American Journal of Mathematical Analysis, 2020, Vol. 8, No.

1, 14-30
Available online at http://pubs.sciepub.com/ajma/8/1/3
Published by Science and Education Publishing
DOI:10.12691/ajma-8-1-3

Essentially Adequate Concept of Holomorphic Functions

in Quaternionic Analysis
Michael Parfenov*

Bashkortostan Branch of Russian Academy of Engineering, Ufa, Russia

*Corresponding author: [email protected]
Received May 25, 2020; Revised June 26, 2020; Accepted July 05, 2020
Abstract The so-called essentially adequate concept of quaternionic holomorphic (ℍ -holomorphic) functions
defined as functions, whose quaternionic derivatives are independent of "the way of their computation", is developed.
It is established that ℍ -holomorphic functions form one remarkable class of quaternionic functions whose properties
are fully similar (essentially adequate) to complex ones: the quaternionic multiplication of these quaternionic
functions behaves as commutative, the left quotient equals the right one, the rules for differentiating sums, products,
ratios, inverses, and compositions are the same as in complex analysis. One can just verify these properties,
constructing ℍ -holomorphic functions from their complex holomorphic counterparts by using the presented
constructing rule. Several examples, confirming the theory in question, are considered. When using this concept
there are no principal restrictions to build a quaternionic analysis similar to complex one. The elementary source
flow and elementary vortex flow, allowing us to construct different 3D steady state fluid flows by superposition, are
considered. To automate the processing of ℍ -holomorphic functions the pack of Mathematica® Programs is
developed, part of which is presented.
Keywords: quaternionic holomorphic functions, quaternionic analysis, quaternionic generalization of Cauchy-
Riemann’s equations, rules for quaternionic differentiation, quaternionic potential, 3D steady state fluid flows
Cite This Article: Michael Parfenov, “Essentially Adequate Concept of Holomorphic Functions
in Quaternionic Analysis.” American Journal of Mathematical Analysis, vol. 8, no. 1 (2020): 14-30.
doi: 10.12691/ajma-8-1-3.

and 𝑖𝑖, 𝑗𝑗, 𝑘𝑘 are quaternionic basis vectors in quaternion

space ℍ.
1. Introduction The quaternionic functions are, respectively, denoted by

Looking for a quaternionic generalization of the ψ ( p ) ψ 1 ( x, y , z , u ) + ψ 2 ( x, y , z , u ) i

=
complex theory [1] seems to be a natural and important +ψ 3 ( x, y, z , u ) j + ψ 4 ( x, y, z , u ) k ∈ ,
issue in a development of mathematical analysis [2,3].
The known "left" or "right" definition of a quaternionic where 𝜓𝜓1 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢) , 𝜓𝜓2 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢) , 𝜓𝜓3 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢) and
derivative (see, e.g. references in [4]) is incomplete (or 𝜓𝜓4 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢) are real-valued functions of real variables
non-essentially adequate), since each of underlying 𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢 in the so-called ℝ4 -representation (see, e.g. [7]).
algebras, viz.: with only the "left" or only the "right" In the Cayley-Dickson doubling form [4,6,8] we have
multiplication does not represent [8] all arbitrary rotations � 𝑏𝑏�� + 𝜙𝜙2 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎,
𝜓𝜓(𝑝𝑝) = 𝜙𝜙1 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � 𝑏𝑏�� ∙ 𝑗𝑗, (1.2)
of vectors in 3D space. Some attempts to unite both
approaches were made, however they were more like where
heuristic rather than systematic (see, e.g. [3,5]). 𝜙𝜙1 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � � = 𝜓𝜓1 + 𝜓𝜓2 𝑖𝑖,
� 𝑏𝑏,
Unlike these attempts, the so-called essentially adequate
concept of the quaternionic differentiability (holomorphy), 𝜙𝜙2 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � � = 𝜓𝜓3 + 𝜓𝜓4 𝑖𝑖.
� 𝑏𝑏,
uniting the left and the right approaches [4,6,7], is
For simplicity, we denote the functions 𝜙𝜙1 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎, ��
� 𝑏𝑏,
complete (or essentially adequate to properties of 3D
space). and 𝜙𝜙2 �𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � � (in the so-called ℂ -representation [7])
� 𝑏𝑏, 2

We denote an independent quaternionic variable by by 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) and 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) or simpler by 𝜙𝜙1 and 𝜙𝜙2 . Upon
the transition to the complex plane 𝑥𝑥 + 𝑧𝑧𝑧𝑧 denoted by 𝜉𝜉
𝑝𝑝 = 𝑥𝑥 + 𝑦𝑦𝑦𝑦 + 𝑧𝑧𝑧𝑧 + 𝑢𝑢𝑢𝑢 = 𝑎𝑎 + 𝑏𝑏𝑏𝑏 ∈ ℍ,
the Cayley-Dickson doubling form becomes the following
where 𝑥𝑥, 𝑦𝑦, 𝑧𝑧, 𝑢𝑢 are real values, the values [6]:
𝑎𝑎 = 𝑥𝑥 + 𝑦𝑦𝑦𝑦, 𝑏𝑏 = 𝑧𝑧 + 𝑢𝑢𝑢𝑢 (1.1) 𝜓𝜓𝐶𝐶 (𝜉𝜉) = 𝜓𝜓1 (𝑥𝑥, 𝑧𝑧) + 𝜓𝜓3 (𝑥𝑥, 𝑧𝑧) ∙ 𝑗𝑗.
are complex constituents of the quaternion representation In accordance with the concept of essentially adequate
in the so-called Cayley-Dickson doubling form [8], differentiability [6] the definition of a derivative is based
 American Journal of Mathematical Analysis 15

on the main idea, viz.: each point of any real line is at the Essentially adequate conditions (1.3) of ℍ-holomorphy
same time a point of some plane and space as a whole, and differ from the known left and right Cauchy - Riemann -
therefore any characterization of differentiability at a point Fueter conditions [2,3], which we can characterize as non-
must be the same regardless of whether we think of that essentially adequate [6].
point as a point on the real axis or a point in the complex Now we recall the following theorems and their
plane, or a point in three-dimensional space. corollary needed for the sequel [6].
Given this, we have to define a quaternionic derivative Theorem 1.3. Let a complex function 𝜓𝜓𝐶𝐶 (𝜉𝜉): 𝐺𝐺2 → ℂ be
∆ψ ℂ -holomorphic everywhere in a connected open set
as a limit of a difference quotient as ∆𝑝𝑝 tends to 0 𝐺𝐺2 ⊆ ℂ, except, possibly, at certain singularities. Then a
∆p
ℍ-holomorphic function 𝜓𝜓𝐻𝐻 (𝑝𝑝) of the same kind as 𝜓𝜓𝐶𝐶 (𝜉𝜉)
[6]. At that the limit is required to be independent not only
can be constructed (without change of a functional
of directions to approach ∆𝑝𝑝 = 0 (as in complex analysis),
dependence form) from 𝜓𝜓𝐶𝐶 (𝜉𝜉) by replacing a complex
but also of the manner of quaternionic division: on the left
variable 𝜉𝜉 ∈ 𝐺𝐺2 (as a single whole) in an expression for
or on the right. Such an independence is called [6] the
𝜓𝜓𝐶𝐶 (𝜉𝜉) by a quaternionic variable 𝑝𝑝 ∈ 𝐺𝐺4 ⊆ ℍ , where
"independence of the way of computation".
𝐺𝐺4 is defined (except, possibly, at certain singularities) by
This requirement corresponds to the requirement usual
the relation 𝐺𝐺4 ⊃ 𝐺𝐺2 in the sense that 𝐺𝐺2 exactly follows
for analysis: the derivative must be unambiguous. At that
an adequate representation of a physical 3D field strength, from 𝐺𝐺4 upon transition from 𝑝𝑝 to 𝜉𝜉.
which is usually described by an unambiguous derivative, Theorem 1.4. It is assumed that a quaternionic function
is achieved [11]. 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝜙𝜙1 + 𝜙𝜙2 𝑗𝑗 ,where 𝜙𝜙1 and 𝜙𝜙2 are differentiable
Definition 1.1 A quaternionic function is said to be the with respect to 𝑎𝑎, 𝑎𝑎, 𝑏𝑏, 𝑏𝑏, is ℍ-holomorphic everywhere in
essentially adequate quaternionic holomorphic function its domain of definition 𝐺𝐺4 ⊆ ℍ. Then its full (uniting the
(𝔼𝔼𝔼𝔼ℍ- holomorphic or, briefly, ℍ-holomorphic function) left and right derivatives) quaternionic derivative, defined
at a point 𝑝𝑝 , if it has a quaternionic derivative by
independent of a way of its computation in some open (′ ) (′ )
𝜓𝜓𝐻𝐻 (𝑝𝑝)′ = 𝜙𝜙1 + 𝜙𝜙2 𝑗𝑗,
connected neighborhood 𝐺𝐺4 ∈ ℍ of a point 𝑝𝑝 ∈ ℍ.
This definition leads to the definition of the necessary where
and sufficient conditions for 𝜓𝜓(𝑝𝑝) to be ℍ-holomorphic (′ )
𝜙𝜙1 = 𝜕𝜕𝑎𝑎 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) + 𝜕𝜕𝑎𝑎 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) = 𝜕𝜕𝑎𝑎,𝑎𝑎 𝜙𝜙1 ,
[6].
Definition 1.2 Suppose that the constituents 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) and (′ )
𝜙𝜙2 = 𝜕𝜕𝑎𝑎 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) + 𝜕𝜕𝑎𝑎 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) = 𝜕𝜕𝑎𝑎,𝑎𝑎 𝜙𝜙2 ,
𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) of a quaternionic function 𝜓𝜓(𝑝𝑝) = 𝜓𝜓(𝑎𝑎, 𝑏𝑏) =
𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) + 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗 possess the continuous first-order is ℍ -holomorphic in 𝐺𝐺4 as well, except, possibly, at
certain singularities. If a quaternion function 𝜓𝜓(𝑝𝑝) is once
partial derivatives with respect to 𝑎𝑎, 𝑎𝑎, 𝑏𝑏, and 𝑏𝑏 in some
ℍ-differentiable in 𝐺𝐺4 , then it possesses derivatives of all
open connected neighborhood 𝐺𝐺4 ∈ ℍ of a point 𝑝𝑝 ∈ ℍ.
orders in 𝐺𝐺4 , each one ℍ-holomorphic.
Then a function 𝜓𝜓(𝑝𝑝) is ℍ-holomorphic (further denoted
Corollary 1.5 All expressions for full derivatives of a ℍ-
by 𝜓𝜓𝐻𝐻 (𝑝𝑝)) at a point 𝑝𝑝 if and only if the functions 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏)
holomorphic function 𝜓𝜓𝐻𝐻 (𝑝𝑝) of the same kind as a
and 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) satisfy in 𝐺𝐺4 the following quaternionic
complex holomorphic (briefly, ℂ -holomorphic) function
generalization of complex Cauchy-Riemann's equations:
𝜓𝜓𝐶𝐶 (𝜉𝜉) have the same forms as the expressions for
1) (𝜕𝜕𝑎𝑎 𝜙𝜙1 | = (𝜕𝜕𝑏𝑏 𝜙𝜙2 |, 2) (𝜕𝜕𝑎𝑎 𝜙𝜙2 | = −(𝜕𝜕𝑏𝑏 𝜙𝜙1 |, corresponding derivatives of a function 𝜓𝜓𝐶𝐶 (𝜉𝜉).
� (1.3) For example, if the first derivative of the ℂ-holomorphic
3) (𝜕𝜕𝑎𝑎 𝜙𝜙1 | = (𝜕𝜕𝑏𝑏 𝜙𝜙2 |, 4) (𝜕𝜕𝑎𝑎 𝜙𝜙2 | = −(𝜕𝜕𝑏𝑏 𝜙𝜙1 |.
function 𝜓𝜓𝐶𝐶 (𝜉𝜉) = 𝜉𝜉 2 is 2𝜉𝜉, where 𝜉𝜉 is a complex variable,
Here 𝜕𝜕𝑖𝑖 , 𝑖𝑖 = 𝑎𝑎, 𝑎𝑎, 𝑏𝑏, 𝑏𝑏 , denotes the partial derivative then the first derivative of the ℍ -holomorphic
with respect to 𝑖𝑖. For details we refer to [6]. function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑝𝑝2 is 2𝑝𝑝 , where 𝑝𝑝 is a quaternionic
The overbars designate the complex (also quaternionic variable.
if needed) conjugation. The brackets (. . | with the closing Theorem 1.4 leads [6] to the following general
vertical bar indicate that the transition 𝑎𝑎 = 𝑎𝑎 = 𝑥𝑥 has expression for the full quaternionic derivatives of a
been already performed in expressions enclosed in ℍ-holomorphic function 𝜓𝜓𝐻𝐻 (𝑝𝑝) of all orders:
brackets. (𝑘𝑘)
Thus, ℍ -holomorphy conditions (1.3) are defined so 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝜙𝜙1(𝑘𝑘) + 𝜙𝜙2(𝑘𝑘) 𝑗𝑗, (1.4)
that during the check of the quaternionic holomorphy of (𝑘𝑘)
any quaternionic function we have to do the transition where a 𝑘𝑘′th derivative of 𝜓𝜓𝐻𝐻 (𝑝𝑝) is denoted by 𝜓𝜓𝐻𝐻 (𝑝𝑝)
(𝑘𝑘) (𝑘𝑘)
𝑎𝑎 = 𝑎𝑎 = 𝑥𝑥 in already computed expressions for the and the constituents 𝜙𝜙1 and 𝜙𝜙2 are expressed by
partial derivatives of the functions 𝜙𝜙1 and 𝜙𝜙2 . (𝑘𝑘) (𝑘𝑘−1) (𝑘𝑘−1) (𝑘𝑘−1)
However, this doesn’t mean that we deal with triplets 𝜙𝜙1 = 𝜕𝜕𝑎𝑎 𝜙𝜙1 + 𝜕𝜕𝑎𝑎 𝜙𝜙1 = 𝜕𝜕𝑎𝑎,𝑎𝑎 𝜙𝜙1 ,
in general, since the transition 𝑎𝑎 = 𝑎𝑎 = 𝑥𝑥 (or 𝑦𝑦 = 0 ) (𝑘𝑘)
𝜙𝜙2 = 𝜕𝜕𝑎𝑎 𝜙𝜙2
(𝑘𝑘−1)
+ 𝜕𝜕𝑎𝑎 𝜙𝜙2
(𝑘𝑘−1) (𝑘𝑘−1)
= 𝜕𝜕𝑎𝑎,𝑎𝑎 𝜙𝜙2 .
cannot be initially done for quaternionic variables and
(𝑘𝑘−1) (𝑘𝑘−1)
functions [6]. Any quaternionic function remains the same Here 𝜙𝜙1 and 𝜙𝜙2 are constituents of the (𝑘𝑘 − 1)′ th
4-dimensional quaternionic function regardless of whether derivative of 𝜓𝜓𝐻𝐻 (𝑝𝑝), represented in the Cayley-Dickson
(𝑘𝑘−1) (𝑘𝑘−1)
we check its holomorphy or not. This transition is needed doubling form as 𝜓𝜓𝐻𝐻 (𝑝𝑝)(𝑘𝑘 −1) = 𝜙𝜙1 + 𝜙𝜙2 ∙ 𝑗𝑗 ,
only to check the holomorphy of any quaternionic (0) (0)
𝑘𝑘 ≥ 1, 𝜙𝜙1 = 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏), 𝜙𝜙2 = 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏). We also denote
function. It is also used when solving the 3D tasks [7].
the full first quaternionic derivative by the prime mark: ′.
16 American Journal of Mathematical Analysis

Using the following equations obtained in [4]: If in an initial expression for 𝜓𝜓𝐶𝐶 (𝜉𝜉) special complex
formulae are used, then in order that the nature of the
1) 𝜕𝜕𝑏𝑏 𝜙𝜙2 = 𝜕𝜕𝑏𝑏 𝜙𝜙2 , 2) 𝜕𝜕𝑎𝑎 𝜙𝜙2 = −𝜕𝜕𝑏𝑏 𝜙𝜙1 , functional dependence does not change we use instead of
(1.5)
3) 𝜕𝜕𝑎𝑎 𝜙𝜙1 = 𝜕𝜕𝑎𝑎 𝜙𝜙1 , 4) 𝜕𝜕𝑎𝑎 𝜙𝜙2 = −𝜕𝜕𝑏𝑏 𝜙𝜙1 , them their analogues valid in the quaternionic area. For
example, we replace the complex Euler formula [1]:
one can construct other expressions for derivatives of a
ℍ-holomorphic function 𝜓𝜓𝐻𝐻 (𝑝𝑝) of all orders. For example, 𝑒𝑒 𝑦𝑦𝑦𝑦 = cos 𝑦𝑦 + 𝑖𝑖 sin 𝑦𝑦
𝜓𝜓′𝐻𝐻 (𝑝𝑝) = �𝜕𝜕𝑎𝑎 𝜙𝜙1 + 𝜕𝜕𝑎𝑎 𝜙𝜙1 � − �𝜕𝜕𝑏𝑏 𝜙𝜙1 + 𝜕𝜕𝑏𝑏 𝜙𝜙1 �𝑗𝑗. by its quaternionic analog:

They generalize in the quaternionic area the known 𝑒𝑒 𝑣𝑣𝑣𝑣 = cos 𝑣𝑣 + 𝑟𝑟 sin 𝑣𝑣.
formula variants for derivatives of ℂ -holomorphic This quaternionic expression follows from the quaternion
functions [1]. representation 𝑝𝑝 = 𝑥𝑥 + 𝑣𝑣𝑣𝑣, where 𝑣𝑣 = �𝑦𝑦 2 + 𝑧𝑧 2 + 𝑢𝑢2 and
The goals of this article are the following: 1) to formalize
yi + zj + uk
a rule for constructing ℍ -holomorphic functions, 2) to r= , that is algebraically equivalent to the
deduce rules for differentiating combinations of ℍ - y2 + z2 + u2
holomorphic functions, 3) to identify properties of a class complex number representation: 𝑧𝑧 = 𝑥𝑥 + 𝑦𝑦𝑦𝑦 . At that
of ℍ -holomorphic functions similar to complex ones, 4) we also replace the imaginary unit 𝑖𝑖 by its analog
to consider the examples of 3D applications, 5) to 𝑟𝑟 (𝑟𝑟 2 = −1, |𝑟𝑟| = 1) , which can be interpreted as a
automate cumbersome manual processing of the ℍ - quaternionic generalization of the complex imaginary unit
holomorphic functions. 𝑖𝑖.
In the sequel, ℍ -holomorphic functions are supposed Example 2.2. To illustrate this rule we construct the
to be defined in domain 𝐺𝐺4 , if nothing other else is quaternionic natural logarithm function from the complex
specified. one. The initial complex natural logarithm function is the
following [1]:
2. Constructing ℍ-holomorphic 𝜓𝜓𝐶𝐶 (𝑧𝑧) = log 𝑧𝑧 = ln|𝑧𝑧| + 𝑖𝑖arg(𝑧𝑧),
Functions where 𝑧𝑧 = 𝑥𝑥 + 𝑖𝑖𝑖𝑖 (do not confuse here with quaternion’s
component 𝑧𝑧 ), arg⁡ (𝑧𝑧) = Arg(z) + 2𝜋𝜋𝜋𝜋 , 𝑛𝑛 is an integer,
Now we formalize a general rule for constructing the z+z
Arg(z) = arccos is the principal value of arg⁡ (𝑧𝑧),
functions 𝜓𝜓𝐻𝐻 (𝑝𝑝) from their complex counterparts 𝜓𝜓𝐶𝐶 (𝜉𝜉) 2z
when replacing a complex variable 𝜉𝜉 as a single whole by
a quaternionic variable 𝑝𝑝 without change of a functional and |𝑧𝑧| = �𝑥𝑥 2 + 𝑦𝑦 2 ≠ 0.
dependence form. The ultimate goal is to obtain from the We consider the principal branch of log 𝑧𝑧 (as usual for
expression for complex holomorphic function 𝜓𝜓𝐶𝐶 (𝜉𝜉) the all multivalued functions): 𝜋𝜋−< Arg𝑧𝑧 ≤ 𝜋𝜋. Then we have
quaternionic expression for its essentially adequate z+z
generalization (Theorem 1.3) in the Cayley-Dickson ψ C ( z=
) L og ( z=) ln z + i ⋅ arccos . (2.2)
2z
doubling form 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) + 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗 without
change of a functional dependence form. Step 1. Replacing 𝑧𝑧 by 𝑝𝑝, we obtain
Constructing rule 2.1. The constructing rule is divided
p+ p
into 2 steps as follows. ψH (=
p ) L og ( =
p ) ln p + i ⋅ arccos ,
Step 1. In an initial expression for ℂ -holomorphic 2 p
function, depending only on complex variables 𝜉𝜉 as a
single whole, we replace this variable by a quaternionic p where
without change of a functional dependence form. By
|𝑝𝑝| = �𝑥𝑥 2 + 𝑦𝑦 2 + 𝑧𝑧 2 + 𝑢𝑢2 = �𝑎𝑎𝑎𝑎� + 𝑏𝑏𝑏𝑏�.
virtue of Theorem 1.3, the obtained function 𝜓𝜓𝐻𝐻 (𝑝𝑝) is
3 1 Step 2. Since when getting (2.2) in complex analysis the
ℍ-holomorphic. For example, the functions 𝑒𝑒 𝑝𝑝 , , Euler formula is used, we replace 𝑖𝑖 by 𝑟𝑟 and have:
1 + p2
5
𝑐𝑐𝑐𝑐𝑐𝑐 𝑒𝑒 𝑝𝑝 , 𝑠𝑠𝑠𝑠𝑠𝑠 𝑝𝑝2 ⋅ 𝐿𝐿𝐿𝐿𝐿𝐿(�𝑝𝑝) are ℍ -holomorphic in their p+ p
ψ H ( p=
) L og ( p=) ln p + r ⋅ arccos
domains of definition. Complex variables can have other 2 p
designations.  yi + zj + uk 
Step 2. We represent the obtained expression for 𝜓𝜓𝐻𝐻 (𝑝𝑝) in = ln aa + bb +   ⋅ arccos p + p
the Cayley-Dickson doubling form (1.2). In order to  y2 + z2 + u2  2 aa + bb
obtain dependencies of complex functions 𝜙𝜙1 and 𝜙𝜙2  
only on complex variables 𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � 𝑏𝑏� we replace real  yi p + p 
components 𝑥𝑥, 𝑦𝑦, z , and 𝑢𝑢 of a quaternion 𝑝𝑝 by the =  ln aa + bb + arccos
 y2 + z2 + u2 2 aa + bb 
equivalent relations: 
a+a a−a b+b b−b zj + uk p+ p
=x = ,y = ,z = ,u (2.1) + arccos ,
2 2i 2 2i y + z +u
2 2 2 2 aa + bb
which follow from (1.1). p + p = a + a = x.
 American Journal of Mathematical Analysis 17

Substituting relations (2.1) and simplifying, we obtain 1) (𝜕𝜕𝑎𝑎 𝑓𝑓1 | = (𝜕𝜕𝑏𝑏 𝑓𝑓2 |, 2) (𝜕𝜕𝑎𝑎 𝑓𝑓2 | = −(𝜕𝜕𝑏𝑏 𝑓𝑓1 |,
finally the following expression for Log(𝑝𝑝): � (3.3)
3) (𝜕𝜕𝑎𝑎 𝑓𝑓1 | = (𝜕𝜕𝑏𝑏 𝑓𝑓2 |, 4) (𝜕𝜕𝑎𝑎 𝑓𝑓2 | = −(𝜕𝜕𝑏𝑏 𝑓𝑓1 |.
� 𝑏𝑏�) + 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏, 𝑎𝑎,
Log(𝑝𝑝) = 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏, 𝑎𝑎, � 𝑏𝑏�)𝑗𝑗, (2.3) Consider the function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑐𝑐𝑓𝑓𝐻𝐻 (𝑝𝑝) = 𝑐𝑐𝑓𝑓1 (𝑎𝑎, 𝑏𝑏) +
where 𝑐𝑐𝑐𝑐2 (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗. Substituting this function into (1.3), we have
a+a 1) 𝑐𝑐(𝜕𝜕𝑎𝑎 𝑓𝑓1 | = 𝑐𝑐(𝜕𝜕𝑏𝑏 𝑓𝑓2 |, 2) 𝑐𝑐(𝜕𝜕𝑎𝑎 𝑓𝑓2 | = −𝑐𝑐(𝜕𝜕𝑏𝑏 𝑓𝑓1 |,
( a − a ) arccos �
2 aa + bb , 3) 𝑐𝑐(𝜕𝜕𝑎𝑎 𝑓𝑓1 | = 𝑐𝑐(𝜕𝜕𝑏𝑏 𝑓𝑓2 |, 4) 𝑐𝑐(𝜕𝜕𝑎𝑎 𝑓𝑓2 | = −𝑐𝑐(𝜕𝜕𝑏𝑏 𝑓𝑓1 |,
φ=
1 ln aa + bb + (2.4)
2v i.e. system of equations which is equivalent to (3.3). Thus
a+a the function 𝑐𝑐𝑓𝑓𝐻𝐻 (𝑝𝑝) is ℍ -holomorphic. The validity of
b ⋅ arccos (3.2) follows from (1.4). This rule is proved.
φ2 = 2 aa + bb , (2.5) By virtue of rule 3.1 formula (3.2) is also valid for the
v function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑐𝑐𝑓𝑓𝐻𝐻 (𝑝𝑝) ± 𝑞𝑞 , where 𝑞𝑞 ∈ ℍ. It is not
superfluous to note that its special case, when 𝑓𝑓𝐻𝐻 (𝑝𝑝) = 𝑝𝑝,
(
4 aa + bb − ( a + a ) ) 2
coincides with the result of the theory based on Cauchy-
v= y + z +u =
2 2 2
(2.6) Riemann-Fueter equations [2,3], however this theory is in
2
principle restricted by this result.
and 𝑝𝑝, 𝑣𝑣 ≠ 0. 3.3 Power rule. The quaternionic derivative of a power
As shown in [6], the function Log(𝑝𝑝) is ℍ-holomorphic, function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑝𝑝𝒶𝒶 , where 𝒶𝒶 ∈ ℝ ∖ 0, is the following:
satisfying generalized Cauchy-Riemann's equations (1.3):
(𝑝𝑝𝒶𝒶 )′ = 𝒶𝒶𝑝𝑝𝒶𝒶−1 . (3.4)
x
arccos
x p3 Proof. The validity of formula (3.4) follows from
1) (𝜕𝜕𝑎𝑎 𝜙𝜙1 � = (𝜕𝜕𝑏𝑏 𝜙𝜙2 � = + , Theorem 1.3 and Corollary 1.5.
2 2b
2 p3 3.4 Sum and difference rule. A sum of a finite number 𝑛𝑛
b of the ℍ -holomorphic functions 𝑓𝑓𝑙𝑙 (𝑝𝑝) ( 𝑙𝑙 = 1, … 𝑛𝑛 is an
2) (𝜕𝜕𝑎𝑎 𝜙𝜙2 � = − (𝜕𝜕𝑏𝑏 𝜙𝜙1 � = − , integer) is also ℍ −holomorphic. The full ℍ-holomorphic
2
2 p3 derivative of the sum 𝜓𝜓𝐻𝐻 (𝑝𝑝) = ∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙 (𝑝𝑝) is the following:
arccos
x 𝜓𝜓′𝐻𝐻 (𝑝𝑝) = (∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙 (𝑝𝑝))′ = ∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙′ (𝑝𝑝).
x p3
=
3) (𝜕𝜕𝑎𝑎 𝜙𝜙1 | = (𝜕𝜕𝑏𝑏 𝜙𝜙2 | + Proof. Let the functions 𝑓𝑓𝑙𝑙 (𝑝𝑝) = 𝑓𝑓1,𝑙𝑙 + 𝑓𝑓2,𝑙𝑙 ∙ 𝑗𝑗 , 𝑙𝑙 =
2 2b
2 p3 1,2 … 𝑛𝑛, be ℍ-holomorphic. Then each of them satisfies
b equations (1.3) as follows:
4) (𝜕𝜕𝑎𝑎 𝜙𝜙2 | = − (𝜕𝜕𝑏𝑏 𝜙𝜙1 | = − ,
2 p3
2
1) (𝜕𝜕𝑎𝑎 𝑓𝑓1,𝑙𝑙 | = (𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,𝑙𝑙 |, 2) (𝜕𝜕𝑎𝑎 𝑓𝑓2,𝑙𝑙 | = −(𝜕𝜕𝑏𝑏 ����
𝑓𝑓1,𝑙𝑙 |,
�
where |𝑝𝑝3 | is |𝑝𝑝| after setting 𝑎𝑎 = 𝑎𝑎 = 𝑥𝑥. 3) (𝜕𝜕𝑎𝑎 𝑓𝑓1,𝑙𝑙 | = (𝜕𝜕𝑏𝑏 𝑓𝑓2,𝑙𝑙 |, 4) (𝜕𝜕𝑎𝑎 𝑓𝑓2,𝑙𝑙 | = −(𝜕𝜕𝑏𝑏 𝑓𝑓1,𝑙𝑙 |.
Adding the functions 𝑓𝑓1 (𝑝𝑝) , 𝑓𝑓2 (𝑝𝑝) ,… 𝑓𝑓𝑛𝑛 (𝑝𝑝) by
component-wise addition, we obtain as follows:
3. Differentiation Rules for Combinations n
of ℍ-holomorphic Functions ∑ fl ( p=) ( f1,1 + f1,2 + ... + f1,n )
l =1
We consider the quaternionic generalizations of known
complex rules and formulae [1] for differentiating powers,
( )
+ f 2,1 + f 2,2 + ... + f 2,n ⋅ j = φ1 + φ2 ⋅ j
sums, products, ratios, inverses, and compositions of whence
ℍ-holomorphic functions.
𝜙𝜙1 = 𝑓𝑓1,1 + 𝑓𝑓1,2 +. . . +𝑓𝑓1,𝑛𝑛 ,
3.1 Constant function. Let a ℍ-holomorphic function be
a constant 𝐶𝐶 = 𝐶𝐶1 + 𝐶𝐶2 𝑗𝑗 ∈ ℍ, where 𝐶𝐶1 , 𝐶𝐶2 ∈ ℂ. Then the 𝜙𝜙2 = 𝑓𝑓2,1 + 𝑓𝑓2,2 +. . . +𝑓𝑓2,𝑛𝑛 .
following formula holds true:
The derivatives which we need to substitute into
𝐶𝐶 ′ = 0. equation (1.3-1) are the following:
Proof. Since the derivative of a constant with respect to 𝜕𝜕𝑎𝑎 𝜙𝜙1 = 𝜕𝜕𝑎𝑎 𝑓𝑓1,1 + 𝜕𝜕𝑎𝑎 𝑓𝑓1,2 +. . . +𝜕𝜕𝑎𝑎 𝑓𝑓1,𝑛𝑛 ,
any variable is zero, this formula is proved.
3.2 Multiplying by a constant. A ℍ -holomorphic 𝜕𝜕𝑏𝑏 𝜙𝜙2 = 𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,1 + 𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,2 +. . . +𝜕𝜕𝑏𝑏 �����
𝑓𝑓2,𝑛𝑛 .
function 𝑓𝑓𝐻𝐻 (𝑝𝑝) multiplied by an arbitrary constant 𝑐𝑐 ∈ ℝ: After performing the transition 𝑎𝑎 = 𝑎𝑎� = 𝑥𝑥 in them, we
𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑐𝑐𝑓𝑓𝐻𝐻 (𝑝𝑝) (3.1) can rewrite equation (1.3-1) as follows:
is ℍ -holomorphic as well. The first quaternionic (𝜕𝜕𝑎𝑎 𝜙𝜙1 | = (𝜕𝜕𝑎𝑎 𝑓𝑓1,1 | + (𝜕𝜕𝑎𝑎 𝑓𝑓1,2 |+. . . +(𝜕𝜕𝑎𝑎 𝑓𝑓1,𝑛𝑛 | =
derivative of 𝜓𝜓𝐻𝐻 (𝑝𝑝) is the following:
(𝜕𝜕𝑏𝑏 𝜙𝜙2 | = (𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,1 | + (𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,2 |+. . . +(𝜕𝜕𝑏𝑏 �����
𝑓𝑓2,𝑛𝑛 |.
′ ′
�𝑐𝑐𝑓𝑓𝐻𝐻 (𝑝𝑝)� = 𝑐𝑐 𝑓𝑓𝐻𝐻 (𝑝𝑝). (3.2)
Since by virtue of equations (3.3) we have (𝜕𝜕𝑎𝑎 𝑓𝑓1,1 | =
Proof. Let a quaternionic function 𝑓𝑓𝐻𝐻 (𝑝𝑝) = 𝑓𝑓1 (𝑎𝑎, 𝑏𝑏) + (𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,1 |, (𝜕𝜕𝑎𝑎 𝑓𝑓1,2 | = (𝜕𝜕𝑏𝑏 ����
𝑓𝑓2,2 |, and so on, we can state that
𝑓𝑓2 (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗 be ℍ-holomorphic. Then it satisfies condition this equation is satisfied for the sum of ℍ-holomorphic
(1.3) of ℍ-holomorphy as follows: functions. Analogously, we prove the validity of the other
18 American Journal of Mathematical Analysis

equations of system (1.3) for the sum ∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙 (𝑝𝑝). Thus the p ⋅ q= ( a1 + b1 ⋅ j ) ⋅ ( a2 + b2 ⋅ j )
sum ∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙 (𝑝𝑝) is ℍ-holomorphic.
( ) ( )
(3.5)
This rule also holds true when subtracting of an = a1a2 − b1 b2 + a1b2 + a2b1 ⋅ j
arbitrary number of ℍ-holomorphic functions.
Using formula (1.4) for 𝑘𝑘 = 1 , we get the full first where quaternions 𝑝𝑝 and 𝑞𝑞 are 𝑝𝑝 = 𝑎𝑎1 + 𝑏𝑏1 ⋅ 𝑗𝑗, 𝑞𝑞 = 𝑎𝑎2 +
𝑏𝑏2 ⋅ 𝑗𝑗 ; components 𝑎𝑎1 = 𝑥𝑥1 + 𝑦𝑦1 𝑖𝑖 , 𝑏𝑏1 = 𝑧𝑧1 + 𝑢𝑢1 𝑖𝑖 , 𝑎𝑎2 =
quaternionic derivative of ∑𝑛𝑛𝑙𝑙=1 𝑓𝑓𝑙𝑙 (𝑝𝑝) as follows:
𝑥𝑥2 + 𝑦𝑦2 𝑖𝑖, 𝑏𝑏2 = 𝑧𝑧2 + 𝑢𝑢2 𝑖𝑖 are complex, and 𝑥𝑥1 , 𝑦𝑦1 , 𝑧𝑧1 , 𝑢𝑢1 ,
 n
'
 and 𝑥𝑥2 , 𝑦𝑦2 , 𝑧𝑧2 , 𝑢𝑢2 are real variables.
ψH
'
( p ) =  ∑ fl ( p )  = ∂ a,a φ1 + ∂ a,a φ2 ⋅ j Performing the transition from quaternionic variables to
 l =1  complex ones, we put 𝑎𝑎1 = 𝑥𝑥1 (𝑦𝑦1 = 0), 𝑏𝑏1 = 𝑧𝑧1 (𝑢𝑢1 = 0),
= (
∂ a,a f1,1 + f1,2 + ... + f1,n ) 𝑎𝑎2 = 𝑎𝑎2 = 𝑥𝑥2 (𝑦𝑦2 = 0) , 𝑏𝑏2 = 𝑏𝑏2 = 𝑧𝑧2 (𝑢𝑢2 = 0) [6].
This leads to two complex variables 𝑝𝑝 = 𝑥𝑥1 + 𝑧𝑧1 𝑗𝑗 and
+∂ a,a ( f 2,1 + f 2,2 + ... + f 2,n ) ⋅ j 𝑞𝑞 = 𝑥𝑥2 + 𝑧𝑧2 𝑗𝑗, where imaginary unit 𝑗𝑗 ( 𝑗𝑗 2 = −1) plays a
n n
role of the habitual complex imaginary unit 𝑖𝑖. At that the
= ∑(∂ a,a f1,l + ∂ a,a f2,l ⋅ =
j ) ∑ f l' ( p ) . rule of quaternionic multiplication (3.5) reduces to the
complex rule of multiplication [1]:
l 1 =l 1

The sum rule is proved. p ⋅ q= ( x1 + z1 j )( x2 + z2 j )

It is evident that this rule remains valid when = ( x1 x2 − z1 z2 ) + ( x1 z2 + x2 z1 ) j.
considering any constants instead of 𝑓𝑓𝑙𝑙 (𝑝𝑝) . When
considering 𝑐𝑐𝑓𝑓𝑙𝑙 (𝑝𝑝), where 𝑐𝑐 ∈ ℝ, instead of 𝑓𝑓𝑙𝑙 (𝑝𝑝) the sum Since the replacement of a complex variable (as a
rule represents the property of linearity of the quaternionic single whole) by a quaternionic one in expressions for
derivative. ℂ-holomorphic functions is a reverse procedure, we have
The following theorem is needed for the sequel. no changing a functional dependence form. Statement (i)
Theorem 3.5 Suppose the quaternionic functions 𝑓𝑓𝐻𝐻 (𝑝𝑝) of the theorem is proved. This holds always true when we
and 𝑔𝑔𝐻𝐻 (𝑝𝑝) are ℍ -holomorphic in 𝐺𝐺4 ⊂ ℍ . Then, in 𝐺𝐺4 consider quaternionic product instead of complex one.
hold true: (i) the quaternionic product 𝜓𝜓𝐻𝐻 (𝑝𝑝) = (ii) As shown in [4], the general expressions for
𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝) is also ℍ-holomorphic, (ii) the quaternionic constituents 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) and 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) of a ℍ -holomorphic
multiplication of the functions 𝑓𝑓𝐻𝐻 (𝑝𝑝) and 𝑔𝑔𝐻𝐻 (𝑝𝑝) behaves function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝜙𝜙1 + 𝜙𝜙2 𝑗𝑗 are the following:
as commutative.
𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) = 𝐴𝐴[𝑎𝑎, 𝑎𝑎, �𝑏𝑏𝑏𝑏�] = 𝐴𝐴, (3.6)
Proof. (i) Consider the quaternionic product 𝜓𝜓𝐻𝐻 (𝑝𝑝) =
𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝). As shown in [6] (when proving Theorem 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) = 𝐵𝐵�(𝑎𝑎𝑎𝑎), (𝑎𝑎𝑎𝑎)𝑚𝑚 , �𝑏𝑏𝑏𝑏��𝑏𝑏 = 𝐵𝐵𝐵𝐵, (3.7)
1.3), there exist only two ways to go from equations (1.3) 0 (𝑚𝑚 −1) 1 2
of ℍ-holomorphy to the Cauchy-Riemann equations, viz.: where 𝐵𝐵 = 𝐵𝐵 , (𝑎𝑎𝑎𝑎)𝑚𝑚 = 𝑎𝑎𝑚𝑚 𝑎𝑎 + 𝑎𝑎 𝑎𝑎 + 𝑎𝑎(𝑚𝑚 −2) 𝑎𝑎
when 𝑝𝑝 = 𝑎𝑎 + 𝑏𝑏𝑏𝑏 → 𝜉𝜉 = 𝑥𝑥 + 𝑧𝑧𝑧𝑧 and when 𝑝𝑝 = 𝑎𝑎 + 𝑏𝑏𝑏𝑏 → 2 (𝑚𝑚 −2) 1 (𝑚𝑚 −1) 0 𝑚𝑚
… + 𝑎𝑎 𝑎𝑎 + 𝑎𝑎 𝑎𝑎 + 𝑎𝑎 𝑎𝑎 or another symmetric
𝜉𝜉 = 𝑥𝑥 + 𝑢𝑢𝑢𝑢. By the 𝜉𝜉 is here denoted a complex variable form invariant under complex conjugation.
with imaginary unit 𝑗𝑗 or 𝑘𝑘 instead of 𝑖𝑖. Assume that the functions 𝑓𝑓𝐻𝐻 = 𝑓𝑓1 + 𝑓𝑓2 ∙ 𝑗𝑗 and 𝑔𝑔𝐻𝐻 =
Both ways mean the same one-to-one correspondence 𝑔𝑔1 + 𝑔𝑔2 ∙ 𝑗𝑗 are ℍ -holomorphic. This means that they
between the set of all ℍ-holomorphic functions and the set satisfy ℍ-holomorphy equations (1.3), where instead of
of all ℂ -holomorphic ones. In other words, upon the functions 𝜙𝜙1 and 𝜙𝜙2 we put, respectively, 𝑓𝑓1 and 𝑓𝑓2 in the
transition from quaternions to complex numbers each case of the function 𝑓𝑓𝐻𝐻 or put 𝑔𝑔1 and 𝑔𝑔2 in the case of the
complex holomorphic function follows uniquely from function 𝑔𝑔𝐻𝐻 . Then, introducing the designations 𝐴𝐴𝑓𝑓 ,
the corresponding ℍ -holomorphic function and vice 𝐵𝐵𝑓𝑓 , 𝐴𝐴𝑔𝑔 , 𝐵𝐵𝑔𝑔 we have in accordance with (3.6) (3.7) the
versa. following valid expressions:
If 𝑓𝑓𝐻𝐻 (𝑝𝑝) and 𝑔𝑔𝐻𝐻 (𝑝𝑝) are ℍ-holomorphic, then 𝑓𝑓𝐶𝐶 (𝜉𝜉) and
𝑔𝑔𝐶𝐶 (𝜉𝜉) are ℂ-holomorphic. In complex analysis the product 𝑓𝑓1 = 𝐴𝐴𝑓𝑓 [𝑎𝑎, 𝑎𝑎, �𝑏𝑏𝑏𝑏�] = 𝐴𝐴𝑓𝑓 ,
of two and more ℂ -holomorphic functions is also a
ℂ -holomorphic function [1]. Then, replacing 𝜉𝜉 by 𝑝𝑝 in ( )
f 2 = B f ( aa ) , ( aa )m , bb  b
(3.8)
the product 𝑓𝑓𝐶𝐶 (𝜉𝜉) ∙ 𝑔𝑔𝐶𝐶 (𝜉𝜉) without change of a functional
= B=
f b, B f Bf
dependence form ( 𝑓𝑓 and 𝑔𝑔 ), we infer by virtue
of Theorem 1.3 that the product 𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝) is
ℍ-holomorphic. =g1 A=  
g  a, a , bb  Ag , ( )
In the expression 𝜓𝜓𝐶𝐶 (𝜉𝜉) = 𝑓𝑓𝐶𝐶 (𝜉𝜉) ∙ 𝑔𝑔𝐶𝐶 (𝜉𝜉) the dot
" ∙ " denotes complex multiplication, however in the
g 2 = Bg ( aa ) , ( aa )m , bb  b ( ) (3.9)
expression 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝) the dot "∙" is already = B=
g b, B g Bg .
associated with quaternionic multiplication. Such a
replacement does not lead to the change of a functional Usual quaternion multiplication is non-commutative for
dependence form, since the forms of rules for the complex arbitrary quaternionic functions 𝑓𝑓 and 𝑔𝑔 : 𝑓𝑓 ∙ 𝑔𝑔 ≠ 𝑔𝑔 ∙ 𝑓𝑓 .
and quaternionic multiplication are the same. We can see However, we now prove that the ℍ-holomorphic functions,
this as follows. satisfying equations (1.3), possess such a property that the
For quaternionic multiplication in the Cayley-Dickson quaternionic multiplication of these functions behaves as
doubling form (1.2) we have the following rule [8]: commutative: 𝑓𝑓𝐻𝐻 ∙ 𝑔𝑔𝐻𝐻 = 𝑔𝑔𝐻𝐻 ∙ 𝑓𝑓𝐻𝐻 .
 American Journal of Mathematical Analysis 19

Using rule (3.5) for quaternion multiplication of the whence

functions 𝑓𝑓 and 𝑔𝑔, we obtain the following expressions:
𝑔𝑔1 = 𝑓𝑓�1 𝐵𝐵𝑔𝑔 + 𝐵𝐵𝑓𝑓 𝑔𝑔1 .
𝑓𝑓1 𝐵𝐵𝑔𝑔 + 𝐵𝐵𝑓𝑓 ��� (3.16)
( f1 + f 2 ⋅ j ) ⋅ ( g1 + g 2 ⋅ j )
fH ⋅ gH =
Further, multiplying both sides of (3.16) by 𝑏𝑏, we have
= ( ) (
f1 g1 − f 2 g 2 + f1 g 2 + f 2 g1 ⋅ j ) ���1 = 𝑓𝑓�1 𝐵𝐵𝑔𝑔 𝑏𝑏 + 𝐵𝐵𝑓𝑓 𝑏𝑏𝑔𝑔1 .
𝑓𝑓1 𝐵𝐵𝑔𝑔 𝑏𝑏 + 𝐵𝐵𝑓𝑓 𝑏𝑏𝑔𝑔
= φ1( f ⋅ g ) + φ2( f ⋅ g ) ⋅ j , Finally, using (3.8), (3.9) in the last expression, we get
the following expression:
g H ⋅ f H = ( g1 + g 2 ⋅ j ) ⋅ ( f1 + f 2 ⋅ j )
𝑔𝑔1 = 𝑓𝑓�1 𝑔𝑔2 + 𝑓𝑓2 𝑔𝑔1 ,
𝑓𝑓1 𝑔𝑔2 + 𝑓𝑓2 ���
( ) (
f1 g1 − f 2 g 2 + f1 g 2 + f 2 g1 ⋅ j
= ) which coincides with (3.11). Statement (ii) of the theorem
= φ1( g ⋅ f ) + φ2( g ⋅ f ) ⋅ j. is proved. This completes the proof of the theorem in
whole.
We will now prove that in the case of ℍ-holomorphic It is evident that by virtue of the associativity law of the
functions 𝑓𝑓𝐻𝐻 and 𝑔𝑔𝐻𝐻 the following equalities are valid: quaternion multiplication, this theorem can be applied to
an arbitrary number of multiplied ℍ -holomorphic
𝑔𝑔2 = 𝑓𝑓1 𝑔𝑔1 − 𝑓𝑓�2 𝑔𝑔2 ,
𝑓𝑓1 𝑔𝑔1 − 𝑓𝑓2 ��� (3.10) functions.
(𝜙𝜙1(𝑓𝑓∙𝑔𝑔) = 𝜙𝜙1(𝑔𝑔∙𝑓𝑓) ), Example 3.6 Consider the quaternionic product of the ℍ-
holomorphic functions Log 𝑝𝑝 and sin 𝑝𝑝 . According to
𝑔𝑔1 = 𝑓𝑓�1 𝑔𝑔2 + 𝑓𝑓2 𝑔𝑔1 ,
𝑓𝑓1 𝑔𝑔2 + 𝑓𝑓2 ��� (3.11) constructing rule 2.1, we have the ℍ -holomorphic
(𝜙𝜙2(𝑓𝑓∙𝑔𝑔) = 𝜙𝜙2(𝑔𝑔∙𝑓𝑓) ). function
Using (3.8) and (3.9), we get the following expressions: erp − e− rp
=
sin p = φ1( sin ) + φ2( sin ) ⋅ j ,
𝑓𝑓2 ���
𝑔𝑔2 = 𝐵𝐵𝑓𝑓 𝑏𝑏𝐵𝐵𝑔𝑔 𝑏𝑏 = 𝐵𝐵𝑓𝑓 𝐵𝐵𝑔𝑔 𝑏𝑏𝑏𝑏, 2r
where
𝑓𝑓�2 𝑔𝑔2 = 𝐵𝐵𝑓𝑓 𝑏𝑏𝐵𝐵𝑔𝑔 𝑏𝑏 = 𝐵𝐵𝑓𝑓 𝐵𝐵𝑔𝑔 𝑏𝑏𝑏𝑏,
whence (e −v
+e
v
) sin a +2 a ( a − a ) ( e−v − ev ) cos
a+a
2
=φ1( sin ) − ,
𝑔𝑔2 = 𝑓𝑓�2 𝑔𝑔2 .
𝑓𝑓2 ��� 2 4v

( e−v − ev ) cos
a+a
Thus, we have proved that equality (3.10), i. e.
𝜙𝜙1(𝑓𝑓∙𝑔𝑔) = 𝜙𝜙1(𝑔𝑔∙𝑓𝑓) , is valid for ℍ-holomorphic functions 𝑓𝑓𝐻𝐻 2 b.
and 𝑔𝑔𝐻𝐻 . The only thing left to do is to prove that the
φ2( sin ) = −
2v
equality (3.11) is also valid for ℍ-holomorphic functions
𝑓𝑓𝐻𝐻 and 𝑔𝑔𝐻𝐻 . The function Log 𝑝𝑝 and its constituents 𝜙𝜙1(Log ) and
According to (3.11), we have for the constituent 𝜙𝜙2(𝑓𝑓∙𝑔𝑔) 𝜙𝜙2(Log ) are defined by (2.3), (2.4), and (2.5).
of the product 𝑓𝑓𝐻𝐻 ∙ 𝑔𝑔𝐻𝐻 the following expression: To simplify all expressions we introduce the following
notation:
𝜙𝜙2(𝑓𝑓∙𝑔𝑔) = 𝑓𝑓1 𝑔𝑔2 + 𝑓𝑓2 ���.
𝑔𝑔1 (3.12)
a+a e − v + ev e − v − ev
Its complex conjugation is = ,γ=ϑ = ,δ ,
2 2 2
��������� �
𝜙𝜙2(𝑓𝑓∙𝑔𝑔) = 𝑓𝑓1 ��� �
𝑔𝑔2 + 𝑓𝑓2 𝑔𝑔1 . (3.13)
(a − a ) a+a
Since, as proved in (i), the quaternionic product 𝑓𝑓𝐻𝐻 ∙ 𝑔𝑔𝐻𝐻= σ = , and ρ arccos .
2v 2 aa + bb
is ℍ-holomorphic, we write the 𝜙𝜙2(𝑓𝑓∙𝑔𝑔) and its conjugation,
according to (3.7), as follows: Using the rule of quaternionic multiplication (3.5), we
get the following expression for the quaternionic product
���������
𝜙𝜙2(𝑓𝑓∙𝑔𝑔) = 𝐵𝐵𝐵𝐵, 𝜙𝜙 2(𝑓𝑓∙𝑔𝑔) = 𝐵𝐵𝑏𝑏 = 𝐵𝐵𝑏𝑏, of the function Log 𝑝𝑝 and sin 𝑝𝑝:
whence L og p ⋅ sin p

=B
φ2( f ⋅ g )
=
φ2( f ⋅ g )
. (3.14)
( )(
= φ1( log ) + φ2( log ) ⋅ j ⋅ φ1( sin ) + φ2( sin ) ⋅ j )
b b = φ1( log⋅sin ) + φ2( log⋅sin ) ⋅ j.
Substituting (3.12) and (3.13) into (3.14), we obtain the
following expression: where
𝜙𝜙1(𝐋𝐋og ∙sin ) = 𝜙𝜙1(Log ) 𝜙𝜙1(sin ) (3.17)
f1 g 2 + f 2 g1 f1 g 2 + f 2 g1
=B = . (3.15)
b b  γ sin ϑ  b ρδ b cos ϑ
( ln p + σρ ) 
−φ2( L og ) φ2( sin ) = +
Substituting expressions (3.8) and (3.9) as well as their  −σδ cos ϑ  v
2

conjugates into (3.15), we have

φ2( L og⋅sin ) φ1( log )φ2( sin ) + φ1( sin )φ2( log )
=
f1Bg b + B f bg1 f1Bg b + B f bg1
=B =
b b
,
=
γ ( sin ϑ ) b ρ
−
( ln p ) δ b cos ϑ . (3.18)

v v
 20 American Journal of Mathematical Analysis

Interchanging the order of multiplication, we obtain the Proof. The ℍ-holomorphy of the function 𝜓𝜓𝐻𝐻 (𝑝𝑝) follows
following expression for quaternionic multiplying sin 𝑝𝑝 by from Theorem 3.5. The differentiation formulae (3.23) and
Log 𝑝𝑝: (3.24) follow from Corollary 1.5.
Assertion 3.10 Let the quaternionic functions 𝑓𝑓𝐻𝐻 (𝑝𝑝) and
( )(
sin p ⋅ L og p = φ1( sin ) + φ2( sin ) ⋅ j ⋅ φ1( L og ) + φ2( L og ) ⋅ j ) 𝑔𝑔𝐻𝐻 (𝑝𝑝) ∈ ℍ ∖ 0 be ℍ-holomorphic. Then the left quotient
of 𝑓𝑓𝐻𝐻 (𝑝𝑝) by 𝑔𝑔𝐻𝐻 (𝑝𝑝) is equal to the right one.
= φ1( sin⋅L og ) + φ2( sin⋅L og ) ⋅ j ,
Proof. By the definition [8] the left and right quotients
of 𝑓𝑓𝐻𝐻 (𝑝𝑝) by 𝑔𝑔𝐻𝐻 (𝑝𝑝) are, respectively, the following:
where
φ1( sin⋅L og ) φ1( sin )φ1( L og ) − φ2( sin ) φ2( L og )
= gH ( p) ⋅ fH ( p) fH ( p) ⋅ gH ( p )
= XL = , XR .
gH ( p ) gH ( p )
2 2
δ b(cos ϑ )bρ (3.19)
= ( γ sin ϑ − σδ cos ϑ ) ( ln p + σρ ) + ,
v2 gH ( p) 1
Since = , the left and right quotients
φ2( sin⋅L og ) = φ1( sin )φ2( L og ) + φ1( L og )φ2( sin ) gH ( p) H ( p)
2 g

γ ( sin ϑ ) b ρ ( ln p ) δ b cos ϑ (3.20) can be represented as follows:

= − .
1
⋅ fH ( p) ,
v v
= XL (3.25)
Comparing expressions (3.17) and (3.19) as well as g H ( p)
(3.18) and (3.20), we prove the equalities:
1
𝜙𝜙1(𝐋𝐋og ∙sin ) = 𝜙𝜙1(sin ∙𝐋𝐋og ) , = X R fH ( p) ⋅ . (3.26)
gH ( p)
𝜙𝜙2(𝐋𝐋og ∙sin ) = 𝜙𝜙2(sin ∙𝐋𝐋og ) .
1
According to reciprocal rule 3.8, the function
Thus, we see that the quaternionic multiplication of the gH ( p)
ℍ -holomorphic functions Log 𝑝𝑝 and sin 𝑝𝑝 behaves as
commutative. is ℍ-holomorphic. Then, by Theorem 3.5, the quaternionic
3.7 Chain rule. Suppose the quaternionic functions 𝑔𝑔𝐻𝐻 (𝑝𝑝) product
1
⋅ f H ( p ) behave as commutative and we
and 𝑓𝑓(𝑝𝑝) are ℍ -holomorphic. Then the composite gH ( p)
function 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑓𝑓�𝑔𝑔𝐻𝐻 (𝑝𝑝)� is also ℍ-holomorphic. The get 𝑋𝑋𝐿𝐿 = 𝑋𝑋𝑅𝑅 .
following differentiation formula holds true: The assertion is proved.
𝜓𝜓′𝐻𝐻 (𝑝𝑝) = 𝑓𝑓𝑔𝑔′ �𝑔𝑔𝐻𝐻 (𝑝𝑝)� ∙ 𝑔𝑔𝐻𝐻′ (𝑝𝑝), (3.21) Given it, the left or right manner of quotient
computation alone may be used in the case of ℍ -
where 𝑓𝑓𝑔𝑔′ �𝑔𝑔𝐻𝐻 (𝑝𝑝)� denotes the derivative of 𝑓𝑓�𝑔𝑔𝐻𝐻 (𝑝𝑝)� with holomorphic functions.
respect to 𝑔𝑔𝐻𝐻 (𝑝𝑝). Example 3.11 We consider the left and right quotients of
Proof. The ℍ-holomorphy of the function 𝜓𝜓𝐻𝐻 (𝑝𝑝) follows yi + zj + uk
from Theorem 1.3. The differentiation formula (3.21) the function 𝑓𝑓𝐻𝐻 (𝑝𝑝) = 𝑝𝑝2 by g H ( p )= r= .
follows from Corollary 1.5. y 2
+ z 2
+ u 2

3.8 Reciprocal Rule. Let a quaternionic function According to constructing rule 2.1, we obtain
𝑔𝑔𝐻𝐻 (𝑝𝑝) ∈ ℍ ∖ 0 be ℍ-holomorphic. Then the multiplicative 𝑝𝑝2 = 𝜙𝜙1�𝑝𝑝 2 � + 𝜙𝜙2�𝑝𝑝 2 � ∙ 𝑗𝑗, (3.27)
1
inverse is also ℍ -holomorphic. The following
gH ( p)
where

differentiation rule holds true: 𝜙𝜙1�𝑝𝑝 2 � = 𝑎𝑎2 − 𝑏𝑏𝑏𝑏, 𝜙𝜙2�𝑝𝑝 2 � = 𝑏𝑏(𝑎𝑎 + 𝑎𝑎). (3.28)
' According to rule 2.1, beginning with Step 2, we get
 1  1
  =
− ⋅ gH
'
( p ). (3.22)
yi + zj + uk
g
 H ( p )  ( gH ( p )) 2
= r = φ1( r ) + φ2( r ) ⋅ j , (3.29)
y2 + z2 + u2
Proof. The ℍ-holomorphy of the quaternionic function
1 where
follows from Theorem 1.3. Formula (3.22)
gH ( p) a−a b
= φ1( r ) = , φ2( r ) , (3.30)
follows from Corollary 1.5. 2v v
3.9 Product rule. Let quaternionic functions 𝑓𝑓𝐻𝐻 (𝑝𝑝) 2
and 𝑔𝑔𝐻𝐻 (𝑝𝑝) be ℍ -holomorphic. Then the full derivative r =φ1( r ) φ1( r ) + φ2( r ) φ2( r ) =1,
of their quaternionic product 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝) =
𝑔𝑔𝐻𝐻 (𝑝𝑝) ∙ 𝑓𝑓𝐻𝐻 (𝑝𝑝) can be calculated by the following and 𝑣𝑣 is defined by (2.6). It is easy to verify that the
formulae: function 𝑟𝑟 as well as its conjugate 𝑟𝑟̅ are ℍ-holomorphic.
′ ′ There exists the following identity:
𝜓𝜓′𝐻𝐻 (𝑝𝑝) = �𝑓𝑓𝐻𝐻 (𝑝𝑝)� ∙ 𝑔𝑔𝐻𝐻 (𝑝𝑝) + 𝑓𝑓𝐻𝐻 (𝑝𝑝) ∙ �𝑔𝑔𝐻𝐻 (𝑝𝑝)� , (3.23)
1 r
′ ′ = == r φ1( r ) − φ2( r ) ⋅ j.
𝜓𝜓′𝐻𝐻 (𝑝𝑝) = �𝑔𝑔𝐻𝐻 (𝑝𝑝)� ∙ 𝑓𝑓𝐻𝐻 (𝑝𝑝) + 𝑔𝑔𝐻𝐻 (𝑝𝑝) ∙ �𝑓𝑓𝐻𝐻 (𝑝𝑝)� . (3.24) r r2
 American Journal of Mathematical Analysis 21

Using (3.25) and (3.5), we get the following Example 3.13 To illustrate the efficiency of the above
expressions: rules we consider the composite ℍ-holomorphic function
2
������
𝑋𝑋𝐿𝐿 = 𝑟𝑟̅ ∙ 𝑝𝑝2 = (𝜙𝜙 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑒𝑒 𝑝𝑝 . According to rule 2.1, we obtain
1(𝑟𝑟) − 𝜙𝜙2(𝑟𝑟) ∙ 𝑗𝑗) ∙ (𝜙𝜙1�𝑝𝑝 2 � + 𝜙𝜙2�𝑝𝑝 2 � ∙ 𝑗𝑗)
2
= 𝜙𝜙1�𝑟𝑟∙�𝑝𝑝 2 � + 𝜙𝜙2�𝑟𝑟∙�𝑝𝑝 2 � ∙ 𝑗𝑗, 𝑒𝑒 𝑝𝑝 = 𝜙𝜙1(e) + 𝜙𝜙2(e) ∙ 𝑗𝑗, (3.35)
where
where
𝜙𝜙1(e) = 𝛽𝛽𝛽𝛽, (3.36)
𝜙𝜙1�𝑟𝑟∙�𝑝𝑝 2 � = ������
𝜙𝜙1(𝑟𝑟) 𝜙𝜙1�𝑝𝑝 2 � + 𝜙𝜙2(𝑟𝑟) ��������
𝜙𝜙2(𝑝𝑝 2 ) , (3.31)
β sin θ
𝜙𝜙2�𝑟𝑟∙�𝑝𝑝 2 � = ������
𝜙𝜙1(𝑟𝑟) 𝜙𝜙2�𝑝𝑝 2 � − ��������
𝜙𝜙1(𝑝𝑝 2 ) 𝜙𝜙2(𝑟𝑟) . (3.32) φ2( e ) = b, (3.37)
v
On the other hand, we have as follows:
������
𝑋𝑋𝑅𝑅 = 𝑝𝑝2 ∙ 𝑟𝑟̅ = (𝜙𝜙1�𝑝𝑝 2 � + 𝜙𝜙2�𝑝𝑝 2 � ∙ 𝑗𝑗) ∙ (𝜙𝜙1(𝑟𝑟) − 𝜙𝜙2(𝑟𝑟) ∙ 𝑗𝑗)
=β
( x 2 − y 2 − z 2 −u 2 )
e= e
a 2 + a 2 − 2bb
2 ,
= 𝜙𝜙1�𝑝𝑝 2 ∙𝑟𝑟̅ � + 𝜙𝜙2�𝑝𝑝 2 ∙𝑟𝑟̅ � ∙ 𝑗𝑗,
( a − a ) sin θ
where =µ cos θ + , (3.38)
2v
𝜙𝜙1�𝑝𝑝 2 ∙𝑟𝑟̅ � = 𝜙𝜙1�𝑝𝑝 2 � ������
𝜙𝜙1(𝑟𝑟) + 𝜙𝜙2�𝑝𝑝 2 � �������
𝜙𝜙2(𝑟𝑟) , (3.33) 𝜃𝜃 = (𝑎𝑎 + 𝑎𝑎)𝑣𝑣,
𝜙𝜙2�𝑝𝑝 2 ∙𝑟𝑟̅ � = 𝜙𝜙1(𝑟𝑟) 𝜙𝜙2�𝑝𝑝 2 � − 𝜙𝜙1�𝑝𝑝 2 � 𝜙𝜙2(𝑟𝑟) . (3.34) and 𝑣𝑣 is defined by (2.6).
2
Substituting (3.28), (3.30) into (3.31), (3.32), (3.33) and Consider the first derivative of the function 𝑒𝑒 𝑝𝑝 . Using
(3.34), we finally get the following identities: formula (1.4), we calculate the first derivative as follows:
2 ′ ′ ′
( 3a + a ) bb − ( a − a ) a 2 �𝑒𝑒 𝑝𝑝 � = 𝜙𝜙1(e) + 𝜙𝜙2(e) ∙ 𝑗𝑗, (3.39)
φ= φ=
2
( ) 1( p ⋅r )
2
1 r⋅p
4bb − ( a − a )
2
,
where
′
𝜙𝜙1(e) = 𝜕𝜕𝑎𝑎 𝜙𝜙1(e) + 𝜕𝜕𝑎𝑎 𝜙𝜙1(e) , (3.40)

φ = φ = −
(
b a + a − 2bb
2 2
). 𝜕𝜕𝑎𝑎 𝜙𝜙1(e) = 𝜇𝜇𝜕𝜕𝑎𝑎 𝛽𝛽 + 𝛽𝛽𝜕𝜕𝑎𝑎 𝜇𝜇,
( )
2 r ⋅ p2 ( )
2 p 2 ⋅r
4bb − ( a − a )
2
𝜕𝜕𝑎𝑎 𝜙𝜙1(e) = 𝜇𝜇𝜕𝜕𝑎𝑎 𝛽𝛽 + 𝛽𝛽𝜕𝜕𝑎𝑎 𝜇𝜇,
′
Thus, the left and right quotients of 𝑝𝑝2 by 𝑟𝑟̅ are equal. 𝜙𝜙2(e) = 𝜕𝜕𝑎𝑎 𝜙𝜙2(e) + 𝜕𝜕𝑎𝑎 𝜙𝜙2(e) , (3.41)
3.12 Quotient rule If the quaternionic functions 𝑓𝑓𝐻𝐻 (𝑝𝑝)
 β sin θ 
and 𝑔𝑔𝐻𝐻 (𝑝𝑝) ∈ ℍ ∖ 0 are ℍ -holomorphic, then the ∂ aφ2( e ) =
b∂ a  ,
following quotient rule holds true:  v 
 β sin θ 
 fH ( p)  f H' ( p ) ⋅ g H ( p ) − f H ( p ) ⋅ g H ( p)
' ' ∂ a φ2( e ) =
b∂ a  .
  = .  v 
 gH ( p)  ( gH ( p )) 2
After cumbersome and quite tedious calculation we get
Proof. Using formula (3.25) for the left quotient, product   a2 − a 2  sin θ 
rule 3.9, reciprocal rule 3.8 and Theorem 3.5 we obtain −  v −  sin θ + 
  4v  2v 
 fH ( p)   1
' '
  
 =   ⋅ f H ( p )   (a − a )  a2 − a 2  
 gH ( p)   gH ( p)  ∂ aφ1( e ) = a βµ + β  +
2 v
v −
 4 v
 cos θ  ,

   
1 1
= − ⋅ gH
'
( p) ⋅ fH ( p) + ⋅ f H' ( p )  
 ( a − a ) sin θ
2
( gH ( p )) H ( ) 
2 g p
+ 3 
 8v 
f ' ( p) ⋅ gH ( p) − fH ( p) ⋅ gH
'
( p)
= H .
  a2 − a 2  sin θ 
( gH ( p )) 2
−  v +  sin θ − 
   4v   2v 
Using formula (3.26) for the right quotient, we have the  
same result. Obviously that all algebraic operations here  (a − a )  a2 − a 2  
are correct, since the quaternionic derivatives 𝑓𝑓𝐻𝐻′ (𝑝𝑝) and ∂ a φ1( e ) = a βµ + β  + v +  cos θ  .
2v   4v  
𝑔𝑔𝐻𝐻′ (𝑝𝑝) by virtue of Theorem 1.4 are also ℍ-holomorphic  
 
 ( a − a ) sin θ
functions and their multiplication behaves also as 2

commutative. The quotient rule is proved. − 3 
Note that the algebraic similarity between ℂ-holomorphic  8v 
and ℍ -holomorphic functions allows proving the Substituting the last two results into (3.40), we get the
presented rules in terms of limits as in the complex ′
following expression for 𝜙𝜙1(e) :
analysis. We see that all got differentiating rules are
′
similar to complex ones. 𝜙𝜙1(e) = (𝑎𝑎 + 𝑎𝑎)𝛽𝛽𝛽𝛽 + 𝛽𝛽[−2𝑣𝑣 sin 𝜃𝜃 + (𝑎𝑎 − 𝑎𝑎) cos 𝜃𝜃]. (3.42)
22 American Journal of Mathematical Analysis

Analogously, we obtain as follows: φ2'' ( e ) 4β b ( a + a ) cos θ

=
 4av 2 + ( a − a ) 
∂ aφ2( e ) =
bβ 
 4v 3
 sin θ
 +
(
2 β b 1 + a 2 + a 2 − 2bb sin θ ) .
(3.46)

v
 a2 − a 2 
+ bβ 1 −  cos θ , On the other hand, given (3.44), and using rules 3.2, 3.7,
 4v 2  and 3.9, we get
2 ′′ 2 ′ ′ 2 ′
 4av 2 − ( a − a )  �𝑒𝑒 𝑝𝑝 � = ��𝑒𝑒 𝑝𝑝 � � = 2�𝑒𝑒 𝑝𝑝 ∙ 𝑝𝑝�
∂ a φ2( e ) =
bβ   sin θ 2 2 2
 4v3  = 2�𝑒𝑒 𝑝𝑝 ∙ 2𝑝𝑝 ∙ 𝑝𝑝 + 𝑒𝑒 𝑝𝑝 � = 2𝑒𝑒 𝑝𝑝 ∙ (2𝑝𝑝2 + 1).
 a2 − a 2  Substituting (3.27) and (3.35), we obtain as follows:
+ bβ 1 +  cos θ .
 4v 2    
( 2φ ( ) + 2φ ( ) ⋅ j ) ⋅ 2φ ( ) + 1 + 2φ ( ) ⋅ j 
''
 p2 
e =  1e 2e 1 p2 2 p2
Combining these expressions and (3.41), we get     
(a + a )
(3.43) = φ1( e ) + φ2( e ) ⋅ j
'' ''
=φ2' ( e ) bβ sin θ + 2bβ cos θ .
v
Further, substituting (3.42), (3.43) into (3.39) we get Using (3.36), (3.37), (3.28) and (3.5), we get, after
′′ ′′
2
the following expression for the first derivative of 𝑒𝑒 𝑝𝑝 : some algebra, expressions for 𝜙𝜙1(e) and 𝜙𝜙2(e) , coinciding
with expressions (3.45) and (3.46). Obviously,
'
 p2  calculations with using the got rules for differentiating
 e  = ( a + a ) βµ + β  −2v sin θ + ( a − a ) cos θ  combinations of ℍ -holomorphic functions are much
 
simplier. In Appendix we present computing programmes,
 (a + a ) 
+ β b sin θ + 2b cos θ  ⋅ j. regarding to this example and essentially simplifying all
 v  considered calculations.
Substituting (3.38), uncovering brackets, and then
uniting the summands, involving the functions sin 𝜃𝜃 and 4. Class of ℍ-holomorphic Functions
cos 𝜃𝜃, we obtain after rather cumbersome calculations the
following expression: The class of the ℍ -holomorphic functions includes
 p2 
'   ( a − a ) sin θ  β sin θ  quaternionic functions that satisfy equations (1.3) of
e  =2  β cos θ + a− bb  the essentially adequate quaternionic generalization
    2v  v  of the complex Cauchy-Riemann equations. Among all
  ( a − a ) sin θ  β sin θ  quaternionic functions they alone possess one remarkable
+ 2  β cos θ + b + ba  ⋅ j , feature: their algebraic and differential operations are fully
  2v  v  identical to complex ones. Each of these functions can be
whence, using (3.36), (3.37) and (3.38), we have obtained from its complex holomorphic counterpart by
2 ′
using rule 2.1. We can list their properties as follows.
�𝑒𝑒 𝑝𝑝 � = 2�𝜙𝜙1(e) 𝑎𝑎 − 𝜙𝜙2(e) 𝑏𝑏� + 2�𝜙𝜙1(e) 𝑏𝑏 + 𝜙𝜙2(e) 𝑎𝑎� ∙ 𝑗𝑗. 1) The quaternionic multiplication of the ℍ-holomorphic
Given the rule of quaternionic multiplication (3.5), we functions behaves as commutative one and the left
can rewrite this expression as follows: quotient of two ℍ-holomorphic functions equals the right
2 ′ 2
one.
�𝑒𝑒 𝑝𝑝 � = 2�𝜙𝜙1(e) + 𝜙𝜙2(e) ∙ 𝑗𝑗� ∙ (𝑎𝑎 + 𝑏𝑏 ∙ 𝑗𝑗) = 2𝑒𝑒 𝑝𝑝 ∙ 𝑝𝑝. 2) The differentiation rules for ℍ -holomorphic
On the other hand, using chain rule 3.7, we can directly functions are the same as for ℂ-holomorphic ones.
get this expression: 3) The constituents of the ℍ -holomorphic functions
(and their derivatives) in the Cayley-Dickson doubling
2 ′ 2 ′ 2
�𝑒𝑒 𝑝𝑝 � = �𝑒𝑒 𝑝𝑝 �𝑝𝑝 2 ∙ (𝑝𝑝2 )′ = 2𝑒𝑒 𝑝𝑝 ∙ 𝑝𝑝. (3.44) form (1.2) have the following general representation form
[4]:
Thus, we see that using chain rule 3.7 reduces
essentially the volume of calculations. 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) = 𝐴𝐴[𝑎𝑎, 𝑎𝑎, �𝑏𝑏𝑏𝑏�] = 𝐴𝐴,
2
Consider the second derivative of 𝑒𝑒 𝑝𝑝 . Using (1.4), we 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) = 𝐵𝐵�(𝑎𝑎𝑎𝑎), (𝑎𝑎𝑎𝑎)𝑚𝑚 , �𝑏𝑏𝑏𝑏��𝑏𝑏 = 𝐵𝐵𝐵𝐵,
obtain after tedious calculations the following expression:
2 ′′
𝐵𝐵 = 𝐵𝐵� .
′′ ′′
�𝑒𝑒 𝑝𝑝 � = 𝜙𝜙1(e) + 𝜙𝜙2(e) ∙ 𝑗𝑗,
Such forms are typical of ℍ -holomorphic functions
where and their derivatives. They could serve as gauge for

( )
correctness of results obtained by using constructing rule
φ1''( e ) = 2β 1 + 2a 2 − 2bb cos θ 2.1.
4) The constituents 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) and 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) of the

+  ( )
β  1 + 2a 2 ( a − a ) − 2 ( 3a + a ) bb  sin θ
 .
(3.45) (1) (1)
ℍ-holomorphic functions (as well as 𝜙𝜙1 and 𝜙𝜙2 , 𝜙𝜙1
(2)

(2)
v and 𝜙𝜙2 of their derivatives, etc) are symmetric in
 American Journal of Mathematical Analysis 23

variables 𝑏𝑏 , 𝑏𝑏 and 𝑎𝑎 , 𝑎𝑎 , respectively [4,6]. Such a Substituting (1.1) and grouping terms with imaginary
symmetry for derivatives is a consequence of uniting units 𝑖𝑖, 𝑗𝑗 and 𝑘𝑘, we get the following expression for the
unsymmetrical parts of the left and right derivatives [4], quaternionic potential Log 𝑝𝑝 in ℝ4 -representation:
reflecting undoubtedly a symmetry of physical space. 𝜓𝜓𝐻𝐻 (𝑝𝑝) = Log 𝑝𝑝 = 𝜓𝜓1 + 𝜓𝜓2 𝑖𝑖 + 𝜓𝜓3 𝑗𝑗 + 𝜓𝜓4 𝑘𝑘,
5) The constituents 𝜙𝜙1 (𝑎𝑎, 𝑏𝑏) and 𝜙𝜙2 (𝑎𝑎, 𝑏𝑏) satisfy also
equations (1.5) [4,6]. where
6) The constituents 𝜓𝜓1 , 𝜓𝜓2 , 𝜓𝜓3 , 𝜓𝜓4 of ℍ-holomorphic  x
functions in ℝ4 -representation are ℍ-harmonic functions y arccos  
in the sense that they satisfy the generalizated quaternionic = ψ 1 ln= p ,ψ 2  p ,
Laplace equations [7]. v
7) ℍ-holomorphic functions have local representations  x  x
by convergent power series [9], i.e. are analytic functions. z arccos   u arccos  
It can be supposed that the commutative behavior of p
  ,ψ  p .
= ψ3 = 4
quaternionic multiplication and the equality of the left and v v
right quotients in the case of ℍ -holomorphic functions After transition 𝑦𝑦 = 0 (to 3D space), we obtain
exist in an "objective reality" independently from whether
any theory exists or not. One can just verify these (Log 𝑝𝑝| = 𝜑𝜑1 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) + 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢)𝑗𝑗 + 𝜑𝜑4 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢)𝑘𝑘,
properties without even knowing any theory, constructing where
ℍ-holomorphic functions from their complex holomorphic
counterparts in accordance with rule of constructing 2.1. 𝜑𝜑1 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = (𝜓𝜓1 | = ln√𝑥𝑥 2 + 𝑧𝑧 2 + 𝑢𝑢2 ,
When using this class there are no principal restrictions
 x 
to build a quaternionic differential analysis similar to z arccos  
complex one.  x 2 + z 2 + u 2 
Given that the ℍ-holomorphic functions are infinitely , u ) (ψ
ϕ3 ( x , z = =3| ,
differentiable, one can denote this class by the symbol ℍ∞ . z2 + u2
Note that due to identity of properties of ℍ- holomorphic
 
and ℂ- holomorphic functions all proofs of limit theorems u arccos 
x

in quaternionic area can be the same as in complex one.  x 2 + z 2 + u 2 

ϕ4 ( x, z= , u ) (ψ = 4| .
z2 + u2
5. Some Applications 1
As shown in [6], we have (Log 𝑝𝑝)′ = . Using (1.1),
We follow the theory of quaternionic potential and the p
notation presented in [7]. Two elementary steady state we obtain the first quaternionic derivative of the function
fluid flows from which it is possible to construct more Log 𝑝𝑝 in ℝ4 -representation as follows:
flows by superposition are here considered.
1 p
( L og p )' = =
2
=ψ 1' + ψ 2' i + ψ 3' j + ψ 4' k
p p
5.1. Elementary Source Flow
The elementary source flow is described in complex where
analysis [11] by the ℂ-holomorphic function or complex x y
potential 𝜓𝜓𝐶𝐶 (𝑧𝑧) = Log 𝑧𝑧 . The quaternionic potential of ψ 1' = ,ψ 2' = −
,
this flow is correspondingly the ℍ-holomorphic function u +x +y +z
2 2 2
u + x + y2 + z2
2 2 2

Log 𝑝𝑝 [7]. We rewrite expressions (2.3) - (2.6) as ψ 3' =

−
z
,ψ 4' =
−
u
.
follows: u +x +y +z
2 2 2 2
u + x + y2 + z2
2 2

𝐋𝐋og 𝑝𝑝 = 𝜙𝜙1(Log 𝑝𝑝) (𝑎𝑎, 𝑏𝑏) + 𝜙𝜙2(Log 𝑝𝑝) (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗, According to the quaternionic potential theory [7], we
where have for the quaternionic potential Log 𝑝𝑝 the following
flow velocity vector (field vector) in 3D space:
a+a
( a − a ) arccos (𝐹𝐹(𝑝𝑝)| = ������������
(Log 𝑝𝑝)′ | = (𝑓𝑓1 | + (𝑓𝑓2 |𝑖𝑖 + (𝑓𝑓3 |𝑗𝑗 + (𝑓𝑓4 |𝑘𝑘 (5.3)
2 p
φ1( L og=
p ) ln p + (5.1)
2v where

f1 | (ψ
a+a x
b ⋅ arccos (= = '
1| , (5.4)
2 p u + x2 + z 2
2
φ2( L og p ) = (5.2)
v (𝑓𝑓2 | = −(𝜓𝜓2′ | = 0,

( )
4 aa + bb − ( a + a )
2
( f3 | =
−(ψ 3' |=
z
, (5.5)
v= y + z +u =
2 2 2
≠ 0, 2
u + x2 + z 2
2
u
|𝑝𝑝| = �𝑥𝑥 2 + 𝑦𝑦 2 + 𝑧𝑧 2 + 𝑢𝑢2 = �𝑎𝑎𝑎𝑎� + 𝑏𝑏𝑏𝑏� ≠ 0. ( f4 | =
−(ψ 4' |= . (5.6)
2
u + x2 + z 2
24 American Journal of Mathematical Analysis

On the other hand, we can directly apply the 3D According to [7], the 3D stream surfaces 𝑆𝑆 are to be
gradient operator [7]: defined by the following generating function:
∇3 = 𝜕𝜕𝑥𝑥 + 𝜕𝜕𝑧𝑧 𝑗𝑗 + 𝜕𝜕𝑢𝑢 𝑘𝑘 (5.7) 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢)𝑗𝑗 + 𝜑𝜑4 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢)𝑘𝑘 = 𝐾𝐾3 𝑗𝑗 + 𝐾𝐾4 𝑘𝑘, (5.9)
to the function 𝜑𝜑1 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢). Then, we obtain the following where 𝐾𝐾3 , 𝐾𝐾4 ∈ ℝ are constants.
result: Equations 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = 𝐾𝐾3 when 𝑢𝑢 = 0 and 𝜑𝜑4 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢)
∇3 𝜑𝜑1 = 𝜕𝜕𝑥𝑥 𝜑𝜑1 +𝜕𝜕𝑧𝑧 𝜑𝜑1 𝑗𝑗 + 𝜕𝜕𝑢𝑢 𝜑𝜑1 𝑘𝑘, = 𝐾𝐾4 when 𝑧𝑧 = 0 give the following traces of the desired
stream surfaces in the planes 𝑥𝑥0𝑧𝑧 ∶ 𝜉𝜉 = 𝑥𝑥 + 𝑧𝑧𝑧𝑧 and
where 𝑥𝑥0𝑢𝑢 ∶ 𝜉𝜉 = 𝑥𝑥 + 𝑢𝑢𝑢𝑢, respectively:
x z  
∂ xϕ1
= = , ∂ zϕ1 , x
u +x +z 22 2
u + x2 + z 2
2 =ϕ3 ( x, z ) arc
= cos   K3 ,
 x 2 + z 2 
u
=∂ uϕ1 , u 2 + x 2 + z 2 ≠ 0.
u + x2 + z 2
2
 x 
=ϕ4 ( x, u ) arc
= cos   K4 .
The last expressions coincide with expressions (5.4),  x 2 + u 2 
(5.5) and (5.6), respectively. Hence the steady state fluid
flow, corresponding to quaternionic potential Log 𝑝𝑝, is a Leaving designations for constants unchanged, we
potential one in 𝐺𝐺4 ∈ ℍ, except at the singularity 𝑝𝑝 = 0. reduce these equations for traces to the following
The 3D flow velocities calculated in accordance with equations:
(5.3) are plotted in Figure 1. x
ϕ3 ( x , z ) = K 3 ,
𝑥𝑥0𝑧𝑧: =
x + z2
2

x
ϕ 4 ( x, u ) = K 4 . .
𝑥𝑥0𝑢𝑢: =
x2 + u 2
Since 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) and 𝜑𝜑4 (𝑥𝑥, 𝑢𝑢) lie in the range of values of
the function 𝐶𝐶𝐶𝐶𝐶𝐶, we have the restrictions 𝐾𝐾3 , 𝐾𝐾4 ≤ 1.
Putting 𝐾𝐾3 = 𝐾𝐾4 = 𝐾𝐾 and rotating counterclockwise the
π
plane 𝑥𝑥0𝑧𝑧 around the 𝑥𝑥 -axis by the angle , we can
2
obtain the trace 𝜑𝜑4 (𝑥𝑥, 𝑢𝑢) from 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧). Hence the desired
stream surfaces 𝑆𝑆 for quaternionic potential Log 𝑝𝑝 in the
simplest case are the surfaces of revolution [12] around
the 𝑥𝑥-axis.
To obtain the surfaces of revolution from 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) we
use the known method of replacing the variable 𝑧𝑧 by
√𝑧𝑧 2 + 𝑢𝑢2 [12] in the expression for 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) and get the
Figure 1. The overall picture of the velocity vectors for quaternionic
potential Log 𝑝𝑝. The vectors near the origin are highlighted by orange following equation for the stream surfaces 𝑆𝑆:
color. The number of depicted flow lines is 6 𝑆𝑆: (𝐾𝐾 2 − 1)𝑥𝑥 2 + 𝐾𝐾 2 𝑧𝑧 2 + 𝐾𝐾 2 𝑢𝑢2 = 0, (5.10)
The flow velocity vectors (𝐹𝐹(𝑝𝑝)| are directed radially 2
where 𝐾𝐾 ≤ 1.
away from the origin and indicated with arrows. The sizes It easy to see that equation (5.10) is the equation of 3D
of depicted arrows depend on the absolute values of the conical surfaces [12] that have the axis 0𝑥𝑥 and vertexes at
1 the origin. The streamlines are situated on these imagined
flow velocity: |(𝐹𝐹(𝑝𝑝)|| = �(𝑓𝑓1 |2 + (𝑓𝑓3 |2 + (𝑓𝑓4 |2 = .
(p| 3D stream surfaces 𝑆𝑆.
From equations (5.8) and (5.10), it follows that the
Note that this dependence corresponds to the analogous
equation of a curve 𝐼𝐼 of intersection of the surfaces 𝜑𝜑1 and
1 𝑆𝑆 is the following:
dependence for complex potential Log 𝑧𝑧 [11].
z
Now by analogy to the example of 3D flow modeling
zI + uI =
2 2
rI2 = K12 − xI2 =
(1 − K 2 ) xI2
,
considered in [7], we find equations of 3D equipotential K2
surfaces 𝜑𝜑1 and stream surfaces 𝑆𝑆 for the quaternionic
potential Log 𝑝𝑝. where 𝑥𝑥𝐼𝐼 , 𝑧𝑧𝐼𝐼 , 𝑢𝑢𝐼𝐼 ∈ 𝐼𝐼, and 𝐼𝐼 is the circle of radius 𝑟𝑟𝐼𝐼 with the
For 3D equipotential surfaces 𝜑𝜑1 we obtain the center (𝑥𝑥𝐼𝐼 , 0,0) on the 𝑥𝑥-axis.
equation Assuming, for definiteness, that 𝑟𝑟𝐼𝐼 = 2, 𝑥𝑥𝐼𝐼 = 3 we get
the following values of constants: 𝐾𝐾 2 = 0,69 and
𝜑𝜑1 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = ln√𝑥𝑥 2 + 𝑧𝑧 2 + 𝑢𝑢2 = 𝐾𝐾1 , 𝐾𝐾12 = 13 for the surfaces 𝜑𝜑1 and 𝑆𝑆 . Then the required
whence it follows that 3D equipotential surfaces 𝜑𝜑1 can equations of the surfaces reduce, respectively, to the
be represented as spheres with centres at 𝑝𝑝 = 0: following ones:
𝜑𝜑1 : 𝑥𝑥 2 + 𝑧𝑧 2 + 𝑢𝑢2 − 𝐾𝐾12 = 0, (5.8) 𝜑𝜑1 : 𝑥𝑥 2 + 𝑧𝑧 2 + 𝑢𝑢2 − 13 = 0, (5.11)
where 𝐾𝐾1 ∈ ℝ is an arbitrary constant. 𝑆𝑆: − 0,31𝑥𝑥 2 + 0,69𝑧𝑧 2 + 0,69𝑢𝑢2 = 0 (5.12)
 American Journal of Mathematical Analysis 25

Figure 2 demonstrates two orthogonal surfaces 𝜑𝜑1 and 𝑟𝑟 ∙ 𝐋𝐋og 𝑝𝑝 = 𝜙𝜙1(𝑟𝑟∙𝐋𝐋og 𝑝𝑝) (𝑎𝑎, 𝑏𝑏) + 𝜙𝜙2(𝑟𝑟∙𝐋𝐋og 𝑝𝑝) (𝑎𝑎, 𝑏𝑏) ∙ 𝑗𝑗,
𝑆𝑆 depicted, according to equations (5.11) and (5.12).
where in accordance with multiplication rule (3.5) we
have
�����������
𝜙𝜙1(𝑟𝑟∙𝐋𝐋og 𝑝𝑝) = 𝜙𝜙1(𝑟𝑟) 𝜙𝜙1(𝐋𝐋og 𝑝𝑝) − 𝜙𝜙2(𝑟𝑟) 𝜙𝜙 2(𝐋𝐋og 𝑝𝑝) ,

𝜙𝜙2(𝑟𝑟∙𝐋𝐋og 𝑝𝑝) = 𝜙𝜙1(𝑟𝑟) 𝜙𝜙2(𝐋𝐋og 𝑝𝑝) + �����������

𝜙𝜙1(𝐋𝐋og 𝑝𝑝) 𝜙𝜙2(𝑟𝑟) .
Substituting expressions (3.30), (5.1) and (5.2) into the
last two expressions, we finally obtain as follows:
 a+a
( a − a )  2v ln p + ( a − a ) arccos 
 2 p 
φ1( r ⋅L og p ) =
4v 2
a+a
bb arccos
2 p
− ,
v2
 a+a 
Figure 2. The quaternionic potential Log 𝑝𝑝 . Orthogonal stream and
b ( a − a ) arccos + v ln p 
2 p
potential surfaces: 𝑆𝑆 for 𝐾𝐾 2 = 0,69 and 𝜑𝜑1 for 𝐾𝐾12 = 13. The curves of φ2( r ⋅L og p ) =  2
.
intersection are highlighted by blue color v
Verifying the orthogonality of 𝜑𝜑1 and 𝑆𝑆 at the points It is easy to verify that these expressions are
𝑝𝑝 ∈ 𝐼𝐼 is essentially the same as in the example of 3D flow equal, respectively, to expressions for 𝜙𝜙1(Log 𝑝𝑝∙𝑟𝑟) and
modeling considered in [7]. We shall not dwell on this 𝜙𝜙2(Log 𝑝𝑝∙𝑟𝑟) , that is, the quaternionic product 𝑟𝑟 ∙ 𝐋𝐋og 𝑝𝑝 is
here. commutative.
To retain an analogy with a "planar" flow in the Using (1.1) and grouping terms with imaginary units
complex plane, we can regard the described 3D flow to 𝑖𝑖 , 𝑗𝑗 and 𝑘𝑘 , we get the following expression for the
the flow with rate of mass transport equal to 1, however a quaternionic potential 𝑟𝑟 ∙ Log 𝑝𝑝 in ℝ4 -representation:
further study of this analogy is beyond the scope of the
present paper. 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑟𝑟 ∙ 𝐋𝐋og 𝑝𝑝 = 𝜓𝜓1 + 𝜓𝜓2 𝑖𝑖 + 𝜓𝜓3 𝑗𝑗 + 𝜓𝜓4 𝑘𝑘,
where
5.2. Elementary Vortex Flow
 x y ln p
In the complex analysis the elementary vortex
ψ1 =
− arccos   , ψ 2 =,
 p  v
flow is represented by the principal value of the function
𝜓𝜓𝐶𝐶 (𝑧𝑧) = 𝑖𝑖Log 𝑧𝑧 , |𝑧𝑧| ≠ 0 , where 𝑖𝑖 denotes the habitual =ψ3
z ln p
= ,ψ 4
u ln p
,
imaginary unit [11]. This function satisfies known v v
complex Cauchy-Riemann's equations [1,11]. However
and 𝑣𝑣 is defined by (2.6), 𝑣𝑣 ≠ 0, |𝑝𝑝| ≠ 0.
the quaternionic function 𝑖𝑖 ∙ Log 𝑝𝑝, |𝑝𝑝| ≠ 0, doesn't satisfy
After transition 𝑦𝑦 = 0 (to 3D space), we have
the quaternionic system of generalized Cauchy-Riemann's
equations (1.3). ( r ⋅ L og p=| ϕ1 ( x, z , u ) + ϕ2 ( x, z , u ) i
As mentioned above (constructing rule 2.1), we can
replace the 𝑖𝑖 by the quaternionic "imaginary vector" + ϕ3 ( x, z , u ) j + ϕ4 ( x, z , u ) k ,
𝑟𝑟 = 𝜙𝜙1(𝑟𝑟) + 𝜙𝜙2(𝑟𝑟) ∙ 𝑗𝑗 defined by (3.29). The function 𝑟𝑟 where
satisfies equations (1.3):
1  x 
1) (𝜕𝜕𝑎𝑎 𝜙𝜙1(𝑟𝑟) | = (𝜕𝜕𝑏𝑏 𝜙𝜙2(𝑟𝑟) | = , ϕ1 ( x, z , u ) = (ψ 1 |= − arccos   , (5.13)
2b  x 2 + z 2 + u 2 
2) (𝜕𝜕𝑎𝑎 𝜙𝜙2(𝑟𝑟) | = − (𝜕𝜕𝑏𝑏 𝜙𝜙1(𝑟𝑟) | = 0,
𝜑𝜑2 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = (𝜓𝜓2 | = 0,
1
3) (𝜕𝜕𝑎𝑎 𝜙𝜙1(𝑟𝑟) | = (𝜕𝜕𝑏𝑏 𝜙𝜙2(𝑟𝑟) | = ,
2b z ln x 2 + z 2 + u 2
𝜑𝜑3 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = (𝜓𝜓3 | = ,
4) (𝜕𝜕𝑎𝑎 𝜙𝜙2(𝑟𝑟) | = − (𝜕𝜕𝑏𝑏 𝜙𝜙1(𝑟𝑟) | = 0. z2 + u2
Hence the function 𝑟𝑟 is ℍ -holomorphic.
Since the functions 𝑟𝑟 and Log 𝑝𝑝 are ℍ -holomorphic, u ln x 2 + z 2 + u 2
the function 𝑟𝑟 ∙ Log 𝑝𝑝 by virtue of Theorem 3.5 is 𝜑𝜑4 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = (𝜓𝜓4 | = .
also ℍ -holomorphic. Considering this function as the z2 + u2
quaternionic potential of elementary vortex flow, rewrite it Combining expression (1.4) for 𝑘𝑘 = 1 and (3.30), we
as follows: calculate the first derivative of 𝑟𝑟 as follows:
 26 American Journal of Mathematical Analysis

( r )' = (∂ φ ( ) + ∂ φ ( ) ) + (∂ φ ( ) + ∂ φ ( ) ) ⋅ j
a 1r a 1r a 2 r a 2 r
According to the quaternionic potential theory [7], we
get the following expression for the flow velocity vector
 4v 2 + ( a − a ) 2 4 v 2 + ( a − a ) 2  (field vector) in 3D space:
=  −  (5.14)

 8 v 3
8 v 3 
 ( F ( p=
) | (( r ⋅ L og p )' | (5.18)
 b (a − a ) b (a − a )  =( f1 | + ( f 2 | i + ( f3 | j + ( f 4 | k ,
+ − ⋅ j = 0
3
 4v 4v 3  where
By virtue of product rule 3.9 and zero result (5.14) the
f1 | (ψ
u2 + z2
first derivative of 𝑟𝑟 ∙ Log 𝑝𝑝 is the following: (= = '
1| , (5.19)
′ u 2 + x2 + z 2
(𝑟𝑟 ∙ Log 𝑝𝑝) = (𝑟𝑟)′ ∙ Log 𝑝𝑝 + 𝑟𝑟 ∙ (Log 𝑝𝑝) = 𝑟𝑟 ∙ 𝑝𝑝 . ′ −1
(𝑓𝑓2 | = −(𝜓𝜓2′ | = 0,
Given the rule of multiplication (3.5), we have as
− (ψ 3' | =
follows: xz
( f3 | = − , (5.20)
𝑟𝑟 ∙ 𝑝𝑝 −1
= �𝜙𝜙1(𝑟𝑟) + 𝜙𝜙2(𝑟𝑟) ∙ 𝑗𝑗� ∙ �𝜙𝜙1�𝑝𝑝 −1 � + 𝜙𝜙2�𝑝𝑝 −1 � ∙ 𝑗𝑗�
2
u +z 2
( u 2 + x2 + z 2 )
= 𝜙𝜙1�𝑟𝑟∙𝑝𝑝 −1 � + 𝜙𝜙2�𝑟𝑟∙𝑝𝑝 −1 � ∙ 𝑗𝑗,
− (ψ 4' | =
xu
( f4 | = − . (5.21)
where (
u 2 + z 2 u 2 + x2 + z 2 )
𝜙𝜙1�𝑟𝑟∙𝑝𝑝 −1 � = 𝜙𝜙1(𝑟𝑟) 𝜙𝜙1�𝑝𝑝 −1 � − 𝜙𝜙2(𝑟𝑟)���������
𝜙𝜙2(𝑝𝑝 −1 ) , (5.15)
Applying 3D gradient operator (5.7) directly to the
𝜙𝜙2�𝑟𝑟∙𝑝𝑝 −1 � = 𝜙𝜙1(𝑟𝑟) 𝜙𝜙2�𝑝𝑝 −1 � + ���������
𝜙𝜙1(𝑝𝑝 −1 ) 𝜙𝜙2(𝑟𝑟) . (5.16) function 𝜑𝜑1 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) defined by (5.13), we obtain the
following expression:
As shown in [6], the constituents of the function 𝑝𝑝−1
are the following: ∇3 𝜑𝜑1 = 𝜕𝜕𝑥𝑥 𝜑𝜑1 +𝜕𝜕𝑧𝑧 𝜑𝜑1 𝑗𝑗 + 𝜕𝜕𝑢𝑢 𝜑𝜑1 𝑘𝑘,
a b where
φ = ,φ = −
( ) ( aa + bb ) ( )
1 p −1 2 p −1 ( aa + bb )   x  u2 + z2
(5.17) ∂ xϕ1 =∂ x  − arccos    =2 ,
a b   x + z 2 + u 2
2   u + x 2 + z 2
φ = ,φ = − 
( ) ( aa + bb ) ( )
1 p −1 2 p −1 ( aa + bb )
.
  x 
Substituting expressions (3.30) and (5.17) into (5.15) ∂ zϕ1 =∂ z  − arccos  
 
 x + z + u  
2 2 2
and (5.16), we obtain the following expressions: 
xz
aa − a 2 + 2bb = − ,
φ
= φ=
( )
'
(
1 r ⋅ p −1 )
1( r ⋅L og p )
(
2v aa + bb )
,
u + z u 2 + x2 + z 2
2 2

b (a + a )   x 
φ
= φ=
' ∂ uϕ1 =∂ u  − arccos  
(
2 r ⋅ p −1 )
2( r ⋅L og p )
(
2v aa + bb )
.

  x + z + u
2 2 2 
 
xu
Substituting (1.1) into the last expressions and grouping = − .
terms with imaginary units 𝑖𝑖, 𝑗𝑗 and 𝑘𝑘, we get the following
expression for the quaternionic derivative of the function
(
u + z u 2 + x2 + z 2
2 2
)
𝑟𝑟 ∙ Log 𝑝𝑝 in ℝ4 -representation: The obtained expressions for 𝜕𝜕𝑥𝑥 𝜑𝜑1 , 𝜕𝜕𝑧𝑧 𝜑𝜑1 , 𝜕𝜕𝑢𝑢 𝜑𝜑1
(𝑟𝑟 ∙ Log 𝑝𝑝) = 𝑟𝑟 ∙ 𝑝𝑝 ′ −1
= 𝜓𝜓1′ + 𝜓𝜓2′ 𝑖𝑖 + 𝜓𝜓3′ 𝑗𝑗 + 𝜓𝜓4′ 𝑘𝑘, coincide with expressions (5.19), (5.20), (5.21),
respectively. Thus, the steady state fluid flow,
where corresponding to the quaternionic potential 𝑟𝑟 ∙ Log(𝑝𝑝) ,
is a potential one in 𝐺𝐺4 ∈ ℍ, except at the singularity point
u2 + y2 + z2 𝑝𝑝 = 0.
ψ 1' = ,
u 2 + x2 + y 2 + z 2 According to [7], we get equations of 3D equipotential
surfaces 𝜑𝜑1 and stream surfaces 𝑆𝑆 for the quaternionic
xy potential Log 𝑝𝑝.
ψ 2' = ,
u +y +z
2 2 2
( u 2 + x2 + y 2 + z 2 ) For 3D equipotential surfaces 𝜑𝜑1 we have the equation
 x 
xz ϕ1 ( x, z , u ) =
− arccos  =
K1 ,
ψ 3' = ,  x 2 + z 2 + u 2 
(
u 2 + y 2 + z 2 u 2 + x2 + y 2 + z 2 ) where 𝐾𝐾1 ∈ ℝ is a constant. Leaving the designations for
xu constants unchanged, we reduce it to the following:
ψ 4' = .
(
u 2 + y 2 + z 2 u 2 + x2 + y 2 + z 2 ) 𝜑𝜑1 : (𝐾𝐾12 − 1)𝑥𝑥 2 + 𝐾𝐾12 𝑧𝑧 2 + 𝐾𝐾12 𝑢𝑢2 = 0. (5.22)
 American Journal of Mathematical Analysis 27

The constant 𝐾𝐾1 is restricted to 𝐾𝐾1 ≤ 1, since it lies in expression becomes opposite in sign (remaining equal in
the range of values of the function 𝐶𝐶𝐶𝐶𝐶𝐶. absolute value), if signs of the variables 𝑥𝑥 and 𝑦𝑦 are
It easy to see that equation (5.22) is the equation of 3D changed simultaneously.
conical surfaces that have the axis 0𝑥𝑥 and the vertexes at Since components (5.19), (5.20), and (5.21) of
the origin. the velocity (𝐹𝐹(𝑝𝑝)| do not change their signs to opposite
To define the stream surfaces 𝑆𝑆 we use the generating ones due to the simultaneous change in the signs of 𝑥𝑥, 𝑧𝑧, 𝑢𝑢,
function (5.9). From the equations 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = 𝐾𝐾3 when expression (5.18) is not good enough for modeling
𝑢𝑢 = 0 and 𝜑𝜑4 (𝑥𝑥, 𝑧𝑧, 𝑢𝑢) = 𝐾𝐾4 when 𝑧𝑧 = 0 it follows that the an elementary 3D vortex flow. To correct this we
desired stream surface has in the planes 𝑥𝑥0𝑧𝑧 ∶ 𝜉𝜉 = 𝑥𝑥 + 𝑧𝑧𝑧𝑧 introduce the piecewise smooth function 𝑟𝑟 ∙ Log 𝑝𝑝 as
and 𝑥𝑥0𝑢𝑢 ∶ 𝜉𝜉 = 𝑥𝑥 + 𝑢𝑢𝑢𝑢 the following equations of traces, follows:
respectively: 𝑟𝑟 ∙ Log 𝑝𝑝, 𝑢𝑢 ≥ 0
𝑥𝑥0𝑧𝑧: 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) = ln√𝑥𝑥 2 + 𝑧𝑧 2 = 𝐾𝐾3 , 𝜓𝜓𝐻𝐻 (𝑝𝑝) = � . (5.26)
−𝑟𝑟 ∙ Log 𝑝𝑝, 𝑢𝑢 < 0
𝑥𝑥0𝑢𝑢: 𝜑𝜑4 (𝑥𝑥, 𝑢𝑢) = ln√𝑥𝑥 2 + 𝑢𝑢2 = 𝐾𝐾4 , It is evident that this function, according to rule 3.2
where 𝐾𝐾3 , 𝐾𝐾4 ∈ ℝ are constants. (at 𝑐𝑐 = 1 , 𝑐𝑐 = −1 ), is piecewise ℍ -holomorphic. The
Leaving the designations for constants unchanged, we
equations of equipotential surfaces 𝜑𝜑1 and stream surfaces
reduce these equations to the following ones:
𝑆𝑆 are the same for both parts of expression (5.26).
𝑥𝑥0𝑧𝑧: 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) = 𝑥𝑥 2 + 𝑧𝑧 2 = 𝐾𝐾3 2 , The corresponding function 𝜑𝜑1 is the following:
𝑥𝑥0𝑢𝑢: 𝜑𝜑4 (𝑥𝑥, 𝑢𝑢) = 𝑥𝑥 2 + 𝑢𝑢2 = 𝐾𝐾4 2 ,
which represent the circles with centres at the origin. ⎡−arccos[ x
], 𝑢𝑢 ≥ 0
Putting 𝐾𝐾3 = 𝐾𝐾4 = 𝐾𝐾 and rotating counterclockwise the ⎢ u + x2 + z 2
2
π 𝜑𝜑1 = ⎢ .
plane 𝑥𝑥0𝑧𝑧 around the 𝑥𝑥 -axis by the angle , we can ⎢ x
2 ⎢ arccos[ ], 𝑢𝑢 < 0
obtain the trace 𝜑𝜑4 (𝑥𝑥, 𝑢𝑢) from 𝜑𝜑3 (𝑥𝑥, 𝑧𝑧). Hence the desired ⎣ u 2 + x2 + z 2
stream surfaces 𝑆𝑆 when considering the quaternionic
potential 𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑟𝑟 ∙ Log 𝑝𝑝, in the simplest case, are the Ultimately, we can write the flow velocity vector
surfaces of revolution [12] around the 𝑥𝑥-axis. Analogously (𝐹𝐹(𝑝𝑝)| as follows:
to the previous subsection, we get these from the trace {(𝑓𝑓1 |, (𝑓𝑓3 |,(𝑓𝑓4 |}, 𝑢𝑢 ≥ 0
(𝐹𝐹(𝑝𝑝)| = ∇3 𝜑𝜑1 = � (5.27)
𝜑𝜑3 (𝑥𝑥, 𝑧𝑧) by replacing the variable 𝑧𝑧 by √𝑧𝑧 2 + 𝑢𝑢2 : {−(𝑓𝑓1 |, −(𝑓𝑓3 |, − (𝑓𝑓4 |}, 𝑢𝑢 < 0
𝑆𝑆: 𝑥𝑥 2 + 𝑧𝑧 2 + 𝑢𝑢2 − 𝐾𝐾 2 = 0. (5.23) where the vector components in braces are defined by
The surfaces (5.23) are 3D spheres with centres at the expressions (5.19), (5.20), (5.21).
origin. The streamlines are situated on these imagined 3D Figure 3 demonstrates two orthogonal surfaces 𝜑𝜑1 and
stream surfaces 𝑆𝑆. 𝑆𝑆 computed by using formulae (5.24) and (5.25).
Comparing expressions (5.8), (5.10) with expressions
(5.22), (5.23), we see that the equations of 𝜑𝜑1 and 𝑆𝑆 for
quaternionic potentials Log 𝑝𝑝 and 𝑟𝑟 ∙ Log 𝑝𝑝 switch places
with each other, just as in complex analysis [11]. At that
the coefficients 𝐾𝐾 and 𝐾𝐾1 is changed with each other.
Analogously to the previous subsection, we obtain the
equation of a curve 𝐼𝐼 of intersection of surfaces 𝜑𝜑1 and 𝑆𝑆
from equations (5.22) and (5.23) as follows:

zI + uI =
2 2
rI2 =K 2
− xI2 =
(1 − K12 ) xI2
.
K12
Assuming, just as in the case of the quaternionic
potential Log 𝑝𝑝, that 𝑟𝑟𝐼𝐼 = 2, 𝑥𝑥𝐼𝐼 = 3 we get the following
values of constants: 𝐾𝐾12 = 0,69 and 𝐾𝐾 2 = 13 for the
surfaces 𝜑𝜑1 and 𝑆𝑆 in the case of quaternionic potential
𝑟𝑟 ∙ Log 𝑝𝑝. Then equations (5.22) and (5.23) reduce to the
following:
𝜑𝜑1 : − 0,31𝑥𝑥 2 + 0,69𝑧𝑧 2 + 0,69𝑢𝑢2 = 0, (5.24)
2 2 2
𝑆𝑆: = 𝑥𝑥 + 𝑧𝑧 + 𝑢𝑢 − 13 = 0. (5.25) Figure 3. The quaternionic potential 𝑟𝑟Log 𝑝𝑝. Example of orthogonal
stream and potential surfaces: 𝑆𝑆 for 𝐾𝐾 2 = 13 and 𝜑𝜑1 for 𝐾𝐾12 = 0,69. The
In complex analysis the velocity vector of the intersection curves are highlighted by blue color
elementary vortex flow is represented [11] by the
i ( x + iy ) The flow velocity vectors (𝐹𝐹(𝑝𝑝)| indicated with arrows
�����������
expression 𝐹𝐹(𝑧𝑧) = (𝑖𝑖Log 𝑧𝑧)′ = − . Moving of on the surface 𝑆𝑆 are computed by using formula (5.27).
2
z They have the same length, since have the same
the flow in a circle in the complex plane is due to the fact absolute value on the sphere, depending only on the radius
that the flow velocity vector 𝐹𝐹(𝑧𝑧) in accordance with this 𝑅𝑅𝑆𝑆 of this sphere:
28 American Journal of Mathematical Analysis

( F ( p ) | = ( f1 |2 + ( f3 |2 + ( f 4 |2 = Appendix
1 1
= .
u +x +z
2 2 2
RS2
We present here the only five computing programs to
Such a vector picture could be imagined as rotating a illustrate processing of the ℍ -holomorphic function
complex plane (with depicted moving of the flow in a 2
𝜓𝜓𝐻𝐻 (𝑝𝑝) = 𝑒𝑒 𝑝𝑝 (see Example 3.13).
circle) counterclockwise around the 𝑥𝑥-axis by the angles Program 1 computes the functions f1 (=𝜙𝜙1 or = 𝜙𝜙1′ for
from 0 to 𝜋𝜋 radians. We do not present an overall picture the derivative) and f2 (= 𝜙𝜙2 or = 𝜙𝜙2′ ) in the case when the
of the velocity vectors for quaternionic potential 𝑟𝑟 ∙ Log 𝑝𝑝, function in question (or its derivative) does not have
since it is the same as on the depicted sphere 𝑆𝑆 distributed operations of the quaternionic multiplication explicitly.
at the whole space. 2
The function 𝑒𝑒 𝑝𝑝 is such a case. The got functions f1 and
Just as in the previous subsection we shall not 2
dwell on verifying the orthogonality of 𝜑𝜑1 and 𝑆𝑆 at the f2 for 𝑒𝑒 𝑝𝑝 are further used as input data for Program 5,
points 𝑝𝑝 ∈ 𝐼𝐼. which tests whether a quaternionic function is
By creating Figure 1, Figure 2, Figure 3 the computing ℍ -holomorphic. The variables 𝑎𝑎, 𝑎𝑎, � 𝑏𝑏, 𝑏𝑏� are denoted
system Wolfram Mathematica® [10] was used. by a, ac, b, bc, respectively. The conjugate functions have
The presented results give a reason for building other extra sign "c".
segments of quaternionic analysis similar to complex ones. In the case when a quaternionic function does contain
To avoid an error-prone cumbersome and tedious explicitly the quaternionic multiplication operations
manual procedure of calculations with ℍ -holomorphic Program 4 is used. Program 4 calculates f1 and f2 to be
functions we have developed the special software pack further used in Program 5 and also tests whether the
written in the programming language Wolfram®. By using quaternionic multiplication in this case behaves as
this pack all calculations can be immediately carried out. commutative. The input data for Program 4 are the
This pack allows us to test the ℍ −holomorphy of any components inia1, inia2 of the first factor of the
quaternionic function, the multiplicative commutativity of quaternionic product (inia1+ inia2∙ 𝑗𝑗) ∙( inib1+inib2∙ 𝑗𝑗) and
ℍ -holomorphic functions, calculate expressions for them also the components inib1, inib2 of the second factor,
2
and their derivatives, including the quaternionic potentials calculated by Program 2 for the first derivative of 𝑒𝑒 𝑝𝑝 or
2
and expressions for 3D potential fields, field divergence Program 3 for the second derivative of 𝑒𝑒 𝑝𝑝 .
and vortex density as well as get pictures of 3D potential All the programs are written each in a separated cell. If
fields such as in Figure 1, Figure 2, Figure 3. A lot of needed the programmes can be directly copied cell by cell
examples of ℍ -holomorphic functions is considered in from this text or its PDF into input cells (only one
this pack. The part of developed programmes is presented program into one cell!) of the opened Notebook blank .nb.
in Appendix below. The programmes are ready for use immediately after
copying. The functions without semicolon at the end of
them must be in separate lines. They can be also used with
References initial data of other functions. To launch each program it
is needed to click on it's cell, and then holding down the
[1] Mathews, J. H., Howell, R. W., Complex Analysis for Shift key while press the Enter key. For the sake of brevity,
Mathematics and Engineering, 3rd ed, Jones and Bartlett the results of only calculating for the second derivative of
Publishers, Boston-Toronto-London-Singapore, 1997. 2
[2] Sudbery, A., "Quaternionic analysis", Math. Proc. Camb. Phil. 𝑒𝑒 𝑝𝑝 are below represented. First the cell with Program 3 is
Soc., 85 (1979), 199-225. evaluated, then the cell with Program 4, and finally the
[3] Dzagnidze, O., "On the differentiability of quaternion functions", cell with Program 5.
arXiv: math.CV, March 2012. Available: arxiv.org/abs/1203.5619. (*Program 1*)
[4] Parfenov, M., "On Properties of Holomorphic Functions in
Quaternionic Analysis", American Journal of Mathematical
(*The function e^p^2, input data for Program 5*)
Analysis, Vol. 5, No. 1, pp. 17-24, July 2017. Clear[f1,f2,f1c,f2c];
[5] Khaled Abdel-Khalek, "Quaternion Analysis", arXiv:hep-th, July quatfunctiontested:=Evaluate[e^p^2];
1996. Available: arxiv.org/abs/hep-th/9607152v2. v=(1/2)*Sqrt[4*(a*ac+b*bc)-(a+ac)^2];
[6] Parfenov, M., "Adequate quaternionic generalization of complex β1=E^((a^2+ac^2-2 b bc)/2);θ=(a+ac)*v;
differentiability", viXra: Functions and Analysis, Jan 2017.
Available: vixra.org/abs/1609.0006 f1[a_,ac_,b_,bc_]:=β1*(Cos[θ]+((a-ac)Sin[θ])/(2*v));
[7] Parfenov, M., "A Quaternionic Potential Conception with f2[a_,ac_,b_,bc_]:=(β1*Sin[θ]*b)/v;
Applying to 3D Potential Fields", American Journal of f1c[a_,ac_,b_,bc_]:=β1*(Cos[θ]+((ac-a)Sin[θ])/(2*v));
Mathematical Analysis, Vol. 7, No. 1, p.p. 1-10, April 2019. f2c[a_,ac_,b_,bc_]:=(β1*Sin[θ]*bc)/v;
[8] Kantor, I. L., Solodovnikov, A. S.. Hypercomplex numbers. An Print["Constituents f1,f2:"];
Elementary Introduction to Algebras. Springer-Verlag, 1989
[9] Parfenov, M., "The Similarity between Rules for Essentially
f1[a,ac,b,bc]
Adequate Quaternionic and Complex Differentiation", viXra: f2[a,ac,b,bc]
Functions and Analysis, 2018. Available: vixra.org/abs/1806.0239.
[10] Wellin, P.R., Gaylord, R. J., Kamin, S.N., An Introduction to (*Program 2*)
Programming with Mathematica, 3rd ed, Cambridge University (*The first derivative (e^p^2)´= e^p^2* 2p, initial data
Press, New York, 2005.
[11] Lavrentiev M. A., Shabat B. V., Methods of the Theory of for Program 4*)
Complex Variable, Nauka, Moscow, 1973. (In Russian) Clear[f1,f2,f1c,f2c,inia1,inia1c,inia2,inia2c,inib1,inib1c,
[12] Podoksenov, M. N., Prokhozhiy, S. A., Analytic Geometry in the inib2,inib2c];
space, "VSU named after P. M. Masherov" Publishers, Vitebsk, quatfunctiontested:=Evaluate[2e^p^2*p];
2013.
v=(1/2)*Sqrt[4*(a*ac+b*bc)-(a+ac)^2];
 American Journal of Mathematical Analysis 29

β1=E^((a^2+ac^2-2 b bc)/2);θ=(a+ac)*v; 1 + 2(𝑎𝑎2 − 𝑏𝑏bc),

inia1[a_,ac_,b_,bc_]:=β1*(Cos[θ]+((a-ac)Sin[θ])/(2*v)); 2(𝑎𝑎 + ac)𝑏𝑏
inia1c[a_,ac_,b_,bc_]:=β1*(Cos[θ]+((ac-
a)Sin[θ])/(2*v)); Remark. The results for inia1, inia2 coincide,
inia2[a_,ac_,b_,bc_]:=(β1*Sin[θ]*b)/v; respectively, with above expressions (3.36), (3.37) for
inia2c[a_,ac_,b_,bc_]:=(β1*Sin[θ]*bc)/v ; 𝜙𝜙1(e) and 𝜙𝜙2(e) multiplied by 2. The results for inib1,
inib1[a_,ac_,b_,bc_]:=2*( a^2-b*bc); inib2 coincide, respectively, with above expressions (3.28)
inib1c[a_,ac_,b_,bc_]:=2*( ac^2-b*bc); for 𝜙𝜙1�𝑝𝑝 2 � and 𝜙𝜙2�𝑝𝑝 2 � multiplied by 2, with addition of +1
inib2 [a_,ac_,b_,bc_]:= 2*(a+ac)*b; to 2𝜙𝜙1�𝑝𝑝 2 � .
inib2c [a_,ac_,b_,bc_]:= 2*(a+ac)*bc;
Print["Constituents inia1,inia2:"]; (*Program 4*)
inia1[a,ac,b,bc] (*Input data for Program 5 and commutativity testing*)
inia2[a,ac,b,bc] vorf1[a_,ac_,b_,bc_]:=inia1[a,ac,b,bc]*inib1[a,ac,b,bc]-
Print["Constituents inib1,inib2:"]; inia2[a,ac,b,bc]*inib2c[a,ac,b,bc];
inib1[a,ac,b,bc] vorf1c[a_,ac_,b_,bc_]:=inia1c[a,ac,b,bc]*inib1c[a,ac,b,
inib2[a,ac,b,bc] bc]-inia2c[a,ac,b,bc]*inib2[a,ac,b,bc];
f1[a_,ac_,b_,bc_]:=FullSimplify[vorf1[a,ac,b,bc]];
(*Program 3*) f1c[a_,ac_,b_,bc_]:=FullSimplify[vorf1c[a,ac,b,bc]];
(*The second derivative (e^p^2)´´= 2e^p^2*(2p^2+1), q1q2f1:=f1[a,ac,b,bc]; q1q2f1c:=f1c[a,ac,b,bc];
initial data for Program 4*) vorf2[a_,ac_,b_,bc_]:=inia1[a,ac,b,bc]*inib2[a,ac,b,bc]
Clear[inia1,inia1c,inia2,inia2c,inib1,inib1c,inib2,inib2c +
]; inia2[a,ac,b,bc]*inib1c[a,ac,b,bc];
v=(1/2)*Sqrt[4*(a*ac+b*bc)-(a+ac)^2];θ=(a+ac)*v; vorf2c[a_,ac_,b_,bc_]:=inia1c[a,ac,b,bc]*inib2c[a,ac,b,
mu:=Cos[θ]+((a-ac)*Sin[θ])/(2*v); muc:=Cos[θ]+((ac-a) bc]+
*Sin[θ])/(2*v); beta:=E^((a^2+ac^2-2 b bc)/2); inia2c[a,ac,b,bc]*inib1[a,ac,b,bc];
quatfunctiontested:=Evaluate[2*e^p^2*(2p^2+1)]; f2[a_,ac_,b_,bc_]:=FullSimplify[vorf2[a,ac,b,bc]];
inia1[a_,ac_,b_,bc_]:=2*beta*mu; f2c[a_,ac_,b_,bc_]:=FullSimplify[vorf2c[a,ac,b,bc]];
inia1c[a_,ac_,b_,bc_]:=2*beta*muc; Print["Result of computing: Constituents f1, f2 of
inia2[a_,ac_,b_,bc_]:=(2*beta*Sin[θ]*b)/v; ",quatfunctiontested, ":"];
inia2c[a_,ac_,b_,bc_]:=(2*beta*Sin[θ]*bc)/v ; f1[a,ac,b,bc]
inib1[a_,ac_,b_,bc_]:= 1+2*(a^2-b*bc); f2[a,ac,b,bc]
inib1c[a_,ac_,b_,bc_]:=1+2*(ac^2-b*bc); q1q2f2:=f2[a,ac,b,bc]; q1q2f2c:=f2c[a,ac,b,bc];
inib2 [a_,ac_,b_,bc_]:= 2*b*(a+ac); vorϕf1[a_,ac_,b_,bc_]:=inia1[a,ac,b,bc]*inib1[a,ac,b,b
inib2c [a_,ac_,b_,bc_]:=2*bc*(a+ac) ; c]-inia2c[a,ac,b,bc]*inib2[a,ac,b,bc];
Print["Constituents inia1, inia2 :"]; vorϕf1c[a_,ac_,b_,bc_]:=inia1c[a,ac,b,bc]*inib1c[a,ac,
inia1[a,ac,b,bc] b,bc]-inia2[a,ac,b,bc]*inib2c[a,ac,b,bc];
inia2[a,ac,b,bc] ϕf1[a_,ac_,b_,bc_]:=FullSimplify[vorϕf1[a,ac,b,bc]];
Print["Constituents inib1, inib2 :"]; ϕf1c[a_,ac_,b_,bc_]:=FullSimplify[vorϕf1c[a,ac,b,bc]];
inib1[a,ac,b,bc] q2q1f1:=ϕf1[a,ac,b,bc]; q2q1f1c:=ϕf1c[a,ac,b,bc];
inib2[a,ac,b,bc] vorϕf2[a_,ac_,b_,bc_]:=inia1c[a,ac,b,bc]*inib2[a,ac,b,
bc]+
Outputs: inia2[a,ac,b,bc]*inib1[a,ac,b,bc];
Constituents inia1, inia2: vorϕf2c[a_,ac_,b_,bc_]:=inia1[a,ac,b,bc]*inib2c[a,ac,b,
bc]+

2e 2
(
1 2
a + ac2 − 2bbc ) inia2c[a,ac,b,bc]*inib1c[a,ac,b,bc];
ϕf2[a_,ac_,b_,bc_]:=FullSimplify[vorϕf2[a,ac,b,bc]];
 1   ϕf2c[a_,ac_,b_,bc_]:=FullSimplify[vorϕf2c[a,ac,b,bc]];
cos  2 ( a + ac ) − ( a + ac ) + 4 ( aac + bbc ) 
2
 q2q1f2:=ϕf2[a,ac,b,bc]; q2q1f2c:=ϕf2c[a,ac,b,bc];
    coins1:=True===FullSimplify[q1q2f1==q2q1f1];
   
⋅  a − ac sin 1 

( )  2 ( a + ac ) − ( a + ac ) 2
+ 4 ( aac + bbc )  
coins2:=True===FullSimplify[q1q2f2==q2q1f2];
coins:=coins1&&coins2;
+ 
− ( a + ac ) + 4 ( aac + bbc )
2
 

1
4be 2
( a2 +ac2 −2bbc)
If[coins,Print["Result of testing: Quaternionic
1  multiplication in the case of ",quatfunctiontested," is
sin  ( a + ac ) − ( a + ac ) + 4 ( aac + bbc ) 
2
 2  commutative"],Print["Result of testing: Quaternionic
⋅ multiplication in the case of ",quatfunctiontested," is not
− ( a + ac ) + 4 ( aac + bbc )
2
commutative"]]

Outputs:
Constituents inib1, inib2:
30 American Journal of Mathematical Analysis

Result of computing: Constituents f1, f2 of df2c:=D[f2c[a,ac,b,bc],{bc}];

2
2𝑒𝑒 𝑝𝑝 ·(2𝑝𝑝2 +1): trdf2c:=df2c/.{a->x,ac->x};
sftrdf2c:=FullSimplify[trdf2c];

2e 2
1
( a2 +ac2 −2bbc) trmdf1da:=PowerExpand[sftrdf1[[1]]]//Expand;
trmdf2cdbc:=PowerExpand[sftrdf2c]//Expand;

( ) 
1 eq1:=True===FullSimplify[trmdf1da==trmdf2cdbc];
 1 + 2a − 2bbc cos[ 2 (a + ac) − ( a − ac ) + 4bbc ]
2 2
trmdf2da:=PowerExpand[sftrdf2[[1]]]//Expand;
 
(( ) )
trmdf1cdbc:=PowerExpand[-sftrdf1c]//Expand;
  1 + 2a 2 ( a − ac ) − 2 ( 3a + ac ) bbc   eq2:=True===FullSimplify[trmdf2da==trmdf1cdbc];
   
⋅    trmdf2db:=PowerExpand[sftrdf2[[3]]]//Expand;

 ⋅ sin ( a + ac ) − ( a − ac ) + 4bbc 
1 2   eq3:=True===FullSimplify[trmdf1da==trmdf2db];
   2   
 trmdf2dac:=PowerExpand[sftrdf2[[2]]]//Expand;
+ 
  trmdf1dbc:=PowerExpand[-sftrdf1[[2]]]//Expand;
 − ( a − ac ) 2
+ 4bbc  eq4:=True===FullSimplify[trmdf2dac==trmdf1dbc];
If[eq1∧eq2∧eq3∧eq4,Print["Result of testing: The
( a + ac ) − ( a − ac )2 + 4bbc  quaternionic function ",quatfunctiontested," is H-
 
  holomorphic"],Print["Result of testing: The quaternionic
4be 2
1 2
( 2
)
a + ac − 2bbc ⋅ cos



2
1
( 2 
a + ac ) − ( a − ac ) + 4bbc  

function ",quatfunctiontested,"
  is not H-holomorphic"]];
2

(
− ( a − ac ) + 4bbc  + 1 + a + ac − 2bbc
2 2
) 

If[eq1,,Print["The equation 1) is not satisfied"]]
If[eq2,,Print["The equation 2) is not satisfied"]]
 1 
⋅ sin  2 ( a + ac ) − ( a − ac ) + 4bbc  
2 If[eq3,,Print["The equation 3) is not satisfied"]]
  If[eq4,,Print["The equation 4) is not satisfied"]]
Result of testing: Quaternionic multiplication in the Output:
2
case of 2𝑒𝑒 𝑝𝑝 ·(2𝑝𝑝2 +1) is commutative. 2
Result of testing: The quaternionic function 2𝑒𝑒 𝑝𝑝 ∙ (1 +
Remark. The results for f1 and f2 (constituents of the 2𝑝𝑝2 ) is H- holomorphic.
2 ′′ ′′ ′′
second derivative �𝑒𝑒 𝑝𝑝 � = f1 + f2 ∙ 𝑗𝑗 = 𝜙𝜙1(e) + 𝜙𝜙2(e) ∙
2 2
𝑗𝑗 = 2𝑒𝑒 𝑝𝑝 ·(2𝑝𝑝2 +1)) coincide with above expressions (3.45) We see that the second derivative of 𝑒𝑒 𝑝𝑝 is ℍ -
′′ ′′
and (3.46) for 𝜙𝜙1(e) and 𝜙𝜙2(e) , respectively. holomorphic.

(*Program 5*) Remark. We can also verify the ℍ -holomorphy of the

2
(*Testing H-holomorphy of quaternionic functions by function 𝑒𝑒 𝑝𝑝 by launching Program 1 and than Program 5
equations (1.3)*) as well as the ℍ -holomorphy of the first derivative
2 ′ 2
df1:=D[f1[a,ac,b,bc],{{a,bc}}]; �𝑒𝑒 𝑝𝑝 � = 2𝑒𝑒 𝑝𝑝 ∙ 𝑝𝑝 by launching Program 2, then Program
trdf1:=df1/.{a->x,ac->x}; 4 and finally Program 5.
sftrdf1:=FullSimplify[trdf1];
df1c:=D[f1c[a,ac,b,bc],{bc}]; The complete pack of programmes entitled "Processing
trdf1c:=df1c/.{a->x,ac->x}; of the ℍ -holomorphic functions" is scheduled to be
sftrdf1c:=FullSimplify[trdf1c]; shortly published.
df2:=D[f2[a,ac,b,bc],{{a,ac,b}}];
trdf2:=df2/.{a->x,ac->x};
sftrdf2:=FullSimplify[trdf2];

© The Author(s) 2020. This article is an open access article distributed under the terms and conditions of the Creative Commons
Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).