Proofs, generalizations and analogs of Menon’s

identity: a survey
arXiv:2110.07271v4 [math.NT] 10 Nov 2023

László Tóth∗
Department of Mathematics
University of Pécs
Ifjúság útja 6, 7624 Pécs, Hungary
E-mail: ltoth@gamma.ttk.pte.hu

Acta Univ. Sapientiae, Mathematica, 15 (2023), no. 1, 142–197

Dedicated to my father, László Gy. Tóth, high school mathematics teacher,
on the occasion of his 90th birthday

Abstract
P
Menon’s identity states that for every positive integer n one has (a − 1, n) = ϕ(n)τ (n),
where a runs through a reduced residue system (mod n), (a − 1, n) stands for the greatest
common divisor of a − 1 and n, ϕ(n) is Euler’s totient function and τ (n) is the number of
divisors of n. Menon’s identity has been the subject of many research papers, also in the last
years. We present detailed, self contained proofs of this identity by using different methods,
and point out those that we could not identify in the literature. We survey the generalizations
and analogs, and overview the results and proofs given by Menon in his original paper. Some
historical remarks and an updated list of references are included as well.

2010 Mathematics Subject Classification: 11A07, 11A25, 22F05

Key Words and Phrases: Menon’s identity, arithmetic function, Euler’s function, Ramanu-
jan’s sum, Dirichlet character, Dirichlet convolution, unitary convolution, group action, orbit
counting lemma, n-even function, finite Fourier representation

Contents
1 Introduction 3
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 The gcd-sum function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
∗
The research was financed by NKFIH in Hungary, within the framework of the 2020-4.1.1-TKP2020 3rd
thematic programme of the University of Pécs.

1
2 Proofs of the classical Menon identity 5
2.1 Method I: Proof by multiplicativity . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Method II: Proof by the inclusion-exclusion principle . . . . . . . . . . . . . . . . 7
2.3 Method III: Proof by convolutional identities . . . . . . . . . . . . . . . . . . . . 8
2.4 Method IV: A variant of Method III . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 Method V: Another variant of Method III . . . . . . . . . . . . . . . . . . . . . . 10
2.6 Method VI: Proof by the orbit counting lemma . . . . . . . . . . . . . . . . . . . 12
2.7 Method VII: Proof by exponential sums . . . . . . . . . . . . . . . . . . . . . . . 13
2.8 Method VIII: Another proof by exponential sums . . . . . . . . . . . . . . . . . . 14
2.9 Method IX: Proof via the finite Fourier representation of n-even functions . . . . 16
2.10 Method X: Proof by Vaidyanathaswamy’s class division of integers (mod n) . . . 19

3 Historical remarks 22
3.1 Remarks on the paper by Menon . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Some more remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 Generalizations and analogs 24

4.1 Identities for f ((a − 1, n)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Identities concerning n-even functions f . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Geometric and harmonic means of (a − 1, n) with (a, n) = 1 . . . . . . . . . . . . 26
4.4 Identities using a polynomial P . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.5 Identities concerning parameters m, n, t . . . . . . . . . . . . . . . . . . . . . . . 28
4.6 Sums over k-reduced residue systems (mod nk ) . . . . . . . . . . . . . . . . . . . 28
4.7 Identities with (a, n) ∈ S, where S is a subset of N . . . . . . . . . . . . . . . . . 29
4.8 Unitary analogs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.9 Regular systems of divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.10 A generalized divisibility relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.11 Regular integers (mod n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.12 Multiple sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.13 Identities for arithmetic functions of several variables . . . . . . . . . . . . . . . . 35
4.14 Identities concerning Dirichlet characters . . . . . . . . . . . . . . . . . . . . . . . 35
4.15 Identities concerning additive characters . . . . . . . . . . . . . . . . . . . . . . . 36
4.16 Identities with respect to sets of units . . . . . . . . . . . . . . . . . . . . . . . . 37
4.17 Identities derived from group actions by the orbit counting lemma . . . . . . . . 37
4.18 Identities concerning subsets of the set {1, . . . , n} . . . . . . . . . . . . . . . . . . 38
4.19 Identities in Dedekind domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.20 Identities in the ring of algebraic integers . . . . . . . . . . . . . . . . . . . . . . 40
4.21 Final remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2
1 Introduction
1.1 Motivation
Menon’s identity states that for every n ∈ N := {1, 2, . . .},
X
M (n) := (a − 1, n) = ϕ(n)τ (n), (1)
a (mod n)
(a,n)=1

where a runs through a reduced residue system (mod n), (k, n) stands P for the greatest common
divisor of k and n, ϕ(n) is Euler’s totient function, and τ (n) = d|n 1 is the divisor function.
Identity (1) was proved by the Indian mathematician P. K. Menon1 [50] in 1965. Note that (1)
is implicitly contained in a more general earlier result, regarding n-even functions, due to E.
Cohen 2
P [18] in 1959. P See Section 2.9. Other notations,
∗
used in the literature, for the above sum
are a∈U(Zn ) and a∈Z∗n , where U(Zn ) = Zn denotes the group of units of the ring Zn = Z/nZ
of residue classes (mod n). However, we prefer to use the notation in (1).
This arithmetic identity is interesting in itself, it is related to certain number-theoretic,
group-theoretic and combinatorial properties, and has several different proofs by using pure
number-theoretic arguments, the orbit counting lemma, the finite Fourier representation of n-
even functions and Vaidyanathaswamy’s3 class division of integers (mod n), respectively. Also,
there are many possibilities to obtain generalizations and analogs, which can also be combined.
For example: How to modify identity (1) if (a − 1, n) is replaced by f ((a − 1, n)), where f is an
arithmetic function, or by (P (a), n), where P is an arbitrary polynomial with integer coefficients,
in particular P (a) = a − s, with s a given integer? As answers to these questions see identities
(7), (32), (52), (48) and (53).
The sum M (n) is related to the gcd-sum function. See Section 1.2. Another identity, similar
to (1), states that for every n ∈ N and s ∈ Z,
X
Rs (n) := cn (a − s) = µ(n)cn (s), (2)
a (mod n)
(a,n)=1

obtained in 1959 by Cohen [18, Cor. 7.2], where µ(n) is the Möbius function and cn (s) is the
Ramanujan sum4 , defined by X
cn (k) = e(ak/n), (3)
a (mod n)
(a,n)=1

with e(x) = e2πix .

For a common generalization of (1) and (2) in terms of n-even functions, in connection with
the Brauer-Rademacher identity see (47), (48) and (49).
1
Puliyakot Kesava Menon (1917-1979)
2
Eckford Cohen (1920-2005), an American mathematician
3
Ramaswamy S. Vaidyanathaswamy (1894-1960), an Indian mathematician
4
named after Srinivasa Ramanujan (1987-1920), the most famous Indian mathematician

3
 Menon’s identity has been reproved and generalized in various directions by several authors,
also in many recent papers. However, it seems that the results and proofs in Menon’s original
paper and in some other papers are not widely known. The goals of the present survey are
to present detailed proofs of this identity by using different methods, and to review its gener-
alizations and analogs. We point out those proofs and formulas that we could not identify in
the literature. We only discuss the directions of generalizations, and present certain identities
obtained by one or two steps of generalization (some of them with proofs), but not the most
general ones. We also overview the results and proofs given by Menon in his paper [50], and
offer some further historical remarks. An updated list of references is included.

1.2 The gcd-sum function

The gcd-sum function, also called Pillai function5 is
X X
P (n) := (a, n) = dϕ(n/d) (n ∈ N). (4)
a (mod n) d|n

The given representation follows by grouping

P the terms according to the values (a, n) = d.
Alternatively, using the Gauss formula n = d|n ϕ(d) we have

n X n X n/d
X X X X X ϕ(d)
P (n) = ϕ(d) = ϕ(d) = ϕ(d) 1=n ,
a=1 d|(a,n) a=1 d|a
d
d|n b=1 d|n
d|n

showing that P = id ∗ϕ, in terms of the Dirichlet convolution. It turns out that the function P
is multiplicative, and direct computations show (see Section 2.1) that for every n ∈ N,
Y
Y ν/(ν + 1)

ν ν−1
P (n) = (ν + 1)p − νp = nτ (n) 1− .
ν ν
p
p ||n p ||n

See the surveys by Haukkanen [31] and Tóth [80] on further properties of the gcd-sum
function and its generalizations and analogs.

1.3 Notation
Throughout this survey we use the following main notations. Those not included here are
explained in the text.
• N = {1, 2, . . .},
• P is the set of primes, Q
• the prime power factorization of n ∈ N is n = p∈P pνp (n) , where all but a finite number
of the exponents νp (n) are zero,
• (n1 , . . . nk ) and gcd(n1 , . . . nk ) denote the greatest common divisor of n1 , . . . , nk ∈ N,
• [n1 , . . . , nk ] and lcm(n1 , . . . , nk ) denote the least common multiple of n1 , . . . , nk ∈ N,
5
named after Subbayya Sivasankaranarayana Pillai (1901-1950), another Indian mathematician

4
 • Cn is the cyclic group of order n,
• Zn = Z/nZ is the group/ring of residue classes (mod n),
• U(Zn ) = Z∗n is the group of units of the ring Zn ,
• 1 is the function 1(n) = 1 (n ∈ N),
• id is the function id(n) = n (n ∈ N),
• µ is the Möbius function,
• ω(n) is the
P number of distinct prime factors of n ∈ N,
s
• σs (n) = d|n d (s ∈ R),
• σ(n) = σ1 (n) is the sum of divisors of n,
• τ (n) = σ0 (n) is the number of divisors of n, Q
• Js is the Jordan function of order s given by Js (n) = ns p|n (1 − 1/ps ) (s ∈ R),
• ϕ = J1 isPEuler’s totient function,
• P (n) = na=1 (a, n) is the gcd-sum function,
• e(x) = e2πix ,
• cq (n) are thePRamanujan sums,
• (f ∗ g)(n) = d|n f (d)g(n/d) is the Dirichlet convolution of the arithmetic functions f and
g, P
• (f ⊗ g)(k) = a+b≡k (mod n) f (a)g(b) is the Cauchy convolution of the n-periodic functions
f and g,
• d || n means that
P d is a unitary divisor of n, i.e., d | n and (d, n/d) = 1,
• (f × g)(n) = d||n f (d)g(n/d) is the unitary convolution of the arithmetic functions f and
g, P
• σs∗ (n) = d||n ds (s ∈ R),
• σ ∗ (n) = σ1∗ (n) is the sum of unitary divisors of n,
• τ ∗ (n) = 2ω(n) is the number of unitary divisors of n.

2 Proofs of the classical Menon identity

In this Section we present direct, self contained proofs of the Menon identity (1). Perhaps
the first idea that comes to mind is to show that the given sum is a multiplicative function
of n (Method I). The condition (a, n) = 1 can be treated by the inclusion-exclusion principle
(Method II). Alternatively, the Möbius function, respectively exponential sums can be used,
combined with certain convolutional identities (Methods III, IV, V, VII, VIII). Application of
the orbit counting lemma provides a different proof (Method VI). Finite Fourier representa-
tions of n-even functions can also be applied (Method IX). Finally, some properties related to
Vaidyanathaswamy’s class division of integers (mod n) lead to another proof (Method X).
A variant of Method I, as well as Methods VI and X have been used by Menon [50]. The
approach of Methods IV and V goes back to Sita Ramaiah [67]. Method IX was first applied
by Cohen [18] and Nageswara Rao [56]. See Section 3 for more details. We are not aware of
references concerning Methods II, III, VII and VIII.

5
2.1 Method I: Proof by multiplicativity
Both sides of identity (1) are multiplicative in n. This is clear for the right hand side, and
we prove it for the left hand side M (n). Let (n1 , n2 ) = 1. We apply the following well-known
property: If a1 runs through a reduced residue system (mod n1 ) and a2 runs through a reduced
residue system (mod n2 ), then a = a1 n2 + a2 n1 runs through a reduced residue system (mod
n1 n2 ). See, e.g., Hardy and Wright [25, Th. 61]. Hence
X X
M (n1 n2 ) = (a1 n2 + a2 n1 − 1, n1 n2 )
a1 (mod n1 ) a2 (mod n2 )
(a1 ,n1 )=1 (a2 ,n2 )=1

X X
= (a1 n2 + a2 n1 − 1, n1 )(a1 n2 + a2 n1 − 1, n2 )
a1 (mod n1 ) a2 (mod n2 )
(a1 ,n1 )=1 (a2 ,n2 )=1
X X
= (a1 n2 − 1, n1 ) (a2 n1 − 1, n2 ) = M (n1 )M (n2 ),
a1 (mod n1 ) a2 (mod n2 )
(a1 ,n1 )=1 (a2 ,n2 )=1

where we use that a1 n2 runs through a reduced residue system (mod n1 ) and a2 n1 runs through
a reduced residue system (mod n2 ).
Now, for a prime power pν (ν ≥ 1) we have
ν−1
pX
X X
ν ν ν
M (p ) = (a − 1, p ) = (a − 1, p ) − (jp − 1, pν )
a (mod pν ) a (mod pν ) j=1
(a,p)=1

ν−1
pX
X
ν
= (a, p ) − 1 = P (pν ) − pν−1 .
a (mod pν ) j=1

To compute P (pν ), which is the value of the gcd-sum function for n = pν , one can use
the convolutional representation (4). To give here a direct argument, let us group the terms
according to the values (a, pν ) = pt with 0 ≤ t ≤ ν. Then a = bpt , where 1 ≤ b ≤ pν−t and p ∤ b.
The number of such values b is ϕ(pν−t ). Therefore,
ν
X ν−1
X
ν t ν−t ν
P (p ) = p ϕ(p )=p + pt (pν−t − pν−t−1 ) = (ν + 1)pν − νpν−1 . (5)
t=0 t=0

We deduce that

M (pν ) = (ν + 1)pν − νpν−1 − pν−1 = (ν + 1)(pν − pν−1 ) = ϕ(pν )τ (pν ).

This completes the proof of (1).

6
2.2 Method II: Proof by the inclusion-exclusion principle
Let n = pν11 · · · pνr r > 1. We have by the inclusion-exclusion principle,
n
X n
X n
X X
M (n) = (a − 1, n) = (a − 1, n) − (a − 1, n)
a=1 a=1 1≤j≤r a=1
(a,n)=1 pj |a

X n
X X n
X n
X
r
+ (a − 1, n) − (a − 1, n) + · · · + (−1) (a − 1, n).
1≤j<k≤r a=1 1≤j<k<ℓ≤r a=1 a=1
pj pk |a pj pk pℓ |a p1 ···pr |a

Here
n
X n
X
(a − 1, n) = (a, n) = P (n),
a=1 a=1

the gcd-sum function. Furthermore, for every j,

n n/pj
X X ν
Sj := (a − 1, n) = (bpj − 1, n/pj j ),
a=1 b=1
pj |a

ν
since (bpj − 1, pj j ) = 1. We deduce
ν −1 νj
pj j
n/p
X Xj  ν ν

Sj = ((k − 1)n/pj j + c)pj − 1, n/pj j
k=1 c=1

ν −1 νj ν −1 νj
pj j
n/p pj j
n/p
X Xj  ν −1 ν
 X Xj  ν

= (k − 1)n/pj j + cpj − 1, n/pj j = cpj − 1, n/pj j
k=1 c=1 k=1 c=1
ν −1
pj j
X ν ν −1 ν
= P (n/pj j ) = pj j P (n/pj j ),
k=1
ν
using that together with c, the values cpj run through a complete residue system (mod n/pj j ).
That is,
n ν −1
X pj j
(a − 1, n) = P (n) ν .
a=1
P (pj j )
pj |a

In a similar way,
n ν −1 νk −1
X pj j
pk
(a − 1, n) = P (n) νj ,
a=1
P (pj )P (pνkk )
pj pk |a

7
 n ν −1 νk −1 νℓ −1
X pj j
pk pℓ
(a − 1, n) = P (n) νj ,
a=1
P (pj )P (pνkk )P (pνℓ ℓ )
pj pk pℓ |a

etc., and we have

ν −1 ν −1 νk −1
X pj j X pj j
pk
M (n) = P (n) − P (n) ν + P (n) νj
1≤j≤r
P (pj j ) 1≤j<k≤r
P (pj )P (pνkk )

ν −1 νk −1 νℓ −1
X pk pℓ pj j p1ν1 −1 · · · prνr −1
r
−P (n) νj + · · · + (−1) P (n)
P (pj )P (pνkk )P (pνℓ ℓ ) P (pν11 ) · · · P (pνr r )
1≤j<k<ℓ≤r
ν −1 !  
Y pj j Y pν−1
= P (n) 1− ν = P (pν ) 1 −
P (pj j ) P (pν )
1≤j≤r pν ||n
Y
Y

= P (pν ) − pν−1 = (ν + 1)pν − νpν−1 − pν−1 =
pν ||n pν ||n
Y
= (ν + 1)(pν − pν−1 ) = ϕ(n)τ (n),
pν ||n

by using (5). This proof is similar to the well known proof for Euler’s ϕ function.

2.3 Method III: Proof by convolutional identities

Slightly more generally, let f be an arbitrary arithmetic function. We have, by grouping the
terms according to the values (a − 1, n) = d,
n
X X n
X
Mf (n) := f ((a − 1, n)) = f (d) 1.
a=1 d|n a=1
(a,n)=1 (a,n)=1
d|a−1
((a−1)/d,n/d)=1

Now using the property of the Möbius µ function, namely,

(
X 1, if n = 1,
µ(d) =
d|n
0, otherwise,

we have
X n
X X X X n
X
Mf (n) = f (d) µ(δ) = f (d) µ(δ) 1
d|n a=1 δ|((a−1)/d,n/d) d|n δ|n/d a=1
(a,n)=1 (a,n)=1
d|a−1 d|a−1
δ|(a−1)/d

8
 X X n
X
= f (d) µ(δ) 1. (6)
d|n δ|n/d a=1
(a,n)=1
a≡1 (mod dδ)

We need the following lemma.

Lemma 2.1. Let n, d ∈ N, d | n and let x ∈ Z. Then
n
( ϕ(n)
X , if (x, d) = 1,
1 = ϕ(d)
a=1 0, otherwise.
(a,n)=1
a≡x (mod d)

For any d and δ such that d | n and δ | n/d we have (dδ) | n. Applying Lemma 2.1 to (6),
with dδ instead of d, and with x = 1, we deduce that
X X µ(δ) X 1 X
Mf (n) = ϕ(n) f (d) = ϕ(n) f (d)µ(δ).
ϕ(dδ) ϕ(t)
d|n δ|n/d t|n dδ=t

We obtain the following result.

Theorem 2.2. Let f be an arbitrary arithmetic function. Then for every n ∈ N,
n
X X (µ ∗ f )(d)
Mf (n) := f ((a − 1, n)) = ϕ(n) , (7)
a=1
ϕ(d)
d|n
(a,n)=1

where ∗ denotes the Dirichlet convolution.

Now, if f = id, where id(n) = n (n ∈ N), then µ ∗ id = ϕ, and we deduce (1).
Lemma 2.1 is well-known in the literature, usually proved by the inclusion-exclusion principle.
See, e.g., Apostol [2, Th. 5.32]. Here we use a different approach.

Proof of Lemma 2.1. For each term of the sum, since (a, n) = 1, we have (x, d) = (a, d) = 1. We
can assume that (x, d) = 1 (otherwise the sum is empty, and equals zero). Using the property
of the Möbius function, the given sum, say T , can be written as
n n/e
X X X X
T = µ(e) = µ(e) 1.
a=1 e|(a,n) e|n j=1
a≡x (mod d) ej≡x (mod d)

Let d and e be such that d | n and e | n. The linear congruence ej ≡ x (mod d) has solutions
in j if and only if (d, e) | x, equivalent to (d, e) = 1, since (x, d) = 1. For (d, e) = 1 there is
one solution (mod d), and there are n/(de) solutions (mod n/e), since n/e = (n/(de)) · d. Here
n/(de) is an integer by the condition (d, e) = 1. We obtain
X  
n n X µ(e) nY 1
T = µ(e) · = = 1−
de d e d p
e|n e|n p|n
(e,d)=1 (e,d)=1 p∤d

9
    
nY 1 Y 1 −1 n ϕ(n) d ϕ(n)
= 1− 1− = · · = .
d p p d n ϕ(d) ϕ(d)
p|n p|d

2.4 Method IV: A variant of Method III

P
The proof by Method III can be shortened by using first P
the identity f (n) = d|n (µ ∗ f )(d).
Note that in the case f = id, this is the Gauss formula n = d|n ϕ(d). We deduce
n
X n
X X
Mf (n) := f ((a − 1, n)) = (µ ∗ f )(d)
a=1 a=1 d|(a−1,n)
(a,n)=1 (a,n)=1

n
X X X n
X
= (µ ∗ f )(d) = (µ ∗ f )(d) 1.
a=1 d|a−1 d|n a=1
(a,n)=1 d|n (a,n)=1
a≡1 (mod d)

Here the inner sum is ϕ(n)/ϕ(d) by Lemma 2.1, which proves identity (7). In the case f = id
we obtain (1).

2.5 Method V: Another variant of Method III

The proof by Method III can also be presented in a slightly different form, by changing the
order of summation. Namely, by using first the property of the Möbius µ function,
n
X n
X X
Mf (n) := f ((a − 1, n)) = f ((a − 1, n)) µ(d)
a=1 a=1 d|(a,n)
(a,n)=1

n
X X X n
X
= f ((a − 1, n)) µ(d) = µ(d) f ((a − 1, n)).
a=1 d|a, d|n d|n a=1
d|a

Here we need the following lemma.

Lemma 2.3. If f is an arithmetic function and n, d ∈ N such that d | n, then

n
X n X (µ ∗ f )(e)
f ((a − 1, n)) = .
a=1
d e
e|n
d|a (e,d)=1

Using Lemma 2.3 we deduce

X n X (µ ∗ f )(e) X (µ ∗ f )(e) X µ(d)
Mf (n) = µ(d) =n , (8)
d e e d
d|n e|n e|n d|n
(e,d)=1 (d,e)=1

10
where the inner sum is (ϕ(n)/n)(ϕ(e)/e)−1 (see the proof of Lemma 2.1), giving
X (µ ∗ f )(e)
Mf (n) = ϕ(n) ,
ϕ(e)
e|n

which is identity (7).

P
Proof of Lemma 2.3. By using that f (n) = d|n (µ ∗ f )(d), we deduce

n n/d n/d
X X X X
U := f ((a − 1, n)) = f ((jd − 1, n)) = (µ ∗ f )(e)
a=1 j=1 j=1 e|jd−1, e|n
d|a

n/d
X X
= (µ ∗ f )(e) 1,
e|n j=1
jd≡1 (mod e)

where the inner sum is, see the proof of Lemma 2.1, n/(de) if (d, e) = 1 and 0 otherwise. This
gives
X n n X (µ ∗ f )(e)
U= (µ ∗ f )(e) · = .
de d e
e|n e|n
(e,d)=1 (e,d)=1

We remark that by grouping the terms in (8) according to the values de = δ we have
X1 X Xn
Mf (n) = n µ(d)(µ ∗ f )(e) = (µ × (µ ∗ f ))(δ)
δ δ
δ|n de=δ δ|n
(d,e)=1

= (id ∗(µ × (µ ∗ f )))(n), (9)

where × is the unitary convolution. If f is multiplicative, then it follows that Mf is also
multiplicative, and for every prime power pν (ν ≥ 1),
ν
X
ν ν
Mf (p ) = (id ∗(µ × (µ ∗ f )))(p ) = pν−j (µ × (µ ∗ f ))(pj )
j=0

ν
X ν
X
= pν + pν−j (µ(pj ) + (µ ∗ f )(pj )) = pν−j (µ ∗ f )(pj ) − pν−1
j=1 j=0
ν ν−1
= (id ∗µ ∗ f )(p ) − p = (ϕ ∗ f )(pν ) − pν−1 ,
giving the identity Y

Mf (n) = (ϕ ∗ f )(pν ) − pν−1 . (10)
pν ||n

11
2.6 Method VI: Proof by the orbit counting lemma
The orbit counting lemma or Cauchy-Frobenius-Burnside lemma can be stated as follows.
See, e.g., Bogart [4], Neumann [60], Rotman [64, Ch. 2], Wright [104] for its proof, history and
applications.

Lemma 2.4. Let G be a finite group and X be a finite set. Let ψ : G × X → X, ψ(g, x) = gx
(g ∈ G, x ∈ X) be an action of G on the set X. Let X/G be the set of orbits and let X g = {x ∈
X : gx = x} be the set of elements x ∈ X fixed by g ∈ G. Then
1 X g
|X/G| = |X |.
|G|
g∈G

Apply this lemma in the case of G = U (Zn ) = {k (mod n) : (k, n) = 1}, the group of units
of the ring Zn of residue classes (mod n), which is a group of order ϕ(n), and X = {1, 2, . . . , n}.
For any k ∈ U (Zn ) and x ∈ {1, 2, . . . , n} let ψ(k, x) = ℓ, where kx ≡ ℓ (mod n), which is a group
action. Here X k = {x : 1 ≤ x ≤ n, kx ≡ x (mod n)} = {x : 1 ≤ x ≤ n, (k − 1)x ≡ 0 (mod n)}.
Note that |X k | = (k − 1, n).
To determine the number of orbits, we need the following lemma.

Lemma 2.5. Let a, b ∈ Z and n ∈ N be given integers. The congruence ax ≡ b (mod n) has
solutions x such that (x, n) = 1 if and only if (a, n) = (b, n).

This lemma shows that the orbits of the considered group action are Od = {k : 1 ≤ k ≤
n, (k, n) = d}, where d is a divisor of n. Therefore, the number of orbits is |X/G| = τ (n). This
completes the proof of identity (1).

Proof of Lemma 2.5. Assume that there is a solution x = t satisfying at ≡ b (mod n) and
(t, n) = 1. Then (at, n) = (b, n). Since (t, n) = 1, we obtain that (a, n) = (b, n).
Conversely, suppose that (a, n) = (b, n) = d. Let us denote A = a/d, B = b/d, N = n/d.
Then (A, N ) = (B, N ) = 1. Hence the linear congruence Ax ≡ B (mod N ) has a solution x = x0
(it has, in fact, one solution x modulo N ) satisfying Ax0 ≡ B (mod N ). Note that (Ax0 , N ) =
(B, N ) = 1, hence (x0 , N ) = 1. It follows that the solutions of the original congruence ax ≡ b
(mod n) are xj = x0 + jN , where 0 ≤ j ≤ d − 1. We have to show that there exist solutions xj
such that (xjQ , n) = 1.
Let C = p|d, p∤N p, where (C, N ) = 1. By the Chinese remainder theorem, for the simulta-
neous congruences y ≡ x0 (mod N ), y ≡ 1 (mod C) there is a solution y = y0 satisfying y0 ≡ x0
(mod N ), y0 ≡ 1 (mod C). Hence (y0 , N ) = (x0 , N ) = 1 and (y0 , C) = 1.
This gives (y0 , n) = 1 and Ay0 ≡ Ax0 ≡ B (mod N ). Hence ay0 ≡ b (mod n), and y0 is a
solution coprime to n.

It can be proved that in the case (a, n) = (b, n) = d, the number of incongruent solutions of
the congruence ax ≡ b (mod n) is

ϕ(n) Y  1

=d 1− . (11)
ϕ(n/d) p
p|n
p∤ n/d

12
 See Bibak et al. [3, Th. 3.1], Grošek and Porubský [24, Th. 2] for the use of two different
approaches to deduce (11). Also see paper Garcia and Ligh [23], defining a generalization of
Euler’s totient function for arithmetic progressions, and deducing the same formula (11).

2.7 Method VII: Proof by exponential sums

We have by denoting a − 1 = b,
X X
M (n) := (a − 1, n) = (b, n).
a (mod n) a,b (mod n)
(a,n)=1 a−1≡b (mod n)
(a,n)=1

By the familiar orthogonality property

(
1 X 1, if n | k,
e(kj/n) = (12)
n 0, otherwise,
j (mod n)

we deduce that X X
1
M (n) = (b, n) e((1 − a + b)j/n)
n
a,b (mod n) j (mod n)
(a,n)=1

1 X X X
= e(j/n) (b, n)e(bj/n) e(−aj/n). (13)
n
j (mod n) b (mod n) a (mod n)
(a,n)=1

Here the last sum is Ramanujan’s sum cn (−j) = cn (j), and we use its Hölder’s evaluation,
namely
ϕ(n)µ(n/(j, n))
cn (j) = (j, n ∈ N). (14)
ϕ(n/(j, n))
We also need the next result. See Tóth [83, Prop. 7].
Lemma 2.6. For every n ∈ N and k ∈ Z,
n
X X
(a, n)e(ka/n) = dϕ(n/d). (15)
a=1 d|(k,n)

According to (13), (14) and (15),

ϕ(n) X X µ(n/(j, n))
M (n) = e(j/n) dϕ(n/d) ,
n ϕ(n/(j, n))
j (mod n) d|(j,n)

and by grouping the terms of the last sum according to the values (j, n) = δ with j = δt,
(t, n/d) = 1 we have
ϕ(n) X µ(n/δ) X X
M (n) = dϕ(n/d) e(t/(n/δ)),
n ϕ(n/δ)
δ|n d|δ t (mod n/δ)
(t,n/δ)=1

13
the last sum being cn/δ (1) = µ(n/δ) by (14). This gives

ϕ(n) X µ2 (n/δ) X ϕ(n) X µ2 (e)

M (n) = dϕ(n/d) = dϕ(ke) (16)
n ϕ(n/δ) n ϕ(e)
δ|n d|δ dek=n

ϕ(n) X X µ2 (e)
= dϕ(m) ,
n ϕ(e)
dm=n ke=m

where the inner sum is m/ϕ(m) (Landau’s identity), and obtain

ϕ(n) X X
M (n) = dm = ϕ(n) 1 = ϕ(n)τ (n).
n
dm=n dm=n
P
Proof of Lemma 2.6. Using the Gauss formula n = d|n ϕ(d) we have
n
X n
X X X n
X
(a, n)e(ka/n) = e(ka/n) ϕ(d) = ϕ(d) e(ka/n)
a=1 a=1 d|(a,n) d|n a=1
d|a

n/d
X X X n X
= ϕ(d) e(kb/(n/d)) = ϕ(d) = dϕ(n/d),
d
d|n b=1 d|n d|(k,n)
n/d |k

applying (12).

2.8 Method VIII: Another proof by exponential sums

Observe that
n
X n
X n
X
M (n) := (a − 1, n) = (a − 1, n), (17)
a=1 b=1 a=1
(a,n)=1 (b,n)=1 ab≡1 (mod n)

since for every fixed b with 1 ≤ b ≤ n and (b, n) = 1 there is exactly one integer a such that
1 ≤ a ≤ n and ab ≡ 1 (mod n), whence (a, n) = 1; and for different values of b the corresponding
values of a are different. Note that the condition (b, n) = 1 can be removed in (17), but we keep
it and use it in what follows.
By property (12) we deduce that
n n n
1 X X X
M (n) = (a − 1, n) e((ab − 1)j/n))
n
b=1 a=1 j=1
(b,n)=1

n n n
1X X X
= e(−j/n) (a − 1, n)e(abj/n)).
n
j=1 b=1 a=1
(b,n)=1

14
 According to (15),
n
X n
X
(a − 1, n)e(abj/n)) = (a, n)e((a + 1)bj/n))
a=1 a=1
X X
= e(bj/n) dϕ(n/d) = e(bj/n) dϕ(n/d),
d|(bj,n) d|(j,n)

with (b, n) = 1. This gives

n n/d
1X X X
M (n) = dϕ(n/d) e((b − 1)t/(n/d)),
n
d|n b=1 t=1
(b,n)=1

where, by (12) again, the last sum is n/d if b ≡ 1 (mod n/d) and 0 otherwise. Hence
X n
X X ϕ(n)
M (n) = ϕ(n/d) 1= ϕ(n/d) = ϕ(n)τ (n),
ϕ(n/d)
d|n b=1 d|n
(b,n)=1
b≡1 (mod n/d)

by using Lemma 2.1. Note that Ramanujan sums are not needed here.
One may wonder whether this method works by using character sums instead of exponential
sums. Well, using the orthogonality property
(
1 X 1, if u ≡ v (mod n),
χ(u)χ(v) = (18)
ϕ(n) χ 0, otherwise,

where the sum is over the Dirichlet characters χ (mod n), from (17) we have
n n
1 XX X
M (n) = (a − 1, n) χ(ab)χ(1)
ϕ(n) χ
b=1 a=1

n
! n !
1 X X X
= χ(b) (a − 1, n)χ(a) . (19)
ϕ(n) χ
b=1 a=1

However, as well-known,
n
(
X ϕ(n), if χ = χ0 , the principal character,
χ(b) = (20)
b=1
0, otherwise,

therefore (19) gives nothing but the definition of the sum M (n). On the other hand, the sum
n
X
(a − 1, n)χ(a)
a=1

can be computed for an arbitrary Dirichlet character (mod n), and it gives a generalization of
Menon’s identity. See Section 4.14.

15
2.9 Method IX: Proof via the finite Fourier representation of n-even functions
A function f : N → C is said to be an n-even function – or even function (mod n) – if
f (k) = f ((k, n)) for all k ∈ N, where n ∈ N is fixed and (k, n) is the gcd of k and n. Note
that if f is n-even, then f is n-periodic, and by this periodicity it can be extended to a function
defined on Z. For example, the function k 7→ (k, n) is n-even. More generally, k 7→ F ((k, n)) is
n-even, where F is an arbitrary arithmetic function. Another example of an n-even function is
k 7→ cn (k), where cn (k) is the Ramanujan sum, defined by (3), and evaluated as
X
cn (k) = dµ(n/d) (k, n ∈ N).
d|(k,n)

The set En of n-even functions forms a τ (n) dimensional linear space, and the Ramanujan
sums k 7→ cd (k) with d | n form a basis of the space En . Consequently, every n-even function f
has a (Ramanujan-)Fourier expansion of the form
X
f (k) = αf (d)cd (k) (k ∈ N), (21)
d|n

where the (Ramanujan-)Fourier coefficients αf (d) (d | n) of f are uniquely determined and given
by
1X
αf (d) = f (e)cn/e (n/d)
n
e|n

1 X
= e(µ ∗ f )(n/e). (22)
n
e|n/d

In order to deduce a Menon-type identity, related to the finite Fourier expansion of n-even
functions, we need the following further preliminaries.
The Cauchy convolution of the n-periodic functions f and g is defined by
X
(f ⊗ g)(k) = f (a)g(b) (k ∈ N).
a+b≡k (mod n)

For example, the Cauchy convolution of the exponential functions f (a) = e(as/n) and g(b) =
e(bt/n) is
(
X ne(ks/n), if s ≡ t (mod n),
(f ⊗ g)(k) = e(as/n)e(bt/n) = (23)
a+b≡k (mod n)
0, otherwise.

We need the following lemmas.

Lemma 2.7. If d | n and e | n, then the Cauchy convolution of the Ramanujan sums a 7→ cd (a)
and b 7→ ce (b) is (
X ncd (k), if d = e,
cd (a)ce (b) = (24)
a+b≡k (mod n)
0, otherwise.

16
Proof of Lemma 2.7. Let n = dd1 = ee1 . Then

X X d
X e
X
cd (a)ce (b) = e(aj/d) e(ℓb/e)
a+b≡k (mod n) a+b≡k (mod n) j=1 ℓ=1
(j,d)=1 (ℓ,e)=1

d
X e
X X
= e(ajd1 /n)e(bℓe1 /n)
j=1 ℓ=1 a+b≡k (mod n)
(j,d)=1 (ℓ,e)=1

d
X e
X X
= ne(kjd1 /n),
j=1 ℓ=1 jd1 ≡ℓe1 (mod n)
(j,d)=1 (ℓ,e)=1

according to (23). Here 1 ≤ d1 ≤ jd1 ≤ dd1 = n, 1 ≤ e1 ≤ ℓe1 ≤ ee1 = n, hence jd1 ≡ ℓe1 (mod
n) holds if and only if jd1 = ℓe1 , that is jd1 de = ℓe1 de, jne = ℓdn, je = ℓd. Taking into account
that (j, d) = 1, (ℓ, e) = 1, we deduce that j = ℓ and d = e. Hence, if d 6= e, then the sum is zero
(empty) and if d = e, then it is
d
X
n e(jk/d) = ncd (k),
j=1
(j,d)=1

completing the proof.

Lemma 2.8. If f and g are n-even functions, then the Cauchy convolution f ⊗ g is also n-even,
and its Fourier coefficients are

αf ⊗g (d) = n αf (d)αg (d) (d | n). (25)

Proof of Lemma 2.8. We have

X
(f ⊗ g)(k) = f (a)g(b)
a+b≡k (mod n)

X X X
= αf (d)cd (a) αg (e)ce (b)
a+b≡k (mod n) d|n e|n
XX X X
= αf (d)αg (e) cd (a)ce (b) = nαf (d)αg (d)cd (k),
d|n e|n a+b≡k (mod n) d|n

by using (24). Since for d | n the function k 7→ cd (k) is n-even we deduce that f ⊗ g is also
n-even, and its Fourier coefficients are nαf (d)αg (d).

See Haukkanen [29], McCarthy [49, Ch. 2], Montgomery [54, Ch. 2], Tóth and Haukkanen
[92] for more details on these notions and properties.

17
 Now, for the proof of a Menon-type identity let f be an arbitrary n-even function with
Fourier coefficients αf (d) (d | n). Consider the sum
X
Sf (k, n) := f (k − b).
b (mod n)
(b,n)=1

If f (b) = (b, n) (b ∈ N) is the gcd function and k = 1, then Sf (1, n) is exactly the sum
appearing in Menon’s identity.
For any n-even function f and any k ∈ N the sum Sf (k, n) can be written as
X X
Sf (k, n) = f (a) = f (a)δ(b, n), (26)
a+b≡k (mod n) a+b≡k(mod n)
(b,n)=1

which is the Cauchy convolution of the function f and the Kronecker δ, given here by δ(b, n) = 1
if (b, n) = 1, and δ(b, n) = 0 if (b, n) > 1. Note that δ is n-even.
The main observation, which leads to the proof of a Menon-type identity is that the function
k 7→ Sf (k, n) is also n-even, being the Cauchy convolution of n-even functions, by Lemma 2.8.
The Fourier coefficients of δ are
1 ϕ(n)µ(d)
αδ (d) = cn (n/d) = (d | n), (27)
n nϕ(d)

the second equality being a consequence of Hölder’s evaluation (14) of the Ramanujan sums.
Now, (26) and (25) imply that
X
Sf (k, n) = n αf (d)αδ (d)cd (k).
d|n

By using (27) we deduce the following general identity, due to Cohen [18] in an implicit form.
See McCarthy [49, p. 85].

Theorem 2.9. Let f be an arbitrary n-even function with Fourier coefficients αf (d) (d | n).
Then
X X µ(d)
Sf (k, n) := f (k − b) = ϕ(n) αf (d) cd (k).
ϕ(d)
b (mod n) d|n
(b,n)=1

In fact, Cohen [18, Th. 1] stated the result of Lemma 2.8, and then presented a series of
corollaries, including formulas for the sums Sf (k, n) of above with special functions f , e.g.,
f (b) = τ ((b, n)), σ((b, n)), cn (b), (n, b)s , where cn (b) is the Ramanujan sum, s is real. See [18,
Cor. 2.2, 2.3, 7.2, 17.1].
Now we show how Menon’s identity is deduced from Theorem 2.9. Consider the function
f (b) = (b, n). If a | n, then we have
X X
(µ ∗ f )(a) = f (d)µ(a/d) = dµ(a/d) = ϕ(a).
d|a d|a

18
 Therefore, by using (22) we deduce that the Fourier coefficients α(d) of the function f (b) =
(b, n) are
1 X
α(d) = eϕ(n/e).
n
e|n/d

We obtain that
X ϕ(n) X µ(d) X
S(k, n) := (k − b, n) = cd (k) eϕ(n/e).
n ϕ(d)
b (mod n) d|n e|n/d
(b,n)=1

Now let k = 1. Then cd (1) = µ(d) and with the notation n = dt, t = n/d = ej,
ϕ(n) X µ2 (d) X ϕ(n) X µ2 (d)
S(1, n) = eϕ(n/e) = eϕ(dj),
n ϕ(d) n ϕ(d)
d|n e|n/d dej=n

which is identical to (16), and gives ϕ(n)τ (n).

2.10 Method X: Proof by Vaidyanathaswamy’s class division of integers (mod

n)
Group the elements a of the set S = {1, 2, . . . , n} according to the values (a, n) = d. Let
Sd = {a ∈ S : (a, n) = d}, where d | n. It is
Pclear that S = P∪d|n Sd and |Sd | = ϕ(n/d) for any
d | n. This leads to the Gauss formula n = d|n ϕ(n/d) = d|n ϕ(d).
Vaidyanathaswamy [93] defined the sum of the subsets Sd1 and Sd2 (d1 | n, d2 | n), denoted
by Sd1 ⊕ Sd2 , as the multiset
Sd1 ⊕ Sd2 = {x + y (mod n) : x ∈ Sd1 , y ∈ Sd2 }.
The main related property proved in paper [93] is that in Sd1 ⊕ Sd2 all elements of a class Sd
occur the same number of times, denoted by γ(d1 , d2 , d), which depends only on d1 , d2 and d.
The coefficients γ(d1 , d2 , d) can be evaluated in terms of the Ramanujan sums, which is a result
of Ramanathan [62].
Lemma 2.10. 1) For any d1 | n and d2 | n one has
[
Sd1 ⊕ Sd2 = γ(d1 , d2 , d)Sd ,
d|n

where γ(d1 , d2 , d)Sd means the multiset where each element occurs
1X
γ(d1 , d2 , d) = cn/d1 (δ)cn/d2 (δ)cn/δ (d). (28)
n
δ|n

times.
2) In the special case d1 = d2 = 1 we have for any d | n,
Y 1

γ(1, 1, d) = ϕ(n) 1− . (29)
p−1
p|n
p∤d

19
 Now, for the proof of Menon’s identity we follow the arguments given by Menon [50], and
explain some more details by using properties of n-even functions. Note that Menon [50] did
not use the concept of n-even functions, which has been introduced and investigated by Cohen
in a series of papers, including [15, 17, 18].
Let f be an arithmetic function, let d1 | n, d2 | n and consider the sum
X
F (n, d1 , d2 ) := f ((a + d2 , n)).
a (mod n)
(a,n)=d1

The following property is needed. See below its proof.

Lemma 2.11. The sum F (n, d1 , d2 ) is unchanged if d2 is replaced by any integer b such that
(b, n) = d2 .
By Lemma 2.11,
X
F (n, d1 , d2 ) = f ((a + b, n)), where (b, n) = d2 . (30)
a (mod n)
(a,n)=d1

If we sum equations (30) over the values of b, note that there are ϕ(n/d2 ) such values of b
(mod n), we obtain
1 X
F (n, d1 , d2 ) = f ((a + b, n)),
ϕ(n/d2 )
a,b (mod n)
(a,n)=d1
(b,n)=d2

and grouping the terms according to the values a + b = c we have

1 X X
F (n, d1 , d2 ) = f ((c, n)) 1.
ϕ(n/d2 )
c (mod n) a,b (mod n)
(a,n)=d1
(b,n)=d2
a+b≡c (mod n)

Here the inner sum is exactly the Cauchy convolution (̺1 ⊗ ̺2 )(c) of the functions ̺1 and
̺2 , where (
1, if (a, n) = dj ,
̺j (a) = (j = 1, 2). (31)
0, otherwise,
Since ̺1 and ̺2 are n-even, ̺1 ⊗ ̺2 is also n-even by Lemma 2.8, that is (̺1 ⊗ ̺2 )(c) =
(̺1 ⊗ ̺2 )((c, n)). We have
1 X
F (n, d1 , d2 ) = f ((c, n))(̺1 ⊗ ̺2 )((c, n))
ϕ(n/d2 )
c (mod n)

1 X
= f (d)ϕ(n/d)(̺1 ⊗ ̺2 )(d),
ϕ(n/d2 )
d|n

by grouping the terms according to the values (c, n) = d. Observe that (̺1 ⊗̺2 )(d) = γ(d1 , d2 , d),
given by (29). This leads to the following general result of Menon [50, Eq. (4.5)].

20
Theorem 2.12. Let f be an arithmetic function, let n ∈ N and d1 , d2 | n. Then
X 1 X
f ((a + d2 , n)) = f (d)γ(d1 , d2 , d)ϕ(n/d).
ϕ(n/d2 )
a (mod n) d|n
(a,n)=d1

If d1 = d2 = 1, then by (29) we deduce the next formula, given by Menon [50, Eq. (4.8)].

Corollary 2.13. Let f be an arithmetic function and n ∈ N. Then

X X Y 1

f ((a − 1, n)) = f (d)ϕ(n/d) 1− . (32)
p−1
a (mod n) d|n p|n
(a,n)=1 p∤d

We remark that it can be showed by direct computations that

Y 1
 X µ(δ)
ϕ(n/d) 1− = ϕ(n) , (33)
p−1 ϕ(dδ)
p|n δ|n/d
p∤d

and that for any function f ,

X X µ(δ) X (µ ∗ f )(d)
f (d) = , (34)
ϕ(dδ) ϕ(d)
d|n δ|n/d d|n

therefore (32), (33) and (34) imply formula (7), which reduces to Menon’s original identity in
the case f = id.

Proof of Lemma 2.10. 1) According to its definition,

X
γ(d1 , d2 , d) = 1 = (̺1 ⊗ ̺2 )(d),
a+b≡d (mod n)
(a,n)=d1
(b,n)=d2

representing the Cauchy convolution of the functions ̺1 and ̺2 , given by (31). As mentioned
above, the function ̺1 ⊗ ̺2 is n-even. Also, the Fourier coefficients of ̺j (j = 1, 2) are
1X 1
α̺j (d) = ̺j (e)cn/e (n/d) = cn/dj (n/d) (d | n).
n n
e|n

Therefore, by Lemma 2.8, the Fourier coefficients of ̺1 ⊗ ̺2 are

1
α̺1 ⊗̺2 (d) = c (n/d)cn/d2 (n/d) (d | n),
n n/d1
and from the Fourier representation (21) we obtain identity (28).
2) If d1 = d2 = 1, then (28) implies (29) by direct computations using properties of the
Ramanujan sums.

21
 Let d1 , d2 , d be divisors of n. It can be shown that γ(d1 , d2 , d) = 0, unless (d1 , d2 ) = (d1 , d) =
(d2 , d) = δ, when  
ϕ(n/d1 )ϕ(n/d2 ) Y 1
γ(d1 , d2 , d) = 1− .
ϕ(n/d) p−1
p|n/δ
p∤d/δ
p∤d1 /δ
p∤d2 /δ

For d1 = d2 = 1 this recovers formula (29). See Haukkanen [28] and Sivaramakrishnan [72,
Ch. XV] for more details.

Proof of Lemma 2.11. Consider the functions g(a) = f ((a, n)) and ̺d1 (a), defined by (31), which
are n-even functions. Also
X X
F (n, d1 , d2 ) = ̺d1 (a)g(a − d2 ) = ̺d1 (a)g(b) = (̺d1 × g)(d2 ),
a (mod n) a+b≡d2 (mod n)

the Cauchy convolution of the functions ̺d1 and g. Therefore the function d2 7→ F (n, d1 , d2 )
is also n-even (where d1 is fixed). So, if (b, n) = d2 , then F (n, d1 , b) = F (n, d1 , (b, n)) =
F (n, d1 , d2 ).

3 Historical remarks
3.1 Remarks on the paper by Menon
Menon [50] gave three different proofs of (1). His first proof is by using the orbit counting
lemma. See Section 2.6. The third proof is by using Vaidyanathaswamy’s class division of
integers (mod n). See Section 2.10. The second proof given by Menon [50] is purely number-
theoretic, by using multiplicative functions of several variables. Namely, he proved the following
result. See the survey paper by the author [86] on multiplicative functions of several variables.
Also see Subbarao [74].
Lemma 3.1 ([50, Lemma]). Let f be a multiplicative arithmetic function of r variables and let
Pi ∈ Z[x] (1 ≤ i ≤ r) be polynomials. Then the function
n
X
F (n) := f ((P1 (a), n), . . . , (Pr (a), n)) (35)
a=1

is multiplicative in the single variable n, where (Pi (a), n) (1 ≤ i ≤ r) is the gcd of Pi (a) and n.
For the proof, he showed that for any n1 , n2 ∈ N with (n1 , n2 ) = 1 one has F (n1 n2 ) =
F (n1 )F (n2 ). Then, by applying Lemma 3.1 in the case r = 2, P1 (x) = x − 1, P2 (x) = x and
f (n1 , n2 ) = f (n1 )E0 (n2 ), where E0 (n) = 1 for n > 1, E0 (1) = 1, the following result is deduced.
Theorem 3.2 ([50, Th. 1]). Let f be a multiplicative arithmetic function of a single variable.
Then  
Xn Y X ν
f ((a − 1, n)) =  ϕ(pν−j )f (pj ) − pν−1  . (36)
a=1 pν ||n j=0
(a,n)=1

22
 Here (36) is the same as identity (10), deduced by different arguments, and it has been
recovered by Sivaramakrishnan [70, Th. 6.1] by the theory of n-even functions.
By selecting f (n) = n in Theorem 3.2 we have the original identity (1). Some other particular
choices of f are also given. For example, see [50, p. 159],
n
X Y 1

σ((a − 1, n)) = n 1+ν− , (37)
a=1
p
pν ||n
(a,n)=1

n
X Y 2ν + 1 ν − 1

ϕ((a − 1, n)) = n 1+ν − + . (38)
p p2
a=1 pν ||n
(a,n)=1

The case f (n) = nk has been treated by Sivaramakrishnan [69]. It has been pointed out by
Tărnăuceanu [78, Cor. 2], applying the weighted form of the orbit counting lemma that
n
X
g((a − 1, n)) = ϕ(n)σ(n), (39)
a=1
(a,n)=1
P
where g = id ϕ ∗ 1, i.e., g(n) = d|n dϕ(d). Here (39) immediately follows from identity (7).
Menon also proved the following result, related to the gcd-sum function (4).

Theorem 3.3 ([50, Th. 2]). Let f be a multiplicative arithmetic function of a single variable.
Then  
Xn Y Xν
f ((a2 − a, n)) = 2 ϕ(pν−j )f (pj ) − pν  . (40)
a=1 pν ||n j=0

Then he presented some of its corollaries. For example, see [50, p. 161],
n
X
σ((a2 − a, n)) = nτ (n2 ).
a=1

Certain generalizations of (40) have been given by Venkatramaiah [96].

Note that by the convolution method, similar to that given in Section 2.4, for the function
(35) one has
X (µr ∗r f )(d1 , . . . , dr )
F (n) = n N (d1 , . . . , dr ), (41)
[d1 , . . . , dr ]
d1 |n,...,dr |n

valid for any function f of r variables, where µr is the r variables Möbius function, defined by
µr (d1 , . . . , dr ) = µ(d1 ) · · · µ(dr ), ∗r is the r variables Dirichlet convolution, and N (d1 , . . . , dr )
denotes the number of solutions (mod [d1 , . . . , dr ]) of the simultaneous congruences P1 (x) ≡ 0
(mod d1 ), ..., Pr (x) ≡ 0 (mod dr ). If f is multiplicative, then the convolution representation
(41) shows that F is also multiplicative, and its values can be computed for special choices of
the polynomials P1 (x), . . . , Pr (x). This is remarked by Haukkanen and Tóth [35, Remark 3.2].

23
3.2 Some more remarks
The orbit counting lemma was also applied by Richards [63] and Fung [22] to deduce Menon’s
identity (1). They do not cite the paper by Menon [50]. Richards [63] also proved the following
result: Let G be a group of order n and let c(G) denote the number of its cyclic subgroups.
Then c(G) ≥ τ (n) with equality if and only if G is cyclic.
It was noted by Sivaramakrishnan [68] that for every n ∈ N,
n
X
b((a, n)) = ϕ(n)τ (n), (42)
a=1
Q
where b(n) = pν ||n (pν − 1), called Venkataraman’s block-function in [68]. We remark that b(n)
is identical to the unitary Euler function ϕ∗ (n), defined by Cohen [20] as the number of integers
a such that 1 ≤ a ≤ n and (a, n)∗ = 1, where (a, n)∗ = max{d ∈ N : d | a, d || n}.
In fact, if f is an arbitrary function, then
n
X X
Pf (n) := f ((a, n)) = f (d)ϕ(n/d), (43)
a=1 d|n

the Dirichlet convolution of the function f and the Euler function ϕ, which follows by grouping
the terms according to the values (a, n) = d. If f = id, then this reduces to identity (4) regarding
the Pillai function P . If f is multiplicative, then Pf is multiplicative as well. Now identity (42)
is an easy consequence of (43).
Another direct consequence of (43) is the polynomial identity
n 
Y  Y ϕ(n/d)
x(a,n) − 1 = xd − 1 , (44)
a=1 d|n

valid for every n ∈ N, which is formally obtained for f (n) = log(xn − 1), and may be compared
to the well-known identity
Y µ(n/d)
Φn (x) = xd − 1
d|n

concerning the cyclotomic polynomials Φn (x).

Menon’s identity (1) and some of its generalizations are included in the following textbooks
and monographs: Freud and Gyarmati [21, Ex. 6.1.16], McCarthy [49, Chs. 1,2], Niven, Zucker-
man and Montgomery [61, Ex. 4.2.25], Sivaramakrishnan [72, Ch. VIII], [73, Ch. 7].

4 Generalizations and analogs

4.1 Identities for f ((a − 1, n))
As already mentioned in Sections 2.10 and 3.1, Menon [50] evaluated the sum
n
X
Mf (n) := f ((a − 1, n)).
a=1
(a,n)=1

24
 In the case f is an arbitrary arithmetic function, he deduced identity (32), which may be
compared to (43), concerning the gcd-sum function. If f is a multiplicative function, then Mf
is also multiplicative, and Menon [50] obtained identity (36).
For an arbitrary function f , identity (7) was proved by Sita Ramaiah [67, Th. 9.1] in a more
general form, concerning regular systems of divisors (see Section 4.9) and k-reduced residue
systems (see Section 4.6). For the special choices of f = σ and f = ϕ see identities (37) and
P(38),
respectively. Several other functions can be discussed. For example, if f (n) = ω(n) = p|n 1,
then one has
Xn X 1
Mω (n) := ω((a − 1, n)) = ϕ(n) ,
p−1
a=1 p|n
(a,n)=1

and if f (n) = log(xn − 1), then one obtains

n
Y   Y
(a−1,n)
x −1 = Φd (x)ϕ(n)/ϕ(d) , (45)
a=1 d|n
(a,n)=1

being a counterpart of identity (44). We are not aware of any references for (44) and (45).
A slightly different type of identity is the following. Let k, m1 , . . . , mk ∈ N, M = [m1 , . . . , mk ].
Tóth [82] proved by number-theoretic arguments, as a special case of more general identities
that
1 X X ϕ(d1 ) · · · ϕ(dk )
(a − 1, m1 ) · · · (a − 1, mk ) = . (46)
ϕ(M ) ϕ([d1 , . . . , dk ])
a (mod M ) d|m1 ,...,dk |mk
(a,M )=1

If k = 1, then (46) reduces to Menon’s identity (1). Note that, according to the orbit
counting lemma,
1 X
(a − 1, m1 ) · · · (a − 1, mk ),
ϕ(q)
a (mod q)
(a,q)=1

where q = m1 · · · mk , represents the number of cyclic subgroups of the group Cm1 × · · · × Cmk .

4.2 Identities concerning n-even functions f

The proof of the next identity only uses the definition of n-even functions and no further
properties of them. It is given by Tóth [87, Th. 2.1] in a more general form.
Theorem 4.1. Let n ∈ N, s ∈ Z and let f be an even function (mod n). Then
X X (µ ∗ f )(d)
f (a − s) = ϕ(n) . (47)
ϕ(d)
a (mod n) d|n
(a,n)=1 (d,s)=1

Proof of Theorem 4.1. Since f is an even function (mod n), for every k ∈ N we have
X
f (k) = f ((k, n)) = (µ ∗ f )(d),
d|(k,n)

25
and we follow Method IV:
X n
X X n
X X
f (a − s) = (µ ∗ f )(d) = (µ ∗ f )(d)
a (mod n) a=1 d|(a−s,n) a=1 d|a−s
(a,n)=1 (a,n)=1 (a,n)=1 d|n

X n
X
= (µ ∗ f )(d) 1.
d|n a=1
(a,n)=1
a≡s (mod d)

By Lemma 2.1 the inner sum is ϕ(n)/ϕ(d) if (d, s) = 1 and 0 otherwise, which proves identity
(47).

If f (a) = (a, n), then (47) gives

X X
(a − s, n) = ϕ(n) 1, (48)
a (mod n) d|n
(a,n)=1 (d,s)=1

which reduces to Menon’s identity (1) for s = 1. If f (a) = cn (a), the Ramanujan sum, then (47)
gives the first identity of the formulas
X X dµ(n/d)
cn (a − s) = ϕ(n) = µ(n)cn (s), (49)
ϕ(d)
a (mod n) d|n
(a,n)=1 (d,s)=1

the second one being the Brauer-Rademacher identity. See [49, Ch. 2]. Compare (49) to (2).

4.3 Geometric and harmonic means of (a − 1, n) with (a, n) = 1

Menon’s identity shows that τ (n) is the arithmetic mean of the values (a−1, n), where a runs
over a reduced residue system (mod n). The following identities are concerning the geometric
and harmonic means of these values, denoted by G(n) and H(n), respectively. One has, for any
n ∈ N,
 Y n 1/ϕ(n) Y
G(n) := (a − 1, n) = pep (n) , (50)
a=1 p∈P
(a,n)=1

where  
p 1
ep (n) = 1−
(p − 1)2 pνp (n)
and  X −1
n
1 Y p

1
−1
H(n) := ϕ(n) = 1− 2 1− , (51)
a=1
(a − 1, n)
p∈P
p −1 p2νp (n)
(a,n)=1
Q νp (n) .
using the notation n = p∈P p

26
 Identity (50) was obtained, in a more general form by Sita Ramaiah [67, Cor. 9.3]. For the
proof choose f (n) = log n in (7) and use that µ ∗ log = Λ, the von Mangoldt function. This gives
n
X X Λ(d)
log G(n) = log (a − 1, n) = ϕ(n) ,
a=1
ϕ(d)
d|n
(a,n)=1

which is not multiplicative, but can be computed, leading to formula (50).

To prove (51) choose f (n) = 1/n in (7) and note that (µ ∗ f )(pν ) = (1 − p)/pν for every
prime power pν (ν ≥ 1).

4.4 Identities using a polynomial P

Richards [63] stated without proof that for any polynomial P with integer coefficients one
has
n
X X
(P (a), n) = ϕ(n) NP (d), (52)
a=1 d|n
(a,n)=1

where NP (d) is the number of incongruent solutions x of the congruence P (x) ≡ 0 (mod d) such
that (x, d) = 1. P
A short proof is the following. By the Gauss formula d|n ϕ(d) = n the left hand side of
(52), say R, is
n
X X X Xn
R= ϕ(d) = ϕ(d) 1.
a=1 d|(P (a),n) d|n a=1
(a,n)=1 (a,n)=1
P (a)≡0 (mod d)

According to Lemma 2.1, to each such solution x one correspond ϕ(n)/ϕ(d) solutions y (mod
n) such that (y, n) = 1. Hence
X ϕ(n) X
R= ϕ(d) NP (d) = ϕ(n) NP (d).
ϕ(d)
d|n d|n

If f (x) = x − s with s ∈ Z, then (52) implies identity (48).

In fact, we have the following more general result of which proof is similar.

Theorem 4.2. If f is an arbitrary function and P (x) is any polynomial with integer coefficients,
then
Xn X (µ ∗ f )(d)
Mf,P (n) := f (P (a), n)) = ϕ(n) NP (d). (53)
ϕ(d)
a=1 d|n
(a,n)=1

If f is multiplicative, then Mf,P is also multiplicative. The case of the polynomial P (x) = xj
(j ∈ N) and general identities concerning systems of polynomials have been discussed by Tóth
[82].

27
4.5 Identities concerning parameters m, n, t
It was proved by Sivaramakrishnan [68], in a slightly different form, that
t
X  
t ϕ(m)τ (n) Y ν
M (m, n, t) := (a − 1, n) = 1− , (54)
a=1
m ν
(ν + 1)p
p ||n1
(a,m)=1

where m, n, t ∈ N such that m | t, n | t and n1 = max{d ∈ N : d | n, (d, m) = 1}. If m = n = t,

then M (n, n, n) = M (n), that is, (54) reduces to (1). Sivaramakrishnan [68] pointed out that
if t = [m, n], the least common multiple of m and n, then M (m, n, [m, n]) is a multiplicative
function of two variables. However, if n | m and t = m, then it follows from (54) that
m
X
(a − 1, n) = ϕ(m)τ (n). (55)
a=1
(a,m)=1

Identity (55) was recently obtained by Jafari and Madadi [38, Cor. 2.2], as a corollary of the
following group theoretic result, without referring to the paper [68], by using the orbit counting
lemma. Let G be a finite group of order m and let n be a divisor of m. Let AG (n) denote the
number of cyclic subgroups of G of which orders divide n and let BG (n) denote the number of
solutions in G of the equation xn = 1. Then, see [38, Th. 2.1],
m
X
1
AG (n) = BG ((a − 1, n)).
ϕ(m)
a=1
(a,m)=1

A short direct proof of (55) can be obtained, e.g., by Method IV:

m
X m
X X X m
X
(a − 1, n) = ϕ(d) = ϕ(d) 1.
a=1 a=1 d|(a−1,n) d|n a=1
(a,m)=1 (a,m)=1 (a,m)=1
a≡1(mod d)

Here d | n and n | m imply that d | m, and the inner sum is ϕ(m)/ϕ(d) by Lemma 2.1, with
m instead of n.

4.6 Sums over k-reduced residue systems (mod nk )

Let k ∈ N and let (a, b)k denote the greatest common k-th power divisor of a, b ∈ N. A
k-reduced residue system (mod nk ) is defined as the subset of elements a of a complete residue
system (mod nk ) such that (a, nk )k = 1. If k = 1, then this is the usual notion of a reduced
residue system (mod n).
The number of elements in a k-reduced residue system (mod nk ) is denoted by φk (n), function
introduced by Cohen [14]. Here φ1 (n) = ϕ(n). Also see Cohen [16].
Note that φk (n) = Jk (n), the Jordan function, for every n, k ∈ N. A direct proof is the
following: Consider the set A = {a ∈ N : 1 ≤ a ≤ nk }. Group the elements a of A according to

28
the values (a, nk )k = dk , where d | n. Here a = dk b, with 1 ≤ b ≤ (n/d)k and (b, (n/d)k )k = 1.
By thePdefinition of the function φk , the number of such elements a isPexactly φk (n/d). Therefore,
nk = d|n φk (n/d). By Möbius inversion we deduce that φk (n) = d|n dk µ(n/d) = Jk (n).
Nageswara Rao [56, Th. 2] proved by using n-even functions that
X
(a − s, nk )k = φk (n)τ (n),
a (mod nk )
(a,nk )k =1

where s ∈ Z, (s, nk )k = 1. For k = 1 one has Menon’s identity.

4.7 Identities with (a, n) ∈ S, where S is a subset of N

Let S be an arbitrary nonempty subset of N. Define the function µS by µS ∗ 1 = ̺S , where
̺S is the characteristic function of S. That is,
(
X 1, if n ∈ S,
µS (n) =
d|n
0, otherwise.

Furthermore, let ϕS (n) denote the number of integers a (mod n) such that (a, n) ∈ S. It
can be shown, see Cohen [19, Cor. 4.1], that ϕS = µS ∗ id, that is,
X
ϕS (n) = dµS (n/d) (n ∈ N).
d|n

If S = {1}, then µ{1} = µ and ϕ{1} = ϕ are the classical Möbius function and Euler function,
respectively. If S is the set of squares, then µS = λ is the Liouville function and φS = β, given
by X
β(n) = d λ(n/d),
d|n

which is identical to the alternating sum-of-divisors function. See the author [85]. Several other
special choices of S can be investigated. To mention a further one, if S is the set of k-th power
free integers, then ϕS is the Klee function. See Klee [40].
If S is an arbitrary subset, n ∈ N and s ∈ Z then
X
(a − 1, n) = (id ∗(µS × ϕ))(n), (56)
a (mod n)
(a,n)∈S

and X
c(a − s, n) = µS (n)cn (s), (57)
a (mod n)
(a,n)∈S

which reduce to Menon’s identity (1), respectively (2) if S = {1}.

29
 A subset S of N is called multiplicative if 1 ∈ S and the characteristic function ̺S is
multiplicative. In this case
X Y
(a − 1, n) = (P (pν ) + ϕS (pν ) − pν ) , (58)
a (mod n) pν ||n
(a,n)∈S

where P is the gcd-sum function.

In particular, if S is the set of squares, then we have
X Y

(a − 1, n) = P (pν ) − β(pν−1 ) (59)
a (mod n) pν ||n
(a,n) is a square

and X
c(a − s, n) = λ(n)cn (s). (60)
a (mod n)
(a,n) is a square

Identities (57) and (58) have been deduced by Haukkanen [26, Th. 3.1, 3.2] by finite Fourier-
expansions. They can be proved by different arguments as well, e.g., by Method V, which also
leads to the general identity (56), to be compared to (9). In the case when S is the set of squares,
identity (59) was obtained by Sivaramakrishnan [71, Th. 6.2], and identity (60) by Haukkanen
and Sivaramakrishnan [33, Sect. 4].
A Menon-type identity related to Klee’s arithmetic function was discussed by Chandran,
Thomas, and Namboothiri [7].

4.8 Unitary analogs

Let (a, b)∗ denote the greatest divisor of a which is a unitary divisor of b. Nageswara Rao
[55] proved via the ,,unitary” version of Method X that
X
(a − 1, n)∗ = ϕ∗ (n)τ ∗ (n), (61)
a (mod n)
(a,n)∗ =1
Q P
where ϕ∗ (n) = pν ||n (pν − 1) is the unitary Euler function and τ ∗ (n) = d||n 1 is the unitary
P
divisor function. Note that τ ∗ (n) = 2ω(n) , where ω(n) = p|n 1.
Short direct proofs can be given by Methods I-V as well, using properties of unitary functions.
Some related identities are
X Y

(a − 1, n) = (ν + 1)pν − νpν−1 − 1 , (62)
a (mod n) pν ||n
(a,n)∗ =1
X Y

(a − 1, n)∗ = 2pν − pν−1 − 1 , (63)
a (mod n) pν ||n
(a,n)=1

where (62) is a special case of identity (64), and we are not aware of any reference for (63).

30
4.9 Regular systems of divisors
Let A(n) be a subset of the set of positive divisors of n for every n ∈ N. The A-convolution
of the functions f, g : N → C is defined by
X
(f ∗A g)(n) = f (d)g(n/d) (n ∈ N).
d∈A(n)

Let A denote the set of arithmetic functions f : N → C. The system A = (A(n))n∈N of

divisors is called regular, cf. Narkiewicz [57], if the following conditions hold true:
(a) (A, +, ∗A ) is a commutative ring with unity,
(b) the A-convolution of multiplicative functions is multiplicative,
(c) the constant 1 function has an inverse µA (generalized Möbius function) with respect to
∗A and µA (pν ) ∈ {−1, 0} for every prime power pν (ν ≥ 1).
It can be shown that the system A = (A(n))n∈N is regular if and only if
(i) A(mn) = {de : d ∈ A(m), e ∈ A(n)} for every m, n ∈ N with (m, n) = 1,
(ii) for every prime power pν (ν ≥ 1) there exists a divisor t = tA (pν ) of ν, called the type
of pν with respect to A, such that

A(pit ) = {1, pt , p2t , . . . , pit }

for every i ∈ {0, 1, . . . , ν/t}.

Examples of regular systems of divisors are A = D, where D(n) is the set of all positive
divisors of n and A = U , where U (n) is the set of unitary divisors of n. For every prime power
pν (ν ≥ 1) one has tD (pν ) = 1 and tU (pν ) = ν. Here ∗D and ∗U are the Dirichlet convolution
and the unitary convolution, respectively.
For properties of regular convolutions and related arithmetical functions we refer to Narkiewicz
[57], McCarthy [48,
P 49] and Sita Ramaiah [67].
Let τA (n) = d∈A(n) 1 denote the corresponding divisor function and let
X
ϕA (n) = 1,
a (mod n)
(a,n)A =1

be the generalized Euler function, where

(a, n)A = max{d ∈ N : d | a, d ∈ A(n)}.

If A is a regular system of divisors, then one has

X
(a − 1, n)A = ϕA (n)τA (n),
a (mod n)
(a,n)A =1

proved by Sita Ramaiah [67], in a more general form, which recovers identities (1) and (61)
in the cases A = D and A = U , respectively. Further generalizations with respect to n-even
functions of several variables were given by Haukkanen and McCarthy [32].

31
4.10 A generalized divisibility relation
Haukkanen [27] defined a generalized divisibility relation on N as follows. For every prime
p let fp : N → N ∪ {0} be a function and let f = (fp )p∈P be the system of these functions. We
say that ≀ is a divisibility relation of type f if
1) 1 ≀ n for all n ∈ N,
2) for every prime p and ν ∈ N ∪ {0}, if d ≀ pν , then d = pi Q for some i ∈ {0, 1, . . Q
. , ν},
3) d ≀ n if and only if pνp (d) ≀ pνp (n) for all primes p, where d = p∈P pνp (d) and n = p∈P pνp (n)
are the canonical factorizations of d and n,
4) for every prime p and ν ∈ N, fp (ν) is the smallest integer i ∈ {1, 2, . . . , ν} such that pi ≀ pν
if such i exists, and fp (ν) = 0 otherwise.
It is clear that if d ≀ n, then d | n. The usual divisibility relation | is a divisibility relation of
type (1, 1, . . .), where 1(ν) = 1 for all ν ∈ N and all p ∈ P. The unitary divisibility relation ||
is a divisibility relation of type (id, id, . . .), where id(ν) = ν for all ν ∈ N and all p ∈ P. More
generally, if A is a regular system of divisors (see Section 4.9), then the A-divisibility relation |A ,
where d |A n if and only if d ∈ A(n) is a divisibility relation of type f = (fp ) with fp (ν) = tA (pν ),
the type of pν . Some more examples were also given in [27].
Let ≀ be a divisibility relation of type f . We denote (m, n)≀ = max{d : d | m, d ≀ n}.
The generalized Euler function φ≀ (n) is defined as the number of integers a (mod n) such that
(a, n)≀ = 1.
Haukkanen [30, Cor. 4.1] proved that for every n ∈ N,
X X ϕ(d)nd
(a − 1, n) = φ≀ (n) , (64)
dφ≀ (nd )
a (mod n) d|n
(a,n)≀ =1
Q
where ϕ is the classical Euler’s function and nd = pν ||n, p|d pν .
Schwab and Bede [66] considered the ”odd-divisibility” relation |⊗ defined by d |⊗ n if and
only if d is odd and d | n. They refer to the paper by Haukkanen [30], remark that this is a
divisibility relation of type (θ, 1, 1, . . .), where θ(ν) = 0 (ν ∈ N), and as an application of (64)
deduce that for every n ∈ N,
X  
1
(a − 1, n) = ϕ⊗ (n) τ (n) − τ2 (n) ,
2
a (mod n)
(a,n)⊗ =1

where (a, n)⊗ denotes the greatest common odd divisor of a and n, ϕ⊗ (n) is the number of
integers a (mod n) such that (a, n)⊗ = 1, τ (n) is the number of divisors of n, and τ2 (n) is the
number of even divisors of n.

4.11 Regular integers (mod n)

An integer a is called regular (mod n), if there exists an integer x such that a2 x ≡ a (mod
n), i.e., the residue class of a is a regular element (in the sense of J. von Neumann) of the ring
Zn of residue classes (mod n).

32
 Let n > 1 be an integer with prime factorization n = pν11 · · · pνr r . It can be shown that the
following three statements are equivalent: (i) a ∈ N is regular (mod n), (ii) (a, n) || n, (iii) for
every i ∈ {1, . . . , r} either pi ∤ a or pνi i | a.
Let Regn = {a : 1 ≤ a ≤ n, a is regular (mod n)} and let ̺(n) = #P Regn denote the number
of regular integers a (mod n) such that 1 ≤ a ≤ n. One has ̺(n) = d||n ϕ(d). See Tóth [79]
for more properties.
Apostol and Tóth [1, Prop. 5.3.1] proved that
X X Y

(a − 1, n) = ϕ(d)τ (d) = pν−1 (p − 1)(ν + 1) + 1 .
a∈Regn d||n pν ||n

4.12 Multiple sums

Nageswara Rao [56, Th. 1] proved by the theory of n-even functions that
n
X
(a1 − s1 , . . . , ak − sk , n)k = Jk (n)τ (n) (k, n ∈ N), (65)
a1 ,...,ak =1
(a1 ,...,ak ,n)=1
Q
where s1 , . . . , sk ∈ Z, (s1 , . . . , sk , n) = 1 and Jk (n) = nk p|n (1 − 1/pk ) is the Jordan function
of order k.
Haukkanen and Sivaramakrishnan [34, Th. 9] deduced, also by using n-even functions that
n
X X 1
(a1 − 1, . . . , ak − 1, n) = Jk (n) (k, n ∈ N), (66)
a1 ,...,ak =1
εk (d)
d|n
(a1 ,...,ak ,n)=1

where εk (d) = Jk (d)/ϕ(d) is the number of cyclic subgroups of order d of the group Zd × · · · × Zd
with k factors.
Sury [75] used the orbit counting lemma to prove the following identity:
n
X
(a1 − 1, a2 , . . . , ak , n) = ϕ(n)σk−1 (n) (k, n ∈ N), (67)
a1 ,a2 ,...,ak =1
gcd(a1 ,n)=1
P
where σs (n) = d|n ds .
For k = 1 all of (65), (66) and (67) recover Menon’s identity.
Direct proofs of these and other similar identities (see, e.g., [95]) can be given by Method
IV, for example. To deduce (65) write
n
X
T := (a1 − s1 , . . . , ak − sk , n)k
a1 ,...,ak =1
(a1 ,...,ak ,n)=1

n
X X
= Jk (d) =
a1 ,...,ak =1 d|(a1 −s1 ,...,ak −sk ,n)
(a1 ,...,ak ,n)=1

33
 X X X
= Jk (d) µ(δ)
d|n aj ≡sj (mod d) δ|(a1 ,...,ak ,n)=1
1≤j≤k
X X X
= Jk (d) µ(δ) 1,
d|n δ|n 1≤bj ≤n/δ
δbj ≡sj (mod d)
1≤j≤k
where the linear congruences δbj ≡ sj (mod d) (1 ≤ j ≤ k) have solutions bj if and only if
(d, δ) | sj (1 ≤ j ≤ k), equivalent to (d, δ) = 1, since (d, δ) | n and (s1 , . . . , sk , n) = 1 by the
given condition. If (d, δ) = 1 is satisfied, then each congruence has one solution (mod d) and
n/(dδ) solutions (mod n/δ). Hence
X X  n k X Jk (d) X µ(δ)
T = Jk (d) µ(δ) = nk
dδ dk δk
d|n δ|n d|n δ|n
(d,δ)=1 (d,δ)=1
X Jk (d) Jk (n)/nk
= nk = Jk (n)τ (n),
dk Jk (d)/dk
d|n
completing the proof of (65).
Tóth [84, Prop. 2.7] proved by arithmetic arguments that for every n, k ∈ N,
n
X
(a1 · · · ak − 1, n) = ϕ(n)k τ (n),
a1 ,...,ak =1
(a1 ···ak ,n)=1

which reduces to Menon’s original identity for k = 1.

Another multidimensional generalization of Euler’s arithmetic function and of Menon’s iden-
tity has been obtained by Tóth [90]. For k ∈ N let define the function ϕk (n) as
n
X
ϕk (n) := 1,
a1 ,...,ak =1
(a1 ···ak ,n)=1
(a1 +···+ak ,n)=1

where ϕ1 (n) = ϕ(n) is Euler’s function.

Let f be an arbitrary arithmetic function. Then for every k, n ∈ N,
Xn X (µ ∗ f )(d)
f ((a1 + · · · + ak − 1, n)) = ϕk (n) . (68)
ϕ(d)
a1 ,...,ak =1 d|n
(a1 ···ak ,n)=1
(a1 +···+ak ,n)=1

If f = id, then
n
X
(a1 + · · · + ak − 1, n) = ϕk (n)τ (n). (69)
a1 ,...,ak =1
(a1 ···ak ,n)=1
(a1 +···+ak ,n)=1
In the case k = 2 (69) has been proved by Sita Ramaiah [67, Cor. 10.4] in the general setting
of regular convolutions and k-reduced residue systems.

34
4.13 Identities for arithmetic functions of several variables
Let f be an arithmetic function of k variables. Haukkanen and Tóth [35] considered (in a
more general form) sums of type
X X
··· f ((a1 − 1, n1 ), . . . , (ak − 1, nk )).
1≤a1 ≤n1 1≤ak ≤nk
(a1 ,n1 )=1 (ak ,nk )=1

In fact, they deduced common generalizations of the Menon-type identities by Sivaramakr-

ishnan [68] and Li, Kim, Qiao [47]. In particular, an explicit formula was proved in the case
f (n1 , . . . , nk ) = [n1 , . . . , nk ], the lcm function. If n1 = · · · = nk = n, then it gives (see [35, Cor.
4.2]): For any n, k ∈ N,
X X
··· [(a1 − 1, n), . . . , (ak − 1, n)]
1≤a1 ≤n 1≤ak ≤n
(a1 ,n)=1 (ak ,n)=1
 
Y k−1
X    
1 + kν + k pj 1 
= ϕ(n)k (−1)j 1− .
j + 1 (p − 1)j (pj − 1) pνj
pν ||n j=1

If k = 1 we recover Menon’s identity (1).

See the paper by Haukkanen and Wang [36] for a general identity with respect to a set of
polynomials.

4.14 Identities concerning Dirichlet characters

Zhao and Cao [105] proved that
n
X
(a − 1, n)χ(a) = ϕ(n)τ (n/d), (70)
a=1

where χ is a Dirichlet character (mod n) with conductor d (n ∈ N, d | n). If χ is the principal

character (mod n), that is d = 1, then (70) reduces to Menon’s identity (1). If χ is a primitive
character (mod n), i.e., d = n, then (70) gives
n
X
(a − 1, n)χ(a) = ϕ(n). (71)
a=1

Note that the sums in (70) and (71) are over (a, n) = 1, since χ(a) = 0 for (a, n) > 1. Zhao
and Cao [105] first proved formula (71) and then deduced identity (70) by using the fact that
every Dirichlet character is induced by a primitive character. They showed that the left hand
sides of (70) and (71) are multiplicative in n and computed their values for prime powers.
Generalizations of (70) involving n-even functions have been deduced by Tóth [87], using
a different approach based on direct manipulations of the corresponding sums, valid for any
integer n ∈ N. We quote here only the following result, having a simple proof. See [87, Th. 2.7].

35
Let χ be a primitive Dirichlet character (mod n), where n ∈ N. Let f be an even function (mod
n) and let s ∈ Z. Then
Xn
f (k − s)χ(k) = (µ ∗ f )(n)χ(s).
k=1

For the proof, use that for every even function f ,

X
f (k) = f ((k, n)) = (µ ∗ f )(d),
d|(k,n)

and that for a primitive character χ (mod n), and for any d | n, d < n and s ∈ Z,
n
X
χ(k) = 0.
k=1
k≡s (mod d)

We can deduce
n
X n
X X
f (k − s)χ(k) = χ(k) (µ ∗ f )(e)
k=1 k=1 e|(k−s,n)

X n
X
= (µ ∗ f )(e) χ(k) = (µ ∗ f )(n)χ(s).
e|n k=1
k≡s(mod e)

Further generalizations have been given by Chandran, Namboothiri, Thomas [6], Chen [10],
Li, Hu, Kim [45, 46], Tóth [88].

4.15 Identities concerning additive characters

Q
We use the notation e(x) = e2πix and n = p∈P p
νp (n) . For every n ∈ N, k ∈ Z,
n
X X
S(n, k) := (a − 1, n) e(ak/n) = e(k/n) dϕ(n/d),
a=1 d|(n,k)

being an identity related to the discrete Fourier transform (DFT) of the gcd function. See
Lemma 2.6.
Li and Kim [44] proved that for every n ∈ N, k ∈ Z such that νp (n) − νp (k) 6= 1 for every
prime p | n one has
n
X
S ∗ (n, k) := (a − 1, n) e(ak/n) = e(k/n)ϕ(n)τ ((n, k)),
a=1
(a,n)=1

which reduces to Menon’s identity (1) in the case n | k. Furthermore, Li and Kim [44, Th. 3.5]
established a formula for S ∗ (n, k), valid for every n ∈ N and k ∈ Z. This was generalized and
proved using a different method by Tóth [89].

36
4.16 Identities with respect to sets of units
Let Z∗n denote the group of units of the ring Zn . Let S be an arbitrary nonempty subset of
c∗ denote the character group of Z∗ . Caiúve and Miguel [5] proved, that
the group Z∗n and let Z n n
for every n ∈ N, X X X
MS (n) := (a − 1, n) = τ (n/dχ ) χ(a), (72)
a∈S c∗n
χ∈Z a∈S

where dχ is the conductor of the character χ, and χ stands for the conjugate of χ. If S = Z∗n ,
then (72) reduces to Menon’s identity (1) by using (20).
Here is the proof: According to relation (18),
(
1 X 1, if x ≡ a (mod n),
δx (a) := χ(x)χ(a) =
ϕ(n) 0, otherwise,
c∗
χ∈Zn

hence (
X 1, if a ∈ S,
δS (a) := δx (a) =
x∈S
0, otherwise,
and deduce that
X X X
MS (n) = (a − 1, n)δS (a) = (a − 1, n) δx (a)
a∈Z∗n a∈Z∗n x∈S

1 X X X
= (a − 1, n) χ(x)χ(a)
ϕ(n) ∗ a∈Zn x∈S χ∈Z
c∗n

1 X X X
= χ(x) (a − 1, n)χ(a).
ϕ(n) ∗
x∈S χ∈Z
c∗ a∈Zn
n

Now application of the result (70) by Zhao and Cao [105] finishes the proof of (72).
We note that Caiúve and Miguel [5] proved (72) in a more general form, namely for multiple
sums and several subsets of Z∗n , and investigated the case when these subsets are subgroups of
Z∗n .

4.17 Identities derived from group actions by the orbit counting lemma
Another kind of generalizations of Menon’s identity, based on applying the orbit counting
lemma to certain group actions have been obtained by Tărnăuceanu [76], Li and Kim [43].
Namely, they considered certain subgroups of GLr (Zn ) acting on Zrn . For example, taking the
subgroup of the upper triangular matrices in GLr (Zn ), Li and Kim [43, Th. 2.4] proved the
following result.
If n, r ∈ N, then
XY r
(dk /dk−1 , n) = nr(r−1)/2 ϕ(n)r τr+1 (n), (73)
aij k=1

37
where the sum is over all aij (1 ≤ i, j ≤ r) such that 1 ≤ aij ≤ n (1 ≤ i, j ≤ r) and (aii , n) = 1
(1 ≤ i ≤ r), dk is the gcd of all k × k minors of the matrix
 
a11 − 1 a12 ... a1r
 0 a22 − 1 . . . a2r 
 
 .. .. . .. .. 
 . . . 
0 0 . . . arr − 1
P
for k ≥ 1, d0 = 1, by convention 0/0 = 0 for dk /dk−1 , and τr+1 (n) = d1 ···dr dr+1 =n 1 is the Piltz
divisor function. Note that this corrects the corresponding result obtained by Tărnăuceanu [76].
If r = 1, then (73) reduces to Menon’s identity (1).
Some other Menon-type identities have been obtained by Wang and Ji [101] and Tărnăuceanu
[77], by using the action of subgroups of U (Zn ) on the set Zn , and the action of the power
automorphism group of G on G, where G is a finite abelian group, respectively.

4.18 Identities concerning subsets of the set {1, . . . , n}

For a nonempty subset A of {1, 2, . . . , n} let (A) denote the gcd of the elements of A. The
subset A is said to be relatively prime if (A) = 1, i.e., the elements of A are relatively prime.
Let f (n) denote the number of relatively prime subsets of {1, 2, . . . , n}. For every n ∈ N one
has
X n  
f (n) = µ(d) 2⌊n/d⌋ − 1 , (74)
d=1

where ⌊x⌋ is the floor function of x.

A similar formula is valid for the number fk (n) of relatively prime k-subsets (subsets with k
elements) of {1, 2, . . . , n}. Namely, for every n, k ∈ N (k ≤ n),
n
X  
⌊n/d⌋
fk (n) = µ(d) . (75)
k
d=1

Furthermore, consider the Euler-type functions Φ(n) and Φk (n), representing the number of
nonempty subsets A of {1, 2, . . . , n} and k-subsets A of {1, 2, . . . , n}, respectively, such that (A)
and n are relatively prime. Observe that Φ1 (n) = ϕ(n) is Euler’s function. One has Φ(1) = 1,
X
Φ(n) = µ(d)2n/d (n ∈ N, n > 1),
d|n

and for every n, k ∈ N (k ≤ n),

X  
n/d
Φk (n) = µ(d) .
k
d|n

The functions f , fk , Φ and Φk have been defined and investigated by Nathanson [58]. Also
see Tóth [81] and its references.

38
 Define the sum M (n) by
X
M (n) := ((A) − 1, n),
∅6=A⊆{1,2,...,n}
((A),n)=1

taken over all nonempty subsets of {1, 2 . . . , n} such that (A) and n are relatively prime, where
((A) − 1, n) denotes the gcd of (A) − 1 and n. Also, for 1 ≤ k ≤ n let
X
M k (n) := ((A) − 1, n),
A⊆{1,2,...,n}
#A=k
((A),n)=1

the sum being over the k-subsets of {1, 2, . . . , n} such that (A) and n are relatively prime.
Observe that the sums M (n) and M k (n) have Φ(n), respectively Φk (n) terms.
Tóth [91] proved the following identities. For every n, k ∈ N,

X X n/δ
X  
n
M (n) = ϕ(d) µ(δ) f ,
jδ
d|n δ|n j=1
(δ,d)=1 δj≡1 (mod d)

X X n/δ
X  
n
M k (n) = ϕ(d) µ(δ) fk , (76)
jδ
d|n δ|n j=1
(δ,d)=1 δj≡1 (mod d)

where the functions f and fk are given by (74) and (75), respectively.
Note that if k = 1, then f1 (n) = 1 (n ∈ N), the last sum in (76) is n/(dδ), and this quickly
leads to Menon’s identity (1). See the proof of Lemma 2.1.

4.19 Identities in Dedekind domains

All generalizations and analogs discussed above are in the setting of rational integers. Miguel
[51, 52] considered residually finite Dedekind domains, that is, Dedekind domains D such that
for every non-zero ideal n of D, the residue class ring D/n is finite. Let N (n) = |D/n| denote
the norm of the ideal n. Let ϕD denote the generalized Euler function defined as
(
1, if n = D,
ϕD (n) =
|U (D/n)|, otherwise.
P
Furthermore, let τD (n) = d|n 1 be the corresponding divisor function. Note that ϕD and
τD are finite valued functions. Miguel [51, Th. 1.1] proved by applying Burnside’s lemma that
for every non-zero ideal n in D,
X
N (ha − 1i + n) = ϕD (n)τD (n), (77)
a∈U(D/n)

39
which reduces to Menon’s identity (1) in the case D = Z.
Miguel [52, Th. 1.1] also deduced, again by Burnside’s lemma, the generalization of Sury’s
identity (67). A further generalization was given by Li and Kim [42].
Identity (77) has been generalized with respect to regular systems of divisors and k-reduced
residue systems by Wang, Zhang and Ji [99]. The Sita Ramaiah identity, which is the case k = 2
of (69) has been generalized in the setting of residually finite Dedekind domains with respect to
regular systems of divisors and k-reduced residue systems by Ji and Wang [39]. Also see Wang,
Hu, and Ji [100] for some different results deduced by the orbit counting lemma, and Chen and
Zheng [13] for some further generalizations.

4.20 Identities in the ring of algebraic integers

Let K be an algebraic number field with the ring of integers denoted by OK . Then OK is
a residually finite Dedekind domain. Let n be a non-zero ideal of OK and let χ be a character
(mod n) with conductor d. Wang and Ji [102, Cor. 1.3] proved, as a corollary of more general
results, that X
N (ha − 1i + n)χ(a) = ϕ(n)τ (n/d), (78)
a∈U (OK /n)

where N , ϕ and τ are the corresponding norm, Euler and divisor functions, respectively. If χ is
the principal character, then (78) reduces to identity (77) in the case of algebraic integers, and
if K = Q, then (78) recovers (70).
Identities (68) and (69) by Tóth have been generalized by Sarkar [65] to the ring of algebraic
integers OK and involving Dirichlet characters. The case k = 2, has been earlier investigated
by Wang and Ji [103], Chattopadhyay and Sarkar [8].

4.21 Final remarks

Extensions of Menon-type identities, in particular of the general identity by Tóth [88], to
the polynomial ring over the a finite field Fq have been obtained Chen and Qi [12].
Some further papers related to Menon-type identities, not mentioned above in this survey
are [9, 11, 37, 41, 53, 59, 94, 97, 98].

References
[1] B. Apostol and L. Tóth, Some remarks on regular integers modulo n, Filomat, 29 (2015),
687–701.

[2] T. M. Apostol, Introduction to Analytic Number Theory, Springer, 1976.

[3] K. Bibak, B. M. Kapron, V. Srinivasan, R. Tauraso, and L. Tóth, Restricted linear congru-
ences, J. Number Theory, 171 (2017), 128–144.

[4] K. A. Bogart, An obvious proof of Burnside’s lemma, Amer. Math. Monthly, 98 (1991),
927–928.

40
 [5] A. Caiúve and C. Miguel, Menon-type identities with respect to sets of units, Ramanujan
J., 55 (2021), 817–822.

[6] A. Chandran, K. V. Namboothiri, and N. E. Thomas, A Menon-type identity concerning

Dirichlet characters and a generalization of the gcd function, Proc. Indian Acad. Sci. (Math.
Sci.), (2021) 131:18, 13 pp.

[7] A. Chandran, N. E. Thomas, and K. V. Namboothiri, A Menon-type identity using Klee’s

function, Czech. Math. J., 72 (147) (2022), 165–176.

[8] J. Chattopadhyay and S. Sarkar, On a generalization of Menon-Sury identity to number

fields involving a Dirichlet character, Ramanujan J., 59 (2022), 979–992.

[9] M. Chen, S. Hu, and Y. Li, On Menon-Sury’s identity with several Dirichlet characters,
Preprint, 2018, arXiv:1807.07241 [math.NT].

[10] M. Chen, Menon-Sury’s identity with a Dirichlet character (Chinese), Pure. Math., 9 (2019),
188–194.

[11] M. Chen, Menon-Sury’s identity with several Dirichlet characters and additive characters
(Chinese) Acta Math. Sinica (Chin. Ser.), 62 (2019), 949–958.

[12] M. Chen and T. Qi, A Menon-Sury-type identity for arithmetic functions on Fq [T ], Publ.
Math. Debrecen, 98 (2021), 1–14.

[13] M. Chen and Z. Zheng, A Menon-type identity with Dirichlet characters in residually finite
Dedekind domains, Ramanujan J., 57 (2022), 1473–1488.

[14] E. Cohen, An extension of Ramanujan’s sum, Duke Math. J., 16 (1949), 85–90.

[15] E. Cohen, A class of arithmetical functions, Proc. Nat. Acad. Sci. U.S.A., 41 (1955), 939–
944.

[16] E. Cohen, Generalizations of the Euler φ-function, Scripta Math., 23 (1958), 157–161.

[17] E. Cohen, Representations of even functions (mod r), I. Arithmetical identities, Duke Math.
J., 25 (1958), 401–421.

[18] E. Cohen, Representations of even functions (mod r), II. Cauchy products, Duke Math. J.,
26 (1959), 165–182.

[19] E. Cohen, Arithmetical functions associated with arbitrary sets of integers, Acta Arith., 5
(1959), 407–415.

[20] E. Cohen, Arithmetical functions associated with the unitary divisors of an integer, Math.
Z., 74 (1960), 66–80.

[21] R. Freud and E. Gyarmati, Number Theory, American Math. Soc., 2020.

41
[22] F. Fung, A number-theoretic identity arising from Burnside’s orbit formula, Pi Mu Epsilon
J., 9 (1994), 647–650.

[23] P. G. Garcia and S. Ligh, A generalization of Euler’s φ-function, Fibonacci Quart., 21

(1983), 26-28.

[24] O. Grošek and Š. Porubský, Coprime solutions to ax ≡ b (mod n), J. Math. Cryptol., 7
(2013), 217–224.

[25] G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers, Sixth Edition,
Edited and revised by D. R. Heath-Brown and J. H. Silverman, Oxford University Press,
2008.

[26] P. Haukkanen, On some set-reduced arithmetical sums, Indian J. Math., 39 (1997), 147–
158.

[27] P. Haukkanen, On a formula for Euler’s totient function, Nieuw Arch. Wisk. (4), 17 (1999),
343–348.

[28] P. Haukkanen, Regular class division of integers (mod r), Notes Number Theory Discrete
Math., 6 (2000), 82–87.

[29] P. Haukkanen, An elementary linear algebraic approach to even functions (mod r), Nieuw
Arch. Wiskd. (5), 2 (2001), 29–31.

[30] P. Haukkanen, Menon’s identity with respect to a generalized divisibility relation, Aequa-
tiones Math., 70 (2005), 240–246.

[31] P. Haukkanen, On a gcd-sum function, Aequationes Math., 76 (2008), 168–178.

[32] P. Haukkanen and P. J. McCarthy, Sums of values of even functions, Portugal. Math., 48
(1991), 53–66.

[33] P. Haukkanen and R. Sivaramakrishnan, Cauchy multiplication and periodic functions (mod
r), Collect. Math., 42 (1991), 33–44.

[34] P. Haukkanen and R. Sivaramakrishnan, On certain trigonometric sums in several variables,

Collect. Math., 45 (1994), 245–261.

[35] P. Haukkanen and L. Tóth, Menon-type identities again: A note on a paper by Li, Kim
and Qiao, Publ. Math. Debrecen, 96 (2020), 487–502.

[36] P. Haukkanen and J. Wang, A generalization of Menon’s identity with respect to a set of
polynomials, Portugal. Math., 53 (1996), 331–337.

[37] P. Haukkanen and J. Wang, High degree analogs of Menon’s identity, Indian J. Math., 39
(1997) 37–42.

[38] M. H. Jafari and A. R. Madadi, On the number of cyclic subgroups of a finite group, Bull.
Korean Math. Soc., 54 (2017), 2141–2147.

42
[39] Ch. Ji and Y. Wang, Another regular Menon-type identity in residually finite Dedekind
domains, Acta Math. Hungar., 162 (2020), 117–129.

[40] V. L. Klee Jr., A generalization of Euler’s φ-function, Amer. Math. Monthly, 55 (1948),
358–359.

[41] Y. Li, R. Chen, and D. Kim, A generalization of Menon-Rao-Sury’s identities to additive

characters by Tóth’s method, Miskolc Math. Notes, 22 (2021), 763–768.

[42] Y. Li and D. Kim, A Menon-type identity with many tuples of group of units in residually
finite Dedekind domains, J. Number Theory, 175 (2017), 42–50.

[43] Y. Li and D. Kim, Menon-type identities derived from actions of subgroups of general linear
groups, J. Number Theory, 179 (2017) 97–112.

[44] Y. Li and D. Kim, Menon-type identities with additive characters, J. Number Theory, 192
(2018), 373–385.

[45] Y. Li, X. Hu, and D. Kim, A generalization of Menon’s identity with Dirichlet characters.
Int. J. Number Theory, 14 (2018), 2631–2639.

[46] Y. Li, X. Hu, and D. Kim, A Menon-type identity with multiplicative and additive charac-
ters, Taiwanese J. Math., 23 (2019), 545–555.

[47] Y. Li, D. Kim, and R. Qiao, A generalization of Menon’s identity to higher exponent, Publ.
Math. Debrecen, 94 (2019), 467–475.

[48] P. J. McCarthy, Regular arithmetical convolutions, Portugal. Math., 27 (1968), 1–13.

[49] P. J. McCarthy, Introduction to Arithmetical Functions, Springer, 1986.

P
[50] P. K. Menon, On the sum (a − 1, n)[(a, n) = 1], J. Indian Math. Soc. (N.S.), 29 (1965),
155–163.

[51] C. Miguel, Menon’s identity in residually finite Dedekind domains, J. Number Theory, 137
(2014) 179–185.

[52] C. Miguel, A Menon-type identity in residually finite Dedekind domains, J. Number Theory,
164 (2016) 43–51.

[53] E. A. Molla and S. Sarkar, A generalization of Tóth identity in Fq [T ] involving a Dirichlet

character, Preprint, 2022, arXiv:2206.03184v1 [math.NT].

[54] H. L. Montgomery, Early Fourier Analysis, Pure and Applied Undergraduate Texts 22,
Providence, RI: American Math. Society, 2014.

[55] K. Nageswara Rao, Unitary class division of integers mod n and related arithmetical iden-
tities, J. Indian Math. Soc., 30 (1966) 195–205.

43
[56] K. Nageswara Rao, On certain arithmetical sums, Lecture Notes Math., 251, Springer, 1972,
181–192.

[57] W. Narkiewicz, On a class of arithmetical convolutions, Colloq. Math., 10 (1963), 81–94.

[58] M. B. Nathanson, Affine invariants, relatively prime sets, and a phi function for subsets of
{1, 2, . . . , n}, Integers, 7 (2007), Paper A1, 7 pp.

[59] M. B. Nathanson, Multiplicative proofs of some generalizations of Menon’s identity,

Preprint, 2021, arXiv:2111.05086v1 [math.NT].

[60] P. Neumann, A lemma that is not Burnside’s, Math. Sci., 4 (1979), 133–141.

[61] I. Niven, H. S. Zuckerman, and H. L Montgomery, An Introduction to the Theory of Num-

bers, 5th ed., John Wiley & Sons, 1991.

[62] K. G. Ramanathan, Some applications of Ramanujan’s trigonometrical sum Cm (n), Proc.

Ind. Acad. Sci. (A), 20 (1944), 62–69.

[63] I. M. Richards, A remark on the number of cyclic subgroups of a finite group, Amer. Math.
Monthly, 91 (1984), 571–572.

[64] J. Rotman, A First Course in Abstract Algebra, Third Edition, Prentice Hall, 2005.

[65] S. Sarkar, A generalization of Tóth identity in the ring of algebraic integers involving a
Dirichlet character, Preprint, 2021, arXiv:2106.13983v1 [math.NT].

[66] E. D. Schwab and B. Bede, A note on a broken Dirichlet convolution, Notes Number Theory
and Discrete Math., 20 (2014), 65–73.

[67] V. Sita Ramaiah, Arithmetical sums in regular convolutions, J. Reine Angew. Math.,
303/304 (1978), 265–283.

[68] R. Sivaramakrishnan, Generalization of an arithmetic function, J. Indian Math. Soc. (N.S.),

33 (1969), 127–132.

[69] R. Sivaramakrishnan, A number-theoretic identity, Publ. Math. Debrecen, 21 (1974), 67–69.

[70] R. Sivaramakrishnan, Multiplicative even functions (mod r), I. Structural properties, J.

Reine Angew. Math., 302 (1978), 32–43.

[71] R. Sivaramakrishnan, Square reduced residue systems (mod r) and related arithmetical
functions, Canad. Math. Bull., 22 (1979), 207–220.

[72] R. Sivaramakrishnan, Classical Theory of Arithmetic Functions, Monographs and Text-

books in Pure and Applied Mathematics, Vol. 126, Marcel Dekker, 1989.

[73] R. Sivaramakrishnan, Certain Number-Theoretic Episodes in Algebra, Pure and Applied

Mathematics (Boca Raton), 286, Chapman & Hall/CRC, Boca Raton, FL, 2007.

44
[74] M. V. Subbarao, Remarks on a paper of P. Kesava Menon, J. Indian Math. Soc. (N.S.), 32
(1968), 317–318.

[75] B. Sury, Some number-theoretic identities from group actions, Rend. Circ. Mat. Palermo
(2), 58 (2009), 99–108.

[76] M. Tărnăuceanu, A generalization of Menon’s identity, J. Number Theory, 132 (2012)

2568–2573.

[77] M. Tărnăuceanu, The number of cyclic subgroups of finite abelian groups and Menon’s
identity, Bull. Aust. Math. Soc., 101 (2020), 201–206.

[78] M. Tărnăuceanu, A new Menon-type identity derived from group actions, Sci. Stud. Res.
Ser. Math. Inform., 31 (2021), no. 2, 73–76.

[79] L. Tóth, Regular integers (mod n), Annales Univ. Sci. Budapest., Sect. Comp., 29 (2008),
263–275.

[80] L. Tóth, A survey of gcd-sum functions, J. Integer Seq., 13 (2010), Article 10.8.1, 23 pp.

[81] L. Tóth, On the number of certain relatively prime subsets of {1, 2, . . . , n}, Integers, 10
(2010), Paper A35, pp. 407–421.

[82] L. Tóth, Menon’s identity and arithmetical sums representing functions of several variables,
Rend. Sem. Mat. Univ. Politec. Torino, 69 (2011) 97–110.

[83] L. Tóth, Weighted gcd-sum functions, J. Integer Seq., 14 (2011), Article 11.7.7, 10 pp.

[84] L. Tóth, Another generalization of the gcd-sum function, Arab. J. Math., 2 (2013), 313–320.

[85] L. Tóth, A survey of the alternating sum-of-divisors function, Acta Univ. Sapientiae Math.,
5 (2013), 93–107.

[86] L. Tóth, Multiplicative arithmetic functions of several variables: a survey, in Mathematics

Without Boundaries, Surveys in Pure Mathematics. Th. M. Rassias, P. Pardalos (Eds.),
Springer, New York, 2014, 483–514.

[87] L. Tóth, Menon-type identities concerning Dirichlet characters, Int. J. Number Theory, 14
(2018), 1047–1054.

[88] L. Tóth, Short proof and generalization of a Menon-type identity by Li, Hu and Kim,
Taiwanese J. Math., 23 (2019), 557–561.

[89] L. Tóth, Menon-type identities concerning additive characters, Arab. J. Math., 9 (2020),
697–705.

[90] L. Tóth, Another generalization of Euler’s arithmetic function and Menon’s identity, Ra-
manujan J., 57 (2022), 811–822.

45
[91] L. Tóth, Menon-type identities concerning subsets of the set {1, 2, . . . , n}, Math. Pannon.
(N. S.), 28 (2022), 65–68.

[92] L. Tóth and P. Haukkanen, The discrete Fourier transform of r-even functions, Acta Univ.
Sapientiae, Math., 3 (2011), 5–25.

[93] R. Vaidyanathaswamy, A remarkable property of integers (mod N ) and its bearing on group
theory, Proc. Ind. Acad. Sci. Section A, (1937), 63–75.

[94] T. Venkataraman, A note on the generalization of an arithmetic function in kth power

residue, Math. Student, 42 (1974), 101–102.

[95] S. Venkatramaiah, On a paper of Kesava Menon, Math. Student, 41 (1973), 303–306.

[96] S. Venkatramaiah, A note on a result of Kesava Menon, Math. Student, 43 (1975), 253–256.

[97] S. Venkatramaiah, A note on certain totient functions, Math. Ed. (Siwan). 18 (1984), 66–71.

[98] S. Venkatramaiah, On generalisations of Menon’s identity, Math. Ed. (Siwan), 31 (1997),

220–225.

[99] Yu-Jie Wang, Xin Zhang, and Chun-Gang Ji, A regular Menon-type identity in residually
finite Dedekind domains, Acta Arith., 188 (2019), 111–123.

[100] Yu-Jie Wang, Yi-Jing Hu, and Chun-Gang Ji, Ideal chains in residually finite Dedekined
domains, Bull. Aust. Math. Soc., 99 (2019), 56–67.

[101] Yu-Jie Wang and Chun-Gang Ji, A new Menon’s identity from groups actions, Bull. Aust.
Math. Soc., 99 (2019), 369–375.

[102] Yu-Jie Wang and Chungang Ji, Another generalization of Menon’s identity in the ring of
algebraic integers, Colloq. Math., 160 (2020), 213–221.

[103] Yujie Wang and Chungang Ji, A generalization of Arai-Carlitz’s identity, Ramanujan J.,
53 (2020), 585–594.

[104] E. M. Wright, Burnside’s Lemma: A historical note, J. Combinat. Theory Ser. B, 30

(1981), 89–90.

[105] X.-P. Zhao and Z.-F. Cao, Another generalization of Menon’s identity, Int. J. Number
Theory, 13 (2017), 2373–2379.

46