with arithmetic functions
Andrzej Nowicki Toru´n 18.06.2017
Contents
1 Introduction 1
2 Notations and preparatory facts 2
3 Gcd-determinants 4
3.1 Smith’s theorem . . . 6
3.2 Smith determinants . . . 8
3.3 Power gcd-determinants . . . 10
3.4 Gcd-determinants with tau . . . 11
3.5 Gcd-determinants with sigma . . . 13
3.6 Other gcd-determinants . . . 15
4 Lcm-determinants 16 4.1 Smith lcm-determinants . . . 17
4.2 Power lcm-determinants . . . 18
4.3 Examples of lcm-determinants . . . 19
4.4 Initial values of some lcm-determinants . . . 19
5 Determinants with some arithmetic functions 20 6 Appendix 21 6.1 Arithmetic functions . . . 21
6.2 Examples of factor-closed sets . . . 26
1 Introduction
A Hankel determinant, named after Hermann Hankel, is the determinant of a square matrix in which each ascending skew-diagonal from left to right is constant, e.g.:
a b c b c d c d e
,
a b c d b c d e c d e f d e f g
.
In this article we consider some Hankel determinants. If a, b are positive integers, then we use the standard notations (a, b) = gcd(a, b) and [a, b] = lcm(a, b).
In 1875, J. H. S. Smith [32] proved that
(1, 1) (1, 2) · · · (1, n) (2, 1) (2, 2) · · · (2, n)
... ... ... (n, 1) (n, 2) · · · (n, n)
= ϕ(1)ϕ(2) · · · ϕ(n),
where ϕ is the Euler totient function. He also showed that
[1, 1] [1, 2] · · · [1, n]
[2, 1] [2, 2] · · · [2, n]
... ... ... [n, 1] [n, 2] · · · [n, n]
= n!Y
p
(1 − p)[n/p],
where p ranges over all primes belonging to the set {1, 2, . . . , n}. Moreover, Smith presented several generalizations of the above determinants. Since Smith, a large number of papers on this topic has been published in the literature (see for exam- ple [1, 4, 5, 14, 18, 19, 22, 23, 30]).
In this article we present our proofs of theorems concerning Smith’s determinants.
Moreover, we present a collections of determinants with great common divisors, least common multiplies and some arithmetic functions.
2 Notations and preparatory facts
We denote by N the set {1, 2, . . . }, of all natural numbers. If a, b ∈ N, then we use the standard notations
(a, b) = gcd(a, b) and [a, b] = lcm(a, b).
A finite set S of positive integers is said to be factor-closed if all positive factors of any element of S belong to S, ([4, 22]). For any n ∈ N, the set S = {1, 2, . . . , n} is factor-closed. Other examples are in Appendix.
A function f is said to be arithmetic, if f is an ordinary function from N to the field C of complex numbers. We denote by A the set of all arithmetic functions.
If n ∈ N, then τ (n) is the number of all natural divisors of n, and σ(n) is the sum of all natural divisors of n. Moreover, ϕ is the Euler totient function, and µ is the M¨obius function. We use also the standard functions I, T , Tr and e. If n is a positive integer, then
I(n) = 1, T (n) = n, Tr(n) = nr, e(n) = 1 n
=
( 1, for n = 1 0, for n > 1.
If f, g : N → C are arithmetic functions, then we denote by f ∗ g the Dirichlet convolution of f and g, that is, f ∗ g is a function from N to C defined by
f ∗ g
(n) =X
d|n
f (d)g
n d
,
for all n ∈ N, where d ranges over all positive divisors of n. Basic properties of the Dirichlet convolution one can find in many papers and books (see for example
[11, 12, 26, 28]). It is well known that the set A (of all arithmetic functions) is a commutative ring with respect to the ordinary addition + and the Dirichlet convolution, as the multiplication. This ring is without zero divisors, and the function e is its identity. An arithmetic function f is invertible in this ring if and only if f (1) 6= 0. If a function f ∈ A is invertible in A, then we denote by f−1 the inverse of f , that is
f ∗ f−1 = e.
In this case we say that f−1is the inverse of f with respect to the Dirichlet convolution.
A multiplicative function is an arithmetic function f with the property f (1) = 1 and whenever a and b are coprime, then f (ab) = f (a)f (b). All the functions e, I, T , ϕ, τ , σ and µ are multiplicative. If f, g are multiplicative functions, then the ordinary product f g, n 7→ f (n)g(n), is also a multiplicative function. The following two propositions are well known ([2, 17, 25, 28, 33]).
Proposition 2.1. The Dirichlet convolution of multiplicative functions is an multi- plicative function.
Proposition 2.2. If f is a multiplicative function, then f has the inverse f−1 (with respect to the Dirichlet convolution), and the function f−1 is multiplicative.
Thus, the set of all multiplicative functions is a subgroup of the multiplicative group of the ring A.
An arithmetic function f is said to be completely multiplicative (or totally multi- plicative) if f (1) = 1 and f (ab) = f (a)f (b) holds for all positive integers a and b, even when they are not coprime. The functions e, I, T and Tr are completely multiplica- tive. The Dirichlet convolution of completely multiplicative functions is not completely multiplicative, in general. For example, I is completely multiplicative, and τ = I ∗ I is not completely multiplicative. Note also that µ = I−1, and hence the inverse of a completely multiplicative function is not, in general, completely multiplicative. If f is a completely multiplicative function, then f has the inverse with respect to Dirichlet convolution and the inverse is equal to µf (see Proposition 6.6 in Appendix).
Note several important equalities.
τ = I ∗ I, σ = T ∗ I, ϕ = T ∗ µ, T = ϕ ∗ I, µ ∗ σ = T, ϕ ∗ τ = σ.
I−1 = µ, T−1 = µT, ϕ−1 = I ∗ µT = I ∗ T−1, τ−1 = µ ∗ µ, σ−1 = T−1∗ µ = µT ∗ µ = ϕ−1∗ τ−1.
For any positive integer n and real r, we define Jr(n) = nrY
p|n
1 − 1
pr
,
where p ranges over all prime divisors of n. The function Jr is usually called Jordan’s totient. It is a multiplicative function. In particular, J1 = ϕ, J0 = e. It is easy to show (see Proposition 6.7) that Jr∗ I = Tr. This implies that
Jr = Tr∗ µ, Jr−1 = µTr∗ I.
where Jr−1 is the inverse of Jr with respect to Dirichlet convolution. It is clear that Jr(1)Jr(2) · · · Jr(n) = (n!)rY
p
1 − 1
pr
[n/p]
.
For any real r, we denote by πrthe multiplicative function defined by πr(pm) = −pr, for each prime power pm. Thus, for n ∈ N, we have
πr(n) =Y
p|n
(−pr),
where p ranges over all prime divisors of n. We denote by δrthe multiplicative function πrJr, that is, for all n ∈ N,
δr(n) = nrY
p|n
(1 − pr).
In particular, δ1(n) = ϕ(n)π1(n) = nQ
p|n
(1 − p).
Let c be a fixed positive integer. We denote by Gc the arithmetic function defined by
Gc(n) = gcd(c, n) = (c, n)
for all n ∈ N. In particular, G1 = I. The function Gcis multiplicative (see Proposition 6.9 in Appendix).
The arithmetic function n 7→ [c, n] is clearly not multiplicative. We denote by Hc the arithmetic function defined by
Hc(n) = 1
c[c, n] = n
(c, n), for n ∈ N.
This function is multiplicative (see Proposition 6.17). We consider also the next mul- tiplicative function, denoted by εc, defined by
εc(n) =
( 1, if n | c, 0, otherwise.
In particular, ε1 = e. For example, ε12(n) = 1 for n ∈ {1, 2, 3, 4, 6, 12} and ε12(n) = 0 in other cases. In Appendix some properties of the inverse ε−1c are given.
3 Gcd-determinants
Let c be a fixed positive integer. Let us recall that if n ∈ N, then Gc(n) = gcd(c, n) = (c, n), εc(n) = 1 when n | c, and εc(n) = 0 when n - c. Observe that if a, b are coprime positive integers, then the numbers Gc(a) and Gc(b) are coprime, too. This implies that if f is a multiplicative function, then the function f ◦ Gc, that is, n 7→ f
(c, n)
, is also multiplicative.
Proposition 3.1. Let f, g be multiplicative functions such that f = g ∗ I. Let c ∈ N, and let Fc= f ◦ Gc. Then Fc∗ µ = εcg, that is, if n ∈ N, then
(Fc∗ µ) (n) =
( g(n), if n | c, 0, otherwise.
Proof. Put h := Fc∗ µ, and let n ∈ N.
Step 1. Assume that n | c. In this case h(n) = g(n). Indeed, h(n) = (Fc∗ µ) (n) =P
d|n
Fc(d)µ nd = P
d|n
f
(c, d)
µ nd
= P
d|n
f (d)µ nd = (f ∗ µ) (n) =
(g ∗ I) ∗ µ (n)
=
g ∗ (I ∗ µ)
(n) = g(n).
Thus, if n | c then h(n) = εc(n)g(n) = (εcg) (n).
Step 2. Assume that n > 2 and (c, n) = 1. Then n - c. In this case h(n) = 0.
Indeed,
h(n) = (Fc∗ µ) (n) =P
d|n
Fc(d)µ nd = P
d|n
f
(c, d)
µ nd
= P
d|n
f (1)µ nd = P
d|n
µ(d) = 0.
We applied the equality µ ∗ I = e. Thus, if n > 2 and (c, n) = 1, then h(n) = 0 = 0 · g(n) = εc(n)g(n) = (εcg) (n).
Step 3. Assume that n is a prime power pα with α > 1, and c < n. We shall show that h(n) = 0. Indeed, if p - c then (c, n) = 1 and the assertion follows from Step 1.
Assume that p | c. Let c = pβu, p - u, 1 6 β < α. Then h(n) = (Fc∗ µ) (n) = P
d|pα
Fc(d)µ pdα = Fc(pα) µ(1) + Fc(pα−1) µ(p)
= f
(c, pα)
− f
(c, pα−1)
= f
(pβu, pα)
− f
(pβu, pα−1)
= f pβ − f pβ = 0.
Thus, if n = pα with α > 1, and c < n, then h(n) = 0 = 0·g(n) = εc(n)g(n) = (εcg) (n).
Step 4. Assume that n > 2 and n | c. We shall show that h(n) = 0. For this aim consider the prime decomposition n = pα11· · · pαss. Since n - c, there exists an i ∈ {1, . . . , s} such that pαii - c. Put p := pi and α := αi. Let c = upα + v, where u, v ∈ Z and 1 6 v < pα. Then h(pα) = 0. Indeed, using Step 3, we have
h(pα) = (Fc∗ µ) (pα) = P
d|pα
Fc(d)µ pdα = P
d|pα
f
(upα+ v, d)
µ pdα
= P
d|pα
f (v, d)
µ pdα = P
d|pα
Fv(d)µ pdα = (Fv ∗ µ) (pα) = 0.
Hence, h (pαii) = 0. But the function h is multiplicative, so
h(n) = h (p1α1· · · pαss) = h (p1α1) · · · h (pαii) · · · h (pαss) = 0.
Thus, in any case we have the equality h(n) = (εcg) (n). This completes the proof. Proposition 3.2. Let f, g be multiplicative functions such that f = g ∗ I, and let S = {x1, . . . , xn} be a factor-closed set of positive integers. Assume that x1 < x2 < · · · < xn. Then for every i ∈ {1, . . . , n} the following equalities hold.
X
xk|xn
f
(xi, xk) µ xn
xk
=
g(xn), if i = n, 0, otherwise.
Proof. Let i ∈ {1, . . . , n}, and let Fxi = f ◦ Gxi, that is, Fxi(m) = f
(xi, m) for all m ∈ N. Then we have
X
xk|xn
f
(xi, xk) µ xn
xk
= X
xk|xn
Fxi(xk)µ xn xk
=X
d|xn
Fxi(d)µxn d
= (Fxi ∗ µ) (xn).
Thus, this assertion follows from Proposition 3.1.
3.1 Smith’s theorem
In 1875 J. H. S. Smith [32] proved the following theorem.
Theorem 3.3 (Smith 1875). Let S = {x1, . . . , xn} (where n > 1) be a factor-closed set of positive integers with x1 < x2 < · · · < xn. If f, g : N → C are multiplicative functions such that f = g ∗ I, then
deth f
(xi, xj)i
16i,j6n
= g(x1)g(x2) · · · g(xn).
In other words, if f, g : N → C are functions such that f (m) = X
d|m
g(d) for m ∈ N, then
f
(x1, x1) f
(x1, x2)
· · · f
(x1, xn) f
(x2, x1) f
(x2, x2)
· · · f
(x2, xn)
... ... ...
f
(xn, x1) f
(xn, x2)
· · · f
(xn, xn)
= g(x1)g(x2) · · · g(xn).
Proof. (Smith [32]). Denote the matrix h f
(xi, xj)i
by Mn. Let Dn = det Mn, and let A1, . . . , An be the columns of the matrix Mn. Replace the last column An by the sum
X
xk|xn
µ xn xk
Ak.
Then the value of Dn is not changed and, by Proposition 3.2, the last new column is equal to [0, 0, . . . , 0, g(xn)]T. Hence,
Dn= g(xn)Dn−1 and consequently, Dn = g(xn)g(xn−1) · · · g(x2)g(x1).
If S = {1, 2, . . . , n}, then the above theorem has the following form.
Theorem 3.4 (Smith 1875). Let n be a positive integer. If f, g : N → C are functions such that f = g ∗ I, then
deth f
(i, j)i
16i,j6n = g(1)g(2) · · · g(n).
In other words, if f, g : N → C are functions such that f (m) = X
d|m
g(d) for m ∈ N,
then
f
(1, 1)
f
(1, 2)
· · · f
(1, n)
f
(2, 1) f
(2, 2)
· · · f
(2, n)
... ... ...
f
(n, 1) f
(n, 2)
· · · f
(n, n)
= g(1)g(2) · · · g(n).
Proofs of this form can be found in several references ([32, 9, 10, 29]). Of course it is an immediate consequence of Theorem 3.3. We present a second proof.
Proof. ([9]). Denote this determinant by D. Consider the numbers aij defined by
aij =
( 1, when j | i, 0, when j - i.
Then, for all i, j ∈ N, we have f i, j)
= X
d|(i,j)
g(d) =
n
X
k=1
aikajkg(k), and this implies that D = A · B, where
A =
a11 a12 · · · a1n a21 a22 · · · a2n ... ... ... an1 an2 · · · ann
, B =
a11g(1) a21g(1) · · · an1g(1) a12g(2) a22g(2) · · · an2g(2)
... ... ...
a1ng(n) a2ng(n) · · · anng(n)
.
Observe that aii = 1 and aij = 0 for i < j. Hence, A = 1 and B = g(1)g(2) · · · g(n).
Therefore, D = g(1)g(2) · · · g(n).
Proposition 3.5. Let Dn = deth f
(i, j)i
16i,j6n, where f : N → C is a function. If f (p) = f (1) for a prime number p, then Dn = 0 for all n > p.
Proof. Observe that f = g ∗ I, where g = f ∗ µ. Since f (p) = f (1), we have g(p) = 0 because g(p) = f (1)µ(p) + f (p)µ(1) = −f (1) + f (p) = 0. Hence, if n > p then, by Theorem 3.3, we have Dn = g(1) · · · g(p) · · · g(n) = 0.
Example 3.6. Let Dn = deth f
(i, j)i
16i,j6n
, where f (x) = x2 − 4x + 20. Then D1 = 17, D2 = −17, and Dn = 0 for n > 3. It follows from Proposition 3.3, because f (1) = f (3).
Example 3.7. Let Dn = deth f
(i, j)i
16i,j6n
, where f (x) = x2 − 8x + 100. Then Dn= 0 for n > 7. It follows from Proposition 3.3, because f (1) = f (7).
3.2 Smith determinants
As a consequence of Theorem 3.3 and the equality T = ϕ∗I, we obtain the following theorem
Theorem 3.8 (Smith 1875).
(x1, x1) (x1, x2) · · · (x1, xn) (x2, x1) (x2, x2) · · · (x2, xn)
... ... ...
(xn, x1) (xn, x2) · · · (xn, xn)
= ϕ(x1)ϕ(x2) · · · ϕ(xn).
If S = {1, 2, . . . , n}, then we obtain the following well known theorem (see for example [32, 14, 4, 22]).
Theorem 3.9 ([32]).
(1, 1) (1, 2) · · · (1, n) (2, 1) (2, 2) · · · (2, n)
... ... ... (n, 1) (n, 2) · · · (n, n)
= ϕ(1)ϕ(2) · · · ϕ(n).
Examples: D1 = D2 = 1, D3 = 2, D4 = 4, D5 = 16, D6 = 32, D7 = 263, D8 = 283, D9 = 2932, D10= 21132, D11 = 212325, D12= 214325.
The above determinant is called the Smith determinant. Many generalizations of Smith determinants have been presented in literature, see [1, 5, 18, 19, 23, 30]. Dickson ([15] 122-129) reports on several papers devoted to proofs and extensions of Smith’s determinant.
Let S = {x1, . . . , xn} be a finite ordered set of distinct positive integers. We do not assume that S is factor-closed. The gcd matrix defined on S is given by [(xi, xj)] and is denoted by (S). In 1989, Scott Beslin and Steve Ligh [4] gave the conjecture that if det(S) = ϕ(x1)ϕ(x2) · · · ϕ(xn) then S is factor-closed. In 1990, Zhongshan Li [22]
proved that this conjecture is true.
Theorem 3.10 (Li [22]). Let S = {x1, . . . , xn} be an ordered set of distinct positive integers. Then det(S) > ϕ(x1)ϕ(x2) · · · ϕ(xn), and the equality
det(S) = ϕ(x1)ϕ(x2) · · · ϕ(xn) holds if and only if the set S is factor-closed.
The matrix (S) is always positive definite ([4, 22]). Moreover (see [22]), det(S) 6 x1x2· · · xn− 1
2n! . Note also:
Proposition 3.11. Let S={x1, . . . , xn} be an ordered set of distinct positive integers.
If S is factor closed, then
ϕ
(x1, x1) ϕ
(x1, x2)
· · · ϕ
(x1, xn) ϕ
(x2, x1) ϕ
(x2, x2)
· · · ϕ
(x2, xn)
... ... ...
ϕ
(xn, x1) ϕ
(xn, x2)
· · · ϕ
(xn, xn)
= h(x1)h(x2) · · · h(xn),
where h = ϕ ∗ µ.
Proof. This is an immediate consequence of Theorem 3.3, because ϕ = h ∗ I.
Example 3.12.
ϕ (1, 1)
ϕ (1, 2)
· · · ϕ
(1, n) ϕ
(2, 1) ϕ
(2, 2)
· · · ϕ
(2, n)
... ... ...
ϕ
(n, 1) ϕ
(n, 2)
· · · ϕ
(n, n)
= 0 for n > 2.
Proof. It follows from Proposition 3.11 that this determinant equals h(1) · · · h(n), where h = ϕ∗µ. But h(2) = ϕ(1)µ(2)+ϕ(2)µ(1) = −1+1 = 0. Hence, the determinant equals 0 for n > 2.
Proposition 3.13. Let f : N → C be the multiplicative function defined by f (pm) = 1 + pm− 1
pm−1(p − 1)2
for each prime power pm. Let S = {x1, . . . , xn} be a factor-closed set of positive inte- gers. Then
f
(x1, x1) f
(x1, x2)
· · · f
(x1, xn) f
(x2, x1) f
(x2, x2)
· · · f
(x2, xn)
... ... ...
f
(xn, x1) f
(xn, x2)
· · · f
(xn, xn)
= 1
ϕ(x1)ϕ(x2) · · · ϕ(xn).
Proof. Denote by g the function µ ∗ f . Then, for each prime power pm, we have g (pm) = µ(1)f (pm) + µ(p)f pm−1 = f (pm) − f pm−1 = 1
pm−1(p − 1) = 1 ϕ(pm). Hence, g(m) = ϕ(m)1 for all positive integer m, because the functions g and ϕ are multiplicative. Note that f = g ∗ I. Therefore, by Theorem 3.3, the determinant is equal to g(x1) · · · g(xn) = ϕ(x 1
1)···ϕ(xn).
3.3 Power gcd-determinants
In this subsection r is a real number. Let us recall (see Section 2) that Jordan’s totient function is defined by
Jr(n) = nrY
p|n
1 − 1
pr
for n ∈ N, where p ranges over all prime divisors of n. In particular, J1 = ϕ and J0 = e.
We know that Jr∗ I = Tr, where Tr(n) = nr for n ∈ N (see Proposition 6.7). This implies that Jr= Tr∗ µ, Jr−1 = µTr∗ I, where Jr−1 is the inverse of Jr with respect to Dirichlet convolution.
Theorem 3.14 ([32, 11]). Let S={x1, . . . , xn} be an ordered set of distinct positive integers, and let r be a real number. If S is factor closed, then
(x1, x1)r (x1, x2)r · · · (x1, xn)r (x2, x1)r (x2, x2)r · · · (x2, xn)r
... ... ...
(xn, x1)r (xn, x2)r · · · (xn, xn)r
= Jr(x1)Jr(x2) · · · Jr(xn).
Proof. Use Theorem 3.3 and the equality Jr∗ I = Tr. For S = {1, 2, . . . , n} the above theorem has the following form.
Theorem 3.15 ([9, 11]). If r is a real number and n is a positive integer, then
(1, 1)r (1, 2)r · · · (1, n)r (2, 1)r (2, 2)r · · · (2, n)r
... ... ...
(n, 1)r (n, 2)r · · · (n, n)r
= Jr(1)Jr(2) · · · Jr(n) = (n!)rY
p
1 − 1
pr
[n/p]
,
where p ranges over all primes belonging to the set {1, 2, . . . , n}.
Note some special cases of the above theorems. We usually denote by Dn every determinant, which appears in the presented theorems and propositions.
3.16. Let Dn= det [(i, j)2]n×n = det [(i2, j2)]n×n. Then Dn= J2(1)J2(2) · · · J2(n) = (n!)2Y
p
1 − 1
p2
[n/p]
, where p ranges over all primes belonging to {1, 2, . . . , n}.
Examples: D1 = 1, D2 = 3, D3 = 24, D4 = 288 = 2532, D5 = 2833, D6 = 21134, D7 = 21535, D8 = 21936, D11 = 2283115.
3.17. Let S={x1, . . . , xn} be an ordered set of distinct positive integers, and let r be a real number. If S is factor closed, then
p(x1, x1) p(x1, x2) · · · p(x1, xn) p(x2, x1) p(x2, x2) · · · p(x2, xn)
... ... ...
p(xn, x1) p(xn, x2) · · · p(xn, xn)
= J1/2(x1)J1/2(x2) · · · J1/2(xn).
3.18. Let Dn= deth
p(i, j)i
n×n. Then Dn =p(n!) Q
p
1 − √1p[n/p]
, where p ranges over all primes belonging to {1, 2, . . . , n}.
Proposition 3.19. Let S={x1, . . . , xn} be an ordered set of distinct positive integers, and let r be a real number. If S is factor closed, then
1 (x1,x1)
1
(x1,x2) · · · (x1
1,xn) 1
(x2,x1) 1
(x2,x2) · · · (x1
2,xn)
... ... ...
1 (xn,x1)
1
(xn,x2) · · · (x 1
n,xn)
= J−1(x1)J−1(x2) · · · J−1(xn).
3.20. Let Dn= deth
1 (i,j)
i
n×n
. Then Dn= 1 n!
Y
p
(1 − p)[n/p].
Examples: D1 = 1, D2 = −12, D3 = 13, D4 = −121, D5 = 151, D6 = 451, D7 = −1052 , D8 = 4201 , D9 = −18901 , D10 = −47251 . ([34], see Theorem 3.15).
3.21. Let Dn= deth
1 (i,j)2
i
n×n
= deth
1 (i2,j2)
i
n×n
. Then Dn= 1 (n!)2
Y
p
1 − p2[n/p]
. Examples: D1 = 1, D2 = −34, D3 = 23, D4 = −81, D5 = 253, D6 = 252, D7 = −122596 , D8 = 24509 , D9 = −110254 , D10= −306258 . (see Theorem 3.15).
3.22. If r is a real number, and {x1, . . . , xn} is a factor-closed set of positive integers, then
Jr
(x1, x1) Jr
(x1, x2)
· · · Jr
(x1, xn) Jr
(x2, x1) Jr
(x2, x2)
· · · Jr
(x2, xn)
... ... ...
Jr
(xn, x1) Jr
(xn, x2)
· · · Jr
(xn,n)
= h(x1)h(x2) · · · h(xn),
where h = Jr∗ µ is the multiplicative function defined, for prime powers pm, by h(p) = pr− 2 and h(pm) = pmr
1 − p1r
2
for m > 2.
3.4 Gcd-determinants with tau
Let us recall that τ (n) is the number of all positive divisors of n.
Proposition 3.23. If {x1, . . . , xn} is a factor-closed set of positive integers, then
τ
(x1, x1) τ
(x1, x2)
· · · τ
(x1, xn) τ
(x2, x1) τ
(x2, x2)
· · · τ
(x2, xn)
... ... ...
τ
(xn, x1) τ
(xn, x2)
· · · τ
(xn, xn)
= 1.
Proof. It follows from Theorem 3.3, because τ = I ∗ I. Proposition 3.24 ([31]). The determinant
τ (2, 2)
τ (2, 3)
· · · τ
(2, n) τ
(3, 2) τ
(3, 3)
· · · τ
(3, n)
... ... ...
τ
(n, 2) τ
(n, 3)
· · · τ
(n, n)
is equal to the number of all square-free numbers belonging to {1, 2, . . . , n}.
Proof. Let dij = τ
(i, j)
for all i, j ∈ {1, . . . , n}. Consider the matrices
D =
d22 d23 · · · d2n
d32 d33 · · · d3n ... ... ... dn2 dn3 · · · dnn
, E =
1 d12 d13 · · · d1n 0 d22 d23 · · · d2n 0 d32 d33 · · · d3n ... ... ... 0 dn2 dn3 · · · dnn
.
Since det D = det E, we need to show that det E is equal to the number of all square- free numbers belonging to {1, 2, . . . , n}.
Denote by Ei the i-th row of E. Beginning with i = n and proceeding towards i = 2, replace each Ei, by
X
k|i
µ i k
Ek.
Let γij = 1 if i | j, and γij = 0 if i - j. Observe that P
k|i
γkj = dij. Hence, using the M¨obius inversion formula we obtainP
k|i
µ ki dkj = γij. This implies that det E = det F , where
F =
1 γ12 γ13 · · · γ1n µ(2) γ22 γ23 · · · γ2n µ(3) γ32 γ33 · · · γ3n ... ... ... µ(n) γn2 γn3 · · · γnn
.
Denote by Fi the i-th row of F . Let G = [gij]16i,j6n be the matrix obtained from F by replacing F1 by
n
P
i=1
µ(i)Fi. Then det G = det F = det D. If j > 1, then g1j = 0. In fact,
g1j =
n
X
i=1
µ(i)γij =X
i|j
µ(i) = 0.
Moreover, g11 =
n
P
i=1
µ(i)2 is equal to the number of all square-free integers belonging to {1, 2, . . . , n}. Observe that [gij]26i,j6n is an upper triangular matrix with 1’s on the diagonal. Hence, det D = det G = g11. Thus det D is equal to the number of all square-free integers belonging to {1, 2, . . . , n}.
The next proposition is an immediate consequence of Theorem 3.3.
Proposition 3.25. Let {x1, . . . , xn} be a factor-closed set of positive integers.
If h = I ∗ I ∗ I = τ ∗ I, then
h
(x1, x1) h
(x1, x2)
· · · h
(x1, xn) h
(x2, x1) h
(x2, x2)
· · · h
(x2, xn)
... ... ...
h
(xn, x1) h
(xn, x2)
· · · h
(xn, xn)
= τ (x1)τ (x2) · · · τ (xn).
3.5 Gcd-determinants with sigma
Let us recall that σ(n) is the sum of all positive divisors of n. If r is a real number, then σr(n) is the sum of the powers r of all positive divisors of n. In particular, σ1 = σ.
Note that σ = T ∗ I and σr= Tr∗ I. Hence, immediately from Theorem 3.3 we obtain the following two propositions.
Proposition 3.26. If r is a real number and {x1, . . . , xn} is a factor-closed set of positive integers, then
σr
(x1, x1) σr
(x1, x2)
· · · σr
(x1, xn) σr
(x2, x1)
σr
(x2, x2)
· · · σr
(x2, xn)
... ... ...
σr
(xn, x1) σr
(xn, x2)
· · · σr
(xn, xn)
= (x1x2. . . xn)r.
Proposition 3.27. If {x1, . . . , xn} is a factor-closed set of positive integers, then
σ
(x1, x1) σ
(x1, x2)
· · · σ
(x1, xn) σ
(x2, x1) σ
(x2, x2)
· · · σ
(x2, xn)
... ... ...
σ
(xn, x1) σ
(xn, x2)
· · · σ
(xn, xn)
= x1x2. . . xn.
Note some special cases of the above propositions, which appear in [32, 14, 9].
3.28.
σ (1, 1)
σ (1, 2)
· · · σ
(1, n) σ
(2, 1) σ
(2, 2)
· · · σ
(2, n)
... ... ...
σ
(n, 1) σ
(n, 2)
· · · σ
(n, n)
= n!.
3.29.
σr (1, 1)
σr (1, 2)
· · · σr
(1, n) σr
(2, 1) σr
(2, 2)
· · · σr
(2, n)
... ... ...
σr
(n, 1) σr
(n, 2)
· · · σr
(n, n)
= (n!)r.
3.30.
σ−1
(1, 1)
σ−1
(1, 2)
· · · σ−1
(1, n) σ−1
(2, 1)
σ−1 (2, 2)
· · · σ−1
(2, n)
... ... ...
σ−1
(n, 1) σ−1
(n, 2)
· · · σ−1
(n, n)
= 1 n!.
3.31. If f : N → C is the multiplicative function defined by f (pm) = ppmm+1(p−1)−1 for each prime power pm, then
f
(1, 1)
f
(1, 2)
· · · f
(1, n)
f
(2, 1) f
(2, 2)
· · · f
(2, n)
... ... ...
f
(n, 1) f
(n, 2)
· · · f
(n, n)
= 1 n!.
Proof. Denote by g the function µ ∗ f . Then, for each prime power pm, we have g (pm) = µ(1)f (pm) + µ(p)f pm−1 = pm+1 − 1
pm(p − 1) − pm− 1
pm−1(p − 1) = 1 pm.
Hence, g(m) = m1 for all positive integer m, because g multiplicative. Note that f = g ∗ I. Therefore, by Theorem 3.3, the determinant is equal to g(1) · · · g(n) = n!1.
The next propositions are immediate consequences of Theorem 3.3.
Proposition 3.32. If r is a real number and {x1, . . . , xn} is a factor-closed set of positive integers, then
h
(x1, x1)
h
(x1, x2)
· · · h
(x1, xn)
h
(x2, x1) h
(x2, x2)
· · · h
(x2, xn)
... ... ...
h
(xn, x1) h
(xn, x2)
· · · h
(xn, xn)
= σr(x1)σr(x2) . . . σr(xn),
where h = σr∗ I.
Proposition 3.33. If h = σ ∗ I, then
h (1, 1)
h (1, 2)
· · · h
(1, n) h
(2, 1) h
(2, 2)
· · · h
(2, n)
... ... ...
h
(n, 1) h
(n, 2)
· · · h
(n, n)
= σ(1)σ(2) · · · σ(n).
3.6 Other gcd-determinants
Recall that e(1) = 1 and e(n) = 0 for n > 2, and moreover, e = µ ∗ I. Hence, by Theorem 3.3, we obtain
Proposition 3.34. If {x1, . . . , xn} is a factor-closed set of positive integers, then
e
(x1, x1)
e
(x1, x2)
· · · e
(x1, xn)
e
(x2, x1) e
(x2, x2)
· · · e
(x2, xn)
... ... ...
e
(xn, x1) e
(xn, x2)
· · · e
(xn, xn)
= µ(x1)µ(x2) . . . µ(xn).
Proposition 3.35.
e (1, 1)
e (1, 2)
· · · e
(1, n) e
(2, 1) e
(2, 2)
· · · e
(2, n)
... ... ...
e
(n, 1)
e
(n, 2)
· · · e
(n, n)
= µ(1)µ(2) · · · µ(n).
In particular: D1 = 1, D2 = −1, D3 = 1, and Dn= 0 for n > 4.
Proposition 3.36. If {x1, . . . , xn} is a factor-closed set of positive integers, then
µ
(x1, x1) µ
(x1, x2)
· · · µ
(x1, xn) µ
(x2, x1) µ
(x2, x2)
· · · µ
(x2, xn)
... ... ...
µ
(xn, x1) µ
(xn, x2)
· · · µ
(xn, xn)
= h(x1)h(x2) . . . h(xn),
where h = µ ∗ µ = τ−1. Colorary 3.37. Let
Dn=
µ (1, 1)
µ (1, 2)
· · · µ
(1, n) µ
(2, 1)
µ
(2, 2)
· · · µ
(2, n)
... ... ...
µ
(n, 1) µ
(n, 2)
· · · µ
(n, n) .
Then D1 = 1, D2 = −2, D3 = 22, D4 = 22, D5 = −23, D6 = −25, and D7 = 26, and Dn= 0 for n > 8.
In the next gcd-determinants appears the multiplicative function πr. The following propositions are consequences of Theorem 3.3.
Proposition 3.38. If r is a real number, and {x1, . . . , xn} is a factor-closed set of positive integers, then
πr
(x1, x1) πr
(x1, x2)
· · · πr
(x1, xn) πr
(x2, x1)
πr
(x2, x2)
· · · πr
(x2, xn)
... ... ...
πr
(xn, x1) πr
(xn, x2)
· · · πr
(xn, xn)
= h(x1)h(x2) . . . h(xn),
where h is the multiplicative function defined, for each prime power pm, by h(pm) =
−(1 + pr) for m = 1, and h(pm) = 0 for m > 1.
Colorary 3.39.
πr (1, 1)
πr (1, 2)
· · · πr
(1, n) πr
(2, 1)
πr
(2, 2)
· · · πr
(2, n)
... ... ...
πr
(n, 1) πr
(n, 2)
· · · πr
(n, n)
= 0 for n > 4.
Proposition 3.40. Let r be a real number and let {x1, . . . , xn} be a factor-closed set of positive integers. Let f = πr ∗ I. Then f is the multiplicative function defined by f (pm) = 1 − mpr, for each prime power pm, and we have
f
(x1, x1) f
(x1, x2)
· · · f
(x1, xn) f
(x2, x1) f
(x2, x2)
· · · f
(x2, xn)
... ... ...
f
(xn, x1)
f
(xn, x2)
· · · f
(xn, xn)
= πr(x1)πr(x2) · · · πr(xn).
Colorary 3.41.et r be a real number and let f = πr∗ I. Then
f
(1, 1)
f
(1, 2)
· · · f
(1, n)
f
(2, 1) f
(2, 2)
· · · f
(2, n)
... ... ...
f
(n, 1) f
(n, 2)
· · · f
(n, n)
= πr(1)πr(2) · · · πr(n) =Y
p
(−pr)[n/p].
4 Lcm-determinants
Let us recall (see Section 2) that if r is a real number, then πr is the multiplicative function defined by πr(pm) = −pr, for each prime power pm. We denote by δr the multiplicative function Jrπr, that is, δr(n) = Jr(n)πr(n) for n ∈ N. In particular,
δ1(n) = ϕ(n)π1(n) = nY
p|n
(1 − p), δr(n) = nrY
p|n
(1 − pr).
4.1 Smith lcm-determinants
Let us start with the following well known theorem (see, for example [32, 3, 7, 27, 34]).
Theorem 4.1 (Smith 1875).
[1, 1] [1, 2] [1, 3] · · · [1, n]
[2, 1] [2, 2] [2, 3] · · · [2, n]
[3, 1] [3, 2] [3, 3] · · · [3, n]
... ... ... ... [n, 1] [n, 2] [n, 3] · · · [n, n]
= δ1(1)δ1(2) · · · δ1(n) = n!Y
p
(1 − p)[n/p],
where p ranges over all primes belonging to the set {1, 2, . . . , n}.
Examples: D1 = 1, D2 = −2, D3 = 12, D4 = −48, D5 = 960 = 26 · 3 · 5, D6 = 11520 = 28· 32 · 5.
Proof. ([34]). Let A, B, N be the n × n matrices defined by A = h [i, j]i
16i,j6n
, B = h
1 (i,j)
i
16i,j6n
and N = h niji
16i,j6n
, where nii = i and nij = 0 for i 6= j. Then A = N BN , det N = n!, and det B = n!1 Q
p
(1 − p)[n/p] (see 3.20). Hence, det A = (n!)2det B = n!Y
p
(1 − p)[n/p].
The equality
n
Q
k=1
δ1(k) = n!Q
p
(1 − p)[n/p] is obvious.
The above determinant is called the Smith lcm-determinant. Several generalizations of this determinants have been presented in literature, see for example [11, 20].
Smith [32] observed that this result remains valid if the set {1, 2, . . . , n} is replaced by a factor-closed set (see also [7, 8, 20]).
Theorem 4.2 ([32]). Let S = {x1, . . . , xn} be an ordered set of distinct positive integers. If S is factor-closed, then
[x1, x1] [x1, x2] [x1, x3] · · · [x1, xn] [x2, x1] [x2, x2] [x2, x3] · · · [x2, xn]
... ... ... ...
[xn, x1] [xn, x2] [xn, x3] · · · [xn, xn]
= δ1(x1)δ2(x2) · · · δ1(xn).
Proof. We do a small modification of the proof of Theorem 4.1. We use the equalities [xi, xj] = (xxixj
i,xj) and we apply results from the previous section.
It follows from this theorem that if S is factor-closed then the above determinant is nonzero. In a general case, when S is not factor-closed, this determinant may not be nonzero (see [7]). For example, if S = {1, 2, 15, 42}, then
det[S] =
1 2 15 42
2 2 30 42
15 30 15 210 42 42 210 42
= 0.
The same we have for S = {1, 2, 3, 4, 5, 6, 10, 45, 180}. This set is gcd-closed but not factor-closed ([19]).
4.2 Power lcm-determinants
Theorem 4.3 ([32]). If r is a real number, then
[1, 1]r [1, 2]r · · · [1, n]r [2, 1]r [2, 2]r · · · [2, n]r [3, 1]r [3, 2]r · · · [3, n]r
... ... ... [n, 1]r [n, 2]r · · · [n, n]r
= δr(1)δr(2) · · · δr(n) = (n!)rY
p
(1 − pr)[n/p],
where δr = πrJr, and where p ranges over all primes belonging to the set {1, 2, . . . , n}.
Proof. Let A, B, N be the n × n matrices defined by A = h [i, j]ri
, B = h
(i, j)−ri
, N = h niji
, where N is the diagonal matrix with nii = ir and nij = 0 for i 6= j. Then A = N BN , det N = (n!)r, and B = (n!)−rQ
p
(1 − pr)[n/p] (see Theorem 3.15). Hence, det A = (n!)2rdet B = (n!)rQ
p
(1 − pr)[n/p].
Colorary 4.4.
1 [1,1]
1
[1,2] · · · [1,n]1
1 [2,1]
1
[2,2] · · · [2,n]1 ... ... ...
1 [n,1]
1
[n,2] · · · [n,n]1
= δ−1(1)δ−1· · · δ−1(n) = 1 n!
Y
p
1 − 1
p
[n/p]
,
where δ−1= π−1J−1, and where p ranges over all primes belonging to the set {1, 2, . . . , n}.
Examples: D1 = 1, D2 = 14, D3 = 181, D4 = 1441 , D5 = 9001 , D6 = 162001 , D7 = 1323001 , D8 = 21168001 . Observe that Dn> 0 for every n.
Note also
Theorem 4.5 ([32, 11]). Let S={x1, . . . , xn} be an ordered set of distinct positive integers, and let r be a real number. If S is factor closed, then
[x1, x1]r [x1, x2]r · · · [x1, xn]r [x2, x1]r [x2, x2]r · · · [x2, xn]r [x3, x1]r [x3, x2]r · · · [x3, xn]r
... ... ...
[xn, x1]r [xn, x2]r · · · [xn, xn]r
= δr(x1)δr(x2) · · · δr(xn).
Proof. We do a small modification of the proof of Theorem 4.3. We use the equalities [xi, xj] = (xxixj
i,xj) and we apply results from the previous section.