Quaternionic and Clifford Analysis 17

and the function

µ : Rn × (0, ∞) → (0, ∞), µ(x, t) = vol (Bnρ [x, t]), x ∈ Rn , t ∈ (0, ∞).
Since Bnρ [x, t] = x + Bnρ [0, t], and because ρ is homogeneous of degree α − n,
we get that Bnρ [0, t] = t1/(α−n) Bnρ [0, 1], hence
µ(x, t) = γ(ρ) t−κ , γ(ρ) = vol (Bnρ [0, 1]), κ = n/(n − α). (4.11)
We refer to κ = n/(n − α) as the exponent of operator P(ρ). Under the
previous assumptions, we have the following result.
Lemma — Sharp L∞ –Estimates of Fractional Integral Transforms. Suppose
that u : Rn → [0, ∞], with u ∈ L1 (Rn ) ∩ L∞ (Rn ). Then
1−1/κ
∥P(ρ) u∥∞ ≤ Γ(ρ) · ∥u∥1/κ
∞ · ∥u∥1 , Γ(ρ) = n/α · γ(ρ)1/κ . (4.12)
Moreover, inequality (4.12) is sharp.
Proof. Associated with u : Rn → [0, ∞] we define the non–increasing function
Z
µu : Rn × (0, ∞) → (0, ∞), µu (x, t) = u(ξ) dvol (ξ), (4.13)
Bn
ρ [x,t]

and observe that by (4.11) we have

µu (x, t) ≤ ∥u∥∞ · µ(x, t) = γ(ρ) · ∥u∥∞ · t−κ , x ∈ Rn , t ∈ (0, ∞). (4.14)
n
Next, suppose t0 ∈ (0, ∞) and introduce ρinn , ρout : R → [0, ∞] given by
ρinn (ξ) = ρ(ξ) − t0 if ξ ∈ Bnρ [0, t0 ], ρinn (ξ) = 0 if ξ ∈ Rn \ Bnρ [0, t0 ],
ρout (ξ) = t0 if ξ ∈ Bnρ [0, t0 ], ρout (ξ) = ρ(ξ) if ξ ∈ Rn \ Bnρ [0, t0 ].
Let Pinn and Pout be the integral operators with kernels ρinn and ρout . Since
ρ = ρinn + ρinn , we get
P(ρ) u(x) = Pinn u(x) + Pout u(x), x ∈ Rn . (4.15)
n
We claim that Pinn u(x) and Pout u(x), where x ∈ R is arbitrary, satisfy
γ(ρ) · ∥u∥∞ −κ+1
Pinn u(x) ≤ · t0 , (4.16)
κ−1
Pout u(x) ≤ ∥u∥1 · t0 . (4.17)
To prove (4.16) we use the non–increasing function µu defined in (4.13) and
express Pinn u(x) as a Stieltjes integral,
Z Z ∞
Pinn u(x) = ( ρ(ξ − x) − t0 ) u(ξ) dvol(ξ) = − (t − t0 ) dµu (x, t).
Bn
ρ [x,t0 ] t0

By (4.14), lim (t − t0 ) µu (x, t) = 0, and integration by parts implies

t→∞
Z ∞
Pinn u(x) = µu (x, t) dt, (4.18)
t0
which in conjunction with (4.14) leads to
Z ∞
γ(ρ) · ∥u∥∞ −κ+1
Pinn u(x) ≤ γ(ρ) · ∥u∥∞ t−κ dt = · t0 .
t0 κ−1
18 Mircea Martin

With regard to inequality (4.17), we just note that

Pout u(x) ≤ ∥ρout ∥∞ · ∥u∥1 ,
and since ∥ρout ∥∞ = t0 , this estimate reduces to (4.17). Assembling (4.15),
(4.16), (4.17) we end up with
γ(ρ) · ∥u∥∞ −κ+1
Pu(x) ≤ · t0 + ∥u∥1 · t0 .
κ−1
The value of t0 that minimizes the right hand side is given by
 1/κ
γ(ρ) · ∥u∥∞
t∗ = .
∥u∥1
Direct computations show that t∗ yields inequality (4.12) with
Γ(ρ) = n/α · γ(ρ)1/κ . (4.19)
We still need to check that inequality (4.12) is sharp. Let u be the character-
istic function of Bnρ [0, 1], and note that by (4.11) we have
Z Z ∞
P(ρ) u(0) = ρ(ξ) dvol (ξ) = − t dµ(0, t)
Bn
ρ [0,1] 1
Z ∞
= κ · γ(ρ) t−κ dt = κ/(κ − 1) · γ(ρ) = n/α · γ(ρ).
1
On the other hand, since ∥u∥∞ = 1 and ∥u∥1 = γ(ρ), by (4.19) we get
α/n
Γ(ρ) · ∥u∥1−α/n
∞ · ∥u∥1 = n/α · γ(ρ)(n−α)/n · γ(ρ)α/n = n/α · γ(ρ). (4.20)
1−α/n α/n
Consequently, ∥P(ρ) u∥∞ ≥ P(ρ) u(0) = Γ(ρ)·∥u∥∞ ·∥u∥1 , an estimate
that singles out u as an extremal function. The poof is complete. □
Part II — Both statements in Theorem B follow from the previous lemma.
Proof of Statement (i). Suppose that kernel Φ : Rn0 → A is continuous, ho-
mogeneous of degree α − n, 0 < α < n, and U ∈ L1 (Rn , E) ∩ L∞ (Rn , E). Let
ρ : Rn0 → [0, ∞) and u : Rn → [0, ∞] be defined as
ρ(ξ) = ∥Φ(ξ)∥, ξ ∈ Rn0 , u(x) = ∥U (x)∥, x ∈ Rn .
Estimate (4.8) in statement (i) follows from inequality (4.12) because
∥P U ∥∞ ≤ ∥P(ρ) u∥∞ , Γ(Φ) = Γ(ρ), ∥U ∥∞ = ∥u∥∞ , ∥U ∥1 = ∥u∥1 .
Proof of Statement (ii). With e ∈ E, ∥e∥ = 1, and u : Rn → [0, ∞) as in
statement (ii), we obviously have
∥P u( · ) e∥∞ = ∥P u( · ) 1A · e∥∞ ≤ ∥P u( · ) 1A ∥∞ .
Therefore, it would be enough to prove (4.9) for U ( · ) = u( · ) 1A . Let x ∈ Rn
be a fixed point and select θ ∈ A⋆ with ∥θ∥ = 1, θ(1A ) = 1, such that
Z
∥P U (x)∥ = θ( P U (x)) = θ ◦ Φ(ξ − x) u(ξ) dvol(ξ). (4.21)
Rn
 Function Spaces in Quaternionic and Clifford Analysis 19

Next, define the kernel ρ : Rn0 → [0, ∞) by ρ(ξ) = max { θ ◦ Φ(ξ), 0 }, ξ ∈ Rn0 .
Since θ( P U (x)) ≥ 0, based on inequality (4.12) we get that
Z
α/n
∥P U (x)∥ ≤ ρ(ξ − x) u(ξ) dvol(ξ) = P(ρ) u(x) ≤ Γ(ρ) · ∥u∥1−α/n
∞ · ∥u∥1 .
Rn

By (4.7) we obviously have Γ(ρ) ≤ Γ+ (Φ), hence the last inequality yields the
1−α/n α/n
poinwise estimate ∥P U (x)∥ ≤ Γ+ (Φ) · ∥u∥∞ · ∥u∥1 , in which the right
hand side is independent of the particular point x ∈ Rn . This observation
clearly implies (4.9) and (4.10). The proof of Theorem B is concluded. □

4.4. Quantitative Hartogs–Rosenthal Theorems

We return to the setting described in Subsection 3.1. Specifically, we let A be
a unital Banach algebra, E a left Banach A–module, and assume that (D, Φ)
is an operator–kernel couple on Rm , m ≥ 2. Suppose Ω ⊆ Rm is a compact
set, let C(Ω, E) be the standard Banach space of continuous functions from
Ω to E, and define the associated solution space S(Ω, E | D) ⊆ C(Ω, E) as the
closure in C(Ω, E) of the subspace formed by restrictions to Ω of functions
U ∈ C ∞ (Λ, E) such that D U (x) = 0, x ∈ Λ, where the domains Λ are open
neighborhood of Ω which dpendend on function U .
Theorem C – Upper Distance Estimates.
Let distC(Ω,E) [ U, S(Ω, E | D) ] denote the distance in C(Ω, E) from the restric-
tion to Ω of a function U ∈ C ∞ (Rm , E) to the solution space S(Ω, E | D).
Assume that either
(i) (D, Φ) is a Cauchy–Pompeiu couple with the additional property DΦ = 0,
or that
(ii) (D, Φ) is a Bochner–Martinelli–Koppelman couple and ai aj · E = {0} for
any coefficients ai , aj ∈ A, 1 ≤ i, j ≤ m, of differential operator D.
Each requirement implies the upper distance estimate
1/m
distC(Ω,E) [ U, S(Ω, E | D) ] ≤ Γ(Φ) · ∥ χΩ D U ∥1−1/m
∞ · ∥ χΩ D U ∥1 , (4.22)
where Γ(Φ) is defined by equation (4.6).
In particular, if U ∈ C ∞ (Rm , E) is a linear function, then D U ∈ E and
distC(Ω,E) [ U, S(Ω, E | D) ] ≤ Γ+ (Φ) · ∥ D U ∥ · vol (Ω)1/m , (4.23)
where Γ+ (Φ) is given by equation (4.7).
Proof. Under the assumptions in Theorem C, we claim that
distC(Ω,E) [ U, S(Ω, E | D) ] ≤ sup { ∥P(Φ) χΛ DU (x)∥ : x ∈ Ω }, (4.24)
where Λ ⊆ Rm is an arbitrary bounded open neighborhood of Ω with a
smooth oriented boundary Σ. To prove (4.24), we note that by Corollary 1
of Theorem A requirement (i) implies
U (x) = CΣ U (x) − P(Φ) χΛ DU (x), x ∈ Λ.
20 Mircea Martin

From assumption DΦ = 0, by equation (3.12) we get DCΣ U (x) = 0, x ∈ Λ,

hence U0 = CΣ U |Ω ∈ S(Ω, E | D) ]. Therefore, (4.24) follows from
distC(Ω,E) [ U, S(Ω, E | D) ] ≤ ∥ U − U0 ∥Ω,∞ = ∥ P(Φ) χΛ DU ∥Ω,∞ , (4.25)
where ∥ · ∥Ω,∞ denotes the norm on C(Ω, E). In their turn, requirement (ii)
and Corollary 2 of Theorem A yield the decomposition
U (x) = CΣ U (x) − P(Φ) χΛ DU (x) − DPΛ U (x), x ∈ Λ.
and by equations (3.12), (3.9) from the special assumption on E we have
DCΣ U (x) = 0, DDPΛ U (x), x ∈ Λ.
Consequently, U0 = CΣ U |Ω − DPΛ U |Ω ∈ S(Ω, E | D) ]. The new function U0
satisfies (4.25), too, and thus we get estimate (4.24).
Next, we select a family Λk , k ≥ 1, of bounded open neighborhoods of
m
T∞R with smooth oriented boundaries, such that Λk+1 ⊆ Λk , k ≥
Ω in
1, k=1 Λk = Ω, use (4.24) for each Λk , k ≥ 1, and then by taking the
limit as k → ∞ we end up with
distC(Ω,E) [ U, S(Ω, E | D) ] ≤ ∥ P(Φ)χΩ D U ∥∞ . (4.26)
It remains to observe that estimates (4.22), (4.23) follow from (4.26) and
statements (i), (ii) in Theorem B. The proof of Theorem C is complete. □
Theorem C generalizes Alexander inequality in a Banach algebra setting, and
leads to the following qualitative Hartogs–Rosenthal theorem.
Corollary 5. Suppose that Ω ⊆ Rm is a compact set with vol (Ω) = 0. If
(D, Φ) and E satisfy requirements (i) or (ii), then S(Ω, E | D) = C(Ω, E).
Direct proofs for the particular operators Dn , n ≥ 2, and ∂¯m , m ≥ 2, could
be found in Brackx, Delanghe, Sommen [11] and Dautov [14]. Using more
ellaborate techniques, Tarkhanov [69] derived qualitative Hartogs–Rosenthal
theorems for elliptic differential operators D acting on spaces of smooth sec-
tions of inner product smooth vector bundles, under the assumption that D
has a fundamental solution Φ. Sharp quantitative counterparts for such cou-
ples (D, Φ) are proved in Martin, Szeptycki [49, 50]. Theorem C in several
variable complex analysis for operator ∂¯m , m ≥ 2, and spaces of ∂¯m –closed
forms of type (p, m − 1), 0 ≤ p ≤ m, on Cm , with best values of the constants
associated with the homogeneous form–valued components of the Bochner–
Martinelli–Koppelman kernel was proved in Martin [44].

4.5. Lower Distance Estimates

We conclude Section 4 with a result that complements Theorem C.
Theorem D – Lower Distance Estimates.
Let D : C ∞ (Rm , E) → C ∞ (Rm , E), m ≥ 2, be a first order constant coeffi-
cient differential operator with symbol mapping σ = s(D) : Rm → A, ∥σ∥ = 1.
Let Ω ⊆ Rm be a smooth compact oriented submanifold of dimension m with
boundary Σ.
 Function Spaces in Quaternionic and Clifford Analysis 21

(i) If U ∈ C ∞ (Rm , E), then

Z
1
distC(Ω,E) [ U, S(Ω, E | D) ] ≥ · D U (ξ) dvol(ξ) , (4.27)
|Σ| Ω
where |Σ| stands for the area of Σ.
(ii) If U : Rm → E is a linear function, then
|Ω|
distC(Ω,E) [ U, S(Ω, E | D) ] ≥ · ∥D U ∥, (4.28)
|Σ|
where |Ω| denotes the volume of Ω.
Proof. The assumptions on Ω allow us to use the space C ∞ (Ω, E) as a domain
of D, and the dense subspace of S(Ω, E | D) of functions U0 ∈ C ∞ (Ω, E) with
property D U0 = 0. Select such a function U0 , assume that U ∈ C ∞ (Rm , E) is
arbitrary, and introduce the E–valued (m−1)–form ω ·(U −U0 ) on Ω where ω
is the A–valued form associated with operator D defined by equation (3.15).
As in Subsection 3.2, note that d(ω · (U − U0 )) = D U , apply Stokes theorem,
and get that
Z Z
σ(ν(ξ)) · (U (ξ) − U0 (ξ)) darea(ξ) = D U (ξ) dvol(ξ).
Σ Ω
Therefore, since ∥σ∥ = 1 and ∥ν(ξ)∥ = 1 for each ξ ∈ Σ, we have
Z Z
D U (ξ) dvol(ξ) = σ(ν(ξ)) · (U (ξ) − U0 (ξ)) darea(ξ)
Ω Σ
Z
≤ ∥σ(ν(ξ))∥ · ∥U (ξ) − U0 (ξ)∥ darea(ξ) ≤ |Σ| · ∥U − U0 ∥Ω, ∞ .
Σ
Statement (i) follows by observing that
inf { ∥U − U0 ∥Ω, ∞ : U0 ∈ C ∞ (Ω, E), D U0 = 0 } = distC(Ω,E) [ U, S(Ω, E | D) ],
and statement (ii) is obvious. The proof of Theorem D is complete. □
Corollary 6. Let (D, Φ) be an operator–kernel couple satisfying requirements
(i) or (ii) in Theorem C, with D ∈ PD1 (Rm , A, E), m ≥ 2, In addition, let
Ω ⊆ Rm be a smooth compact oriented submanifold of dimension m with
boundary Σ. Then,
|Ω|(m−1)/m
≤ Γ+ (Φ), (4.29)
|Σ|
where Γ+ (Φ) is given by equation (4.7).
The proof is straightforward. We just combine the lower and upper distance
estimates (4.28), (4.23) assuming that U : Rm → E is a linear function such
that ∥DU ∥ = 1, and get the next chain of inequalities,
|Ω|
≤ distC(Ω,E) [ U, S(Ω, E | D) ] ≤ Γ+ (Φ) · |Ω|1/m .
|Σ|

The two sides of inequality (4.29) are independent of each other, a rather
intriguing feature. In the framework of Clifford analysis, the natural choice
22 Mircea Martin

of (D, Φ) is the Dirac operator–kernel couple (Dm , Φm ) defined in Section 2.

To determine the value of Γ+ (Φm ), we recall that by equation (2.14) we have
1 σ̄m (η) 1 η
Φm (η) = m−1 · m
= − m−1 · m , η ∈ Rm
0 . (4.30)
|S | |η| |S | |η|
We next note that related to definition (4.7) of Γ+ (Φm ) it would be enough to
use functionals θ ∈ Cm (R)⋆ given by θ( · ) = ⟨ · , τ ⟩, with ⟨ · , · ⟩ the standard
inner product on Cm (R) and τ ∈ Rm ⊆ Cm (R) a unit vector. For each such θ
we have to determine the volume of Bm m
Φm ,θ = {η ∈ R0 : θ ◦ Φ(η) ≥ 1} ∪ {0},
an algebraic compact set with the explicit description
n ⟨ η, τ ⟩ o
BmΦm ,θ = η ∈ R m
: − m−1 ≥ |η|m
. (4.31)
|S |
Since the volumes of these sets are independent of vector τ, we may take τ =
−e1 , use a parametrization of the resulting set, and get the value of Γ+ (Φm )
by setting up and evaluating an integral. The simplest case m = 2 deserves
a special mention. The set we need is the disc given by { η = (η1 , η2 ) ∈
R2 : (2π)−1 η1 ≥ η12 + η22 } whose area equals (16π)−1 , and by using (4.7) we
conclude that
π 1/2 |B2 |1/2
Γ+ (Φ2 ) = 2 · (16π)−1/2 = = ,
2π |S1 |
where B2 ⊆ R2 is the unit disc. Corrolary 6 reduces to the next result.
Isoperimetric Inequality. For any compact set Ω ⊆ R2 with a smooth oriented
boundary Σ, the area |Ω| of Ω and the length |Σ| of Σ satisfy the inequality

|Ω|1/2 |B2 |1/2

≤ .
|Σ| |S1 |

5. Comments and References

Quaternionic and Clifford analysis reached their current status due to work
done by many authors. To assist the reader in further explorations of re-
search projects that illustrate the scope of past and current developments,
this section will briefly comment on a few contributions related to themes
addressed in our article, with references. For convenience, no previously men-
tioned references are included. As highly recommended additional sources of
information we refer to Anglès [5], Chevalley [12], Colombo, Sabadini, Som-
men, Struppa [13], Delanghe, Sommen, Souček [20], Gilbert, Murray [26],
Gürlebeck, Sprössig [27], Kravchenko, Shapiro [35], Louenesto [38], Mitrea
[52], Rocha–Chavez, Shapiro, Sommen [57], and the lists of references therein.
The articles due to Bernstein [7], Bory–Reyes, Pérez–Regalado, Shapiro [9],
Bory–Reyes, González–Cervantes, Sabadini [10], Delange, Krausshar, Mal-
onek [19], Gustafsson, Khavinson [28], Hile [33], Malonek [39], Martin [41],
Rocha–Chavez, Shapiro, Sommen [54–56], Ryan [58–61], Shapiro [63], Som-
men [66], Vasilevski, Shapiro [70] are also connected with issues in our article.
 Function Spaces in Quaternionic and Clifford Analysis 23

Clifford analysis has natural extension in the framework of spin geometry.

The objects of interest are smooth vector bundles over real or complex man-
ifolds, equipped with linear connections and geometric structures that make
it possible to introduce special kinds of Laplace operators and explicitlry con-
struct their square roots, which by definition are refered to as Dirac operators.
The imposed requirements prove critical in setting up Bochner–Weitzenböck
and Bochner–Kodaira–Nakano curvature identities connecting two such Dirac
operators, and eventually deriving, or recovering, significant topological con-
sequences in global differential geometry, based on the method discovered by
Bochner [8]. The monographs by Berline, Getzler, Vergne [6] and Lawson,
Michelsohn [37] serve as excellent references. A complete analyis of all lin-
ear connections on inner product vector bundles compatible with prescribed
Clifford structures is included in Martin [48].
As a final comment, without ellaborating on details, it seems aproppriate to
note that Clifford analysis and spin geometry concepts have counterparts in
multivariable operator theory. Specific results that make a point centered on
the study of seminormal systems of Hilbert space operators are reported in
Martin [42, 43], and some proofs rely on techniques presented in this article.

References
[1] Ahlfors, L., Beurling, A.: Conformal invariants and function theoretic null sets.
Acta Math. 83, 101–129 (1950)
[2] Aizenberg, L. A., Dautov, Sh. A.: Differential Forms Orthogonal to Holomor-
phic Functions or Forms and Their Properties. Transl. Math. Monographs, vol.
56. Amer. Math. Soc. (1983)
[3] Alexander, H.: Projections of polynomial hulls. J. Funct. Anal. 13, 13–19 (1973)
[4] Alexander, H.: On the area of the spectrum of an element of a uniform algebra.
In: Complex Approximation Proceedings, pp. 3–12, Quebec, 3–8 July 1978.
Birkhäuser, Basel (1980)
[5] Anglès, P.: Conformal Groups in Geometry and Spin Structures. Progress in
Mathematical Physics, vol. 50. Birkhäuser, Boston (2008)
[6] Berline, N., Getzler, E., Vergne, M.: Heat kernels and Dirac operators.
Grundlehren der mathematischen Wissenschaften, vol. 298. Springer, Berlin
(1992)
[7] Bernstein, S.: A Borel–Pompeiu formula in Cn and its applications to inverse
scattering theory. In: Clifford Algebras and Their Applications in Mathemat-
ical Physics. Progress in Mathematical Physics Series, vol. 19, pp. 117–185.
Birkhäuser, Boston (2000)
[8] Bochner, S.: Curvature and Betti numbers I, II. Ann. of Math. 49, 50, 379–390,
79–93 (1948, 1949)
[9] Bory–Reyes, J., Pérez–Regalado, C.O., Shapiro, M.: Cauchy Type Integral
in Bicomplex Setting and Its Properties. Complex Anal. Oper. Theory 13,
2541–2573 (2019)
[10] Bory–Reyes, J., González–Cervantes, J. O., Sabadini, I. : A Borel–Pompeiu
formula in a (q, q‘)–model of quaternionic analysis. Preprint (2024)
24 Mircea Martin

[11] Brackx, F., Delanghe, R., Sommen, F.: Clifford Analysis. Pitman Research
Notes in Mathematics Series, vol. 76. Pitman, Massachusetts (1982)
[12] Chevalley, C.: The Algebraic Theory of Spinors. Springer, Berlin (1997)
[13] Colombo, F., Sabadini, I., Sommen, F., Struppa, D.C.: Analysis of Dirac Sys-
tems and Computational Algebra. Progress in Mathematical Physics, vol. 39.
Birkhäuser, Boston (2004)
¯
[14] Dautov, Sh. A.: On ∂–closed forms of type (p, n−1) as analogues of holomorphic
functions of one complex variable. Holomorphic Functions of Several Complex
Variables, Inst. Fiz. SO Akad. Nauk SSSR, 21–36 (1972)
[15] Deavours, C.A.: The quaternion calculus. Am. Math. Mon. 80, 995–1008 (1973)
[16] Delanghe, R.: On regular–analytic functions with values in a Clifford algebra.
Math. Ann. 185, 91–111 (1970)
[17] Delanghe, R.: On regular points and Liouville’s theorem for functions with
values in a Clifford algebra. Simon Stevin 44, 55–66 (1970–1971)
[18] Delanghe, R.: On the singularities of functions with values in a Clifford algebra.
Math. Ann. 196, 293–319 (1972)
[19] Delanghe, R., Krausshar, R.S., Malonek, H.R.: Differentiability of functions
with values in some real associative algebras: Approaches to an old problem.
Bulletin de la Société Royale des Sciences de Liège 70(4–6), 231–249 (2001)
[20] Delanghe, R., Sommen, F., Souček, V.: Clifford Algebra and Spinor–Valued
Functions. Kluwer, Dordrecht (1982)
[21] Fueter, R.: Analytische Funktionen einer Quaternionenvariablen. Comment.
Math. Helv. 4, 9–20 (1932)
[22] Fueter, R.: Die Funktionentheorie der Differentialgleichungen und mit vier
reellen Variablen. Comment. Math. Helv. 7, 307–330 (1935)
[23] Fueter, R.: Über die analytische Darstellung der regulären Funktionen einer
Quaternionen variablen. Comment. Math. Helv. 8, 371–378 (1936)
[24] Fueter, R.: Die Singularitäten der eindeutigen regulären Funktionen einer
Quaternionen-variablen. Comment. Math. Helv. 9, 320–335 (1937)
[25] Gamelin, T.W.: Uniform Algebras. Prentice Hall, Englewood Cliffs (1969)
[26] Gilbert, J. E., Murray, M. A. M.: Clifford Algebras and Dirac Operators in
Harmonic Analysis. Cambridge Studies in Advanced Mathematics, vol. 26.
Cambridge University Press, Cambridge (1991)
[27] Gürlebeck, K., Sprössig, W.: Quaternionic and Clifford Calculus for Physicists
and Engineers. Wiley, Chichester (1997)
[28] Gustafsson, B., Khavinson, D.: On approximation by harmonic vector fields.
Houston J. Math. 20, 75–92 (1994)
[29] Haefeli, H.: Hyperkomplexe differentiale. Comment. Math. Helv. 20, 382–420
(1947)
[30] Hamilton, W.R.: Elements of Quaternions. Longmans Green, London (1866)
[31] Hartogs, F., Rosenthal, A.: Über Folgen analytischer Functionen. Math. Ann.
104, 606–610 (1931)
[32] Hausdorff, F.: Zür Theorie der Systeme complexer Zahlen. Leipz. Ber. 52, 43–61
(1900)
[33] Hile, G.N.: Representations of solutions of a special class of first order systems.
J. Differ. Equ. 25, 410–424 (1977)
 Function Spaces in Quaternionic and Clifford Analysis 25

[34] Iftimie, V.: Fonctions hypercomplexes. Bull. Math. Soc. Sci. Math RSR 57,
279–332 (1965)
[35] Kravchenko, V. V., Shapiro, M. V.: Integral Representations for Spatial Models
of Mathematical Physics. Pitman Research Notes in Mathematics Series, vol.
351. Longman, Harlow (1996)
[36] Krylov, N. M.: On Rowan Hamilton’s quaternions and the notion of mono-
genicity (in Russian). Dokl. Akad. Nauk SSSR 55, 799–800 (1947)
[37] Lawson, H. B., Michelsohn, M–L.: Spin Geometry. Princeton Mathematical
Series, vol. 38, Princeton University Press, Priceton (1989)
[38] Louenesto, P.: Clifford Algebras and Spinors. Cambridge University Press,
Cambridge (2001)
[39] Malonek, H.R.: Representations of solutions of a special class of first order
systems. A new hypercomplex structure of the Euclidean space and the concept
of hypercomplex differentiability. Complex Var. 14, 25–33 (1990)
[40] Martin, M.: Higher–dimensional Ahlfors–Beurling inequalities in Clifford anal-
ysis. Proc. Am. Math. Soc. 126, 2863–2871 (1998)
[41] Martin, M.: Convolution and maximal operator inequalities. In: Clifford Alge-
bras and Their Applications in Mathematical Physics. Progress in Mathemat-
ical Physics, vol. 19, pp. 83–100. Birkhäuser, Boston (2000)
[42] Martin, M.: Spin geometry, Clifford analysis, and joint seminormality. In: Ad-
vances in Analysis and Geeometry. New Developments Using Clifford Algebras.
Trends in Mathematics Series, vol. 14, pp. 227–255. Birkhäuser, Boston (2004)
[43] Martin, M.: Uniform approximation by solutions of elliptic equations and semi-
normality in higher dimensions. Operator Theory: Advances and Applications,
vol. 149, pp. 387–406. Birkhäuser, Basel (2004)
[44] Martin, M.: Uniform approximation by closed forms in several complex vari-
ables. Adv. Appl. Clifford Algebras 19(3–4), 777–792 (2009)
[45] Martin, M.: Deconstructing Dirac operators. I: Quantitative Hartogs-Rosenthal
theorems. In: More Progress in Analysis. Proceedings of the Fifth International
Society for Analysis, Its Applications and Computation Congress, ISAAC 2005,
pp. 1065–1074. World Scientific, Singapore (2009)
[46] Martin, M.: Deconstructing Dirac operators. III: Dirac and semi-Dirac pairs.
In: Operator Theory: Advances and Applications, vol. 203, pp. 347–362.
Birkhäuser, Basel (2009)
[47] Martin, M.: Deconstructing Dirac operators. II: Integral representation for-
mulas. In: Hypercomplex Analysis and Applications. Trends in Mathematics.
Proceedings of the Seventh International Society for Analysis, Its Applications
and Computation Congress, ISAAC 2009, London, pp. 195–211. Springer, Basel
(2011)
[48] Martin, M.: Derivations and linear connections on Clifford vector bundles. Adv.
Appl. Clifford Algebras 27, 2585–2605 (2017)
[49] Martin, M., Szeptycki, P.: Sharp inequalities for convolution operators with
homogeneous kernels and applications. Indiana Univ. Math. J. 46, 975–988
(1997)
[50] Martin, M., Szeptycki, P.: Integral transforms controlled by maximal func-
tions. In: Operator Theory: Advances and Applications, vol. 153, pp. 169–184,
Birkhäuser, Basel (2004)
26 Mircea Martin

[51] Mejlihzon, A.Z.: On the notion of monogenic quaternions (in Russian). Dokl.
Akad. Nauk SSSR 59, 431–434 (1948)
[52] Mitrea, M.: Singular Integrals, Hardy Spaces, and Clifford Wavelets. Lecture
Notes in Mathematics, vol. 1575, Springer, Berlin (1994)
[53] Putinar, M.: Extreme hyponormal operators.In: Operator Theory: Advances
and Applications, vol. 28, pp. 249–265. Birkhäuser, Basel (1988)
[54] Rocha–Chavez, R., Shapiro M., Sommen, F.: On the singular Bochner–
Martinelli integral. Integr. Equ. Oper. Theory 32, 354–365 (1998)
[55] Rocha–Chavez, R., Shapiro M., Sommen, F.: Analysis of functions and dif-
ferential forms in Cm . In: Proceedings of the Second International Society
for Analysis, Its Applications and Computation Congress, ISAAC 1999, pp.
1457–1506. Kluwer, Dordrecht (2000)
[56] Rocha–Chavez, R., Shapiro M., Sommen, F.: Integral theorems for solutions
of the complex Hodge-Dolbeault system. In: Proceedings of the Second In-
ternational Society for Analysis, Its Applications and Computation Congress,
ISAAC 1999, pp. 1507–1514. Kluwer, Dordrecht (2000)
[57] Rocha–Chavez, R., Shapiro M., Sommen, F.: Integral Theorems for Functions
and Differential Forms in Cm . Research Notes in Mathematics, vol. 428. Chap-
man & Hall, Boca Raton (2002)
[58] Ryan, J.: Applications of complex Clifford analysis to the study of solutions to
generalized Dirac and Klein–Gordon equations, with holomorphic potential. J.
Differ. Equ. 67, 295–329 (1987)
[59] Ryan, J.: Cells of harmonicity and generalized Cauchy integral formulae. Proc.
Lond. Math. Soc. 60, 295–318 (1990)
[60] Ryan, J.: Plemelj formulae and transformations associated to plane wave de-
compositions in complex Clifford analysis. Proc. Lond. Math. Soc. 64, 70–94
(1991)
[61] Ryan, J.: Intrinsic Dirac operators in Cn . Adv. Math. 118, 99–133 (1996)
[62] Scheffers, G.: Verallgemeinerung der Grundlagen der gewöhnlichen complexen
Zahlen. Berichte kgl. Sächs. Ges. der Wiss. 52, 60 (1893)
[63] Shapiro, M.: Some remarks on generalizations of the one–dimensional complex
analysis: hypercomplex approach. In: Functional Analytic Methods in Complex
Analysis and Applications to Partial Differential Equations, pp. 379–401. World
Scientific, Singapore (1995)
[64] Stein, E.M.: Singular Integrals and Differentiability Properties of Functions.
Princeton University Press, Princeton (1970)
[65] Stein, E.M.: Harmonic Analysis: Real-Variable Methods, Orthogonality, and
Oscillatory Integrals. Princeton University Press, Princeton (1993)
[66] Sommen, F.: Martinelli–Bochner formulae in complex Clifford analysis.
Zeitschrift für Analysis und ihre Anwendungen 6, 75–82 (1987)
[67] Sudbery, A.: Quaternionic analysis. Math. Proc. Camb. Philos. Soc. 85,
199–225 (1979)
[68] Tarkhanov, N. N.: The Cauchy Problem for Solutions of Elliptic Equations.
Akademie, Berlin (1995)
[69] Tarkhanov, N. N.: The Analysis of Solutions of Elliptic Equations. Mathematics
and Its Applications, vol. 406, Kluwer Academic Publishers (1997)
 Function Spaces in Quaternionic and Clifford Analysis 27

[70] Vasilevski, N., Shapiro, M.: Some questions of hypercomplex analysis. In: Com-
plex Analysis and Applications, Sofia, 1987, pp. 523–531 (1989)

Mircea Martin
Department of Mathematics
Baker University
Baldwin City, KS 66006, USA
e-mail: [email protected]

View publication stats