1 Introduction
Multiplicative functions play a central role in number theory, and a fundamental problem is to understand how their behaviour is constrained by conditions imposed on special sets of integers. A classical direction, initiated by Spiro [Reference Spiro11], is to impose additive relations on structured subsets of the natural numbers and study the extent to which such conditions determine the function. For instance, Spiro [Reference Spiro11] showed that a multiplicative function
$f : \mathbb {N} \rightarrow \mathbb {C}$
satisfying
$f(p + q) = f(p) + f(q)$
for primes p and q is, under mild hypotheses, the identity function. Related problems have since been studied for other structured sets.
Let
$k \geq 2$
,
$S \subseteq \mathbb {N}$
, and let
$\mathcal {F}$
be a class of arithmetic functions. For
$f \in \mathcal {F}$
, consider the functional equation
for arbitrary
$x_1, \ldots , x_k \in S$
. One then asks whether such a relation determines f. We say that S is a k-additive uniqueness set for
$\mathcal {F}$
if a function
$f \in \mathcal {F}$
satisfying condition (1.1) is uniquely determined. In recent years, this problem has been studied in connection with polygonal numbers. In particular, results have been obtained for triangular numbers [Reference Chung and Phong2, Reference Park8], squares [Reference Chung1, Reference Park7], generalised pentagonal and hexagonal numbers [Reference Hasanalizade3, Reference Kim, Kim, Lee and Park5, Reference Kim, Kim, Lee and Park6], and generalised octagonal numbers [Reference Hasanalizade and Park4]. More recently, the
$2$
-additive uniqueness of generalised nonagonal numbers was considered in [Reference Park9].
The generalised s-gonal numbers are given by
We consider the case
$s=10$
, the generalised decagonal numbers
We study multiplicative functions satisfying
for all nonzero integers m and n, and determine all such functions.
Since
$D_n \equiv 1 \pmod {3}$
if
$3 \nmid n$
, while
$9 \mid D_n$
if
$3 \mid n$
, integers of the form
$3d$
with
$3 \nmid d$
cannot be expressed as sums of two elements of
$\mathcal {D}$
. This suggests that
$f(3)$
may be arbitrary.
Define a function
$g: \mathbb {N} \rightarrow \mathbb {C}$
by
The main result of this work is as follows.
Theorem 1.1. If a multiplicative function f satisfies
for all nonzero generalised decagonal numbers a and b, then
$f = g$
.
2 Proof of Theorem 1.1
We prove Theorem 1.1 by induction, beginning with small values of n. If
$N = ab$
with
$a,b \geq 2$
and
$\gcd (a,b)=1$
, then
$f(N) = f(a)f(b)$
. Thus, it suffices to show that
$f(3^k) = 3^k$
for
$k\geq 2$
and
$f(p^k)=p^k$
for primes
$p\neq 3$
.
We repeatedly use the relation
$f(D_n + D_m) = f(D_n) + f(D_m)$
for suitable m and n. In general, we cannot write
$f(m(4m-3)) = f(m)f(4m-3)$
and we must account for the divisibility of the factors.
We first evaluate
$f(n)$
for small n.
Lemma 2.1. For
$n \leq 32$
, we have
$f(n)=g(n)$
.
Proof. We have
$f(1) = 1$
and
$f(2) = 2 f(1) = 2$
. Note that
Thus,
$f(5) = 0$
if and only if
$f(11) =1$
.
If
$f(5) = 0$
, then
$f(11) = 1$
, so
$f(23) = f(1) + f(2) f(11) = 3$
. However, this contradicts
Hence,
$f(5) \neq 0$
and
$f(4)=4$
. Moreover,
From
we obtain
Thus,
Combining (2.1) and (2.2) gives
so
$f(5) = -1/2$
or
$f(5)=5$
.
From
we obtain
The latter case leads to a contradiction. Indeed,
Using
we conclude that
$f(n) = g(n)$
for
$n \leq 32$
.
We now extend this to higher powers of
$3$
.
Lemma 2.2. Let
$k \geq 2$
. If
$f(n) = n$
for
$n \leq 3^{k-1}$
such that
$n \not \equiv \pm 3 \pmod {9}$
, then
$f(3^k) =~3^k$
.
Proof. We already know that
$f(9)=9$
.
Suppose
$k=2r$
, with
$r \geq 2$
. Then,
Since
$3^{r+1} \leq 3^{2r-1}$
and
$4 \cdot 3^{r-1} \pm 1 < 3^{2r-1}$
, it follows that
$f(3^{2r}) = 3^{2r}$
.
Suppose
$k=2r+1$
. Since we have
$f(27)=27$
, we may assume
$r \geq 2$
. Note that
$3^{2r} \equiv \pm 1 \pmod {10}$
. Assume that
$3^k = 3(10m + 1)$
for some integer
$m \geq 8$
. Consider
Since
$9 \mid 10m+1$
, we have
$m \equiv -1 \pmod {9}$
and
$4m+1 \equiv -3 \pmod {9}$
. Thus,
Rewriting (2.3),
Since
$27, m, 3m+3, (4m+1)/3 < 10m+1 = 3^{2r}$
, we obtain
$f(3^k) = 3^k$
.
A similar argument using
shows that if
$3^k = 3(10m - 1)$
for some integer
$m> 1$
, then
$f(3^k) = 3^k$
.
We now turn to powers of
$5$
.
Lemma 2.3. Let
$k \geq 2$
. If
$f(n) = n$
for
$n < 5^{k}$
such that
$n\not \equiv \pm 3 \pmod {9}$
, then
$f(5^k) =~5^{k}$
.
Proof. We already know that
$f(5) =5$
and
$f(25) = 25$
.
Suppose
$k = 2r$
, with
$r \geq 2$
. Consider
Since
$5^r, 4 \cdot 5^{r} \pm 3 < 5^{r+1} \leq 5^{2r-1}$
, it follows that
$f(5^{2r}) = 5^{2r}$
.
Suppose
$k = 2r+1$
, where
$r \geq 1$
. Write
$5^{2r} = 4m+1$
for some
$m \geq 6$
with
$5 \nmid m$
. Consider
where
$9m, 4m+1, 5m+2, 17m+5 < 20m+5 = 5^{k}$
.
If
$3 \mid m$
, we write
$m = 3^u d$
for some
$d \geq 1$
,
$3 \nmid d$
. Then,
$f(9m) = f(3^{u+2})f(d) = 3^{u+2}d = 9m$
, and since
$3 \nmid (4m+1)(5m+2)(17m+5)$
, we have
$f(a) = a$
for
$a \in \{ 4m+1, 5m+2, 17m+5 \}$
.
If
$3 \nmid m$
, then
$f(9m) = f(9)f(m) = 9m$
. If
$m \equiv 2 \pmod {3}$
, then
which is a contradiction. Thus,
$m \equiv 1 \pmod {3}$
and
$3 \nmid (4m+1)(5m+2)(17m+5)$
.
In all cases, it follows that
$f(5^k) = 5^k$
.
Before proceeding to the next case, we record the following useful observation.
Lemma 2.4.
-
(1) The only nonnegative solution to
(2.4) $$ \begin{align} 2^x + 3 = 5^y \end{align} $$is $(x,y)=(1,1)$
.
-
(2) The only nonnegative solutions to
(2.5) $$ \begin{align} 2^x - 3 = 5^y \end{align} $$are $(x,y) = (2,0), (3,1), (7,3)$
.
Proof. It is easy to show that (2.4) has the unique solution
$(x,y)= (1,1)$
. Indeed, assuming
$x \geq 2$
and reducing the equation modulo
$4$
leads to a contradiction. For (2.5), it is readily verified that
$(x,y) = (2,0), (3,1)$
and
$(7,3)$
are solutions. By [Reference Scott and Styer10, Theorem 4], the equation has at most one solution in positive integers
$(x,y)$
with
$y>1$
. Hence, these constitute all solutions.
We now consider powers of
$2$
.
Lemma 2.5. Let
$k \geq 1$
. If
$f(n) = n$
for
$n \leq 2^{k-1}$
such that
$3 \nmid n$
, then
$f(2^k) = 2^{k}$
.
Proof. Since
$f(2^k) = 2^k$
for
$1 \leq k \leq 5$
, we may assume
$k \geq 6$
.
Suppose
$k = 2r+1$
, with
$r\geq 3$
. Consider
with
$2^{r-1} < 2^{r+1} - 3< 2^{r+1} + 3 < 2^{k-1}$
. Thus,
$f(2^k) = 2^k$
.
Suppose k is even. Then,
$2^{k+1} \equiv \pm 3 \pmod {5}$
.
If
$2^{k+1} \equiv 3 \pmod {5}$
, consider
By Lemma 2.4, the only nonnegative solutions of
$2^x - 3 = 5^y$
are
$(x,y) \in ~\{ (2,0), (3,1), (7,3) \}$
. In addition,
$q \nmid 2^{k+1} - 3$
for
$q=2,3$
. We write
$2^{k+1} - 3 = 5^u d$
with
$u \geq 1$
,
$d \geq 7$
and
$5 \nmid d$
. Then,
where
Thus,
$f(2^k) = 2^k$
.
If
$2^{k+1} \equiv -3 \pmod {5}$
, a similar argument using
and the fact that
$2^x + 3 = 5^y$
has the unique solution
$(x,y)=(1,1)$
yields
$f(2^k) = 2^k$
.
Lemma 2.6. Let
$p \geq 7$
be prime and
$k \geq 2$
. If
$f(n) = n$
for
$n < p^{k}$
with
$n \not \equiv \pm 3 \pmod {9}$
, then
$f(p^k) = p^{k}$
.
Proof. Suppose
$k=2r$
, with
$r \geq 1$
. Consider
with
$p^r, 4p^r \pm 3 < p^{r+1} < p^k$
. Hence,
$f(p^k) = p^k$
.
Suppose k is odd, so
$p^k \equiv \pm 1, \pm 3 \pmod {10}$
.
Assume that
$p^k =10m+1$
for some positive integer m. Since we already know that
$f(q) = q$
for
$q = 11, 31$
, we may assume that
$m \geq 4$
. Consider
with
$m+1, 4m+1, 9m < 10m+1 = p^k$
. If
$m \equiv 2 \pmod {3}$
, then
$p^k = 10m+1 \equiv 0 \pmod {3}$
, which is contrary to our assumption on p. Hence,
$m \not \equiv 2 \pmod {3}$
and
Thus,
$f(p^k) = p^k$
.
By a similar argument using the equation
it can be shown that if
$p^k = 10m-1$
, then
$f(p^k) = p^k$
.
Next, assume that
$p^k = 10m+3$
for some positive integer m. Since we have
$f(q)=~q$
for
$q=13,23$
, we may assume that
$m \geq 3$
. Consider
where
$m, 4m, 4m+1, 4m+3, 9m < 10m+3 = p^k$
. Since
$3 \nmid m$
,
so
$f(p^k) = p^k$
.
By a similar argument using the equation
it can be shown that if
$p^k = 10m-3$
, then
$f(p^k) = p^k$
. This completes the proof of the lemma and of the main theorem.