1 Notation
Let
$\mathbb {N}$
be the set of all natural numbers and
$\mathbb {Z}$
be the set of all integers. The set
$\mathcal {P}(\mathbb {N})$
is the collection of all subsets of
$\mathbb {N}$
. For a set A,
$|A|$
denotes the cardinality of A. For any real-valued function g and any set-valued function f, we assume
$\sum _{s\in S}g(s) = 0$
and
$\bigcup _{s\in S}f(s) = \varnothing $
if
$S = \varnothing $
. If A is a set of natural numbers, then the indicator function
$1_{A} $
of the set A is defined by
$$ \begin{align*} 1_{A}: \mathbb{N} \to \{0,1\}, \quad 1_{A}(x) = \begin{cases} 1 & \mbox{for } x\in A,\\ 0 & \mbox{for } x\notin A. \end{cases} \end{align*} $$
By
$\max (A)$
, we mean the largest element of the set A. For a given real number x,
$\lfloor x \rfloor $
denotes the largest integer less than or equal to x. For given real numbers a and b,
$$ \begin{align*} (a,b):=\{x\in \mathbb{Z}: a<x<b \}, \quad & (a,b]:=\{x\in \mathbb{Z}: a<x\leq b \}, \\ [a,b):=\{x\in \mathbb{Z}: a\leq x<b \}, \quad & [a,b]:=\{x\in \mathbb{Z}: a\leq x\leq b \}. \end{align*} $$
For given subsets A and B of
$\mathbb {N}$
, and an integer u, we define
$$ \begin{align*} B+A = A+B:=\{x+ y : x & \in A,y\in B\}, \quad B+u = u+B:=\{u+y:y\in B\}, \\ & u-B:=\{u-y:y\in B\}.& \end{align*} $$
For given sets X and Y, we set
$X\setminus Y := \{x\in X: x \notin Y\}$
. The set
$X \setminus Y$
is also known as the complement of Y in X. For any subset A of
$\mathbb {N}$
, ‘the complement of the set A’ means
$\mathbb {N}\setminus A$
throughout the article.
2 Introduction
A simple Diophantine equation
$x+y = n$
always has a solution among nonnegative integers for every natural number n. However, finding solutions may not be so easy if we choose x and y from prescribed sets of natural numbers. This motivates us to look for sets A and B of nonnegative integers such that the equation
$x+y=n$
has solutions with
$x \in A$
and
$y \in B$
for all sufficiently large integers n. In general, we are interested in studying those pairs
$(A, B)$
for which
$\mathbb {N}\setminus (A+B)$
is a finite set. This sparks curiosity in the following question.
Question 2.1. Let A be a nonempty set of natural numbers. Does there exist a set of natural numbers, B, such that
$\mathbb {N}\setminus (A+B)$
is a finite set?
This leads to the following definition.
Definition 2.2. Let A be a set of natural numbers. A set B of natural numbers is an additive complement of A if the complement of
$A+B$
in the set of all natural numbers is a finite set.
Note that the set of all natural numbers is always an additive complement to every nonempty set of natural numbers. So, one can be curious to look for other special additive complements with certain structures. Examples of such additive complements could be sparse additive complements, additive complements which are the union of disjoint infinite arithmetic progressions, additive complements of a set inside the complement of that set and so on. By considering a set with zero density (see Definition 3.1) as a sparse set, Erdős proposed the conjecture that every infinite set of natural numbers has a sparse additive complement and, in 1954, Lorentz [Reference Lorentz11] proved this conjecture. Additive complements of the form of the union of disjoint infinite arithmetic progressions are described by Mohan et al. [Reference Mohan, Patil and Pandey12]. See also [Reference Chen and Fang1–Reference Leonetti and Tringali10, Reference Ruzsa13, Reference Ruzsa14] for similar results. In this article, we study the additive complement of a set in the complement of that set with the following definition.
Definition 2.3. Let A be a set of natural numbers. A set B of natural numbers is an additive complement of A in the complement of A if B is an additive complement of A and
$B\cap A=\varnothing $
.
For example, the set of all primes is a sparse additive complement of the set of all composite natural numbers. However, the existence of an additive complement of a set in the complement of that set is not necessary; for example, there is no additive complement of the set of all even natural numbers in the set of odd natural numbers. Mohan et al. [Reference Mohan, Patil and Pandey12, Theorem 2.3, page 57] gave an infinite family of sets A such that every additive complement of the set A intersects the set A.
Proposition 2.4 ([Reference Mohan, Patil and Pandey12], Mohan et al.).
Let
$\alpha $
be a real number such that
$0 \leq \alpha \leq 1$
. Then, there exists a set
$A \subseteq \mathbb {N}$
of density
$\alpha $
such that no additive complement of A is in the complement of A.
Consequently, we are interested in the following question.
Question 2.5. Is it possible to characterise sets A of natural numbers such that there exists an additive complement of the set A in the complement of the set A?
We answer this question through the following theorem which provides such a family of sparse sets of natural numbers.
Theorem 2.6 (Ratio test for the existence of an additive complement in the complement).
Let
$A = \{a_{i} \in \mathbb {N} : i\in \mathbb {N}\}$
be an infinite set of natural numbers such that
$a_{i}<a_{i+1}$
for all
$i \in \mathbb {N}$
and
$\liminf _{n\rightarrow \infty }(a_{n+1}/a_{n})>1$
. Then, there exists a set B of natural numbers with density zero, such that B is an additive complement of A in the complement of A.
The proof of the theorem is given in Section 4. As a consequence, every sufficiently large natural number can be expressed as the sum of a factorial number and a nonfactorial number, the sum of a Fibonacci number and a non-Fibonacci number, and so on.
The conclusion of the above theorem may not hold if
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n}) =1$
. Using Proposition 2.4, we can construct examples of such sets A.
Corollary 2.7. Let
$\alpha $
be a real number such that
$0 \leq \alpha \leq 1$
. Then, there exists a set
$A = \{a_{i} \in \mathbb {N}: i\in \mathbb {N}\}$
with
$a_{i}<a_{i+1}$
,
$d(A) = \alpha $
and
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n}) =1$
such that there is no additive complement in the complement of A.
3 Preliminaries
Definition 3.1. Let A be a set of natural numbers. The upper asymptotic density and lower asymptotic density, denoted by
$\overline {d}(A)$
and
$\underline {d}(A)$
, respectively, are defined by
If
$\overline {d}(A) = \underline {d}(A) $
, then we say that the density of A, denoted by
$d(A)$
, exists where
$d(A)=\overline {d}(A) = \underline {d}(A)$
.
Lemma 3.2. Let
$(x_i)_{i=1}^{\infty }$
be an increasing sequence of natural numbers. Then,
$$ \begin{align*}\liminf_{n\rightarrow \infty }\dfrac{1}{n}\sum_{i=1}^{n}\dfrac{\log x_i}{x_i}=0.\end{align*} $$
Proof. Let
$\epsilon>0$
be a real number. Since
$\lim _{n\rightarrow \infty }{\log n}/{n}=0$
by L’Hopital’s rule, there exists
$n_0\in \mathbb {N}$
such that
Therefore, for each
$t> n_0$
,
$$ \begin{align*}\dfrac{1}{t}\sum_{i=1}^{t}\dfrac{\log x_i}{x_i}\leq\dfrac{1}{t} \sum_{i\leq n_0} \dfrac{\log x_i}{x_i}+\dfrac{1}{t}\sum_{n_0<i\leq t}\dfrac{\log x_i}{x_i}<\dfrac{1}{t} \sum_{i\leq n_0} \dfrac{\log x_i}{x_i}+\frac{t-n_0}{t}\epsilon\end{align*} $$
and so
$$ \begin{align*}\lim_{t\rightarrow \infty}\dfrac{1}{t}\sum_{i=1}^{t}\dfrac{\log x_i}{x_i} \leq \epsilon.\end{align*} $$
Since
$\epsilon $
is an arbitrary positive real number,
$$ \begin{align*} \lim_{t\rightarrow \infty}\dfrac{1}{t}\sum_{i=1}^{t}\dfrac{\log x_i}{x_i}=0. \\[-44pt] \end{align*} $$
Lemma 3.3. Let
$(a_{i})^{\infty }_{i=1}$
be an increasing sequence of real numbers such that
Then, there exists
$n_{0} \in \mathbb {N}$
and a real number
$\alpha>1$
such that
${a_{n+1}}/{a_{n}}> \alpha $
for all
$n \geq n_{0}$
.
Proof. We prove the result by contradiction. If possible, assume that for each
$\alpha>1$
and for each
$k\in \mathbb {N}$
, there exists
$n(k,\alpha )>k$
such that
Let
$n_{0} = 1$
. For every
$t\in \mathbb {N}$
, define
$\alpha _{t} = 1+{1}/{t} \ \text {and} \ n_{t} = n(n_{t-1}, \alpha _{t}).$
Since
$(a_{i})^{\infty }_{i=1}$
is an increasing sequence, by (3.1), we get a strictly increasing sequence
$(n_{t})_{t=1}^{\infty }$
such that
${1 {\kern-1pt}\leq{\kern-1pt} {a_{n_{t}+1}}/{a_{n_{t}}} {\kern-1pt}\leq{\kern-1pt} 1{\kern-1pt}+{\kern-1pt} {1}/{t}}$
. This implies
$\lim _{t\rightarrow \infty }{a_{n_{t}+1}}/{a_{n_{t}}} {\kern-1pt}={\kern-1pt}1$
and so
${\liminf _{n\rightarrow \infty } {a_{n+1}}/{a_{n}} {\kern-1pt}\leq{\kern-1pt} 1}$
, which is a contradiction.
4 Proof of Theorem 2.6
The proof of Theorem 2.6 is divided into two auxiliary propositions.
Proposition 4.1. Let m and n be two natural numbers with
$n\geq 2m$
, and let
$l>1$
be a real number. Let A be an infinite set of natural numbers such that
$|A\cap [1,m]|> |A\cap (m,ln]|$
. Then,
Proof. Since
$|A\cap [1,m]|> |A\cap (m,ln]|$
, we have
$A\cap [1,m]\neq \varnothing $
. Let u be the largest element in the set
$A\cap [1,m]$
. From the hypothesis
$n\geq 2m$
, we have
${(n,ln]\subseteq u+(m,ln].}$
This gives
as
$u\in A$
. Therefore, to prove the proposition, we need to show that
where
$X:=\{x\in (m,ln]\cap A : u+x\in (n,ln] \}.$
Now, we prove (4.1). If
$X = \varnothing $
, we are done. Otherwise, let
$y\in X$
and define
Since
$|A\cap [1,m]|> |A\cap (m,ln]|$
, we have
Since
$y\in X$
and
$u=\max (A\cap [1,m])$
,
This gives
Let
$v\in U_y$
with
$v\in (m,ln]\setminus A$
. Then,
for some
$b\in A$
. This proves (4.1) and completes the proof of the proposition.
Proposition 4.2. Let
$x_1$
,
$x_2$
, m and n be natural numbers with
$m+n\leq x_2$
. Let A and B be nonempty sets of natural numbers with
$B \subseteq (x_{1}, x_{2}]$
and
$|A \cap [1,m-x_1)|> x_{2}-x_{1}-|B|+1$
. If
$(m, m+n]\subseteq \bigcup _{i\in B} (A+ i)$
, then there exists a set
$S\subseteq B $
such that
$(m, m+n]\subseteq \bigcup _{i\in S} (A+ i)$
and
where C is a constant which is independent from m, n,
$x_1$
,
$x_2$
, A and B.
The proof of this proposition is given in Section 5.
Corollary 4.3. Let q be a natural number and A be an infinite set of natural numbers such that
$|A\cap [1,q)|>|A \cap (q,4q]|+1$
. If
$(2q,4q]\subset A+ ((q,4q]\setminus A)$
, then there exists a set
$S\subset ((q,4q]\setminus A) $
such that
$$ \begin{align*} (2q,4q]\subset A+ S \quad \text{and} \quad |S|\leq C\dfrac{q\log(|A \cap [1,q)|- |A\cap (q,4q]|)}{|A \cap [1,q)|- |A\cap (q,4q]|}, \end{align*} $$
where C is a constant which is independent of q and A.
Proof. The corollary follows from Proposition 4.2 by taking
$x_1 = q, x_2 =4q, m=2q, n=2q$
and
$B = (q,4q]\setminus A$
.
Proof of Theorem 2.6.
Since
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n})>1$
, by Lemma 3.3, there exists a natural number
$n_{0}$
and a real number
$\alpha>1$
such that
Since
$\alpha>1$
, we can choose
$r\in \mathbb {N}$
such that
Let p be a positive integer such that
$p>\max \{a_{n_0}, a_{2r+1}\}$
and let q be a positive integer such that
$q\geq p$
. Then, there exist positive integers
$u_q$
and
$m_q$
such that
$u_q \geq n_0$
and
This, together with (4.4) and (4.5), gives
This implies that
$a_{u_q+r+1}\not \in A\cap (q,4q]$
and so
$|A\cap (q,4q]|=m_q\leq r$
. Therefore,
Since
$ p>a_{2r+1}$
, we have
$|A\cap [1,x)|\geq 2r+1$
for every
$x\geq p$
. This, along with inequality (4.6), implies
By Proposition 4.1,
From this expression, (4.7) and Corollary 4.3, for every
$m\geq p$
, there exists a set
${S_m \subseteq (m,4m]\setminus A }$
such that
and
Further, using (4.6) and the fact that
${\log x}/{x}$
is a decreasing function on
$(1,\infty )$
,
Since
$p>a_{2r+1}$
, we have
$|A\cap [1,m]|\geq 2r+1$
for every
$m\geq p$
. Therefore,
Define
$$ \begin{align*}B:=\bigcup_{i=2+\lfloor\log_2 (p)\rfloor}^{\infty} S_{2^i}.\end{align*} $$
By taking m as a power of
$2$
in (4.8) and (4.9), we see that
$S_{2^{l-1}}\subset (2^{l-1},2^{l+1}]\setminus A$
and
$(2^l,2^{l+1}] \subset A+ S_{2^{l-1}}$
for every positive integer
$l>1+\lfloor \log _2 (p)\rfloor $
. This proves that B is an additive complement of A in the complement of A.
Now, we claim that the density of B is zero. Observe that
$$ \begin{align*}B\cap [1,n]\subset \bigcup_{i=\gamma}^{\beta_n} S_{2^i} \quad\mbox{for all } n\in \mathbb{N},\end{align*} $$
where
$\gamma =2+\lfloor \log _2 (p)\rfloor $
and
$\beta _n=\max \{j:2^j<n\}.$
This gives
$$ \begin{align*}|B\cap [1,n]|\leq \sum_{i=\gamma}^{\beta_n} 2 C 2^i \dfrac{\log (|A\cap [1,2^i]|)}{|A\cap [1,2^i]|}\leq 4C\sum_{i=\gamma}^{\beta_n} \sum_{j=2^{i-1}+1}^{2^i} \dfrac{\log (|A\cap [1,j]|)}{|A\cap [1,j]|},\end{align*} $$
which implies that
$$ \begin{align*}|B\cap [1,n]|\leq 4C \sum_{j=p}^{n} \dfrac{\log (|A\cap [1,j]|)}{|A\cap [1,j]|}.\end{align*} $$
Thus, by Lemma 3.2,
Therefore, the density of the set B is zero.
5 Proof of Proposition 4.2
We shall use the following lemmas to prove this proposition.
Lemma 5.1. Let A be an infinite set of natural numbers and B be a nonempty finite set of natural numbers such that
$B \subseteq (a,b]$
. For every
$n\in \mathbb {N}$
,
Proof. We have
$$ \begin{align*} \sum_{i\in B} 1_{A+i}(n)&=| A \cap (n-B)| =|A\cap [n-b,n-a)|-|A\cap (n-((a,b]\setminus B))|\\ &\geq |A\cap [n-b,n-a)|-|(a,b]\setminus B|. \\[-2.9pc] \end{align*} $$
Lemma 5.2. Let A be an infinite set of natural numbers and B be a nonempty finite set of natural numbers. Let R be a nonempty finite set of natural numbers such that every translate of the form
$A+x$
, where
$x\in B$
, contains at most r elements of R. Then,
Proof. Since every translate
$A+x$
, where
$x\in B$
, contains at most r elements of R,
Therefore,
Proof of Proposition 4.2.
Define
$f: \mathbb {N}\rightarrow \mathcal {P}(\mathbb {N})$
by
Since
$(m, m+n]\subset \bigcup _{i\in B} (A+ i)$
, by induction, we can choose a sequence
$(b_i)_{i=1}^t$
of distinct elements of B such that for every
$j\in [1,t]$
,
$$ \begin{align} &\bigg|f(b_j)\setminus \bigg(\bigcup_{i=1}^{j-1}f(b_i)\bigg) \bigg|= \max\bigg\{ \bigg|f(r) \setminus \bigg(\bigcup_{i=1}^{j-1}f(b_i)\bigg)\bigg|: r\in B\setminus \bigcup_{i=1}^{j-1}\{b_i\}\bigg\}, \end{align} $$
where t is the smallest positive integer such that
$$ \begin{align*} (m,m+n] \subseteq \bigcup_{i=1}^{t} f(b_i). \end{align*} $$
Let
$S=\bigcup _{i=1}^t \{b_i\}$
. Then,
$(m,m+n] \subseteq A+S$
.
To complete the proof of the proposition, we shall now compute the cardinality of the set S. Let
$q:=|f(b_1)|$
. For every
$j \in \mathbb {N}$
, define
$$ \begin{align} T_j:=\bigg\{r\in [1,t]: \bigg|f(b_r)\setminus \bigg(\bigcup_{i=1}^{r-1}f(b_i)\bigg) \bigg|=j\bigg\} \quad \text{and}\quad K_j :=|T_j|. \end{align} $$
From (5.1),
$T_j=\varnothing $
and
$ K_j=0 $
for each integer
$j>q$
. Also, from (5.1) and (5.2), we obtain a decreasing sequence
$(n_i)_{i=0}^q$
of positive integers such that
$n_0=t$
,
$n_q=1$
, and
Therefore,
$$ \begin{align} |S|=&\bigg|\bigcup_{x=1}^{q} T_x\bigg|=\sum_{x=1}^{q}K_x \end{align} $$
and computation of
$K_x$
will help us to compute
$|S|$
.
Define
$H: [0,q]\rightarrow \mathcal {P}(\mathbb {N})$
by
$$ \begin{align*}H(x):=(m,m+n]\setminus \bigg(\bigcup_{i=x+1}^{q}\bigcup_{r\in T_i} f(b_r)\bigg).\end{align*} $$
One can easily see that
$H(x-1)\subset H(x)$
for
$x\geq 1$
. This implies that
$$ \begin{align*} |H(x)|-|H(x-1)|&= |H(x)\setminus H(x-1)|\\ &=\bigg|\bigg(\bigcup_{i=x}^{q}\bigcup_{r\in T_x} f(b_r)\bigg)\setminus \bigg(\bigcup_{i=x+1}^{q}\bigcup_{r\in T_i} f(b_r)\bigg)\bigg|\\ &=\bigg|\bigg(\bigcup_{r\in T_x} f(b_r)\bigg)\setminus \bigg(\bigcup_{i=x+1}^{q}\bigcup_{r\in T_i} f(b_r)\bigg)\bigg|\\ &=\bigg|\bigg(\bigcup_{r=n_x+1}^{n_{x-1}}f(b_r)\bigg)\setminus \bigg(\bigcup_{r=1}^{n_x} f(b_r)\bigg)\bigg|\\ &=\bigg|\bigg(\bigcup_{r=n_x+1}^{n_{x-1}}(f(b_r)\setminus \bigcup_{i=n_x+1}^{r-1}f(b_i))\bigg)\setminus \bigg(\bigcup_{r=1}^{n_x} f(b_r)\bigg)\bigg|\\ &=\bigg|\bigcup_{r=n_x+1}^{n_{x-1}}\bigg(f(b_r)\setminus \bigcup_{i=1}^{r-1}f(b_i)\bigg)\bigg|=\bigg|\bigcup_{r\in T_x}\bigg( f(b_r)\setminus \bigg(\bigcup_{i=1}^{r-1} f(b_i)\bigg)\bigg)\bigg|\\ &=\displaystyle\sum_{r\in T_x}\bigg|f(b_r)\setminus \bigg(\bigcup_{i=1}^{r-1} f(b_i)\bigg)\bigg|=x|T_x|. \end{align*} $$
Therefore,
From this equation, we can determine
$K_x$
by computing the cardinality of
$H(x)$
.
We compute the cardinality of
$H(x)$
by computing the lower and upper bounds of
$\sum _{i\in B}\sum _{t\in H(x)} 1_{A+i}(t)$
. To get the upper bound, we count the number of elements of the set
$f(u)$
inside the set
$H(x)$
for every
$x\geq 1$
and for every
$u\in B$
. Observe that for every
$u\in B$
and
$x\in [0,q]$
,
$$ \begin{align*} |f(u)\cap H(x)|&=\bigg|f(u)\setminus \bigg(\bigcup_{i=x+1}^{q}\bigcup_{b\in T_i} f(b)\bigg)\bigg| =\bigg|f(u)\setminus \bigg(\bigcup_{i=1}^{n_x} f(b_i)\bigg)\bigg|. \end{align*} $$
From (5.1), this gives
$$ \begin{align*} |f(u)\cap H(x)|\leq \bigg|f(b_{n_x+1})\setminus \bigg(\bigcup_{i=1}^{n_x} f(b_i)\bigg)\bigg|\leq x, \end{align*} $$
because
$n_x+1\in T_j$
for some
$j\in [1,x]$
. Therefore,
This, together with Lemma 5.2, gives
$$ \begin{align} \sum_{i\in B}\sum_{t\in H(x)} 1_{A+i}(t)\leq x|B| \quad\mbox{for all } x\in [0,q]. \end{align} $$
We use Lemma 5.1 to get a lower bound for
$\sum _{i\in B}\sum _{t\in H(x)} 1_{A+i}(t)$
. Since
$B \subseteq (x_{1}, x_{2}]$
, Lemma 5.1 implies that
$$ \begin{align*} \sum_{i\in B}\sum_{t\in H(x)}1_{A+i}(t)&=\sum_{t\in H(x)}\sum_{i\in B} 1_{A+i}(t)\\ &\geq \sum_{t\in H(x)}(|A \cap [t-x_2,t-x_1)|- (x_2-x_1-|B|)) \end{align*} $$
for every
$x \in [0,q]$
. Note that
$t-x_2\leq 0$
for every
$t\in H(x)$
, because
$H(x)\subset (m,m+n]$
and
$m+n\leq x_2$
. Thus, for every
$x\in [0,q]$
,
$$ \begin{align} \sum_{i\in B}\sum_{t\in H(x)}1_{A+i}(t)&\geq \sum_{t\in H(x)}(|A \cap [1,t-x_1)|- (x_2-x_1 -|B|))\nonumber\\ & \geq |H(x)| (|A \cap [1,m-x_1)|- (x_2-x_1 -|B|)). \end{align} $$
Combining (5.5) and (5.6) gives
Since
$|A \cap [1,m-x_1)|> (x_2-x_1-|B|)$
, for every
$x\in [0,q]$
, we obtain
Now, we can compute the upper bound of the cardinality of the set S. Let
$q_0$
be an arbitrary positive integer. From (5.3),
$$ \begin{align} |S|=\begin{cases} \displaystyle\sum_{x=1}^{q_0}K_x+\sum_{x=q_0+1}^{q}K_{x} & \mbox{for } q_0<q, \\ \displaystyle\sum_{x=1}^{q}K_x & \mbox{for }q_0\geq q. \end{cases} \end{align} $$
Since
$f(b_i)\subset (m,m+n]$
for each
$i\in \bigcup _{x=1}^{q} T_x$
,
$$ \begin{align*} n\geq \bigg|\bigcup_{x=q_0+1}^{q}\bigcup_{j\in T_x}\bigg(f(b_{j})\setminus \bigg(\bigcup_{i=1}^{j-1} f(b_i)\bigg)\bigg)\bigg|&=\sum_{x=q_0+1}^{q}\sum_{j\in T_x}\bigg|f(b_{j})\setminus \bigg(\bigcup_{i=1}^{j-1} f(b_i)\bigg)\bigg|\\ &=\sum_{x=q_0+1}^{q}\sum_{j\in T_x}x =\sum_{x=q_0+1}^{q}xK_x\\ &>q_0\sum_{x=q_0+1}^{q}K_x. \end{align*} $$
Therefore,
$$ \begin{align} \sum_{x=q_0+1}^{q}K_{x}<\dfrac{n}{q_0}. \end{align} $$
From (5.4),
$$ \begin{align*} \sum_{x=1}^{q'}K_{x} &= \sum_{x=1}^{q'}\dfrac{|H(x)|-|H(x-1)|}{x} = \sum_{x=1}^{q'-1}\dfrac{|H(x)|}{x(x+1)} + \dfrac{1}{q'}|H(q')| \quad\mbox{for all } q'\leq q. \end{align*} $$
From this equation and the upper bound for
$|H(x)|$
in (5.7),
$$ \begin{align} \sum_{x=1}^{q'}K_{x} \leq \frac{|B|}{(|A \cap [1,m-x_1)|- (x_2-x_1-|B|))}\bigg(\sum_{x=1}^{q'}\dfrac{1}{x}\bigg) \quad\mbox{for all } q'\leq q. \end{align} $$
By combining (5.8), (5.9) and (5.10),
$$ \begin{align*} |S|\leq \frac{|B|}{|A \cap [1,m-x_1)|- (x_2-x_1-|B|)}\bigg(\sum_{x=1}^{q_0}\dfrac{1}{x}\bigg)+\frac{n}{q_0}. \end{align*} $$
Since
$\sum _{i=1}^{j} {1}/{i}<C\log j$
for every
$j\in \mathbb {N}$
, where C is a constant independent of j, this gives
$$ \begin{align} |S|\leq \frac{C|B|\log q_0}{|A \cap [1,m-x_1)|- (x_2-x_1-|B|)}+\frac{n}{q_0}. \end{align} $$
Since
$q_0$
is an arbitrary positive integer, we choose
$$ \begin{align*} q_0=\bigg\lfloor\dfrac{|A \cap [1,m-x_1)|- (x_2-x_1-|B|)}{\log({|A \cap [1,m-x_1)|- (x_2-x_1-|B|)})}\bigg\rfloor. \end{align*} $$
Putting this value of
$q_0$
in (5.11),
6 Conclusion
The density of the set A in Theorem 2.6 is zero. So Theorem 2.6 gives a partial answer to the following problem posed in [Reference Mohan, Patil and Pandey12] when
$\alpha =0$
.
Problem 6.1. Given
$\alpha \in [0,1]$
, characterise the sets
$A\subseteq \mathbb {N}$
such that
$d(A)=\alpha $
and
$\mathbb {N}\setminus A$
is an additive complement of A.
This problem is still open even for a single
$\alpha>0$
. Theorem 2.6 does not apply to sets
$A = \{a_{i} \in \mathbb {N}: i\in \mathbb {N}~ a_{i}< a_{i+1}\}$
such that
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n}) =1$
. When
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n}) =1$
, there are examples for the nonexistence of any additive complement in the complement of the set A by Corollary 2.7. However, if we take the strictly increasing sequence
$(c_n)_{n=1}^{\infty }$
formed by all the composite numbers, then
$\liminf _{n\rightarrow \infty } (c_{n+1}/c_{n}) =1$
and the set of all prime numbers is an additive complement in the complement of the set of composite numbers. This motivates us to pose the following problem.
Problem 6.2. Let
$A = \{a_{i} \in \mathbb {N}: i\in \mathbb {N}~ a_{i}< a_{i+1}\}$
. Give a sufficient condition on such a set A with
$\liminf _{n\rightarrow \infty } (a_{n+1}/a_{n}) =1$
such that there exists an additive complement of the set A in the complement of A.
In particular, it would be interesting to find an additive complement of the set of all primes in the set of all composite numbers with density zero. We are also interested in the following problem related to the root test for additive complements.
Problem 6.3. Let
$A = \{a_{i} \in \mathbb {N} : i\in \mathbb {N}\}$
with
$a_{i}<a_{i+1}$
for all
$i \in \mathbb {N}$
. If
$\liminf _{n\rightarrow \infty } (a_{n})^{{1}/{n}}>1$
, does there exist a set B of natural numbers with density zero such that B is an additive complement of A in the complement of A.
In this article, we gave a family of sets A of natural numbers such that there exists a set B of natural numbers, which is an additive complement of A and
$A\cap B = \varnothing $
. We generalise this in the following problem.
Problem 6.4. Characterise all infinite sets A and C of natural numbers such that there exists a set B of natural numbers which is an additive complement of A and
$B\cap C = \varnothing $
.
In particular, if
$A = C$
, then by Theorem 2.6, we have a family of sets A which satisfy the conclusion of this problem.








