Article

On the Relation between Lambert W-Function and Generalized

Hypergeometric Functions
Pushpa Narayan Rathie 1,† and Luan Carlos de Sena Monteiro Ozelim 2, *,†

1 Department of Statistics, University of Brasilia, Brasília 70910-900, Brazil

2 Department of Civil and Environmental Engineering, University of Brasilia, Brasília 70910-900, Brazil
* Correspondence: [email protected] or [email protected]
† These authors contributed equally to this work.

Abstract: In the theory of special functions, finding correlations between different types of functions is
of great interest as unifying results, especially when considering issues such as analytic continuation.
In the present paper, the relation between Lambert W-function and generalized hypergeometric func-
tions is discussed. It will be shown that it is possible to link these functions by following two different
strategies, namely, by means of the direct and inverse Mellin transform of Lambert W-function and
by solving the trinomial equation originally studied by Lambert and Euler. The new results can be
used both to numerically evaluate Lambert W-function and to study its analytic structure.

Keywords: Lambert W-function; H-function; R-function; Mellin transform; Lagrange’s inversion

theorem

1. Introduction
The study of special functions and their applications has increasingly grown in the
Citation: Rathie, P.N.; Ozelim, last 50 years. This comes together with the advent of powerful computation techniques
L.C.d.S.M. On the Relation between
and devices which have allowed applied scientists to envision real-world applications to
Lambert W-Function and
this class of functions.
Generalized Hypergeometric
Even though great work has been carried out in applying the existing theory of special
Functions. Stats 2022, 5, 1212–1220.
functions to physical and engineering problems, theoretical studies have also been taken
https://doi.org/10.3390/
into account.
stats5040072
About 25 years ago, the interest in the so-called Lambert W-function tremendously
Academic Editor: Wei Zhu increased as important physical applications and the basic theory of this function were dis-
Received: 2 October 2022
cussed in the seminal work of [1]. The fact that Lambert W-function is linked to exponential-
Accepted: 19 November 2022
like equations provides a great number of applications in almost every branch of science,
Published: 23 November 2022
as this type of equation is common.
Apparently unrelated to Lambert W-function, the so-called generalized hypergeo-
Publisher’s Note: MDPI stays neutral
metric functions have been considered as solutions to ordinary and partial differential
with regard to jurisdictional claims in
equations [2]. This generalized class of functions has as special cases a great number
published maps and institutional affil-
of elementary and special functions, leading researchers to recognize such functions as
iations.
higher-order special functions.
So far, no relation between Lambert W-function and hypergeometric-type functions
has been established. In the theory of special functions, finding such correlations is of great
Copyright: © 2022 by the authors.
interest, leading to more compact and rigorous representations. This approach is frequent
Licensee MDPI, Basel, Switzerland. in the literature as seen, for example, in [3–5]. In addition, by connecting functions which
This article is an open access article apparently have no relation, important analytical continuation strategies can be envisioned.
distributed under the terms and In the present paper, the relation between Lambert W-function and generalized hyper-
conditions of the Creative Commons geometric functions is discussed. It will be shown that this link can be accomplished in
Attribution (CC BY) license (https:// two different ways, namely, by solving the trinomial equation originally studied by Euler
creativecommons.org/licenses/by/ in [6] and by Lambert; and by means of the direct and inverse Mellin transform of Lambert
4.0/). W-function.

Stats 2022, 5, 1212–1220. https://doi.org/10.3390/stats5040072 https://www.mdpi.com/journal/stats

Stats 2022, 5 1213

The paper is organized as follows: Section 2 presents some general concepts about
generalized hypergeometric functions, while Section 3 introduces the Lambert-W function
and its application to solve Lambert’s trinomial equation. Section 4 presents the solution of
Lambert’s trinomial equation both in terms of hypergeometric functions and in terms of
the Lambert-W function. Then, Section 5 indicates how general hypergeometric functions
and the Lambert-W function can be related while represented as contour integrals. Finally,
Section 6 presents the conclusions of the present work.

2. Generalized Hypergeometric Functions

In the present paper, the H-function is the first class of generalized hypergeometric
functions which will be explored. This function is defined as a contour complex integral
by [2]:
  
m,n
 ( a 1 , A ), . . . , ( a n , A n ), ( a n +1 , A n +1 ), . . . , ( a p , A p )
H p,q z 

(b1 , B1 ), . . . , (bm , Bm ), (bm+1 , Bm+1 ), . . . , (bq , Bq )
1
Z
= h(s)z−s ds, (1)
2πi L
where:
m n
∏ Γ ( b j + B j s ) ∏ Γ (1 − a j − A j s )
j =1 j =1
h(s) = q p (2)
∏ Γ (1 − b j − B j s ) ∏ Γ( a j + A j s)
j = m +1 j = n +1

and p, q, m, n are integers such that 0 ≤ m ≤ q and 0 ≤ n ≤ p, A j and Bj are positive real
quantities, and all the a j and b j are complex numbers. The contour L runs from c − i∞ to
c + i∞ such that the poles of Γ(b j + Bj s), j = 1, . . . , m lie to the left of L and the poles of
Γ(1 − a j − A j s), j = 1, . . . , n lie to the right of L.
By performing the variable change s → −r and adjusting the contour L to L∗ , where
the integral runs from c∗ − i∞ to c∗ + i∞, the H-function can be alternatively defined as:
  
m,n
 ( a 1 , A ), . . . , ( a n , A n ), ( a n +1 , A n +1 ), . . . , ( a p , A p )
H p,q z  
(b1 , B1 ), . . . , (bm , Bm ), (bm+1 , Bm+1 ), . . . , (bq , Bq )
m n
∏ Γ ( b j − B j r ) ∏ Γ (1 − a j + A j r )
1
Z
j =1 j =1
= q p zr dr, (3)
2πi L∗
∏ Γ (1 − b j + B j r ) ∏ Γ( a j − A j r )
j = m +1 j = n +1

for which the same parameter domains’ restrictions apply.

By considering the definition in (3), the H-function can be expressed in computable
form as [2]:
m
When the poles of ∏ Γ(b j − Bj r ) are simple, we have:
j =1
Stats 2022, 5 1214

m  
bh + ν
m ∞
∏ Γ b j − B j
Bh
j =16 = h
H p q (z) = ∑ ∑
mn
q  ×
bh + ν
h =1 ν =0
∏ Γ 1 − b j + B j Bh
j = m +1
!
n
bh + ν
∏ Γ 1 − a j + A j Bj
j =1 (−1)ν z(bh +ν)/Bh
× p   (4)
b +ν ν!Bh
∏ Γ a j − A j hBh
j = n +1

p q
for z 6= 0 if δ > 0 and for 0 < |z| < D −1 if δ = 0, where δ = ∑ j=1 Bj − ∑ j=1 A j and
p Aj q Bj
D = ∏ j =1 A j / ∏ j =1 B j .
n
When the poles of ∏ Γ(1 − a j + A j r ) are simple, we have
j =1

n
1 − ah + ν
 
∏ Γ 1 − a j − A j Ah
n ∞ j =16 = h
H p q (z) = ∑ ∑
mn
p  ×
1 − ah + ν

h =1 ν =0
∏ Γ a j + A j Ah
j = n +1
m 
1 − ah + ν

∏ Γ b j + Bj Ah
j =1 (−1)ν (1/z)(1−ah +ν)/Ah
× q (5)
1−a +ν
 
ν!Ah
∏ Γ 1 − bj − Bj Ahh
j = m +1

for z 6= 0 if δ < 0 and for |z| > D −1 if δ = 0.

Both representations above apply when the poles of the gamma function in the numer-
ator of the quotients are simple. When this simplification does not hold, residue theorem
has to be applied. For details about this theorem, one may refer to [7].
Another well-known example of generalized hypergeometric function, which is a
special case of the H-function, is the Meijer G-function [2]. More recently, the R-function
has also shown its applicability to a number of statistical problems [8].
The R-function (see [8]) is defined as a contour complex integral which contains
powers of Gamma functions in their integrands by:
  
α,β m,n  ( a 1 , A ), . . . , ( a n , A n ), ( a n +1 , A n +1 ), . . . , ( a p , A p )
R
γ,δ p,q z 
 (b1 , B1 ), . . . , (bm , Bm ), (bm+1 , Bm+1 ), . . . , (bq , Bq )
1 h(s)
Z
= z−s ds, (6)
2πi L (α + βs)γ+δs

in which h(s) is defined in Equation (2) and α, β, γ, and δ are chosen such that the integral
on the right-hand side of Equation (6) exists.
Several functions are special cases of both the H-function and the R-function, for
example, generalized hypergeometric series p Fq , MacRobert’s E-function, generalized
Bessel–Maitland function, Kratzel function, Wright generalized hypergeometric function,
Bessel functions, Whittaker function, Mittag–Leffler function, trigonometric functions,
exponentials, and so on [2]. In actuality, several authors have studied the properties and
generalizations of such hypergeometric functions, such as [9–14].
Stats 2022, 5 1215

3. Lambert W-Function and Lambert’s Trinomial Equation

Formally, Lambert W-function of a real variable x can be defined as [1]:

W ( x )eW ( x) = x, x ∈ [−1/e, ∞). (7)

It is worth noticing that for x ∈ [−1/e, 0], there are two possible branches of W ( x ),
namely, W0 ( x ) stands for the branch in which W ( x ) ≥ −1, while W−1 ( x ) stands for the
branch in which W ( x ) ≤ −1. Figure 1 shows the behavior of the function as well as its real
branches. The latter are represented by a full line (W0 ( x )) and a dashed line (W−1 ( x )).

WHxL

x
1 2 3 4

W0 HxL
-1
W-1 HxL

-2

-3

Figure 1. Lambert W-function.

When the Lambert W-function of a complex variable is considered, there are infinitely
many branches. This comes from multi-branches of the logarithm function involved in the
inversion of Equation (7).
While studying previous works by Lambert, Euler [6] considered the solution of the
following trinomial equation, where α, β and ν ∈ R:

x α − x β = (α − β)νx α+ β (8)

By means of taking the limit β → α in Equation (8), the following is obtained [1]:

log x = νx α (9)

The solution to Equation (9) is easily obtained in terms of Lambert W-function as

x α = −W (−αν)/αν [1].
In the present paper, in order to show the relation between Lambert W-function and
the H-function, Equation (8) will be solved and the limit will be applied later on.

4. Solution to Lambert’s Trinomial Equation

As pointed out in [1], Equation (8) can be brought back to Lambert’s original trinomial
equation by setting y = x − β , q = (α − β)ν and m = α/β. This way, as in the case of
Equation (8), α, β and ν ∈ R and we can rearrange (8) using these variable changes to obtain

x − β = (α − β )ν + x −α , (10)

In order to solve equations similar to (8) and (10), Rathie and Ozelim [15] applied
Lagrange’s inversion theorem [16]. After a series solution was obtained, those authors
compared such series to computable representations of the H-function. This way, following
the same rationale, we will use the results from [15] to find the solution of Lambert’s
transcendental equation in terms of the H-function.
Stats 2022, 5 1216

Theorem 1. The transformed version of Lambert’s trinomial equation, as presented in (10), can be
solved analytically in terms of the H-function as:
(a) If α/β > 1:
   
β+α α
−α ,β
  
α −1 
α (( α − β ) ν ) β
1,1 β
xα = − H1,2 −((α − β)ν) β    ,
(0, 1), αβ , α−β β
β 
(11)
  α
β α− β 1
i f |ν| < α | β|
   
β−α α− β
−α  (1, 1), β , β
  
1− αβ
− α((α− ββ)ν)
β
1,1 
xα = H2,1 −((α − β)ν)   ,

− αβ , αβ (12)
otherwise
(b) If 0 < α/β < 1 :

−α
"     β+α α   β−α β−α
 #
β , β , β ,
α −1
− α((α− ββ)ν)
β
1,1

xα = H2,1 −((α − β)ν) β  β ,
(0, 1)

(13)
  α
β α− β 1
i f |ν| < α | β|
" #
−α (1, 1)
  
1− αβ
− α((α− ββ)ν)
β
1,1
 
xα = H1,2 −((α − β)ν) 
 −α, α , α, β−α
 ,
β β β β (14)
otherwise
m,n
In the expressions above, H p,q stands for the H-function (see [2]).

Proof. Rathie and Ozelim [15] obtained general solutions, in terms of the H-function, to
trinomial equations. In their work, they considered that α∗ , β∗ , γ∗ , andδ∗ ∈ R and solved
the following equation:
∗ ∗
x α = β∗ + γ∗ x δ , (15)
Thus, by comparing (10) and (15), it is possible to see that when α∗ = − β,
β∗ = (α − β)ν, γ∗ = 1 and δ∗ = −α, the solutions to the trinomial equation can be
given as pointed out in [15].

4.1. Particular Case When β → α

We are interested in the particular case where β → α, in order to match our H-function
solutions to the Lambert W-function solution.

Corollary 1. The W−1 branch of the W-function can be related to the H-function solutions in
equations from (11) to (14) as:
   
−α β+α α
α2 ν((α − β)ν) β 1,1  ,β
  
α −1 
β
W−1 (−αν) = lim H1,2 −((α − β)ν) β     (16)
β→α− β (0, 1), α , α− β

β β

if |ν| < 1/(e|α|) or

− αβ
"   #
β−α α− β
α2 ν((α − β)ν)
  
1,1 1− αβ  (1, 1), β , β
W−1 (−αν) = lim H2,1 −((α − β)ν)    (17)
β→α− β  −α, α
β β

otherwise. Here, W−1 (z) denotes the complex conjugate of W−1 (z). Additionally,
Stats 2022, 5 1217

− αβ
α2 ν((α − β)ν)
     β+α α   β−α  
α −1 β−α
W−1 (−αν) = lim 1,1
H2,1 −((α − β)ν) β β , β , β , (18)

 β
β→α + β  (0, 1)

if |ν| < 1/(e|α|) or

− αβ
α2 ν((α − β)ν)
    
1,1 1− αβ (1, 1)
W−1 (−αν) = lim H1,2 −((α − β)ν) (19)
  

 −α, α , α, β−α
β→α+ β β β β β

otherwise.

Proof. When β → α, as pointed out by [1], the trinomial Equation (8) reduces to Equation (9),
whose solution can be given in terms of Lambert W-function as x α = −W (−αν)/αν. We can
see that the branches satisfying the solutions are x α = −W−1 (−αν)/αν when 0 < αν < 1/e
and x α = −W0 (−αν)/αν when αν < 0. Finally, numerical checks reveal that the correct
branch which matches the equations is W−1 .

4.2. New Series Expansion Formulas for W−1

So far, the relation between Lambert W-function and the H-function has been presented
in terms of closed-form representations. On the other hand, it is possible to explore
numerical alternatives to such representation, as subsequently shown.

Corollary 2. By considering the alternative representation of W−1 in terms of the H-function, the
following new series representations for W−1 arise:
 
α −1 l
 
α2 ν((α − β)ν)
− αβ ∞ ((α − β)ν) β Γ α
βl − α

∑
β
W−1 (−αν) = lim 
β−α α− β
 , if |αν| < 1/e (20)
β→α β l =0 Γ β + β l l!

    βl −1
α
1− αβ
 
Γ
β
∞ −((α − β)ν) αl −1
− αβ
W−1 (−αν) = lim αν((α − β)ν) ∑ (−1)l   , otherwise. (21)
β→α Γ α l − l l!
β
l =0

Proof. Both the H-functions in Equations (16) and (18) can be converted to a series by
directly using the results in (4) and (5). The same is also possible for the H-functions in
Equations (17) and (19), which ends the proof.

The Gamma functions in both Equations (20) and (21) show up after applying La-
grange’s inversion theorem, as indicated in [15]. Such ratios may present numerical
challenges to some software, especially when the values of the Gamma functions increase.

Corollary 3. In order to avoid calculating the ratios of Gamma functions in (20) and (21), such
equations can be rewritten as:
h
−α
W−1 (−αν) = lim −αν((α − β)ν) β +
β→α
  
α −1
− αβ ∞ l −2  l
α2 ν((α − β)ν) ((α − β)ν)
 β
α α
β ∑∏ β
l− −1−k
β l!
, |αν| < 1/e (22)
l =1 k =0

and
Stats 2022, 5 1218

h
W−1 (−αν) = lim α(α − β)−1 +
β→α

∞ l −2    βl −1
(−1)l
  
α
− αβ β 1− αβ
αν((α − β)ν) ∑∏ α
l−2−k
l!
−((α − β)ν) , otherwise (23)
l =1 k =0

Proof. Each ratio of Gamma functions can be further simplified by noticing that [15]:
a −1
Γ ( b + 1)
Γ ( b − a + 1)
= ∏ (b − k) (24)
k =0

By combining (24) with (20) and (21), (22) and (23) follow.

It can be seen that, not only it was possible to show that Lambert W-function is a
limiting case of the H-function, but also that a new series could be obtained for W−1 (z). In
addition, when z > −1/e, W−1 (z) = W−1 (z), since we are dealing with the real values of
the branch.

5. Relation to Generalized Hypergeometric Functions of Higher Rank

We can establish another relation between the Lambert W-function and generalized
hypergeometric functions by looking at its Mellin transform. As pointed out in [17], the
Mellin transform of the W-function can be given as:
Z ∞
(−s)−s Γ(s)
M (W (z)) = x s−1 W ( x )dx = (25)
0 s

for −1 < R(s) < 0.

It is possible to invert the Mellin transform of the W-function to obtain its contour
integral representation. This leads to:

(−s)−s Γ(s) −s Γ(s)

  
1 1
Z Z
−s 0,1 1,0

W (z) = z ds = (− z ) ds = R
1,1 0,1 − z  (26)
2πi L s 2πi L ss+1  (0, 1)

By applying the residue Theorem [2] to calculate the contour integral in Equation (26),
the series for W0 is obtained.
Linking the Lambert-W function to such high class hypergeometric function as the
R-function can shed light onto problems which could not be addressed in terms of the
H-function. These problems can be found in many areas, such as complex multivariate dis-
tributions in the area of the inference on multiple time series [18–20], multivariate statistical
analysis [21], and distribution of the likelihood ratio criterion for testing specific conditions
for the covariance matrix and mean vector of multivariate normal distributions [22–25].
In addition, in physics, this type of special function can be applied to problems
involving the energy loss of fast particles by ionization [26,27]. Therefore, there are plenty
of future works which may benefit from this relation, both in terms of new analytical
results and representations, as well as in new computational calculations made easier by
the Lambert-W function.

6. Conclusions
In the present paper we could establish relations between H-functions and R-function
and the branches W−1 and W0 of the W-function, respectively. These results were obtained
for the first time, establishing a connection between the two classes of functions. Further
studies may lead to other relations, allowing interesting new analytic continuation strategies
for both classes of functions. The new series are also able to compute W−1 (z) for arbitrary
precision, even when z < −1/e.
Stats 2022, 5 1219

While the new relations directly follow from previous results presented in the literature,
they have not been made available to date. The limiting procedure carried out is novel
for these classes of functions, which may shed light on the relation between apparently
unrelated functions.

Author Contributions: Conceptualization, L.C.d.S.M.O.; methodology, P.N.R. and L.C.d.S.M.O.;

software, L.C.d.S.M.O.; validation, P.N.R. and L.C.d.S.M.O.; formal analysis, P.N.R. and L.C.d.S.M.O.;
investigation, L.C.d.S.M.O.; writing—original draft preparation, L.C.d.S.M.O.; writing—review and
editing, P.N.R.; supervision, P.N.R. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding and the APC was gently waived by the Editorial
Office of Stats.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data used are available upon request to the corresponding author.
Acknowledgments: The authors acknowledge the support provided by the University of Brasilia (UnB).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Corless, R.M.; Gonnet, G.H.; Hare, D.E.G.; Jeffrey, D.J.; Knuth, D.E. On the LambertWfunction. Adv. Comput. Math. 1996,
5, 329–359. [CrossRef]
2. Mathai, A.M.; Saxena, R.K.; Haubold, H.J. The H-Function: Theory and Applications; Springer: New York, NY, USA, 2010.
3. Pathan, M.A.; Shahwan, M.J.S. New representations of the voigt functions. Demonstr. Math. 2006, 39, 75–80. [CrossRef]
4. Alzer, H.; Richards, K.C. Series representations for special functions and mathematical constants. Ramanujan J. 2016, 40, 291–310.
[CrossRef]
5. Rathie, A.K.; Rathie, P.N.; Ozelim, L.C.S.M. On a new identity for the H-function with applications to the summation of
hypergeometric series. Turk. J. Math. 2018, 42, 924–935. [CrossRef]
6. Euler, L. De serie Lambertina plurimisque eius insignibus proprietatibus. Acta Acad. Sci. Imp. Petropolitanae 1783, 1779, 29–51.
7. Springer, M.D. The Algebra of Random Variables; John Wiley: New York NY, USA, 1979.
8. Rathie, P.N. Generalized Hypergeometric Functions and Exact Distributions of Test Statistics. Am. J. Math. Manage. Sci. 1988,
9, 155–172. [CrossRef]
9. Dziok, J.; Srivastava, H. Classes of analytic functions associated with the generalized hypergeometric function. Appl. Math.
Comput. 1999, 103, 1–13. [CrossRef]
10. Srivastava, H.M.; Agarwal, P.; Jain, S. Generating functions for the generalized Gauss hypergeometric functions. Appl. Math.
Comput. 2014, 247, 348–352. [CrossRef]
11. Srivastava, H. Generalized hypergeometric functions with integral parameter differences. Indag. Math. (Proc.) 1973, 76, 38–40.
[CrossRef]
12. Saxena, R.K.; Kumar, R. A basic analogue of the generalized H-function. Le Mat. 1995, 50, 263–271.
13. Nisar, K.S.; Rahman, G.; Choi, J.; Mubeen, S.; Muhammad, A. Generalized hypergeometric k-functions via (k, s)-fractional
calculus. J. Nonlinear Sci. Appl. 2017, 10, 1791–1800. [CrossRef]
14. Nisar, K.S. On the generalized modified k-bessel functions of the first kind. Commun. Korean Math. Soc. 2017, 32, 909–914.
[CrossRef]
15. Rathie, P.N.; Ozelim, L.C.S.M. General Solutions to Certain Real Degree Equations and their Applications. Adv. Stud. Contemp.
Math. 2012, 22, 325–341.
16. Whittaker, E.T.; Watson, G.N. A Course of Modern Analysis, 4th ed.; Cambridge Mathematical Library, Cambridge University Press:
Cambridge, UK, 1927.
17. Corless, R.M.; Jeffrey, D.J.; Knuth, D.E. A sequence of series for the Lambert W function. In Proceedings of the 1997 International
Symposium on Symbolic and Algebraic Computation-ISSAC ’97, Maui, HI, USA, 21–23 July 1997; ACM Press: New York, NY,
USA, 1997; pp. 197–204. [CrossRef]
18. Krishnaiah, P. Some recent developments on complex multivariate distributions. J. Multivar. Anal. 1976, 6, 1–30. [CrossRef]
19. Arora, K. A Study of G, H and R-Functions and Their Applications to Statistics. Ph.D. Thesis, Maharshi Dayanand Saraswathi
University, Ajmer, Rajasthan, India, 1993.
20. Rathie, P.N. On the R-function and the distribution of a LRC in multinormal complex case. Int. J. Math. Stat. Sci. 1994, 3, 133–146.
21. Anderson, T.W. An Introduction to Multivariate Statistical Analysis; John Wiley: New York, NY, USA, 1984.
Stats 2022, 5 1220

22. Nagarsenker, B.N.; Pillai, K.C.S. Distribution of the likelihood ratio criterion for testing a hypothesis specifying a covariance matrix.
Biometrika 1973, 60, 359–364. Available online: https://academic.oup.com/biomet/article-pdf/60/2/359/619118/60-2-359.pdf
(accessed on 1 November 2022). [CrossRef]
23. Nagarsenker, B.N.; Pillai, K.C.S. Distribution of the likelihood ratio criterion for testing Σ = Σ0 , µ = µ0 . J. Multivar. Anal. 1974,
4, 114–122. [CrossRef]
24. Korin, B.P. On the distribution of a statistic used for testing a covariance matrix. Biometrika 1968, 55, 171–178. [CrossRef]
25. Rathie, A.k. A generalization of generalized hypergeometric functions. Int. J. Math. Stat. Sci. 1996, 5, 119–133.
26. Landau, L.D. On the energy loss of fast particles by ionization. J . Phys. (USSR) 1944, 8, 201–205.
R σ−i∞
27. Borsch-Supan, W. On the evaluation of the function φ(λ) = (2πi )−1 σ+i∞ eu log u+λu du for real values of λ. J. Res. Natl. Bur.
Stand. Sect. B Math. Math. Phys. 1961, 65, 245–250. [CrossRef]