ISSN: 0025-5742

THE
MATHEMATICS
STUDENT
Volume 93, Nos. 1-2, January - June (2024)
(Issued: May, 2024)

Editor-in-Chief
G. P. Youvaraj

EDITORS

Bruce C. Berndt George E. Andrews M. Ram Murty

B. Sury Siddhartha Gadgil Gadadhar Misra
Sukanta Pati Kaushal Verma Subhojoy Gupta
0
S. K. Tomar Clare D Cruz L. Sunil Chandran
Aparna Dar C. S. Aravinda Atul Dixit
Indranil Biswas Timothy Huber T. S. S. R. K. Rao
Debjani Chakraborty Safique Ahmad T. Raja Sekhar

PUBLISHED BY
THE INDIAN MATHEMATICAL SOCIETY
Website: https://indianmathsoc.org
 THE MATHEMATICS STUDENT
Edited by G. P. Youvaraj

In keeping with the current periodical policy, THE MATHEMATICS STUDENT seeks
to publish material of interest not just to mathematicians with specialized interest but
to the postgraduate students and teachers of mathematics in India and abroad. With
this in view, it will ordinarily publish material of the following type:
1. research papers,
2. the texts (written in a way accessible to students) of the Presidential Addresses, the
Plenary talks and the Award Lectures delivered at the Annual Conferences.
3. general survey articles, popular articles, expository papers and Book-Reviews.
4. problems and solutions of the problems,
5. new, clever proofs of theorems that graduate / undergraduate students might see in
their course work, and
6. articles that arouse curiosity and interest for learning mathematics among readers and
motivate them for doing mathematics.
Articles of the above type are invited for publication in THE MATHEMATICS
STUDENT. Manuscripts intended for publication should be submitted online in the
LATEX and .pdf file including figures and tables to the Editor-in-Chief on the E-mail:
[email protected].
Manuscripts (including bibliographies, tables, etc.) should be typed double spaced on
A4 size paper with 1 inch (2.5 cm.) margins on all sides with font size 11 pt. in LATEX.
Sections should appear in the following order: Title Page, Abstract, Text, Notes and
References. Comments or replies to previously published articles should also follow this
format. In LATEX the following preamble be used as is required by the Press:
\ documentclass[11 pt,a4paper,twoside,reqno]{amsart}
\ usepackage {amsfonts, amssymb, amscd, amsmath, enumerate, verbatim, calc}
\ renewcommand{\ baselinestretch}{1.2}
\ textwidth=12.5 cm
\ textheight=20 cm
\ topmargin=0.5 cm
\ oddsidemargin=1 cm
\ evensidemargin=1 cm
\ pagestyle{plain}
The details are available on Society’s website: https://indianmathsoc.org
Authors of articles / research papers printed in the Mathematics Student as well as in
the Journal shall be entitled to receive a soft copy (PDF file) of the paper published.
There are no page charges for publication of articles in these journals.
All business correspondence should be addressed to Prof. M. M. Pawar, Treasurer,
Indian Mathematical Society, Formerly at K.B.C. North Maharastra University, Jalgon,
Maharashtra - 425001, India, on the E-mail: [email protected].
In case of any query, one may contact the Editor through the e-mail.
Copyright of the published articles lies with the Indian Mathematical Society.
 ISSN: 0025-5742

THE
MATHEMATICS
STUDENT
Volume 93, Nos. 1-2, January - June (2024)
(Issued: May, 2024)

Editor-in-Chief
G. P. Youvaraj

EDITORS

Bruce C. Berndt George E. Andrews M. Ram Murty

B. Sury Siddhartha Gadgil Gadadhar Misra
Sukanta Pati Kaushal Verma Subhojoy Gupta
0
S. K. Tomar Clare D Cruz L. Sunil Chandran
Aparna Dar C. S. Aravinda Atul Dixit
Indranil Biswas Timothy Huber T. S. S. R. K. Rao
Debjani Chakraborty Safique Ahmad T. Raja Sekhar

PUBLISHED BY
THE INDIAN MATHEMATICAL SOCIETY
Website: https://indianmathsoc.org
 ISSN: 0025-5742

ii

c THE INDIAN MATHEMATICAL SOCIETY, 2024.

This volume or any part thereof should not be

reproduced in any form without the written
permission of the publisher.

This volume is not to be sold outside the

Country to which it is consigned by the
Indian Mathematical Society.

Member’s copy is strictly for personal use.

It is not intended for sale or circulation.

Published by Prof. M. M. Shikare for the Indian Mathematical Society,

type set by Prof. G. P. Youvaraj, Former Director and Head, Ramanu-
jan Institute for Advanced Study in Mathematics, Chepauk, Chennai -
600005, India and printed by Dinesh Barve at Parashuram Process, Shed
No. 1246/3, S. No. 129/5/2, Dalviwadi Road, Barangani Mala, Wadgaon
Dhayari, Pune-411041 (India).

Printed in India.
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024)

CONTENTS

1. M. M. Shikare Remembering the Legacy of Professor J. R. 01–02

Patadia: A Pillar of Dedication To The IMS
2. R. Rangarajan, A KDV-Type Equation: 03–13
Bharatha, K. Lax Pair and Traveling Wave Solutions
3. Amrik Singh On Gosper’s Accelerated Series For π 14–23
Nimbran
4. Aritram Dhar Proofs of Two Formulas of Viladet Jovovic 24–29
5. Renukadevi, S.D., Uniqueness of An Entire Function and 30–41
Deepa, N.A. Its Linear Difference Polynomial
Sharing Small Functions
6. Manjunath, B.E., Difference Polynomials of Entire Functions 42–53
Harina, P.W. Sharing a Polynomial Of Certain Degree
7. Gautam Patel, New Method of Lines Solutions For 54–66
Kaushal Patel Nonlinear Sine Gordon Equation With
Reproducing Kernel Hilbert Space Method
8. Padmapriya,V.P., A Characterization of Chaotic 67–70
Ali Akbar, K. Group Actions
9. John M. Cambell A Proof of ζ(2) = π 2 /6 Related To 71–75
Fractional Calculus
10. Teena Thomas On Kakutani’s Characterization of The 76–82
Closed Linear Sublattices of C(X)-Revisted
11. Abhijeet G., A Note On Flexible Immersions 83–100
Suhas Pandit
12. Devendra Tiwari A Shimizu-Type Theorem For 101–115
Subgroups in Sp(N, 1)
13. Asma Ali, Characterization of Rough 3-Prime Ideals 116–131
Arshad Zishan 3-Prime Fuzzy Ideals In Near Rings
14. Nandan Sai A Solution To An Interesting Sum Involving 132–137
Dasireddy Classical Harmonic Number and
Central Binomial Coefficient
 iv

15. Phunphayap,P.N., Niven Numbers and A 138–148

Khemaratchata- Unique Property of 2023
kumthorn, T.,
Pongsrham, P.
16. Rather, Z.A.,, Solving Wiener-Hopf Equation 149–159
Rais Ahmad Involving XOR-Operation
17. Richard D. What Is The Volume of The 160– 172
Carmichael Chambered Nautilus Shell?
18. Niichel, R.J., Uniqueness of Memoryless 173– 180
Ferrell, S., Distributions of Periodic Time Scales
Jokerst, C.
19. Ajit Barman Generalized Ricci Solitons 181– 190
20. Anish Ghosh An Introduction To The Mathematics 191– 199
of Hillel Furstenberg
21. S. G. Dani A Brief Introduction To Certain Dynamical 200– 212
Systems Related To Number Theory
22. Dinesh S. Thakur Elementary Prime Counting 213– 215
23. Problem Section 216– 224

*******
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2023), 01–02

REMEMBERING THE LEGACY OF

PROFESSOR J. R. PATADIA:
A PILLAR OF DEDICATION TO THE IMS

PROF. M. M. SHIKARE

In memoriam of the late Professor J. R.

Patadia, esteemed former faculty member
at Maharaj Sayajirao University, Vado-
dara, Gujarat, and former Editor-in-Chief
of The Mathematics Student, it is with
heavy hearts that we acknowledge his pass-
ing on November 1, 2023, after battling ill-
ness. I had the privilege of hosting Prof.
Patadia and his daughter at my home in
Pune back in June 2023, where we shared
Prof. J. R. Patadia
pleasant conversations on various familial
and IMS-related matters. His vitality during our meeting gave no indica-
tion of the forthcoming loss we would all mourn. As a tribute to his memory,
I reflect on his profound contributions to the Indian Mathematical Society
(IMS).
Among his myriad achievements, Prof. Patadia’s stewardship as Editor-
in-Chief of The Mathematics Student and his instrumental role in develop-
ing and maintaining the IMS website stand out. Assuming the editorship
in 2004, he dedicated 15 years to this role, despite initial reluctance. Prof.
Patadia inherited a publication in disarray, epitomized by the disconcerting
delay in releasing volumes; however, through his relentless efforts, including
addressing copyright hurdles and transitioning to online publishing, he not
only cleared the backlog but also ensured timely releases. The arduous task
of typesetting, once outsourced, became his personal endeavor, showcasing
his dedication to the cause. His selfless act of reimbursing the IMS for his
typesetting services underscores his commitment.
1
2 PROF. M. M. SHIKARE

Prof. Patadia’s editorial acumen was not confined to logistics but ex-
tended to enhancing the scholarly quality of The Mathematics Student.
His discerning decision to discontinue abstracts of conference papers ele-
vated the journal’s stature as a research publication, a sentiment echoed
by esteemed mathematicians like Professor Bruce Berndt. Under his guid-
ance, the journal’s format evolved into a model of international standards,
featuring distinct sections catering to various academic interests.
His tireless advocacy for IMS extended beyond print to digital realms,
exemplified by his pivotal role in securing recognition for The Mathematics
Student in esteemed databases like SCOPUS. Moreover, his meticulous cu-
ration of the IMS website, featuring historical archives and vital updates,
ensures his legacy endures.
Outside his professional endeavors, Prof. Patadia’s unwavering pursuit
of justice, as exemplified by his protracted legal battle for teachers’ rights,
speaks volumes of his character. Even in the face of personal adversity, his
altruism shone through, epitomized by his generous donation to the IMS
amid unresolved pension disputes.
In honoring Prof. Patadia’s memory, we acknowledge the profound loss
to the IMS community. His refusal of a dedicated volume in his honor
epitomizes his humility and underscores his legacy as a cherished friend,
esteemed office-bearer, and unwavering supporter of the IMS.
May his soul find eternal peace.

Prof. M. M. Shikare, General Secretary, IMS,

Emeritus Professor, JSPM University Pune.
[email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 03–13

A KDV-TYPE EQUATION : LAX PAIR AND

TRAVELLING WAVE SOLUTION

R. RANGARAJAN AND BHARATHA K.

(Received : 13 - 08 - 2021 ; Revised : 02 - 02 - 2023)

Abstract. In the present paper, a KdV-type equation is derived with

the help of scaling technique applied to the dependent variable as well
as the Lax pair of the KdV equation. The equation is shown to ad-
mit a Miura-type transformation which yields an m-KdV-type equa-
tion. Standard travelling wave solution is obtained by direct integration
method as well as Adomian decomposition series method. Also lines
on which nonlinear and dispersive terms get equated are computed.

1. Introduction

KdV equation is one of the model equations well studied in the

field of Partial Differential Equations, Nonlinear Dynamics and their appli-
cations. For more details please refer [1, 2, 5, 7, 8, 12]. The KdV equation
∂u ∂u ∂ 3 u
− 6u + =0 (1.1)
∂t ∂x ∂x3
is an integrable equation with the standard Lax pair:
∂2 ∂3 ∂ ∂u
L= − uI and M = −4 + 6u + 3 I, (1.2)
∂x 2 ∂x3 ∂x ∂x
where I is the identity operator.
The equation can be derived in the form
∂L
+ (LM − M L) = 0,
∂t
where one has to make use of the expansion
∂ ∂ ∂f
× (f (x, t)I) = f (x, t) + I.
∂x ∂x ∂x

2010 Mathematics Subject Classification: 35C07, 35C10, 35Q53

Key words and phrases: KdV-type equation, travelling wave solution, the Adomian
decomposition series

© Indian Mathematical Society, 2024 .

3
4 R. RANGARAJAN AND BHARATHA K.

For more details on Lax pair please refer [6, 9].

∂v
It is well known result that the Miura transformation [10] u = v 2 +
∂x
when applied to KdV equation
∂u ∂u ∂ 3 u
− 6u + = 0, yields
∂t ∂x ∂x3
∂3v ∂3v
   
∂v 2 ∂v ∂ ∂v 2 ∂v
2v − 6v + + − 6v + = 0.
∂t ∂x ∂x3 ∂x ∂t ∂x ∂x3
The modified equation is called mKdV equation.
Given an evolution equation, which is a typical nonlinear PDE of the
form
∂u ∂ 2 u ∂ 3 u
 
∂u
+ F u, , , , · · · = 0,
∂t ∂x ∂x2 ∂x3
the travelling wave solution is given by

u(x, t) = U (z), z = x − ct, c > 0,

where c is called velocity of the wave. The term U (z) can be computed as
exact or approximate solution of the corresponding nonlinear ODE. There
are many travelling wave solutions such as solitary waves, kinks, periodic
waves, compactons and so on. For more details please refer [12].
Adomian decomposition series guided by linear and nonlinear terms of a
differential or integral equation provides a lot of flexibility to use series with
powers of well known elementary functions such as identity function, expo-
nential function and so on. Many times it is more simpler than Maclaurin
series. For more details please refer [3, 4, 11, 12].
In the present paper, section 2 mainly has the following results:

Theorem 1.1. The KdV-type equation

∂u 3 ∂u β ∂ 3 u
− αu + =0 (1.3)
∂t 2 ∂x 4 ∂x3
has Lax pair
∂2 α ∂3 3 ∂ 3 ∂u
L= 2
− uI and M = −β 3
+ αu + α I, (1.4)
∂x β ∂x 2 ∂x 4 ∂x
where α, β are non-zero real parameters.

Corollary 1.2. For α = β = 4 in the Theorem 1.1, one gets back the
original KdV equation (1.1) and its Lax pair (1.2).
 A KDV-TYPE EQUATION : LAX PAIR AND TRAVELLING WAVE SOLUTION 5
 
β ∂v
Theorem 1.3. The Miura-type transformation u = v2 + trans-
α ∂x
forms KdV-type equation (1.3) into mKdV-type equation
∂v 3 2 ∂v β ∂3v
− βv + = 0, (1.5)
∂t 2 ∂x 4 ∂x3
where α, β are non-zero real parameters.

By employing direct integration in section 3, the exact solitary wave

solutions are derived for KdV-type equation (1.3) as well as mKdV-type
equation (1.5). In section 4, hypergeometric expansions of solutions of (1.3)
and (1.5) equations are obtained. In the ensuing section, Adomian decom-
position technique is applied to derive solutions in the series form for both
(1.3) and (1.5). In the concluding section, balancing of nonlinear terms with
dispersive terms are discussed.

2. Proofs of the Main Theorems

In order to do scaling manipulations, it is convenient to take the depen-

dent variable ũ and the Lax pair (L̃, M̃ ) so that the KdV equation
∂ ũ ∂ ũ ∂ 3 ũ
− 6ũ + =0
∂t ∂x ∂x3
has the Lax-Pair
∂2
L̃ = − ũI and
∂x2
∂3 ∂ ∂ ũ
M̃ = −4 3 + 6ũ + 3 I.
∂x ∂x ∂x
The scalings u = kũ, L = lL̃ and M = mM̃ where k, l, m are nonzero real
parameters help to derive a KdV-type equations as follows:
∂2
L=l − kluI,
∂x2
∂3 ∂ ∂u
M = −4m + 6kmu + 3km I,
∂x3 ∂x ∂x
∂L ∂u
= −kl I,
∂t ∂t
∂5 ∂3 ∂u ∂ 2 ∂2u ∂
 
∂
LM = lm −4 5 + 10ku 3 + 15k 2
+ 12k 2
− 6k 2 u2
∂x ∂x ∂x ∂x ∂x ∂x ∂x
 3 
∂ u ∂u
+ klm 3 3 − 3ku ,
∂x ∂x
6 R. RANGARAJAN AND BHARATHA K.

∂5 ∂3 ∂u ∂ 2 ∂2u ∂
 
∂
M L = ml −4 5 + 10ku 3 + 15k 2
+ 12k 2
− 6k 2 u2
∂x ∂x ∂x ∂x ∂x ∂x ∂x
 3 
∂ u ∂u
+ mlk 4 3 − 9ku ,
∂x ∂x
and
∂L
+ (LM − M L) = 0 yields
∂t
∂u ∂u ∂3u
− 6mku + m 3 = 0. (2.1)
∂t ∂x ∂x
By taking α = 4mk and β = 4m in (2.1), we have proved Theorem 1.1.
As a special case, for α = β = 4 one gets back the original KdV equation,
which proves Corollary 1.2.
The derived KdV-type equation admits a Miura-type transformation
β ∂v
u= v2 + .
α ∂x
∂u 3 ∂u β ∂ 3 u
It transforms the KdV-type equation (1.3) − αu + = 0 into
∂t 2 ∂x 4 ∂x3
β ∂3v β ∂3v
   
β ∂v 3 2 ∂v β ∂ ∂v 3 2 ∂v
(2v) − βv + + − βv + = 0.
α ∂t 2 ∂x 4 ∂x3 α ∂x ∂t 2 ∂x 4 ∂x3
Hence the resulting mKdV-type equation (1.5) is
∂v 3 2 ∂v β ∂3v
− βv + = 0,
∂t 2 ∂x 4 ∂x3
which proves Theorem 1.3.

3. Travelling Wave Solution

There are many methods available in the literature to derive travelling

wave solutions of KdV and mKdV equations [1, 2, 5, 7, 8, 12].

3.1. KdV-type equation. Following direct integration method, the KdV-

type equation (1.3)
∂u 3 ∂u β ∂ 3 u
− αu + =0
∂t 2 ∂x 4 ∂x3
transforms into
β dU 2
   α 
= U2 c + U
4 dz 2
when applied to the transformation u(x, t) = U (z), z = x − ct, c > 0 and
setting the arbitrary integration constants to zero.
 A KDV-TYPE EQUATION : LAX PAIR AND TRAVELLING WAVE SOLUTION 7
r r
dU 2α 2c
Case 1 α > 0, β > 0: Final equation reduces to = U + U.
dz β α
−2c
The change of variable U = sech2 (V ) readily gives the solution
α
r 
−2c 2 c
U= sech z .
α β
r r
dU 2|α| 2c
Case 2 α < 0, β > 0 : The final equation reduces to = − U.
dz β |α|
2c
Again, by the change of variable U = sech2 (V ) readily gives the solution
|α|
r 
2c 2 c
U= sec h z .
|α| β
Case 3 β < 0, α > 0: Writing β = −|β| and changing U to (−V ), one can
recast the final form
s r
dV 2α 2c
= V V − .
dz |β| α
2c
The change of variable V = sec2 (U ) readily gives the solution
r  α
2c c
U = − sec2 z .
α |β|
Case 4 β < 0, α < 0: The final form of the equation is
s s
dU 2|α| 2c
= U U−
dz |β| |α|
and the solution is r 
2c c
U= sec2 z .
|α| |β|
√
By observing sec(x) = sech(ix), i = −1 one may express all the four
cases simply as
r 
−2c 2 c
U= sech z , α 6= 0, β 6= 0.
|α| β
3.2. mKdV-type equation. The standard application of the transforma-
tion
u(x, t) = V (z), z = x − ct, c > 0
8 R. RANGARAJAN AND BHARATHA K.

to the mKdV-type equation (1.5)

∂v 3 2 ∂v β ∂3v
− βv + =0
∂t 2 ∂x 4 ∂x3
yields the corresponding ODE
dV 3 dV β d3 V
−c − βV 2 + = 0.
dz 2 dz 4 dz 3
Following standard integration method and setting arbitrary integration
constants to zero, one can finally obtain
dV 2
   
2 4c 2
=V +V .
dz β
√
Case 1 β > 0: put V = iW, i = −1.
The equation simplifies to
r
dW 4c
=W − W 2.
dz β
r
4c
The change of variable W = sech(s) and it has the solution
β
r
β
z=− s, which gives
4c
r r 
4c 4c
W (z) = sech z .
β β
√
Case 2 β < 0: put V (z) = W (iz), i = −1.
The equation simplifies to
s
dW 4c
=W − W2
dz |β|
and its solution is
s s !
4c 4c
W (iz) = sec h i z
|β| |β|
s r 
4c 4c
= sech z .
|β| β
Hence in both the cases, the final solution is
r r 
4c 4c
V (z) = i sech z .
β β
 A KDV-TYPE EQUATION : LAX PAIR AND TRAVELLING WAVE SOLUTION 9

4. Hypergeometric expansions of solutions of KdV-type and

mKdV-type equations

By looking at same form in all the four cases of KdV-type equation (1.3)
it is enough to work with Case 1, namely α > 0 and β > 0. Similarly, it is
enough to work with β > 0 for mKdV-type equation (1.5).

4.1. KdV-type equation. By writing

2βk 2
 
k
U (z) = − sech2 z ;
α 2
c
r
where k = 2 , one can express
β
−2βk 2 e−kz

; if z ≥ 0


α (1 + e−kz )2

U (z) =
−2βk 2 ekz
; if z ≤ 0


α (1 + ekz )2


−2βk 2 e−k|z|
= .
α (1 + e−k|z| )2
Put ξ = e−k|z| and note that 0 < ξ < 1.
2βk 2
U (ξ) = − [ξ − 2ξ 2 + 3ξ 3 − 4ξ 4 + · · · ]
α
2βk 2
=− ξ 2 F1 (2, 1; 1; −ξ)
α
∞ (α) (β) z n
P n n
where 2 F1 (α, β; γ; z) = 1 + with (a)n = a(a + 1) · · · (a +
n=1 (γ)n n!
n − 1), a = α, β, γ, |z| < 1 is the standard Gauss Hypergeometric series.

4.2. mKdV-type Equation. Similarly, for β > 0

W (z) = |V (z)| = k sech(kz)

e−k|z|
= 2k
1 + (e−k|z| )2
= 2k [ξ − ξ 3 + ξ 5 − + · · · ]
= 2kξ 2 F1 (1, 1; 1; −ξ 2 ),
c
r
where ξ = e−k|z| , k = 2 and 0 < ξ < 1.
β
10 R. RANGARAJAN AND BHARATHA K.

5. Adomian Decomposition Series Method

5.1. KdV-type equation. For executing the method for KdV-type equa-
tion (1.3) the suitable nonlinear equation is
d2 U α
2
= k2 U + 3 U 2 ;
dz β
r
c β
k=2 , α > 0, β > 0, o < z < ∞, U (0) = − k 2 , U (∞) = 0.
β 2α
The Adomian decomposition series is

U = U0 + U1 + U2 + · · · + Un + · · ·
d2 U0
where − k 2 U0 = 0,
dz 2
n−1
d2 Un 2 αX
− k Un − 3 Um Un−1−m = 0, n = 1, 2, 3, . . .
dz 2 β
m=0
β β
Choose U0 = −2 k 2 e−kz and U1 = 4 k 2 e−2kz .
α  α
β
Let us assume that Um = (−1)m (m + 1) −2 k 2 e−(m+1)kz and work-
α
out the major step of principle of mathematical induction.
m
d2 Um+1 αX
Consider 2
− k 2 Um+1 − 3 Ul Um−l = 0.
dz β
l=0
By choosing Um+1 = am+1 e−(m+2)kz , we have

 "X m
 #
β
am+1 [(m + 2)2 − 1] =(−1)m+1 6 −2 k2 (l + 1)(m + 1 − l)
α
i=0
   2
m+1 β 2 m (m + 1)
=(−1) −2 k 6 + (m + 1)2
α 2

m(m + 1(2m + 1))
−
6
 
m+1 β 2
=(−1) −2 k (m + 1)(m + 2)(m + 3)
α
Hence
 
m+1 β 2
am+1 =(−1) −2 k (m + 2)
α
 A KDV-TYPE EQUATION : LAX PAIR AND TRAVELLING WAVE SOLUTION 11

and
  ∞
β X
U (z) = −2 k 2 e−kz (−1)n (n + 1)e−(n+1)kz
α
n=0
 
β 2 k
=− k sech2 z , which is the desired solution.
2α 2
5.2. mKdV-type Equation. For solving mKdV-type equation (1.5) by
Adomian decomposition series method, the convenient ODE is
d2 W
r
2 3 c
2
−k W +2W = 0, k = 2 , β > 0, 0 < z < ∞, W (0) = k, W (∞) = 0.
dz β
The Adomian series W = W0 + W1 + · · · + Wn + · · · decomposes into the
following equations :
d2 W0
2
− k 2 W0 = 0
dz
 
2 n−1 l
d Wn X X
− k 2 Wn + 2 Wl  Wj Wn−1−l−j  = 0, n = 1, 2, 3, . . .
dz 2
j=0 j=0

Choose W0 = 2ke−kz and W1 = −2ke−3kz . Let us assume that Wm =

(−1)m 2k e−(2m+1)kz and workout the major steps of principle of mathe-
matical induction.
 
2 m l
d Wm+1 X X
− k 2 Wm+1 + 2 Wl  Wj Wm−l−j  = 0.
dz 2
i=0 j=0

Choose Wm+1 = am+1 e−(2m+3)kz then

m
X
am+1 [(2m + 3)2 − 1] =(−1)m+1 (2k) 8 (l + 1)
l=0
m+1
=(−1) (2k) 4(m + 1)(m + 2).
m+1
Hence, am+1 =(−1) 2k and
∞
X
W =2ke−kz (e−2kz )n
n=0
=k sech(kz), which is the desired solution.
12 R. RANGARAJAN AND BHARATHA K.

6. Balancing property of travelling wave solutions

6.1. KdV-type equation. Consider the travelling wave solution of the

KdV-type equation (1.3)
r
2c 2 c
U (z) = − sech z, α 6= 0, β 6= 0
|α| β
which satisfies the following two relations:
dU 2
 
β h α i
U = U4 c + U
4 dz 2
and 2
β3 d3 U

3 α
= U 2 [c + αU ]2 [c + U ].
64 dz 3 2 2
As a result,
 3 2
dU 2
 
d U
U = whenever,
dz dz 3
β2 4 α 3 α
U [c + U ] = U 2 [c + αU ]2 [c + U ]
16 2 2 2
That is,
−2c β2 2 3
U = 0, U = or U = [c + αU ]2 .
α 16 2
Hence
2 2
d3 U
 
dU
U = is possible in the following cases :
dz dz 3
i) z = 0
ii) β > 0, z = ±∞ r
−1 2|α|
iii) 0 < |β| < 6c and z = sech
6c ∓ β

6.2. mKdV-type equation. Consider the travelling wave solution of mKdV-

type equation (1.5)
r
c
W (z) = k sech(kz), k = 2 , β>0
β
which satisfies two important relations
 2
2 dW
W = W 6 (k 2 − W 2 )
dz
 3 2
d W 2 2 2
W 2 (k 2 − W 2 ).


and = k − 6W
dz 3
 A KDV-TYPE EQUATION : LAX PAIR AND TRAVELLING WAVE SOLUTION 13

It is straight forward to deduce that

 2  3 2
2 dW d W
W = , which is possible whenever
dz dz 3
1 1 1 1
z = 0, ±∞, sech−1 √ or sech−1 √ .
k 5 k 7
In conclusion, the paper demonstrates that all the standard results of
KdV and mKdV equations are passed on to KdV-type and mKdV-type
equations, respectively in a nontrivial manner.

7. Acknowledgement

The authors are thankful to University Grants Commission (UGC),

Govt. of India for financial support under the grant UGC-SAP-DRS-II,
No.F.510/12/DRS-II/2018(SAP-I) dated 9 April 2018. We thank the anony-
mous referee for the valuable comments, corrections and suggestions for the
improvement of the paper.
References
[1] Ablowitz M.J and Clarkson P.A, Solitons, Nonlinear Evolution Equations and In-
verse Scattering, London Mathematical Society Lecture Notes, 149, Cambridge Uni-
versity Press, 1991.
[2] Ablowitz M.J and Segur H, Solitons and Inverse Scattering Transform, SIAM Stud-
ies in Applied Mathematics, SIAM, 1981.
[3] Adomian G., Solving Frontier Problems of Physics: The Decomposition Method,
Kluwer Academic Publisher, 1994.
[4] Adomian G., The decomposition method and some recent results for nonlinear equa-
tions, Math.Comput.Modelling, 13(7) (1992),17-43.
[5] Drazin P. G. and Johnson R.S, Solitons: an Introduction, Cambridge University
Press,1992.
[6] Griffiths G.W, Lax pairs, Online City University, UK Notes 2012.
[7] Infield E and Rowlands G, Nonlinear Waves, Solitons and Chaos, Cambridge Uni-
versity Press, 2000.
[8] Lakshmanan M. and Rajashekar S., Nonlinear Dynamics, Integrability Chaos and
Patterns, Springer International edition 2003.
[9] Lax P, Integrals of nonlinear equations of evolution and Solitary Waves, Comm.Pure
Applied Math., 21(5), (1968), 467-490.
[10] Miura R.A, Korteweg-deVries Equation and Generalization I: A remarkable explicit
nonlinear transformation, J.Math.Phys., 9 (1968) 1202-1204.
[11] Peter J Collins, Differential and Integral Equations, Oxford University Press, 2010.
[12] Wazwaz A.M., Partial Differential Equations and Solitary Waves Theory, Springer,
2009.
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 14–23

ON GOSPER’S ACCELERATED SERIES FOR π

AMRIK SINGH NIMBRAN

(Received : 24 - 11 - 2021 ; Revised : 22 - 06 - 2022)

Abstract. This paper gives alternative derivation of Gosper’s accel-

erated series for π obtained by him via transformation of slow series.
Binomial expansions and Euler’s beta function are used for this.

1. Introduction

1.1. Background. The circular solar disc, the full moon and the semi-
circular rainbow must have fascinated our ancient ancestors. Pythagoras
considered circle as the most perfect of all plane figures. Circularity also
had practical applications. The invention of the wheel stimulated the ad-
vance of civilization. Circle is defined as a plane figure enclosed by a curved
line every point on which is equidistant from some fixed interior point. The
line forming the edge of a circle is called its circumference (C), the common
distance the radius (r), and the fixed point the centre. Any straight-line
passing through the centre and joining two points on the circumference is
C
called a diameter (D) of the circle. Numerical estimates of the ratio D
found in the records of most ancient civilizations (3 in the Old Testament
of the Bible, 3 81 in Babylon, 3 1381 in Egypt) point to an early realization
C
that D is constant irrespective of the circle’s size. It is proved in Euclid’s
Elements (Bk. XII, Prop. 2): Circles are to one another as the squares on
(their) diameters. The ratio is denoted by the Greek letter π, introduced
(most likely as an abbreviation for ‘perimeter’ of circle with unit diame-
ter) by William Jones in his Synopsis Palmariorum Matheseos: or, a New
Introduction to the Mathematics (1706). The notation became a standard
symbol once adopted by Euler in his paper Variae observationes circa series
infinitas (1744) and his book Introductio in Analysin Infinitorum (1748).

2010 Mathematics Subject Classification: 11Y60, 33B15, 40A25, 65B10

Key words and phrases: Gosper’s Pi series, convergence acceleration, Beta function

© Indian Mathematical Society, 2024 .

14
 ON GOSPER’S ACCELERATED SERIES FOR π 15

It is not known how the ancients estimated the value of π. Archimedes

(287–212 BC) demonstrated rigorously, by means of Eudoxus’ method of
exhaustion using an inscribed regular polygon and a circumscribing polygon,
10
that 3 71 < π < 3 17 . Afterward, mathematicians used polygons with more
sides to refine Archimedes’ calculations. The method remained in vogue
until the discovery (around 1670) of the series for the inverse sine function
by Isaac Newton (1642–1727) and for the arctangent function by James
Gregory (1638–1675) which revolutionized the process of computing π.

1.2. Two examples of transformed series. G. W. Leibniz (1646–1716)

discovered in 1673 an alternating series that provided a practical tool to
approximate π.
π 1 1 1 1 1
=1− + − + − + −··· . (1.1)
4 3 5 7 9 11
(1.1) can be obtained by putting x = 1 in the series for arctan x. The
Leibniz series converges very slowly making it useless for computational
purpose.
L. Euler (1707–1783) introduced in [4, Part II, Ch. 1] a technique to
enhance the rate of convergence of an alternating series. We reproduce here
Euler’s transform of (1.1) in [4, Part II, Ch. 1, p.295].

Euler took a series S = a0 − a1 + a2 − a3 + − . . . (ai > 0) and computed

successive rows of differences using his difference operator.
16 AMRIK SINGH NIMBRAN

Let ∆nk = (∆n−1 n−1

k+1 − ∆k ) with ∆0k = (−1)k ak . Then

∆00 = a0 ; ∆10 = −a0 + a1 ; ∆20 = ∆11 − ∆10 = a0 − 2a1 + a2 ;

and in general,
n  
n−k n
X
∆n0 = ∆n−1
1 − ∆0n−1 = (−1) ak .
k
k=0

He thus showed that

∞
X (−1)n ∆n 0
S= .
2n+1
n=0
1
Euler’s series with rate of convergence 2 can be written as
∞
π X 2k−1
= 2k

. (1.2)
4 (2k + 1) k
k=0

The following series is ascribed to Newton by Euler in [3, §14]:

∞ n(n−1)
1 1 1 1 1 1 1 X (−1) 2 π
1+ − − + + − − + + − −··· = = √ .
3 5 7 9 11 13 15 2n + 1 2 2
n=0

Applying what he calls ‘the Nilakantha transformation’ to this series, Brink

2
[2, eq.(A.5)] obtained a series with rate of convergence 27 :
∞
X (5n + 4) (2n)!2 (3n)! 23n 3π
= √ .
(6n + 1)(6n + 5) n! (6n)! 8 2
n=0

2. Gosper’s accelerated series for π

Let us now take up R. W. Gosper’s series [1] with rate of convergence

2
27 :
∞
π X 25n − 3
=
. (2.1)
2 2n 3n
n
n=0
Not knowing how Gosper obtained (2.1), we will derive it with the help of
Euler’s beta function which is defined by the integral [6, p.263]:
Z 1
B(p, q) = tp−1 (1 − t)q−1 dt (<(p), <(q) > 0). (2.2)
0
The beta function is related to the Gamma function (a generalization of the
Γ(p) Γ(q)
factorial function) by: B(p, q) = . [6, p.264]
Γ(p + q)
 ON GOSPER’S ACCELERATED SERIES FOR π 17
P∞ 1
2.1. Derivation of n=0 2n (3n) . Differentiating the geometric series:
n

∞
1 X
= yn (|y| < 1).
1−y
n=0

results in
∞
1 X
= n y n−1 (|y| < 1). (2.3)
(1 − y)2
n=0
One can easily see using the definition of beta function that
1 n!(2n)! n(n − 1)!(2n)!
3n

= = = n B(n, 2n + 1),
n
(3n)! (3n)!
so that Z 1
1
3n = n

tn−1 (1 − t)2n dt.
n 0
Multiplying both sides by n
x and taking the sum from n = 0 to ∞
yields
∞ ∞ 1
xn
X X Z
n
3n =

nx tn−1 (1 − t)2n dt
n=0 n n=0 0
Z 1 ∞
X
= x(1 − t)2 nxn−1 tn−1 {(1 − t)2 }n−1 dt
0 n=0

interchanging the order of summation and integration.

Now setting y = xt(1 − t)2 in (2.3) we get:
∞
1 X
2 2
= nxn−1 tn−1 (1 − t)2n−2 ,
{1 − xt(1 − t) }
n=0

where |xt(1 − t)2 |

< 1 and 0 < t < 1.
Thus on taking x = 21 , we have:
∞ Z 1
X 1 (1 − t)2
= 2 dt.
2n 3n 2 2
0 {2 − t(1 − t) }


n=1 n

We may write the integrand on RHS as

(1 − t)2 (1 − t)2
=
{2 − t(1 − t)2 }2 (t − 2)2 (1 + t2 )2
whose partial fraction expansion is:
−2t − 9 2(3t − 4) 2 1
− + + .
125(t2 + 1) 25(t2 + 1)2 125(t − 2) 25(t − 2)2
18 AMRIK SINGH NIMBRAN

The resulting integrals are easy to calculate and equal:

   
9π log(2) π 1 2 log(2) 1 11π − 12 log(2) + 20
− + + + − + = .
500 125 25 50 125 50 500
We thus get:
∞
X 1 11π − 12 log(2) + 20

= . (2.4)
n=0
2n 3n
n
250

P∞ n
2.2. Derivation of n=0 2n (3n) . Multiplying both sides of (2.3) by y, and
n
then differentiating gives
∞
1+y X
− 3
= n2 y n−1 (|y| < 1). (2.5)
(1 − y)
n=0

Proceeding as earlier, we have

∞ Z 1X ∞
X n xn
3n

= n2 xn tn−1 (1 − t)2n dt
n=0 n 0 n=0
Z 1 ∞
X
2
= x(1 − t) n2 xn−1 tn−1 {(1 − t)2 }n−1 dt.
0 n=0

Now setting y = xt(1 − t)2 in (2.5) we get:

∞
1 + xt(1 − t)2 X
− 2 3
= n2 xn−1 tn−1 (1 − t)2n−2 ,
(1 − xt(1 − t) )
n=0

where |xt(1 − t)2 |

< 1 and 0 < t < 1.
Thus on taking x = 21 , we have:
∞ Z 1
X n (1 − t)2 (2 + t(1 − t)2 )
= −2 dt.
2n 3n {2 − t(1 − t)2 }3


n=1 n 0

Since {2 − t(1 − t)2 }3 = −(t − 2)3 (t2 + 1)3 , we may write the RHS as
Z 1
(1 − t)2 (t3 − 2t2 + t + 2)
2 .
0 (t − 2)3 (t2 + 1)3
The partial fraction expansion of the integrand is:
8(2t − 11) 2(41t − 148) 6t − 73
2 3
− 2 2
+
125(t + 1) 625(t + 1) 3125(t2 + 1)
4 17 6
+ 3
+ 2
− .
125(t − 2) 625(t − 2) 3125(t − 2)
 ON GOSPER’S ACCELERATED SERIES FOR π 19

The resulting integrals can be calculated with the help of standard integral
tables and hence the sum:
∞
X n 79π − 18 log(2) + 405
3n

= . (2.6)
2n
n
3125
n=0

25 × (2.6) − 3 × (2.4) yields (2.1).

3. An akin series

We now give a series with two linear factors in the denominator.

∞
π X 10n + 3
=1+ 3n

. (3.1)
2 n
2 n (3n + 1) n
n=1

Proof. To establish (3.1), we split the sum on the RHS into two parts:
∞ ∞
X 3 X 1
3n

+ 3n

.
n=1
2n n n n=1
2n (3n + 1) n
∞
X1
Sum 1. It is easy to compute the sum
by proceeding as in the
2 n n 3n
n=1 n
last section. Using the binomial expansion and the Beta function, we have
Z 1
xn
3n =

xn tn−1 (1 − t)2n
n n 0

and so
∞ Z 1 "∞ #
X xn X
3n =

x(1 − t)2 xn−1 tn−1 (1 − t)2n−2 dt
n=0
n n 0 n=0
Z 1 2
x(1 − t)
= 2
dt.
0 1 − xt(1 − t)

Taking x = 12 , we get
∞ 1 1
(1 − t)2 1 − t)2
Z Z
X 1

= dt = − dt
n=1
2n n 3n
n 0 2 − t(1 − t)2 0 (t − 2)(t2 + 1)
and using the partial fraction expansion
∞ Z 1 Z 1
X 1 2(2t − 1) 1
3n = − 2 + 1)
dt + − dt
−


2 n n n 0 5(t 0 5(t 2)
n=1

where the two integrals are easy to calculate. We thus obtain:

∞
X 1 π log(2)

= − . (3.2)
n=1
2n n 3n
n
10 5
20 AMRIK SINGH NIMBRAN

Sum 2. For the second sum, we proceed as follows.

1 Γ(n + 1)Γ(2n + 1)
3n

= (3n + 1)
n
Γ(3n + 2) Z1
= (3n + 1)B(n + 1, 2n + 1) = (3n + 1) tn (1 − t)2n dt.
0

Z1  n
1 t(1 − t)2
Hence, we have:
=
3n dt.
2n (3n + 1) n
2
0

∞ Z1 Z1
X 1 2 2
3n

= dt = − dt
n
2 (3n + 1) n
2 − t(1 − t)2 (t − 2)(t2 + 1)
n=0 0 0
Z1   Z1 Z1 Z1
2 t+2 1 1 2t 4 1 2 1
= 2
− dt = 2
+ 2
− dt,
5 t +1 t−2 5 t +1 5 t +1 5 t−2
0 0 0 0
log(2) π 2 log(2) π 3 log(2)
= + + = + .
5 5 5 5 5
We took Sum 2 from n = 0. So subtract 1 for the sum from n = 1. Multi-
plying Sum 1 by 3 and adding that to Sum 2, we get:

∞
X 10n + 3 3π 3 log(2) 3 log(2) π π

=
3n − −1+ + = −1 + .
2n n (3n + 1) n
10 5 5 5 2
n=1

4. Two more series of Gosper

Gosper illustrates in [5] how the rate of convergence of infinite series can
be accelerated by a suitable splitting of each term into two parts and then
combining the second part of the n-th term with the first part of the n+1-th
term and leaving the first part of the first term. Repeated application of
this process yields a new series which approaches 0 and the series of the left
out first parts (‘orphans’) that converges faster than the original series. We
will now discuss two other series obtained by Gosper via transformation of
slow series.
1 1 3(2n + 1)
Series I. It is easy to see that + = . Fur-
3n + 1 3n + 2 (3n + 1)(3n + 2)
ther,
1 (3n + 1)(3n + 2)Γ(n + 1) Γ(2n + 2)
3n =


n
(2n + 1)Γ(3n + 3)
 ON GOSPER’S ACCELERATED SERIES FOR π 21

or,
2n + 1 Γ(n + 1) Γ(2n + 2)
3n

= = B(n + 1, 2n + 2).
(3n + 1)(3n + 2) n
Γ(3n + 3)
Multiplying both sides by xn and using the integral for the beta function
as earlier, we have
Z 1
(2n + 1)xn

= xn tn (1 − t)2n+1 dt
(3n + 1)(3n + 2) 3n
n 0
1
and taking x = 2 and summing the series from 0 to infinity as earlier, we
get
∞ 1
2(1 − t)
Z
X 2n + 1
3n = 2 − t(1 − t)2
dt


n=0
(3n + 1)(3n + 2)2n n 0
1 Z 1
2(t − 3)
Z
2
= dt − 2
dt
0 5(t − 2) 0 2(1 + t)
3
= (π − 2 log(2))
10
and hence we deduce
∞
X 1 7π 12 log(2)
3n

= − . (4.1)
(3n + 2) 2n n
10 5
n=0

Combining this result with Sum 1, we obtain the series which occurs
in Gosper’s paper [5, p.32]:
∞
X 5n + 3 1 π
3n

= . (4.2)
(3n + 1)(3n + 2) n 2n 2
n=0

Brink calls it the Nilakantha transform of the Leibniz series [2, eq(3)].
We also get:
∞
X 5n + 4
3n

= 3 log(2). (4.3)
n=0
(3n + 1)(3n + 2) 2n n

By a shift of the index n from 0 to 1, the last formula becomes

∞ ∞ ∞
X 5n − 2 X 2 X 1 π
3n n

= 3n n

+ 3n

= .
n(2n − 1) n 2 n n 2 (2n − 1) n 2n 6
n=1 n=1 n=1

Using value of the first sum obtained in the previous section we deduce
∞
X 1 2 log(2) π
3n

= − . (4.4)
(2n − 1) 2n n
5 30
n=1
22 AMRIK SINGH NIMBRAN

Combining this result and the earlier result gives a series for log(2)
∞
X 5n − 1
3n

= log(2). (4.5)
n=1
n(2n − 1) 2n n

Series II and associated series. Lastly, we take up another series of

Gosper [5, p.33]:
∞
X 5

= 4 − π. (4.6)
n=0
(2n + 1) 2n 3n+4
n+2
The LHS can be decomposed and the formula can be written as
∞  
X 100 20 10 1
− −
= 4 − π.
81(3n + 1) 27(3n + 2) 81(3n + 4) 2n 3n n
n=0

Using the known values, we deduce from the last expression:

∞
X 1 59π 102 log(2) 162
3n

= + − . (4.7)
(3n + 4) 2n n
10 5 5
n=0

Combining this with earlier results gives

∞
X 11n + 15 π 18
3n

= + . (4.8)
(3n + 1)(3n + 4) 2n n
10 5
n=0
∞
X 19n + 24 79π 108
3n

= − . (4.9)
(3n + 2)(3n + 4) 2n n
10 5
n=0
And 19 × (4.8) − 11 × (4.9) yields
∞
X 8n + 9
3n

= 18 − 5π. (4.10)
n=0
(3n + 1)(3n + 2)(3n + 4) 2n−1 n

Concluding remarks

We derived here Gosper’s series for π by means of Euler’s beta function.

We first developed the binomial coefficient 3n


n into associated integral, then
evaluated the resulting integral and lastly eliminated certain constants, such
as logarithms that arose during the process, to obtain the desired series
involving only π. We touched upon Euler’s transformation of the Leibniz
series which involved the central binomial coefficient 2n


n . However, other
4n 6n



binomial coefficients, like 2n and 3n , may not be that easy to handle and
may require much more effort. The technique demonstrated here is readily
 ON GOSPER’S ACCELERATED SERIES FOR π 23

applicable to sums with small binomial coefficient in the denominator and

the associated integral easy to manipulate and evaluate.

References
[1] G. Almkvist, C. Krattenthaler and J. Petersson, Some new formulas for π, Experi-
mental Mathematics, 12.4(2003): 441–456.
[2] D. Brink, Nilakantha’s accelerated series for π. Acta Arith. 171.4(2015): 293–308.
[3] L. Euler, De summis serierum reciprocarum. Originally published in Comment. Acad.
Sci. Petropol. 7(1740): 123–134.
[4] L. Euler, Institutiones calculi differentialis cum eius vsu in analysi finitorum ac doc-
trina serierum. Academiae Imperialis Scientiarum Petropolitanae, 1755.
[5] R. W. Gosper, Acceleration of Series, M.I.T., 1974. Available online at
http://dspace.mit.edu/bitstream/handle/1721.1/6088/AIM-304.pdf
[6] E. T. Whittaker and G. N. Watson, A Course of Modern Analysis ed. Victor H. Moll.
Cambridge University Press, 5th ed., 2021.

Amrik Singh Nimbran

B3/304, Palm Grove Heights, Ardee City, Sector 52,
Grurugram, Haryana, INDIA.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 24–29

PROOFS OF TWO FORMULAS OF VLADETA

JOVOVIC

ARITRAM DHAR
(Received : 15 - 12 - 2021 ; Revised : 09 - 09 - 2022)

Abstract. In this paper, we first provide an analytic and a bijective

proof of a formula stated by Vladeta Jovovic in the OEIS sequence
A117989. We also provide a bijective proof of another interesting result
stated by him on the same page concerning integer partitions with fixed
differences between the largest and smallest parts.

1. Introduction

A partition π of a positive integer n is a non-increasing sequence of

Xr
natural numbers λ1 ≥ λ2 ≥ . . . ≥ λr such that λi = n. λ1 , λ2 , . . . , λr
i=1
are called the parts of the partition π. We call λ1 and λr to be the largest
and smallest parts of the partition π respectively and denote p(n) to be the
number of (unrestricted) partitions of n. For example, p(5) = 7 where the
seven partitions of 5 are 5, 4 + 1, 3 + 2, 3 + 1 + 1, 2 + 2 + 1, 2 + 1 + 1 + 1,
and 1 + 1 + 1 + 1 + 1. Their corresponding largest parts are 5, 4, 3, 3, 2, 2,
and 1 respectively and smallest parts are 5, 1, 2, 1, 1, 1, and 1 respectively.
Throughout the paper, we consider |q| < 1 and adopt the usual notation
for the conventional q-Pochammer symbols:
n−1
Y
(a)n = (a; q)n := (1 − aq k ),
k=0

(a)∞ = (a; q)∞ := lim (a; q)n .

n→∞
We define (a)n for all real numbers n by
(a)∞
(a)n := .
(aq n )∞

2010 Mathematics Subject Classification: 05A17, 05A19, 11P81, 11P82

Key words and phrases: partition, part, largest part, smallest part, fixed difference

© Indian Mathematical Society, 2024 .

24
 PROOFS OF TWO FORMULAS OF VLADETA JOVOVIC 25

By convention, we take (a)0 = 1 and (0)∞ = 1.

2. Main Formulas

Our starting point is the sequence A117989 in the On-Line Encyclopedia

of Integer Sequences [3]. The sequence in question, a(n), counts the number
of partitions of n where the smallest part occurs at least twice. For example,
a(6) = 7 where the relevant partitions are 4 + 1 + 1, 3 + 3, 3 + 1 + 1 + 1,
2 + 2 + 2, 2 + 2 + 1 + 1, 2 + 1 + 1 + 1 + 1, and 1 + 1 + 1 + 1 + 1 + 1.
Also, on the page of A117989, we find the following formula of a(n)
given by Vladeta Jovovic:

Formula 1: (Vladeta Jovovic, July 21 2006)

a(n) = 2p(n) − p(n + 1) ∀ n ≥ 1 (2.1)

where p(n) denotes the partition function.

We denote b(n) = 2p(n) − p(n + 1). By (2.1), we have a(6) = 7 =

22 − 15 = 2p(6) − p(7) = b(6).
We also find another interesting result on the same page posted by
Vladeta Jovovic which states:

Formula 2: (Vladeta Jovovic, May 09 2008)

a(n) = p(2n, n) ∀ n ≥ 1 (2.2)

where p(2n, n) denotes the number of partitions of 2n with fixed difference

equal to n between the largest and smallest parts.

In sections 3 and 4, we give a q-theoretic and a bijective proof of (2.1)

respectively and finally we provide a bijective proof of (2.2) in section 5.

3. Analytic Proof of Formula 1

In this section, we prove (2.1) using generating functions and elementary

infinite series-product identities from the theory of q-hypergeometric series.
To begin with, we define the generating functions for a(n) and b(n) to
be
X∞ X∞
n
A(q) := a(n)q and B(q) := b(n)q n (3.1)
n=1 n=1
26 ARITRAM DHAR

respectively. We then have

∞
X
A(q) = q k+k (1 + q k + q 2k + · · · )(1 + q k+1 + q 2(k+1) + · · · ) · · ·
k=1
∞
X q 2k
=
(q k )∞
k=1
∞
1 X
= (q)k−1 q 2k
(q)∞
k=1
∞
q2 X (q)k (q)k q 2k
=
(q)∞ (q)k
k=0
∞
q 2 (q 3 )∞ X (q 2 )k q k
= (3.2)
(q 2 )∞ (q)k (q 3 )k
k=0
∞
X q k+2
=
(q)k (1 − q k+2 )
k=0
∞
X q k+2 (1 − q k+1 )
=
(q)k+2
k=0
∞ ∞
X q k+2 X q 2k+3
= −
(q)k+2 (q)k+2
k=0 k=0
∞ ∞
X q k+2 1 X (q 2 )k+2
= −
(q)k+2 q (q)k+2
k=0 k=0
1 − q + q2
   
1 1 1 1
= − − − (3.3)
(q)∞ 1 − q q (q 2 )∞ 1−q
1 1 1−q 1 − q + q2
= − − +
(q)∞ 1 − q q(q)∞ q(1 − q)
2 1 1
= − + −1
(q)∞ q(q)∞ q
which is equal to B(q), the generating function for b(n) = 2p(n) − p(n + 1)
∀ n ≥ 1.
Note that (3.2) follows by replacing a = q, b = q, c = 0, and t = q 2 in
Heine’s transformation [1], p.19, Corollary 2.3] and (3.3) follows by replacing
a = 0, t = q and a = 0, t = q 2 respectively in Cauchy’s identity [[1], p.17,
Theorem 2.1]. 
 PROOFS OF TWO FORMULAS OF VLADETA JOVOVIC 27

4. Bijective Proof of Formula 1

In this section, we provide a bijective proof of (2.1).

From (2.1), we have

a(n) = 2p(n) − p(n + 1)

= p(n) − (p(n + 1) − p(n))

which implies
p(n) − a(n) = p(n + 1) − p(n). (4.1)
Let us now define c(n) = p(n) − a(n) and d(n) = p(n + 1) − p(n). Thus,
from (4.1), it suffices to prove that c(n) = d(n) ∀ n ≥ 1.
We note that c(n) denotes the number of partitions of n where the
smallest part occurs exactly once. This follows straightforward from the
definition of a(n). We also note that d(n) denotes the number of partitions
of n + 1 which do not contain 1 as a part because every partition of n + 1
which contains 1 as a part can be obtained by adjoining 1 as a part to every
partition of n.
Let Cn be the set of all partitions of n where the smallest part occurs
exactly once and Dn be the set of all partitions of n + 1 not containing 1
as a part. So, #Cn = c(n) and #Dn = d(n). Thus, it is clear that we will
now produce a bijection between the sets Cn and Dn to obtain the desired
result.
Firstly, we consider a partition π ∈ Cn and consider two cases pertaining
to π: If 1 is a part of π, since it is the smallest part, it occurs exactly once.
Now, add 1 to the 1 already in π to get a new partition π 0 ∈ Dn whose
smallest part now is 2. Hence, π 0 does not contain 1 as a part. Now, if 1
is not a part of π, then the smallest part of π is greater than or equal to 2
and hence it does not contain 1 as a part. On adding 1 to the smallest part
of π, we get a new partition π 0 of n + 1 which does not contain 1 as a part.
Hence, π 0 ∈ Dn .
Now, we consider a partition π 0 ∈ Dn . Thus, π 0 does not contain 1 as
a part which implies that the smallest part of π 0 is greater than or equal
to 2. Again, we consider two cases concerning π 0 : If the smallest part of π 0
occurs exactly once, we subtract 1 from it to get a new partition π ∈ Cn and
we are done. On the other hand, if the smallest part of π 0 occurs at least
twice, subtract 1 from any one of the smallest parts to get a new partition
π ∈ Cn .
28 ARITRAM DHAR

Thus, the process is reversible and hence Cn is bijection with Dn ∀ n ≥ 1.

So, we have our desired result. 

5. Bijective Proof of Formula 2

In this section, we provide a bijective proof of (2.2).

Let An be the set of all partitions of n where the smallest part occurs
at least twice and Fn be the set of all partitions of 2n where the difference
between the largest and smallest parts is equal to n. So, #An = a(n) and
#Fn = p(2n, n). Now, we will provide a bijection between the sets An and
Fn to show that a(n) = p(2n, n).
Firstly, we consider a partition π ∈ An . Then, we add n to any one of
the smallest parts of π (since the smallest part of π occurs at least twice)
to get a new partition π 0 . π 0 ∈ Fn because the largest part of π 0 now is
equal to n+ the smallest part of π and the smallest part of π 0 is equal to
the smallest part of π thus making the difference equal to n.
For the other way, we now consider a partition π 0 ∈ Fn . Note that the
largest part of π 0 occurs exactly once. We then subtract n from the largest
part of π 0 to get a new partition π whose smallest part is equal to the
smallest part of π 0 and consider two cases pertaining to π 0 : If the smallest
part of π 0 occurs at least twice, we are done, i.e., π ∈ An since subtracting
n from the largest part of π 0 does not affect the frequency of the smallest
part of π (= the smallest part of π 0 ) which still remains at least 2. Lastly, if
the smallest part of π 0 occurs exactly once, subtracting n from the largest
part of π 0 makes it equal to the the smallest part of π 0 and thus, the new
partition π that we obtain has smallest part occuring at least twice since
smallest parts of π 0 and π are equal. Hence, π ∈ An .
Thus, the process is reversible and hence, we have a bijection between
An and Fn giving our desired result. 

6. Conclusion

Thus, we have p(2n, n) = a(n) = 2p(n) − p(n + 1) ∀ n ≥ 1. We

speculate that there is an interesting proof of (2.2) using the generating
function approach. Although in [2], Andrews, Beck, and Robbins gave
the generating function for p(n, t) (which is the number of partitions of n
with fixed difference equal to t between the largest and smallest parts), the
generating function of p(2n, n) does not follow straightforward.
 PROOFS OF TWO FORMULAS OF VLADETA JOVOVIC 29

∞
X
In the same spirit, we define Gm (q) := am (n)q n where am (n) denotes
n=1
the number of partitions of n where the smallest part occurs at least m
∞
X q mk
times. On the page of A117989, we see that Gm (q) = . It will
(q k )∞
k=1
be very interesting to see if there is a closed formula analogous to (2.1) for
am (n) ∀ m ≥ 3 and if there exists such a formula, then it would be nice to
provide a combinatorial proof of it.
Acknowledgement: The author would like to thank George E. Andrews
for suggesting him to prove the two formulas of Vladeta Jovovic which came
up during an ongoing project with him.

References
[1] G. E. Andrews, The Theory of Partitions, Cambridge Mathematical Library, Cam-
bridge University Press, 1998.
[2] G. E. Andrews, M. Beck, and N. Robbins, Partitions with fixed differences between
largest and smallest parts, Proc. Amer. Math. Soc., 143(10): 4283-4289 (2015).
[3] The On-Line Encyclopedia of Integer Sequences, https://oeis.org/A117989.

Aritram Dhar
Department of Mathematics
University of Florida,
Gainesville, FL 32611.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 30–41

UNIQUENESS OF AN ENTIRE FUNCTION AND ITS

LINEAR DIFFERENCE POLYNOMIAL SHARING
SMALL FUNCTIONS

RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

(Received : 16 - 03 - 2022 ; Revised : 27 - 04 - 2023)

Abstract. This article investigates the uniqueness problem of a finite

order entire function and its linear difference polynomial sharing a set
consisting of two distinct entire functions of smaller orders. The result
obtained in this paper generalises the result by Jianming Qi, Yanfeng
Wang, and Yongyi Gu [14].

1. Introduction

We adopt the following standard notations of the Nevanlinna theory

through out the paper, which can be found in [18].
Let f (z) be a meromorphic function and not a constant in the disc |z| ≤
R(0 < R < ∞). For 0 < r < R, Nevanlinna defined the following functions:

Proximity function of f (z):

Z 2π
1
m(r, f ) = m(r, ∞, f ) = log+ f (reiθ ) dθ,
2π 0

where log += max {logx, 0} for all x ≥ 0. Here m(r, f ) is the average
of the positive logarithm of |f (z)| on the circle |z| = r.

Counting function of poles of f (z):

r
n(t, f ) − n(0, f )
Z
N (r, f ) = dt + n(0, f ) log r,
0 t

Corresponding author: Renukadevi S. Dyavanal

2010 Mathematics Subject Classification: 30D35, 32H30
Key words and phrases: Entire function f (z), order of f (z), Linear difference
polynomial of f (z), Small function of f (z), Meromorphic function, Nevanlinna theory

© Indian Mathematical Society, 2024 .

30
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 31
...

Reduced counting function of poles of f (z):

r
n(t, f ) − n(0, f )
Z
N (r, f ) := dt + n(0, f )logr,
0 t
where n(t, f ) denotes the number of poles of f (z) in the disc | z |≤ t,
multiple poles are counted according to their multiplicities, and n(t, f ) de-
notes the number of distinct poles of f (z) in the disc | z |≤ t.

Characteristic function of f (z):

T (r, f ) = m(r, f ) + N (r, f ).

The characteristic function T (r, f ) is obviously a non-negative function

and plays a cardinal role in the whole theory of meromorphic functions.

1
Let a be a complex number. Obviously, f (z)−a is meromorphic in the
disk |z| ≤ R. Similar to the above definitions, R. Nevanlinna defined the
following functions:

1
Proximity function of f (z)−a :
  Z 2π
1 1 1
m r, = m(r, f = a) = m(r, a) = log+ dθ,
f −a 2π 0 |f (reiθ ) − a|
1
which is the average of the positive logarithm of |f (z)| on the circle |z| = r.

Counting function of f (z) at the value a:

1 1
r n(t, f −a ) − n(0, f −a )
  Z
1
N r, = N (r, f = a) = dt
f −a 0 t
 
1
+ n 0, log r,
f −a
1
where n(t, f −a ) denotes the number of zeros of f (z) − a in the disc |z| ≤ t
1
counting multiplicities and n(0, f −a ) the multiplicity of zeros of f (z) − a at
the origin.
32 RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

1
The characteristic function of f (z)−a :
     
1 1 1
T r, = m r, + N r, .
f −a f −a f −a

Order of a function: Let f (z) be a meromorphic function on the whole

complex plane. The order ρ(f ) and lower order µ(f ) are defined, respec-
tively, by the order and lower order of T (r, f ), that is

log T (r, f )
ρ(f ) = limr→∞ ,
log r
log T (r, f )
µ(f ) = limr→∞ .
log r
The exponent of convergence of zeros λ(f ) of f (z) is defined by

log N (r, f1 )
λ(f ) = lim .
r→∞ log r
Nevanlinna established the following two fundamental theorems.

First fundamental theorem of Nevanlinna: Let f (z) be a meromorphic

function on the whole complex plane, and a is any complex number. Then

 
1
T r, = T (r, f ) + O(1).
f −a
 
1
This means that, for any complex number a, the difference of T r, f −a
and T (r, f ) is a bounded quantity.
The Second fundamental theorem of Nevanlinna for q(≥ 3) values:
Suppose that f (z) is a meromorphic function in the complex plane and
a1 , a2 , · · · , aq are q(≥ 3) distinct values in C. Then
q  
X 1
(q − 2)T (r, f ) < N r, + S(r, f ).
f − aj
j=1

The Second fundamental theorem of Nevanlinna for three small

functions: Suppose that f (z) is a meromorphic function in the complex
plane and a1 (z), a2 (z), a3 (z) are three distinct small functions of f (z).
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 33
...

Then
3  
X 1
T (r, f ) < N r, + S(r, f ).
f − aj (z)
j=1

The error term S(r, f ) in the second fundamental theorem of Nevanlinna is

given by
S(r, f ) = o(T (r, f )), (r → ∞), if the order of f (z) is finite and if the order
of f (z) is infinite, then S(r, f ) = o(T (r, f )), (r → ∞, r 6∈ E) , where E is
a set, with a finite linear measure.
Small function of f (z):
If f (z) and a(z) are meromorphic functions in the complex plane, then a(z)
is called a small function of f (z) if it satisfies T (r, a(z)) = o(T (r, f )) =
S(r, f ).

Sharing small functions: Let f and g be two meromorphic functions,

and a and b are two small functions of f and g. We say that f and g share
a pair of small functions (a, b) CM (resp. IM ) if f − a and g − b have
the same zeros counting multiplicities (resp. ignoring multiplicities). When
a = b, we say that f and g share a CM (resp. IM ) if f − a and g − a have
the same zeros, counting multiplicities (resp. ignoring multiplicities).
Sharing a Set: Let S be a finite set of some functions, and f (z) is a
meromorphic function. Then a set Ef (S) is defined as

Ef (S) = ∪a∈S {z | f (z) − a(z) = 0, counting multiplicities(CM )}.

Assume that g is another function. We say that f and g share the set S
CM, provided that Ef (S) = Eg (S).

For a meromorphic function f (z), we define its shift by fc = f (z + c)

and its difference operators by

4c f = f (z + c) − f (z),

4nc f = 4n−1
c (4c f (z)), n ∈ N, n ≥ 2.

A linear difference polynomial of f (z) is denoted by L(z, f ) and de-

fined as

L(z, f ) = a0 f (z) + a1 f (z + t1 ) + · · · + ak f (z + tk ),
34 RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

where ti ’s are finite complex numbers and the coefficients ai ’s are small
functions of f (z).
In 2009, Liu [16] proved the following result:
Theorem 1.1. Let a be a non-zero complex number and f be a transcen-
dental entire function with finite order. If f and 4c f share {a, −a} CM,
then 4c f (z) = f (z) for all z ∈ C.
In 2012, Li [15] proved the following theorem:
Theorem 1.2. Suppose that a, b are two distinct entire functions, and f (z)
is a non-constant entire function with ρ(f ) 6= 1 and λ(f ) < ρ(f ) < ∞ such
that ρ(a) < ρ(f ) and ρ(b) < ρ(f ). If f (z) and 4c f (z) share {a, b} CM,
then f (z) = 4c f (z) for all z ∈ C.
In 2019, Jianming Qi, Yanfeng Wang, and Yongyi Gu [14] removed the
condition ρ(f ) 6= 1 in the Theorem and proved the following result:
Theorem 1.3. Suppose that a, b are two distinct entire functions, and f (z)
is a non-constant entire function of finite order with λ(f ) < ρ(f ) such that
ρ(a) < ρ(f ) and ρ(b) < ρ(f ). If f (z) and 4c f (z) share {a, b} CM, then
f (z) = Aeµz , where A, µ are two non-zero constants satisfying eµc = 2.
Furthermore, f = 4c f
In this paper, we extend the above Theorem 1.3 for L(z, f ), a linear
difference polynomial of f (z).
The following is the main result of this paper.
Theorem 1.4. If f (z) and L(z, f )(6≡ 0) share {c, d} CM, where c and d
are two distinct entire functions, and f (z) is a non-constant entire function
of finite order with λ(f ) < ρ(f ) such that ρ(c) < ρ(f ) and ρ(d) < ρ(f ),
then f (z) = KeCz , where K is a non-zero constant and a0 + a1 eCt1 + · · · +
ak eCtk = 1. Furthermore, f (z) = L(z, f ).

2. Some preliminary results

We need the following results to prove our main result.

Lemma 2.1. [2]) Let f (z) be a meromorphic function of finite order, and let
ω1 and ω2 are two arbitrary complex numbers, such that ω1 6= ω2 . Assume
that σ is the order of f , then for each  > 0, we have
 
f (z + ω1 )
m r, = O(rσ−1+ ). (2.1)
f (z + ω2 )
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 35
...

Lemma 2.2. ([18] Theorem 1.51) Suppose that f1 (z), f2 (z) · · · fn (z)(n ≥ 2)
are meromorphic functions, and g1 (z) · · · gn (z) are entire functions satisfy-
ing the following conditions:
Xn
1) fj (z)egj (z) = 0
j=1
2)gj (z) − gk (z) are not constants for 1 ≤ j < k ≤ n
3)For 1 ≤ j ≤ n, 1 ≤ h < k ≤ n, T (r, fj ) = o(T (r, egh −gk )(r −→ ∞, r 6∈ E).
Then fj (z) = 0(j = 1, 2, · · · , n).

Lemma 2.3. [1]) Let g(z) be a transcedental and meromorphic function in

the plane of order is less than 1. If h > 0, then there exists an -set E such
that
g(z + ω)
→ 1, when z → ∞ in C \ E,
g(z)
uniformly in ω for | ω |6 h.

Lemma 2.4. ([18] Theorem 1.44) Let h(z) be a non-constant entire func-
tion, and f (z) = eh(z) . Let ρ and µ be the order and the lower order of
f (z), respectively. Then we have
(i) If h(z) is a polynomial of degree p, then ρ = µ = p.
(ii) If h(z) is a transcidental entire function, then ρ = µ = ∞.

Lemma 2.5. ([18] Theorem 1.18) Let f (z) and g(z) be two non-constant
meromorphic functions in the complex plane with, ρ(f ) as the order of f (z)
and µ(g) as the lower order of g(z). If ρ(f ) < µ(g), then
T (r, f ) = o(T (r, g)), (r → ∞).

Lemma 2.6. ([18] Theorem 1.14) Suppose f (z) and g(z) are two non-
constant meromorphic functions in the complex plane with, ρ(f ) and ρ(g)
as their orders, respectively. Then ρ(f ·g) ≤ max {ρ(f ), ρ(g)} and ρ(f +g) ≤
max {ρ(f ), ρ(g)}.

3. Proof of the Theorem 1.4

Proof. Since f and L(z, f ) share {c, d} CM, we can write

(L(z, f ) − c)(L(z, f ) − d)
= eφ , (3.1)
(f − c)(f − d)
where φ is an entire function. Furthermore, it deduces from (3.1) and
max {ρ(c), ρ(d)} < ρ(f ) < ∞ that φ is polynomial.
36 RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

By the Hadmard Factorization Theorem, we have f (z) = h(z)ep(z) , where

h(6≡ 0) is an entire function, and p(z) is a polynomial satisfying

λ(f ) = ρ(h) < ρ(f ) = ρ(ep(z) ) = deg(p(z)). (3.2)

Hence

L(z, f ) = a0 h(z)ep(z) + a1 h(z + t1 )ep(z+t1 ) + · · · + ak h(z + tk )ep(z+tk )

 
= ep(z) a0 h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) + · · · + ak h(z + tk )ep(z+tk )−p(z) .
(3.3)
Substituting this in (3.1), we get
n o
ep(z) [a0 h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) + · · · + ak h(z + tk )ep(z+tk −p(z) ] − c(z)
n o
ep(z) [a0 h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) + · · · + ak h(z + tk )ep(z+tk )−p(z) )] − d(z)

= (h(z)ep(z) − c)(h(z)ep(z) − d)eφ , (3.4)

where a0 , a1 , a2 · · · ak are small functions of f (z) and t1 , t2 , · · · , tk are finite
complex constants.
Set w1 = a0 h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) + · · · + ak h(z + tk )ep(z+tk )−p(z) .

If w1 ≡ 0, then L(z, f ) ≡ 0. It contradicts L(z, f ) 6≡ 0. Thus, w1 6≡ 0.

Using (3.2), Lemma 2.4, and Lemma 2.5, we have

T (r, h) = o(T (r, ep(z) )), (r → ∞).

This implies that h(z) is a small function of ep(z) .

Using (3.2) and repeated application of Lemma 2.6 to w1 , we obtain that

ρ(w1 ) ≤ ρ(ep(z+ti )−p(z) ) = deg(p(z) − 1) < degp(z) = ρ(ep(z) ).

Using this with Lemma 2.4, and Lemma 2.5, we get that

T (r, w1 ) = o(T (r, ep(z) )), (r → ∞).

This implies that w1 is a small function of ep(z) .

From (3.4), following can be written
c p d
w12 [ep − ][e − ]
w1 w1
eφ = . (3.5)
2 p
c p d
h [e − ][e − ]
h h
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 37
...

Note that c 6= d. Without loss of generality, we suppose that c 6≡ 0. Assume

c
that z0 is a zero of ep − but not a zero of w1 . It follows from (3.5) and
h
c d
the assumption about sharing that z0 is a zero of ep − or ep − . We
w1 w1
p
denote by N1 (r, e ) the reduced counting function of those common zeros of
c c
ep − and ep − . Similarly, we denote by N2 (r, ep ) the reduced counting
h w1
c d
function of those common zeros of ep − and ep − . Note that h is a small
h w1
function with respect to ep . Applying the second fundamental theorem to
ep gives  
1
T (r, ep ) ≤ N r, c + S(r, f )

ep −
h
= N1 (r, ep ) + N2 (r, ep ) + S(r, ep ). (3.6)
This implies that either N1 (r, ep ) 6= S(r, ep ) or N2 (r, ep ) 6= S(r, ep ). We
consider the following two cases:

Case 1 : N1 (r, ep ) 6= S(r, ep ).

c c
Let c0 be the common zero of ep − and ep − . Then it is clear that c0
h w1
c c c c
is a zero of − . If − 6≡ 0, then
h w1 h w1
 
 
p p 1 c c
S(r, e ) 6= N1 (r, e ) 6 N r , c c  6 T r, h − w = S(r, ep ),
 
− 1
h w1
a contradiction. Thus h ≡ w1 .
It leads to

a0 h(z)+a1 h(z+t1 )ep(z+t1 )−p(z) +· · ·+ak h(z+tk )ep(z+tk )−p(z) = h(z). (3.7)

That is,

a0 h(z)ep(z) + a1 h(z + t1 )ep(z+t1 ) + · · · + ak h(z + tk )ep(z+tk ) = h(z)ep(z) .

This implies that

a0 f (z) + a1 f (z + t1 ) + a2 f (z + t2 ) + .... + ak f (z + tk ) = f (z).

38 RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

That is, L(z, f ) = f (z) and also equation (3.7) can be written as

(a0 − 1)h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) + · · · + ak h(z + tk )ep(z+tk )−p(z) ) = 0.

(3.8)
If p(z +t1 )−p(z), p(z +t2 )−p(z), · · · , p(z +tk )−p(z) are not constants,
then by Lemma 2.2, we get that h(z) ≡ 0, a contradiction.

So, at least one of p(z + t1 ) − p(z), p(z + t2 ) − p(z), · · · , p(z + tk ) − p(z)

is constant, say p(z + ti ) − p(z) = di f or some (i ∈ {1, 2, · · · , k}), where di
is a constant.
This gives p0 (z + ti ) − p0 (z) = 0 and hence p0 (z) is a periodic function
of period ti . Also, p0 (z) is a polynomial. Since a polynomial can not be
a periodic function unless it is a constant function, we get that p0 (z) is a
constant function. Thus p(z) is a polynomial of degree 1, say p(z) = Cz+D,
where C and D are two constants and C 6= 0. Hence
ep(z) = eCz+D = eD eCz and ep(z+ti ) = eCz eCti +D for i = 1, 2, · · · , k.
Hence, equation (3.8) becomes

(a0 − 1)h(z)eCz eD + a1 h(z + t1 )eCt1 +D eCz + · · · + ak h(z + tk )eCtk +D eCz = 0.

(3.9)
That is,

a1 h(z + t1 )eCt1 +D eCz + · · · + ak h(z + tk )eCtk +D eCz = (1 − a0 )h(z)eCz eD .

This gives
h(z + t1 ) Ct1 h(z + tk ) Ctk
a1 e + · · · + ak e = (1 − a0 ). (3.10)
h(z) h(z)
Using ρ(h) < ρ(f (z)) = ρ(ep(z) ) = 1 and by Lemma 2.3, we obtain
h(z + t1 ) h(z + tk )
→ 1, · · · , → 1 and using this, (3.10) leads to
h(z) h(z)
a0 + a1 eCt1 + · · · + ak eCtk = 1.

Here h(z) is a periodic function of periods t1 , t2 , · · · tk and using h(z)

as an entire function whose growth is small compared to the growth of
ep(z) = eCz+D gives that h(z) is a non-zero constant function, say α. Using
this, finally we get f (z) = α eCz+D = K eCz , where K = α eD is a non-zero
constant and C satisfies a0 + a1 eCt1 + · · · + ak eCtk = 1.
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 39
...

Case 2: N2 (r, ep ) 6= S(r, ep ).

c d
Let d0 be the common zero of ep − and ep − . Then, it is clear that
h w1
c d c d
d0 is a zero of − . If − 6≡ 0, then
h w1 h w1
 
 
p p 1 c d
= S(r, ep ),
 
S(r, e ) 6= N2 (r, e ) 6 N 
r , c
6T r, −
d  h w1
−
h w1
a contradiction. Thus
c d
− ≡ 0. (3.11)
h w1
c
If d ≡ 0, then ≡ 0, which implies c ≡ 0, which is in contradiction to the
h
fact that c and d are distinct functions. So d 6≡ 0.
d
We assume that t0 is a zero of ep − but not a zero of w1 . It follows
h
c d
from (3.5), that t0 is a zero of ep − or ep − . We denote by N3 (r, ep )
w1 w1
d
the reduced counting function of those common zeros of ep − and ep −
h
c
. Similarly, we denote by N4 (r, ep ) the reduced counting function of
w1
d d
those common zeros of ep − and ep − . Again, applying the second
h w1
fundamental theorem to ep gives
 
1
T (r, ep ) ≤ N r,  + S(r, e )
p
 
d
ep −
h
= N3 (r, e ) + N4 (r, ep ) + S(r, ep ),
p
(3.12)
which implies that either N3 (r, ep ) 6= S(r, ep ) or N4 (r, ep ) 6= S(r, ep ).
Subcase 2(i): If N4 (r, ep ) 6= S(r, ep ), then as in case 1, we get the conclu-
sion of the Theorem 1
Subcase 2(ii): If N3 (r, ep ) 6= S(r, ep ), then as in the above Case 2, we can
deduce that
d c
− ≡ 0. (3.13)
h w1
It follows from (3.11), and (3.13)

c2 = d2 . (3.14)
40 RENUKADEVI S. DYAVANAL AND DEEPA N. ANGADI

Note that c 6≡ d. Thus, c = −d. Again, by (3.13), one has w1 = −h, which
is written as
(a0 + 1)h(z) + a1 h(z + t1 )ep(z+t1 )−p(z) , · · · , ak h(z + tk )ep(z+tk )−p(z) ) ≡ 0
By Lemma 2.2, if p(z + t1 ) − p(z), p(z + t2 ) − p(z), · · · , p(z + tk ) − p(z) are
not constants, then h(z) ≡ 0, a contradiction.
So, at least one of p(z + t1 ) − p(z) , p(z + t2 ) − p(z), · · · , p(z + tk ) − p(z)
is constant. Continuing as in Case 1, we get the conclusion of the Theorem
1.4. 

Acknowledgement: The authors are grateful to the referees and editors

for their valuable comments, and suggestions for improving the manuscript.
The University Research Seed Grant Policy at Karnatak University, Dhar-
wad, India, provided financial support for the first author’s research project
with the reference number KU/PMEB/2021-22/76, and the second author
is supported by a URS fellowship at the Department of Mathematics at the
same institution with the reference number KU/Sch/URS/2019-2020/616.

References
[1] Bergweiler W., Langley J. K., Zeros of differences of meromorphic functions, Math.
Proc. Camb. Philos. Soc 142(2007),133–147.
[2] Chiang Y. M., Feng S. J., On the Nevanlinna characteristic of f (z +η) and difference
equations in the complex plane, Ramanujan J 16(2008), 105–129.
[3] Dyavanal R. S., Applications of difference analogue of Cartan’s second main theorem
for holomorphic curves, The J Anal 28(2020), 615-622.
[4] Dyavanal R. S., Desai R. V., Uniqueness of difference polynomials of entire functions,
Appl. J. Math 8(2014), 3419-3424.
[5] Dyavanal R. S., Mathai M. M., Uniqueness of Difference-differential Polynomials
of Meromprphic Functions Ukrainian Mathematical Journal (UKR MATH J+) 71,
December, (2019). DOI 10.1007/s11253-019-01695-8
[6] Dyavanal R. S., Mathai M. M., Value distribution of general diffential-difference
polynomials of meromorphic functions, J Anal 27(2019), 931-942.
[7] Dyavanal R. S., Mathai M. M., Uniqueness of Relaxed Weakly Weighted Sharing of
Differential –Difference Polynomials of Entire Functions The Mathematics Student
92(2023), 205-220.
[8] Dyavanal R. S., Hattikal A. M., Uniqueness of difference-differential polynomials of
entire functions sharing one value, Tamkang J. Math 47(2016), 193-206.
[9] Dyavanal R. S., Mathai M. M., Hattikal A. M., Unicity Theorems of Linear Dif-
ference Polynomial of Entire and Meromorphic Functions, Indian Journal of Math-
ematics 61(2019), 141-152.
UNIQUENESS OF AN ENTIRE FUNCTION AND ITS LINEAR DIFFERENCE POLYNOMIAL 41
...

[10] Dyavanal R. S., Mathai M. M., Hattikal A. M., Unique range set of a meromor-
phic function and it’s linear difference polynomial, Journal of the Indian Math. Soc
89(2022), 44-57.
[11] Hayman W. K., Meromorphic Functions, Clarendon, Oxford, 1964.
[12] Halburd R. G., Korhonen R. J., Nevanlinna theory for the difference operators, Ann.
Acad. Sci. Fenn., Math 31(2)
[13] Halburd R. G., Korhonen R. J., Difference analogue of the lemma on the loga-
rithmic derivative with applications to difference equations, J. Math. Anal. Appl
314(2)(2006), 477–487.
[14] Jianming, Q., Yanfeng, W., Yongyi, G., A note on entire functions sharing a finite
set with their difference operators, Adv. Differ. Equ (2019), 1-7, Article ID 114.
[15] Li, X. M., Entire functions sharing a finite set with their difference operators, Com-
put. Methods Funct. Theory 12(2012),307–328.
[16] Liu, K., Meromorphic functions sharing a set with applications to difference equa-
tions, J. Math. Anal. Appl 359(2009),384–393.
[17] Yi, H. X., A question of Gross and the uniqueness of entire function, Nagoya Math.
J 138 (1995), 169–177.
[18] Yang C. C., Y, H. X, Uniqueness Theory of Meromorphic Functions, Science Press,
Beijing, 2003.

Renukadevi S. Dyavanal
Department of Mathematics
Karnatak University,
Dharwad - 580003, India.
E-mail: [email protected] ; [email protected] ;

Deepa N. Angadi
Department of Mathematics
Karnatak University,
Dharwad - 580003, India.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 42–53

DIFFERENCE POLYNOMIALS OF ENTIRE

FUNCTIONS SHARING A POLYNOMIAL OF
CERTAIN DEGREE

MANJUNATH B. E., HARINA P. WAGHAMORE

(Received : 07 - 05 - 2022 ; Revised : 07 - 08 - 2022)

Abstract. In this paper we use the notion of weighted sharing to

investigate the uniqueness relation of entire functions when difference
products say f n P (f )∆c f and g n P (g)∆c g share a non-zero polynomial
p(z). We also note the other occurrence for P (z), which generalizes the
result of Qi, Yang, and Liu [15].

1. Introduction

f (z) is meromorphic if it is analytic in the complex plane except at

isolated poles; if there are no poles, then f (z) reduces to an entire func-
tion. In what follows, we assume the reader understands Nevanlinna’s basic
results and notation [8, 16]. Let f and g be two non-constant meromor-
phic functions in the complex plane C. We say that f, g share a counted
multiplicities (CM) if f − a, g − a have the same zeros with the same mul-
tiplicities and we say that f, g share a ignoring multiplicities (IM) if we do
not consider the multiplicities, where a is a small function of f and g.
The relaxation has been done based on the notion of weighted sharing ob-
tained by I. Lahiri as follows.

Definition 1.1. [7] Let k ∈ N ∪ {0} ∪ {∞}. For a ∈ C ∪ {∞} we denote

by Ek (a; f ) the set of all a-points of f where an a-point of multiplicity m is
counted m times if m ≤ k and k + 1 times if m > k. If Ek (a; f ) = Ek (a; g),
we say that f, g share the value a with weight k.

The definition implies that if f, g share a value a with weight k, then

z0 is an a-point of f with multiplicity m(≤ k) if and only if it is an a-point
2010 Mathematics Subject Classification: 30D35
Key words and phrases: Difference operator, uniqueness, Entire function, weighted
sharing

© Indian Mathematical Society, 2024 .

42
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 43

of g with multiplicity m(≤ k) and z0 is an a-point of f with multiplicity

m(> k) if and only if it is an a-point of g with multiplicity m(> k), where
m is not necessarily equal to n. We write f, g share (a, k) to mean that
f, g share the value a with weight k. Clearly if f, g share (a, k) then f, g
share (a, p) for any integer p, 0 ≤ p < k. Also we note that f, g share a
value a IM or CM if and only if f, g share (a, 0) or (a, ∞) respectively.
For a finite value z0 , if f (z0 ) = z0 or z0 is a zero of f (z) − z then we say
that a finite value z0 is called a fixed point of f .
We use the notion (m∗ ) defined by

0, if m = 0,
m∗ =
m, if m ∈ N,
for the sake of simplicity.

Let f (z) be a transcendental meromorphic function, n be a positive

integer. Many authors have investigated the value distributions of f n f 0 .
In 1959, Hayman [4] proved that f n f 0 takes every non-zero complex value
infinitely often if n ≥ 3. A similar proposition was proved by Mues [12]
in 1979 for n = 2. Bergweiler and Eremenko [1] showed that f f 0 − 1 has
infinitely many zeros.
Laine and Yang [9, Theorem 2], investigated, corresponding to the above
results, the value distribution of difference product of entire functions, and
obtained the following result.

Theorem 1.2. [9] Let f be a transcendental entire function of finite order,

and c be a non-zero complex constant. Then, for n ≥ 2, f n (z)f (z + c)
assumes every non-zero value a ∈ C infinitely often.

It is interesting to recall Yang and Hua [14] results corresponding to

Theorem 1.2, which may be regarded as a gateway to new research to be
conducted on the sharing of values between differential polynomials.

Theorem 1.3. [11] Let f amd g be two non-constant entire functions, n ∈

N such that n ≥ 6. If f n f 0 and g n g 0 share 1 CM, then either f (z) =
c1 ecz , g(z) = c2 c−cz , where c1 , c2 , c ∈ C statisifying 4(c1 c2 )(n+1) c2 = −1,
or f ≡ tg for a constant t such that tn+1 = 1.
44 MANJUNATH B. E., HARINA P. WAGHAMORE

In 2001, Fang and Hong studied the uniqueness of differential polyno-

mials of the form f n (f − 1)f 0 and g n (g − 1)g 0 and proved the following
uniqueness result.

Theorem 1.4. [3] Let f amd g be two non-constant entire functions, and
let n ≥ 11 be a positive integer. If f n (f − 1)f 0 and g n (g − 1)g 0 share the
value 1 CM, then f = g.

In 2004, Lin and Yi extended the above result in the fixed point view,
and they proving the following.

Theorem 1.5. [10] Let f amd g be two transcendental entire functions,

and let n ≥ 7 be a positive integer. If f n (f − 1)f 0 and g n (g − 1)g 0 share z
CM, then f = g.

In 2010, Zhang got an analog result in the difference.

Theorem 1.6. [17] Let f amd g be two transcendental entire functions of

finite order and α(z) be a small function with respect to both f (z) and g(z).
Suppose that c is a non-zero complex constant and n ≥ 7 is an integer. If
f n (z)(f (z) − 1)f (z + c) and g n (z)(g(z) − 1)g(z + c) share α(z) CM, then
f (z) ≡ g(z).

In 2010, Qi, Yang and Liu [15] proved the following uniqueness theorem
regarding shift operator, which is a difference counterpart of Theorem 1.3.

Theorem 1.7. [15] Let f and g be transcendental entire functoins of finite

order, let c be a non-zero complex constant, and let n ≥ 6 be an integer. If
f n f (z + c) and g n g(z + c) share z CM, then f (z) ≡ tg(z) for a constant t
satisifying tn+1 = 1.

Theorem 1.8. [15] Let f and g be transcendental entire functoins of finite

order, let c be a non-zero complex constant, and let n ≥ 6 be an integer.
If f n f (z + c) and g n g(z + c) share 1 CM, then f g ≡ t2 or f ≡ t3 g for a
constant t2 and t3 that satisfy tn+1
3 = 1.

So we have seen that there are many generalizations of the difference

polynomial. The purpose of this paper is to study the uniqueness problem
for more general difference polynomials f n P (f )∆c f (z) and g n P (g)∆c g(z)
sharing a non-zero polynomial p(z). Here we also note the other occurrence
for P (z).
We now present the following theorem, which is the main result of the paper.
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 45

Theorem 1.9. Let f and g be transcendental entire functoins of finite or-

der, let c ∈ C \ {0} be a complex constant and let p(z) be a non-zero polyno-
mial with deg(p) ≤ n − 1, n(≥ 2), m∗ be two integers such that n > m∗ + 7.
Let P (z) = am z m + am−1 z m−1 + · · · + a1 z + a0 be a non-zero polynomial.
If f n P (f )∆c f (z) − p(z) and g n P (g)∆c g(z) − p(z) share (0, 2) then

(1) When P (z) = am z m + am−1 z m−1 + · · · + a1 z + a0 is a non-zero

polynomial, one of the following three cases holds,
(a) f (z) = tg(z) for a constant t such that td = 1,
where d = gcd(n + 1, · · · , n + m + 1 − i, · · · , n + m + 1),
(b) f and g satisify the algebraic equation R(f, g) ≡ 0, where

R(w1 , w2 ) = w1n P (w1 )∆c w1 (z) − w2n P (w2 )∆c w2 (z),

(c) P (z) reduces to a non-zero monomial, namely P (z) = ai z i 6≡ 0,

for i ∈ {0, 1, 2, · · · , m} if p(z) is a non-zero constant b, then
f (z) = c1 ed1 z , g(z) = c2 ed2 z where d1 , d2 , c1 , c2 are non-zero
constants such that ai (c1 c2 )(n+1+i) ed1 z + e−d1 z − 2 = −b2 ;

(2) When P (z) = z m − 1, then f ≡ tg for some constant t such that

tm = 1;
(3) When P (z) = (z−1)m (m ≥ 2), one of the following two cases holds:
(a) f (z) = g(z),
(b) f and g satisify the algebraic equation R(f, g) ≡ 0, where

R(w1 , w2 ) = w1n P (w1 )∆c w1 (z) − w2n P (w2 )∆c w2 (z),

(4) When P (z) ≡ c0 , then f ≡ tg for some constant t such that tn+1 =
1.

2. Necessary Lemmas

Lemma 2.1. [2] Let f be a meromorphic function of finite order σ, and let
c ∈ C \ {0} be fixed. Then for each  > 0, we have
   
f (z + c) f (z)
m r, + m r, = O(rσ−1+ ) = S(r, f ).
f (z) f (z + c)
The following lemma has a few modifications to the original version [2,
Corollary 2.5]
46 MANJUNATH B. E., HARINA P. WAGHAMORE

Lemma 2.2. [2] Let f be a transcendental meromorphic function of finite

order, c ∈ C \ {0} be fixed. Then

T (r, f (z + c)) = T (r, f ) + S(r, f ).

Lemma 2.3. [5] Let f be a non-constant meromorphic function of finite

order and c ∈ C. Then N (r, 0; f (z + c)) ≤ N (r, 0; f (z)) + S(r, f ),
N (r, ∞; f (z +c)) ≤ N (r, ∞; f )+S(r, f ), N (r, 0; f (z + c)) ≤ N (r, 0; f (z))+
S(r, f ),
N (r, ∞; f (z + c)) ≤ N (r, ∞; f ) + S(r, f ).

Lemma 2.4. [16] Let f be a non-constant meromorphic function and let

an (z)(6≡ 0), an−1 (z), ..., a0 (z) be meromorphic functions such that
T (r, ai (z)) = S(r, f ) for i = 0, 1, 2, ..., n. Then

T (r, an f n + an−1 f n−1 + ... + a1 f + a0 ) = nT (r, f ) + S(r, f ).

Lemma 2.5. Let f be an entire function of finite order ρ, c be a fixed non-

zero complex constant and let n ∈ N and P (z) be defined as in Theorem 1.9
then for each  > 0, we have

T (r, f n P (f )∆c f (z)) = (n + m∗ + 1)T (r, f ) + O(rρ−1+ ).

Proof. By Lemma 2.1,

T (r, f n P (f )∆c f ) = m (r, f n P (f )∆c f ) ,
≤ T r, f n+1 P (f ) + O rρ−1+ ,

≤ (n + m∗ + 1)T (r, f ) + s(r, f ).

(n + m∗ + 1)T (r, f ) = T r, f n+1 P (f ) ,

 
n f
≤ m (r, f P (f )∆c f ) + m r, ,
∆c f
≤ m (r, f n P (f )∆c f ) + S(r, f ).
Therefore,

T (r, f n P (f )∆c f (z)) = (n + m∗ + 1)T (r, f ) + S(r, f ).

Lemma 2.6. Let f be a transcendental entire function of finite order ρ,

c ∈ C \ {0} be a complex constant , n(≥ 1), m∗ (≥ 0) be two integer and let
a(z)(6≡ 0 , ∞) be a small function of f . If n > 2, then f n P (f )∆c f − a(z)
has infinitely many zero’s.
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 47

Proof. Let Ψ = F n P (f )∆c f , now from Lemma 2.3, and [13], we get
   
1 1
T (r, ψ) ≤ N r, + N (r, ψ) + N r, + ( + o(1)) ,
ψ ψ − a(z)
     
1 1 1
≤ 3N r, + N r, + N r, + ( + o(1)) ,
f P (f ) ψ − a(z)
 
∗ 1
≤ (3 + m )T (r, f ) + N r, + ( + o(1)) ,
ψ − a(z)
from Lemma 2.4 and 2.5,
 
∗ ∗ 1
(n + m + 1) T (r, f ) ≤ (3 + m )T (r, f ) + N r, .
ψ − a(z)
Take  < 1. Since n > 2 from above, one can easily say that ψ − a(z) has
infinitely many zero’s. 

Lemma 2.7. [Hadamard Factorization Theorem] Let f be an entire func-

tion of finite order ρ with zero’s a1 , a2 , · · · , an each zero is counted as often
as its multiplicity. Then f can be expressed in the form f (z) = Q(z)eα(z) ,
where α(z) is a polynomial of degree not exceeding [ρ] and Q(z) is the canon-
ical product formed with the zeros of f .

Lemma 2.8. [6] Let f and g be two non-constant meromorphic functions

sharing (1, 2). Then one of the following holds:
   
(1) T (r, f ) ≤ N2 r, f1 + N2 r, g1 + N2 (r, f ) + N2 (r, g) + s(r, f ) +
s(r, g),
(2) f g = 1,
(3) f ≡ g.

Lemma 2.9. Let f, g be two transcendental entire functions of finite order,

c ∈ C \ {0} be a complex constant and n ∈ N and p be a non-zero polynomial
such that deg (p) ≤ (n − 1) where n ∈ N. Let P (z) be a non-zero polynomial
de?ned as in Theorem 1.9. Suppose

(f n P (f )∆c f (z)) (g n P (g)∆c g(z)) ≡ p2 ,

then P (z) reduces to a non-zero monomial namely P (z) = ai z i 6= 0. For i ∈

{0, 1, 2, · · · , m}. If p(z) = b ∈ C \ {0} then f (z) = c1 ed1 (z) , g(z) = c2 ed2 (z)
where d1 , d2 , c1 , and c2 are non-zero constants such that,
 
a2i (c1 c2 )(n+1+i) ed1 c + e−d1 c − 2 = −b2
48 MANJUNATH B. E., HARINA P. WAGHAMORE

Proof. Let F = f n P (f )∆c f and G = g n P (g)∆c g,

suppose
F G = p2 , (2.1)
we consider the following cases,
Case I. Let deg p(z) = l(≥ 1), since  F and G are transcendental
  entire
1 1
functions, we deduce by 2.1 that N r, f = O(logr) = N r, g .
Suppose P (z) is not a non-zero monomial. For the sake of simplicity let
P (z) = z − a where a ∈ C \ {0}, clearly Θ(0, f ) + Θ(a, f ) = 2, which is
impossible for a entire function. Thus P (z) reduces to a non-zero monomial,
namely P (z) = ai z i 6= 0 for some i ∈ {0, 1, 2, ..., m} and so 2.1 reduces to

a2i f n+i ∆c f g n+i ∆c g ≡ p2 . (2.2)

   
From 2.2 it follows that N r, f1 = O (logr) = N r, g1 . From Lemma 2.7,
we obtain that f = h1 eα1 and g = h2 eβ1 , where h1 , h2 are two nonconstant
ploynomials and α1 , β1 are two non constant polynomials. By virtue of the
polynomial p(z), from 2.2 we arrive at a contradiction.
Case II. Let p(z) = b ∈ C \ {0}. Then from 2.1, we have

(f n P (f )∆c f (z)) (g n P (g)∆c g(z)) ≡ b2 , (2.3)

Now from the assumption that f and g are two nonconstant entire functions,
we deduce by 2.3 that f n P (f ) 6= 0 and g n P (g) 6= 0. By Picard’s theorem,
we claim that P (z) = a1 z i 6≡ 0 for i ∈ {0, 1, 2, · · · , m}, otherwise the
Picard’s exception values are atleast three which is contradiction. Then 2.3
reduces to
(a2i f n+i ∆c f )(g n+i ∆c g) ≡ b2 . (2.4)
Hence by Lemma 2.7, we obtain that

f = eα , g = eβ (2.5)
  
a2i e(α+β)(n+i) eα(z+c) − eα(z) eβ(z+c) − eβ(z) = b2 ,
  
eα(z+c)−α(z) − 1 eβ(z+c)−β(z) − 1 = d2 e(α(z)+β(z))(n+i+1) (2.6)
we conclude that from 2.6 that eα(z+c)−α(z) − 1 has no zero’s.
Let ψ = eα(z+c)−α(z) , then ψ 6= 0, 1, ∞ for any z ∈ C. By picard theorem,
ψ is constant, so deg(α(z)) = 1. Similarly we can prove that deg(β) = 1.
Assume now that f (z) = c1 ed1 (z) , g(z) = c2 ed2 (z) where d1 , d2 , c1 , c2 are
non-zero constants.
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 49

From 2.3, we get d1 = −d2 and a2i (c1 c2 )(n+1+i) ed1 c + e−d1 c − 2 = −b2 .

Hence the proof. 

Lemma 2.10. Let f and g be two transcendental entire functions of finite

order c ∈ C \ {0} and n ∈ N with n > 1. If f n P (f )∆c f ≡ g n P (g)∆c g,
where P (z) is defined as in the Theorem 1.9 then
(1) When P (z) = am z m + am−1 z m−1 + · · · + a1 z + a0 is a non-zero
polynomial, one of the following three cases holds,
(a) f (z) = tg(z) for a constant t such that td = 1,
where d = gcd(n + 1, · · · , n + m + 1 − i, · · · , n + m + 1),
(b) f and g satisify the algebraic equation R(f, g) ≡ 0, where

R(w1 , w2 ) = w1n P (w1 )∆c w1 (z) − w2n P (w2 )∆c w2 (z),

(2) When P (z) = z m − 1, then f ≡ tg for some constant t such that

tm = 1;
(3) When P (z) = (z −1)m (m ≥ 2), one of the following two cases holds:
(a) f (z) = g(z),
(b) f and g satisify the algebraic equation R(f, g) ≡ 0, where

R(w1 , w2 ) = w1n P (w1 )∆c w1 (z) − w2n P (w2 )∆c w2 (z),

(4) When P (z) ≡ c0 , then f ≡ tg for some constant t such that tn+1 =
1.

Proof. Suppose
f n P (f )∆c f ≡ g n P (g)∆c g. (2.7)
Since g is transcendental entire function, hence g(z), g(z+c) 6= 0 we consider
following cases,
Case 1. P (z) = c0 ,
let h = fg , if h is a constant by putting f = hg in 2.7, we get
     
am g m h(n+m+1) − 1 +am−1 g m−1 h(n+m) − 1 +· · ·+a0 h(n+1) − 1 ≡ 0,

which implies that hd = 1, where d is such that td = 1, where d is the gcd

of the elements of J = {p ∈ I : ap 6= 0} and I = {n + 1, n + 2, · · · , n + m +
1 − i, · · · , n + m + 1}.
Thus f = tg for a constant t such that td = 1, where d is the GCD of the
elements of J = {p ∈ I : ap 6= 0} and I = {n + 1, n + 2, · · · , n + m + 1 −
i, · · · , n + m + 1} i ∈ {0, 1, 2, · · · , m}.
50 MANJUNATH B. E., HARINA P. WAGHAMORE

If h is not a constant then we know that by 2.7 that f and g satisifying

the algebraic equation R(f, g) = 0, where R(w1 , w2 ) = w1n P (w1 )∆c w1 (z) −
w2n P (w2 )∆c w2 (z). We now discuss the following subcases,
Subcase 1.1. P (z) = z m − 1, Then from 2.7, we have

f n (f m − 1)∆c f ≡ g n (g m − 1)∆c g. (2.8)

f
Let h = g and from 2.8, we get
(g m − 1)∆c g (f m − 1)∆c f
hn+1 = .
g f
If h is not constant, then we have
(g m − 1)∆c g (f m − 1)∆c f
   
(n + 1)T (r, h) ≤ T r, + T r, + S(r, f )
g f
+ S(r, g),
(n + 1)T (r, h) ≤ (m + 1) [T (r, f ) + T (r, g)] + S(r, f ) + S(r, g).
Combining the above inequality with,
 
f
T (r, h) = T r, = T (r, f ) + T (r, g) + S(r, f ) + S(r, g),
g
we obtain,

(n − m) (T (r, f ) + T (r, g)) ≤ S(r, f ) + S(r, g),

which is impossible. Therefore h is a constant, then by substituting f = gh

into 2.8, we get

(gh)n ((gh)n − 1)∆c (gh) = g n (g n − 1)∆c g,

 
g m h(n+m+1) − 1 = hn+1 − 1. (2.9)
Since h is constant and g is transcendental entire function, from 2.9, we
have h(n+m+1) − 1 = 0 if and only if hn+1 − 1 = 0 and so hm = 1. Thus
f (z) = tg(z) for a constant t such that tm = 1.
Subcase 1.2. Let P (z) = (z − 1)m . Then from 2.7, we have

f n (f − 1)m ∆c f ≡ g n (g − 1)m ∆c g. (2.10)

Let h = fg , if m = 1 then the result follows from subcase 1.1

For m ≥ 2, first we suppose that h is non-constant then from 2.10, we can
say that f and g satisify the algebraic equation R(f, g) = 0,
where R(w1 , w2 ) = w1 (w1 −1)m ∆c w2 −w2 (w2 −1)m ∆c w2 . Next we suppose
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 51

that h is constant, then from 2.10,we get

m   m  
n
X
i m m−i n
X
i m
f ∆c f (z) (−1) f = g ∆c g(z) (−1) g m−i
m−i m−i
i=0 i=0
(2.11)
Now substitute f = gh in 2.11, we get
m  
X
m m
g m−i hn+m+1−i − 1 ≡ 0,
 
(−1)
m−i
i=0

which implies that h = 1. Hence f ≡ g.

Case.2 P (z) = c0 .
Let h = fg , then from 2.7, we have

f n c0 ∆c f = g n c0 ∆c g,
∆c g f
hn+1 = . . (2.12)
g ∆c f
Therefore from Lemma 2.3 and 2.4,
   
∆c g f
(n + 1)T (r, h) = T r, + T r, ,
g ∆c f
≤ 2 [T (r, f ) + T (r, g)] + S(r, f ) + S(r, g),
(n + 1) (T (r, f ) + T (r, g)) ≤ 2 [T (r, f ) + T (r, g)] + S(r, f ) + S(r, g).
Contradicts with n ≥ 2. Hence h must be constant, which implies that
hn+1 = 1, thus f ≡ tg and tn+1 = 1. Which completes the proof. 

3. Proof of Theorem
f n P (f )∆
cf
n
Proof. Let F = p(z) and G = g Pp(z)
(g)∆c g
Then F, G share (1, 2) except
the zero’s of p(z). Now applying Lemma 2.8, we see that one of the following
three cases holds
Case 1. Suppose
   
1 1
T (r, F ) ≤ N2 r, + N2 r, + S(r, F ) + S(r, G),
F G
      
1 1 1
≤ 2 N r, + N r, + N r,
f g P (f )
     
1 1 1
+ N r, + N r, + N r,
∆c f P (g) ∆c g
+ S(r, f ) + S(r, g),
52 MANJUNATH B. E., HARINA P. WAGHAMORE

T (r, F ) ≤ (m∗ + 4) (T (r, f ) + T (r, g)) + S(r, f ) + S(r, g),

≤ (2m∗ + 8)T (r) + S(r).

From Lemma 2.5,

(n + m∗ + 1)T (r, f ) ≤ (2m∗ + 8)T (r) + S(r). (3.1)

Similarly, we have

(n + m∗ + 1)T (r, g) ≤ (2m∗ + 8)T (r) + S(r). (3.2)

Combining 3.1 and 3.2, we get

(n + m∗ + 1)T (r) ≤ (2m∗ + 8)T (r) + S(r)

which contradicts with n > m∗ + 7.

Case 2. F ≡ G,
f n P (f )∆c f = g n P (g)∆c g and so the result follows from Lemma 2.10.
Case 3. F G ≡ 1.
Then we have (f n P (f )∆c f ) (g n P (g)∆c g) ≡ p2 and so the result follows
from Lemma 2.9.

Which completes the proof.



Acknowledgement: The author is also grateful to the anonymous referee

for his/her valuable suggestions which considerably improved the presenta-
tion of the paper.

References
[1] W. Bergweiler and A. Eremenko, On the singularities of the inverse to a meromorphic
function of finite order, Rev. Mat. Iberoamericana 11 (1995), no. 2, 355–373.
[2] Y.-M. Chiang and S.-J. Feng, On the Nevanlinna characteristic of f (z + η) and
difference equations in the complex plane, Ramanujan J. 16 (2008), no. 1, 105–129.
[3] M.-L. Fang and W. Hong, A unicity theorem for entire functions concerning differ-
ential polynomials, Indian J. Pure Appl. Math. 32 (2001), no. 9, 1343–1348.
[4] W. K. Hayman, Picard values of meromorphic functions and their derivatives, Ann.
of Math. (2) 70 (1959), 9–42.
[5] J. Heittokangas et al., Value sharing results for shifts of meromorphic functions, and
sufficient conditions for periodicity, J. Math. Anal. Appl. 355 (2009), no. 1, 352–363.
[6] I. Lahiri, Weighted value sharing and uniqueness of meromorphic functions, Complex
Variables Theory Appl. 46 (2001), no. 3, 241–253.
 DIFFERENCE POLYNOMIALS OF ENTIRE FUNCTIONS... 53

[7] I. Lahiri, Weighted sharing and uniqueness of meromorphic functions, Nagoya Math.
J. 161 (2001), 193–206.
[8] I. Laine, Nevanlinna theory and complex differential equations, De Gruyter Studies
in Mathematics, 15, Walter de Gruyter & Co., Berlin, 1993.
[9] I. Laine and C.-C. Yang, Value distribution of difference polynomials, Proc. Japan
Acad. Ser. A Math. Sci. 83 (2007), no. 8, 148–151.
[10] W. Lin and H. Yi, Uniqueness theorems for meromorphic functions concerning fixed-
points, Complex Var. Theory Appl. 49 (2004), no. 11, 793–806.
[11] K. Liu and L.-Z. Yang, Value distribution of the difference operator, Arch. Math.
(Basel) 92 (2009), no. 3, 270–278.
[12] E. Mues, Über ein Problem von Hayman, Math. Z. 164 (1979), no. 3, 239–259.
[13] K. Yamanoi, The second main theorem for small functions and related problems,
Acta Math. 192 (2004), no. 2, 225–294.
[14] C.-C. Yang and X. Hua, Uniqueness and value-sharing of meromorphic functions,
Ann. Acad. Sci. Fenn. Math. 22 (1997), no. 2, 395–406.
[15] X.-G. Qi, L.-Z. Yang and K. Liu, Uniqueness and periodicity of meromorphic func-
tions concerning the difference operator, Comput. Math. Appl. 60 (2010), no. 6,
1739–1746.
[16] C.-C. Yang and H.-X. Yi, Uniqueness theory of meromorphic functions, Mathematics
and its Applications, 557, Kluwer Academic Publishers Group, Dordrecht, 2003.
[17] J. Zhang, Value distribution and shared sets of differences of meromorphic functions,
J. Math. Anal. Appl. 367 (2010), no. 2, 401–408.

Manjunath B.E.
Department of Mathematics
Bangalore University
Jnanabharathi Campus
Bangalore 560056, INDIA.
E-mail: [email protected]

Harina P. Waghamore
Department of Mathematics
Bangalore University
Jnanabharathi Campus
Bangalore 560056, INDIA.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2 , January-June (2024), 54–66

NEW METHOD OF LINES SOLUTION FOR THE

NONLINEAR SINE GORDON EQUATION WITH
REPRODUCING KERNEL HILBERT SPACE METHOD

GAUTAM PATEL AND KAUSHAL PATEL

(Received : 28 - 08 - 2022 ; Revised : 15 - 06 - 2023)

Abstract. The paper reports a method of lines (MOL) for gener-

ating a solution for the sine Gordon equation in one dimension with
suitable initial and boundary conditions based on the reproducing ker-
nel Hilbert space method (RKHSM). Two valid test examples are given
to determine the validity and effectiveness of the current technique. We
observed that the presented method has better accuracy and efficiency
compared to the other methods in the literature.

1. Introduction

Nonlinearity appear in many branches of scientific fields such as plasma

physics, solid state physics, mathematical biology, fluid dynamics and chem-
ical kinetics, can be modeled by partial differential equations (PDEs). A
broad class of analytical and numerical solution methods have been used to
handle these problems.
In this article, we consider the following one dimensional sine Gordon equa-
tion
∂2u ∂2u
= − sin(u), (x, t) ∈ Ω × Γ, (1.1)
∂t2 ∂x2
where u = u(x, t) represents the wave displacement at position x and time
t and sin(u) is the nonlinear force, with initial and boundary conditions

u(x, 0) = Θ1 (x), ut (x, 0) = Θ2 (x), x ∈ Ω, (1.2)

u(a, t) = Φ1 (t), u(b, t) = Φ2 (t), t ∈ Γ, (1.3)

2010 Mathematics Subject Classification: 46C99, 46E22, 37N99, 37N25

Key words and phrases: Mathematical model, Nonlinear system of ODEs,
Reproducing kernel Hilbert space method, Method of Lines

© Indian Mathematical Society, 2024 .

54
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 55

where Ω = (a, b), Γ = (0, T ] with 0 < T < ∞.

The sine Gordon equation is a nonlinear hyperbolic equation, which is a
model of soliton wave. It was first introduced by Edmond Bour [6] in the
study of surfaces of constant negative curvature and rediscovered by Frenkel
and Kontorova [12] in the study of the Frenkel-Kontorova model. Also, the
sine Gordon equation appears in many applications like stability of fluid
motion, differential geometry, applied sciences and nonlinear physics [3],
propagation of fluxion in Josephson junctions [25].
The study of the solution of nonlinear sine Gordon wave equation has been
gained much attention in the past several decades. It has been solved by
Ablowitz et al. [1] using the inverse-scattering method. Ben-Yu et al. [4]
presented a numerical solution by two difference schemes. Bratsos [7] solved
the equation using a fourth-order rational approximation to the matrix ex-
ponential term in a three time level recurrence relation for the numerical
solution. Dehghan and Shokri [11] used radial basis functions with collo-
cation points to approximate the solution. Uddin et al. [27] evaluated nu-
merical solution using a meshfree approach based on radial basis function.
Mittal and Bhatia [19] discussed for the numerical solution using modified
cubic B-spline collocation method based on collocation of modified cubic
B-splines over finite elements.
In this paper we present a semi analytical method of lines (MOL) solu-
tion. The MOL were applied to solve the PDEs in fully numerical based by
Schiesser et al. [26]. In 2004, Koto [15] applied this technique to approxi-
mations of delay differential equations using Runge Kutta method. Hamdi
et al. [14] gave basic idea of MOL. The semi analytic MOL is used actively
for solving linear PDEs. For example see [21, 22, 23].
In our work a different approach is used, the usual finite difference scheme
is employed for spatial discretization in the nonlinear initial boundary val-
ued PDE to convert nonlinear initial valued system of ordinary differential
equations (ODEs) and then using RKHSM to find a solution. The theory
of reproducing kernels [24] dates to the first half of the 20th century, and
its roots go back to the pioneering papers by S. Zeremba [28], Mercer [18],
and Bergman [5]. In 1950, N. Aronszajn [2] outlined the past works and
gave a systematic reproducing kernel theory and laid a good foundation for
the research of each special case and greatly simplified the proof. This the-
ory has been successfully applied on linear and nonlinear applications with
56 GAUTAM PATEL AND KAUSHAL PATEL

different type conditions by many researchers [8, 9, 13, 16]. The analytic
solution of RKHSM is represented in the form of series. The RKHSM is
easily implemented, grid free and without time discretization. Also, we can
evaluate the solution for finite number of points and use it often.
The paper is laid out as follows. In the next section, we show how we use
MOL to solve the sine Gordon equation. The results of numerical exper-
iments are presented in Section 3. Final Section is dedicated to a brief
conclusion. Finally the references are listed at the end.

2. Method of Lines

In this section, we derive MOL to solve the sine Gordon equation using
RKHSM. To do this, we divide the section into two subsections. In the first
subsection, we discretize the spatial derivatives to obtain a system of ODEs
in the time variable. In the second subsection, we explain RKHSM to solve
the system of ODEs.

2.1. Discretization. To use the MOL for solving (1.1)-(1.3), we discretize

the spatial coordinate x with m − 1 grid points xi = xi−1 + h, h = (b −
a)/m, x0 = a, xm = b, i = 1, 2, . . . , m − 1. We apply a second-order dif-
ference approximation for the second-order derivative in x in grid points
x1 , xm−1 and a fourth-order difference approximation to the second-order
derivative in x in grid points xi , i = 2, 3, . . . , m − 2.
Let us consider ui (t) approximate u(xi , t). Here, using the central differ-
ence approximations in second-order and forth-order for the second order
derivative in x of (1.1), we get
d2 u1 u0 − 2u1 + u2
= − sin(u1 ),
dt2 h2
d2 um−1 um−2 − 2um−1 + um
= − sin(um−1 ),
dt2 h2 (2.1)
d2 ui −ui−2 + 16ui−1 − 30ui + 16ui+1 − ui+2
= − sin(ui ),
dt2 12h2
i = 2, 3, . . . , m − 2.
Further, the conditions (1.2) and (1.3) become
dui (0)
ui (0) = Θ0 (xi ), = Θ1 (xi ), i = 1, 2, 3, . . . , m − 1,
dt
u0 = Φ1 (t), um = Φ2 (t).
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 57

We change system (2.1) to a system of the first-order ODEs by using

dui
vi =, i = 1, 2, . . . , m − 1.
dt
Consequently, we get the following initial values system of 2m − 2 equations
dui
=vi i = 1, 2, . . . , m − 1,
dt
dv1 Φ1 (t) − 2u1 + u2
= − sin(u1 ),
dt h2
dv2 −Φ1 (t) + 16u1 − 30u2 + 16u3 − u4
= − sin(u2 ),
dt 12h2
dvi −ui−2 + 16ui−1 − 30ui + 16ui+1 − ui+2
= − sin(ui ),
dt 12h2
i = 3, 4, . . . , m − 3, (2.2)
dvm−2 −um−4 + 16um−3 − 30um−2 + 16um−1 − Φ2 (t)
=
dt 12h2
− sin(um−2 ),
dvm−1 um−2 − 2um−1 + Φ2 (t)
= − sin(um−1 ),
dt h2
with initial conditions
ui (0) = Θ0 (xi ), vi (0) = Θ1 (xi ), i = 1, 2, . . . , m − 1. (2.3)

In the next section, we will discuss the RKHSM to solve the first-order
nonlinear system of ODEs with homogeneous initial conditions.

2.2. The Reproducing Kernel Hilbert Space Method. In this section,

firstly we construct a homogeneous initial values system of ODEs from the
equations (2.2) and (2.3).
To do this, we take ui = Θ0 (xi )(1 − t) + wi , vi = Θ1 (xi )(1 − t) + wm−1+i ,
i = 1, 2, . . . , m − 1, we get
dwi
=Θ0 (xi ) + Θ1 (xi )(1 − t) + wm−1+i i = 1, 2, . . . , m − 1,
dt
dwm Φ1 (t) − 2(Θ0 (x1 )(1 − t) + w1 ) + Θ0 (x2 )(1 − t) + w2
=Θ1 (x1 ) +
dt h2
− sin(Θ0 (x1 )(1 − t) + w1 ),
dwm+1 −Φ1 (t) + 16(Θ0 (x1 )(1 − t) + w1 ) − 30(Θ0 (x2 )(1 − t) + w2 )
=Θ1 (x2 ) +
dt 12h2
16(Θ0 (x3 )(1 − t) + w3 ) − (Θ0 (x4 )(1 − t) + w4 )
+ − sin(Θ0 (x2 )(1 − t) + w2 ),
12h2
dwm−1+i −(Θ0 (xi−2 )(1 − t) + wi−2 ) + 16(Θ0 (xi−1 )(1 − t) + wi−1 )
=Θ1 (xi ) +
dt 12h2
−30(Θ0 (xi )(1 − t) + wi ) + 16(Θ0 (xi+1 )(1 − t) + wi+1 ) − (Θ0 (xi+2 )(1 − t) + wi+2 )
+
12h2
58 GAUTAM PATEL AND KAUSHAL PATEL

− sin(Θ0 (xi )(1 − t) + wi ), i = 3, 4, . . . , m − 3, (2.4)

dw2m−3 −(Θ0 (xm−4 )(1 − t) + wm−4 ) + 16(Θ0 (xm−3 )(1 − t) + wm−3 )
=Θ1 (xm−2 ) +
dt 12h2
−30(Θ0 (xm−2 )(1 − t) + wm−2 ) + 16Θm−1 − Φ2 (t)
+
12h2
− sin(Θ0 (xm−2 )(1 − t) + wm−2 ),
dw2m−2 Θ0 (xm−2 )(1 − t) + wm−2 − 2(Θ0 (xm−1 )(1 − t) + wm−1 ) + Φ2 (t)
=Θ1 (xm−1 ) +
dt h2
− sin(Θ0 (xm−1 )(1 − t) + wm−1 ),

with homogeneous initial conditions

wi (0) = 0, i = 1, 2, . . . , 2m − 2. (2.5)

Now, we introduce the reproducing kernel Hilbert spaces W22 [0, T ] and
W21 [0, T ] with corresponding reproducing kernel functions R(t, s) and G(t, s),
respectively, to generate the algorithm of the RKHSM.

Definition 2.1. [10] Consider H is a Hilbert space of real or complex valued

functions defined on a nonempty set X endowed with hf, giH .
A function R(·, ·) : X × X → F ( F denotes R or C) is called a reproducing
kernel of the Hilbert space H if it satisfies,

hf, R(·, s)iH = f (s), (2.6)

for each fixed s ∈ X.

The equation (2.6) is known as “the reproducing property”.

Definition 2.2. [20] The inner product space W22 [0, T ] is defined as W22 [0, T ] =
{w : w, w0 are absolutely continuous real valued functions on [0, T ], w00 ∈
L2 [0, T ], and w(0) = 0} with the inner product and the norm of W22 [0, T ]
are defined, respectively, by
Z T 2
dw(0) dy(0) d w(t) d2 y(t)
hw, yiW22 = w(0)y(0) + + dt,
dt dt 0 dt2 dt2
q
kwkW22 = hw, wiW22 .

Theorem 2.3. [20] The Hilbert space W22 [0, T ] is a reproducing kernel
Hilbert space and its reproducing kernel function R(·, ·) can be written as

 s (6t + 3ts − s2 ), s ≤ t,
R(t, s) = 6
 t (6s + 3ts − t2 ), s > t.
6
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 59

Definition 2.4. [17] The inner product space W21 [0, T ] is defined as W21 [0, T ] =
{w : w is absolutely continuous real valued function on [0, T ], w0 ∈ L2 [0, T ]}
with the inner product and the norm of W21 [0, T ] are defined, respectively,
by
Z T
dw(t) dy(t)
hw, yiW21 = w(0)y(0) + dt,
0 dt dt
q
kwkW21 = hw, wiW21 .

Theorem 2.5. [17] The Hilbert space W21 [0, T ] is a reproducing kernel
Hilbert space and its reproducing kernel function G(·, ·) can be written as

1 + s, s ≤ t,
G(t, s) =
1 + t, s > t.

We interpret a differential operator for Lwi (t) = dw dt , i = 1, 2, . . . , 2m−2

i

as L : W22 [0, T ] → W21 [0, T ], such that equations (2.4) and (2.5) become
Lwi (t) = fi (t, w1 (t), w2 (t), . . . , w2m−2 (t)), 0 < t < T , subject to the initial
conditions, wi (0) = 0, i = 1, 2, . . . , 2m − 2, where wi (·) ∈ W22 [0, T ] and
fi (·, w1 (·), w2 (·), . . . , w2m−2 (·)) ∈ W21 [0, T ].

Theorem 2.6. The operator L : W22 [0, T ] → W21 [0, T ] is bounded and lin-
ear.

Proof. For the proof, we refer to [10]. 

Now, we consider function in the form ψj (·) = L∗ G(·, tj ), j = 1, 2, 3, . . .,

where {tj }∞ ∗
j=1 is dense on [0, T ] and L is the adjoint operator of L. Like,

hwi (·), ψj (·)iW22 = hwi (·), L∗ G(·, tj )iW22

= hLwi (·), G(·, tj )iW21
= Lwi (tj ), j = 1, 2, . . . , i = 1, . . . , 2m − 2.

Since,

ψj (·) = L∗ G(·, tj )
= hL∗ G(·, tj ), R(t, s)iW22
= hG(·, tj ), Ls R(·, s)iW21
= hLs R(·, s), G(·, tj )iW21
60 GAUTAM PATEL AND KAUSHAL PATEL

= Ls R(·, s)|s=tj , j = 1, 2, . . . .

Thus, ψj (·) can be evaluated by ψj (·) = Ls R(·, s)|s=tj , j = 1, 2, 3, . . ., where

Ls is a differential operator with respect to s.
Theorem 2.7. If {tj }∞ ∞
j=1 is dense on [0, T ], then {ψj (·)}j=1 is a complete
system of the space W22 [0, T ].
Proof. For the proof, we refer to [10]. 
Now, we will derive the method of analytical solution of the equations
(2.4) and (2.5) in the reproducing kernel Hilbert space W22 [0, T ].
The orthonormal system Ψj (·) of W22 [0, T ], j = 1, 2, 3, . . . can be constructed
from Gram-Schmidt orthogonalization process of ψj (·), j = 1, 2, 3, . . . with
coefficients βjk , as
j
X
Ψj (·) = βjk ψk (·), j = 1, 2, 3, . . . . (2.7)
k=1

Theorem 2.8. If {tj }∞ j=1 is dense on [0, T ], then the analytic solution of
(2.4) and (2.5) represented by
j
∞ X
X
wi (·) = βjk fi (·, w1 (·), w2 (·), . . . , w2m−2 (·))Ψj (·), i = 1, 2, . . . , 2m − 2.
j=1 k=1
(2.8)
Proof. Let wi (·), i = 1, 2, . . . , 2m − 2 be the solution of (2.4) and (2.5) in
W22 [0, T ]. Therefore, ∞
P
j=1 hwi (·), Ψj (·)iW22 Ψj (·), i = 1, 2, . . . , 2m−2 are the
expansion of about orthonormal system Ψj (·) and convergent in the sense
of k.kW22 . On the other hand, using (2.7) yields that

∞
X
wi (·) = hwi (·), Ψj (·)iW22 Ψj (·)
j=1
∞
X j
X
= hwi (·), βjk ψk (·)iW22 Ψj (·)
j=1 k=1
j
∞ X
X
= βjk hwi (·), ψk (·)iW22 Ψj (·)
j=1 k=1
j
∞ X
X
= βjk hwi (·), L∗ G(·, tk )iW22 Ψj (·)
j=1 k=1
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 61

j
∞ X
X
= βjk hLwi (·), G(·, tk )iW21 Ψj (·)
j=1 k=1
j
∞ X
X
= βjk hfi (·, w1 (·), w2 (·), . . . , w2m−2 (·)), G(·, tk )iW21 Ψj (·)
j=1 k=1
j
∞ X
X
= βjk fi (tk , w1 (tk ), w2 (tk ), . . . , w2m−2 (tk ))Ψj (%),
j=1 k=1

i = 1, 2, . . . , 2m − 2.

Thus, (2.8) is the analytical solution of (2.4) and (2.5). 

Here, equation (2.8) is the analytic solution of (2.4) and (2.5). Now,
for the solution, we can define initial conditions as wi,0 (t1 ) = wi (t1 ) and
set n terms approximations to wi (t), i = 1, 2, . . . , 2m − 2 as wi,n (t) =
Pn {i} {i}
j=1 Aj Ψj (t), i = 1, 2, . . . , 2m−2, where the coefficients Aj of Ψj (t) are
{i} P j
given as Aj = k=1 βjk fi (tk , w1,k−1 (tk ), w2,k−1 (tk ), . . . , w2m−2,k−1 (tk )),
i = 1, 2, . . . , 2m − 2.

3. Numerical Experiments

In this section, we consider two nonlinear time dependent sine Gordan

type equations on a finite interval that are implemented to demonstrate the
accuracy and capability of the proposed algorithm, and all of them were
performed on the computer using a program written in Matlab. To show
the efficiency of the presented scheme we calculate the error norms L2 and
L∞ as
v
u m
u X
L2 = kuexact − uM OL k2 = th |(uexact )i − (uM OL )i |2 ,
i=1

L∞ = kuexact − uM OL k∞ = max |(uexact )i − (uM OL )i | .

i

Example 3.1. To test the MOL in the domain [−1, 1], we consider the
initial conditions Θ0 (x) = 0, Θ1 (x) = 4 sech(x) and boundary conditions
Φ1 (t) = 4 arctan(t sech(−1)), Φ2 (t) = 4 arctan(t sech(1)). The analytic so-
lution [19] is given as u(x, t) = 4 arctan(t sech(x)).

In Table 1, we introduce the L2 and L∞ errors between the MOL and

analytic solutions in the domain [−1, 1] with h = 0.04 and ∆t = 0.0001 for
62 GAUTAM PATEL AND KAUSHAL PATEL

Figure 1. The result of the MOL with h = 0.04. (a) The

analytic solution. (b) The MOL solution. (c) The absolute
error.

Table 1. L2 and L∞ errors of Example 3.1 with h = 0.04

at t = 0.25, 0.50, 0.75, 1.00.

MOL Mittal and Bhatia [19] Dehghan and Shokri [11]

t L2 L∞ L2 L∞ L2 L∞
0.25 1.7478 × 10 3.7809 × 10 1.18 × 10 2.32 × 10 3.91 × 10 5.89 × 10−6
−7 −7 −5 −5 −5

0.50 4.8636 × 10−7 6.8728 × 10−7 4.19 × 10−5 4.11 × 10−5 1.30 × 10−4 2.01 × 10−5
0.75 7.0941 × 10−7 7.7246 × 10−7 7.78 × 10−5 1.02 × 10−4 2.35 × 10−4 3.63 × 10−5
1.00 6.7936 × 10−7 6.7041 × 10−7 1.30 × 10−4 1.64 × 10−4 3.27 × 10−4 5.07 × 10−5

Example 3.1. Also, a comparison of some past works is given in the table,
which gives that the proposed method is accurate. The analytic solution,
the MOL solution and the absolute errors of Example 3.1 with h = 0.04 are
displayed in Figure 1.

Example 3.2. The MOL solution of (1.1) is calculated inthe computa-

 
x
tional domain [−3, 3] with initial conditions Θ0 (x) = 4 arctan exp √1−c 2
,
−4cγexp(γx)
Θ1 (x) = 1+exp(2γx) and boundary conditions Φ1 (t) = 4 arctan(exp(γ(−3 −
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 63

ct))), Φ2 (x) = 4 arctan(exp(γ(3 − ct))). The analytic solution [19] is given

√
as u(x, t) = 4 arctan(exp(γ(x − ct))), where γ = 1/ 1 − c2 .

Figure 2. The result of the MOL with h = 0.04. (a) The

analytic solution. (b) The MOL solution. (c) The absolute
error.

Table 2. L2 and L∞ errors of Example 3.2 with h = 0.04

at t = 0.25, 0.50, 0.75, 1.00.

MOL Mittal and Bhatia[19] Dehghan and Shokri [11]

t L2 L∞ L2 L∞ L2 L∞
0.25 5.7347 × 10 6.4305 × 10 3.66 × 10 4.90 × 10 1.76 × 10 4.95 × 10−6
−8 −8 −5 −5 −5

0.50 1.4646 × 10−7 1.5131 × 10−7 9.00 × 10−5 7.55 × 10−5 4.31 × 10−5 8.42 × 10−6
0.75 2.2560 × 10−7 2.4219 × 10−7 1.60 × 10−4 1.43 × 10−4 8.25 × 10−5 1.65 × 10−5
1.00 2.7964 × 10−7 3.0977 × 10−7 2.27 × 10−4 2.10 × 10−4 1.27 × 10−4 2.51 × 10−5

We consider h = 0.04, ∆t = 0.0001 and c = 0.5. The analytic solution,

the MOL solution and the absolute error are plotted in Figure 2. Table
2 illustrates the L2 and L∞ errors and a comparison with the method in
[11, 19] and it is clear that our method is accurate.
64 GAUTAM PATEL AND KAUSHAL PATEL

4. Conclusion

In this paper, we proposed an efficient algorithm to solve a nonlinear

time dependent sine Gordon equation. The equation is reduced to a system
of ODEs and the system is solved by RKHSM. The main benefits of the
method are simplicity of performance, high-order accuracy, and low com-
putational cost. The efficiency and superiority of the scheme for a solution
have been shown through experimental examples. It may be extended to
solve different types of PDEs.

Acknowledgement: We wish to thank the editor-in-chief and the anony-

mous referees for their valuable suggestions to improve the quality of the
paper.

References
[1] Ablowitz, M. J., Kaup, D. J., Newell, A. C. and Segur, H., Method for solving the
sine-Gordon equation, Physical Review Letters 30(25) (1973), 1262.
[2] Aronszajn, N., Theory of reproducing kernels, Transactions of the American math-
ematical society 68(3) (1950), 337–404.
[3] Barone, A., Esposito, F., Magee, C. J. and Scott, A. C., Theory and applications of
the sine-Gordon equation, La Rivista del Nuovo Cimento (1971-1977) 1(2) (1971),
227–267.
[4] Ben-Yu, G., Pascual, P. J., Rodriguez, M. J. and Vázquez, L., Numerical solution
of the sine-Gordon equation, Applied Mathematics and Computation 18(1) (1986),
1–14.
[5] Bergman, S., The approximation of functions satisfying a linear partial differential
equation, Duke Mathematical Journal 6(3) (1940), 537–561.
[6] Bour, E., Théorie de la déformation des surfaces, Journal de l’École Impériale Poly-
technique 19 (1861), 1–48.
[7] Bratsos, A. G., A fourth order numerical scheme for the one-dimensional sine-Gordon
equation, International Journal of Computer Mathematics 85(7) (2007), 1083–1095.
[8] Cui, M. and Deng, Z., On the best operator of interpolation, Math. Numer. Sin 8
(1986), 209–216.
[9] Cui, M. and Geng, F., Solving singular two-point boundary value problem in repro-
ducing kernel space, Journal of Computational and Applied Mathematics 205(1)
(2007), 6–15.
[10] Cui, M. and Lin, Y., Nonlinear numerical analysis in reproducing kernel space, Nova
Science Pub, 2009.
 NEW METHOD OF LINES SOLUTION FOR THE NONLINEAR SINE GORDON. . . 65

[11] Dehghan, M. and Shokri, A., A numerical method for one-dimensional nonlinear
sine-Gordon equation using collocation and radial basis functions, Numerical Meth-
ods for Partial Differential Equations: An International Journal 24(2) (2008), 687–
698.
[12] Frenkel, J. and Kontorova, T., On the theory of plastic deformation and twinning,
Izv. Akad. Nauk, Ser. Fiz. 1 (1939), 137–149.
[13] Geng, F. and Cui, M., A reproducing kernel method for solving nonlocal fractional
boundary value problems, Applied Mathematics Letters 25(5) (2012), 818–823.
[14] Hamdi, S., Schiesser, W. E. and Griffiths, G. W., Method of lines, Scholarpedia 2(7)
(2007), 2859.
[15] Koto, T., Method of lines approximations of delay differential equations, Computers
& Mathematics with Applications 48(1-2) (2004), 45–59.
[16] Li, X. and Wu, B., New algorithm for nonclassical parabolic problems based on the
reproducing kernel method, Mathematical Sciences 7(1) (2003), 1–5.
[17] Lin, Y., Cui, M. and Yang, L., Representation of the exact solution for a kind of
nonlinear partial differential equation, Applied Mathematics Letters 19(8) (2006),
808–813.
[18] Mercer, J., Functions of positive and negative type and their connection with the
theory of integral equations, Philosophical Transsaction of the Royal Society of Lon-
don Ser 209 (1909), 415–446.
[19] Mittal, R. C. and Bhatia, R., Numerical solution of nonlinear sine-Gordon equa-
tion by modified cubic B-spline collocation method, International Journal of Partial
Differential Equations 2014 (2014).
[20] Mokhtari, R., Toutian, F. and Mohammadi, M., Reproducing kernel method for solv-
ing nonlinear differential-difference equations, Abstract and Applied Analysis 2012
(2012).
[21] Patel, G. and Patel, K., The method of lines for solution of one dimensional heat
equation, Proceeding of the International Conference on Emerging Trends in Scien-
tific Research, 1 (2015), 200-208.
[22] Patel, G. and Patel, K., The method of lines for solution of the one dimensional
second order wave equation, Mathematical Sciences International Research Journal,
4(2) (2015), 98-104.
[23] Patel, G. and Patel, K., The method of lines for solution of the two dimensional
elliptic equation, Annals of Faculty Engineeering Hunedoara International Journal
of Enginnering, XIV(1) (2016), 225–230.
[24] Patel, G. and Patel, K., Reproducing kernel for Robin boundary conditions, The
Mathematics Student 90(3–4) (2021), 143–158.
[25] Perring, J. K. and Skyrme, T., A model unified field equation, Selected Papers, With
Commentary, Of Tony Hilton Royle Skyrme (1994), 216–221.
[26] Saucez, P. and Schiesser, W., Adaptive method of lines, CRC Press, 2001.
[27] Uddin, M., Haq, S. and Qasim, G., A meshfree approach for the numerical solu-
tion of nonlinear sine-Gordon equation, International Mathematical Forum 7(21-
24) (2012), 1179–1186.
66 GAUTAM PATEL AND KAUSHAL PATEL

[28] Zaremba, S., Sur Le Calcul Numeric Des Fonctions Demandees Dans le Problem de
Dirichlet et le Probleme hydrodynamique, Bulletin International de L’Academie des
Sciences de Cracovie (1908), 125–195.

Gautam Patel
Department of Mathematics
Veer Narmad South Gujarat University,
Surat, Gujarat, India.
E-mail: [email protected]

Kaushal Patel
Department of Mathematics
Veer Narmad South Gujarat University,
Surat, Gujarat, India.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93 , Nos. 1-2, January-June (2024), 67–70

A CHARACTERIZATION OF CHAOTIC GROUP

ACTIONS

PADMAPRIYA V P AND ALI AKBAR K

(Received : 12 - 08 - 2022 ; Revised : 03 - 04 - 2023)

Abstract. In this paper, we prove that a group acts chaotically on

an infinite Hausdorff space X if and only if any finite collection of
nonempty open subsets of X shares infinitely many periodic orbits.

1. Introduction

Chaotic dynamical systems have been the topic of intensive research.

Most of the work (see [2]) in chaotic dynamics has been concerned with the
iteration of self-maps. In [1], the authors developed a theory in its most
general setting as an action of a group on some topological space. Though
the groups acting chaotically on Hausdorff spaces have been characterized
earlier in [1], no further characterizations were done related to the space
on which the group action takes place. The objective of this paper is to
provide further characterizations of chaotic group actions in terms of the
underlying space and the periodic orbits shared by open subsets of the
underlying space.
In this paper, we prove that the following are equivalent for an action of a
group G on an infinte Hausdorff space X.
(1) action is chaotic.
(2) any two nonempty open subsets of X share a periodic orbit.
(3) any finite collection of nonempty open subsets of X share infinitely
many periodic orbits.
We prove these results with the help of basic ideas involved in the areas
of group theory and topology. These characterizations will be helpful for
further studies of the dynamics of continuous group actions.
2020 Mathematics Subject Classification: 37B05, 37B02
Key words and phrases: group actions, periodic orbits, topological transitivity,
periodic point, chaotic

© Indian Mathematical Society, 2024 .

67
68 PADMAPRIYA V P AND ALI AKBAR K

Definition 1.1. Let G be a group acting on a Hausdorff topological space

X. The orbit of a point x ∈ X under G is defined as Gx := {gx : g ∈ G}.

Definition 1.2. Let G be a group acting on a Hausdorff topological space

X. The action of G on X is continuous if for each g ∈ G, the map
g : X → X is continuous.

Throughout this paper, we denote G for a group and X for a Hausdorff

topological space. The group action here is always considered as continuous.

Definition 1.3. The action of G on X is called topologically transitive

if for any two nonempty open sets U and V of X, there exists an element
g ∈ G such that gU ∩ V 6= φ.

Definition 1.4. A point x ∈ X is called a periodic point of the action

if the orbit of x under G is finite. The number of elements of the orbit is
called the period of the orbit.

Definition 1.5. An action is called chaotic if it is topologically transitive

and the set of periodic points forms a dense subset of X.

In [3], a collection of nonempty open sets shares a periodic orbit for a

self-map is defined. Analogous to this we define the following for a group
action.

Definition 1.6. A finite collection of nonempty open sets V1 , V2 , ..., Vn

share a periodic orbit if there exists a periodic point whose orbit meets
each Vi for i = 1, 2, ..., n.

2. Main Results

The following proposition can be considered as another definition of

chaotic group actions.

Proposition 2.1. The action of G on X is chaotic if and only if any two

nonempty open subsets of X share a periodic orbit.

Proof. If any two nonempty open subsets share a periodic orbit, then every
open set contains a periodic point and therefore periodic points are dense in
X. To prove the topological transitivity, let U and V be any two nonempty
open subsets of X. By assumption, U and V share a periodic orbit. That
is, there exists x ∈ X such that the orbit Gx = {gx : g ∈ G} intersects
 A CHARACTERIZATION OF CHAOTIC GROUP ACTIONS 69

U and V . Therefore there exists g1 , g2 ∈ G such that g1 x ∈ U ∩ Gx and

g2 x ∈ V ∩ Gx. Let g = g2 g1 −1 . This implies gU ∩ V 6= ∅ and hence the
action is topologically transitive. Conversely, if the action is chaotic, for
any two nonempty open sets U, V there exists x ∈ U and g ∈ G such that
gx ∈ V . Then W := g −1 V ∩ U is nonempty, open, and gW ⊂ V . This W
must contain a periodic point p since the periodic points are dense in X.
Observe that p ∈ U and gp ∈ V . Hence U and V share a periodic orbit. 
The following lemma proves the existence of a periodic orbit of a chaotic
action of a group G on a Hausdorff topological space X shared by finitely
many non-empty open sets.
Lemma 2.2. The action of G on X is chaotic if and only if any finite
collection of nonempty open sets shares a periodic orbit.
Proof. The proof follows by Proposition 2.1 and by applying induction on
the number of nonempty open sets. 
Now we are ready to consider our main theorem.
Theorem 2.3 (Main Theorem). The action of G on an infinite X is chaotic
if and only if every finite collection of nonempty open subsets of X shares
an infinite number of periodic orbits.
Proof. Let U1 , U2 , ..., Un be nonempty open subsets of X. Suppose this
collection of open sets share only finitely many periodic orbits. Let P be
the set of all points in the periodic orbits shared by U1 , U2 , ..., Un . Then
P is finite since it is the finite union of periodic orbits that are finite. We
define Vi = Ui \P for i = 1, ..., n. Then each Vi is a nonempty open subset of
X since X is Hausdorff, P is finite and the action is chaotic. Observe that
Vi ⊂ Ui for all i. Now V1 , V2 , ..., Vn forms a finite collection of nonempty
open subsets of X. Therefore by the assumption, they share a periodic
orbit under the action of G. But this orbit is not contained in P and the
orbit intersects the finite collection {U1 , U2 , ....., Un }, a contradiction. The
converse follows from Lemma 2.2. 
We conclude our short article by providing the following remark.
Remark 2.4. Let G be a group acting chaotically on an infinite Hausdorff
topological space X. Because of Proposition 2.1 and Theorem 2.3, we can
conclude that the following are equivalent for any two non-empty subsets
U and V of X.
70 PADMAPRIYA V P AND ALI AKBAR K

(1) U and V of X share a periodic orbit.

(2) U and V of X share finitely many periodic orbits.
(3) U and V share an infinite number of periodic orbits.

Acknowledgement: The authors are grateful to the referee for valuable

suggestions. The first author acknowledges CSIR, India and the second
author acknowledges SERB-MATRICS Grant No. MTR/2018/000256 for
financial support.

References

[1] Cairns, G., Davis, G., Kolganova, A. and Perversi, P., Chaotic group actions,
L’Enseign. Math., 41(1995)123–133.
[2] Devaney, R., An introduction to chaotic dynamical systems, Addison-Wesley, Cali-
fornia, 1989.
[3] Touhey, P., Yet another definition of chaos, American Mathematical Monthly,
104(1997)411–414.

Padmapriya v P
Research Scholar
Department of Mathematics
Central University of Kerala,
E-mail: [email protected]

Ali Akbar K
Associate Professor
Department of Mathematics
Central University of Kerala.
E-mail: [email protected], [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93 (1-2), January-June (2024), 71–75

π2
A PROOF OF ζ(2) = 6
RELATED TO FRACTIONAL
CALCULUS

JOHN M. CAMPBELL
(Received : 24 - 10 - 2022 ; Revised : 23 - 01 - 2023)

Abstract. We offer a new and appealingly simple proof related to

2
fractional calculus of the famous equality ζ(2) = π6 .

1. Introduction

Euler’s closed-form formula

1 1 π2
+ + · · · = (1.1)
12 22 6
is one of the most celebrated and most beautiful formulas in all of mathe-
matics. It appears that this Euler formula is the one classical formula with
the most published research articles specifically devoted to new and different
proofs of it. For example, the lengthy list of References in our recent con-
tribution [2], in which Euler’s formula was proved using the Wilf–Zeilberger
method [8], lists many articles based on proofs of Euler’s formula, including
the recent Mathematics Student article [6] that has inspired our current ar-
ticle. Our new proof of Euler’s formula relies on a very recently published
identity [1] that was derived using an area in mathematics that is known as
fractional calculus and that is described below, and our new proof also relies
on a special function known as the dilogarithm, which is reviewed below.

1.1. Fractional calculus. The repeated application of differential or of

differential-type operators is used in virtually every form of mathematical
analysis. The use of the phrase “repeated application” suggests that one’s
applying such operators over and over again is of a discrete nature, in terms
of an integer number of times such operators may be applied. Fractional

2010 Mathematics Subject Classification: 11Y60, 33B30

Key words and phrases: Basel problem, Riemann zeta function, fractional calculus

© Indian Mathematical Society, 2024 .

71
72 JOHN M. CAMPBELL

calculus may be thought of as being given by how one may generalize this
notion using non-integer values.
d
If we let D denote a linear, differential operator such as the operator dx
d n
or Dn


(defined on some suitable domain), then expressions such as dx
would typically refer to self-iterations of the form

| ◦ D {z
D ◦ · · · ◦ D} .
n

If we consider how such self-iterations may be generalized and applied using

analytic functions, this leads us to consider the relation

◦ · · · ◦ D}(xn ) = n!
|D ◦ D {z (1.2)
n
d
for D = dx and how it is typically generalized in the field of fractional
calculus.
To generalize (1.2) for non-integer values n, one is led toward the use
of the Γ-function (see Rainville [9, §8]) given by the Euler integral Γ(x) =
R ∞ x−1 −u
0 u e du. For example, for α > − 12 , we let the Caputo operators D1/2
and D−1/2 be such that
Γ(α + 1) α− 1
D1/2 xα =
x 2
Γ α + 12
and
Γ(α + 1) α+ 1
D−1/2 xα =
x 2 , (1.3)
Γ α + 32
following the references [1, 3] that are to be later involved in our work. In
view of (1.2), we are led to find that
 
D1/2 ◦ D1/2 xα = αxα−1

and similarly with respect to (1.3), and hence a standard way of generalizing
identities such as (1.2) using non-integer values n. The recent article [1]
concerning Caputo operators introduced an identity via such operators that
we briefly review below and that we are to prove in a self-contained way to
2
provide a new proof of 112 + 212 + · · · = π6 .

1.2. The dilogarithm function. The study of higher logarithm functions

constitutes a large area in special functions theory, with reference to the
33B30 Mathematics Subject Classification. Something of a prototype or
 π2
A PROOF OF ζ(2) = 6
73

main instance of a so-called higher logarithm function is the famous dilog-

arithm function, which we are to review below. For excellent and relevant
exposition on the dilogarithm and the history concerning this special func-
tion, we refer to Stewart’s recent article [10], and the following background
material is partly inspired by [10].
xn
We write Li2 (x) := ∞
P
n=1 n2 to denote the dilogarithm function, letting
this function be defined for |x| ≤ 1. As in [10], we provide [5] as a reference
for the history of the use of the function Li2 in mathematics. For a succinct
review of many of the fundamental properties concerning Li2 , such as its
usual integral formula
Z x
log(1 − t)
Li2 (x) = − dt
0 t
and the Euler reflection formula
π2
Li2 (x) + Li2 (1 − x) = − log(x) log(1 − x),
6
we again refer to Stewart’s article [10]. For our present purposes, we are
more interested in an integral identity involving Li2 given in [1] that is to
be proved and applied in a self-contained way.
R1
Let the inner product h·, ·i be such that hf (x), g(x)i = 0 f (x)g(x) dx
for (Riemann) integrable functions f and g. The method of semi-integration
by parts (SIBP) was introduced in [3] and formulated in the following way:
D    E
hf, gi = D1/2 τ f, τ D−1/2 g ,

provided that both sides of this equality are well-defined, and where τ de-
notes the operator such that h(x) 7→ h(1 − x). We let f and g respectively
denote the analytic functions given by the power series expansions with
2
n 2n
2 1 n 2n 2
summands given by an = α16



n and bn = 16 n ; we may then
apply SIBP, as in [1], to prove that
√
sin−1 ( x)
Z 1
Li2 (α) − Li2 (−α)
√ p dx = . (1.4)
0 1 − x α2 (x − 1) + 1 α
We are to prove this identity in a self-contained way, using “Feynman’s
2
favorite trick” [7], to prove that ζ(2) = π6 , writing ζ(x) = 11x + 21x + · · · to
denote the Riemann zeta function.

1.3. Feynman’s favorite trick. The interchange of limiting operations

arises in virtually all areas of classical and modern analysis. For example,
74 JOHN M. CAMPBELL

exchanging the order of integration and infinite summation is a common

kind of “trick” used in the determination of closed-form evaluation of infinite
series, and may be justified, for reasonably well-behaved integrands, subject
to the conditions of famous results in mathematical analysis such as the
Monotone Convergence Theorem; see classic texts such as [4, p. 317]. To
justify differentiating under the integral sign, one would typically appeal to
the famous Leibniz integral rule. According to this rule, for integrals with
upper and lower limits that are constant, and with an integrand f (α, x) such
∂
that f (α, x) and ∂α f (α, x) are continuous in α and x (for the closed interval
given by the upper and lower limits of the integral), one may differentiate
under the integral sign (with respect to α). Although the integrand in (2.1)
is only continuous with respect to x for [0, 1) and not for [0, 1], one may use
a standard generalization of the Leibniz integral rule for improper integrals.

π2
2. A new proof of ζ(2) = 6

Although (1.4) had initially been discovered via fractional calculus, we

are to prove (1.4) in a self-contained way, to provide a new and self-contained
2
proof that ζ(2) = π6 . The famous problem of evaluating ζ(2) in closed form
is referred to as the Basel problem, as below.
π2
Theorem 2.1. The solution to the Basel problem whereby ζ(2) = 6 holds
true.

Proof. To prove that (1.4) holds for |α| ≤ 1, we instead prove the equivalent
formulation
√
sin−1 ( x) α
Z 1
√ p dx = Li2 (α) − Li2 (−α). (2.1)
0 1 − x α2 (x − 1) + 1
By differentiating with respect to α on both sides of the purported iden-
tity in (2.1) (using the generalization of the Leibniz integral rule indicated
above), we can show that (2.1) is equivalent to
√
sin−1 ( x)
Z 1
log(α + 1) log(1 − α)
√ 3/2
dx = − . (2.2)
0 1 − x (1 + (x − 1)α2 ) α α
However, we may easily check that an antiderivative of the integrand of the
left-hand side of the above equality is
 p √  √ √
2 log α α2 (x − 1) + 1 + α x 2 1 − x sin−1 ( x)
− p ,
α α2 (x − 1) + 1
 π2
A PROOF OF ζ(2) = 6
75

so taking the required limits then gives us a proof of (2.1). From (2.1), we
R 1 sin−1 (√x)
obtain that 0 √1−x√x dx = Li2 (1) − Li2 (−1). We may easily check that
√ 2
an antiderivative for the integrand on the left-hand side is sin−1 ( x) , so
2
this gives us a proof that π4 = Li2 (1) − Li2 (−1). So, we have proved that
π2 P∞ 1 P∞ (−1)n
4 = n=1 n2 − n=1 n2 , and the use of series bisections easily gives us
2
that this is equivalent to ζ(2) = π6 . 
Acknowledgement: The anonymous reviewer of this article offered many
suggestions that significantly improved this article.

References
[1] Campbell, J. M., Applications of Caputo operators in the evaluation of Clebsch–
Gordan-type multiple elliptic integrals. Integral Transforms Spec. Funct. (2022).
[2] Campbell, J. M., A Wilf-Zeilberger-based solution to the Basel problem with appli-
cations, Discrete Math. Lett., 10 (2022), 21-27.
[3] Campbell, J. M., Cantarini, M., D’Aurizio, J., Symbolic computations via Fourier-
Legendre expansions and fractional operators, Integral Transforms Spec. Funct., 33
(2022), 157-175.
[4] Carothers, N. L., Real analysis, Cambridge University Press, Cambridge, 2000.
[5] Maximon, L. C., The dilogarithm function for complex argument, Proc. R. Soc.
Lond. A, 459 (2003), 2807-2819.
[6] Murty, M. R., A simple proof that ζ(2) = π 2 /6, Math. Student, 88 (2019), 113-115.
[7] Nahin, P. J., Feynman’s favorite trick, Inside Interesting Integrals Undergraduate
Lecture Notes in Physics. Springer, Cham. (2020), 73-116.
[8] Petkovšek, M., Wilf, H. S., Zeilberger, D., A = B, A K Peters, Ltd., Wellesley, MA,
1996.
[9] Rainville, E. D., Special functions, The Macmillan Co., New York, 1960.
[10] Stewart, S. M., Some simple proofs of Lima’s two-term dilogarithm identity, Ir.
Math. Soc. Bull., 89 (2022), 43-49.

John M. Campbell
Department of Mathematics
Toronto Metropolitan University,
Toronto, Ontario, CANADA.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 76–82

ON KAKUTANI’S CHARACTERIZATION OF THE

CLOSED LINEAR SUBLATTICES OF
C(X)—REVISITED

TEENA THOMAS
(Received : 18 - 01 - 2023 ; Revised : 27 - 05 - 2023)

Abstract. In his paper [Concrete representation of abstract (M )-

spaces (A characterization of the space of continuous functions), Annals
of Mathematics, No. (4), 42 (1941), 994–1024.], S. Kakutani gave an
interesting representation of the closed linear sublattices of the space of
real-valued continuous functions on a compact Hausdorff space, which
is determined by a set of algebraic relations. In this short note, we
present a simple proof of this representation without using any pro-
found lattice theory or functional analysis results, making this proof
accessible even to undergraduate students.

1. Preliminaries

Let X be a compact Hausdorff space. We denote the Banach space (in

other words, a complete normed linear space) of all real-valued continuous
functions on X, equipped with the supremum norm, by C(X). It is well-
known that C(X) is a lattice under the operation of pointwise maximum or
minimum of a pair of functions in C(X) and an algebra under the operation
of pointwise multiplication of a pair of functions in C(X). All the subspaces
in this note are assumed to be closed with respect to the norm topology.
We say that a closed linear subspace A of C(X) is a sublattice of C(X) if
A is a lattice in its own right under the same lattice operation as in C(X).
A closed linear subspace A of C(X) is said to be a subalgebra of C(X)
if A is an algebra in its own right under the same algebra operation as in
C(X). Kakutani presented an algebraic characterization of the closed linear
sublattices of C(X) as follows :

2010 Mathematics Subject Classification: 46E15, 46B42

Key words and phrases: space of real-valued continuous functions, compact Hausdorff
space, sublattices, subalgebras

© Indian Mathematical Society, 2024 .

76
 ON KAKUTANI’S CHARACTERIZATION 77

Theorem 1.1 ([2, Theorem 3, pg. 1005]). Let X be a compact Hausdorff

space. Let A be a closed linear subspace of C(X). Then A is a sublattice of
C(X) if and only if there exists an index set I and co-ordinates (ti , si , λi ) ∈
X × X × [0, 1], for each i ∈ I such that

A = {f ∈ C(X) : f (ti ) = λi f (si ), for each i ∈ I}. (1.1)

In Section 2, we provide an elementary proof of the result above. In

fact, we exploit the structure of the closed linear sublattices of the two-
dimensional vector space, R2 , in this proof. Moreover, in this proof, we
apply a few auxiliary results whose proofs assume the knowledge of a basic
course in real analysis and point-set topology.
Let us first understand what the sublattices look like in R2 under the
lattice operation of co-ordinate-wise maximum or minimum of a pair of vec-
tors in R2 . For simplicity, we denote the linear span of a vector (a, b) ∈ R2
as span{(a, b)}. Let A be a closed linear subspace of R2 . Basic knowledge of
linear algebra tells us that the possible dimensions of A are 0, 1 or 2. If the
dimension of A is either 0 or 2, then A = {(0, 0)} or A = R2 respectively;
obviously, these subspaces are sublattices of R2 . Assume that A is a one-
dimensional subspace of R2 . We know that if A = span{(a, b)} for some
(a, b) ∈ R2 , then A = span{(λa, λb)} for each λ ∈ R\{0}. Therefore, with-
out loss of generality, there exists a, b ∈ [−1, 1] such that A = span{(a, b)}.
It is easy to see that if −1 ≤ a, b ≤ 0 or 0 ≤ a, b ≤ 1, then A is a sublattice
of R2 . Furthermore, if −1 ≤ a < 0 < b ≤ 1 then the minimum of (a, b) and
(2a, 2b) is (2a, b) 6∈ A; hence A is not a sublattice of R2 . Using a similar ar-
gument, if −1 ≤ b < 0 < a ≤ 1 then A is not a sublattice of R2 . Therefore,
without loss of generality, we arrive at the following conclusion.

Lemma 1.2. Consider R2 as a lattice under the operation of co-ordinate-

wise maximum or minimum of a pair of vectors in R2 . Then the only
closed linear sublattices of R2 are {(0, 0)}, R2 and span{(a, b)}, for those
(a, b) ∈ R2 satisfying 0 ≤ a, b ≤ 1.

We can also list all possible closed linear subalgebras of R2 as follows :

Lemma 1.3 ([1, Lemma 4.46]). Consider R2 as an algebra under co-

ordinate-wise addition and multiplication of a pair of vectors in R2 . Then
the only subalgebras of R2 are {(0, 0)}, R2 , span{(1, 0)}, span{(0, 1)} and
span{(1, 1)}.
78 TEENA THOMAS

We now recall a few known facts and definitions needed to prove Theo-
rem 1.1. For a compact Hausdorff space X, we say a closed linear subspace
A of C(X) separates the points of X if for every two distinct points x, y ∈ X,
there exists f ∈ A such that f (x) 6= f (y).
The following result shows the interconnection between a subalgebra
and a sublattice of C(X).

Lemma 1.4 ([1, Lemma 4.48]). Let X be a compact Hausdorff space. If A

is a closed linear subalgebra of C(X), then A is a sublattice of C(X).

Furthermore, for a compact Hausdorff space X and a closed linear sub-

lattice A of C(X), a sufficient condition for a function in C(X) to be in A
is

Lemma 1.5 ([1, Lemma 4.49]). Let X be a compact Hausdorff space. Let
A be a closed linear sublattice of C(X) and f ∈ C(X). If for every x, y ∈ X
there exists gxy ∈ A such that gxy (x) = f (x) and gxy (y) = f (y), then f ∈ A.

Let T be a locally compact Hausdorff space. We denote the Banach

space of real-valued continuous functions on T vanishing at infinity by
C0 (T ). Further, let us denote T∞ as the one-point compactification of
T and t∞ be the “point at infinity”. We consider the subspace Jt∞ :=
{f ∈ C(T∞ ) : f (t∞ ) = 0} of C(T∞ ). Now, consider the restriction map
Φ : Jt∞ → C0 (T ) defined as Φ(f ) = f |T , for f ∈ Jt∞ . Then the map Φ
is an isometry and lattice, algebra and linear isomorphism (in other words,
a map which preserves the distances as well as the lattice, algebraic and
linear structures). The identification above and Theorem 1.1 help us to al-
gebraically characterize the closed linear sublattices and subalgebras of the
space C0 (T ) in Section 2.

2. Main Results

Let us now prove our main result.

Proof of Theorem 1.1. We assume that X contains at least two points be-
cause if X is a singleton set, then C(X) is simply R and the only closed
linear sublattices or subalgebras are {0} and R. If A has the description as
in (1.1), then clearly A is a sublattice of C(X).
 ON KAKUTANI’S CHARACTERIZATION 79

Now, assume that A is a sublattice of C(X). For every two distinct

points x, y ∈ X, define

Axy = {(g(x), g(y)) ∈ R2 : g ∈ A}.

Since A is a lattice, Axy is a sublattice of R2 (under the operation of co-

ordinate-wise maximum of a pair of vectors in R2 ).
Case 1 : Assume A separates the points of X.
If for every x, y ∈ X, Axy = R2 then by Lemma 1.5, A = C(X). Hence
A has the description as in (1.1). Otherwise, there exists two distinct points
x0 , y0 ∈ X such that Ax0 y0 is a proper sublattice of R2 . Consider the
following collection :

I = {(t, s, λ) ∈ X × X × [0, 1] : f (t) = λf (s), for each f ∈ A}.

We now show that I 6= ∅. Since A separates the points of X, Ax0 y0

cannot be {(0, 0)} or span{(1, 1)}. Thus Ax0 y0 = span{(a, b)} for some
0 ≤ a, b ≤ 1 and a 6= b. If a = 0 and b > 0 then for each g ∈ A, g(x0 ) = 0
and hence (x0 , y0 , 0) ∈ I. If a > 0 and b = 0 then for each g ∈ A, g(y0 ) = 0
and hence (y0 , x0 , 0) ∈ I. If without loss of generality 0 < a < b, then for
each g ∈ A, there exists rg ∈ R such that (g(x0 ), g(y0 )) = (rg a, rg b). It
follows that g(x0 ) = ab g(y0 ). Thus (x0 , y0 , ab ) ∈ I.
We index I by I itself, that is, each element of I is indexed by itself.
Further, define

A0 = {f ∈ C(X) : f (ti ) = λi f (si ), for each i ∈ I}.

By the definition of I, it is clear that A ⊆ A0 .

We next show that A0 ⊆ A. Let f ∈ A0 . In order to show f ∈ A, by
Lemma 1.5, it suffices to show that for each x, y ∈ X, there exists gxy ∈ A
such that gxy (x) = f (x) and gxy (y) = f (y). Therefore, it suffices to show
that for each x, y ∈ X, (f (x), f (y)) ∈ Axy .
Let x, y ∈ X. Since A separates the points of X, Axy cannot be {(0, 0)}
or span{(1, 1)}. We remark here that in this case, we use the assumption
that A separates the points of X only to prove that I 6= ∅ and to rule out
the above two possibilities of Axy .
If Axy = R2 , then clearly (f (x), f (y)) ∈ R2 = Axy . If Axy = span{(0, 1)}
then (0, f (y)) ∈ Axy and for each g ∈ A, g(x) = 0. Thus (x, y, 0) ∈ I. Since
f ∈ A0 , f (x) = 0. Hence (f (x), f (y)) ∈ Axy . Similar arguments hold true
if Axy = span{(1, 0)}.
80 TEENA THOMAS

Without loss of generality, let 0 < a < b < 1. Consider Axy =

span{(a, b)}. Then for each g ∈ A, g(x) = ab g(y). Thus (x, y, ab ) ∈ I.
Since f ∈ A0 , let f (x)
a = f (y)
b = r (say). It follows that (f (x), f (y)) ∈
span{(a, b)} = Axy .
Case 2 : Assume that A does not separate the points of X. Thus
there exists two distinct points x0 , y0 ∈ X such that f (x0 ) = f (y0 ), for
each f ∈ A. Consider the following collection :

I = {(t, s, λ) ∈ X × X × [0, 1] : f (t) = λf (s), for each f ∈ A}.

Since (x0 , y0 , 1), (y0 , x0 , 1) ∈ I, clearly I 6= ∅. We index I by I itself, that

is, each element of I is indexed by itself. Let us define

A0 = {f ∈ C(X) : f (ti ) = λi f (si ), for each i ∈ I}.

Clearly A ⊆ A0 . In order to show A0 ⊆ A, by Lemma 1.5, it suffices to

show that for each f ∈ A0 and each x, y ∈ X, (f (x), f (y)) ∈ Axy .
Let x, y ∈ X. We first consider the following possibilities of Axy : R2 ,
span{(0, 1)}, span{(1, 0)} and span{(a, b)} for those a, b ∈ R satisfying 0 <
a < b < 1 (without loss of generality). For each of the above possibilities, we
apply arguments similar to that used in Case 1 to show that (f (x), f (y)) ∈
Axy .
In this case, due to our assumption that A does not separate the points
of X, we need to consider the two possibilities which we rule out in Case 1.
They are as follows : If Axy = {(0, 0)} then for each g ∈ A, g(x) = 0 = g(y).
Hence, (x, y, 0), (y, x, 0) ∈ I. Since f ∈ A0 , (f (x), f (y)) = (0, 0) ∈ Axy . If
Axy = span{(1, 1)} then for each g ∈ A, g(x) = g(y). Thus (x, y, 1) ∈ I.
Since f ∈ A0 , (f (x), f (y)) = (f (x), f (x)) ∈ Axy . 

The closed linear subalgebras of the C(X) space has a representation

similar to that in (1.1) with the value of the coefficients λi being either 0 or
1. With the help of Lemmas 1.3, 1.4 and 1.5, the following result is proved
using a similar argument as in Theorem 1.1, and hence we omit it.

Theorem 2.1. Let X be a compact Hausdorff space. Let A be a closed

linear subspace of C(X). Then A is a subalgebra of C(X) if and only if
there exists an index set I and co-ordinates (ti , si , λi ) ∈ X × X × {0, 1}, for
each i ∈ I such that

A = {f ∈ C(X) : f (ti ) = λi f (si ), for each i ∈ I}. (2.1)

 ON KAKUTANI’S CHARACTERIZATION 81

The following characterizations of the sublattices and subalgebras of

C0 (T ) for a locally compact Hausdorff space T follows directly from The-
orems 1.1 and 2.1 and the fact that C0 (T ) is isometrically lattice, algebra
and linear isomorphic to the subalgebra Jt∞ of C(T∞ ).

Corollary 2.2. Let T be a locally compact Hausdorff space. Let A be a

closed linear subspace of C0 (T ). Then
(1) A is a sublattice of C0 (T ) if and only if there exists an index set I
and co-ordinates (ti , si , λi ) ∈ T × T × [0, 1], for each i ∈ I such that

A = {f ∈ C0 (T ) : f (ti ) = λi f (si ), for each i ∈ I}. (2.2)

(2) A is a subalgebra of C0 (T ) if and only if there exists an index set

I and co-ordinates (ti , si , λi ) ∈ T × T × {0, 1}, for each i ∈ I such
that

A = {f ∈ C0 (T ) : f (ti ) = λi f (si ), for each i ∈ I}. (2.3)

Remark 2.3. Kakutani provided a precise identification of a much general

class of Banach spaces, namely abstract (M )-spaces (see [2, pg. 994] for the
definition), with a subspace of C(X) for some compact Hausdorff space X.
Let X be a compact Hausdorff space and A be a closed subspace of C(X).
If A is an abstract (M )-space, then by [2, Theorem 1, pg. 998], there exists
a compact Hausdorff space Ω such that A is isometrically lattice and linear
isomorphic to the subspace, {f ∈ C(Ω) : f (ti ) = λi f (si ), for each i ∈ I}, of
C(Ω) for some index set I and co-ordinates (ti , si , λi ) ∈ Ω × Ω × [0, 1] for
each i ∈ I. However, it is not necessary that A, being a subspace of C(X),
has a description in C(X) as given in (1.1). For example, consider A to
be the space of real-valued affine continuous functions on [0, 1]. Now, A is
a closed subspace of C([0, 1]) but not a sublattice of C([0, 1]). Therefore,
by Theorem 1.1, A does not have a description as in (1.1), for any given
subfamily of co-ordinates in [0, 1] × [0, 1] × [0, 1]. Nevertheless, it is easy to
see that A is isometrically lattice and linear isomorphic to C({0, 1}) and
hence is an abstract (M )-space.

Acknowledgement: The author would like to thank Prof. Dirk Werner

for indicating Remark 2.3. The author is also grateful to the editor and
the referee for their helpful suggestions to improve the quality of this man-
uscript.
82 TEENA THOMAS

References
[1] Folland, G. B., Real analysis, Pure and Applied Mathematics (New York), Second
Edition, John Wiley & Sons, Inc., New York, pg. xvi+386, 1999.
[2] Kakutani, S., Concrete representation of abstract (M )-spaces (A characterization
of the space of continuous functions), Annals of Mathematics, No. (4), 42 (1941),
994–1024.

Teena Thomas
Department of Mathematics
Indian Institute of Technology Hyderabad
Sangareddy, Telangana, India – 502284.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January-June (2024), 83–100

A NOTE ON FLEXIBLE IMMERSIONS

ABHIJEET GHANWAT AND SUHAS PANDIT

Abstract. In this note, we introduce the notion of a flexible immer-

sion of a compact surface in a 4–manifold and give its applications. As
a first application, we show that every open book of every closed 3–
manifold open book immerses in the 5–sphere S 5 with the trivial open
book. As a second application, we show that given a second integral
homology class d of CP 2 and given a closed oriented surface Σ, there
is a flexible immersion f : Σ ↬ CP 2 such that [f ] = d. We also show
that each of the classes 0 and ±2 ∈ Z = H2 (CP 2 , Z) can be represented
by a flexible embedding f : Σ ,→ CP 2 of any closed oriented surface Σ
into CP 2 .

1. Introduction

Hirose and Yasuhara [8] discussed the notion of flexible embeddings of

closed surfaces in compact 4–manifolds. A proper embedding f : Σ ,→ X of
a compact surface Σ in a compact 4–manifold X is said to be flexible if for
given a diffeomorphism ϕ : Σ → Σ, there exists an isotopy ψt : X → X of
X such that ψ1 maps f (Σ) to itself and f −1 ◦ ψ1 ◦ f = ϕ. We consider the
diffeomorphisms and isotopies of a 4–manifold which are the identity near
the boundary.
Proper flexible embeddings of compact surfaces in compact 4–manifolds
are used in [4], [13] and [6] to study open book embeddings of closed 3–
manifolds in S 2 × S 3 as well as in S 2 ×S
e 3 . Proper flexible embeddings of
compact surfaces in compact 4–manifolds are also used in [5] and [14] to
study Lefschetz fibration embeddings of 4–manifolds in 6–manifolds.
An open book of a closed manifold M is a way to express M as a
locally trivial fiber bundle π : M \ B → S 1 in the complement of a co-
dimension–2 submanifold B having trivial normal bundle such that the
closure of the boundary of each fiber is B. Alexander [1] proved that every

2010 Mathematics Subject Classification: 57R45, 57R15, 57R65

Key words and phrases: 3–manifolds, open book, immersion

© Indian Mathematical Society, 2024 .

83
84 ABHIJEET GHANWAT AND SUHAS PANDIT

closed orientable 3–manifold admits an open book. The existence of an

open book for a closed non-orientable 3–manifold is proved in [2] as well
as in [7]. For various proofs of the existence of an open book for closed
3–manifolds, refer [3].
Roughly speaking, we say that a smooth manifold M with a given open
book admits an open book embedding in an open book of a smooth manifold
N , provided there is an embedding of M in N such that –as a submanifold
of N – the given open book decomposition on M is compatible with the open
book of N . It is not known that whether every closed 3–manifold admits
an open book which open book embeds in an open book of S 5 .
In this exposition, we define the notion of a flexible immersion of a
compact surface as well as the notion of an open book immersion, see Defi-
nition 3.1 and Definition 4.2. We show that given a compact surface Σ, there
is a proper flexible immersion f : Σ ,→ D4 of Σ into D4 , see Lemma 3.7
and Lemma 3.8. As a first application of this, we show that every open of
every closed 3–manifold open book immerses in the 5–sphere S 5 with the
trivial open book, see Theorem 4.4.
As a second application, we show that given a second integral homology
class d of CP 2 and given a closed oriented surface Σ, there is a flexible
immersion f : Σ ↬ CP 2 such that [f ] = d, see Theorem 5.3. We also show
that each of the classes 0 and ±2 ∈ Z = H2 (CP 2 , Z) can be represented by
a flexible embedding f : Σ ,→ CP 2 of any closed oriented surface into CP 2 ,
see Theorem 5.5.

2. Preliminaries

In this section, we recall the necessary definitions and the results re-
quired for this article. Unless otherwise specified, we assume all the diffeo-
morphisms and the isotopies of a compact manifold are the identity near
the boundary. Let us begin with the definition of the mapping class group.

Definition 2.1. The mapping class group MCG(Σ, ∂Σ) of a compact ori-
ented surface Σ is a group of orientation preserving diffeomorphisms of Σ
upto isotopy. In case, the surface Σ is non-orientable, the mapping class
group is a group of diffeomorphisms of Σ up to isotopy.

Let us recall a few well-known facts about the generators of the mapping
class group of a compact surface which we require for this article. Licko-
rish [10], [12] showed that the mapping class group of a compact oriented
 A NOTE ON FLEXIBLE IMMERSIONS 85

σi,j
σ1 σi σj σk

α1 α2 α3 αg
β1 β2 βg−1

γ1 γ2 γ3 γg

Figure 1. The Dehn twists along the curves shown in the

above figure generates the mapping class group
MCG(Σ, ∂Σ).

surface Σ is generated by the Dehn twists along the simple closed curves on
Σ and the mapping class group of a compact non-orientable surface N is
generated by the Dehn twists along the two-sided simple closed curves on
N and the y–homeomorphisms along embedded Y –spaces in N . A Dehn
twist along a two-sided simple closed curve c on a surface S, we mean a
diffeomorphism dc : S → S supported in a regular neighborhood of c in
S obtained by cutting S along c, twisting 2π to right or left and regluing.
Recall that a Y –space can be obtained from a Möbius band µ by removing
one disc D from the interior and attaching a Möbius band M along ∂D. A
y–homeomorphism of a Y –space is a homeomorphism y : Y → Y which is
obtained by sliding the Möbius band M on Y once along the core of the
Möbius band µ.
In fact, Lickorish showed that the mapping class group MCG(Σg,k )
of a compact oriented surface of genus g with k boundary components is
generated by the set of all the Dehn twists along the simple closed curves
αi ’s, βi ’s, γi ’s, σi ’s and σi,j ’s as shown in Figure 1. We call these curves
Lickorish generating curves.
Now, we describe a finite generating set for the mapping class group of
a compact non-orientable surface given by Omori and Kobayashi [11]. To
do this, we first describe a compact non-orientable surface Ng,k of genus g
with k–boundary components as follows. Consider the surface S obtained
by removing g + k disjoint open discs D1 , D2 , . . . , Dg+k from the 2–sphere.
Then, the non-orientable surface Ng,k is obtained from S by attaching g
Möbius bands µ1 , µ2 , . . . , µg along ∂D1 , ∂D2 , . . . ∂Dg , respectively as shown
in Figure 2. The cross marks in Figure 2 depicts the g Möbius bands
µ1 , µ2 , . . . , µg . Kobayashi and Omori [11] proved that the mapping class
86 ABHIJEET GHANWAT AND SUHAS PANDIT

β1 β2 βk

µ1 µ2 µ4 µ4 µg

α1 α2 α3 α4 αg−1

αij
ρij

i i+1 j i j

σij ′
σij

i j 1 i j

µ1 µ2

α1
1 2 3 4

Y -Space
Figure 2. Kobayashi and Omori generators for the map-
ping class group of a compact non-orientable surface with
boundary.

group MCG(Ng,k , ∂Ng,k ) is generated by the Dehn twists along the simple
′ , γ as shown in Figure 2 and the y–
closed curves αi , βi , αi,j , ρi,j , σi,j , σi,j
homeomorphism obtained by sliding the Möbius band along α1 once on
the Y –space as shown in Figure 2. We call these curves Kobayashi-Omori
generating curves.
 A NOTE ON FLEXIBLE IMMERSIONS 87

3. Flexible Immersions in D4

In this section, we define the notion of a flexible immersion of a compact

surface in a compact 4–manifold and show that the 4–ball D4 admits a
proper flexible immersion of every compact surface. Let us begin by the
following definition.

Definition 3.1. Let S be a compact surface and X be a compact 4–

manifold. A proper immersion f : (S, ∂S) ↬ (X, ∂X) is said to be a
flexible immersion if given any diffeomorphism ϕ : S → S, there exists a
family ft : S ↬ X of proper immersions such that f0 = f , f1 = f ◦ ϕ and
all the immersions ft agree on a collar neighborhood of ∂Σ. In this case, we
say that the diffeomorphism ϕ is flexible with respect to the immersion f .

Remark 3.2. In the above definition, if each ft is an embedding, then

the embedding f is called a flexible embedding in the sense of Hirose and
Yasuhara [8].

Definition 3.3. Let f : S ↬ X be a proper immersion of a compact

surface S into a compact 4–manifold X. We say that a two-sided simple
closed curve c on S is in standard position with respect to the immersion f
if
(1) the map f restricted to an annular neighborhood A of c in S is an
embedding.
(2) the curve c′ = f (c) bounds a disc D in X which intersects f (S) only
in c′ ,
(3) a tubular neighborhood N (D) of D in X intersects f (S) in a regular
neighborhood ν(c′ ) = f (A) of c′ in f (S),
(4) there is a coordinate chart ϕ : C2 → N (D) such that ϕ(g −1 (1)) =
ν(c′ ) and ∂(ϕ ◦ g −1 (z)) = ∂(ϕ ◦ g −1 (w)) for all z, w ∈ C, where the
map g : C2 → C is given by g(z1 , z2 ) = z1 z2 .

Remark 3.4. The standard position of a curve c on a compact surface S

with respect to an embedding f of S into a compact 4–manifold is defined
in [5].

Now, we have the following Lemma:

Lemma 3.5. Let f be a proper immersion of a compact surface S into a

compact 4–manifold X. Let c be a two-sided simple closed curve in S which
88 ABHIJEET GHANWAT AND SUHAS PANDIT

is in standard position with respect to the immersion f . Then, the Dehn

twist dc along c is flexible with respect to the immersion f .

Proof. We show there is a smooth family ft of immersions of S in X such

that f0 = f and f1 = f ◦ dc .
As c is in standard position, it bounds a disc D such that the disc D
and a tubular neighborhood N (D) of D satisfies all the properties in Defi-
nition 3.3. Let A be an annular neighborhood of c in S as in Definition 3.3.
Note that the regular fiber of the Lefschetz fibration π = g◦ϕ−1 : N (D) → C
is an annulus with ∂π −1 (y) = h1 ∪ h2 , for all y ∈ C in X, where g and ϕ are
as in Definition 3.3. The monodromy of the fibration π along any simple
closed curve γ : [0, 1] → C which goes around the origin O of C anticlockwise
is a Dehn twist along the central curve of the annulus π −1 (γ(0)). Therefore,
there exists a smooth map ψ : [0, 2π] × N (D) → N (D) such that
• for each θ, we have a diffeomorphism ψθ : N (D) → N (D) given by
ψθ (x) = ψ(θ, x).
• ψθ (π −1 (1)) = π −1 (eiθ ).
• ψθ is the identity on h1 ∪ h2 .
• ψ2π |π−1 (1) : π −1 (1) → π −1 (1) is a Dehn twist along the central curve
of the annulus π −1 (1).
Now, we consider the family ft : S ↬ D4 of proper immersions of S
into D4 as follows.

f (x) if x ∈ S \ A
ft (x) =
ψ2πt ◦ f (x) if x ∈ A

Note that f0 = f and f1 = ψ2π ◦ f = f ◦ dc . Thus, ft is the required family

of immersions.
□

Remark 3.6. In the definition of the family ft described in the above

−1
lemma, if we use ψ2πt instead of ψ2πt , then we get flexibility of the Dehn
−1
twist dc with respect to f .

Now, we prove the following important lemma of this article.

Lemma 3.7. Let Σ be a compact orientable surface. Then, there exists a

proper flexible immersion f : Σ ↬ D4 of Σ into D4 .
 A NOTE ON FLEXIBLE IMMERSIONS 89

S3 × {5}

D′ l′ × {4}

D l × {3}
S3 × [1, 5]

l1 ln l l′ S3 × {2}

f (β)
f (β)#CH

S3 × {1}

Figure 3. Proper flexible immersion f of Σg,n into D4

Proof. We regard S 3 × [1, 5] as a collar of ∂D4 in D4 with ∂D4 = S 3 × {5}.

Let Σg be the boundary of a standardly embedded genus g handlebody in
S 3 ×{2}. Now, we remove the interiors of n+1 disjoint discs D1 , D2 , . . . Dn+1
from Σg and attach one full twisted band along ∂Dn+1 . This produces
an embedded genus g surface Σg,n+2 with n + 2 boundary components
l1 , l2 , . . . ln , l, l′ in S 3 × {2} as shown in Figure 3. Finally, the embedded
surface Σg,n+1 together with n cylinders l1 × [2, 5], . . . ln × [2, 5] and the
properly embedded discs D and D′ in S 3 × [2, 3) and S 3 × [2, 4), respec-
tively with ∂D = l and ∂D′ = l′ produces a proper immersion of the genus g
surface Σg,n with n boundary components in D4 . We denote this immersion
by the map f : Σg,n ,→ D4 . After a small perturbation, we can make f
smooth.
Now, we show that the map f is a flexible immersion. Recall that the
mapping class group MCG(Σg,n ) is generated by the Dehn twists along
the Lickorish generating curves αi , βi , γi , σi and σi,j as shown in Figure 1.
Therefore, it is enough to show that Dehn twists along the Lickorish gener-
ating curves are flexible with respect to f . Let γ be a Lickorish generating
curve. Since the simple closed curve l in f (Σg,n ) bounds the disc D in
f (Σg,n ), the curve γ can be isotoped in Σg,n to a simple closed curve β
such that f (β) is the band connected sum f (γ)#CH of f (γ) and the curve
CH ⊂ f (Σg,n ) ∩ S 3 × {2}, where CH is a simple closed curve parallel to l
and passes once through the full twisted band as shown in Figure 3. No-
tice that a regular neighborhood of f (β) in f (Σg,n ) is a Hopf annulus H in
S 3 × {2}. Note that f (β) is an unknot in S 3 × {2}. Hence, it bounds a
90 ABHIJEET GHANWAT AND SUHAS PANDIT

disc D in S 3 × (1, 2]. Since, the regular neighborhood of f (β) in f (Σg,n )

is a Hopf annulus H in S 3 × {2}, the disc D and a tubular neighborhood
N (D) of D in D4 satisfy all the properties required for the curve β to be
in standard position with respect to the immersion f . Let ϕt be an isotopy
of Σg,n which sends γ to β. Now, it is clear that the curve γ is in standard
position with respect to the immersion f ◦ ϕ1 .
Therefore, by Lemma 3.5, there is a family ht : Σg,n ↬ D4 of Σg,n into
D4 of immersions with h0 = f ◦ ϕ1 and h1 = f ◦ ϕ1 ◦ dγ .
Now, we consider the family ft : Σg,n ↬ D4 of Σg,k into D4 as follows.

f ◦ ϕt

 0≤t≤1
ft = ht−1 1≤t≤2


f ◦ ϕ3−t ◦ dγ 2 ≤ t ≤ 3


Observe that f0 = f and f3 = f ◦ dγ . This completes the proof of the

Lemma.
□

Now, we discuss the similar Lemma for compact non-orientable surfaces.

Lemma 3.8. Let N be a compact non-orientable surface. Then, there exists

a proper flexible immersion f : N ↬ D4 of N into D4 .

Proof. We regard S 3 × [1, 5] as a collar of ∂D4 in D4 with ∂D4 = S 3 × {5}.

Let S be the boundary of a 3–ball B 3 in S 3 × {2}. Now, we remove the
interiors of g + k + 1 disjoint discs D1 , D2 , . . . Dg+k+1 from S and attach
g half twisted bands along ∂D1 , ∂D2 , . . . , ∂Dg and one full twisted band
along ∂Dg+k+1 . This produces an embedded genus surface S ′ with g + k + 2
boundary components b1 , b2 , . . . , bg , l1 , l2 , . . . lk , l, l′ in S 3 × {3} as shown in
Figure 3. Finally, the embedded surface S ′ together with n cylinders l1 ×
[2, 5], . . . lk × [2, 5], the properly embedded discs ∆2 , . . . ∆g , D′ in S 3 × [2, 4)
with ∂∆i = bi , ∂D′ = l′ and the properly embedded discs ∆1 , D′ in S 3 ×[2, 3)
with ∂∆1 = b1 , ∂D = l produces a proper immersion of the non-orientable
genus g surface Ng,k with k boundary components in D4 . We denote this
proper immersion by the map f : Ng,k :,→ D4 . After a small perturbation,
we can make f smooth.
Now, we show that the map f is a flexible immersion. Recall that the
mapping class group MCG(Ng,k ) is generated by the Dehn twists along the
Kobayashi-Omori generating curves αi , βi , αi,j , ρi,j , σi,j , σi,j ′ , γ as shown in
 A NOTE ON FLEXIBLE IMMERSIONS 91

S3 × {5}

∆2 ∆g D′ l′ × {4}

∆1 D l × {3}

S3 × [1, 5]
b1 b2 lk l l′ S3 × {2}
bg l1

S3 × {1}

Figure 4. Proper flexible immersion f of Ng,k into D4

Figure 2 and the y–homeomorphism obtained by sliding the Möbius band

µ1 along α1 once. Therefore, it is enough to show that Dehn twists along the
Kobayashi-Omori generating curves and the y–homeomorphism are flexible
with respect to f . Notice that each Kobayashi-Omori generating curve can
be isotped to a curve on Ng,k which is in standard position with respect
to f . Therefore, by similar arguments used in the proof of Lemma 3.7, we
can see that the Dehn twists along Kobayashi-Omori generators are flexible
with respect to f .
Now, we show that the y–homeomorphism is flexible with respect to f .
Let D3 be a 3–disc in S 3 × {2} containing only the first half twisted band
and ∂D1 from the immersed surface f (Ng,k ). One can easily see that there
is an isotopy ϕt of S 3 × {2} which slides B 3 once along α1 . We use this
isotopy ϕt to define the diffeomorphism ψ : S 3 × [1, 5] → S 3 × [1, 5] given
by

(ϕt−1 (x), t), 1 ≤ t ≤ 2


ψ(x, t) = (ϕ1 (x), t), 2≤t≤3


(ϕ4−t (x), t), 3 ≤ t ≤ 4


Note the diffeormphism ψ is identity on S 3 × {1} and S 3 × {4}. There-

fore, the diffeomorphism ψ can be extended by the identity map in the
complement of S 3 × [1, 4] in D4 to a diffeomorphism Ψ : D4 → D4 which is
isotopic to the identity map. Let ψt be an isotopy of D4 with Ψ0 = Id and
Ψ1 = Ψ. Now, we consider the family ft : Ng,k → D4 of proper immersions
92 ABHIJEET GHANWAT AND SUHAS PANDIT

given by ft = Ψt ◦ f . One can easily see that f0 = f and f1 = Ψ1 ◦ f = f ◦ y.

This completes the proof of the Lemma.
□

4. Open book immersions of all closed 3–manifolds in the

trivial open book of S 5

In this section, we recall the definition of an open book decomposition

and define the notion of an open book immersion. Then, we show that all
closed 3–manifolds open book immerses in the trivial open book of S 5 . Let
us begin with the following definition.

Definition 4.1. An open book decomposition of a closed manifold M is a

pair (B, π), where
(1) B is a closed co-dimension 2 submanifold of M with trivial normal
bundle,
(2) π : M \ B → S 1 is a fiber bundle,
(3) there is a trivialization of a tubular neighborhood N (B) = B × D2
of B in M such that the map π : B × D2 \ B × {0} → S 1 sends
(x, r, θ) to θ, where x is a coordinate on B and (r, θ) are the polar
coordinates on D2 .
Here, B is called the binding, and the closure of π −1 (θ) is called the page
of the open book (B, π).

Note that an open book (B, π) of M is completely determined upto

diffeomorphism of M by the page S = π −1 (θ) of the open book (B, π)
and the monodromy ϕ of the fibration π : M \ B → S 1 . i.e. there is
diffeomorphism F : MS,ϕ → M , where

[ S × [0, 1]
MS,ϕ = MT (S, ϕ) ∂S × D2 , where MT (S, ϕ) =
(x, 1) ∼ (ϕ(x), 0)
Id

such that
(1) the map π ◦ F : MT (S, ϕ) → S 1 is given by (π ◦ F )(y, θ) = eiθ ,
(2) the map π ◦F : ∂S ×D2 \∂S ×{0} → S 1 is given by (π ◦F )(x, r, θ) =
θ, where x is a coordinate on ∂S and (r, θ) are the polar coordinates
on D2 .
 A NOTE ON FLEXIBLE IMMERSIONS 93

In this case, we call the pair (S, ϕ) an abstract open book of M associated
to the open book of (B, π). Now, we state the notion of an open book
immersion.

Definition 4.2. Let M and N be closed manifolds with the open books
(B, π) and (B ′ , π ′ ). We say that an immersion f : M → N is an open book
immersion if
(1) the map f restricted to B is an embedding of B into B ′ ,
(2) the following diagram commutes:

f
M \B N \ B′
π π′
Id
S1 S1
Remark 4.3. Note that the condition 2 in the above definition implies for
each θ ∈ S 1 , f : π −1 (θ) → π ′ −1 (θ) is a proper immersion, i.e. f properly
immerses the page of the open book (B, π) into the page of the open book
(B ′ , π ′ ).

Recall that the n–sphere S n admits the trivial open book (B ′ , π ′ ) with
pages the n − 1 disc Dn−1 and the monodromy the identity map, where

B ′ = {(x1 , x2 , . . . , xn+1 ) ∈ S n | x1 = x2 = 0}

and the map π ′ : S n \ B → S 1 is given by

(x1 , x2 )
π ′ ((x1 , x2 , . . . , xn+1 )) = .
x21 + x22
Now, we prove the following theorem as a first application of Lemma 3.7.

Theorem 4.4. Every open book of every closed 3–manifold open book im-
merses in the 5–sphere S 5 with the trivial open book.

Proof. Let M be a closed 3–manifold with an open book (B, π). Let (S, ϕ)
be an abstract open book of M associated to (B, π). Let F : MS,ϕ → M be
a diffeomorphism, where
[
MS,ϕ = MT (S, ϕ) ∂S × D2
Id
such that
(1) the map π ◦ F : MT (S, ϕ) → S 1 is given by (π ◦ F )(y, θ) = eiθ ,
94 ABHIJEET GHANWAT AND SUHAS PANDIT

(2) the map π ◦F : ∂S ×D2 \∂S ×{0} → S 1 is given by (π ◦F )(x, r, θ) =

θ, where x is a coordinate on ∂S and (r, θ) are the polar coordinates
on D2 .
Similarly, the trivial open book decomposition (B ′ , π ′ ) of S 5 decomposes
S 5 as
D4 × S 1 ∪Id S 3 × D2
such that
(1) the map π ′ : D4 × S 1 → S 1 is given by π ′ (y, θ) = θ,
(2) the map π ′ : S 3 × D2 \ S 3 × {0} → S 1 is given by π(x, r, θ) = θ,
where x is a coordinate on S 3 and (r, θ) are the polar coordinates
on D2 .
Our first aim is to construct a proper immersion G : MT (S, ϕ) ↬
D4 × S 1 such that π ′ ◦ F = π. By Lemma 3.7 for orientable surfaces and
by Lemma 3.8 for non-orientable surfaces, there exists a proper flexible
immersion g : S ↬ D4 . Therefore, there exists a family gt : S ↬ D4 of
proper immersions such that g0 = g and g2π = g ◦ ϕ. This family gt allows
us to define the desired map G as G(y, θ) = gθ (y).
Note that the map G restricted to the boundary ∂MT (S, ϕ) of MT (S, ϕ)
is an embedding. Now, it is easy to see that the proper immersion G can
be extended to an open book immersion of MS,ϕ into the trivial open book
of S 5 . Thus, G ◦ F −1 : M → S 5 is the required open book immersion of M
into S 5 . □

5. Representing the homology classes by flexible immersions

It is well known that every d ∈ H2 (CP 2 , Z) can be represented by an

embedding of a closed-oriented surface. Therefore, it is interesting to know
that given d ∈ H2 (CP 2 , Z), can we represent it by a flexible embedding of
a closed-oriented surface? In [9], Kronheimer and Mrowka showed that if a
closed oriented smoothly embedded genus g surface Σ in CP 2 represents the
same homology class as an algebraic curve of degree d then g ≥ (d−1)(d−2)2 .
2
Therefore, it is not possible to represent a given d ∈ H2 (CP , Z) by a closed-
oriented genus g embedded surface for every g. Hence, it is not possible to
represent a given d ∈ H2 (CP 2 , Z) by a flexible embedding of the closed
oriented surface of genus g for every g. Here, we ask the same question for
flexible immersions.
 A NOTE ON FLEXIBLE IMMERSIONS 95

In this section, we give a second application of flexible immersions of

compact surfaces in 4–manifolds. We show that given a second integral
homology class d of CP 2 and given a closed oriented surface Σ, there is a
flexible immersion f : Σ ↬ CP 2 such that f represents the class d. We
also show that each of the classes 0 and ±2 ∈ Z = H2 (CP 2 , Z) can be
represented by a flexible embedding f : Σ ,→ CP 2 of any oriented surface
Σ into CP 2 . Let us begin by the following Lemma.
Lemma 5.1. Let g : S 2 ↬ X be an immersion of the 2–sphere S 2 into
a closed oriented 4-manifold X such that g represents the second integral
homology class d ∈ H2 (X, Z). Suppose g is injective in the complement of
finitely many points on S 2 . Then, given a closed-oriented surface Σ, there
is a flexible immersion f : Σ ↬ X which represents the class d.
Proof. Let D4 be a 4–disc in X disjoint from the immersion g, i.e. D4 ∩
g(S 2 ) = ∅. Let h : Σ ↬ D4 ⊂ X be a flexible immersion of the closed
oriented surface Σ in D4 as constructed in Lemma 3.7. Therefore, the
immersion h represents the class 0 ∈ H2 (X, Z). Note that that the the
flexible immersion h is in a collar S 3 × [1, 5] of ∂D4 in D4 with ∂D4 =
S 3 × {5}. Also note that the immersion h is injective in the complement
of two points on Σ. Therefore, there is a point p1 on S 2 and a point
p2 on Σ such that the g −1 (g(p1 )) = p1 and h−1 (h(p2 )) = p2 . Let α be an
embedded arc in (X \D4 )∪S 3 ×[2, 5] with the end points g(p1 ) and g(p2 ) on
g(S 2 ) and h(Σ), respectively such that the interior of the arc α is disjoint
from both the immersions g and h. Now, we take appropriate ambient
connected sum of the immersions g and h in X along a tube contained
in a tubular neighborhood N (α) of the arc α in X to get the immersion
f = g#h : Σ ≈ S 2 #Σ ↬ X such that the immersion f represents the class
d. Now, it is easy to see that the flexibility of the immersion f follows by
the similar arguments used in Lemma 3.7. □
Remark 5.2. Any generic immersion g : S 2 ↬ X 4 is injective in the
complement of finitely many points of S 2 .
Now, we prove the following theorem.
Theorem 5.3. Let d ∈ H2 (CP 2 , Z) be a second integral homology class
of the complex projective 2–space CP 2 and Σ be a closed oriented surface.
Then, there exists a flexible immersion f : Σ ↬ CP 2 which represents the
class d.
96 ABHIJEET GHANWAT AND SUHAS PANDIT

Proof. By Lemma 5.1, it is enough to construct an immersion g : S 2 ↬ CP 2

which is injective in the complement of finitely many points of S 2 and
represents the class d.
To construct the required immersion g, we consider the following conve-
nient handle decomposition of CP 2 . The handle decomposition consists of
one 0–handle H0 , one 2–handle H2 and one 4–handle H4 . In this decompo-
sition, the 2–handle H2 is attached to the 0–handle along an unknot in ∂H0
with the framing −1 and finally the 4–handle H4 is attached to H0 ∪ H2
along ∂(H0 ∪ H2 ) = S 3 to get CP 2 .
Now, we describe the desired immersion g of S 2 in CP 2 . We regard S 3 ×
[1, 5] as a collar of ∂H0 = ∂D4 in the 0–handle H0 = D4 with ∂D4 = S 3 ×
{5}. Consider an embedded 2–sphere S as the boundary of a 3–disc in S 3 ×
{2}. Then, we remove the interiors of d disjoint discs D1 , D2 , . . . , Dd from
S and attach d cylinder ∂D1 × [2, 5], . . . , ∂Dd × [2, 5] along ∂D1 , . . . , ∂Dd ,
respectively to get a properly embedded surface S ′ of genus 0 in H0 with
d boundary components l1 , l2 , . . . , ld . Let K be the attaching circle of the
2–handle H2 with the framing −1 and let N (K) = S 1 × D2 be an attaching
region of H2 in ∂H0 . Notice that each li is an unknot in ∂H0 = S 3 without
linking each other. Therefore, without loss of generality, we can assume
that li = S 1 × {( 1i , 0)} ⊂ S 1 × D2 = N (K). Now, it is easy to see that each
li bounds a disc Di in the 2–handle H2 and intersects the co-core of H2
once. Since the 2–handle H2 is attached along K with −1 framing, all the
discs Di ’s intersect at the origin of the co-core of H2 . Thus, the embedded
surface S ′ together with the d discs D1 , D2 , . . . , Dd produces an immersion
of S 2 in CP 2 which is injective on S 2 in the complement of 2d points. We
denote this immersion by the map g : S 2 → CP 2 . By choosing the correct
orientation on S 2 , we can make sure that [g] = d ∈ Z = H2 (CP 2 , Z). □

Remark 5.4. Let d ∈ H2 (#k CP 2 , Z) be a second integral homology class of

the connected sum #k CP 2 of k copies of the complex projective 2–spaces
CP 2 ’s and Σ be a closed oriented surface. Then, there exists a flexible
immersion f : Σ ↬ #k CP 2 which represents the class d.

Finally, we prove the following theorem.

Theorem 5.5. Let Σg be a closed-oriented surface of genus g. Then, each

of the second integral homology classes 0 and ±2 ∈ Z = H2 (CP 2 , Z) can be
represented by a flexible embedding f : Σg ,→ CP 2 .
 A NOTE ON FLEXIBLE IMMERSIONS 97

Proof. We consider the same handlebody decomposition of CP 2 as de-

scribed in the proof of Theorem 5.3 and regard S 3 × [1, 5] as a collar of
∂H0 = ∂D4 in the 0–handle H0 = D4 with ∂D4 = S 3 × {5}.

Representing the class 0 ∈ H2 (CP 2 , Z) by a flexible embedding of

Σg :
Consider the embedded surface Σg as a boundary of standardly embed-
ded genus g handlebody Hg in S 3 × {2}. We denote this embedding of Σg
in CP 2 by the map f : Σg ,→ CP 2 . In order to show the embedding f is
flexible, it is enough to show that the Dehn twists along each of the Lickor-
ish generating curves on Σg is flexible with respect to the embedding f . Let
γ be a Lickorish generating curve on Σg and A be a regular neighborhood
of γ in Σg . Observe that if there is an isotopy ψt : CP 2 → CP 2 of CP 2
such that ψ0 = Id and the curve γ is in standard position with respect to
the embedding ψ1 ◦ f , then the curve γ is also in standard position with
respect to the embedding f . Then, the flexibility of the Dehn twist dγ with
respect to f follows by Lemma 3.5.
Let us show that such an isotopy ψt exists. Notice that a regular neigh-
borhood f (A) of f (γ) in f (Σg ) is a planar annulus A in S 3 × {2}, i.e. ∂A
is a link consisting of two unknots with the linking number 0 in S 3 × {2}.
Let K be the attaching circle of the 2–handle H2 with the framing −1 and
let N (K) = S 1 × D2 be an attaching region of H2 in ∂H0 . Now, one can
observe the following:
(1) There is family ft : Σg → H0 ⊂ CP 2 , 0 ≤ t ≤ 1 of embeddings such
that f0 = f and f1 (A) = S 1 × [−1, 1] ⊂ S 1 × D2 = N (K) ⊂ ∂H0
and all ft agree with f in the complement of a small neighborhood
of A in Σg . This can be achieved by pushing the planar annulus A
towards the attaching region N (K) of the 2–handle H2 .
(2) Since, the 2–handle H2 is attached to H0 along the unknot K with
the framing −1, f1 (A) = S 1 × [−1, 1] becomes a Hopf annulus H ⊂
∂H2 in the boundary of the 2–handle H2 .
(3) Therefore, there is family ht : Σg → H0 ∪ H2 ⊂ CP 2 of embeddings
such that h0 = f1 and h1 (A) is a properly embedded annulus in H2
with h1 (A) = g −1 (1), where g : H2 = D4 → C is Lefschetz fibration
given by g(z1 , z2 ) = z1 z2 with complex coordinates (z1 , z2 ) on D4 .
The family ht can be achieved by pushing the interior of H ⊂ ∂H2
into the interior of the 2–handle H2 .
98 ABHIJEET GHANWAT AND SUHAS PANDIT

∂H0 S3 × {5}

S3 × [1, 5]

l l′ S3 × {2}

S3 × {1}

Figure 5. Figure depicts the proper embedding g of Σg,2

into the 0–handle H0 .

Now, we consider the following family of embeddings gt : Σg → CP 2

defined by.

f
t 0≤t≤1
gt =
ht−1 1≤t≤2
Note that the curve γ is in standard position with respect to the embedding
g2 = h1 . Since gt is a family of embeddings of Σg in CP 2 , there is an isotopy
ψt : CP 2 → CP 2 such that gt = ψt ◦ f. Thus, as the curve γ is in standard
position with respect to ψ1 ◦ f = g2 , we have the required isotopy ψt .

Representing the classes ±2 ∈ H2 (CP 2 , Z) by a flexible embeddings

of Σg into CP 2 :
Let Σg be the boundary of a standardly embedded genus g handlebody
in S 3 ×{2}. Now, we remove the interior of a disc D from Σg and attach one
full twisted band along ∂D. This produces an embedded genus g surface Σg,2
with 2 boundary components l, l′ in S 3 × {2} as shown in Figure 5. Finally,
the embedded surface Σg,2 together with 2 cylinders l × [2, 5], l′ × [2, 5]
produces a proper embedding of the genus g surface Σg,2 with 2 boundary
components in D4 . We denote this embedding by the map g : Σg,2 ,→ CP 2 .
Now, we describe a flexible embedding f2 : Σg ,→ CP 2 which represents the
class 2 ∈ H2 (CP 2 , Z).
Since, the link l ∪ l′ is a Hopf link in ∂H0 , we can assume that the
2–handle H2 is attached to H0 along the unknot l with the framing −1 and
the unknot l′ corresponds to the framing −1 on the attaching region N (l)
of the 2–handle in ∂H0 . As the unknot l′ links once to the attaching circle l
 A NOTE ON FLEXIBLE IMMERSIONS 99

of the 2–handle H2 in ∂H0 and the 2–handle H2 is attached to the 0–handle

H0 along l with the framing −1, the unknot l bounds a disc D (the core
of the 2–handle H0 ) in the 2–handle and the unknot l′ bounds a disc D′
which is parallel to the core disc D. Notice that the discs D and D′ are
disjoint and each of them intersects once positively to the co-core of the
2–handle H0 . Thus, the embedding g : Σg,2 ,→ CP 2 together with the discs
D and D′ gives an embedding f2 : Σg ,→ CP 2 which represents the class
2 ∈ H2 (CP 2 , Z). The flexibility of the embedding f2 follows by the similar
arguments used in the Lemma 3.7.
Let ϕ : Σg → Σg be an orientation reversing diffeomorphism of Σg .
Then, the flexible embedding f2 ◦ ϕ represents the class −2 in H2 (CP 2 , Z).
□

Remark 5.6. Let Σg be a closed oriented surface of genus g. Then, each of

the second integral homology classes (d1 , d2 , . . . , dk ) ∈ Zk = H2 (#k CP 2 , Z)
can be represented by a flexible embedding f : Σ ,→ #k CP 2 of the oriented
surface Σg in #k CP 2 , where each di ∈ {0, ±2}.

References
[1] Alexander, J., A lemma on systems of knotted curves, Proc. Nat. Acad. Sci., 9
(1923), 93-95.
[2] Berstein, I. and Edmonds, A., On the construction of branched coverings of low-
dimensional manifolds, Trans. Amer. Math. Soc., 247 (1979), 87-124.
[3] Etnyre, J., Lectures on open book decompositions and contact structures, Lecture
notes for Clay Summer School at the Alfred Renyi Institute of Mathematics (Hun-
garian Academy of Sciences) Budapest, Hungary, (2004).
[4] Etnyre, J. and Lekili, Y., Embedding all contact 3-manifolds in a fixed contact
5-manifold. J. Lond. Math. Soc. (2), 99(1) (2019), 52–68.
[5] Ghanwat, A. and Pancholi, D., Embeddings of 4–manifolds in CP 3 .
arXiv.2002.11299 [math.GT] (2020).
[6] Ghanwat, A., Pandit, S. and Selvakumar, A., Open book embeddings of closed
non-orientable 3–manifolds. Rocky Mountain J. Math., 49 (2019), no. 4, 1143–1168.
[7] Ghanwat, A., Pandit, S. and Selvakumar, A., Open books for closed non-orientable
3-manifolds, Glasg. Math. J., 62 (2020), 584–599.
[8] Hirose, S. and Yasuhara, A., Surfaces in 4-manifolds and their mapping class groups,
Topology, 47 (1) (2008), 41–50.
[9] Kronheimer, P. and Mrowka, T., The genus of embedded surfaces in the projective
plane, Math. Res. Lett., 1(6) (1994), 797–808.
[10] Lickorish, W., A representation of orientable combinatorial three-manifolds, Ann.
of Math. (2), 76 (1962), 531–540.
100 ABHIJEET GHANWAT AND SUHAS PANDIT

[11] Kobayashi, R. and Omori, G., An infinite presentation for the mapping class group
of a non-orientable surface with boundary, Osaka J. Math. , 59 (2) (2022), 269–314.
[12] Lickorish, W., Homeomorphisms of non-orientable two-manifolds, Proc. Cambridge
Philos. Soc., 59 (1963), 307–317.
[13] Pancholi, D., Pandit, S. and Saha, K., Embeddings of 3–manifolds via open books,
J. Ramanujan Math. Soc. 36 (3) (2021), 243–250.
[14] Pandit, S. and Selvakumar, A., Embeddings of 4–manifolds in S 2 × S 4 and S 2 × S 4
using bordered Lefschetz fibration, Houston Journal of Mathematics, 48(3) (2022),
691–723.

Tata Institute of Fundamental Research, 1st Homi Bhabha Road, Co-

laba, Mumbai-400 005, India.
Email address: [email protected]

Indian Institute of Technology Madras, IIT PO. Chennai- 600 036, Tamil
Nadu, India.
Email address: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 101–115

A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN

Sp(N, 1)

DEVENDRA TIWARI
(Received : 23 - 04 - 2023 ; Revised : 13 - 07 - 2024)

Abstract. In this work we shall provide a proof of the Shimizu type

inequality for two generator subgroups of Sp(n, 1) with a unipotent
parabolic generator. We will also compare our result with some existing
Shimizu type results and as an application discuss the extremality of
the inequality. Our main result is useful in proving discreteness criteria
for Zariski dense subgroups of Sp(n, 1) and SU(n, 1) such as [GMT19].

1. Introduction

To motivate the reader, for Shimizu lemma in Sp(n, 1), we will start with
the classical Shimizu lemma and other discreteness criteria for subgroups
in SL(2, C). We also mention some application of related ideas in classical
case.

1.1. Historical Background. The study of discrete subgroups of the group

of orientation preserving isometries of hyperbolic spaces is closely related
with the study of geometry and topology of corresponding orbit spaces i.e.
of hyperbolic manifolds. Here a group G is said to be discrete if it is a
discrete set in the usual matrix topology or equivalently by discreteness we
mean that for a sequence of transformations Tn → I in the group G, we
must have Tn = I ∀ large n.
The classical Jørgensen inequality, see [Jor76], gives a necessary criterion
to check discreteness of a two generator subgroup (say G) of SL(2, C) that
acts by Möbius transformations on the Riemann sphere. Jørgensen in 1976
proved that the question of discreteness of arbitrary groups in PSL(2, C)
can be reduced to the question of discreteness of two generator subgroups.
2010 Mathematics Subject Classification: Primary 20H10; Secondary 51M10
Key words and phrases: hyperbolic space, Jørgensen inequality, discreteness,
quaternions.

© Indian Mathematical Society, 2024 .

101
102 DEVENDRA TIWARI

Now question arises when the two generator group G = hf, gi considered as
a subgroup of PSL(2, C) is discrete?
Only necessary or only sufficient conditions for the discreteness have
been obtained so far. For the necessary condition, we have the remark-
able Jørgensen inequality, the inequality of Shimizu-Leutbecher and their
analogues. A sufficient condition has been discussed in [GMMR97].
The Jörgensen inequality is a celebrated result concerning the classi-
cal problem to understand discreteness of subgroups of SL(2, C). It pro-
vides the necessary condition of discreteness for a two generator subgroup
of SL(2, C) (see [Jor76]). Jörgensen’s inequality, which generalises the clas-
sical Shimizu’s lemma, for two generator subgroups of PSL(2, C) is among
the most important results and tools in the study of three dimensional
manifolds.
An element f ∈ SL(2, C) is elliptic if it has a fixed point on the hyper-
bolic 3-space. It is parabolic, respectively loxodromic, if it is non-elliptic and
has exactly one, respectively two fixed points on the boundary Ĉ. For ellip-
tic transformations canonical form of f is given by z 7→ λz, |λ| = 1 such that
tr(f ) ∈ R with tr2 (f ) < 4. For parabolic transformations canonical form is
given by z → z + k, k ∈ C. such that tr(f ) = ±2. For loxodromic trans-
formations canonical form is given by z 7→ λz, λ ∈ C and |λ| 6= 1. From
1 1
the canonical form, of loxodromic element, note that, tr(f ) = λ 2 + λ− 2
with tr(f ) 6= ±2. This subdivides loxodromic elements into: (purely) hy-
perbolic when λ ≥ 0 such that tr(f ) ∈ R, with tr2 (f ) > 4 and loxodromic:
tr(f ) ∈
/ R. With this background we will now state Jörgensen inequality:

Theorem 1.1. [Jor76]. If G =< f, g > is discrete group of Möbius trans-

formations, acting on Riemann Sphere Ĉ then we have

ktr2 (f ) − 4k + ktr(f gf −1 g −1 ) − 2k ≥ 1 (1.1)

except in the following three cases, which are elementary groups:
(A) G cyclic or a finite abelian extension of a cyclic group and |tr2 (f ) −
4| < 1.
(B) f is loxodromic or elliptic with |tr2 (f ) − 4| < 1/2, while g inter-
changes the fixed points of f .
(C) f is parabolic while g is parabolic or elliptic of order 2, 3, 4 or 6
and fixes the fixed point of f .
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 103

Remark 1.2. If one of the generators, of two generator subgroup for which
Jörgensen’s inequality holds, is parabolic then the discreteness condition
given above is just a reformulation of the classical Shimizu’s lemma. In this
sense Jörgensen’s inequality generalises Shimizu’s lemma and deals also with
groups with loxodromic and elliptic generators.

More precisely classical Shimizu’s lemma [Shi63] gives a necessary con-

dition for a subgroup of PSL(2, R) containing a parabolic element to be
discrete. This result is also very important in studying the structure of
cusp(s)in Riemann surfaces. Shimizu proved that:

Theorem 1.3. [Shi63] Let G be a discrete subgroup of PSL(2, R) containing

the parabolic map T (z) = z + t for some t > 0 and any S(z) = az+b
cz+d ∈ G
with c 6= 0 then |c| ≥ 1/t.

This result Shimizu obtained, as a by-product, in his work of comput-

ing the dimension of the space of cusp forms corresponding to any irre-
ducible, discrete subgroup of a product of copies of SL(2, R). Geometrically
Shimizu’s lemma says that the radius rS of the isometric sphere of isometry
S satisfies rS ≤ t, where t is the translational length of the T . Therefore
Shimizu lemma provides a uniform bound on the radii of isometric spheres
of those elements of the group not having infinity as their fixed point. This
uniform bound is the translation length (in Euclidean metric) of the para-
bolic map fixing infinity. For the modular group PSL(2, Z), Shimizu lemma
is sharp with t = 1, and S(z) = −1/z has rS = 1. There is a cusp neigh-
borhood C which cannot be enlarged. It has area equal to 1, which is large
compared to the area H2 /PSL(2, Z). In another way other hand this can
also be considered as a special case of another important result known as
Collar lemma. Later these discreteness criteria have been vastly generalized;
a striking example is Margulis’s lemma for non-positively curved manifolds.
We will mention a couple of generalisations:

1.2. Some Applications. The Margulis lemma (named after Grigory Mar-
gulis) is a result about discrete subgroups of isometries of a non-positively
curved Riemannian manifold (e.g. the hyperbolic n-space). Roughly, it
states that within a fixed radius, usually called the Margulis constant, the
structure of the orbits of such a group cannot be too complicated. More
precisely, within this radius around a point all points in its orbit are in fact
104 DEVENDRA TIWARI

in the orbit of a nilpotent subgroup (in fact a bounded finite number of

such). The Margulis lemma can be formulated as follows.
Theorem 1.4. [Rat2006] Let X be a simply-connected manifold of non-
positive bounded sectional curvature. There exist constants C, ε > 0 with the
following property. For any discrete subgroup Γ of the group of isometries
of X and any x ∈ X, if Fx is the set:

Fx = {g ∈ Γ : d(x, gx) < ε}

then the subgroup generated by Fx contains a nilpotent subgroup of index
less than C.

Here d is the distance induced by the Riemannian metric. A particularly

well studied family of examples of negatively curved manifolds are given by
the symmetric spaces associated to semisimple Lie groups. In this case
the Margulis lemma can be given the following more algebraic formulation
which dates back to Hans Zassenhaus.
Theorem 1.5. [Rag72] If G is a semisimple Lie group there exists a neigh-
bourhood Ω of the identity in G and a C > 0 such that any discrete subgroup
Γ which is generated by Γ ∩ Ω contains a nilpotent subgroup of index ≤ C.

Such a neighbourhood Ω is called a Zassenhaus neighbourhood in G.

1.3. Shimizu Lemma in Sp(n, 1). In this paper we prove a generalisation

of the Shimizu lemma in Sp(n, 1). A complex hyperbolic version of this
lemma was obtained by Hersonsky and Paulin in (proposition (A.1)[HP96]).
More precisely, in complex case, Kamiya [Kam83] generalised Shimizu’s
lemma for vertical Heisenberg translations and Parker [Par92, Par97] gener-
alised Shimizu’s lemma for non-vertical Heisenberg translation or some spe-
cial type of screw parabolic map. Jiang, Parker and Kamiya [JP03, KP08]
generalised Shimizu’s lemma to groups containing screw parabolic map. In
quaternionic case, Kim and Parker [KP03] generalised Shimizu’s lemma
to groups containing non-vertical Heisenberg translation. In case n = 2,
Daeyong Kim [Kim04] extended the results of [JKP03] to quaternionic case.
Cao-Parker in [CP18] proved a general version of the Shimizu’s lemma that
involves an arbitrary parabolic map. Other related problems and results on
Shimizu type inequality and discreteness see [BM98, HP96, Par98, GMT19,
GMT21] and references therein.
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 105

Though the methods of Cao and Parker works for an arbitrary parabolic
map, it is pretty technical and we find it difficult for application e.g. to
obtain discreteness criteria for Zariski dense subgroups in Sp(n, 1) as in
[GMT19] the version of Shimizu type lemma proved by Cao-Parker is not
suitable. Hence, here we provide a much simpler form of the quaternionic
Shimizu’s lemma that is quite analogous to the result of Hersonsky and
Paulin. We must remark that though this version is simpler, it is weaker
than the versions of Kim and Parker, and Cao and Parker mentioned above.
We have also noted a corollary that gives us some information about the
extremality of the inequality.

Let H denote the division ring of Hamilton’s quaternions. Let HnH de-
note the n-dimensional quaternionic hyperbolic space. Let Sp(n, 1) be the
linear group that acts on HnH by isometries. Up to conjugacy, we assume
that (see following section for the details) an Heisenberg translation fixes
the boundary point 0, i.e. it is of the form
 
1 0 0
Ts,ζ = s 1 ζ ∗  , (1.2)
 

ζ 0 I
where ζ is column vector known as translation length of Ts,ζ , s is a scalar
with Re(s) = 12 |ζ|2 and I is (n − 1) × (n − 1) identity matrix. ζ ∗ denotes the
conjugate transpose of ζ. Now we state a version of Shimizu type lemma:

Theorem 1.6. Suppose Ts,ζ is a Heisenberg translation in Sp(n, 1) and

A ∈ Sp(n, 1) is given by
 
a b γ∗
A =  c d δ∗ 
 

α β U
Suppose A does not fix 0. Set

t = Sup{|b|, |β|, |γ|, |U − I|}, M = |s| + 2|ζ|.

If
M t + 2|ζ| < 1,
then the two generator group hA, Ts,ζ i is either non-discrete or fixes the
point 0.
106 DEVENDRA TIWARI

After some preliminaries in section 2, we will prove the above theorem

in section 3 and in section 4 we will note some applications of this theorem.

2. Preliminaries

We begin with some background material on quaternionic hyperbolic

geometry. Much of this can be found in [CG74, KP03].
Let Hn,1 be the right vector space over H of quaternionic dimension
(n+1) (so real dimension 4n+4) equipped with the quaternionic Hermitian
form for z = (z0 , ..., zn ), w = (w0 , ..., wn ),

hz, wi = −(z̄0 w1 + z̄1 w0 ) + Σni=2 z̄i wi

Thus the Hermitian form is defind by the matrix

 
0 −1 0
J = −1 0 0 
 

0 0 In−1
Following Section 2 of [CG74], let
n o n o
V0 = z ∈ Hn,1 − {0} : hz, zi = 0 , V− = z ∈ Hn,1 : hz, zi < 0 .

It is obvious that V0 and V− are invariant under Sp(n, 1). We define an

equivalence relation ∼ on Hn,1 by z ∼ w if and only if there exists a non-
zero quaternion λ so that w = zλ. Let [z] denote the equivalence class of z.
Let P : Hn,1 −{0} −→ HPn be the right projection map given by P : z 7−→ z,
where z = [z]. The n dimensional quaternionic hyperbolic space is defined
to be HnH = P(V− ) with boundary ∂HnH = P(V0 ).

In our chosen model there are two distinct points 0 and ∞ on ∂HnH . For
z1 6= 0, the projection map P is given by

P(z1 , z2 , . . . , zn+1 ) = (z2 z1−1 , . . . , zn+1 z1−1 ),

and accordingly we choose boundary points

P(0, 1, . . . , 0, 0)t = 0. (2.1)

P(1, 0, . . . , 0, 0)t = ∞. (2.2)

 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 107

2.1. Bergmann and Cygan Metric. The Bergmann metric on HnH is

given by the distance formula
ρ(z, w) hz, wihw, zi
cosh2 = , where z, w ∈ HnH , z ∈ P−1 (z), w ∈ P−1 (w).
2 hz, zihw, wi
The above forumula is independent of the choice of z and w.

The finite points in the boundary of HnH naturally carry the structure of
the generalised Heisenberg group Nn , which is defined to be Nn = Hn−1 H ×
=(H), with the group law

(ζ1 , v1 )(ζ2 , v2 ) = (ζ1 + ζ2 , v1 + v2 + 2=(ζ2∗ ζ1 )). (2.3)

We define horospherical coordinates on quaternionic hyperbolic space as

follows.
ψ : Nn × R+ → V 0 ∪ V −

   
  01
(−|ζ|2 − u + v)/2  . 
0
 .. 
 
 
ψ(ζ, v, u) =  ζ  , ψ(∞) =   , ψ(o) = 
    .. 
 0 
.
 
1  
0 1
(2.4)
The Cygan metric on Heisenberg group corresponding to the norm
1 1
|(ζ, v)|H = ||ζ|2 + v| 2 = (|ζ|4 + |v|2 ) 4

is given by

1
dH ((ζ1 , v1 ), (ζ2 , v2 )) = |(ζ1 , v1 )−1 (ζ2 , v2 )|H = |2hψ(ζ1 , v1 , 0), ψ(ζ2 , v2 , 0)i| 2
= |(ζ2 − ζ1 , v2 − v1 − 2=(ζ2∗ ζ1 )|H
1
= ||ζ2 |2 + |ζ1 |2 − 2ζ2∗ ζ1 + v2 − v1 | 2 .

Now consider the non-compact linear Lie group

Sp(n, 1) = {A ∈ GL(n + 1, H) : A∗ JA = J}.

n
An element g ∈ Sp(n, 1) acts on HH = HnH ∪ ∂HnH as g(z) = PgP−1 (z).
Thus the isometry group of HnH is given by PSp(n, 1) = Sp(n, 1)/{I, −I}.
108 DEVENDRA TIWARI

2.2. Classification of Isometries. As with real and complex hyperbolic

isometries, non-trivial element g of Sp(n, 1) are classified as:
(i) elliptic if it has a fixed point in HnH .
(ii) parabolic if it has exactly one fixed point which lies in ∂HnH .
(iii) loxodromic if it has exactly two fixed points which lie in ∂HnH .

Definition 2.1. Let Hs be a parabolic element of Sp(n, 1). Up to conju-

gacy, we may assume it to fix a point 0 on the boundary. Using Jordan
decomposition, we can see it to be of the form, up to conjugacy,

Hs = R(µ,U ) T(ζ,s)

The element with the form T(ζ,s) is called a Heisenberg translation and
it fixes the boundary point 0. Whereas R(µ,U ) is an elliptic element fixing
0.

Up to conjugacy, Heisenberg translation Tζ,s ∈ Sp(n, 1) is given by (see

[CG74, p. 70])  
1 0 0
Ts,ζ = s 1 ζ ∗  . (2.5)
 

ζ 0 I
here ζ is column vector known as translation length of Ts,ζ , s is a scalar
with Re(s) = 21 |ζ|2 and I is (n − 1) × (n − 1) identity matrix. ζ ∗ denotes the
conjugate transpose of ζ. If ζ = 0, it is a vertical Heisenberg translation,
otherwise it is a non-vertical Heisenberg translation.

Identifying the boundary ∂HnH with N4n−1 , the (4n − 1)-dimensional

generalized Heisenberg group with 3-dimensional center. There is a natural
metric, the Cygan metric, on N4n−1 . Let T be a parabolic element fixing
∞. It is a Cygan isometry on N4n−1 . The natural projection from N4n−1
to Hn−1 defines the vertical projection of T , which is a Euclidean isometry
of Hn−1 .

Definition 2.2. Let g ∈ Sp(n, 1) not fixing ∞, then, there exists a Cygan
sphere, called the isometric sphere, in N4n−1 centered at g −1 (∞),

Ig = {z ∈ HH : |hψ(z), ψ(∞)i| = hψ(z), g −1 (ψ(∞))i|}

which is sent by g to the Cygan sphere centered at g(∞) having the
same radius.
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 109

2.3. Some Computation. Now we shall use the Hermitian form J to note
down a few equations to be used later. Letting z and w vary over a basis
for Hn,1 , we see that J = A∗ JA. From this we find that A−1 = J −1 A∗ J for
A ∈ Sp(n, 1) given by
 
a b γ∗
A =  c d δ∗  (2.6)
 

α β U
 
d¯ b̄ −β ∗
A−1 =  c̄ ā −α∗ 
 

−δ −γ U ∗
where a, b, c, d are scalars, γ, δ, α, β are column matrices and U is an
element in U (n − 1, H).
Since AA−1 = I, equating both sides we have the following relations:

ad¯ + bc̄ − γ ∗ δ = 1
ab̄ + bā − γ ∗ γ = 0
−aβ ∗ − bα∗ + γ ∗ U ∗ = 0
cd¯ + dc̄ − δ ∗ δ = 0
cb̄ + dā − δ ∗ γ = 1
−cβ ∗ − dα∗ + δ ∗ U ∗ = 0
αd¯ + βc̄ − U δ = 0
αb̄ + βā − U γ = 0
−αβ − βα∗ + U U ∗ = I
∗

¯ + b̄c − β ∗ α = 1
da
¯ + b̄d − β ∗ β = 0
db
¯ ∗ + b̄δ ∗ − β ∗ U
dγ = 0
c̄a + āc − α∗ α = 0
c̄b + ād − α∗ β = 1
c̄γ ∗ + āδ ∗ − α∗ U = 0
∗
−δa − γc + U α = 0
−δb − γd + U ∗ β = 0
−δγ ∗ − γδ ∗ + U ∗ U = I
110 DEVENDRA TIWARI

2.4. Some Shimizu type Theorems. Before proving our main result, we
will note here some other Shimizu type theorems.

Theorem 2.3. [KP08] Let A be a positively oriented screw parabolic ele-

ment of P U (2, 1) fixing ∞. Let eiθ ∈ U (1) denote the rotational part of A
and suppose that |eiθ − 1| < 1/4. Let B be any element of P U (2, 1) not
protectively fixing ∞ and let rB denote the radius of the isometric sphere of
B. If

p !2
ρ0 (B(∞), AB(∞))ρ0 (B −1 (∞), AB −1 (∞)) 1+ 1 − 4|eiθ − 1|
2 <
rB 2

then hA, Bi is not discrete.

Theorem 2.4. [KP03] Let G be a discrete subgroup of PSp(n, 1) that con-

tains a Heisenberg translation g by (τ, t) ∈ Nn . Let h be any element of G
not fixing ∞ with isometric sphere of radius rh . Then

rh2 ≤ dH (h−1 (∞), gh−1 (∞))dH (h(∞), gh(∞)) + 4|τ |2

3. A Shimizu-Type Theorem for Subgroups in Sp(n, 1)

The following theorem follows essentially mimicking the argument of

Hersonsky and Paulin in Appendix of [HP96].

Theorem 3.1. Suppose Ts,ζ be a Heisenberg translation in Sp(n, 1) and A

be an element in Sp(n, 1) of the form (2.6). Suppose A does not fix 0. Set

t = Sup{|b|, |β|, |γ|, |U − I|}, M = |s| + 2|ζ|. (3.1)

If
M t + 2|ζ| < 1, (3.2)
then the group generated by A and Ts,ζ is either non-discrete or fixes 0.

When ζ = 0, as a corollary to the above theorem we obtain theorem

(3.2)[Kam83].
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 111

Proof. Consider the sequence A0 = A, Ak+1 = Ak Ts,ζ A−1

k . Thus
 
ak+1 bk+1 γk+1 ∗
Ak+1 =  ck+1 dk+1 δk+1 ∗  (3.3)
 

αk+1 βk+1 Uk+1

   
ak bk γk ∗ 1 0 0 d¯k b̄k −βk∗
=  ck dk δk ∗  s 1 ζ ∗   c̄k āk −αk∗  (3.4)
   

αk βk Uk ζ 0 I −δk −γk Uk∗

(3.5)

Using the above set of relations, equating both sides we have:

ak+1 = 1 + bk sd¯k − bk ζ ∗ δk + γk∗ ζ d¯k (3.6)

bk+1 = bk sb̄k + 2Im(γk∗ ζ b̄k ) (3.7)
ck+1 = dk sd¯k + 2Im(δk∗ ζ d¯k ) (3.8)
dk+1 = 1 + dk sb̄k + δk∗ ζ b̄k − dk ζ ∗ γk (3.9)
∗
γk+1 = −bk sβk∗ − γk∗ ζβk∗ + bk ζ ∗ Uk∗ (3.10)
∗
δk+1 = −dk sβk∗ − δk∗ ζβk∗ + dk ζ ∗ Uk∗ (3.11)
αk+1 = βk sd¯k − βk ζ ∗ δk + Uk ζ d¯k (3.12)
∗
βk+1 = βk sb̄k − βk ζ γk + Uk ζ b̄k (3.13)
Uk+1 = I − βk sβk∗ − Uk ζβk∗ + βk ζ ∗
Uk∗ . (3.14)

Set
tk = Sup{|bk |, |βk |, |γk |, |Uk − I|}, M = |s| + 2|ζ|.
Now it follows from the expressions above that

|bk+1 | ≤ M t2k (3.15)

|γk+1 | ≤ M t2k + |bk ||ζ| (3.16)
|βk+1 | ≤ M t2k + |bk ||ζ| (3.17)
|Uk+1 − I| ≤ M t2k + 2|βk ||ζ|. (3.18)

These relations imply that

tk+1 ≤ M t2k + 2|ζ|tk , . (3.19)

Choose  > 0 such that

M t0 + 2|ζ| ≤  < 1.
112 DEVENDRA TIWARI

From (3.19) it follows that t1 ≤ t0 , t2 ≤ t0 (M t1 + 2|ζ|). Now note that
M t1 ≤ M t0 < M t0 , hence M t1 + 2|ζ| ≤ M t0 + 2|ζ| < , hence t2 ≤ 2 t0 .
Repeating this proces we have

tk ≤ k t0 .

Since  < 1, hence tk → 0. In particular, Uk → I, βk → 0, γk → 0, bk → 0.

Now note that for k ≥ 3

|dk+1 − 1| < ktk (|s| + |ζ| + |ζ|2 )Ck ,

|δk+1 − ζ| ≤ ktk (|s| + |ζ| + |ζ|2 )Ck + |ζ||dk − 1|,

where Ck = Sup{|dk − 1|, |δk − ζ|, 1}. If required, after passing to a subse-
quence, we have
dk → 1, δk → ζ.
Similarly,
ak → 1, ck → s, αk → ζ.
If t0 6= 0, this implies that Ak → A∗ , where A∗ ∈ Sp(n, 1). Thus the
group hA, Ts,ζ i can not be discrete.
Suppose t0 = 0. Then b = 0, γ = 0, β = 0. This implies that A fixes
the fixed point f1 = (1, 0, ..., 0)t of Ts,ζ .
This proves the theorem. 

4. Application of the main result

Now we will discuss some applications and comparison of the main result
we just proved.

Corollary 4.1. Let Ts,ζ be a Heisenberg translation in Sp(n, 1) and A ∈

Sp(n, 1). Suppose A(∞) 6= ∞ set M , t, as in the previous theorem. Suppose
hA, Ts,ζ i is non-elementary and discrete and

M t + 2|ζ| = 1

Consider Shimizu-Leutbecher sequence

A0 = A, Ak+1 = Ak Ts,ζ A−1

k

then for each k, hAk , Ts,ζ i is non-elementary discrete subgroup of Sp(n, 1)

and
M tk + 2|ζ| = 1 ∀ k.
In particular, tk = t for all k.
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 113

Proof. Note that

Ak Ts,ζ A−1
k (Ak (∞)) = Ak (∞)
So, the limit set of hAk , Ts,ζ i can form {Ak (∞), ∞}. Now, if hA1 , Ts,ζ i
is elementary, it must preserve {A(∞), ∞}. But Ts,ζ can not fix A(∞).
This implies that hA1 , Ts,ζ i is non-elementary. By induction, hAk , Ts,ζ i is
non-elementary. They are also discrete. So,

M tk + 2|ζ| ≥ 1.

Note that if M t0 + 2|ζ| = 1, then

M tk + 2|ζ| ≤ M t0 + 2|ζ| = 1.

⇒ M tk + 2|ζ| = 1 = M t + 2|ζ| ∀ k.
In particular, tk = t. 

Remark 4.2. Observe that theorem (4.8) of [KP03] says that if,
dH (A−1 (∞), Ts,ζ A−1 (∞))dH (A(∞), Ts,ζ A(∞)) + 4|ζ|2
1> 2
rA
then hTs,ζ , Ai is not discrete. Here dH is a Cygan metric and rA is the
q radius
2
of the isometric sphere of A. If A is of the form (2.6), then rA = |b| and
the above inequality becomes

1 1
1 > |b||s + ζ ∗ βb−1 − b̄−1 β ∗ ζ| 2 |s − ζ ∗ γ b̄−1 − b−1 γ ∗ ζ| 2 + 2|ζ|2 |b| (4.1)

Suppose now that

M t + 2|ζ| < 1
then
1 − 2|ζ|
t<
|s| + 2|ζ|
now right hand side of the (4.1) can be seen (thorough computations)
to be less than 1. Hence whenever our theorem holds, Kim and Parker’s
theorem also holds, accordingly the theorem of Cao and Parker also holds.
Therefore Theorem (3.1) is weaker than the theorem of Kim and Parker
[KP03], or Cao and Parker [CP11], though it is simpler.

Remark 4.3. Similarly one can show that our theorem implies Cao and
Parker’s Shimizu’s lemma for Heisenberg translation.
114 DEVENDRA TIWARI

We will conclude the paper by stating the main application of the result
(3.1). A subgroup G of Sp(n, 1) is called Zariski dense if it does not fix a
point on HnH ∪ ∂HnH , and neither it preserves a totally geodesic subspace
of HnH . With the above notations, we apply our main result to prove the
following part (3) of the following theorem in [GMT19]:

Theorem 4.4. [GMT19] Let G be a Zariski dense subgroup of Sp(n, 1).

(1) Let g ∈ Sp(n, 1) be a regular elliptic element such that δ(g) < 1. If
hg, hgh−1 i is discrete for every loxodromic element h ∈ G, then G
is discrete.
(2) Let g ∈ Sp(n, 1) be loxodromic element such that Mg < 1. If
hg, hgh−1 i is discrete for every loxodromic element h ∈ G, then
G is discrete.
(3) Let g ∈ Sp(n, 1) be a Heisenberg translation such that |ζ| < 12 . If
hg, hgh−1 i is discrete for every loxodromic h in G, then G is discrete.

Acknowledgement: The author thanks the anonymous referee for many

useful comments and suggestions, which helped improving the manuscript.
During the work the author was supported by NBHM postdoctoral fellow-
ship.

References
[BM98] A. Basmajian, and R. Miner, Discrete subgroups of complex hyperbolic motions.
Invent. Math. 131, 85-136 (1998).
[CP11] Wensheng Cao and John R. Parker. Jørgensen’s inequality and collars in n-
dimensional quaternionic hyperbolic space. Q. J. Math., 62(3):523–543, 2011.
[CP18] Wensheng Cao and John R. Parker. Shimizu’s lemma for quaternionic hyperbolic
space. Comput. Methods Funct. Theory, 18(1):159–191, 2018.
[CG74] S. S. Chen and L. Greenberg. Hyperbolic spaces. In Contributions to analysis (a
collection of papers dedicated to Lipman Bers), Academic Press, New York. pages
49–87, 1974.
[GMT19] Krishnendu Gongopadhyay, Mukund Madhav Mishra and Devendra Tiwari.
On Generalised Jørgensen Inequality in SL(2, C). Siberian Electronic Mathematical
Reports, Vol 16 (2019), 542–546.
[GMT19] Krishnendu Gongopadhyay, Mukund Madhav Mishra and Devendra Tiwari.
On discreteness of subgroups of quaternionic hyperbolic isometries. Bull. Aust.
Math. Soc. 101 (2020), no. 2, 283–293.
[GMT21] Krishnendu Gongopadhyay, Abhishek Mukherjee and Devendra Tiwari. Dis-
creteness of hyperbolic isometries by test maps. Osaka J. Math. 58 (2021), no. 3,
697-710.
 A SHIMIZU-TYPE THEOREM FOR SUBGROUPS IN Sp(n, 1) 115

[GMMR97] F. W. Gehring, C. Maclachlan, G. J. Martin and A. W. Reid. Arithmeticity,

discreteness and volume. Trans Am. Math. Soc 349 (1997).
[HP96] Sa’ar Hersonsky and Frédéric Paulin. On the volumes of complex hyperbolic
manifolds. Duke Math. J., 84(3):719–737, 1996.
[JP03] Y. P. Jiang and J. R. Parker, Uniform discreteness and Heisenberg screw motions.
Math. Z. 243, 653-669 (2003)
[Kam83] S. Kamiya, Notes on non-discrete subgroups of U (1, n; F). Hiroshima Math. J.
13, 501-506 (1983)
[Kim04] D. Kim, Discreteness criterions of isometric subgroups for quaternionic hyper-
bolic space. Geometriae Dedicata 106, 51-78 (2004).
[JKP03] Yueping Jiang, Shigeyasu Kamiya, and John R. Parker. Jørgensen’s inequality
for complex hyperbolic space. Geom. Dedicata, 97:55–80, 2003.
[Jor76] Troels Jørgensen. On discrete groups of Möbius transformations. Amer. J. Math.,
98(3):739–749, 1976.
[KP08] S. Kamiya, J. R. Parker, Discrete subgroups of P U (2, 1) with screw parabolic
elements. Math. Proc. Cambridge Philos. Soc. 144, 443-445 (2008).
[KP03] Inkang Kim and John R. Parker. Geometry of quaternionic hyperbolic manifolds.
Math. Proc. Cambridge Philos. Soc., 135(2):291–320, 2003.
[Par92] J. R. Parker, Shimizu’s lemma for complex hyperbolic space. Int. J. Math. 3,
291-308 (1992)
[Par97] J. R. Parker, Uniform discreteness and Heisenberg translations. Math. Zeit. 225,
485-505 (1997).
[Par98] J. R. Parker, On the volumes of cusped complex hyperbolic manifolds. and orb-
ifolds. Duke Math. J. 94, 433-464 (1998).
[Rag72] M. S. Raghunathan, Discrete subgroups of Lie groups. Ergebnisse der Math-
ematik und ihrer Grenzgebiete, Band 68. Springer-Verlag, New York-Heidelberg,
1972. ix+227 pp.
[Rat2006] John Ratcliffe, Foundations of hyperbolic manifolds, Second edition. Springer.
pp. xii+779, (2006).
[Shi63] H. Shimizu. On discontinuous subgroups operating on the product of the upper
half planes. Ann. of Math. 77 (1963), pp. 33-71.

Devendra Tiwari
Bhaskaracharya Pratishthana
56/14, Erandavane, Off Law College Road
Pune 411 004, India.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 116–131

CHARACTERIZATION OF ROUGH 3-PRIME IDEALS

AND 3-PRIME FUZZY IDEALS IN NEAR RINGS

ASMA ALI AND ARSHAD ZISHAN

(Received : 01 - 06 - 2023 ; Revised : 06 - 03 - 2024)

Abstract. Our aim in this paper is to introduce and study rough

prime ideals and semiprime ideals of a near ring N as a generalization of
rough prime and semiprime ideals in commutative rings. We also define
rough fuzzy ideal of N and carry out properties of rough 3-prime fuzzy
ideals. In addition, we examine the conditions under which the upper
and lower rough 3-prime ideals and upper and lower approximations of
their homomorphic images are related.

1. Introduction

The idea of fuzzy sets offers a useful mechanism for explaining the be-
haviour of systems that are either too complicated or too poorly specified
to allow for precise mathematical study using conventional tools and tech-
niques. It has demonstrated great potential for managing uncertainties to
a manageable degree, particularly in decision-making models under various
types of risks, subjective assessment, ambiguity, and vagueness. Expert sys-
tems, control systems, pattern recognition, and image processing are just a
few of the many disciplines in which this idea has already been widely ap-
plied. Zadeh [8] was the first to present the theory of fuzzy sets. After that
Rosenfeld [1] and Zaid [17] introduced and studied fuzzy group and fuzzy
ring respectively. In [5, 7, 10, 16, 19, 20, 21, 22], researchers introduced and
studied fuzzy prime, semiprime, maximal and radical of a fuzzy ideal of a
ring R.
Pawlak [27] was the first to present the theory of rough sets. Pawlak in-
troduced rough set theory as a potent mathematical tool for handling data
2010 Mathematics Subject Classification: 03E72, 16N60, 16W25, 16Y30.
Key words and phrases: 3-prime ideals, 3-semiprime ideals, rough 3-prime ideals, fuzzy
ideals, rough 3-prime fuzzy ideal

© Indian Mathematical Society,

2024 .
116
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 117

uncertainty. Rough set theory begins with the premise that components of
a universe with the same characterization are indistinguishable in light of
the information at hand. An extension of set theory known as rough set
theory describes a subset of the universe as being defined by two classical
sets known as the lower and upper approximations. The foundation for cre-
ating the lower and the upper approximations are the equivalence classes
[23, 26]. For basic ideas without clear boundaries, such as those that are
hazy and ambiguous, rough sets offer an appropriate mathematical repre-
sentation. Rough set theory is becoming into a potent tool for working with
unreliable data. Recently, it has drawn a lot of attention from the research
community in both the theory and practical applications. Rough subgroups
and rough ideals in semigroup discussed by Biswas et al. [15] and Kuroki
[11] respectively. Some characteristics of the lower and upper approxima-
tions with regard to the normal subgroups were represented by Kuroki and
Wang[12]. Kazanci and Davvaz [13] brought up the ideas of rough prime
and rough fuzzy prime ideals of commutative rings in 2008.
In this paper we define rough fuzzy 3-prime and 3-semiprime ideals in a
near ring N . We also discuss some of their features. This in-depth investi-
gation could perhaps offer a useful tool for approximative reasoning, in our
opinion. When it comes to the theory as well as applications of fuzzy sets
and rough sets, we think the rough near rings presented here will be more
beneficial. As a generalisation of rings, Pilz[6] proposed the idea of near
rings in 1983. In which, only one distributive law is required, hence the
addition operation is not required to be commutative. It would be unfair to
characterise near ring theory as merely a collection of unimportant findings
pertaining to particular pathological systems with little relevance to other
areas of mathematics. In contrast to applications in axioms and geometry,
novel and extremely effective classes of balanced incomplete block designs
are provided by particular classes of finite near rings (finite planar near
ring).

2. Preliminaries

Definition 2.1. A full congruence relation (F.C.R.) R on a near ring N is

an equivalence relation that preserves the algebraic operations on that set
i.e., for p, q ∈ N , (p, q) ∈ R implies that (p + r, q + r), (r + p, r + q), (pr, qr)
and (rp, rq) ∈ R, for all r ∈ N .
118 ASMA ALI AND ARSHAD ZISHAN

Lemma 2.2. If R is F.C.R. on a near ring N , then (p, q), (r, s) ∈ R imply
(p + r, q + s), (pr, qs) and (−p, −q) ∈ R.

Proof. Consider R is a F.C.R. on a near ring N and (p, q), (r, s) ∈ R. Then
(p + r, q + r), (q + r, q + s) ∈ R. By definition of F.C.R. (p + r, q + s) ∈ R.
In similar manner, we can see that (pr, qs), (−p, −q) ∈ R.


Lemma 2.3. Let R be F.C.R. on a near ring N . If m, n ∈ N , then

(i) [m]R + [n]R = [m + n]R
(ii) [−m]R = −[m]R
(iii) {ab | a ∈ [m]R , b ∈ [n]R } ⊆ [mn]R .

Proof. (i) Let x = p+q ∈ [m]R +[n]R , p ∈ [m]R , q ∈ [n]R . Then (p, m), (q, n) ∈
R. By lemma 2.2, (p + q, m + n) ∈ R i.e., p + q = x ∈ [m + n]R . Now, take
z ∈ [m + n]R i.e., (z, m + n) ∈ R or (−m + z, n) ∈ R i.e., (−m + z) ∈ [n]R
entails that z ∈ [m]R + [n]R . Thus [m]R + [n]R = [m + n]R .
(ii) For any a ∈ [−m]R or equivalently

(a, −m) ∈ R ⇐⇒ (0, −m − a) ∈ R

⇐⇒ (m, −a) ∈ R
⇐⇒ −a ∈ [m]R
⇐⇒ a ∈ −[m]R .

This show that [−m]R = −[m]R .

(iii) Let c = ab such that a ∈ [m]R , b ∈ [n]R . Then (a, m) ∈ R and
(b, n) ∈ R. Since R is F.C.R., hence (ab, mn) ∈ R i.e., ab = c ∈ [mn]R . This
implies that [m]R [n]R ⊆ [mn]R . 

Definition 2.4. A F.C.R. R on N is said to be complete (C.C.R.) if

[mn]R = {pq | p ∈ [m]R , q ∈ [n]R }, for all p, q ∈ N .

Definition 2.5. Let R be a F.C.R. on a near ring N and S ⊆ N . Then the

sets R− (S) = {n ∈ N | [n]R ⊆ S} and R − (S) = {n ∈ N | [n]R ∩ S =6 φ}
are referred to R-lower and R-upper approximation of the set S.

R(S) = (R− (S), R − (S)) is called a rough set with respect to R if

R− (S) 6= R − (S). S is said to be upper rough ideal if R − (S) is an ideal of
N.
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 119

Lemma 2.6. Let R be a F.C.R. on N . Then R − (J ) is an ideal of N , if

J is an ideal of N .

Proof. Let x, y ∈ R − (J ) and n ∈ N . Then [x]R ∩ J 6= φ and [y]R ∩ J 6= φ.

So, there exist u ∈ [x]R ∩J and v ∈ [y]R ∩J . This implies that u−v ∈ J and
nu ∈ J . By lemma 2.3, u−v ∈ [x]R −[y]R = [x−y]R . Thus u−v ∈ [x−y]R
i.e., [x − y]R ∩ J =
6 φ. Which implies that

x − y ∈ R − (J ). (2.1)

Since (u, x) ∈ R, hence (nu, nx) ∈ R. Thus nu ∈ [nx]R also nu ∈ J . So,

[nx]R ∩ J =6 φ. Which implies that

nx ∈ R − (J ). (2.2)

Moreover, n + x − n ∈ R − (J ). Since x ∈ R − (J ), then [x]R ∩ J 6= φ. So,

there exist w ∈ [x]R and w ∈ J . Since J is an ideal and R is F.C.R., hence
n − w + n ∈ [n − x + n]R and n − w + n ∈ J . Thus [n − x + n]R ∩ J = 6 φ or

n − x + n ∈ R − (J ). (2.3)

Consider n1 , n2 ∈ N and i ∈ R − (J ). Then [i]R ∩J = 6 φ. So, there exist k ∈

[i]R ∩J i.e., (i, k) ∈ R and k ∈ J . Since J is an ideal of N and R is F.C.R.,
hence ((n1 +i)n2 −n1 n2 , (n1 +k)n2 −n1 n2 ) ∈ R and (n1 +k)n2 −n1 n2 ∈ N .
Thus (n1 + k)n2 − n1 n2 ∈ [(n1 + i)n2 − n1 n2 ]R ∩ J . Which implies that
[(n1 + i)n2 − n1 n2 ]R ∩ J =6 φ. Therefore

(n1 + i)n2 − n1 n2 ∈ R − (J ). (2.4)

Equations (2.1)-(2.4) show that R − (J ) is an ideal of N . 

Remark 2.7. The converse statement of the previous lemma is not true.

Example 2.8. Consider (N , +, ·) be a near ring under following operations

+ 0 1 2 3 · 0 1 2 3
0 0 1 2 3 0 0 0 0 0
1 1 0 3 2 and 1 0 1 2 3
2 2 3 0 1 2 0 2 0 0
3 3 2 1 0 3 0 3 2 3.
Let R be a F.C.R. on N with equivalence classes [0]R = {0, 2} and [1]R =
{1, 3}. For J = {0, 1, 3}, we can see that J is not an ideal of N . But,
R − (J ) = N is an ideal of N .
120 ASMA ALI AND ARSHAD ZISHAN

Lemma 2.9. Let R be a F.C.R. on a near ring N and J be an ideal of N .

If R− (J ) is a nonempty set, then R− (J ) = J .
Proof. Assume that R− (J ) is nonempty set. Then there exists j ∈ R− (J ).
Clearly, we can see that R− (J ) ⊆ J . Now, we will only show that J ⊆
R− (J ). Suppose that i, j ∈ J . Then [0]R = [j + (−j)]R = [j]R + [−j]R ⊆
J + J = J and

j ∈ i + [0]R ⇐⇒ j − i ∈ [0]R
⇐⇒ (j − i, 0) ∈ R
⇐⇒ (j, i) ∈ R
j ∈ [i]R .

We get [i]R ⊆ J . This implies that i ∈ R− (J ). 

Definition 2.10. Let X be a subset of a near ring N and (R− (X ), R − (X ))
a rough set. If R− (X ) and R − (X ) are ideals of N , then we call (R− (X ), R − (X ))
a rough ideal.
Corollary 2.11. If J is an ideal of N and R− (J ) is a nonempty set, then
(R− (J ), R − (J )) is a rough ideal.
Proof. From lemmas 2.6 and 2.9. 
Corollary 2.12. If J and K are ideals of N such that R− (J ∩ K) 6= φ,
then (R− (J ∩ K), R − (J ∩ K)) is a rough ideal of N .
Proof. From corollary 2.11, (R− (J ∩ K), R − (J ∩ K)) is a rough ideal of
N. 
Proposition 2.13. Let f : N1 → N2 be an epimorphism of a near ring N1
to a near ring N2 . Then
(i) R1 = {(n1 , n2 ) ∈ N1 × N2 | (f (n1 ), f (n2 )) ∈ R2 } is F.C.R. on N1 ,
if R2 is F.C.R.
(ii) If R2 is complete and f is injective, then R1 is complete.
(iii) For any subset X of N1 , f (R1− (X )) = R2− (f (X )).
(iv) f (R1− (X )) ⊆ R2− (f (X )). Equality holds if f is injective.
Proof. (i) Let (n1 , n2 ) ∈ R1 and n ∈ N1 and n ∈ N1 . Then (f (n1 ), f (n2 )) ∈
R2 . Since R2 is F.C.R., hence (f (n)+f (n1 ), f (n)+f (n2 )), (f (n1 )+f (n), f (n2 )+
f (n)), (f (n)f (n1 ), f (n)f (n2 )) (f (n1 )f (n), f (n2 )f (n)) ∈ R2 . As f is homo-
morphism, (f (n + n1 ), f (n + n2 )), (f (n1 + n), f (n2 + n)),(f (nn1 ), f (nn2 )),
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 121

(f (n1 n), f (n2 n)) ∈ R2 . This implies that (n + n1 , n + n2 ), (n1 + n, n2 +

n),(nn1 , nn2 ), (n1 n, n2 n) ∈ R1 .

(ii) Let l ∈ [mn]R1 . Then (l, mn) ∈ R1 . By (i), (f (l), f (mn) ∈ R2 . So,

f (l) ∈ [f (mn)]R2 = [f (m)f (n)]R2

0 0 0 0
= {f (m )f (n ) | f (m ) ∈ [f (m)]R2 , f (n ) ∈ [f (n)]R2 }.

Thus there exist a, b ∈ N1 such that f (l) = f (a)f (b) = f (ab) and f (a) ∈
[f (m)]R2 , f (b) ∈ [f (n)]R2 . f (l) = f (a)f (b) implies that l = ab, since f is
injective and a ∈ [m]R1 , b ∈ [n]R1 . Hence

l ∈ {ab | a ∈ [m]R1 , b ∈ [n]R1 }. (2.5)

From equation 2.5 and (iii) part of lemma 2.3, equality holds.

(iii) Let z ∈ f (R1− (X )). Then one can find y ∈ R1− (X )) so that f (y) = z.
Since [y]R1 ∩ X 6= φ, hence there exists x ∈ [y]R1 ∩ X , which implies that
(x, y) ∈ R1 and x ∈ X or (f (x), f (y)) ∈ R2 . So, [f (y)]R2 ∩f (X ) 6= φ implies
that z = f (y) ∈ R2− (f (X )) i.e.,

f (R1− (X )) ⊆ R2− (f (X )) (2.6)

Conversely, assume that a ∈ R2− (f (X )). Then there is b ∈ N1 so that

f (b) = a. So, [f (b)]R2 ∩ f (X ) 6= φ. Therefore, one can find c ∈ X in a way
that f (c) ∈ f (X ), f (c) ∈ [f (b)]R2 and c ∈ [b]R1 . Thus, [b]R1 ∩ X =
6 φ implies
− −
that b ∈ R1 (X ). So, a = f (b) ∈ f (R1 (X )) i.e.,

R2− (f (X )) ⊆ f (R1− (X )). (2.7)

From equations (2.6) and (2.7), f (R1− (X )) = R2− (f (X )).

(iv) Suppose that a ∈ f (R1− (X )), then there is b ∈ R1− (X ) in such

a manner that f (b) = a. So, we have [b]R1 ⊆ X . Now, we assume that
c ∈ [a]R2 , then there is d ∈ N1 so that f (d) = c and f (d) ∈ [f (b)]R2 . Hence,
d ∈ [b]R1 ⊆ X and so c = f (d) ∈ f (X ). Thus [a]R2 ⊆ f (X ) which yields
that a ∈ R2− (f (X )). This demonstrate that

f (R1− (X )) ⊆ R2− (f (X )). (2.8)

122 ASMA ALI AND ARSHAD ZISHAN

Now assume that f is one to one and a ∈ R2− (f (X )), then there exists
b ∈ N1 such that f (b) = a and [f (b)]R2 ⊆ f (X ). Suppose c ∈ [b]R1 , then
f (c) ∈ [f (b)]R2 ⊆ f (X ) and so c ∈ X . Thus [b]R1 ⊆ X , which implies that
b ∈ R1− (X ). Then a = f (b) ∈ f (R1− (X )), so,

R2− (f (X )) ⊆ f (R1− (X )). (2.9)

From equations (2.8) and (2.9), f (R1− (X )) = R2− (f (X )). 

3. Rough 3-prime ideals

Definition 3.1. An ideal J in a near ring N is 3-prime if a, b ∈ N and

aN b ⊆ J implies that a ∈ J or b ∈ J .

Definition 3.2. Let R be a C.R. on a near ring N . Then a subset X of N

is called a lower rough 3-prime ideal of N , if R− (X ) is 3-prime ideal of N .

In similar manner, upper rough 3-prime ideal can be defined.

Definition 3.3. An ideal J in near ring N is 3-semiprime if a ∈ N and

aN a ⊆ J implies that a ∈ J .

Definition 3.4. An ideal J of a near ring N is rough 3-prime (semiprime)

ideal of N if it is both a lower and an upper rough 3-prime (semiprime)
ideal of N .

Lemma 3.5. Let R be C.C.R. on near ring N and L is a 3-prime ideal of

N such that R − (L) 6= N . Then L is an upper rough prime ideal of N .

Proof. By lemma2.6, R − (L) is an ideal. Now, we will only show that R − (L)
is 3-prime. Assume that a, b ∈ N such that anb ∈ R − (L), for all n ∈ N .
Then [anb]R ∩ L 6= φ. Since R is complete, hence {xmy | x ∈ [a]R , m ∈
[n]R , y ∈ [b]R }∩L 6= φ. Suppose that there exist x ∈ [a]R , m ∈ [n]R , y ∈ [b]R
such that xmy ∈ [anb]R ∩ L. Since L is 3-prime, hence xmy ∈ L implies
that x ∈ L or y ∈ L. Therefore, [a]R ∩ L 6= φ or [b]R ∩ L 6= φ. This implies
that a ∈ R − (L) or b ∈ R − (L). 

Lemma 3.6. Let R be C.C.R. on a near ring N and L is a 3-prime ideal

of N such that R− (L) 6= N . Then L is a lower rough prime ideal of N .

Proof. Direct from lemma 2.9. 

 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 123

Lemma 3.7. Let R be a C.C.R. on a near ring N and J be a 3-semiprime

ideal of N such that R − (J ) 6= N . Then J is an upper rough 3-semiprime
ideal of N .

Proof. Let a ∈ N such that ana ∈ R − (J ), for all n ∈ N . Then [ana]R ∩J 6=

φ. Since R is C.C.R., hence {xmx | x ∈ [a]R , m ∈ [n]R }. Thus there exist
x ∈ [a]R , m ∈ [n]R such that xmx ∈ J . As J is 3-semiprime, x ∈ J . This
6 φ i.e., a ∈ R − (J ).
implies thta [a]R ∩ J = 

Lemma 3.8. Let R be a C.C.R. on N and J a 3-semiprime ideal of N

such that R− (J ) 6= N . Then J is a lower rough 3-semiprime ideal of N .

Proof. Direct from lemma 2.9 

Theorem 3.9. Let f : N1 → N2 be an epimorphism from a near ring N1

to a near ring N2 and R2 be a F.C.R. on N2 . Let X be a subset of N1 . If
R1 = {(n1 , n2 ) ∈ N1 × N2 | (f (n1 ), f (n2 )) ∈ R2 }, then
(i) R1 − (X ) is an ideal of N1 if and only if R2 − (f (X )) is an ideal of
N2 .
(ii) R1 − (X ) is 3-prime ideal of N1 if and only if R2 − (f (X )) is 3-prime
ideal of N2 , provided R2 is complete.
(iii) R1 − (X ) is 3-semiprime ideal of N1 if and only if R2 − (f (X )) is
3-semiprime ideal of N2 , provided R2 is complete.

Proof. (i) Suppose that R2 − (f (X )) is an ideal of N2 . Let x1 , x2 ∈ R1 − (X ),

n ∈ N1 . Then f (x1 ), f (x2 ) ∈ f (R1 − (X )), f (n) ∈ N2 . By proposition
2.13, f (x1 ), f (x2 ) ∈ R2 − (f (X )). As R2 − (f (X )) is an ideal of N2 , f (x1 ) −
f (x2 ), f (n) − f (x1 ) + f (n), f (n)f (x1 ) ∈ R2 − (f (X )), also for f (n1 ) ∈ N2 ,
(f (n) + f (x1 ))f (n1 ) − f (n)f (n1 ) ∈ R2 − (f (X )). Since f is homomorphism,
hence f (x1 − x2 ), f (n − x1 + n), f (nx1 ), f ((n + x1 )n1 − nn1 ) ∈ R2 − (f (X ))
or f (x1 − x2 ), f (n − x1 + n), f (nx1 ), f ((n + x1 )n1 − nn1 ) ∈ f (R1 − (X )). So,
there exist a, b, c, d ∈ R1 − (X ) such that f (x1 − x2 ) = f (a), f (n − x1 + n) =
f (b), f (nx1 ) = f (c), f ((n + x1 )n1 − nn1 ) = f (d) and we get [a]R1 ∩ X = 6
φ, [b]R1 ∩X = 6 φ, [c]R1 ∩X = 6 φ, [d]R1 ∩X = 6 φ and x1 −x2 ∈ [a]R1 , n−x1 +n ∈
[b]R1 , nx1 ∈ [c]R1 , (n + x1 )n1 − nn1 ∈ [d]R1 . Thus [x1 − x2 ]R1 ∩ X 6=
φ, [n − x1 + n]R1 ∩ X 6= φ, [nx1 ]R1 ∩ X 6= φ, [(n + x1 )n1 − nn1 ]R1 ∩ X = 6 φ.
This implies that x1 − x2 , n − x1 + n, nx1 , (n + x1 )n − nn1 ∈ R1 − (X ). Thus
R1 − (X ) is an ideal of N1 .
Conversely, suppose that R1 − (X ) is an ideal of N1 . Let y1 , y2 ∈ R2 − (f (X ))
124 ASMA ALI AND ARSHAD ZISHAN

0
and n2 = f (n2 ) ∈ N2 . Given that f is epimorphism, so by proposition 2.13,
y1 , y2 ∈ R2 − (f (X )) = f (R1 − (X )) i.e., there exist a, b ∈ R1 − (X ) such that
0 0
y1 = f (a), y2 = f (b). Then y1 −y2 = f (a−b), n2 y1 = f (n2 )f (a) = f (n2 a),
0 0 0 0 0 0
n2 +y1 −n2 = f (n2 +a−n2 ), and ((n+y1 )n2 −nn2 ) = f ((n +a)n2 −n n2 ).
0 0 0 0 0
Since R1 − (X ) is an ideal of N1 , hence a − b, n2 a, n2 + a − n2 , (n + a)n2 −
0 0
n n2 ∈ R1 − (X ). This show that R2 − (f (X )) is an ideal of N2 .

(ii) Suppose that R2 − (f (X )) is a 3-prime ideal of N2 and for a, b ∈ N1 ,

an1 b ∈ R1 − (X ), for all n1 ∈ N1 . Then proposition 2.13, f (an1 b) ∈ f (R1 − (X ))
i.e., f (a)f (n1 )f (b) ∈ R2 − (f (X )). As R2 − (f (X )) is 3-prime ideal, f (a) ∈
R2 − (f (X )) or f (b) ∈ R2 − (f (X )) i.e., f (a) ∈ f (R1 − (X )) or f (b) ∈ f (R1 − (X )).
Then there exist x, y ∈ R1 − (X ) such that f (a) = f (x) and f (b) = f (y).
Thus [x]R1 ∩ X = 6 φ and a ∈ [x]R1 or [y]R1 ∩ X = 6 φ and b ∈ [y]R1 . Which
− −
implies that a ∈ R1 (X ) or b ∈ R1 (X ).
Conversely, suppose that R1 − (X ) is a 3-prime ideal of N1 . Let a, b ∈ N2 such
0 00 0
that an2 b ∈ R2 − (f (X )), for all n2 ∈ N2 . Then there exist n1 , n1 , n2 ∈ N1
0 00 0 0 0 00
such that f (n1 ) = a, f (n1 ) = b and f (n2 ) = n2 . So, f (n1 )f (n2 )f (n1 ) ∈
0 0 00 0 0 00
R2 − (f (X )) or f (n1 n2 n1 ) ∈ R2 − (f (X )). Thus [f (n1 )f (n2 )f (n1 )]R2 ∩
0 0
f (X ) 6= φ. Since R2 is complete, hence there exist f (m1 ) ∈ [f (n1 )]R2 ,
0 0 00 00 0 0 00
f (m2 ) ∈ [f (n2 )]R2 , f (m1 ) ∈ [f (n1 )]R2 such that f (m1 )f (m2 )f (m1 ) =
0 0 00 0 0 0
f (m1 m2 m1 ) ∈ f (X ). Then by proposition 2.13, m1 ∈ [n1 ]R1 , m2 ∈
0 00 00 0 0 00
[n2 ]R1 , m1 ∈ [n1 ]R1 and there exist z ∈ X such that f (m1 m2 m1 ) =
0 0 00 0 0 00 0 0 00
f (z). Hence m1 m2 m1 ∈ [n1 n2 n1 ]R1 and z ∈ [m1 m2 m1 ]R1 or z ∈
0 0 00 0 0 00 0 0 00
[n1 n2 n1 ]R1 . Thus [n1 n2 n1 ] ∩ X 6= φ. This implies that n1 n2 n1 ∈
R1 − (X ). Since R1 − (X ) is 3-prime ideal, hence
0 00
n1 ∈ R1 − (X ) or n1 ∈ R1 − (X ). (3.1)

By equation 3.1 and proposition 2.13,

0
a = f (n1 ) ∈ f (R1 − (X )) = R2 − (f (X ))
or
00
b = f (n1 ) ∈ f (R1 − (X )) = R2 − (f (X )).

This show that R2 − (f (X )) is 3-prime ideal.

(iii) Proof running parallel to (ii) 
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 125

Theorem 3.10. Let f : N1 → N2 be an isomorphism from a near ring N1

to a near ring N2 and R2 be a C.C.R. on N2 . Let X be a subset of N1 . If
R1 = {(n1 , n2 ) ∈ N1 × N1 | (f (n1 ), f (n2 )) ∈ R2 }, then
(i) R1− (X ) is an ideal of N1 if and only if R2− (f (X )) is an ideal of
N2 .
(ii) R1− (X ) is 3-prime ideal of N1 if and only if R2− (f (X )) is 3-prime
ideal of N2 , provided R2 is complete.
(iii) R1− (X ) is 3-semiprime ideal of N1 if and only if R2− (f (X )) is
3-semiprime ideal of N2 , provided R2 is complete.

Proof. Firstly, from proposition 2.13, we get f (R1− (X )) = R2− (f (X )).

After that proof is running parallel to theorem 3.9. 

4. Rough 3-prime fuzzy ideals

Definition 4.1. Let R be a F.C.R. on a near ring N and ζ be fuzzy ideal

(F.I.) of N . Then we define the fuzzy sets R− (ζ) and R − (ζ) as follow:
R− (ζ)(n) = inf {ζ(a)} and R − (ζ)(n) = sup {ζ(a)}.
a∈[n]R b∈[n]R

R− (ζ) and R − (ζ) are known as R-lower and R-upper approximation of

fuzzy set ζ.

Definition 4.2. Let R− (ζ) and R − (ζ) are R-lower and R-upper approx-
imation of a fuzzy ideal ζ. Then R(ζ) = (R− (ζ), R − (ζ)) is said to be a
rough fuzzy set (R.F.S) if R− (ζ) 6= R − (ζ).

Definition 4.3. Let ζ be a fuzzy set of a near ring N . Then ζ is called

upper rough F.I., if R − (ζ) is a F.I. of N .

Definition 4.4. A F.I. ζ of N is said to be 3-prime if for n1 , n2 ∈ N ,

∨{ζ(n1 ), ζ(n2 )} ≥ ∧ ζ(n1 nn2 ).
n∈N

Definition 4.5. A rough fuzzy set (R− (ζ), R − (ζ)) is said to be a rough
F.I., if R− (ζ) and R − (ζ) is a F.I. of a near ring N .

Example 4.6. Consider (N , +, ·) be a near ring under following operations

126 ASMA ALI AND ARSHAD ZISHAN

+ 0 1 2 3 4 5 6 7 + 0 1 2 3 4 5 6 7
0 0 1 2 3 4 5 6 7 0 0 0 0 0 0 0 0 0
1 1 2 3 0 5 6 7 4 1 0 1 0 1 0 1 1 0
2 2 3 0 1 6 7 4 5 2 0 2 0 2 0 2 2 0
3 3 0 1 2 7 4 5 6 and 3 0 3 0 3 0 3 3 0
4 4 7 6 5 0 3 2 1 4 4 4 4 4 4 4 4 4
5 5 4 7 6 1 0 3 2 5 4 5 4 5 4 5 5 4
6 6 5 4 7 2 1 0 3 6 4 6 4 6 4 6 6 4
7 7 6 5 4 3 2 1 0 7 4 7 4 7 4 7 7 4

Let R be a F.C.R. on N with equivalence classes [0]R = {0, 2, 4, 6} and

[1]R = {1, 3, 5, 7}. Take 3-prime F.I. ζ on N which is defined by ζ(0) =
ζ(1) = ζ(2) = ζ(3) = 0.1, ζ(4) = ζ(5) = ζ(6) = ζ(7) = 0.2. We can easily
find that R− (ζ)(a) = 0.1 6= 0.2 = R − (ζ)(b) for all a, b ∈ N . Since R − (ζ)
and R− (ζ) are F.I., hence rough fuzzy set R(ζ) is rough F.I..

Lemma 4.7. Let R be a F.C.R. on a near ring N . If ζ is a F.I. of N , then

(i) R − (ζ) is a F.I. of N , and

(ii) R− (ζ) is a F.I. of N .

Proof. (i) For i, j ∈ N ,

R − (ζ)(i − j) = sup {ζ(k)} = sup {ζ(k)}

k∈[i−j]R k∈[i]R −[j]R

= sup {ζ(k1 − k2 )} ≥ sup {ζ(k1 ) ∧ ζ(k2 )}

k1 ∈[i]R ,k2 ∈[j]R k1 ∈[i]R ,k2 ∈[j]R

= sup {ζ(k1 )} ∧ sup {ζ(k2 )}

k1 ∈[i]R k2 ∈[j]R
− −
= R (ζ)(i) ∧ R (ζ)(j). (4.1)

R − (ζ)(ij) = sup {ζ(k)} ≥ sup {ζ(k1 k2 )}

k∈[ij]R k1 ∈[i]R ,k2 ∈[j]R

= sup {ζ(k1 )} ∧ sup {ζ(k2 )}

k1 ∈[i]R k2 ∈[j]R
− −
= R (ζ)(i) ∧ R (ζ)(j). (4.2)
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 127

R − (ζ)(j + i − j) = sup {ζ(k)} = sup {ζ(k2 + k1 − k2 )}

k∈[j+i−j]R k1 ∈[i]R ,k2 ∈[j]R

= sup {ζ(k1 )}
k1 ∈[i]R

= R − (ζ)(i). (4.3)

R − (ζ)(ij) = sup {ζ(k)} ≥ sup {ζ(k1 k2 )}

k∈[ij]R k1 ∈[i]R ,k2 ∈[j]R

≥ sup {ζ(k2 )} = R − (ζ)(j). (4.4)

k2 ∈[j]R

R − (ζ)((i + l)j − ij) = sup {ζ(k)} = sup {ζ(k)}

k∈[(i+l)j−ij]R k∈[(i+l)j]R −[ij]R

≥ sup {ζ((k1 + k2 )k3 − k1 k2 )}

k1 ∈[i]R ,k2 ∈[l]R ,k3 ∈[j]R

≥ sup {ζ(k2 )} = R − (ζ)(l). (4.5)

k2 ∈[l]R

Equations (4.1)-(4.5) show that R − (ζ) is a F.I. of N .

(ii) For i, j ∈ N ,

R− (ζ)(i − j) = inf {ζ(k)} = inf {ζ(k)}

k∈[i−j]R k∈[i]R −[j]R

≥ inf {ζ(k1 ) ∧ ζ(k2 )}

k1 ∈[i]R ,k2 ∈[j]R

= R− (ζ)(i) ∧ R− (ζ)(j). (4.6)

R− (ζ)(ij) = inf {ζ(k)} ≥ inf {ζ(k1 k2 )}

k∈[ij]R k1 ∈[i]R ,k2 ∈[j]R

= inf {ζ(k1 )} ∧ inf {ζ(k2 )}

k1 ∈[i]R k2 ∈[j]R

= R− (ζ)(i) ∧ R− (ζ)(j). (4.7)

R− (ζ)(j + i − j) = inf {ζ(k)} = inf {ζ(k2 + k1 − k2 )}

k∈[j+i−j]R k1 ∈[i]R ,k2 ∈[j]R

= inf {ζ(k1 )} = R− (ζ)(i). (4.8)

k1 ∈[i]R
128 ASMA ALI AND ARSHAD ZISHAN

R − (ζ)(ij) = inf {ζ(k)} ≥ inf {ζ(k1 k2 )}

k∈[ij]R k1 ∈[i]R ,k2 ∈[j]R

≥ inf {ζ(k2 )} = R− (ζ)(j). (4.9)

k2 ∈[j]R

R− (ζ)((i + l)j − ij) = inf {ζ(k)} = inf {ζ(k)}

k∈[(i+l)j−ij]R k∈[(i+l)j]R −[ij]R

≥ inf {ζ((k1 + k2 )k3 − k1 k2 )}

k1 ∈[i]R ,k2 ∈[l]R ,k3 ∈[j]R

≥ inf {ζ(k2 )} = R− (ζ)(l). (4.10)

k2 ∈[l]R

Equations (4.6)-(4.10) show that R − (ζ) is a F.I. of N . 

Corollary 4.8. A rough fuzzy set (R− (ζ), R − (ζ)) is F.I. if ζ is F.I..

Proof. By lemma 4.7, (R− (ζ), R − (ζ)) is F.I.. 

Theorem 4.9. Let ζ be a 3-prime F.I. of a near ring N . If R is C.C.R.

on N , then

(i) R− (ζ) is 3-prime F.I. of N .

(ii) R − (ζ) is 3-prime F.I. of N .

Proof. By lemma 4.7, R− (ζ) is F.I. of N . Now, we will only show that
3-primeness of R− (ζ). Let a, b ∈ N . Then

inf R− (ζ)(anb) = inf { ∧ ζ(z)}

n∈N n∈N z∈[anb]R

= inf { ∧ ζ(z1 mz2 )}

n∈N z1 ∈[a]R ,m∈[n]R ,z2 ∈[b]R

= ∧ { inf ζ(z1 mz2 )}

z1 ∈[a]R ,m∈[n]R ,z2 ∈[b]R m∈N

≤ ∧ {max(ζ(z1 ), ζ(z2 ))}

z1 ∈[a]R ,z2 ∈[b]R

= max{( ∧ ζ(z1 ), ∧ ζ(z2 ))}

z1 ∈[a]R z2 ∈[b]R

= max{R− (ζ)(a), R− (ζ)(b)}. (4.11)

 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 129

This shows that R− (ζ) is 3-prime F.I. of N .

(ii) By lemma 4.7, R − (ζ) is F.I. of N . Now, we will only show that 3-
primeness of R − (ζ). Let a, b ∈ N . Then

inf R − (ζ)(anb) = inf { ∨ ζ(z)}

n∈N n∈N z∈[anb]R

= inf { ∨ ζ(z1 mz2 )}

n∈N z1 ∈[a]R ,m∈[n]R ,z2 ∈[b]R

= ∨ { inf ζ(z1 mz2 )}

z1 ∈[a]R ,m∈[n]R ,z2 ∈[b]R m∈N

≤ ∨ {max(ζ(z1 ), ζ(z2 ))}

z1 ∈[a]R ,z2 ∈[b]R

= max{( ∨ ζ(z1 ), ∨ ζ(z2 ))}

z1 ∈[a]R z2 ∈[b]R

= max{R − (ζ)(a), R − (ζ)(b)}. (4.12)

This shows that R − (ζ) is 3-prime F.I. 

5. Conclusion

Rough fuzzy ideals give a useful and effective tool for modeling complex
systems with ambiguity, imprecision, and uncertainty. The rough fuzzy
ideal of a near ring is a generalization of an ideal of near ring. So, we
replaced a universe set by a near ring and introduced the notion of rough
3-prime (3-semiprime) ideals and rough fuzzy 3-prime ideal of near rings.
We discussed the conditions under which the upper and lower rough 3-prime
ideals and upper and lower approximations of their homomorphic images
are related.
Acknowledgement: We are grateful to the referee for the comments which
improved the quality of the paper.

References
[1] Rosenfeld, A., Fuzzy Groups, J. Math. Anal. Appl. 35 (1971), 512-517.
[2] Kedukodi, B.S.; Bhavanari, S.; and Kuncham, S.P. C-prime fuzzy ideals of near
rings, Soochow J. Mathematics 33 (2007), 891-901.
[3] Davvaz, B., Roughness in Rings, Inform. Sci. 164 (2004), 147-163.
[4] Davvaz, B., Roughness based on fuzzy ideals, Inform.Sci. 176 (2006), 2417-2437.
[5] Malik, D.S. and Mordeson, J.N. Fuzzy maximal,radical, and primary ideals of rings,
Inform. Sci. 53 (1991), 237-250.
[6] Pilz, G., Near-rings, Revised edn. North Holland (1983).
[7] Kumbhojkar, H.V. and Bapat, M.S., On prime and primary fuzzy ideals and their
radicals, Fuzzy Sets and Systems 53 (1993), 203-216.
130 ASMA ALI AND ARSHAD ZISHAN

[8] Zadeh, L.A. Fuzzy Sets, Information and Control 8 (1965), 338-353.
[9] Kondo, M. On the structure of generalized rough sets, Inform. Sci. 176 (2006),
589–600.
[10] Zahedi, M.M. A note on L-fuzzy primary and semiprime fuzzy ideals, Fuzzy Sets
and Systems 51 (1992),243-247.
[11] Kuroki, N. Rough ideals in semigroups, Inform. Sci. 100 (1997), 139–163.
[12] Kuroki, N.; . Wang, P.P. The lower and upper approximations in a fuzzy group,
Inform. Sci. 90 (1996) 203–220.
[13] Kazanci,O.; Davvaz, B. On the structure of rough prime (primary) ideals and
rough fuzzy prime (primary) ideals in commutative rings, Inform. Sci. 178 (2008),
1343–1354.
[14] Xiao, Q.M.; Zhang, Z.L. Rough prime ideals and rough fuzzy prime ideals in semi-
groups, Inform. Sci. 176(2006), 725–733.
[15] Biswas, R.; Nanda, S. Rough groups and rough subgroups, Bull. Polish Acad. Sci.
Math. 42 (1994), 251-254.
[16] Kumar, R., Fuzzy nil radicals and fuzzy primary ideals, Fuzzy Sets and Systems 43
(1991),81-93.
[17] Zaid, S.A. On Fuzzy Subnear-Rings and Ideals, Fuzzy Sets and Systems 44
(1989),139-146.
[18] Deng, T.; Chen, Y.; Xu, W.; Dai, Q. A novel approach to fuzzy rough sets based on
a fuzzy covering, Inform. Sci. 177 (2007), 2308–2326.
[19] Mukherjee, T.K. and Sen, M.K. On Fuzzy Ideals of a Ring I, Fuzzy Sets and Systems
21 (1987), 99-104.
[20] Dixit, V.N.; Kumar, R. and Ajmal, N. On Fuzzy rings, Fuzzy Sets and Systems 49
(1992),205-213.
[21] Dixit, V.N.; Kumar, R. and Ajmal, N. Fuzzy ideals and fuzzy prime ideals of a ring,
Fuzzy Sets and Systems 44 (1991),127-138.
[22] Liu, W., Fuzzy Invariant Subgroups and Fuzzy Ideals, Fuzzy Sets and Systems 8
(1982),133-139.
[23] Wu, W.Z. and Zhang, W.X. Neighbourhood operator systems and approximations,
Inform. Sci. 144 (2002), 201-217.
[24] Zhu, W. Generalized rough sets based on relations, Inform. Sci. 177 (2007),
4997–5011.
[25] Jun, Y.B., Roughness of ideals in BCK-algebras, Sci. Math. Japonica 57(1) (2003),
165–169.
[26] Yao, Y.Y., Relational interpretation of neighbourhood operators and rough set ap-
proximation operator, Inform. Sci. 111 (1998), 239-259.
[27] Pawlak, Z., Rough sets, Int. J. Inf. Comp. Sci. 11 (1982), 341-356.
[28] Yue, Z., L-fuzzy ideals and primary L-fuzzy ideals, Fuzzy Sets and Systems 27
(1988), 345-350.
 CHARACTERIZATION OF ROUGH 3-PRIME IDEALS..... 131

Prof. Asma Ali

Department of Mathematics
Faculty of Sciences
Aligarh Muslim University,
Aligarh, INDIA.
E-mail: [email protected]

Arshad Zishan
Department of Mathematics
Faculty of Sciences
Aligarh Muslim University,
Aligarh, INDIA.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 132–137

A SOLUTION TO AN INTERESTING SUM

INVOLVING CLASSICAL HARMONIC NUMBER AND
CENTRAL BINOMIAL COEFFICIENT

NANDAN SAI DASIREDDY

(Received : 09 - 07 - 2023 ; Revised : 14 - 10 - 2023)

Abstract. In this paper we solve an interesting sum recently consid-

ered by A.S.Nimbran concerning the calculation of a series involving the
classical harmonic number and central binomial coefficient.The key in-
gredients for obtaining our infinite series result are some of the difficult
and obscure definite integral formulas due to K.S.Kölbig and Cornel
Ioan Vălean.

1. Introduction

Amrik Singh Nimbran left the following problem of evaluating a series

involving the classical harmonic number and central binomial coefficient as
an interesting sum in [4, p. 134], which we will provide a solution to in this
paper:
∞ 2n


X Hn n
S=
n2 22n
n=1
Throughout this paper Hn denotes the nth-classical harmonic num-
n  
X 1 2n
ber de?ned by Hn = , denotes the central binomial coeffi-
k n
k=1  
2n (2n)!
cient, which is defined for n ≥ 1 by = , ζ (s) denotes the
n (n!)2
∞
X 1
Riemann zeta function, which is defined by ζ (s) = , <(s) > 1 and
ns
n=1

2010 Mathematics Subject Classification: Primary 40A25, 65B10, 05A10; Secondary

11M06
Key words and phrases: Harmonic numbers, central binomial coefficients, Riemann
zeta function, infinite summation formulas

© Indian Mathematical Society, 2024 .

132
 A SOLUTION TO AN INTERESTING ... CENTRAL BINOMIAL COEFFICIENT 133

Lin (x) denotes the polylogarithm function, which is defined for |x| ≤ 1 by
∞
X xk
Lin (x) = , n ∈ N, n ≥ 2.
kn
k=1

To evaluate S, we shall establish a lemma.

Lemma 1.1. The following identity holds:

∞
n
X Hn cos2 θ
= − Li3 sin2 θ + Li3 cos2 θ +



n 2
n=1

Li2 sin2 θ ln sin2 θ +

1
ln cos2 θ ln2 sin2 θ + ζ (3) .



2
Proof. In the recent article [1, Lemma 4.5], Seán Mark Stewart had evalu-
ated the following generating function involving classical harmonic number:

∞
X Hn xn
= − Li3 (1 − x) + Li3 (x) + Li2 (1 − x) ln (1 − x) +
n2
n=1
1
ln (x) ln2 (1 − x) + ζ (3) . (1.1)
2

Substituting x = cos2 θ on both sides of (1.1) , we attain the following

equation:

∞
n
X Hn cos2 θ
= − Li3 1 − cos2 θ + Li3 cos2 θ +



n2
n=1

Li2 1 − cos2 θ ln 1 − cos2 θ +

1
ln cos2 θ ln2 1 − cos2 θ + ζ (3)



2

= − Li3 sin2 θ + Li3 cos2 θ +

Li2 sin2 θ ln sin2 θ +

1
ln cos2 θ ln2 sin2 θ + ζ (3) .



2
134 NANDAN SAI DASIREDDY

Theorem 1.2. The infinite series

∞ 2n


X Hn n
n2 22n
n=1

admits the symbolic form

9
ζ(3) − 4 ln 2ζ(2).
2
∞ 2n

∞   
X Hn n
X Hn 1 2n
Proof. =
n2 22n n2 22n n
n=1 n=1

Invoking the Wallis’ well-known integral formula, namely,

Z π  
2 2 (cos θ)2n dθ = 1 2n . we have,
π 0 22n n
Z π
 
∞    X ∞
X Hn 1 2n Hn  2 2
= (cos θ)2n dθ
n2 22n n n2 π 0
n=1 n=1
 π
! 
Z ∞
H cos 2θ n
2 n
=  2
X
dθ
π 0 n2
n=1
 π 
Z
2
=  2 − Li3 sin2 θ + Li3 cos2 θ + Li2 sin2 θ ln sin2 θ dθ +






π 0

 π 
Z  
2 2 1
ln cos2 θ ln2 sin2 θ + ζ (3) dθ



π 0 2
Z π Z π
 
2
− 2 Li3 sin2 θ dθ + 2 Li3 cos2 θ dθ +



=
π 0 0
 π
Z π

Z
2 2 1 2
Li2 sin2 θ ln sin2 θ dθ + ln cos2 θ ln2 sin2 θ dθ +





π 0 2 0
 π 
Z
2 2
ζ (3) dθ .
π 0
 A SOLUTION TO AN INTERESTING ... CENTRAL BINOMIAL COEFFICIENT 135

In this expression, due to symmetry, recognizing that,

Z π Z π
2 Li sin2 θ
dθ = 2 Li cos2 θ
dθ
3 3
0 0

Z π Z π
and 2 ln cos2 θ ln2 sin2 θ dθ = 2 ln sin2 θ
ln2 cos2 θ
dθ, we



0 0

conclude:
 π 
∞ 2n


Hn
Z
2
=  2 Li2 sin2 θ ln sin2 θ dθ +
X
n



n2 22n
π 0
n=1

Z π Z π
 
2 1 2
ln sin2 θ ln2 cos2 θ dθ + 2 ζ (3) dθ



π 2 0 0

Z π
 
2
2 2 Li2 sin2 θ ln (sin θ) dθ +


=
π 0

Z π
 
2 π
4 2 ln (sin θ) ln2 (cos θ) dθ + ζ (3)
π 0 2
 π   π 
Z Z
4 8
=  2 Li2 sin2 θ ln (sin θ) dθ +  2 ln (sin θ) ln2 (cos θ) dθ


π 0 π 0

+ζ (3)
 
a 1
Using Landen's identity, Li2 (−a) = − Li2 − ln2 (1 + a) for
1+a 2
a ≥ 0 [2], Cornel Ioan Vălean [3, (1.97)], had recently evaluated the definite
integral

Z π
2 ln (sin θ) Li sin2 θ
dθ = 5π ζ(3) − π ln 2ζ(2) + π ln3 2
2
0 8

and
136 NANDAN SAI DASIREDDY

using the derivative of the Euler's beta function and the Leibniz formula
for the differentiation of products, K.S.Kölbig [5, p. 25], had evaluated the
definite integral

Z π
2 ln (sin θ) ln2 (cos θ) dθ = π ζ(3) − 4 ln3 2
.
0 8

giving us that the following equality holds:

∞ 2n

 
X Hn n 4 5π 3
= ζ(3) − π ln 2ζ(2) + π ln 2 +
n2 22n π 8
n=1
8 π

ζ(3) − 4 ln3 2 + ζ (3)
π 8
5
= ζ(3) − 4 ln 2ζ(2) + 4 ln3 2+
2

ζ(3) − 4 ln3 2 + ζ (3)

9
= ζ(3) − 4 ln 2ζ(2).
2

Here in the proof of Theorem 1.2, Bernstein's theorem [6, Thm. 9.30,
p. 243] justifies interchanging the order of integration and summation be-
cause of the positivity of the coefficients.


Acknowledgement: I express my sincere thanks to the anonymous referee

for providing constructive comments and invaluable suggestions on the first
version of this paper.

References
[1] Stewart, Seán. M., Explicit Evaluation of Some Quadratic Euler-Type Sums Con-
taining Double-Index Harmonic Numbers, Tatra Mt. Math. Publ. 77(1) (2020),
73–98. doi: 10.2478/tmmp-2020-0034
[2] Landen, J., Mathematical Memoirs Respecting a Variety of Subjects, with an Ap-
pendix Containing Tables of Theorems for the Calculation of Fluents, 1 (1780),
112.
[3] Vălean, C. I., More (Almost) Impossible Integrals, Sums, and Series: A New Collec-
tion of Fiendish Problems and Surprising Solutions, Problem Books in Mathematics,
Springer, Cham, First Edition 2023. doi: 10.1007/978-3-031-21262-8
 A SOLUTION TO AN INTERESTING ... CENTRAL BINOMIAL COEFFICIENT 137

[4] Nimbran, A. S., Sums of series involving central binomial coefficients and har-
monic numbers, Math. Student 88(1-2) (2019), 125-135. Available from: https:
//indianmathsoc.org/ms/mathstudent-part-1-2019.pdf
[5] Kölbig, K. S., On the value of a logarithmic-trigonometric integral, BIT 11(1)
(1971), 21-28. doi: 10.1007/BF01935325
[6] Apostol, T. M., Mathematical Analysis, 2nd edn. Addison Wesley, 1974.

Nandan Sai Dasireddy

House No. 3-11-363, Pavani Plaza, Flat No:401,
Road No. 2, Shivaganga Colony, Siris Road,
L.B. Nagar, Hyderabad, Telangana, India.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 138–148

NIVEN NUMBERS AND A UNIQUE PROPERTY OF

2023

PHAKHINKON NAPP PHUNPHAYAP, TAMMATADA

KHEMARATCHATAKUMTHORN, AND PRAPANPONG PONGSRIIAM∗

Abstract. A positive integer n is called a Niven number if n is divisi-

ble by the sum of its decimal digits. In this article, we study Niven num-
bers of special types connecting with the values of generalized happy
functions, and show that 2023 is a unique Niven number of a particular
special type.

1. Introduction, literature review, and main results

At the end of the year 2022, the authors thought about finding some
unique properties of the integer 2023. We came up with an idea connecting
2023, Niven numbers, divisors, the sum of digits, and the happy and Alladi-
Erdős functions.
For each n ∈ N, let s1 (n) be the sum of the decimal digits of n and
let s2 (n) be the sum of the squares of the decimal digits of n. A positive
integer n is called a Niven number or a harshad number if n is divisible by
s1 (n) and n is called a happy number if
(k)
s2 (n) converges to 1 as k → ∞,
(k)
where s2 is the k-fold composition of s2 . While the history of happy
numbers is unclear, harshad numbers can be traced back to D. R. Kaprekar-
an Indian mathematician, and they were made more popular in the paper
by Ivan Niven [11].
Niven and happy numbers and the functions s1 and s2 have been stud-
ied by many mathematicians. For example, if N (x) is the number of Niven
numbers not exceeding x, then Kennedy and Cooper [12] showed in 1984

2010 Mathematics Subject Classification: 11A63, 11A25, 11A41, 11B83

Key words and phrases: Niven number, happy number, digital problem, prime,
factorization
* Corresponding author

© Indian Mathematical Society, 2024 .

138
 NIVEN NUMBERS AND A UNIQUE PROPERTY OF 2023 139

that N (x)/x converges to zero as x → ∞, which was later generalized

to other integer sequences by the same authors [13] in 1989. Cooper and
Kennedy [3] also proved in 1992 that there does not exist a sequence of more
than 20 consecutive integers which are Niven numbers, and 20 is sharp in
the sense that there are infinitely many sequences of 20 consecutive Niven
numbers. Cooper and Kennedy’s result [3] was extended by Grundman [7],
Cai [1], and Wilson [22] to Niven numbers in the general bases b ≥ 2 during
1994 to 1997. Then De Koninck, Doyon, and Kátai [14] obtained in 2003 an
asymptotic formula N (x) ∼ cx/ log x as x → ∞, where c = (14 log 10)/27.
Results concerning the distribution of the function s1 on arithmetic pro-
gressions were obtained by Gelfond [5] and investigated further by several
researchers, which were collected and explained in Morgenbesser’s thesis
[15]. Answering a question in Guy’s book [9], El-Sedy and Siksek [4] proved
in 2000 that there are arbitrarily long consecutive happy numbers. Other
mathematicians have also made some contributions to happy numbers; see
for example in the work of Chase [2], Gilmer [6], Grundman and Teeple
[8], Hargreaves and Siksek [10], Noppakeaw, Phoopha, and Pongsriiam [16],
Pan [17], Phoopha, Pongsriiam, and Phunphayap [18], Styer [20], Subwat-
tanachai and Pongsriiam [21], and Zhou and Cai [23]. For more information
on Niven and happy numbers, we refer the reader to the sequences A005349
and A007770 in OEIS [19].
However, in this article, we consider a more elementary problem that
connects the integer 2023, Niven numbers, prime factorizations, the sum of
digits, the sum of divisors, and the happy and Alladi-Erdős functions. We
have s1 (2023) = 2+0+2+3 = 7 is a prime, s2 (2023) = 22 +02 +22 +32 = 17
is also a prime, and the canonical factorization of 2023 into powers of primes
is 2023 = 7 × 172 = s1 (2023)s2 (2023)2 , and we show in Corollary 1.2 that
2023 is the only positive integer with this property. In addition, we show in
Theorem 1.4 that the only positive integer n satisfying n = s1 (n)s1 (σ(n))2
or n = s1 (n)s2 (A(n))2 where s1 (n) and s2 (A(n)) are prime is n = 2023.
Here σ(n) is the sum of positive divisors of n and A(n) is the sum of prime
divisors (with multiplicity) of n.
We can extend the investigation to a larger set of Niven numbers of simi-
lar types by connecting them with the values of generalized happy functions.
For each integers α ≥ 1 and b ≥ 2, let sα,b : N → N be defined by

sα,b (n) = aαk + aαk−1 + · · · + aα0

140 P. N. PHUNPHAYAP, T. KHEMARATCHATAKUMTHORN, AND P. PONGSRIIAM

if n = (ak ak−1 · · · a0 )b = ki=0 ai bi is the b-adic expansion of n with ak 6= 0

P

and 0 ≤ ai < b for all i = 0, 1, 2, . . . , k. We say that n is a Niven number in

base b or a b-adic Niven number if n is divisible by s1,b (n), and n is a b-adic
Niven number of type 2 if n can be written as

n = sα1 ,b (n)a1 sα2 ,b (n)a2 · · · sαk ,b (n)ak , (1.1)

where 1 = α1 < α2 < · · · < αk and a1 , a2 , . . . , ak are positive integers. In

addition, we call n a b-adic Niven number of type 3 or a special b-adic Niven
number if (1.1) is the canonical factorization of n into powers of primes, that
is, n satisfies (1.1) and sαi ,b (n) is a prime number for each i. Furthermore,
if a1 + a2 + · · · + ak = d, then we say that n is of degree d. If the sαi ,b (n)
are distinct and αk = `, then n is said be at level `. Finally, we remark that
b-adic Niven numbers are usually called b-Niven numbers for short, but for
clarity, we do not use the short name in this article.
For simplicity, we focus only on the case of special Niven numbers where
the base b = 10 and the degree d and the level ` are small. We also write
sα instead of sα,10 . Then we obtain the following theorem.

Theorem 1.1. Let n be a positive integer. Then the following statements

hold.

(i) (Special Niven numbers of degree 1)

n = s1 (n) if and only if n ≤ 9.
Consequently, n is a special Niven number of degree 1 if and only
if n = 2, 3, 5, 7.
(ii) (Special Niven numbers of degree 2 at level ` ≤ 3)
n = s1 (n)2 if and only if n = 1 or n = 81;
n = s1 (n)s2 (n) if and only if n = 1, 133, 315, 803, 1148, 1547, 2196;
n = s1 (n)s3 (n) if and only if n = 1, 1215, 3700, 11680, 13608, 87949.
Among these integers, 133 and 803 are the special Niven numbers
with degree 2 at level 2, and 87949 is the only special Niven number
of degree 2 at level 3.
(iii) (Special Niven numbers of degree 3 at level ` ≤ 2)
n = s1 (n)3 if and only if n = 1, 512, 4913, 5832, 17576, 19683;
n = s1 (n)2 s2 (n) if and only if n = 1, 2511, 24624;
n = s1 (n)s2 (n)2 if and only if n = 1, 2023, 2400, 52215, 615627,
938600, 1648656.
 NIVEN NUMBERS AND A UNIQUE PROPERTY OF 2023 141

Consequently, 4913 is the only special Niven number of degree 3

at level 1, and 2023 is the only special Niven number of degree 3 at
level 2.
(iv) (Special Niven numbers of degree 3 at level 3)
n = s1 (n)2 s3 (n) if and only if n = 1, 32144, 37000, 111616, 382360;
n = s1 (n)s2 (n)s3 (n) if and only if n = 1, 3163617, 3822147;
n = s1 (n)s3 (n)2 if and only if n = 1, 4147200, 12743163, 21147075,
39143552, 52921472, 156754936, 205889445, 233935967.
In particular, there is no special Niven number of degree 3 at level
3.
(v) (Special Niven numbers of degree 2 at level ` = 4, 5, 6)
n = s1 (n)s4 (n) if and only if n = 1, 182380, 444992;
n = s1 (n)s5 (n) if and only if n = 1, 415000, 3508936, 3828816,
4801896, 5659875;
n = s1 (n)s6 (n) if and only if n = 1, 21268584.
In particular, there is no special Niven number of degree 2 at level
` = 4, 5, 6.

Before proving Theorem 1.1, let us record some arithmetic curiosities

that can be obtained straightforwardly from our theorem.

Corollary 1.2. The only positive integer n satisfying n = s1 (n)s2 (n)2

where s1 (n) and s2 (n) are prime is n = 2023.

Corollary 1.3. We have

(A) 81 = (8 + 1)2 , 133 = (1 + 3 + 3)(12 + 32 + 32 ),
315 = (3 + 1 + 5)(32 + 12 + 52 ), 803 = (8 + 0 + 3)(82 + 02 + 32 ),
1148 = (1 + 1 + 4 + 8)(12 + 12 + 42 + 82 ),
1547 = (1 + 5 + 4 + 7)(12 + 52 + 42 + 72 ),
2196 = (2 + 1 + 9 + 6)(22 + 12 + 92 + 62 ),
1215 = (1 + 2 + 1 + 5)(13 + 23 + 13 + 53 ),
3700 = (3 + 7 + 0 + 0)(33 + 73 + 03 + 03 ),
11680 = (1 + 1 + 6 + 8 + 0)(13 + 13 + 63 + 83 + 03 ),
13608 = (1 + 3 + 6 + 0 + 8)(13 + 33 + 63 + 03 + 83 ),
87949 = (8 + 7 + 9 + 4 + 9)(83 + 73 + 93 + 43 + 93 ),
(B) 512 = (5 + 1 + 2)3 , 4913 = (4 + 9 + 1 + 3)3 ,
5832 = (5 + 8 + 3 + 2)3 , 17576 = (1 + 7 + 5 + 7 + 6)3 ,
142 P. N. PHUNPHAYAP, T. KHEMARATCHATAKUMTHORN, AND P. PONGSRIIAM

19683 = (1 + 9 + 6 + 8 + 3)3 ,
2511 = (2 + 5 + 1 + 1)2 (22 + 52 + 12 + 12 ),
24624 = (2 + 4 + 6 + 2 + 4)2 (22 + 42 + 62 + 22 + 42 ),
2023 = (2 + 0 + 2 + 3)(22 + 02 + 22 + 32 )2 ,
2400 = (2 + 4 + 0 + 0)(22 + 42 + 02 + 02 )2 ,
52215 = (5 + 2 + 2 + 1 + 5)(52 + 22 + 22 + 12 + 52 )2 ,
615627 = (6 + 1 + 5 + 6 + 2 + 7)(62 + 12 + 52 + 62 + 22 + 72 )2 ,
938600 = (9 + 3 + 8 + 6 + 0 + 0)(92 + 32 + 82 + 62 + 02 + 02 )2 ,
1648656 = (1 + 6 + 4 + 8 + 6 + 5 + 6)(12 + 62 + 42 + 82 + 62 + 52 + 62 )2 ,

(C) 32144 = (3 + 2 + 1 + 4 + 4)2 (33 + 23 + 13 + 43 + 43 ),

37000 = (3 + 7 + 0 + 0 + 0)2 (33 + 73 + 03 + 03 + 03 ),
111616 = (1 + 1 + 1 + 6 + 1 + 6)2 (13 + 13 + 13 + 63 + 13 + 63 ),
382360 = (3 + 8 + 2 + 3 + 6 + 0)2 (33 + 83 + 23 + 33 + 63 + 03 ),
(D) 182380 = (1 + 8 + 2 + 3 + 8 + 0)(14 + 84 + 24 + 34 + 84 + 04 ),
444992 = (4 + 4 + 4 + 9 + 9 + 2)(44 + 44 + 44 + 94 + 94 + 24 )

and this is the complete list of positive integers larger than 1 with this prop-
erty.

Finally, let us record one more theorem connecting 2023, Niven numbers,
the sum of divisors function σ(n), and Alladi-Erdős function A(n).

Theorem 1.4. Let n be a positive integer. Then

n = s1 (n)s1 (σ(n))2 if and only if n = 1, 100, 2023, 7500, 23328;

n = s1 (n)s2 (A(n))2 if and only if n = 48, 1521, 2023, 3600, 154652, 798950.

In particular, a positive integer n satisfies the first equation with s1 (n) being
a prime if and only if n = 2023. The only positive integer n satisfying the
second equation where s1 (n) and s2 (A(n)) are prime is n = 2023.

We give some lemmas and prove Theorems 1.1 and 1.4 in the next
section.
 NIVEN NUMBERS AND A UNIQUE PROPERTY OF 2023 143

2. Lemmas and the proof of the main theorems

Throughout, for each a, b ∈ Z, a ≥ 1, b ≥ 0, c ≥ 4, let fa,b , Fc , and ga,b

be the functions defined by

fa,b (n) = s1 (n)a s2 (n)b , Fc (n) = s1 (n)sc (n), ga,b = s1 (n)a s3 (n)b

for all n ∈ N. In addition, let h(n) = s1 (n)s2 (n)s3 (n) for all n ∈ N. Then
we have the following result.
Lemma 2.1. For each n ∈ N, we have
(i) f2,0 (n) ≤ f1,1 (n) ≤ f3,0 (n) ≤ f2,1 (n) ≤ f1,2 (n);
(ii) g2,1 (n) ≤ h(n) ≤ g1,2 (n);
(iii) g1,1 (n) ≤ F4 (n) ≤ F5 (n) ≤ F6 (n).
Proof. This follows from the fact that s1 (n) ≤ s2 (n) ≤ s1 (n)2 , and s2 (n) ≤
s3 (n) ≤ s4 (n) ≤ s5 (n) ≤ s6 (n) for all n ∈ N. 
Lemma 2.2. We have 9(k + 1) < 10k for all k ≥ 2, 95 (k + 1)3 < 10k for all
k ≥ 8, 97 (k + 1)2 < 10k for all k ≥ 9, and 97 (k + 1)3 < 10k for all k ≥ 10.
Proof. It is easy to verify that the second inequality holds when k = 8. If
k ≥ 8 and the second inequality holds for k, then
 3  3
5 3 5 3 1 5 3 10
9 (k + 2) = 9 (k + 1) 1 + ≤ 9 (k + 1) < 10k+1 ,
k+1 9
which proves the second inequality by induction. The others are similar. 
From this point, we apply Lemma 2.2 without further reference.
Lemma 2.3. We have s1 (n) < n for all n ≥ 10.
Proof. If n ≥ 100 and n = (ak ak−1 · · · a0 )10 where k ≥ 2, then
k
X
s1 (n) = ai ≤ 9(k + 1) < 10k ≤ n.
i=1

It is easy to check that s1 (n) < n for 10 ≤ n ≤ 99. So the proof is

complete. 
Lemma 2.4. We have f1,2 (n) < n for all n ≥ 2 × 107 and 2 × 107 is
sharp in the sense that if m = 2 × 107 − 1, then f1,2 (m) > m. In addition,
g1,2 (n) < n for all n ≥ 3 × 109 , and if m = 3 × 109 − 1, then g1,2 (m) > m.
Furthermore, F6 (n) < n for all n ≥ 3 × 108 and if m = 3 × 108 − 1, then
F6 (m) > m.
144 P. N. PHUNPHAYAP, T. KHEMARATCHATAKUMTHORN, AND P. PONGSRIIAM

Proof. Let n ≥ 108 . Then n = (ak ak−1 · · · a0 )10 where k ≥ 8, ak 6= 0, and

0 ≤ ai ≤ 9 for all i. For convenience, we write f instead of f1,2 . Then we
obtain
k
! k !2
X X
2

2
f (n) = ai ai ≤ 9(k + 1) 92 (k + 1) < 10k ≤ n.
i=0 i=0

Therefore f (n) < n for all n ≥ 108 . Next, if 30233088 < n ≤ 108 − 1, then

f (n) ≤ s1 (108 − 1)s2 (108 − 1)2 = 30233088 < n.

If 21192665 < n ≤ 30233088, then

f (n) ≤ s1 (3 × 107 − 1)s2 (3 × 107 − 1)2 = 21192665 < n.

Finally, if 2 × 107 ≤ n ≤ 21192665, then

f (n) ≤ s1 (20999999)s2 (20999999)2 = 13445600 < n.

Thus f (n) < n for all n ≥ 2 × 107 . It is easy to check that f (m) > m if
m = 2 × 107 − 1. The inequalities for g1,2 and F6 can be proved similarly.
So the proof is complete. 

With these lemmas, we are now ready to give a proof of Theorem 1.1.

Proof of Theorem 1.1. It is straightforward to verify that the converse of

each statement in (i) to (v) of Theorem 1.1 holds. By Lemma 2.3, it is easy
to see that s1 (n) = n if and only if n ≤ 9, and so (i) is proved. Solving the
equations n = s1 (n)2 and n = s1 (n)s2 (n) in (ii) are equivalent to solving
f2,0 (n) = n and f1,1 (n) = n. By Lemmas 2.1 and 2.4, the integers larger
than or equal to 2 × 107 are not the solutions. The number of positive
integers less than 2 × 107 is finite and not too large, and we can therefore
use a computer to check whether or not an integer n < 2 × 107 is a solution.
Similarly, n = s1 (n)s3 (n) if and only if g1,1 (n) = n, and by Lemmas
2.1 and 2.4, we only need to search for the solutions in positive integers
n < 3 × 108 . This leads to the solutions to each equation in (ii). Then it is
easy to check that among these solutions, both s1 (n) and s2 (n) are prime
numbers only when n = 133, 803, 87949. This proves (ii).
Similarly, by Lemmas 2.1 and 2.4, the proof of the other parts can be
reduced to finding positive integers n < M such that f (n) = n where M is
2 × 107 , 3 × 108 , or 3 × 109 as appropriate and f = f3,0 , f2,1 , f1,2 , F4 , F5 ,
F6 , g2,1 , h, g1,2 , respectively. This completes the proof. 
 NIVEN NUMBERS AND A UNIQUE PROPERTY OF 2023 145

Theorem 1.4 can be proved by using the same idea as follows.

Proof of Theorem 1.4. Let f (n) = s1 (n)s1 (σ(n))2 and g(n) = s1 (n)s2 (A(n))2
for all n ∈ N. We first observe that for each n ∈ N, we have
X X
σ(n) = d≤n 1 ≤ n2 . (2.1)
d|n d|n

If n ≥ 107 , then n = (ak ak−1 · · · a0 )10 where k ≥ 7, ak 6= 0, and 0 ≤ ai ≤ 9

for every i, which implies σ(n) ≤ 102k+2 , and therefore s1 (σ(n)) ≤ 9(2k+2).
Thus if n ≥ 107 , then

f (n) ≤ 9(k + 1) · 92 (2k + 2)2 < 10k ≤ n,

where the second inequality can be proved by induction. So the solutions

to f (n) = n can be found only in the range n < 107 , which can be done by
using a computer search. This proves the first statement.
Next, let n > 1 and write n = pa11 pa22 · · · pakk where p1 , p2 , . . . , pk are
distinct primes and a1 , a2 , . . . , ak are positive integers. We observe that if
a ≥ 1 and q ≥ 2 are integers, then q a−1 ≥ 2a−1 ≥ a, and so aq ≤ q a .
Therefore

A(n) = a1 p1 + a2 p2 + · · · + ak pk ≤ pa11 + pa22 + · · · + pakk . (2.2)

We also observe that if c1 , c2 ≥ 2, then (c1 − 1)(c2 − 1) ≥ 1, which implies

c1 + c2 ≤ c1 c2 . So if c1 , c2 , . . . , ck ≥ 2, then c1 + c2 + c3 + · · · + ck ≤
c1 c2 c3 · · · ck . From this and (2.2), we see that A(n) ≤ n. Now if n ≥ 108 ,
n = (ak ak−1 · · · a0 )10 , k ≥ 8, ak 6= 0, and 0 ≤ ai ≤ 9 for every i, then
s2 (A(n)) ≤ 81(k + 1), and so

g(n) ≤ 9(k + 1)(81(k + 1))2 < 10k ≤ n.

Since g(n) < n for all n ≥ 108 , we can use a computer to search for the so-
lution to g(n) = n in the range n < 108 . This proves the second statement.
The rest follows immediately. This completes the proof. 

It is possible to solve the equation n = fa,b (n) and n = ga,b (n) when a, b
are large, but the upper bound may be much larger than 3×109 , which takes
a longer computational time and requires a more powerful computer and
a better skill in computer programming. More generally, if a1 , a2 , . . . , ak ,
α1 , α2 , . . . , αk are fixed, it is possible to solve the equation

n = sα1 (n)a1 sα2 (n)a2 · · · sαk (n)ak .

146 P. N. PHUNPHAYAP, T. KHEMARATCHATAKUMTHORN, AND P. PONGSRIIAM

We leave this problem to the interested reader.

Table 1 shows special Niven numbers of degree d at level `. By the
definition, when d = 1, we automatically have ` = 1. So there are no result
for d = 1 and ` ≥ 2. The case d = 3 and ` = 4, 5, 6 is not included in
Theorem 1.1, so we put ∗ in the table to mean that it is open. It seems
that there may be an infinite number of Niven numbers of type 2 but the
Niven numbers of type 3 are very rare.

d, ` `=1 `=2 `=3 `=4

d = 1 2, 3, 5, 7 - - -
d = 2 None 133 and 803 87949 None
d=3 4913 2023 None ∗
Table 1

Problem Determine whether or not there are infinitely many Niven num-
bers of type 2. Prove or disprove that the number of Niven numbers of type
3 is finite. If d and ` are very large, is there a special Niven number of degree
d at level `? Are there infinitely many n such that n = s1 (n)a sα (σ(n))b or
n = s1 (n)a sα (A(n))b for some a, b, α ∈ N distinct from those in Theorem
1.4?

Acknowledgement

Prapanpong Pongsriiam’s research project is supported jointly by the

Faculty of Science Silpakorn University and the National Research Council
of Thailand (NRCT), grant number NRCT5-RSA63021-02. He was also
supported by the Tosio Kato Fellowship given by the Mathematical Society
of Japan during his visit at Nagoya University in July 2022 to July 2023.
This work was started when he was in Japan and finished when he came back
to Thailand. He is thankful to Professor Kohji Matsumoto for being his host
professor and finding a very convenient and comfortable accommodation
and office for him during his visit at Nagoya University.

References
[1] Cai, T., On 2-Niven numbers and 3-Niven numbers, Fibonacci Quart. 34(2) (1996),
118–120.
[2] Chase, Z., On the iterates of digit maps, Integers 18 (2018), #A86.
 NIVEN NUMBERS AND A UNIQUE PROPERTY OF 2023 147

[3] Cooper, C. and Kennedy, R., On consecutive Niven numbers, Fibonacci Quart. 31
(1993), 146–151.
[4] El-Sedy, E. and Siksek, S., On happy numbers, Rocky Mountain J. Math. 30 (2000),
565–570.
[5] Gelfond, A. O., Sur les nombres qui ont des propriété additives et multiplicatives
données, Acta Arith. 13 (1968), 259–265.
[6] Gilmer, J., On the density of happy numbers, Integers 13 (2013), #A48.
[7] Grundman, H., Sequences of consecutive n-Niven numbers, Fibonacci Quart. 32(2)
(1994), 174–175.
[8] Grundman, H. G. and Teeple, E. A., Iterated sums of fifth powers of digits, Rocky
Mountain J. Math. 38 (2008), 1139–1146.
[9] Guy, R. K., Unsolved Problems in Number Theory, Springer-Verlag, First Edition,
1981.
[10] Hargreaves, K. and Siksek, S., Cycles and fixed points of happy functions, J. Comb.
Number Theory 2 (2010), 65–77.
[11] Harshad number, Wikipedia, available at https://en.wikipedia.org/wiki/Harshad
_number
[12] Kennedy, R. and Cooper, C., On the natural density of the Niven numbers, College
Math. J. 15 (1984), 309–312.
[13] Kennedy, R. and Cooper, C., Chebyshev’s inequality and natural density, Amer.
Math. Monthly 96 (1989), 118–124.
[14] De Koninck, J. M., Doyon, N. and Kátai, I., On the counting function for the Niven
numbers, Acta Arith. 106 (2003), 265–275.
[15] Morgenbesser, J., Gelfond’s Sum of Digits Problems, Diploma
Thesis, Vienna University of Technology, 2008. Available at
https://dmg.tuwien.ac.at/drmota/morgenbesserda.pdf
[16] Noppakeaw, P. Phoopha, N. and Pongsriiam, P., Composition of happy functions,
Notes Number Theory Discrete Math. 25(3) (2019), 13–20.
[17] Pan, H., On consecutive happy numbers, J. Number Theory 128 (2008), 1646–1654.
[18] Phoopha, N. Pongsriiam, P. and Phunphayap, P., Digit Maps, Math. Gaz. 107(568)
(2023), 35–43.
[19] Sloane, N. J. A., The On-Line Encyclopedia of Integer Sequences, https://oeis.org
[20] Styer, R., Smallest examples of strings of consecutive happy numbers, J. Integer Seq.
13 (2010), Article 10.6.3.
[21] Subwattanachai, K. and Pongsriiam, P., Composition of happy functions and digit
maps, Int. J. Math. Comput. Sci. 16(1) (2021), 169–178.
[22] Wilson, B., Construction of 2n consecutive n-Niven numbers, Fibonacci Quart. 35
(1997), 122–128.
[23] Zhou, X. and Cai, T., On e-power b-happy numbers, Rocky Mountain J. Math. 39
(2009), 2073–2081.
148 P. N. PHUNPHAYAP, T. KHEMARATCHATAKUMTHORN, AND P. PONGSRIIAM

Phakhinkon Napp Phunphayap

Department of Mathematics
Faculty of Sciences
Burapha University
Chonburi, 20131, Thailand
E-mail: [email protected], [email protected]

Tammatada Khemaratchatakumthorn
Department of Mathematics
Faculty of Sciences
Silpakorn University
Nakhon Pathom, 73000, Thailand
E-mail: [email protected], [email protected]

Prapanpong Pongsriiam
Department of Mathematics
Faculty of Sciences
Silpakorn University
Nakhon Pathom, 73000, Thailand
and
Graduate School of Mathematics
Nagoya University
Nagoya, 464-8602, Japan
E-mail: [email protected], [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 149–159

SOLVING WIENER-HOPF EQUATION INVOLVING

XOR-OPERATION

ZAHOOR AHMAD RATHER AND RAIS AHMAD

Abstract. In this study, we explore a novel application of the Wiener-

Hopf equation by incorporating the XOR-operation. Our findings re-
veal that the Wiener-Hopf equation involving XOR-operation is math-
ematically equivalent to a variational inequality problem. To solve
this equation, we propose an iterative algorithm specifically tailored
for XOR-based Wiener-Hopf equations. By utilizing the Bounded In-
verse Theorem, we successfully obtain the solution. Furthermore, we
discussed the convergence criteria associated with this approach.

1. Introduction

The Wiener-Hopf method is a well-established mathematical technique

that finds extensive applications in applied mathematics. It is particularly
useful for solving two-dimensional partial differential equations with mixed
boundary conditions on a shared boundary. The introduction of Wiener-
Hopf equations can be attributed to Robinson [21] and Shi [26, 27], who
initially employed them in specific contexts. These equations are employed
in various scenarios to analyze the properties of solutions for different classes
of variational inequalities and their extensions.
Variational inequalities find applications in a wide range of fields, in-
cluding fluid flow through porous media, moving boundary value problems,
traffic assignment problems, optimization problems, economic equilibrium
and more. For detailed information on variational inequalities and Wiener-
Hopf equations, one can refer to works such as [1, 2, 3, 5, 6, 7, 8, 9, 10, 11,
12, 13, 16, 17, 18, 19, 20, 22, 23, 24, 25] and the references provided therein.
In logic, there exists a connective called ’exclusive or’ also known as
exclusive disjunction. It evaluates to true when exactly one (but not both)
of two conditions is true. The XOR operation often represented as ⊕ is used
2010 Mathematics Subject Classification: 49J40, 47H05
Key words and phrases: Inequality, equation, convergence, solution, algorithm

© Indian Mathematical Society, 2024 .

149
150 ZAHOOR AHMAD RATHER AND RAIS AHMAD

to express this connective. By applying XOR operation to the values of n

and m, the binomial coefficient m


n mod 2 can be efficiently calculated. This
property simplifies the construction of Pascal’s triangle mod 2. The XOR
Cipher utilizes the XOR logical operation for data encryption. The data
is encrypted by performing XOR operations with a specific key resulting
in encrypted data. To decrypt the data, the same key is used and XOR
operation is applied again. Notably, the XOR operation employs the same
key for both encryption and decryption processes.
Motivated by the aforementioned considerations, this paper presents an
introduction to a novel Wiener-Hopf equation that incorporates the XOR
operation. Additionally, we discuss the corresponding variational inequality
problem and establish an equivalence between the two. To solve the Wiener-
Hopf equation involving XOR operation we propose an iterative algorithm.
Finally, we demonstrate the existence and convergence of a solution by
employing the Bounded Inverse Theorem.

2. Preliminaries

In this paper, we consider a real ordered Hilbert space denoted as H

equipped with a norm k.k and an inner product h., .i. We also assume the
existence of a closed convex subset K and a cone C within H. To establish
the main result of this paper, we draw upon several established concepts
and results many of which can be found in references such as [4, 14, 15] by
Du and Li.

Definition 2.1. Consider a cone C in the set H. For any elements l and
m belonging to H, the inequality l ≤ m is true if and only if l − m belongs
to the cone C. This relationship denoted by "≤" is referred to as a partial
ordered relation.

Definition 2.2. Any two elements l and m belonging to the set H are
considered comparable if either l ≤ m or m ≤ l holds (represented as
l ∝ m).

Definition 2.3. For any elements l and m belonging to the set H, we

use the notations lub{l, m} and glb{l, m} to represent the least upper
bound and the greatest lower bound of the set {l, m}, respectively. If both
lub{l, m} and glb{l, m} exist, we can define certain binary operations as
follows:
 SOLVING WIENER-HOPF EQUATION INVOLVING XOR-OPERATION 151

(i) l ∨ m = lub{l, m},

(ii) l ∧ m = glb{l, m},
(iii) l ⊕ m = (l − m) ∨ (m − l),
(iv) l m = (l − m) ∧ (m − l).
The symbols ∨, ∧, ⊕, and refer to the OR, AND, XOR, and
XNOR operations, respectively.

Proposition 2.4. Let ⊕ be an XOR-operation and be an XNOR-operation.

Then the following relations hold:
(i) l l = 0, l m = m l = −(l ⊕ m) = −(m ⊕ l),
(ii) if l ∝ 0, then −l ⊕ 0 ≤ l ≤ l ⊕ 0,
(iii) (λl) ⊕ (λm) = |λ|(l ⊕ m),
(iv) 0 ≤ l ⊕ m, if l ∝ m,
(v) if l ∝ m, then l ⊕ m = 0 if and only if l = m,
(vi) k0 ⊕ 0k = k0k = 0,
(vii) kl ∨ mk ≤ klk ∨ kmk ≤ klk + kmk,
(viii) kl ⊕ mk ≤ kl − mk,
(ix) if l ∝ m, then kl ⊕ mk = kl − mk.

Lemma 2.5. Given t ∈ H, s ∈ K satisfies the inequality

hs − t, v − si ≥ 0, f or all v ∈ K (2.1)

if and only if
u = PK (t)
where K is a nonempty closed convex set in H and PK is the projection
operator.

Theorem 2.6 (Bounded Inverse Theorem). A bijective, bounded, linear

operator T from one Banach space to another has bounded inverse T −1 .

3. Formulation of the problem and equivalence result

Consider the mappings T, g : H → H, where g is bijective, bounded, lin-

ear, Lipschitz continuous and T is linear. We explore a novel Wiener-Hopf
equation involving XOR-operation as:
Find s, t ∈ H such that

ρT g −1 PK (t) ⊕ (QK (t)) = 0, (3.1)

152 ZAHOOR AHMAD RATHER AND RAIS AHMAD

Here, t = g(s) − ρT (s), QK (t) = (PK − I)(t), I denotes the identity

operator, ρ > 0 is a constant and PK represents the projection of H onto
K.
We now introduce its equivalent variational inequality problem as:
Find s ∈ H such that g(s) ∈ K satisfying the inequality:

hT (s), g(v) − g(s)i ≥ 0, for all g(v) ∈ K. (3.2)

Both equations (3.1) and (3.2) are interconnected and the study of the
Wiener-Hopf equation involving XOR-operation reveals its equivalence with
the variational inequality problem (3.2). This connection offers new per-
spectives and opportunities for analyzing and solving complex mathematical
challenges in various applications. The following Lemma ensures that the
variational inequality problem (3.2) is equivalent to a fixed point equation.

Lemma 3.1. The variational inequality problem (3.2) admits a solution

s ∈ K, g(s) ∈ K if and only if it satisfies the following equation

g(s) = PK [g(s) − ρT (s)],

where ρ > 0 is a constant.

Proof. Proof is easy and hence omitted. 

Now in following Lemma we show that Wiener-Hopf equation inolving

XOR-operation (3.1) and variational inequality problem (3.2) are equiva-
lent.

Lemma 3.2. The Wiener-Hopf equation involving XOR-operation (3.1)

admits a solution s, t ∈ H if and only if s ∈ H, g(s) ∈ K satisfies the
variational inequality problem (3.2), given that ρT g −1 PK (t) ∝ QK (t), where
C denotes a cone in H,

g(s) = PK (t), t = g(s) − ρT (s), QK (t) = (PK − I)(t) (3.3)

and ρ > 0 is a constant.

Proof. Let s, t ∈ H be a solution of Wiener-Hopf equation inolving XOR-
operation (3.1). Then for some ρ > 0, we have

ρT g −1 PK (t) ⊕ QK (t) = 0.
 SOLVING WIENER-HOPF EQUATION INVOLVING XOR-OPERATION 153

Since ρT g −1 PK (t) ∝ QK (t), by (v) of Proposition 2.4, we have

ρT g −1 PK (t) = QK (t)
= PK (t) − t. (3.4)

By using Lemma 2.5, Lemma 3.1, (3.4) and the fact that t = g(s)−ρT (s), we
have

hg(s) − t, g(v) − g(s)i ≥ 0,

hPK (t) − t, g(v) − g(s)i ≥ 0,
hρT g −1 PK (t), g(v) − g(s)i ≥ 0,
ρhT g −1 PK (t), g(v) − g(s)i ≥ 0,

which gives us
hT (s), g(v) − g(s)i ≥ 0,
the required variational inequality problem (3.2). Conversely, suppose that
the variational inequality problem (3.2) holds, that is s ∈ H, g(s) ∈ K such
that
hT (s), g(v) − g(s)i ≥ 0, f or all g(v) ∈ K.
Using Lemma 3.1, we have

g(s) = PK [g(s) − ρT (s)] = PK (t), (3.5)

where t = g(s) − ρT (s).

Using (3.5), we have

t = g(s) − ρT (s)
= PK (t) − ρT g −1 PK (t),

so,

ρT g −1 PK (t) = PK (t) − t
= QK (t).

Thus, we have

ρT g −1 PK (t) ⊕ QK (t) = QK (t) ⊕ QK (t) = 0,

that is, ρT g −1 PK (t) ⊕ QK (t) = 0,

the required Wiener-Hopf equation involving XOR-operation 3.1. 

154 ZAHOOR AHMAD RATHER AND RAIS AHMAD

4. Algorithm, Existence and Convergence

In this section, we present an original iterative algorithm and establish

both the existence and convergence results for the Wiener-Hopf equation
involving XOR-operation (3.1). By simple manipulation equation (3.1) can
be rewritten as:

ρT g −1 PK (t) ⊕ (PK (t) − t) = 0, where t = g(s) − ρT (s)

or ρT g −1 PK (t) = PK (t) − (g(s) − ρT (s))
ρT g −1 PK (t) = PK (t) + ρT (s) − g(s)
ρT g −1 PK (t) ⊕ (PK (t) + ρT (s)) = (PK (t) + ρT (s)) ⊕ (PK (t) + ρT (s)) − g(s)
ρT g −1 PK (t) ⊕ (PK (t) + ρT (s)) = −g(s),

thus, we have

g(s) = ρT g −1 PK (t) (PK (t) + ρT (s)), as l m = −(l ⊕ m). (4.1)

This fixed point formulation 4.1 enables us to develop the following iterative
algorithm.

Algorithm 4.1. For s0 , t0 ∈ H, compute the sequences {sn } and {tn }

by the following scheme:

g(sn+1 ) = ρT g −1 PK (tn ) (PK (tn ) + ρT (sn )). (4.2)

where tn = g(sn ) − ρT (sn ), n = 0, 1, 2, ... . (4.3)

Theorem 4.1. Let H be a real ordered Hilbert space, K be a closed con-
vex subset of H and C be a cone in H with partial ordering “ ≤ ”. Let
T, g : H → H be the single-valued mappings such that g is bijective, bounded,
linear, Lipschitz continuous with constant λg , strongly monotone with con-
stant δg and T is linear, Lipschitz continuous with constant λT . Suppose
that g(sn+1 ) ∝ g(sn ), ρT g −1 PK (t) ∝ Q(t), where QK (t) = (PK −I)(t), t =
g(s) − ρT (s), g(s) ∈ K, n= 0,1,2,... .
If the following condition is satisfied
s
n (ξ − ρλ )2 o
2 2 g T 2 2
|λg + ρλT | < ρ λT + − ρ λT . (4.4)
(1 + ρλT δg )2
Then, there exist s, t ∈ H the solution of Wiener-Hopf equation involving
XOR-operation (3.1) and the sequences {sn } and {tn } generated by Algo-
rithm 4.1 converge strongly to s and t, respectively.
 SOLVING WIENER-HOPF EQUATION INVOLVING XOR-OPERATION 155

Proof. Using 4.2 of Algorithm 4.1, (i) and (iii) of proposition 2.4, we have

0 ≤ g(sn+1 ) ⊕ g(sn ) = [ρT g −1 PK (tn ) (PK (tn ) + ρT (sn )]

−1
⊕[ρT g PK (tn−1 ) (PK (tn−1 ) + ρT (sn−1 )]
= [−(ρT g −1 PK (tn ) ⊕ (PK (tn ) + ρT (sn ))]
⊕[−(ρT g −1 PK (tn−1 ) ⊕ (PK (tn−1 ) + ρT (sn−1 ))]
= | − 1|[ρT g −1 PK (tn ) ⊕ (PK (tn ) + ρT (sn )]
⊕[ρT g −1 PK (tn−1 ) ⊕ (PK (tn−1 ) + ρT (sn−1 )].(4.5)

From 4.5, (viii) of proposition 2.4, we have

kg(sn+1 ) ⊕ g(sn )k = k[ρT g −1 PK (tn ) ⊕ (PK (tn ) + ρT (sn )]

⊕[ρT g −1 PK (tn−1 ) ⊕ (PK (tn−1 ) + ρT (sn−1 )]k
= k[ρT g −1 PK (tn ) ⊕ ρT g −1 PK (tn−1 )]
⊕[(PK (tn ) + ρT (sn )) ⊕ (PK (tn−1 ) + ρT (sn−1 ))]k
≤ k[ρT g −1 PK (tn ) − ρT g −1 PK (tn−1 )]k
+k[(PK (tn ) + ρT (sn )) − (PK (tn−1 ) + ρT (sn−1 ))]k
= ρkT g −1 PK (tn ) − T g −1 PK (tn−1 )]k
+k[(PK (tn ) + ρT (sn )) − (PK (tn−1 ) + ρT (sn−1 ))]k
= ρkT g −1 PK (tn ) − T g −1 PK (tn−1 )]k
+k(PK (tn ) − PK (tn−1 )) + ρ(T (sn ) − T (sn−1 )k.
(4.6)

Using λT -Lipschitz continuity of T , non-expansiveness of the projection

operator and Cauchy-Bunyakovsky-Schwarz inequality, we have
k(PK (tn ) − PK (tn−1 )) + ρ(T (sn ) − T (sn−1 ))k2

= kPK (tn ) − PK (tn−1 )k2 + 2hρ(T (sn ) − T (sn−1 )), PK (tn ) − PK (tn−1 )i
+ρ2 kT (sn ) − T (sn−1 )k2
≤ ktn − tn−1 k2 + 2ρλT ksn − sn−1 kktn − tn−1 k + ρ2 λ2T ksn − sn−1 k2
= (ktn − tn−1 k + ρλT ksn − sn−1 k)2 ,

thus, we have

k(PK (tn )−PK (tn−1 ))+ρ(T (sn )−T (sn−1 ))k ≤ ktn −tn−1 k+ρλT ksn −sn−1 k.
(4.7)
156 ZAHOOR AHMAD RATHER AND RAIS AHMAD

Since T is λT -Lipschitz continuous, the projection operator is non-expansive

and using Bounded Inverse Theorem 2.6 for g, we have

kT g −1 (PK (tn )) − T g −1 (PK (tn−1 ))k ≤ λT kg −1 (PK (tn ) − PK (tn−1 )k

≤ λT kg −1 kkPK (tn ) − PK (tn−1 )k
≤ λT δg ktn − tn−1 k. (4.8)

Combining (4.7) and (4.8) with (4.6), we have

kg(sn+1 ) ⊕ g(sn )k ≤ ρλT δg ktn − tn−1 k + ktn − tn−1 k + ρλT ksn − sn−1 k
= (1 + ρλT δg )ktn − tn−1 k + ρλT ksn − sn−1 k. (4.9)

Using (4.3) of Algorithm 4.1, we have

ktn − tn−1 k = k[g(sn ) − ρT (sn )] − [g(sn−1 ) − ρT (sn−1 )]k

= kg(sn ) − g(sn−1 ) − ρ(T (sn ) − T (sn−1 ))k. (4.10)

Since g is λg -Lipschitz and T is λT -Lipschitz, we have

kg(sn ) − g(sn−1 ) − ρ(T (sn ) − T (sn−1 ))k2

= kg(sn ) − g(sn−1 )k2 + 2ρhg(sn ) − g(sn−1 ),

T (sn ) − T (sn−1 )i + ρ2 kT (sn ) − T (sn−1 )k2
≤ λ2g ksn − sn−1 k2 + 2ρkg(sn ) − g(sn−1 )k
kT (sn ) − T (sn−1 )k + kT (sn ) − T (sn−1 )k2
≤ λ2g ksn − sn−1 k2 + 2ρλg λT ksn − sn−1 k2
+ρ2 λ2T ksn − sn−1 k
= (λ2g + 2ρλg λT + ρ2 λ2T )ksn − sn−1 k2 .

That is

kg(sn ) − g(sn−1 ) − ρ(T (sn ) − T (sn−1 ))k ≤ P (θ)ksn − sn−1 k, (4.11)

q
where P (θ) = λ2g + 2ρλg λT + ρ2 λ2T .
Combining (4.10) with (4.11), we have

ktn − tn−1 k ≤ P (θ)ksn − sn−1 k. (4.12)

Using (4.12), (4.9) becomes

kg(sn+1 ) ⊕ g(sn )k ≤ (1 + ρλT δg )P (θ)ksn − sn−1 k + ρλT ksn − sn−1 k. (4.13)

 SOLVING WIENER-HOPF EQUATION INVOLVING XOR-OPERATION 157

Since g(sn+1 ) ∝ g(sn ), from (4.13), we have

kg(sn+1 ) − g(sn )k ≤ (1 + ρλT δg )P (θ)ksn − sn−1 k + ρλT ksn − sn−1 k. (4.14)

where P (θ) is defined in (4.11).

Since g is strongly monotone with constant ξg , we have

kg(sn+1 ) − g(sn )k ≥ ξg ksn+1 − sn k

1
that is ksn+1 − sn k ≤ kg(sn+1 ) − g(sn )k. (4.15)
ξg
Using (4.15), (4.14) becomes
1
ksn+1 − sn k ≤ [(1 + ρλT δg )P (θ) + ρλT ]ksn − sn−1 k,
ξg
ksn+1 − sn k ≤ ξ(θ)ksn − sn−1 k, (4.16)
q
where ξ(θ) = ξ1g [(1+ρλT δg )P (θ)+δλT ] and P (θ) = λ2g + 2ρλg λT + ρ2 λ2T .
By (4.4), we have ξ(θ) < 1, thus from (4.16) it follows that {sn } is a Cauchy
sequence. Since H is complete, we can suppose that sn → s ∈ H. From
(4.12), it follows that {tn } is also a Cauchy sequence, thus there exist t ∈ H
such that tn → t. Thus by Lemma 3.1 and Lemma 3.2, we conclude that
s, t ∈ H is a solution of Wiener-Hopf equation involving XOR-operation
(3.1). 

Conclusion
In conclusion, this study introduces a novel application of the Wiener-
Hopf equation by incorporating the XOR-operation. The research demon-
strates that the Wiener-Hopf equation involving XOR-operation is math-
ematically equivalent to a variational inequality problem. To effectively
address this equation, a dedicated iterative algorithm is proposed, specifi-
cally designed for XOR-based Wiener-Hopf equations. The application of
the Bounded Inverse Theorem enables the successful derivation of the solu-
tion.
Furthermore, the study extensively discusses the convergence criteria
associated with this tailored approach. These findings not only advance
the understanding of the Wiener-Hopf method in XOR-based scenarios but
also establish a powerful framework for solving intricate mathematical chal-
lenges in diverse applications. The proposed iterative algorithm and con-
vergence analysis contribute significantly to the potential application of the
Wiener-Hopf equation in various domains, promising new avenues for future
158 ZAHOOR AHMAD RATHER AND RAIS AHMAD

research and practical implementations.

Acknowledgement: The authors are grateful to the anonymous refree for

his favourable evaluation and suggestions for improvements.

References
[1] Brezis, H., Operateurs Maximaux Monotone et Semigroups de Contractions dans
les Espaces de Hilbert, North-Holland,Amsterdam, 1973.
[2] Cottle, R.W., Giannessi, F., and Lions, J.L. , Variational Inequalities and Com-
plementarity Problems: Theory and Applications, J. Wiley and Sons, New York,
1980.
[3] Crank, J., Free and Moving Boundary Problems, Clarendon press, Oxford, U.K.
1984.
[4] Du, Y.H., Fixed points of increasing operators in ordered Banach spaces and appli-
cations , Appl. Anal., 38, 1-20(1990).
[5] Abrahams, I.D., A brief and Gentle Introduction to the Wiener-Hopf technique, De-
livered to the summer school of the Isaac Newton Institute Programme on “Bringing
pure and applied analysis together via the Wiener-Hopf technique", 2019.
[6] Giannessi, F., and Maugeri,A., Variational inequalities and Network Equilibrium
Problems, Plenum Press, New York, 1995.
[7] Glowinski, R., Lions, J.L., and Trémolières, R., Numerical Analysis of Variational
inequalities, North-Holland, Amsterdam, 1981.
[8] Harker, P.T., and Pang, J.S., Finite-dimensional variatioal inequalities and comple-
mentarity problems: a survey of theory, algorithms and applications, Math. Pro-
gramming 48(1990), 161-220.
[9] Haubruge, S., Nguyen, V.H. and Strodiot, J.J., Convergence analysis and applica-
tions of the Glowinski-Le Tallec splitting method for finding a zero of the sum of
two maximal monotone operators, J. Optim. Theory Appl. 97(1998), 645-673.
[10] He, B., A class of projection and contraction methods for monotone variational
inequalities, Appl. Math. Optim., 35(1997), 69-76.
[11] He, B., A class of new methods for monotone variational inequalities, Preprint,
Institute of Mathematics, Nanjing University, Nanjing, China, 1995.
[12] Hyres, D.H., Isac, G. and Rassias,Th. M., Topics in Nonlinear Analysis and Appli-
cations, World Scientific Publications, Singapore, 1997.
[13] Kinderlehrer, D. and Stampacchia, G., An introduction to Variational Inequalities
and Their Applications, Academic Press, New York, (1980).
[14] Li, H.G., A nonlinear inclusion problem involving (α, λ)-NODM set-valued mappings
in ordered Hilbert space , Appl. Math. Lett. 25, 2012, 1384-1388.
[15] Li, H.G., Pan, X.B., Deng, Z.Y., Wang, C.Y., Solving frameworks involving (γG, λ)-
weak-GRD set-valued mappings in positive Hilbert spaces, Fixed Point Theory Appl.
, 2014.
 SOLVING WIENER-HOPF EQUATION INVOLVING XOR-OPERATION 159

[16] Mosco, U., Implicit variational prolems and quasi-variational inequalities, in, Non-
linear Operators and the Calculus of Variations, Lecture Notes Math. No. 543,
Springer-Verlag, Berlin, (1976), 83-126.
[17] Noor, M. A., Wiener-Hopf equations and variational inequalities, J. Optim. Theory
Appl. 79(1993), 197-206.
[18] Noor, M. A., Sensitivity analysis for quasi variational inequalities, J. Optim. Theory
Appl. 95(1997), 399-407.
[19] Pang, J.S., and Ralph, D., Piecewise smoothness, local invertibility and parametric
analysis of normal maps, Math. Opers. Research 21(1996), 401-426.
[20] Pitonyak, A., Shi, P., and Shillor, M., On an iterative method for variational in-
equalities, Numer. Math. 58(1990), 231-242.
[21] Robinson, S. M., Normal maps induced by linear transformations, Math. Opers.
Research 17(1992), 691-714.
[22] Robinson, S. M., Newton’s Method for a class of nonsmooth functions, Set-Valued
Analysis 2(1994), 291-305.
[23] Robinson, S. M., Nonsingularity and symmetry for linear normal maps, Math. Pro-
gramming 62(1993), 415-425.
[24] Robinson, S. M., Sensitivity analysis of variational inequalities by normal maps
techique, in: Variational Inequalities and Network Equilibrium Problems(eds. F.
Giannessi and A. Maugeri), Plenum Press, New York, (1995),257-269.
[25] Sellami, H.,and Robinson, S.M., Implementation of a continuous method for normal
maps, Math. Programming 76(1997), 563-578.
[26] Shi, P., Equivalence of variational inequalities with Wiener-Hopf equations, Proc.
Amer. Math. Soc. 111(1991), 339-346.
[27] Shi, P., An iterative method for obstacle problem via Green’s functions, Nonlinear
Analysis 15(1990), 339-344.

Zahoor Ahmad Rather

Department of Mathematical Sciences
Islamic University of Science and Technology
Awantipora 192122, India.
E-mail: zahoorrather348gmail.com

Rais Ahmad
Department of Mathematics
Aligarh Muslim University, Aligarh 202002, India.
E-mail: [email protected]
 The Mathematics Student ISSN: 0025 - 5742
Vol. 93, Nos. 1-2, January - June (2024), 160 - 172

WHAT IS THE VOLUME OF THE CHAMBERED

NAUTILUS SHELL ?

RICHARD D. CARMICHAEL
(Received: 25 - 08 - 2023; Revised 25 - 02 - 2024)

Abstract. Elementary calculus and geometry are used to compute

an approximation of the volume of the chambered nautilus shell.

1. Introduction

The chambered nautilus (CN) (and the nautilus in general) is a member

of the marine mollusc family Nautilidae and is found in the Indian and
Pacific oceans. The CN has a hard shell. As the CN matures it periodically
seals off the previous shell component and creates a new, continuing, and
larger living chamber. The growth process yields a spiral structure which is
consistently characterized in relation to a logarithmic or equiangular spiral.
Descriptions of the nautilus in general and of the CN and its shell can be
found in many articles which are available on-line. See articles on-line using
the computer with phrases such as "nautilus", "chambered nautilus", "nau-
tilus shell", "mathematics of the nautilus shell", "geometry of the nautilus
shell", "logarithmic spiral", and "equiangular spiral". Sketches and discus-
sion of the CN shell and of logarithmic spirals are given in many articles. In
particular [1], [2], and [3] of this paper contain relevant sketches and discus-
sion of the CN shell in relation to the logarithmic spiral. We further discuss
the logarithmic spiral concept and state a formula in polar coordinates with
corresponding discussion for logarithmic spiral below in this paper.
Recently a high school teacher of one-variable calculus inquired of this
author if a value of the volume of the CN shell could be determined math-
ematically. The teacher was using this question and its possible solution
as a project for the students to consider in the calculus class. Because of
the shape of the CN shell, this inquiry possed an interesting problem to
2020 Mathematics Subject Classification: 26A06, 51M04, 00A05, 00A08
Key words and phrases: logarithmic spiral, nautilus shell, volume, cone

©, Indian Mathematical Society, 2024 .

160
 WHAT IS THE VOLUME OF THE CHAMBERED NAUTILUS SHELL ? 161

this author. The purpose of this paper is to obtain a technique to mathe-

matically compute a good approximation to the volume of a CN shell using
one-variable calculus and intuitive geometry. The author has related the
contents of this paper to the above noted inquiring high school teacher.

2. Volume of a CN Shell

A logarithmic spiral in the Euclidean plane is defined by the rule that

for a given rotation angle, such as one revolution or 2π radians, the distance
from the pole (origin or spiral origin) to a point on the spiral is multiplied by
a fixed real constant. When this fixed multiplication amount is the "golden
ratio" of (1+51/2 )/2 a particular type of logarithmic spiral is formed. See [3]
and references in [3] for a discussion of the questioned relationship between
the shape of the CN shell and the logarithmic spiral based on the "golden
ratio".
As described in analysis [2, p. 733] the general formula for the logarith-
mic spiral in the plane in polar coordinates (r, θ) is r = αekθ , k 6= 0. To
be specific with respect to the CN shell we take the polar equation for the
associated logarithmic spiral to be

r = αecot(β)θ (1)
as in reference [1]. Here α > 0 and β, 0 < β < π/2, are arbitrary but fixed
real constants. As α and β appropriately vary as fixed constants, different
logarithmic spirals are sketched in the plane. Since we take β, 0 < β < π/2,
then the cotangent of β is greater than 0, cot(β) > 0, and the logarithmic
spiral is outward or a counter clockwise spiral. As the constant α > 0
increases the associated spiral expands. For a fixed constant α > 0 the
fixed constant β, 0 < β < π/2, is the angle between the radial line outward
from the origin of the spiral (center of the plane) and the tangent lines to
each associated spiral (for all points on the spirals corresponding to the fixed
α > 0) with the view of the angle from above the spiral. (We emphasize
that the angle β will be determined by the contour of the nautilus shell.
In a given logarithmic spiral the angle β is constant at all points on the
spiral; see [2, problem 37, p. 784] and [1]. For a fixed value of α, different
values of β can occur for different logarithmic spirals. See the sketches in
[1].) See reference [1] for various sketches of logarithmic spirals associated
162 RICHARD D. CARMICHAEL

with given specific values of α and β. See also sketches in many articles
noted previously that can be obtained on-line.
Before beginning the description of our process to calculate the volume
of the CN shell we suggest that the reader note the illustrations given at
the end of this paper after the references; these illustrations will be helpful
in the visualization of the descriptions of our process. In the discussion that
follows in this paper Illustration I refers to the second sheet of illustrations
after the references; Illustration II refers to the third sheet of illustrations;
Illustration III refers to the fourth sheet of illustrations. The first sheet
of illustrations presents a summary of Illustrations I, II, and III. Also we
suggest that the reader note the picture of the nautilus shell in reference [3]
if possible. This picture clearly displays the spiral form of the nautilus shell,
the form of the constructed chambers, and the relationship of the boundary
of the shell in relation to the spirals.
Using the model of equation (1) for logarithmic spiral we now obtain
an approximation of the volume of the CN shell. To begin our process to
calculate the volume first we visualize a line (curved line or arc) starting
at the spiral origin of the CN shell and proceeding through the spirals at
the center of the spirals until reaching the end of the chambers (Illustration
I). Another manner to visualize this curved line (or arc) is to cut the shell
in half thus displaying the logarithmic spiral nature of the CN shell. The
curved line (or arc) previously described is equivalently obtained by again
starting at the spiral origin of one-half of the CN shell and following the
spiral of the shell through the center of the spiral structure until reaching
the end of the last one-half chamber. This curved line (or arc) is then a
logarithmic spiral against the plane face of the cutaway of the one-half of
the CN shell. Let (r, θ) be the polar coordinates of this plane face of the
cutaway of the shell. Thus for the described curved line (or arc), which is
a logarithmic spiral, there is a fixed real number α > 0 and a fixed real
number β, 0 < β < π/2, such that the curved line (or arc) is given in
(r, θ) form on the plane face as equation (1). A given CN shell could have
numerous spirals of 2π radians each before the end of the shell is reached.
We call the positive integer n to be the number of spirals of 2π radians
each of the CN shell; thus the θ variable in the equation (1) will vary over
0 ≤ θ ≤ 2πn. The opening at the end of the CN shell is approximately
circular in nature; we approximate this ending as a circle of radius fixed
 WHAT IS THE VOLUME OF THE CHAMBERED NAUTILUS SHELL ? 163

R > 0 (Illustration I). This circular opening at the end of the CN shell is
where the nautilus emerges.
The question now is how to put all of this information together con-
cerning the logarithmic spiral r = αecot(β)θ , 0 ≤ θ ≤ 2πn, with α > 0 and
β, 0 < β < π/2, both fixed and with the positive integer n fixed and with
the fixed radius R > 0 at the end of the spirals in order to compute the
volume of the CN shell. Envision the boundary of the winding chambers
as they are constructed on top of the previous chamber as a type of clay
or putty. By the construction of the CN shell in the form of a spiral, if
the chambers are unwound (Illustration II) and the object is placed upright
with the circular opening of the shell at the bottom and the spiral origin at
the top, the resulting object is approximately a right circular cone (Illus-
tration III). We know that the volume V of such a cone is V = (π/3)q 2 h
where q is the radius of the base and h is the height of the cone.
Before proceeding with the visualization of the resulting right circular
cone in the preceding paragraph let us further describe the process to "un-
wind" the chambers in order to obtain a right circular cone as stated in the
preceding paragraph. To visualize this process let us "unwind" the logarith-
mic spiral that defines the shape of the CN shell and that is a line (curved
line or arc) starting at the spinal origin of the CN shell and proceeding
through the chambers at the center of the chambers until reaching the end
of the chambers. This line (curved line or arc) is a logarithmic spiral which
we assume as above to be an outward spiral or a counter clockwise spiral.
Assume now that we have such a spiral which is constructed from flexible
and bendable wire. Place the wire on a plane surface and fix the end point
of the wire that coincides with the spinal origin of the curved line. Now
grasp the other end of the wire and turn the end in the clockwise direction
on the plane surface until all of the spirals are "unwound" (Illustration II);
the result will be a straight line wire with one end at the spinal origin and
the other end coinciding with the tip end of the spiral. This is the exact
process to "unwind" the chambers of the CN shell as described in the pre-
ceding paragraph. (The spinal origin of the CN shell corresponds to the
spinal origin of the curved line. The opening end of the CN shell corre-
sponds to the other end of the wire that is to be grasped.) When the CN
shell is "unwound" by clockwise rotation parallel to a plane the resulting
object will be approximately a right circular cone with the spinal origin as
164 RICHARD D. CARMICHAEL

the tip of the cone and the circular end of the CN shell as the base of the
cone (Illustration III).
Let us now proceed in our analysis of the relationship between the CN
shell and the right circular cone. In the present case the radius of the base
of our constructed approximate cone from the CN shell is the above stated
fixed R > 0, the radius of the opening at the end of the CN spirals as noted
above (Illustrations I and III). The height of our constructed approximate
cone from the CN shell is the arc length of the above described logarithmic
spiral curved line of form (1) for the fixed α > 0 and fixed β, 0 < β < π/2,
with 0 ≤ θ ≤ 2πn for a fixed positive integer n. That is, in transforming the
CN shell to the approximate right circular cone as described, the described
logarithmic spiral becomes the height line of the cone (Illustrations I and
III). This height line has length equal to the arc length of the logarithmic
spiral following the middle of the chambers from the spiral origin to the end
of the chambers of the CN shell as described above (Illustrations I and III).
(The spirals of the boundary of a CN shell will not be smooth like that of
the boundary of a right circular cone. Thus in transforming the CN shell to
the cone by the unwinding process as described above we call the resulting
object an approximate cone. The resulting object will be in the shape of a
right circular cone, and the volume of the resulting object will be a close
approximation to the volume of the CN shell. Thus we use the formula
for the volume of a right circular cone below to approximate the volume of
the resulting cone like object and hence of the CN shell.) h The arc length
of the above described logarithmic spiral of form (1) can be computed by
calculus using, for example, the integral formula in [2, p. 740] or [4, p. 697]
for arc length in polar coordinates. Thus the height h of our constructed
approximate cone from the CN shell can be computed as follows and is the
arc length of the described logarithmic spiral r = αecot(β)θ , 0 ≤ θ ≤ 2πn,
where α > 0 and β, 0 < β < π/2, and the positive integer n are fixed for
the specific CN shell. We have from [2, p. 740] or [4, p. 697]
 WHAT IS THE VOLUME OF THE CHAMBERED NAUTILUS SHELL ? 165

Z 2πn
dr 2 1/2
h = arc length = (r2 + ( ) ) dθ (2)
0 dθ
Z 2πn
= (α2 e2cot(β)θ + α2 (cot(β))2 e2cot(β)θ )1/2 dθ
0
Z 2πn
= (α2 e2cot(β)θ (1 + (cot(β))2 ))1/2 dθ
0
Z 2πn
2 1/2
= α(1 + (cot(β)) ) ecot(β)θ dθ
0
α(1 + (cot(β))2 )1/2 2πncot(β)
= (e − 1).
cot(β)
Thus the volume V of a given CN shell having logarithmic spiral form
r = αecot(β)θ for fixed α > 0 and β, 0 < β < π/2; having positive integer
n spirals of 2π radians each; and having an approximate circle of radius
R > 0 at the end (or base) of the spirals is thus approximately

πR2 α(1 + (cot(β))2 )1/2 2πncot(β)

V = (e − 1) (3)
3cot(β)
cubic units.
Since 0 < β < π/2 and n = 1, 2, 3, 4, ... in (3), the formula for V yields
a positive number; and both intuitively and through the formula (3), the
volume V increases in value as n increases.
As stated in section 2 we chose β in (1) to be such that 0 < β < π/2
thus making cot(β) > 0 in (1) and yielding the logarithmic spiral to be
outward or a counter clockwise spiral. We then proceeded to obtain the
volume formula in (3) for the CN shell. Mathematically we can obtain a
value as in (3) from the logarithmic spiral equation (1) for π/2 < β < π
and hence cot(β) < 0; in this case the logarithmic spiral (1) is inward or a
clockwise spiral. Now assuming π/2 < β < π so that cot(β) < 0 we start
at (1) and proceed with the same procedure and analysis to compute the
corresponding values as in (2) and (3). In this case corresponding to (2) we
obtain

α(1 + (cot(β))2 )1/2

h = arc length = (1 − e2πncot(β) ); (4)
−cot(β)
and corresponding to (3) we have
166 RICHARD D. CARMICHAEL

πR2 α(1 + (cot(β))2 )1/2

V = (1 − e2πncot(β) ) (5)
−3cot(β)
cubic units.
For the case π/2 < β < π and thus cot(β) < 0 notice that

α(1 + (cot(β))2 )1/2

lim h = (6)
n→∞ −cot(β)
and

πR2 α(1 + (cot(β))2 )1/2

lim V = . (7)
n→∞ −3cot(β)
These limits yield mathematical upper bounds on the height (arc length) h
and the volume V for the case π/2 < β < π and cot(β) < 0. (In the case
that π/2 < β < π and hence cot(β) < 0, the CN shell and its associated
logarithmic spiral is inward or a clockwise spiral. In this case the unwind-
ing process described in Section 2 would be done in the counter clockwise
direction to transform the CN shell into a right circular cone.)

3. Computation Examples

We consider the three logarithmic spirals in [1, p. 22] in which three

sets of values for α and β are stated with an accompanying sketch for each
of the α, β pairs. We recommend observing these sketches in [1, p. 22]
as we have the following discussion. For each of these three α, β pairs the
logarithmic spiral is of the type leading to formula (3) for the volume of
the associated CN shell; each of these spirals defines the shape of the asso-
ciated CN shell. In each of these cases the stated β value, which is stated
in radians, satisfies 0 < β < π/2 in radians; hence the sprial rotation is
counter clockwise in each case. For each of these three α, β pairs we state
the number n of rotations of 2π radians that we consider and a value of the
radius R of the CN shell opening at the bottom of the corresponding right
circular cone and use equation (3) to compute the approximating volume of
the associated CN shell for each of the three pairs of α, β. We leave the com-
putational details to compute each of the volumes to the reader if desired;
we have stated the three approximate volumes that we have obtained below.
 WHAT IS THE VOLUME OF THE CHAMBERED NAUTILUS SHELL ? 167

For α = 1.2, β = 1.4 radians, n = 4, and R = 7 centimeters (cm) we have

V = 29403.776 cm3 .

For α = 0.8, β = 1.4 radians, n = 4, and R = 6 cm we have V = 14401.692

cm3 .

For α = 0.8, β = 1.45 radians, n = 4, and R = 5 cm we have V = 3606.460

cm3 .

Comparing the first two spirals we have β = 1.4 radians is the same for
both , but α = 1.2 is larger than α = 0.8 for the second spiral. For the
same β, as α increases the associated spiral expands as we have mentioned
before; hence the volume in the associated CN shell expands as α increases
for the same β. Because of this the radius at the end of the CN shell gets
larger as α increases for the same number n of 2π radian revolutions. This
is the reason for choosing R = 7 cm for the first spiral and R = 6 cm for the
second spiral. The volume for the associated first CN shell is larger than
that for the second associated CN shell for the same n = 4 revolutions of
2π radians each. (For each of the three sets of values for α, β, n, and R
the corresponding spiral is the mid line for the corresponding CN shell as
described before.)
In comparing the second and third spirals, α = 0.8 in each whereas
β = 1.4 radians in the second spiral while β = 1.45 radians in the third
spiral. When β increases for the same α value the spirals are compressed (or
flattened) and the amount of volume in the associated CN shell is decreased.
Seen otherwise, as the angle β decreases for the same α the tangent lines to
the spiral have increasing positive slope which creates larger volume in the
CN shell. For this reason R = 5 cm was chosen corresponding to β = 1.45
radians while R = 6 cm was chosen corresponding to β = 1.4 radians for the
same number n = 4 revolutions of 2π radians each. The volume of the CN
shell corresponding to the second spiral is larger than that corresponding
to the third spiral for the same n = 4 revolutions of 2π radians each.
In comparing the first spiral with the third spiral, the first spiral has the
advantage for larger volume of the CN shell in both the value of α and the
value of β as seen in the commentary of the previous two paragraphs. Thus
the volume of the associated CN shell for the first spiral would naturally be
168 RICHARD D. CARMICHAEL

much larger than that for the third spiral for the same number of 2π radian
revolutions in each.

Dedication: This paper is dedicated in remembrance of the author’s re-

search colleague Dr. Ram Shankar Pathak, Banaras Hindu University,
Varanasi, India.

References
[1] Allen, Ashley. Mathematical modeling of the nautilus shell, Undergrad-
uate Mathematics Exchange 1, No. 1 (2003), 21 - 23.
https://digitalresearch.bsu.edu/2021/02/Mat./mathexchange(Vol.1,No.1)

[2] Anton, H., Bivens, I., Davis, S. Calculus, 7th ed., John Wiley and Sons,
2002.

[3] Peterson, Ivars. Sea shell spirals.

https://sites.millersville.edu/Math.457/nautilus

[4] Stewart, J., Clegg, D., Watson, S. Calculus (Early Transcendentals),

Ninth Edition, Cengage Learning Inc., 2021.

Dr. Richard D. Carmichael

Department of Mathematics
Wake Forest University
Wiston-Salem, NC 27109, U. S. A.
E-mail: [email protected]
169
170
171
172
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 173–180

UNIQUENESS OF MEMORYLESS DISTRIBUTIONS

ON PERIODIC TIME SCALES

ROBERT J. NIICHEL, SCOTT FERRELL, AND CONNOR JOKERST

Abstract. The classical memoryless property induces probability dis-

tributions for both discrete and continuous random variables. On the
positive real line, the only continuous random variables with the mem-
oryless property are the exponentials, while on N the geometric random
variables are the only memoryless ones. In their recent paper, Cuchta
and Niichel explored various memoryless properties, leaving an open
problem as to whether a certain type of “memoryless” criterion on pe-
riodic time scales implies a unique family of distributions. We answer
that question in the affirmative herein.

1. Introduction

The memoryless property in the classical literature defines a random

variable X to be “memoryless" if whenever t, s > 0 and t, s are in the
sample space,
P (X > s + t|X > t) = P (X > s), (1.1)
or equivalently,

P (X > s + t) = P (X > t)P (X > s). (1.2)

This property gives rise to two different families of distributions. On R+ ,

the exponential is the only continuous random variable with the memory-
less property, while on N0 , the geometric random variables are the only
memoryless ones.
Time scales calculus has developed significantly over the past several
decades, and a description of the most salient results can be found in [?].
A time scale T is an arbitrary nonempty closed subset of R. The function
gX (t) = g(t) = P (X > t) is called the survival function of X. Throughout
this paper, we will assume that inf T = 0, and similar to the exponential

2010 Mathematics Subject Classification: 60a05

Key words and phrases: Memoryless property, time scales, commensurable

© Indian Mathematical Society, 2024 .

173
174 ROBERT J. NIICHEL, SCOTT FERRELL, AND CONNOR JOKERST

and geometric, the survival function g(t) always satisfies g(0) = 1. A time
scale is said to be periodic with period ω > 0 if whenever t ∈ T, t + ω ∈ T.
We say that T is n-periodic with periods ω1 , . . . ωn > 0 if whenever t ∈ T
and m1 , . . . , mn ∈ N0 ,
Xn
t+ mj ωj ∈ T.
j=1
With multiple periods in play, there is the possibility that some of the
periods will have common multiples. We say that two periods ω1 and ω2
are commensurable if ω1 /ω2 ∈ Q; otherwise, ω1 and ω2 are said to be non-
commensurable.
The main problem with the memoryless property on time scales is the
potential for lacuna in the time scale itself. That is, t, s ∈ T does not
imply that t + s ∈ T. A number of alternative memoryless properties have
been introduced, including those by Poulsen et al. [?, Definition V.2], and
Matthews [?, Definition 34]. The definition by Poulsen et al. is pertinent
to this paper:

∀t ∈ T P (X > t + ω|X > t) = P (X > ω). (1.3)

Note that this definition of memorylessness eliminates the freedom of choice

of s in the classical definition, (1.1); now only ω is allowed in this position.
It is straightforward to show that (1.3) implies

∀t ∈ T, k ∈ N0 P (X > t + kω|X > t) = P (X > ω)k . (1.4)

Cuchta and Niichel [?] have recently explored the implications of these
new definitions. They also presented a modified version of the memoryless
property (1.3) with (1.4) for n-periodic time scales:
n n
!
X Y
P X >t+ mk ωk = P (X > t) P (X > ωk )mk , (1.5)
k=1 k=1

where t ∈ T and m1 , . . . , mn ∈ N0 . In that paper, they also left open the

following question:

Let ω1 < ω2 be non-commensurable real numbers and con-

sider the time scale T = {n1 ω1 +n2 ω2 |n1 , n2 ∈ N0 }. Suppose
that the random variable X has the memoryless property
(1.5), so defining g(x) = P (X > x) yields g(n1 ω1 + n2 ω2 ) =
 UNIQUENESS OF MEMORYLESS DISTRIBUTIONS ON PERIODIC TIME SCALES175

g(ω1 )n1 g(ω2 )n2 . If we choose g(ω1 ) ∈ (0, 1), then what re-
strictions are forced upon the value of g(ω2 )?
They conjectured that g(ω2 ) = g(ω1 )ω2 /ω1 is necessary because of the mem-
oryless property (1.5) together with the non-commensurability of ω1 and
ω2 , but did not prove it. We have taken up that task and shown it to be
true. We also explore a few other related questions.

2. The Commensurable Case

Our first result pertains to commensurable periods:

Theorem 2.1. If g(t) is the survival function of a memoryless random

variable on a 2-periodic (or “doubly” periodic) time scale T = {m1 ω1 +
m2 ω2 |m1 , m2 ∈ N0 } with commensurable periods ω1 and ω2 , then g(ω2 ) =
g(ω1 )ω2 /ω1

Proof. Since ω1 and ω2 are commensurable, there exist m1 , m2 ∈ N such

that m1 ω1 = m2 ω2 , implying that m1 /m2 = ω2 /ω1 . Hence by (1.5),

g(m2 ω2 ) = g(m1 ω1 ),
g(ω2 )m2 = g(ω1 )m1 ,
g(ω2 ) = g(ω1 )m1 /m2 ,
g(ω2 ) = g(ω1 )ω2 /ω1 ,

completing the proof. 

nP o
n
If T = m
j=1 j jω |m 1 , . . . , m n ∈ N 0 is an n-periodic time scale with
periods 0 < ω1 < ω2 < · · · < ωn , all of which are commensurable with ω1
(and hence with each other), by the same proof as in Theorem 2.1, it must
be the case that the survival function g satisfies g(ωj ) = g(ω1 )ωj /ω1 . This
implies part of the following result.
nP o
n
Theorem 2.2. If T = m
j=1 j j ω |m 1 , . . . , m n ∈ N 0 is an n-periodic
time scale with periods 0 < ω1 < ω2 < · · · < ωn , all of which are commen-
surable with ω1 then
g(ωj ) = g(ω1 )ωj /ω1 (2.1)
and therefore g(t) = g(ω1 )t/ω1 for all t ∈ T.
176 ROBERT J. NIICHEL, SCOTT FERRELL, AND CONNOR JOKERST

Proof. The discussion before the theorem proves that g(ωj ) = g(ω1 )ωj /ω1 .
It must be shown that g(t) = g(ω1 )t/ω1 for all t ∈ T. To that end, suppose
that t = nj=1 mj ωj . By (1.5),
P

n
Y
g(t) = g(0) g(ωj )mj
j=1
n
Y
= g(ω1 )ωj mj /ω1 (2.2)
j=1
−1 Pn
= g(ω1 )ω1 j=1 ωj mj

= g(ω1 )t/ω1 ,
which completes the proof. 

Theorem 2.2 implies the following:

Corollary 2.3. Given an n-periodic time scale

 
X n 
T= mj ωj |m1 , . . . , mn ∈ N0
 
j=1

with pairwise commensurable periods ω1 < · · · < ωn , if a T-valued ran-

dom variable X satisfies (1.5), then it is part of a one-parameter family of
distributions on T.

3. The Non-Commensurable Case

Turning now to a doubly-periodic time scale T = {m1 ω1 +m2 ω2 |m1 , m2 ∈

N0 } with ω1 and ω2 non-commensurable, is it still true that g(ω2 ) =
g(ω1 )ω2 /ω1 ? This is the question posed by Cuchta and Niichel. If so, it would
imply that Theorem 2.3 can be extended to encompass non-commensurable
periods as well. Indeed, the proof of Theorem 2.2 did not require anything in
particular about the periods, just that the survival function satisfied (2.1).
The proof of the non-commensurable case proved more slippery. How-
ever, there were good reasons to believe it was true. We ran simulations
in Python using Tom Cuchta’s timescalescalculus package.1 We first
generated the values of a doubly-periodic time scale using the two periods
p p
ω1 = 1/10 and ω2 = 1/2. We then assigned different values to g(ω1 )
and g(ω2 ); values of g on the rest of the time scale were then determined
1Commit 5252076. See https://github.com/tomcuchta/timescalecalculus
 UNIQUENESS OF MEMORYLESS DISTRIBUTIONS ON PERIODIC TIME SCALES177

using the property (1.5). We then graphed g(t), assigning the color orange
to a point ti if g increased from the previous point in the time scale (thus
indicating that g fails to be non-increasing, and so cannot be a survival
function). Several of these graphs are below:

Figure 1. g(ω1 ) = 0.9; g(ω2 ) = 0.99. The graph fails to be

non-increasing. Clearly, g(ω2 ) should not be larger than
g(ω1 ).

Figure 2. g(ω1 ) = 0.9; g(ω2 ) = 0.53. The graph fails to

be non-increasing, even though g(ω1 ) > g(ω2 ). Arbitrary
values of g(ω2 ) that are less than g(ω1 ) are not guaranteed
to work.

These graphs appear to indicate that when g(ω2 ) is not the conjectured
value, different “threads” appear in the graph. When g(ω1 ) and g(ω2 ) are
closer a similar phenomenon occurs, though it is more subtle and it can
take longer for the first increasing value to appear.

Theorem 3.1. Let T = {m1 ω1 + m2 ω2 |m1 , m2 ∈ N0 } be a doubly-periodic

time scale with non-commensurable periods ω1 < ω2 . If a random variable
178 ROBERT J. NIICHEL, SCOTT FERRELL, AND CONNOR JOKERST

√
Figure 3. g(ω1 ) = 0.9; g(ω2 ) ≈ 0.9 5 . This is the conjec-
tured value for g(ω2 ). The graph appears to be non-
increasing.

satisfies (1.5) on T, then the survival function g(t) of the random variable
satisfies g(ω2 ) = g(ω1 )ω2 /ω1 .

ω2
Proof. By way of contradiction, assume g(ω2 ) 6= g(ω1 ) ω1 . For x∈ {nω1 |n ∈
1 x 1 y
N0 }, g(x) = g(ω1 ) ω1 and similarly for y = mω2 , g(y) = g(ω2 ) ω2 .
ω2  1   1 
Since g(ω2 ) 6= g(ω1 ) ω1 , g(ω1 ) ω1 6= g(ω2 ) ω2 . We now have need of the
following lemma:

Lemma 3.2. Let 0 < a < b < 1 and consider the curves ax and bx
∗
on [0, ∞). Let G(x) = x∗ − x, where x∗ is such that bx = ax . Then
limx→∞ G(x) = ∞ monotonically.

Proof of lemma 5. Compute:

∗
bx = ax
∗
ln(bx ) = ln(ax )
x∗ ln(b) = x ln(a)
ln(a)
x∗ = x
ln(b)
 UNIQUENESS OF MEMORYLESS DISTRIBUTIONS ON PERIODIC TIME SCALES179

Since 0 < a < b < 1, ln a < ln b < 0. This implies that ln a/ ln b > 1, hence,

lim G(x) = lim x∗ − x

x→∞ x→∞
ln(a)
= lim x −x
x→∞ ln(b)
 
ln(a)
= lim −1 x
x→∞ ln(b)

= ∞,

completing the proof of lemma 5. 

 1   1 
Suppose that g(ω1 ) ω1 > g(ω2 ) ω2 . Since the limit in lemma 3.2 is
monotonic, and using the notation from that theorem, there is an M ∈ N
such that G(M ω2 ) > ω1 . There is therefore a multiple of ω1 , say N ω1 ,
such that M ω2 < N ω1 < (M ω2 )∗ . However, this implies that g(N ω1 ) =
 1 N ω1  1 M ω2
g(ω1 ) ω1 > g(ω2 ) ω2 = g(M ω2 ), which in turn implies that g
increases, a contradiction
 to the fact that g isa survival function.
1  1
The argument for g(ω1 ) ω 1 < g(ω2 ) ω2 is symmetric, and the proof
is complete. 

We now know that for a 2-periodic time scale T, whether or not a period
is commensurable or non-commensurable with ω1 , it must be the case that
g(ωj ) = g(ω1 )ωj /ω1 . Thus we have the following extension of Corollary 2.3:

Corollary 3.3. Given an n-periodic time scale

 
X n 
T= mj ωj |m1 , . . . , mn ∈ N0
 
j=1

with periods ω1 < · · · < ωn , if the T-valued random variable X satisfies

(1.5), then it is part of a one-parameter family of distributions on T.

Proof. The proof follows from Theorems 2.1 and 3.1; each period is consid-
ered individually in terms of its relationship to ω1 . Each of them, whether
commensurable or non-commensurable, must satisfy g(ωj ) = g(ω1 )ωj /ω1 .
Using a similar argument as that in Theorem 2.2 (i.e., the progression (2.2)),
this implies that for every element t ∈ T, g(t) can be expressed in terms of
g(ω1 ) alone. 
180 ROBERT J. NIICHEL, SCOTT FERRELL, AND CONNOR JOKERST

Acknowledgement: We are grateful to the referee for the comments which

improved the quality of the paper.This project was supported by National
Science Foundation grant DMS-2150226.

Robert J Niichel
Department of Computer Science and Mathematics
Fairmont State University,
Fairmont, WV, USA
E-mail: [email protected]

Scott Ferrell
Shawnee State University

Connor Jokerst
Lindenwood University
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 181–190

GENERALIZED RICCI SOLITONS

AJIT BARMAN
(Received : 13 - 09 - 2023 ; Revised : 04 - 04 - 2024)

Abstract. In this paper, we have investigated the condition for which

the generalized Ricci soliton will not be an Einstein manifold and will
be Ricci-semi-symmetric whenever the vector field is a conformal killing
vector field. We studied the effects of the electromagnetic field, per-
fect fluids and charged fluids of the energy-momentum tensor on the
generalized Ricci soliton without cosmological constant. Finally, we
derive the isotropic pressure without cosmological constant fluid in the
generalized Ricci soliton from a 3-dimensional example.

1. Introduction

The importance and application of soliton in manifolds (M) are growing

day by day in the physics of relativity and cosmology and it is an impor-
tant chapter in modern geometry. In modern times, the solution of solitons,
the singularity model of geometric flow, has been considered an important
topic. Authors who have worked on solitons are Hamilton [5], Barman ([1],
[2]), Duggal and Sharma [3], Erken [4], Nurowski and Randall [8], O’Neill
[9], Ozen [10] and many others.

The Ricci tensor is a constant multiple of the Riemannian metric g and

the Ricci soliton [7] is a generalization of the Einstein metric. A Ricci soliton
can be written as a manifold such that

LV g(X, Y ) + 2S(X, Y ) + 2µg(X, Y ) = 0,

where L denotes the Lie derivative of Riemannian metric g along a vector

field V, µ is a constant and S is a Ricci tensor of type (0, 2) of the manifolds.

2020 Mathematics Subject Classification: 53C15, 53C25

Key words and phrases: Ricci soliton, generalized Ricci soliton, Conformal killing
vector field, Ricci-semi-symmetric, Einstein’s field equation, Electromagnetic field,
Perfect fluid.

© Indian Mathematical Society, 2024 .

181
182 AJIT BARMAN

Depending on whether it (µ) is negative, zero, or positive, a Ricci soliton

is said to be shrinking, stable, or expanding. Recent progress has been
made in the study of the generalized Ricci soliton [8]. The generalized Ricci
soliton of an n-dimensional semi-Riemannian manifold can be defined as

LV g(X, Y ) = −2C1 X1b (X) X1b (Y ) + 2C2 S(X, Y ) + 2λg(X, Y ) (1.1)

on a vector field X and X1b is a non-zero 1-form such that

< X, X1b >= g(X, X)

and C1 , C2 and λ are arbitrary real constants and is the Tensor products.

The generalized Ricci soliton equation (1.1) contains some important

equations in differential geometry. The equations are as follows:

1. Killing’s equation:- If C1 = C2 = λ = 0 in equation (1.1), then the

generalized Ricci soliton equation is known as Killing’s equation.

2. Equation for homotheties:- If C1 = C2 = 0, λ = 6 0 in equation (1.1),

then the generalized Ricci soliton equation is known as an equation for ho-
motheties.

3. Ricci Solitons:- If C1 = 0, C2 = −1, λ 6= 0 in equation (1.1), then the

generalized Ricci soliton equation is known as the Ricci Solitons.

1
4. Cases of Einstein-Weyl:- If C1 = 1, C2 = − n−2 , λ 6= 0 in equa-
tion (1.1), then the generalized Ricci soliton equation is known as Cases of
Einstein-Weyl.

5. Metric projective structures with skew-symmetric Ricci tensor in

1
projective class:- If C1 = 1, C2 = − n−1 , λ = 0 in equation (1.1), then the
generalized Ricci soliton equation is known as Metric projective structures
with skew-symmetric Ricci tensor in projective class.

6. Vacuum near-horizon geometry equation:- If C1 = 1, C2 = 12 , λ 6= 0

in equation (1.1), then the generalized Ricci soliton equation is known as
 GENERALIZED RICCI SOLITONS 183

the Vacuum near-horizon geometry equation.

Let (M n , g) be a n-dimensional spacetime of the generalized Ricci soli-

ton which admits a conformal killing vector field [11] V ∗ with conformal
function ρ∗ satisfying

LV g(X, Y ) = 2ρ∗ g(X, Y ), (1.2)

which reduces to a homothetic or killing vector field whenever ρ∗ is a non-

zero constant or zero respectively.

The General relativity given by Einstein’s equation [9] of the lower form
of flow is as follows:
1 ∗
S(X, Y ) − rsc g(X, Y ) + λ∗cc g(X, Y ) = κ∗gc T (X, Y ),
2
where rsc∗ is the scalar curvature, T (X, Y ) is the energy-momentum tensor

of type (0, 2), λ∗cc is the cosmological constant and κ∗gc is the gravitational
constant. Einstein’s equation gave the equation without the cosmological
constant as follows:
1 ∗
S(X, Y ) − rsc g(X, Y ) = κ∗gc T (X, Y ). (1.3)
2
We now present a brief description of two energy-momentum tensor fields.

(i) Electromagnetic field :- The electric and magnetic fields can be

combined into a single tensor field F (X, Y ) on M, that is, F (X, Y ) is a skew-
symmetric (2-form) tensor of type (0, 2). The energy-momentum tensor of
an electromagnetic field [3] is

1
T (X, Y ) = g(X, Y ) − F (X, Y ). (1.4)
4
(ii) Perfect fluids :- The energy-momentum tensor describes a Perfect
fluid [6] if
∗
T (X, Y ) = (σed + p∗ip )A(X)A(Y ) + p∗ip g(X, Y );
∗
σed + p∗ip 6= 0; σed
∗
> 0, (1.5)
∗ is the energy density and p∗ is the isotropic pressure of the fluid.
where σed ip
184 AJIT BARMAN

A perfect fluid is said to be isentropic if it is subject to a barotropic

∗ = σ ∗ (p∗ ). Cases which are of particular physical in-
equation of state σed ed ip
∗
σed
terest are the dust matter field (p∗ip = 0), the radiation field (p∗ip = 3 ) and
the stiff matter (p∗ip = σed
∗ ).

The energy-momentum tensor of the charged fluid [3] is then given

∗ 1
T (X, Y ) = (σed + p∗ip )A(X)A(Y ) + (p∗ip + )g(X, Y )
4
−F (X, Y ). (1.6)

The present paper is organised as follows: In introduction, we briefly

discussed the Ricci soliton, generalized Ricci soliton, conformal killing vec-
tor field and two energy-momentum tensor fields which are the electromag-
netic field and perfect fluids of the Einstein field equation. The condition
for which the generalized Ricci soliton will not be an Einstein manifold and
will be Ricci-semi-symmetric whenever the vector field is a conformal killing
vector field is defined in Section 2. In Section 3, we studied the generalized
Ricci soliton of the energy-momentum tensor without cosmological con-
stants and formed the relation of the electromagnetic field operator and also
the non-zero 1-form is positive of the perfect fluid with the corresponding
vector field. Finally, we derive the isotropic pressure from a 3-dimensional
example in a generalized Ricci soliton without cosmological constant fluid.

2. Conformal motions in generalized Ricci soliton

Theorem 2.1. For killing equations and equation for homothetes, the gen-
eralized Ricci soliton does not exist in Einstein manifolds whenever the vec-
tor field is a conformal killing vector field.

Proof. Combining equation (1.1) and equation (1.2), the resulting represen-
tation is

2ρ∗ g(X, Y ) = −2C1 X1b (X) X1b (Y ) + 2C2 S(X, Y )

+2λg(X, Y ). (2.1)

Since X1b is a non-zero 1-form and U ∗ is a corresponding vector field

defined by

X1b (Y ) = g(Y, U ∗ ). (2.2)

 GENERALIZED RICCI SOLITONS 185

If we substitute equation (2.2) into equation (2.1), we get exactly this

b
2C2 S(X, Y ) = (2ρ∗ + 2C1 U1∗ − 2λ)g(X, Y ). (2.3)

Equation (2.3) can be rearranged and written like this

b
(ρ∗ + C1 U1∗ − λ)
S(X, Y ) = g(X, Y ). (2.4)
C2
The proof of Theorem is completed. 

Theorem 2.2. For cases of Einstein-Weyl, metric projective structures with

skew-symmetric Ricci tensor in projective class and Vacuum near-horizon
geometry equation, the Ricci tensor of the generalized Ricci soliton is flat
whenever the vector field is a conformal killing vector field.
b b λ−ρ∗
Proof. If ρ∗ + C1 U1∗ − λ = 0 or U1∗ = C1 in equation (2.4), then this
equation will be

S(X, Y ) = 0. (2.5)
That means the Ricci tensor of the generalized Ricci soliton is flat, for cases
of Einstein-Weyl, metric projective structures with skew-symmetric Ricci
tensor in projective class and Vacuum near-horizon geometry equation. The
Theorem has proven. 

Definition 2.3. A manifold is said to be locally Ricci symmetric if it sat-

isfies
∇S = 0,
where ∇ denotes the Levi-Civita connection.

Definition 2.4. A manifold is said to be Ricci-semi-symmetric if the rela-

tion
R(X, Y ) · S = 0
holds, where R is the curvature operator.

Theorem 2.5. For cases of Einstein-Weyl, metric projective structures with

skew-symmetric Ricci tensor in projective class and Vacuum near-horizon
geometry equation, the generalized Ricci soliton is a Ricci-semi-symmetric
whenever the vector field is a conformal killing vector field.
186 AJIT BARMAN

Proof. Since every Ricci flat manifold is locally Ricci symmetric and every
locally Ricci symmetric manifold is Ricci-semi-symmetric.

Using the above analysis and Theorem 2.2, we can be written the The-
orem 2.5. 

3. Einstein’s field equation in generalized Ricci soliton

Theorem 3.1. The generalized Ricci soliton of the energy-momentum ten-

sor without the cosmological constant for the electromagnetic tensor field,
perfect fluid and charged fluid will be equation (3.2), equation (3.3) and
equation (3.4) respectively.

Proof. From equation (1.1) and equation (1.3) it follows that

∗
LV g(X, Y ) = −2C1 X1b (X) X1b (Y ) + (C2 rsc + 2λ)g(X, Y )
+2C2 κ∗gc T (X, Y ), (3.1)

which is the equation of generalized Ricci soliton of the energy-momentum

tensor without the cosmological constant.

Combining equation (1.4) and equation (3.1) shows that

∗ 1
LV g(X, Y ) = −2C1 X1b (X) X1b (Y ) + (C2 rsc + 2λ + C2 κ∗gc )g(X, Y )
2
−2C2 κ∗gc F (X, Y ),(3.2)

which is the equation of generalized Ricci soliton of the electromagnetic

tensor field without the cosmological constant.

Adding equation (1.5) and equation (3.1) together gives

∗
LV g(X, Y ) = −2C1 X1b (X) X1b (Y ) + (C2 rsc + 2λ + 2C2 κ∗gc p∗ip )g(X, Y )
+2C2 κ∗gc (σed
∗
+ p∗ip )A(X)A(Y ),(3.3)

which is the equation of generalized Ricci soliton of perfect fluid without

the cosmological constant.

Substituting equation (1.6) into equation (3.1) implies that

 GENERALIZED RICCI SOLITONS 187

∗ 1
LV g(X, Y ) = −2C1 X1b (X) X1b (Y ) + (C2 rsc + 2λ + C2 κ∗gc
2
+2C2 κ∗gc p∗ip )g(X, Y ) + 2C2 κ∗gc (σed
∗
+ p∗ip )A(X)A(Y )
−2C2 κ∗gc F (X, Y ), (3.4)

which is the equation of generalized Ricci soliton of the energy-momentum

tensor of the charged fluid without the cosmological constant.
The proof is completed. 

Theorem 3.2. If the generalized Ricci soliton resides in the energy-momentum

tensor without the cosmological constant, then the electromagnetic field op-
erator will have the relation f Y = ( 41 − p∗ip ))Y.

Proof. Subtracting equation (3.4) from equation (3.3), we obtain

1
F (X, Y ) = ( − p∗ip )g(X, Y ). (3.5)
4
Since F (X, Y ) = g(X, f Y ), where f be an electromagnetic field opera-
tor.

Equation (3.5) can improve this

1
g(X, f Y ) = ( − p∗ip )g(X, Y ).
4
Or
1
f Y = ( − p∗ip )Y.
4


Theorem 3.3. If generalized Ricci soliton lies in the energy-momentum

tensor without the cosmological constant, then a perfect fluid is isentropic.

Proof. Again subtracting Equation (3.4) from Equation (3.2), we have

p∗ip g(X, Y ) + (σed

∗
+ p∗ip )A(X)A(Y ) = 0. (3.6)

Since A(X) is a non-zero 1-form of the perfect fluid and W1 is the

corresponding vector field such that A(X) = g(X, W1 ).
Putting X = W1 in equation (3.6), it follows that

{p∗ip + σed
∗
A(W1 ) + p∗ip A(W1 )}A(Y ) = 0. (3.7)
188 AJIT BARMAN

Since A(X) 6= 0, then the equation (3.7) can written as

∗
[1 + A(W1 )]p∗ip
σed =− , (3.8)
A(W1
∗ = σ ∗ (p∗ )).
which is a barotropic equation (σed ed ip
The proof of the Theorem is completed. 

Theorem 3.4. If the dust-matter field of a perfect fluid lies in the general-
ized Ricci soliton of the energy-momentum tensor without the cosmological
constant, then the non-zero 1-form of the perfect fluid with the corresponding
vector field is positive.
∗ > 0, then equation (3.8) will be
Proof. Since σed
A(W1 )
p∗ip < .
1 + A(W1 )
For the dust matter field p∗ip = 0, the above equation is A(W1 ) > 0.
The Theorem has proven here. 

4. Example

In this section, we consider a 3-dimensional example and let M be a

3-dimensional manifold defined by

M = {(X ∗ , Y ∗ , V ∗ ) ∈ R3 },
where (X ∗ , Y ∗ , V ∗ ) are the standard coordinates in R3 . Let’s consider
3 vector fields on M as follows;

∂ ∗ ∂ ∂
E∗1 = E ∗−2V ∗
, E∗2 = ∗
, E∗3 = .
∂X ∂Y ∂V ∗
Define a Riemannian metric g by g(E∗i , E∗j ) = 0 for i 6= j, 1 ≤ i, j ≤ 3
and g(E∗i , E∗i ) = −1. Thus, the vector fields (E∗1 , E∗2 , E∗3 ) have formed
an orthonormal vector field set which is the basis of M .
Then we obtain

[E∗1 , E∗3 ] = 2E∗1 , [E∗1 , E∗2 ] = [E∗2 , E∗3 ] = 0.

Recall the classical Kozsul’s formula from the Differential geometry,

2g(∇X ∗ Y ∗ , X ∗ ) = X ∗ g(Y ∗ , Z ∗ )+Y ∗ g(X ∗ , Z ∗ )−Z ∗ g(X ∗ , Y ∗ )−g(X ∗ , [Y ∗ , Z ∗ ])

−g(Y ∗ , [X ∗ , Z ∗ ]) + g(Z ∗ , [X ∗ , Y ∗ ]),

 GENERALIZED RICCI SOLITONS 189

for all X ∗ , Y ∗ , Z ∗ on M. Using Koszul’s formula we get the following

∇E∗1 E∗1 = −2E∗3 , ∇E∗1 E∗2 = 0, ∇E∗1 E∗3 = 2E∗1 ;

∇E∗2 E∗1 = 0, ∇E∗2 E∗2 = 0, ∇E∗2 E∗3 = 0;

∇E∗3 E∗1 = 0, ∇E∗3 E∗2 = 0, ∇E∗3 E∗3 = 0.

By using the above Levi-Civita connection results, we can find the non-
zero components of the curvature tensor (R) on the manifolds as follows:

R(E∗1 , E∗3 )E∗1 = 4E∗3 ,

and
R(E∗3 , E∗1 )E∗3 = 4E ∗1 .
With the help of the above curvature tensor (R) results, we can conclude
the non-zero Ricci tensor (S) as follows:

S(E∗1 , E∗1 ) = S(E∗3 , E∗3 ) = −4.

The scalar curvature (rsc∗ ) of the manifolds will be r ∗ = S(E∗ , E∗ ) +
sc 1 1
S(E∗2 , E∗2 ) + S(E∗3 , E∗3 ) = −8.
Since the gravitational constant κ∗gc is 9.81 and using the non-zero Ricci
tensor (S) and scalar curvature (rsc ∗ ), we get the energy-momentum tensor

T as follows

T (E∗1 , E∗1 ) = T (E∗3 , E∗3 ) = −0.815494.

From the energy-momentum tensor T, we have the electromagnetic tensor
field F as follows:

F (E∗1 , E∗1 ) = F (E∗3 , E∗3 ) = 0.565494

With the help of the above electromagnetic tensor field F results and
equation (3.5), the isotropic pressure p∗ip without cosmological constant fluid
in the generalized Ricci soliton is 0.815494.

Acknowledgement: We are grateful to the referee for the comments which

improved the quality of the paper.
190 AJIT BARMAN

References
[1] Barman, A., Some properties of a semiconformal curvature tensor on a Riemannian
manifold, The Mathematics Student 91(2022), 201-208.
[2] Barman, A., Spacetime classification by Ricci soliton and Gradient Ricci soliton,
Palestine Journal of Mathematics 12(2023), 361-365.
[3] Duggal, K. L. and Sharma, R., Symmetries of spacetimes and Riemannian manifolds,
Springer-Science Business media, 1999, pages 46, 61-63.
[4] Erken, I. K., Yamabe solitons on three-dimensional normal almost paracontact met-
ric manifolds, Periodica Mathematica Hungarica 2019.
[5] Hamilton, R. S., The Ricci Flow on Surfaces, Contemporary Mathematics 71(1988),
237-261.
[6] Hawking, S. W. and Ellis, G. F. R., The large scale structure of spacetime, Cam-
bridge University Press, Cambridge, 1973.
[7] Mondal, C. K. and Shaikh, A. A., On Ricci solitons whose potential is convex, Proc.
Indian Acad.Sci. (Math. Sci.) 130(55)(2020), 1-7.
[8] Nurowski, P. and Randall, M., Generalized Ricci Solitons, The Journal of Geometric
Analysis 26(2016), 1280-1345.
[9] O’Neill, B., Semi-Riemannian geometry with applications to relativity, Academic
Press, pages-79, 221, 223.
[10] Ozen, F. Z., m-projecttively flat spacetimes, Math. Reports 14(2012), 363-370.
[11] Yano, K., Integral formulas in Riemannian geometry, Marcel Dekker, New York,
1970.

AJIT BARMAN
Department of Mathematics
Ramthakur College
P.O.:- A. D. Nagar-799003
Agartala, Dist- West Tripura
Tripura, India.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 191–199

AN INTRODUCTION TO THE MATHEMATICS OF

HILLEL FURSTENBERG

ANISH GHOSH

Abstract. This is a popular exposition of some of the beautiful math-

ematics of Hillel Furstenberg. It is based on a talk given by the author
at the Bangalore Probability Seminar.

1. Introduction

As a first year undergraduate student with an interest in mathematics

but no background to speak of, I came across the book “Proofs from the
Book" by Aigner and Ziegler. There, I found an enchanting proof of the
infinitude of primes. This proof, which Hillel Furstenberg published [2]
while he was an undergraduate, uses a beautiful ‘topological’ idea1. I liked
it so much that I gave a talk about it to my fellow students. Little was I to
know then that I would have the privilege to read many beautiful and deep
works of mathematics by Furstenberg.
The Abel prize for the year 2020 was awarded to two mathematical
giants, and heroes of mine, Hillel Furstenberg and Gregory Margulis for
“pioneering the use of methods from probability and dynamics in group
theory, number theory and combinatorics". This article is meant to be
an introduction to the work of Hillel Furstenberg on the occasion of his
winning the Abel prize. To summarize all or indeed a representative portion
of Furstenberg’s work in a few pages is a daunting task, well beyond the
abilities of the author of this article. Even if one were to make the difficult
choice of discussing one or two papers, it is still a steep uphill climb. A quote
from the Abel prize website aptly captures the author’s predicament: “When
Hillel Furstenberg published one of his early papers, a rumor circulated
that he was not an individual but instead a pseudonym for a group of

2010 Mathematics Subject Classification: 00A09, 11B30, 37B20

Key words and phrases: Ergodic Theory, Topological Dynamics, Ramsey theory.

© Indian Mathematical Society, 2024 .

1
Much later, I recognized this to be the profinite topology.
191
192 ANISH GHOSH

mathematicians. The paper contained ideas from so many different areas,

surely it could not possibly be the work of one man?". Accordingly, I
will limit myself and focus on two themes in this article which illustrate
Furstenberg’s seminal contributions to the interactions between dynamics,
combinatorics and number theory. Namely I will discuss commuting maps
on the torus, and ergodic Ramsey theory. The article ends with a conjecture
of Furstenberg’s which he has popularized in the last few years and which
I personally find very appealing. I have tried to make the article accessible
to a broad scientific audience and included a bibliography with suggestions
for future reading; any omissions should be blamed on my own intellectual
limitations and to a lesser extent, on space constraints.

2. times 2 (×2 ) and times 3 (×3 )

We will discuss Furstenberg’s work in an area of mathematics called

ergodic theory, which is closely related to the field of dynamical systems.
Very loosely speaking, the latter involves the mathematical study of moving
objects, whilst the former comprises the study of statisticial or measure
theoretic properties of dynamical systems. Thus in the study of dynamical
systems we have a space, for example a metric or topological space, along
with a rule which tells us how points in the space move, while in ergodic
theory, we are also given a measure on the space and we are interested in
the long term behaviour of typical as well as exceptional orbits with respect
to the given measure.
Consider the circle doubling map

×2 : x → 2x mod 1.

Here we write the circle additively as T = R/Z and denote by B its Borel
sigma algebra. The idea is to iterate this map and study the resulting
dynamics. It turns out that the dynamics is quite rich, namely this map
is a classical example of what one would call a “chaotic" system. It has
periodic points, i.e. points which return to themselves after finitely many
iterations, as well as points with dense orbits. In fact, it is an easy exercise
to check that there are 2n −1 points of period n and moreover that the set of
periodic points is dense in T. A periodic orbit is an example of an invariant
set. A subset X of T is said to be invariant under ×2 if ×2 (X) ⊂ X.
There are many (closed) invariant subsets for such maps. For instance, the
circle is invariant, as is a periodic orbit. A more interesting example is the
 MATHEMATICS OF HILLEL FURSTENBERG 193

middle third Cantor set which is invariant for ×3 2. This brings us to a

fundamental question: is it possible to concretely describe the dynamics
of such maps? For instance, can one list all the possible closed invariant
subsets? It turns out that one can model ×2 (and also ×3 and so on) using
symbolic dynamical systems. In the case at hand, this relationship is easy
to describe. Every real number x ∈ [0, 1] has a base n expansion
∞
X
x = 0.x1 x2 · · · = xi n−i , xi ∈ {0, 1, . . . , n − 1}.
i=1

This expansion is unique (except for a countable set of numbers which have
two). If we take x ∈ [0, 1] with base n expansion x = 0.x1 x2 x3 . . . then
×n (x) = 0.x2 x3 x4 . . . . Let Σ = {0, . . . , n − 1}N denote the set of sequences
of elements in {0, . . . , n − 1}. On this space, we define the shift map as
follows:
σ((x1 , x2 , x3 , . . . )) = (x2 , x3 , . . . ).
In other words, σ takes as input a sequence in Σ and outputs another se-
quence obtained by removing the first element of the input sequence and
shifting the remaining elements one place to the left. The base n expan-
sion connects our circle map to the shift in the following natural manner.
Consider the map φ : Σ → [0, 1] given by
x1 x2 x3
φ((x1 , x2 , . . . )) = + 2 + 3 + ...
n n n
We can think of φ as a map from Σ to T by identifying 0 and 1 and it
provides us with a semi-conjugacy between the circle map ×n and the shift
σ. This means that φ ◦ σ = ×n ◦ φ. Moreover, φ is onto and 1 − 1.3 The
upshot is that one can try and understand the dynamics of a complicated
system like (T, ×n ) using a (hopefully) easier dynamical system (Σ, σ) via
the semi-conjugacy φ.
We now turn to the ergodic theory of the doubling map. A map T :
T → T is called a measure-preserving transformation if

µ(T −1 B) = µ(B) for all B ∈ B.

Alternatively, one says that T preserves µ; it turns out that ×n preserves

Lebesgue measure. We say that a measure preserving transformation T of

2For the purposes of our discussion, one can similarly consider the × map for any integer
n
n ≥ 2.
3provided of course, that one ignores a countable subset of sequences.
194 ANISH GHOSH

a probability space (X, B, µ) is ergodic, or equivalently that µ is an ergodic

measure for T , if it is “indecomposable" namely if

T −1 B = B implies that µ(B) = 0 or 1.

One could ask how many ergodic measures a dynamical system has? Is it
possible to list them? As the reader might have guessed already, it turns out
that the ×2 map has infinitely many (in fact uncountably many) ergodic
measures.
Let us now consider two maps together, say ×2 and ×3 . We have seen
that each of them has many invariant subsets. Are there any subsets of T
which are jointly invariant under both? In other words, do expansions in
base 2 and base 3 have anything in common? The answer is provided in a
beautiful theorem of Furstenberg which can be found in his seminal work
[3]. Two integers p, q > 1 are multiplicatively independent if they are not
both rational powers of a single integer, i.e. log p/ log q ∈
/ Q. For instance,
2 and 3 are multiplicatively independent but 4 and 8 are not. Then Sp,q be
the semigroup of the natural numbers defined as follows

Sp,q = {pm q n : m, n ∈ Z, m, n ≥ 0}.

Theorem 2.1. (Furstenberg 1967): Let p, q be multiplicatively independent

integers and suppose that X ⊂ T is closed and invariant under Sp,q . Then
either X = T or X is finite (and so X ⊂ Q).

Furstenberg’s theorem is an example of a (topological) rigidity result

in mathematics. Rigidity results in ergodic theory often have striking con-
sequences in other fields such as number theory and geometry. Here is an
interesting Diophantine consequence of the Theorem above, see [1] for an
elementary proof of this result.

Theorem 2.2. (Furstenberg): If Sp,q is as above and α ∈ R is an irrational,

then
{sα mod 1 : s ∈ Sp,q }
is dense in [0, 1].

The reader will recognize this to be a generalization of Weyl’s classical

equidistribution theorem. In [3], Furstenberg introduced a new property
of dynamical systems called disjointness. This idea has since proved to be
very influential in dynamics. Furstenberg introduced this notion both in the
 MATHEMATICS OF HILLEL FURSTENBERG 195

topological as well as the measure theoretic setting. We will not go into the
precise definition here but remark that it can be thought of as the opposite
of isomorphism in some sense. Furstenberg showed that dynamical systems
with many periodic points are disjoint from those which have the property
that every orbit is dense.
What about measure theoretic rigidity? This turns out to be deeper and
more subtle than the topological question. Furstenberg made the following
influential conjecture in the same paper.

Conjecture 2.3. (Furstenberg 1967): Let p, q be two multiplicative in-

dependent positive integers. Any Borel measure µ on T ergodic under the
action of Sp,q is either Lebesgue measure or an atomic measure supported
on finitely many rational points.

Furstenberg’s conjecture is possibly the most famous open problem in

ergodic theory. His work has had far reaching consequences in the dynamics
of group actions on homogeneous spaces. We refer the reader to [10, 11] for
details and references.

3. Ergodic Ramsey Theory

Let us begin with a basic result in dynamical systems, namely the

Poincaré Recurrence Theorem. It states that if (X, B, µ, T ) is a measure
preserving system and A ∈ B has positive measure, then there exists n ∈ N
such that µ(A ∩ T −n A) > 0. In his seminal paper [4], Furstenberg proved
a far reaching generalization of the Poincaré Recurrence Theorem, namely
he considered multiple recurrences. Furstenberg proved:

Theorem 3.1. Let (X, B, µ, T ) be a measure preserving system and k be a

positive integer. Then for any A ∈ B with µ(A) > 0 there exists n > 0 such
that
µ(A ∩ T n (A) ∩ · · · ∩ T (k−1)n A) > 0.

Furstenberg has spectacularly demonstrated how recurrence results are

related to deep problems in additive number theory. The branch of additive
number theory we are interested in is called Ramsey theory and starts with
van der Waerden’s famous result.

Theorem 3.2. (van der Waerden 1927): Let k and r be given. There
exists a number N = N (k, r) such that if the integers in [1, N ] are coloured
196 ANISH GHOSH

using r colours, then there is a non-trivial monochromatic k term arithmetic

progression.

In 1936, Erdős and Turán made a famous conjecture generalizing van

der Waerden’s theorem above. This conjecture was eventually settled in a
landmark work by Szemerédi in 1974. In part for this work, Szemerédi was
awarded the Abel prize in the year 2012. In order to state his result we need
the notion of upper density of a subset of the integers. The upper density
of a subset S of the integers is defined as
1
d∗ (S) := lim sup |A ∩ {−n, −n + 1, . . . , n − 1, n}|.
n→∞ 2n + 1

Theorem 3.3. (Szemerédi 1974):Any subset of Z with positive upper den-

sity contains arbitrarily long arithmetic progressions.

The first non-trivial case of the conjecture is the case k = 3 which was
settled by Roth [12] using analytic methods, namely exponential sums. The
case k = 4 was then settled by Szemerédi before he went on to settle the
general case. Szemerédi’s proof is based on intricate combinatorial methods.
Subsequently, an analytic proof was found by Gowers [7] in a significant
breakthrough.
Furstenberg [4] in 1977 gave a new, completely different proof of Sze-
merédi’s theorem using ergodic theory and thereby introduced a new branch
of mathematics called ergodic Ramsey theory. The main idea in Fursten-
berg’s proof is to translate the problem into dynamics via the Furstenberg
correspondence principle. What is the dynamical system to consider? It
turns out to be the symbolic system considered above. Namely, Fursten-
berg’s multiple recurrence theorem applied to a Bernoulli shift implies Sze-
merédi’s theorem. For a beautiful discussion of these topics, I refer the
reader to Furstenberg’s article [6] and his book [5]. Furstenberg’s work has
had very far reaching consequences and it’s influence can be seen in diverse
important breakthroughs in the area including, for instance, the Green-Tao
theorem which states that the primes4 contain arbitrarily long non-trivial
arithmetic progressions. We refer the reader to [9] and [16] for an overview
of the many exciting recent developments, many of which can be traced
back to this seminal paper of Furstenberg.

4which have zero density.

 MATHEMATICS OF HILLEL FURSTENBERG 197

4. Well approximable numbers and homogeneous dynamics

I’ll end this article with a beautiful problem that Furstenberg has posed
on many occasions in recent times and has to do with the interaction be-
tween Diophantine approximation and dynamics on homogeneous spaces of
Lie groups. The study of Diophantine approximation deals with the approx-
imation of real numbers by rational numbers. This is an ancient subject
with connections to many branches of mathematics and the sciences. Yet
many basic questions in the area remain unanswered and pose a funda-
mental challenge to modern mathematics. In particular, understanding the
Diophantine properties of individual numbers is a very difficult problem in
general. A basic and indispensable tool in this regard is the continued frac-
tion expansion of a real number. We point the reader to [8] for a classic
exposition. We all know from our school days that 22/7 is a good approxi-
mation for π. In fact, every real number has a continued fraction expansion.
For example, the one for π is

π = [3; 7, 15, 1, 292, 1, 1, 1, 2, . . . ],

which is shorthand for

1
π =3+
1
7+
1
15 +
1
1+
292 + · · ·
The entries in the continued fraction expansion are called partial quotients.
One can read off rational approximations for π by looking at truncations of
the continued fraction expansion, for example
1 22
[3; 7] = 3 + = .
7 7
√
Similarly, the continued fraction expansion for 2 is
√
2 = [1; 2, 2, 2, . . . ].

This is an instance of a theorem of Lagrange which states that α is a qua-

dratic irrational number if and only if its continued fraction expansion is
eventually periodic. Similarly, one can prove that a real number α is badly
198 ANISH GHOSH

approximable by rational numbers if and only if it has bounded partial quo-

√
tients, 2 presents an example. Badly approximable numbers have zero
Lebesgue measure but are nevertheless plentiful in that they form a set of
full Hausdorff dimension. On the other end of the spectrum, a number
α ∈ R is called well approximable if it has unbounded partial quotients.
These form a full Lebesgue measure subset of the real line. One might
then wonder about concrete examples of well approximable numbers. This
turns out to be very difficult problem as illustrated by the following folklore
conjecture:
√
Conjecture 4.1. 3 2 is well approximable.
In fact, see [15] questions 2.9, 2.10, it is unknown whether any algebraic
number of degree greater than 2 has bounded partial quotients or whether
any algebraic number of degree greater than 2 has unbounded partial quo-
tients. Here are the first few terms in the continued fraction expansion of
√3
2.
√3
2 = [1; 3, 1, 5, 1, 1, 4, 1, 1, 8, 1, 14, . . . ]
In fact, Furstenberg has presented the above problem as a challenge to
the ergodic theory community by converting it into a problem about the
dynamics of certain flows on homogeneous spaces. The precise formulation
will carry us too far from this article, so I’ll refer the interested reader
to appendix A in [13] for the statement and a discussion of Furstenberg’s
conjecture.
5. Concluding Remarks

This article presents the tip of the iceberg as far as Furstenberg’s math-
ematical achievements are concerned. Thanks to giants like Furstenberg,
who have devoted a lifetime to developing and nurturing mathematics at
the very highest level, Israel is now the centre of the world as far as ergodic
theory is concerned. As a mathematician from a developing country, with
some understanding of the serious issues and difficulties involved in such
matters, this amazing achievement fills me with admiration.
Acknowledgement: This article is an expanded version of a talk I first
gave at the Bangalore Probability Seminar. I am grateful to the organizers
of the seminar for the invitation to give the talk. I thank C.S. Aravinda for
his kind invitation to write this article, and for his patience as I subjected
the deadline to a particularly chaotic dynamical system.
 MATHEMATICS OF HILLEL FURSTENBERG 199

References
[1] Boshernitzan, Michael D. Elementary proof of Furstenberg’s Diophantine result.
Proc. Amer. Math. Soc. 122 (1994), no. 1, 67–70.
[2] Furstenberg, Harry On the infinitude of primes. Amer. Math. Monthly 62 (1955),
353.
[3] Furstenberg, Harry Disjointness in ergodic theory, minimal sets, and a problem in
Diophantine approximation. Math. Systems Theory 1 (1967), 1–49.
[4] H. Furstenberg, Ergodic behaviour of diagonal measures and a theorem of Szemerédi
on arithmetic progressions, J. Analyse Math. 31 (1977), 204–256.
[5] Furstenberg, H. Recurrence in ergodic theory and combinatorial number theory. M.
B. Porter Lectures. Princeton University Press, Princeton, N.J., 1981. xi+203 pp.
[6] Furstenberg, Harry Poincaré recurrence and number theory. Bull. Amer. Math. Soc.
(N.S.) 5 (1981), no. 3, 211–234.
[7] Gowers, W. T. A new proof of Szemerédi’s theorem. Geom. Funct. Anal. 11 (2001),
no. 3, 465–588.
[8] Khinchin, A. Ya. Continued fractions. The University of Chicago Press, Chicago,
Ill.-London 1964 xi+95 pp.
[9] Kra, Bryna. The Green-Tao Theorem on arithmetic progressions in the primes: an
ergodic point of view. Bull. Amer. Math. Soc. (N.S.) 43 (2006), 3–23
[10] Lindenstrauss, Elon Rigidity of multiparameter actions. Probability in mathematics.
Israel J. Math. 149 (2005), 199–226.
[11] Lindenstrauss, Elon Equidistribution in homogeneous spaces and number theory.
Proceedings of the International Congress of Mathematicians. Volume I, 531–557,
Hindustan Book Agency, New Delhi, 2010.
[12] Roth, Klaus Friedrich. On certain sets of integers. Journal of the London Mathe-
matical Society. 28 (1): 104–109.
[13] Sargent, Oliver; Shapira, Uri Dynamics on the space of 2-lattices in 3-space. Geom.
Funct. Anal. 29 (2019), no. 3, 890–948.
[14] Szemerédi, Endre. On sets of integers containing no k elements in arithmetic pro-
gression. Acta Arithmetica. 27: 199–245.
[15] Waldschmidt, Michel Open Diophantine problems. Mosc. Math. J. 4 (2004), no. 1,
245–305, 312.
[16] Ziegler, Tamar Linear equations in primes and dynamics of nilmanifolds. (English
summary) Proceedings of the International Congress of Mathematicians-Seoul 2014.
Vol. II, 569–589, Kyung Moon Sa, Seoul, 2014.

Anish Ghosh
School of Mathematics
Tata Institute of Fundamental Research
Mumbai, 400005, INDIA
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 200–212

A BRIEF INTRODUCTION TO CERTAIN

DYNAMICAL SYSTEMS RELATED TO NUMBER
THEORY

S.G. DANI

Abstract. The area of homogeneous dynamics has been one of the

rapidly developing areas in recent decades, It intertwines ideas from
dynamics, group theory, ergodic theory and number theory, and has
led to several new insights and resolution of long-standing problems.
The aim of this article is to give a first introduction to the topic, with
minimal background requirements.

1. Introduction

One way of thinking about what ‘intuitively’ constitutes a ‘dynamical

system’ is to realize it as a space, say X, together with a family of transfor-
mations of X into itself, encoding the changes over time. The main aspect
that we shall concern ourselves here is the cumulative effect of repeated
changes over unit time over a (relatively) large time scale, asymptotically.
For convenience, we shall choose time to shift discretely and the change to
be governed by the same rule at each stage. We shall further assume the
space to have a topology in which it is locally compact and the transfor-
mation to be a homeomorphism. Thus a dynamical system in our context
is nothing but a homeomorphism of a locally compact space, except that
our main interest is in the asymptotic behaviour over a large number of
iterations of the homeomorphism (and not in the topological aspects of the
homeomorphism). We shall denote such a system as (X, T ), where X is the
space and T is the homeomorphism.
Let us begin with the following example. Let X = S1 , the circle, say
realised as {z ∈ C | |z| = 1}, where C stands for complex numbers. Let

2010 Mathematics Subject Classification: 37A17, 11J25.

Key words and phrases: Homogeneous dynamics, periodic and dense orbits, recurrence,
uniform distribution, Diophantine approximation.

© Indian Mathematical Society, 2024 .

200
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 201

θ ∈ R, namely a real number. Let T : S1 → S1 be defined by T (z) = eiθ z,

for all z ∈ S1 ; it is easy to see that T is a homeomorphism of S1 . Thus we
have a dynamical system (X, T ). The transformation at each stage consists
of ‘rotating’ a wheel (the circle) by an angle θ; if θ is positive then the
rotation is considered, by convention, to be anticlockwise by an angle θ and
if θ is negative the rotation is considered to be clockwise by an angle |θ|.
θ
Suppose first that 2π θ
is rational, say 2π = pq , where p and q are integers
and q is positive. Then we see that, for any z ∈ S1 , T q (z) = eiqθ z =
e2πip z = z, which shows that if we keep applying the transformation T , after
q steps we reach the same point where we started. The whole behaviour
(no matter what the starting point) then repeats after q iterations. Such a
transformation is said to be periodic. The minimum number of times after
which the repetition occurs is called the period of the dynamical system.
In the above case it can be verified that q would be the period, provided p
and q are co-prime.
θ
Next suppose that 2π is not rational. Let us start from a point z ∈ S1
and consider the sequence of points z, T (z), T 2 (z), ..., T k (z), ..., to which the
point will get transformed after successive iterations. Since T k (z) = eikθ z
and kθ is not a multiple of 2π, T k (z) is not equal to z, for any k. Thus the
system is not periodic. We shall next see that this system has the following
interesting property:

Proposition 1.1. Let θ be as above. Then for any z and w in S1 and  > 0
there exists a natural number k such that |T k (z) − w| < .

Proof: lt is enough to show that there exists a j such that T j is a

rotation by an angle ψ such that |ψ| < , since for any z and w the desired
assertion would then hold for a suitable multiple k of j, depending on the
relative positions of z and w; according to the sign of ψ the rotation would
be clockwise or anticlockwise, but the preceding contention holds in either
case. There is no loss of generality in assuming θ to be such that |θ| < π.
Let m be the unique positive integer m|θ| < 2π < (m + 1)|θ|. Then at least
one of (m + 1)|θ| − 2π and 2π − m|θ| has absolute value less than | 2θ | and
we choose θ1 to be one of them such that |θ1 | < | 2θ |. On account of the
choice as above there exists j1 , viz. m sign θ or (m + 1) sign θ, such that T j1
is, in effect (setting aside the full angle 2π), a rotation by an angle θ1 , with
|θ1 | < 2θ . Repeating the procedure we find j2 , j3 , . . . such that T j1 j2 ...jr is a
202 S.G. DANI

rotation by an angle θr such that |θr | < | 2θr |. Choosing j = j1 j2 . . . jr with

r such that | 2θr | <  we get the assertion as desired above. This completes
the proof.

2. Recurrence and uniform distribution

Given a dynamical system (X, T ), a point x ∈ X is said to be recurrent

if there exists a sequence {ni } tending to infinity such that T ni x converges
to x as i → ∞; in other words, for such a point the iterates will keep coming
arbitrarily close to the starting point. A point x ∈ X is said to be periodic
if there exists a positive integer p such that T p x = x. A periodic point
is always recurrent. In the above examples if πθ is rational all points are
periodic and if πθ is irrational then no point is periodic but all points are
recurrent.
Let (X, T ) be a dynamical system. For x ∈ X the sequence of points
x, T x, ..., T j x, ... is called the orbit of x under T ; often the set underlying the
sequence is also called the orbit. Proposition 1.1 shows that for a rotation
of the circle by an angle which is an irrational multiple of 2π, all orbits
are dense in the circle. In an intuitive sense, they fill up the space as we
pick more and more points from any given orbit. It turns out that they
even fill it quite uniformly, as we go along: we mean the following: consider
any angular arc say A in S1 . Start with a z ∈ S1 and consider the orbit
z, T (z), T 2 (z), ..., T j (z), ... where T is the rotation by an angle θ. For any
positive integer k let Ok = {z, T (z), ..., T k−1 (z)} be the part of the orbit
considering of the first k elements. Let nk be the number of elements of Ok
contained in A. Then as k → ∞, the average nkk tends to 2π α
where α is
the angular width of the arc A. This means in particular that if we take
two different arcs of the same angular width, then the average frequency of
visiting them approaches the same value, as we consider longer and longer
parts of the orbit. To show that this holds it is enough to show that for any
continuous function f on S1
k−1 Z 2π
1X j 1
f (T (z)) −→ f (cos t + i sin t) dt. (2.1)
k 2π 0
j=0

To deduce the preceding assertion one may apply (2.1) to functions

which are 1 on most part of the arc, taper off to 0 at the ends and vanish
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 203

on the complement of the arc, and also to similar functions with respect to
the complementary arc.
Before going to the proof of the above, for continuous functions, we
formulate a notion of uniform distribution of orbits in a general dynamical
system. Let (X, T ) be a dynamical system and let x ∈ X. Let C(X)
denote the space of all bounded continuous complex-valued functions on
X. The orbit of x is said to be uniformly distributed if for any f ∈ C(X)
the sequence k1 k−1 j
P
j=0 f (T x) converges as k → ∞ and the limit is positive
for all non-negative functions f ∈ C(X) for which f (x) > 0; if cf is the
limit of the sequence for f ∈ C(X), then the function I : C(X) → C
defined by I(f ) = cf will be called the asymptotic integral corresponding
to the (uniformly distributed) orbit. We note that a point whose orbit is
uniformly distributed in this sense is necessarily recurrent. It may however
not be dense in the whole space; the latter will be ensured if the invariant
integral is positive for all nonzero non-negative functions in C(X). By the
Lebesgue integral on S1 we mean the function I : C(S1 ) → C defined by
setting, for f ∈ C(S1 ), I(f ) to be the right hand side of 2.1. Our claim
above can now be restated as follows:

Proposition 2.1. Let X = S1 and let T be the rotation by an angle θ such

that πθ is irrational. Then the orbit of any z ∈ S1 is uniformly distributed
and the corresponding asymptotic integral is the Lebesgue integral on S1 .

Proof: We have to show that (2.1) holds for all continuous functions on
S1 .
First consider any function of the form f (z) = z m , where m ∈ Z − (0).
Then
k−1 k−1
1X
1X z m 1 − eikmθ
f T j (z) = eijmθ z m = ,
k k k 1 − eimθ
j=0 j=0
where the denominator in the last term is non-zero since mθ is not a multiple
of 2π. As k → ∞ the right hand side clearly tends to 0, which is the same
as the Lebesgue integral of the function z → z m . Also equation (2.1)
evidently holds for all constant functions. It follows therefore that equation
(2.1) holds for all trigonometric polynomials, namely functions of the form
Pn m
m=−n am z , where n is any positive integer and am , −n ≤ m ≤ n,
are complex numbers. Now let f ∈ C(X) and  > 0 be arbitrary. By the
Weierstrass approximation theorem there exists a trigonometric polynomial
204 S.G. DANI


ϕ such that |f (x) − ϕ(x)| < 3 for all x ∈ X. Then
k−1 k−1
1X j 1X
f (T (z)) − ϕ(T j (z)) < /3.
k k
j=0 j=0

Since equation (2.1) holds for ϕ in the place of f and since |I(f )−I(ϕ)| < 3 ,
this implies that k1 k−1 j
P

j=0 f T (z) − I(f ) <  for all large k, I being the
Lebesgue integral. Since  is arbitrary this means that equation (2.1) holds
for f . This completes the proof.

For any real number t we denote by hti the fractional part of t, namely
the unique s ∈ [0, 1) such that t = s + n for some integer n. A sequence
{tj } in R is said to be uniformly distributed mod 1 if for any interval (a, b)
contained in (0, 1)

1 
j | 1 ≤ j ≤ k and < tj >∈ (a, b) → (b − a),
k
as k → ∞; here | · | stands for the cardinality.

Corollary 2.2. Let α ∈ R be an irrational number. Then the sequence

{jα} is uniformly distributed mod 1.

Proof: follows immediately from Proposition 2.1 and the fact that for
t ∈ R and a, b ∈ (0, 1), hti ∈ (a, b) if and only if e2πit belongs to the arc
{e2πis | a < s < b} in S1 .
We mention the following interesting consequence of the Corollary 2.2.
Let {dj } be the sequence formed by taking the leading digit in the (usual)
expression for 2j in base 10; d1 = 2, d2 = 4, d3 = 8, d4 = 1, d5 = 3, · · · . How
frequently will we see the digit 1 in this sequence? Specifically this question
would mean the following: if we count the number of 1’s in {d1 , . . . , dk }
and divide it by k to get the average, will the ratio have a limit as k tends
to infinity and if so what is the limit? The answer is that it indeed tends
to log10 2; so we will see 1’s a little over 30% of the time if we follow the
sequence long enough. This readily follows from the above corollary 2.2
applied to α = log10 2.
The rotations as above form a rather special class of dynamical systems
and such ‘neat’ behaviour does not occur in general. However a large class
of dynamical systems come close to having such properties in some weak
ways and some results of this kind constitute an important theme in ergodic
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 205

theory and dynamical systems. We now sketch some of these properties. A

crucial role is played in this by finite invariant measures for the dynamical
systems. (From this point on it would be necessary to have some familiarity
with general measure theory).
Let (X, T ) be a dynamical system. By a measure on X we shall mean
a Borel measure, namely a measure defined on the σ-algebra of all Borel
sets. A measure µ is said to be finite if µ(X) is finite, and it is said to be
T -invariant if µ(T −1 (E)) = µ(E) for any Borel set E; T −1 (E) denotes the
set {x | T x ∈ E}).
Theorem 2.3. (Poincaré recurrence lemma) Let (X, T ) be a dynamical
system and let R be the set of its recurrent points. Suppose that the topology
of X has a countable base. Let µ be any finite T -invariant measure on X.
Then for the complement X\R of R in X we have µ(X\R) = 0.

When the space X is noncompact a transformation T need not admit a

finite invariant measure; e.g. X = R and T x = x+1 for all x ∈ R. However,
when X is compact the existence of such a measure is assured, namely we
have the following theorem.
Theorem 2.4. (Krylov and Bogoliubov) Let (X, T ) be a dynamical system
where X is compact. Then there exist finite T -invariant measures on X

The two results together imply in particular that any dynamical system
on a compact second countable space admits recurrent points. This may
not hold on non-compact spaces; this is again illustrated by the example
X = R and T (x) = x + 1 for all x ∈ R, for which there are no recurrent
points.

3. Some ergodic theory

We next recall a fundamental theorem from ergodic theory. In the

formulation we shall restrict to dynamical systems as above, though the
theorem holds in the generality of measure-preserving transformations of
measure spaces. We recall that L1 (X, µ) denotes the space of all measurable
R
functions f on X such that X |f |dµ < ∞.
Theorem 3.1. (Birkhoff’s ergodic theorem) Let (X, T ) be a dynamical sys-
tem and let µ be a T -invariant measure on X. Let f ∈ L1 (X, µ). Then as
Pk−1
k → ∞, k1 j=0 f (T j x) converges µ-almost everywhere.
206 S.G. DANI

We note that in general the set of points where the convergence holds
is a proper subset and depends on the function f .
Given a dynamical system (X, T ), a T -invariant measure on X is said to
be ergodic if for any Borel subset E such that T −1 (E) = E either µ(E) =
0 or µ(X − E) = 0. The condition means that the system can not be
‘partitioned’ in to two measure-theoretically nontrivial parts. Typically
finer and finer partitions into invariant sets may be possible.
In general the class of ergodic invariant measures of a dynamical system
can be large, and even uncountable. A general T -invariant measure µ such
that µ(X) = 1 can be expressed, in a certain canonical way, as a ‘continuous
sum’ (or ‘integral’) or ergodic invariant measures each with total measure
1.
It is not difficult to deduce that if µ as in the hypothesis of Birkhoff’s
ergodic theorem is an ergodic T -invariant measure and µ(X) = 1, then the
R
limit in the conclusion coincides with X f dµ for µ-almost all points. To-
gether with the results recalled above one can deduce from this the following
result about uniformly distributed orbits.

Theorem 3.2. Let (X, T ) be a dynamical system and suppose that X has a
countable base. Let µ be a T -invariant measure on X such that µ(X) < ∞.
Then for µ-almost all x the orbit of x is uniformly distributed. If µ is ergodic
then the associated integral is the integral corresponding to µ, for µ-almost
all x.

The asymptotic integral corresponding to the uniformly distributed or-

bits as in the theorem would of vary with the measure under consideration.
We note in particular that if µ is an invariant measure and there is a com-
pact invariant subset A such that µ(A) > 0 and µ(X − A) > 0, there would
be x ∈ A whose orbit is uniformly distributed, for which the correspond-
ing asymptotic integral would have to be different from the integral with
respect to µ.

4. Some simple dynamical systems related to Number theory

The above discussion shows that in some sense ‘most’ orbits in a general
dynamical system on a compact second countable space behave like those
of the rotations of the circle. In many problems, especially those related to
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 207

Number Theory, it is important to know precisely which orbits have this

property. We shall now discuss various examples and describe the situation
in this respect.
Let n be a positive integer and let X = Tn = S1 × S1 × . . . × S1
(n copies), the n-dimensional torus. Let θ1 , θ2 , . . . , θn ∈ R and let T :
X → X be defined by T (z1 , z2 , , . . . , zn ) = (eiθ1 z1 , eiθ2 z2 , . . . , eiθn zn ) for all
z1 , z2 , . . . , zn ∈ S1 . It is a homeomorphism consisting of (possibly different)
rotations in each coordinate. These systems are called translations of the
torus. This case is somewhat similar to the rotation of the circle.

Proposition 4.1. Let the notation be as above. Then all orbits are uni-
formly distributed. If there do not exist any integers k1 , . . . , kn with at least
one kj nonzero such that nj=1 kj θj is a multiple of π, then the asymptotic
P

integral corresponding to the orbits is the integral with respect to the product
1
measure λ × λ × . . . × λ, where λ is 2π times the angle measure on S1 .
There exist dense orbits if and only if the preceding condition holds and in
that case all orbits are dense.

Unlike in the one-dimensional case where the systems are either peri-
odic or have all orbits dense, in higher dimensions there are ‘intermediate’
possibilities, where the closure of the orbit could be a finite union of tori
of some intermediate dimension. Using the above Proposition 4.1 one can
get the following result on simultaneous Diophantine approximation with
linear forms; we recall that by a linear form on Rn one means a function
P
of the form L(x1 , . . . , xn ) = aj xj for all x1 , . . . , xn ∈ R, where a1 , . . . , an
are real constants (coefficients).

Corollary 4.2. Let L1 , L2 , . . . , Lk be linear forms on Rn , where 1 ≤ k ≤

n − 1. Suppose that there do not exist any c1 , . . . , ck ∈ R, except all 0’s,
such that c1 L1 + c2 L2 + . . . + ck Lk is a linear form with rational coefficients.
Then for any a1 , a2 , . . . ak ∈ R  > 0 there exist integers x1 , . . . xn such that
for all j = 1, 2, . . . , k

|Lj (x1 , . . . , xn ) − aj | < .

Next let X = T2 = S1 × S1 . Let A = (aij ) be a 2 × 2 matrix with integer

entries and determinant 1. Let T : X → X be defined by

T (z1 , z2 ) = (z1a11 z2a12 , z1a21 z2a22 ) ∀z1 , z2 ∈ S1 .

208 S.G. DANI

Then T is a homeomorphism of X and thus (X, T ) is a dynamical system.

These systems are automorphisms of the group T2 = S1 × S1 ; (similar
systems are studied in higher dimensions and also on more general groups,
but for simplicity we restrict ourselves to T2 ). In this case the behaviour
depends crucially on the eigenvalues of the matrix A. If A is unipotent
(namely 1 is the only eigenvalue or, equivalently, (A − I)2 = 0 where I
denotes the identity matrix), then T consists of rotations along certain
embedded family of circles (at different angles) and some fixed points; when
a11 = a12 = a22 = 1 and a21 = 0, the second coordinate is unaltered and
in the circle along the first coordinate the transformation is a rotation by
as much as the second coordinate. In this case all orbits are uniformly
distributed but no orbit is dense in T2 . If both eigenvalues of A are roots
of unity other than 1 then T is periodic.
When there exists an eigenvalue which is not a root of unity then there
have to be two real eigenvalues, one of absolute value greater than 1 and
other of absolute value less than 1; such an automorphism (when there is
no eigenvalue of absolute value 1) is called a hyperbolic automorphism. In
this case there is great variety of orbits. Some orbits are periodic, some are
dense, while there are also others whose closures are totally disconnected
Cantor-like sets. While there have to exist uniformly distributed orbits,
there are also orbits which are not uniformly distributed and even points
which are not recurrent. For these transformations explicit description of
the set of points for which recurrence or density or uniform distribution etc.
holds seems to be a hopeless task.
One can also consider composites of translations and automorphisms.
When the automorphism is hyperbolic, the behaviour of the composite is
like the hyperbolic automorphism. However when the automorphism is
defined by a unipotent matrix we can have composites all whose orbits are
dense and uniformly distributed. For simplicity we note only the following
illustrative example.

Proposition 4.3. (Furstenberg) Let X = T2 = S1 × S1 . Let θ ∈ R be such

that πθ is irrational and let T : X → X be the homeomorphism defined by
T (z1 , z2 ) = (z1 z2 , eiθ z2 ) for all z1 , z2 ∈ S1 . Then every orbit of T is uni-
formly distributed and the corresponding asymptotic integral is the integral
1
with respect to the product measure λ × λ, where λ is 2π times the angle
1
measure on S . In particular every orbit is dense in T . 2
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 209

A similar result holds in greater generality and in all dimensions and it

has the following interesting consequence.

Corollary 4.4. (H. Weyl) Let P (t) = nj=0 aj tj be a polynomial with real
P

coefficients aj , j = 0, 1, · · · , n, and suppose that at least one of a1 , · · · , an

is irrational. Then the sequence {P (k)} is uniformly distributed mod 1.

5. Homogeneous dynamics and Diophantine approximation

Next let X be the space of lattices in Rn , where n ≥ 2 is fixed; (a lattice

in Rn is a subgroup generated by a set of n vectors which form a basis of
Rn ); the subgroup would consist of integral combinations of the vectors.
We define a topology on X as follows: a subset O of X is said to be open
if for any Λ ∈ O and any basis B of Rn generating Λ there exists an  > 0
such that for any basis B 0 of Rn formed by choosing one point each from the
-neighbourhoods of the points of B, the lattice generated by B 0 belongs to
O. It can be verified that this indeed defines a topology on X. Further,
X is a locally compact space with respect to this topology; in fact it is a
manifold of dimensions n2 (that is, each point has a neighbourhood which
2
is homeomorphic to Rn ).
On X there is a canonical class of dynamical systems arising from invert-
ible linear transformations of Rn : for any invertible linear transformation A
we define TA : X → X by setting, for any lattice Λ, TA (Λ) to be the lattice
A(Λ) = {A(v) | v ∈ Λ}. It is easy to see that each TA is a homeomorphism
of X.
The study of the dynamics of these systems has some very important
applications to questions in Diophantine approximation. A longstanding
conjecture due to A. Oppenheim about values of quadratic forms at integral
points was settled by G. A. Margulis in 1987 via such a study, proving the
following.
Pn Pn
Theorem 5.1. (G.A.Margulis) Let Q(x1 , · · · , xn ) = i=1 j=1 aij xi xj ,
where n ≥ 3, and (aij ) is a symmetric non-singular matrix. Suppose that
there exist i, j, k and l such that akl 6= 0 and aij /akl is irrational. Then the
set {Q(p1 , · · · , pn ) | p1 , · · · , pn ∈ Z} is dense in R; namely for any α ∈ R
and  > 0 there exist p1 , · · · , pn ∈ Z such that

|Q(p1 , · · · , pn ) − α| < .
210 S.G. DANI

In the 1990s some remarkable results were proved by Marina Ratner,

on the theme of invariant measures and uniform distribution of orbits of
actions on ‘homogeneous spaces’ by translation by ‘unipotent elements’;
the subspace X1 of the space X of lattices in Rn (as above), consisting of
those lattices whose discriminant (namely the volume of the paralleloped in
Rn corresponding to the generating vectors together with the zero vector, or
equivalently the determinant of the matrix corresponding to the n vectors)
is 1, is a typical illustrative example of a homogeneous space as involved in
Ratner’s theory, and the translating ‘unipotent’ elements in the general case
correspond to unipotent linear transformations (namely those for which 1 is
the only complex eigenvalue) in this case. The space X1 is a manifold and
admits a natural finite measure which is invariant under actions of all linear
transformations with determinant 1, enabling study of orbits via methods
discussed above. Ratner’s results yield in particular the following.

Theorem 5.2. (M. Ratner) If A is a unipotent linear transformation then

all orbits of TA on X (notation as above) are uniformly distributed.

Ratner’s results also give a description of all invariant measures of the

systems and in particular of the asymptotic integrals of the orbits and settle
a conjecture of M.S. Raghunathan, about closures of the orbits. Note that
the orbits of the systems as above are not dense in X. For Λ ∈ X the set of
lattices with the same discriminant as Λ is a closed set invariant under TA
as above; in particular if Λ ∈ X1 then the closure is contained in X1 ; while
for most Λ ∈ X1 the closure is X1 itself, for a class of exceptional lattices
it could be still smaller.
To give a description of orbits of TA in X1 we first recall the following
definition: a subgroup H of GL(n, R) is said to be an algebraic subgroup
defined over rationals if there exist some polynomials P1 , · · · Pl in the n2

variables having rational coefficients and such that H = x = (xij ) ∈
GL(n, R) | Pk (xij ) = 0 for all k = 1, · · · , l .

Corollary 5.3. Let Λ ∈ X1 , B be a basis of Rn contained in Λ and let

(aij ) be the matrix of the transformation A with respect to the basis B.
The TA -orbit of Λ is dense in X1 if and only if any algebraic subgroup H
of GL(n, R) which is defined over rational and contains (aij ) contains all
matrices of determinant 1.
 CERTAIN DYNAMICAL SYSTEMS RELATED TO NUMBER THEORY 211

Subsequently Margulis and I (jointly) strengthened Ratner’s results,

where we study also the dependence on the starting point while considering
uniform distribution, and applied to get lower estimates for the number of
solutions in large balls, for the inequalities as in Theorem 5.1. We will not
go in to the details here.
There has been a great deal of activity in the last three decades following
Ratner’s results spurred by Ratner’s milestone results and its applications
in various areas, in Number theory, Geometry as also certain more applied
areas. The area has now been christened as “Homogeneous Dynamics”.
We will not go into more details here but mention one of the problems in
Number theory that now stands as a challenge: a conjecture of Littlewood.
For any t ∈ R let δ(t) denote the distance of t from the nearest inte-
ger. Littlewood conjectures that for any s, t ∈ R, lim inf n→∞ nδ(s)δ(t) = 0.
Consider the set, say E, of (s, t) ∈ R × R for which this is not true; con-
jecturally E is empty. There have been results asserting various specific of
pairs (s, t) not being contained in E. It is relatively straightforward to show
that E is a set of (planar) Lebesgue measure 0. Considerable effort has now
gone into understanding E by techniques of homogeneous dynamics. One of
the notable results, by M. Einsiedler, A. Katok and E. Lindenstrauss in this
respect (which was featured in the Fields medal citation of the last named
author!) has been to show that E is a countable union of compact sets of
‘box dimension’ 0; the box dimension is a measure of size of a set and the
smallness established in the result is a remarkable stride in showing the set
to be very small (if nonempty), as also in terms of the techniques used; if
the dynamical results involved in proving it could be strengthened to avoid
a certain ‘entropy condition’ used, that would prove the full conjecture.

6. Suggestions for follow up

The following books are recommended basic material in the area:

(1) P. R. Halmos, Lectures in Ergodic Theory, The Mathematical Society
of Japan, 1956.
(2) Peter Walters, An introduction to ergodic theory. Graduate Texts
in Mathematics, 79. Springer-Verlag, New York-Berlin, 1982.
212 S.G. DANI

(3) M. Bachir Bekka and Matthias Mayer, Ergodic theory and topo-
logical dynamics of group actions on homogeneous spaces. London Math-
ematical Society Lecture Note Series, 269. Cambridge University Press,
Cambridge, 2008.
(4) Manfred Einsiedler and Thomas Ward, Ergodic theory with a view
towards number theory, Graduate Texts in Mathematics, 259. Springer-
Verlag London, Ltd., London, 2011.
I would also suggest the following paper, giving a proof of the Oppen-
heim conjecture (Theorem 12 above) involving relatively little background.
(5) Shrikrishna G. Dani, On the Oppenheim conjecture on values of
quadratic forms. Essays on geometry and related topics, Vol. 1, 2, 257–270,
Monogr. Enseign. Math., 38, Enseignement Math., Geneva, 2001.

S.G. Dani
UM-DAE Centre for Excellence in Basic Sciences
University of Mumbai Campus, Santracruz
Mumbai 400098, INDIA
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 213–215

ELEMENTARY PRIME COUNTING

DINESH S. THAKUR

Abstract. This is a short note describing some fun examples of ele-

mentary counting of primes.

1. Introduction

The first two chapters of [HW75], my first and fond interaction with
number theory, give several proofs of the infinitude of primes together with
lower bounds for the prime counting function π(x) (being the number of
primes ≤ x) arising from making the proofs constructive. For example, the
Euclid’s argument producing a new prime as a divisor of p1 · · · pn + 1, or the
relative primality of Fermat numbers argument gives the lower bound (es-
sentially) log log(x) and a little more sophisticated proof [HW75, Sec. 2.6]
(parts credited to Euler 1737 and Erdos 1938) gives log(x)/ log(4) lower
bound.

While Tchebyshef’s 1852 famous argument leading to essentially opti-

mal lower bound cx/ log(x) has now been simplified to just a few lines, this
note describes intermediate bounds the author derived (in 1979-1980) es-
sentially from the Euler product (from 1737) for the Riemann zeta (which
essentially leads to infinitude of primes by divergence of harmonic series, or
perversely by irrationality of π 2 and Euler’s evaluation of ζ(2)).

The author has not seen these simple arguments anywhere in the liter-
ature, and hopes that they might still be of some interest.

2. Euler product arguments

(1) Since positive integers are made up of primes, we see (without even
using unique factorization into primes) that, for s, x > 1,
2010 Mathematics Subject Classification: 11N05, 11A41
Key words and phrases: Primes, Euler product

© Indian Mathematical Society, 2024 .

213
214 DINESH S. THAKUR

x+1
(x + 1)1−s
Z
1 dx X 1 Y 1
− = s
≤ s
≤ (1 − s )−1
1−s 1−s 1 x n p
n≤x p≤x

Y 1 −1 e 1/ts
P 1−s
≤ (1 − ) ≤ e ≤ ee((π(x)+1) /(1−s)−1/(1−s)) .
ts
2≤t≤π(x)+1
p √
With 1 − s = e/ log(x), we get π(x) + 1 ≥ (1 + e) log(x)/e , for large x.

Details and variants of the manipulations: (i) In more detail, with

P := π(x) + 1 and 1 − s = 1/z, the inequality, for large z, is ez(P 1/z − 1) >
log(z((x + 1)1/z − 1), so that ezP 1/z > ez + (log(x)/z) equivalent to the
claim.
(ii) For simplicity, we have used ebu ≥ 1/(1 − u), for b = e, 0 ≤ u ≤ 1/2,
but could have improved the bound by a constant power by optimising for
b > 1 when it works for sufficiently small positive u, by splitting the sum
over t’s into small and large t’s. We leave it to the interested reader.

(2) A different elementary argument: Taking logarithm of the Euler

P
product inequality shows 1+ p≤x (1/p) > log log(x). Raise this to the n-th
power and note that in the expanded product there are (π(x)+1)n = P n re-
ciprocals, and none can occur more than n! times (by unique factorization),
giving
1 1
n!(1 + + · · · + n ) > (log log(x))n
2 P
so that roughly n!n log(P ) > (log log(x))n , which with n = log log(x), by
3/2+
the Stirling approximation of n! gives π(x) > elog(x)/(log log(x)) , for large
x.

Details of the manipulations: Writing `k for the k-th iterated loga-

rithm, we have n!n(`1 (P ) + (γ + )/n) > (`2 (x))n . So

C + n`1 (n) − n + (3/2)`1 (n) + `2 (P ) > `1 (n!) + `1 (n) + `2 (P ) > n`3 (x).

So, n = `2 (x) gives `1 `1 (P ) = `2 (P ) > `2 (x) − (3/2)`3 (x) − C >

`1 (`1 (x)/(`2 (x))3/2+ )

Remarks The Euler product heuristically (under very strong regularity

assumptions) leads to the prime number theorem itself.
 ELEMENTARY PRIME COUNTING 215

Acknowledgements and dedication I dedicate this paper (on his 75th

birthday) to B. Godse, who introduced [HW75] book to me in school, to
Mangesh Rege, to S. S. Rangachari and to the memory of S. Srinivasan, K.
G. Ramanathan and K. Ramachandra from TIFR for their encouragement
in my school and undergraduate years. In fact, S. Srinivasan had pointed out
to me then that my first counting did not even use the unique factorization
into primes.

References
[HW75] G. H. Hardy and E. M. Wright. Introduction to the theory of numbers, The
English Language Book Society and Oxford University Press (1975), 4th edition.

Dinesh S. Thakur
Department of Mathematics
University of Rochester
Rochester, NY 14627, USA.
E-mail: [email protected]
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos. 1-2, January - June (2024), 216–224

PROBLEM SECTION
In the Mathematics Student, volume 92(3-4) 2023, we had invited solutions
for Problems 5, 6 from MS 92(1-2) 2023 and for the eight new problems of
MS 92(3-4) 2023 till February, 2024.
We have received solutions to Problems 5, 6 of MS 92 (1-2) 2023 by Dr.
Andrés Ventas, Santiago de Compostela, Spain; these problems were pro-
posed by Dr. Anup Dixit, The Institute of Mathematical Sciences, India.
For the new problems from MS 92(3-4) 2023, we have received solutions for
Problems 1, 3, 4, 5. Problem 1 was suggested by Dr. Shpetim Rexhepi and
Dr. Ilir Demiri, Mother Teresa University, North Macedonia and the solu-
tion was provided by Dr. Henry Ricardo, Westchester Area Math Circle,
New York, USA. Problems 2-5 were proposed by Dr. B. Sury, Indian Sta-
tistical Institute, India and Problems 6-8 were proposed by Dr. Chudamani
Pranesachar Anil Kumar, KREA University, India. We didn’t receive any
solutions for Problems 2, 6, 7, 8. Below we present the solutions received
based on the recommendations of the proposers and the experts. We ap-
preciate the contributions from the proposers and sincerely acknowledge all
solutions received from the readers.
First we present new problems for this volume. We invite solutions for
these and for problems 2, 6, 7, 8 of MS 92 (3-4) 2023 from the readers
till September 30, 2024. Correct solutions received by this date will be
published in volume 93 (3-4) 2024 of The Mathematics Student, which is
scheduled to be published in October 2024.

New Problems
Problems 1-3. are proposed by Dr. Chudamani Pranesachar Anil
Kumar, KREA University, India.

© Indian Mathematical Society, 2024 .

216
 PROBLEMS SECTION 217

MS 93(1-2) 2024 : Problem 1.

Let m, n be two positive integers such that 2 ≤ m ≤ n. For a permutation
σ ∈ Sn , 0 ≤ i ≤ m − 1, let

Ti = {σ ∈ Sn | inv(σ) ≡ i mod m}.

Show that | T0 |=| T1 |= · · · =| Tm−1 |, that is, the sets Ti have equal
cardinality for 0 ≤ i ≤ m − 1 and find this cardinality.

MS 93 (1-2) 2024 : Problem 2.

In a 2023 × 2023 chess-board (not the usual 8 × 8), the four corner
squares are removed.
(1) Can the rest be covered by a combination of 5 × 1 dominoes, (that
is, 5 square boxes) by putting them horizontally, vertically on the
board?

(Type 1) , (Type 2)

(2) Can the rest be covered by a combination of 5 × 1 dominoes, (that

is, 5 square boxes) by putting them horizontally, vertically or
diagonally in both ways (as shown in the figure) on the board?

(Type 1) , (Type 2) , (Type 3) , (Type 4)

MS 93(1-2) 2024 : Problem 3.

Let P = {p1 = 2 < p2 < . . . < pl } be a finite set of l primes. Let

T = {n ∈ Z | gcd(n , pi ) = 1, 1 ≤ i ≤ l} = {. . . < a−2 < a−1 < a0 < a1 < a2 < . . .}.

Then prove the following:

(1) The sequence xj = aj+1 − aj , j ∈ Z is bounded.
(2) There exists j ∈ Z such that 2l ≤ xj = aj+1 − aj .
218 PROBLEMS SECTION

Problems 4-6 are proposed by Dr. B. Sury, ISI, Bengaluru, India.

MS 93(1-2) 2024 : Problem 4.

Let f : [0, 1] → R be differentiable. Suppose f, f 0 have no common zeroes.

Prove that the zero set of f :{x ∈ [0, 1] | f (x) = 0} must be finite.

MS 93(1-2) 2024 : Problem 5.

Find all triangles with vertices A = (0, 0), B = (4, 3) and C = (u, v) where
u, v are integers and AB, AC have integer lengths.

MS 93(1-2) 2024 : Problem 6.

Let G be a group on which the m-th power map and n-th

power,m, n ∈ N \ {1}, map are both homomorphisms. If m(m − 1)/2 and
n(n − 1)/2 are relatively prime, prove that G must be abelian.
Conversely, if the greatest common divisor of m(m − 1)/2 and n(n − 1)/2
is > 1, show there exist non-abelian groups G on which the m-th power
map and the n-th power maps are homomorphisms.

MS 93(1-2) 2024 : Problems 7. (Proposed by Dr. Shpetim Rexhepi

and Dr. Ilir Demiri, Mother Teresa University, North Macedonia.)

Let α, β and γ be any real numbers satisfying

α2 (β + γ) + β 2 (γ + α) + γ 2 (α + β) = 0. Let
   
4 4 4α + β + 2γ 4 α+β
A := sin α + sin β − 16 cos sin ,
2 2
   
3 3 3 3α + β + 2γ 3 α+β
B := sin α + sin β − 8 cos sin
2 2
and
   
7 7 7 α + β + 2γ 7 α+β
C := sin α + sin β − 128 cos sin .
2 2
AB 6
Prove that = .
C 7
 PROBLEMS SECTION 219

MS 93(1-2) 2024 : Problems 8. (Proposed by Dr. Andrés Ventas

Santiago de Compostela, Spain.)

Given the following two constants defined by continued fractions with all
their coefficients repeated, one with a positive sign, the Golden Section,
ϕ = [1, 1, 1, 1, · · · ], and the other with a negative sign, the Pena Trevinca
constant, τ ,
1
τ =3−
1
3−
1
3−
3 − ···
Prove that τ = ϕ + 1.

Solutions to the New Problems

MS 92 (3-4) 2023 : Problem 1. (Proposed by Dr. Shpetim Rexhepi
and Dr. Ilir Demiri, Mother Teresa University, Skopje, North Macedo-
nia)

If d1 and d2 are metrics on the metric space X and x1 , x2 , ..., xn are in X,

then prove that:
n−1
X
d21 (xi , xi+1 )
i=1 d21 (x1 , xn )
≥ n−1
!.
1 + d2 (x1 , xn ) + n X
(n − 1) 1 + n + d2 (xi , xi+1 )
i=1

Solution: (by Dr. Henry Ricardo, Westchester Area Math Circle, New
York, USA).
We use the Cauchy-Schwarz inequality in the form
n
ai2 ( ni=1 ai ) 2
X P
≥ Pn (1)
bi i=1 bi
i=1

and the generalized triangle inequality for metrics d:

d(x1 , xn ) ≤ d(x1 , x2 ) + d(x2 , x3 ) + · · · + d(xn−1 , xn ). (2)

220 PROBLEMS SECTION

Now we have
Pn−1 2 n−1
i=1 d1 (xi , xi+1 )
X d12 (xi , xi+1 )
=
1 + d2 (x1 , xn ) + n 1 + d2 (x1 , xn ) + n
i=1
P 2
n−1
(1) i=1 d 1 (x i , xi+1 )
≥
(n − 1) + (n − 1)d2 (x1 , xn ) + n(n − 1)
(2) d12 (x1 , xn )
≥  .
(n − 1) 1 + n + n−1
P
i=1 d2 (xi , xi+1 )

Problems 3, 4, 5 of 92(3-4) were proposed by Dr. B. Sury, ISI, Ben-

galuru, India.
92 (3-4) 2023 Problem 3.

Start with a unit square. To its right, adjoin a square of unit area. Then,
we have a rectangle of base 2 and height 1. Now, adjoin a rectangle on
top of the earlier one which has the same base (2 in this case) and area 1.
Thus, the new rectangle would have height 1/2. In this manner,
recursively adjoin rectangles to the right and on the top to the previous
one, each time with unit area as in the figure.

Find the limit of the ratio of the base to the height of the large rectangle
formed at each stage as n → ∞.

Solution: (by Dr. Andrés Ventas, Santiago de Compostela, Spain).

 PROBLEMS SECTION 221

We start with a rectangle with base 2 and height 1. Next we build a small
rectangle on top of this with base 2 and area 1. So its height must be 1/2.
Now the new larger rectangle has base 2 and height 3/2 = 1 + 1/2. Next
we construct a small rectangle on the side of the new large rectangle so
that the area of this small rectangle is 1; hence its base must be equal to
the inverse of 3/2, which is 2/3. Thus the base of the new larger rectangle
at this stage is equal to 2 + 2/3. Next we construct a small rectangle on
top of this with area 1 and calculate the height of the new larger
rectangle. So

height of the large rectangle = height of the large rectangle two stages before
+ inverse of the base of the previous large rectangle.

So starting with base 2, at each stage we calculate height or the base of

the new larger rectangle alternatively. Let us calculate first elements of
this sequence and here it is preferable not to simplify:
a0 = 2;
1 3
a1 = +1= ;
2 2
2 8 2·4
a2 = +2= = ;
3 3 3
3 3 3·5
a3 = + = ;
2·4 2 2·4
2·4 2·4·6
a4 = + 2 · 43 = ;
3·5 3·5
3·5 3·5 3·5·7
a5 = + = ;
2·4·6 2·4 2·4·6
···
Looking for a pattern in the above we see that the double factorial
notation will be useful. Denoted by n!!, the double factorial is the product
of all the positive integers up to n that have the same parity (odd or
even) as n. That is,
dn
2
e−1
Y
n!! = (n − 2k) = n(n − 2)(n − 4) · · ·
k=0

n!!
. Thus we have an = , with n even for the base and n odd for the
(n − 1)!!
an
height. We need to calculate lim . For which, we need the Stirling
n→∞ an−1
222 PROBLEMS SECTION

asymptotic approximation,

√πn n
 n/2
, if n is even,

e



n!! ∼
√  n n/2


 2n
 , if n is odd.
e
Using the above we calculate,
√  n n/2
πn
e
√  n n/2 √  n n/2 2
an 2n πn
lim = lim √  e n/2 = lim √  e n/2 2
n→∞ an−1 n→∞ n n→∞ n
2n 2n
e e
√  n n/2
πn
e
 n n
πn π
= lim  ne n = .
n→∞
2n 2
e

92 (3-4) 2023 Problem 4.

For each real number x ≥ 0, let Nx = {n : bnxc is even }. If
X 1 4
f (x) = for x ∈ [0, 1), prove that Inf{f (x) : x ∈ [0, 1)} = .
2n 7
n∈Nx
Solution: (by Dr. Andrés Ventas, Santiago de Compostela, Spain).
2
We shall show Inf{f (x) : x ∈ [0, 1)} = .
3
When x goes to 0, the sum has all members for each n, because
bn · 0 = 0c = 0 is even, then we get the supremum value of the sum.
When x goes to 1, we tend to alternate the members between odd and
even, because bn · 1c = n. We would have Nx = {2, 4, 6, 8, · · · }, starting at
X 1 1/4 1
1/4, thus = = .
22n 1 − 1/4 3
n∈Nx
But 1 is excluded from the interval, so our infimum sum, the greater lower
bound, has alternating members Nx = {1, 3, 5, 7, · · · }, starting at 1/2, this
∞
X 1
is 2n−1
.
2
n=1
 PROBLEMS SECTION 223

1 1
Thus, our geometric series has a = and r = .
2 4
X 1 1/2 2
Inf{f (x) : x ∈ [0, 1)} = 2n−1
= = .
2 1 − 1/4 3
n∈2n−1

92 (3-4) 2023 Problem 5.

If f is a function on R satisfying f (U ) = R for every nonempty open set

U , then show that f is discontinuous at EVERY point, Further, show
there exist such functions f .
Solution: (by Dr. Henry Ricardo, Westchester Area Math Circle,
New York.)
Solution 1. The function in the hypothesis of the problem is said to be
strongly Darboux. The “Conway base 13 function” is a Darboux function,
an everywhere surjective function, that is discontinuous at every point of
R (https://en.wikipedia.org/wiki/Conway base 13 function)

Solution 2. In 2018, A. Bergfeldt devised a simpler example: If

{xi }i ∈ Z+ is the binary expansion of x, so that each xi ∈ {0, 1}, define
f : R → R by
∞
X (−1)xk
f (x) = if the series converges, and f (x) = 0 otherwise.
k
i=1

This function is open and (according to Bergfeldt)

”clearly not continuous at any point”.
https://en.wikipedia.org/wiki/Conway base 13 function

Proposer’s Notes:
(1) In addition to the two nice solutions above, the following third
solution is due essentially to Brian Scott.
Consider a basis B of the Q-vector space R/Q. Map B bijectively
onto R and extend it Q-linearly to a surjection from R/Q to R.
Composing this surjection with the natural quotient map from R
to R/Q of Q-vector spaces, one has a function f : R → R with the
asserted property.
(2) For Conway’s base-13 function, readers may also refer to an article
by A. Ayyer, B. Sury entitled John Horton Conway,
224 PROBLEMS SECTION

The Magical Genius Who Loved Games, Resonance, May 2021,

p. 595-601.
(3) Also, for further aspects of this problem, readers are advised to
refer to a paper by Israel Halperin entitled Discontinuous
functions with the Darboux property, published in Canadian
Mathematical Bulletin, Vol. 2, No. 2, May 1959, pp. 111-118.
 FORM IV
1. Place of Publication: PUNE
2. Periodicity of
publication: QUARTERLY
3. Printer’s Name: DINESH BARVE
Nationality: INDIAN
Address: PARASURAM PROCESS
38/8, ERANDWANE
PUNE-411 004, INDIA
4. Publisher’s Name: M. M. SHIKARE
Nationality: INDIAN
Address: GENERAL SECRETARY
THE INDIAN MATHEMATICAL SOCIETY
EMERITUS PROFESSOR
DEPARTMENT OF MATHEMATICS
JSPM UNIVERSITY, PUNE 412207
MAHARASHTRA, INDIA
5. Editor’s Name: G. P. YOUVARAJ
Nationality: INDIAN
Address: RAMANUJAN INSTITUTE FOR ADVANCED
STUDY IN MATHEMATICS, UNIVERSITY OF
MADRAS, CHENNAI 600005, TN, INDIA
6. Names and addresses THE INDIAN MATHEMATICAL SOCIETY
of individuals who own
the newspaper and
partners or sharehold-
ers holding more than
1% of the total capital:

I, M. M. Shikare, the General Secretary of the IMS, hereby declare that the par-
ticulars given above are true to the best of my knowledge and belief.

M. M. Shikare
Dated: May 15, 2024 Signature of the Publisher

Published by Prof. M. M. Shikare for the Indian Mathematical Society, type set
by Prof. G. P. Youvaraj, Former Director and Head, Ramanujan Institute for Ad-
vanced Study in Mathematics, Chepauk, Chennai - 600005, India and printed by
Dinesh Barve at Parashuram Process, Shed No. 1246/3, S. No. 129/5/2, Dalvi-
wadi Road, Barangani Mala, Wadgaon Dhayari, Pune 411 041 (India).

Printed in India
The Mathematics Student ISSN: 0025-5742
Vol. 93, Nos.1-2, January - June (2024)
EDITORIAL BOARD
G. P. Youvaraj (Editor-in-Chief)
Ramanujan Institute for Advanced Study in Mathematics
University of Madras, Chennai- 600 005, Tamil Nadu, India
E-mail : [email protected]
Bruce C. Berndt George E. Andrews
Dept. of M athematics, U niversity Dept. of M athematics, T he P ennsylvania
of Illinois 1409 W est Green St. State U niversity, U niversity P ark
U rbana, IL 61801, U SA P A 16802, U SA
E − mail : [email protected] E − mail : [email protected]
M. Ram Murty B. Sury
Dept. of M athematics and Statistics T heoretical Stat. and M at. U nit
Jef f ery Hall, Queens U niversity Indian Statistical Institute
Kingston, Ontario, K7L3N 6, Canada Bangalore 560059, India
E − mail : [email protected] E − mail : [email protected]
Siddharth Gadgil Gadadhar Misra
Indian Institute of Science Indian Statistical Institute
Bangalore − 560012, India Bangalore − 560059, India
E − mail : [email protected] E − mail : [email protected]
Sukanta Pati Atul Dixit
IIT Guwahati IIT Gandhinagar
Guwahati 781039, Assam, India Gandhinagar − 382355, Gujarat, India
E − mail : [email protected] E − mail : [email protected]
S. K. Tomar Aparna Dar
P anjab U niversity IIT Kanpur
Chandigarh − 160014, India Kanpur − 208016, U.P., India
E − mail : [email protected] E − mail : [email protected]
Clare D0 Cruz L. Sunil Chandran
Dept. of M athematics, CM I Dept. of Computer Science&Automation
IT P ark P adur P.O., Siruseri Indian Institute of Science
Kelambakkam − 603103, T.N., India Bangalore − 560012, India
E − mail : [email protected] E − mail : [email protected]
Subhojoy Gupta T. S. S. R. K. Rao
Indian Institute of Science M anipalU niversity
Bangalore − 560012, India M Jaipur 303007, Rajasthan, India
E − mail : [email protected] E − mail : [email protected]
Kaushal Verma C. S. Aravinda
Dept. of M athematics T IF R Centre f or Applicable M athematics
Indian Institute of Science P. B. N o. 6503, GKV K P ost Sharadanagara
Bangalore − 560012, India Chikkabommasandra, Bangalore − 560065, India
E − mail : [email protected] E − mail : [email protected]
Indranil Biswas Timothy Huber
School of M athematics, T ata Institute School of M athematics and statistical Sciences
of F undamental Research, Homi Bhabha U niversity of T exas Rio Grande V alley, 1201
Rd., M umbai − 400005, India W est U niv. Avenue, Edinburg, T X78539 U SA
E − mail : [email protected] E − mail : [email protected]
Debjani Chakrabory Safique Ahmad
Department of M athematics Department of M athematics
IIT Kharagpur, Kharagpur 721302, India IIT Indore, Indore 452020, India
E − mail : [email protected] E − mail : [email protected]
T. Raja Sekhar
Department of M athematics
IIT Kharagpur, Kharagpur 721302, India
E − mail : [email protected]
 THE INDIAN MATHEMATICAL SOCIETY
Founded in 1907
Registered Office: Center for Advanced Study in Mathematics
Savitribai Phule Pune University, Pune - 411007

COUNCIL FOR THE SESSION 2024-2025

PRESIDENT: Madhu Raka, NASI Senior Scientist Fellow, Department of ,
Mathematics, Punjab University, Chandigarh - 160014, Punjab, India
IMMEDIATE PAST PRESIDENT: R. Balasubramanian, Former Director,
The Institute of Mathematical Sciences, Taramani, Chennai- 600113, India
GENERAL SECRETARY: M. M. Shikare, Emeritus Professor, Department of
Mathematics, JSPM University, Pune 412207, Maharashtra, India
ADMINISTRATIVE SECRETARY: B. N. Waphare, Center for Advanced Study in
Mathematics, S. P. Pune University, Pune-411007, Maharashtra, India
ACADEMIC SECRETARY: G. P. Raja Sekhar, Department of Mathematics
IIT Kharagpur, Kharagpur 721302, West Bengal, India
TREASURER: M. M. Pawar, Formerly at K.B.C. North Maharastra University,
Jalgon, Maharashtra, India
EDITOR: J. of the Indian Math. Society: Peeyush Chandra, Professor (Retired)
Department of Mathematics, I. I. T. Kanpur 208016, Kanpur (U. P.), India
EDITOR: The Mathematics Student: G. P. Youvaraj, Former Director,
Ramanujan Inst. for Advanced Study in Mathematics, University of Madras,
Chennai 600005, India
LIBRARIAN: M. Pitchaimani, Director, Ramanujan Inst. for Advanced Study
in Mathematics, University of Madras, Chennai 600005, India

OTHER MEMBERS OF THE COUNCIL:

• B. Rushi Kumar, Dept. of Mathematics, Vellore Institute of Technology, Vellore,
Tamil Nadu
• Jitendra Kumar, Department of Mathematics, IIT Kharagpur, Kharagpur
• Pratulananda Das, Dept. of Math. and Comp. Science, Jadhavpur Uni., Kolkata,
West Bengal
• Ajay Kumar Shukla, Department of Mathematics, Sardar vallabhai NIT, Surat,
Gujarat
• S. B. Nimse, Former Vice-Chancellor, SRTM University Nanded and Lucknow
University, Lucknow, Uttar Pradesh
• Shiv Datta Kumar, Motilal Nehru NIT, Allahabad, Uttar Pradesh
• Gajanan Lomte, MGM University, Aurangabad, Maharashtra
• Sameer Chavan, Dept. of Mathematics and Statistics, IIT Kanpur, Kanpur, Uttar
Pradesh
• Swaroop Nandan Bora, Dept. of Mathematics, IIT Guwahati, Guwahati, Assam
• Seshadev Padhi, Dept. of Mathematics, Birla Institute of Technology, Mesra,
Ranchi, Jarkhand
Back volumes of our periodicals, except for a few numbers out of stock, are available.
Edited by G. P. Youvaraj and published by M. M. Shikare, the General Secretary of
the IMS, for the Indian Mathematical Society.
Typeset by Prof. G. P. Youvaraj, Former Director and Head, Ramanujan Institute
for Advanced Study in Mathematics, Chepauk, Chennai - 600005, India and printed
by Dinesh Barve at Parashuram Process, Shed No. 1246/3, S. No.129/5/2, Dalvi-
wadi Road, Barangani Mala, Wadgaon Dhayari, Pune - 411041, Maharashtra, India.
Printed in India
Copyright c The Indian Mathematical Society, 2024.