1. Introduction
The Jacobsthal sequence
is defined by
and the first terms are given by 0, 1, 1, 3, 5, 11, 21, 43, 85, 171, 341, 683, 1365, 2731, 5461, 10,923, 21,845, 43,691, …([
1] and ([
2], A001045)). Jacobsthal [
3] may never have seen the actual values of this sequence. However, Horadam [
4] uses the name “Jacobsthal sequence”; such an important sequence needs a name. Many properties and identities can be found in ([
1], Ch.44).
In this paper, for fixed integers
, we consider the linear Diophantine equation of the form
For example, when
and
, the equation
has three different non-negative solutions (representations):
,
and
. However,
is not the largest integer that has three solutions, but
is the largest. This implies that for all
, the linear equation
has at least four solutions. We are interested in an explicit form of such a largest integer
N.
More generally, for positive integers
with
and
N, we shall consider the linear equation
For a given non-negative integer
p, the largest integer
N, such that the linear Equation (
2) has at most
p different non-negative solutions (representations), is called the
p-Frobenius number and denoted by
[
5,
6,
7]. When
, 0-Frobenius number is called the Frobenius number, and such a problem is called the
linear Diophantine problem of Frobenius. This problem, also known as the stamp exchange problem or the chicken McNuggets problem, has long fascinated many people in various ways. Much research has been done on this issue from various perspectives, but what we are interested in is the expression of closed formulas. Indeed, when
, for two variables, Sylvester obtained the explicit form
. For three (or more) variables, however, no explicit form of
has been found, but the Frobenius number cannot be given by any set of closed formulas which can be reduced to a finite set of certain polynomials [
8]. Only some special cases, explicit closed formulas have been found, including arithmetic, geometric-like, Fibonacci, Mersenne, and triangular (see [
9,
10,
11,
12,
13,
14,
15] and references therein). When
, the situation is even harder. For two variables, it is known that
. However, for three or more variables, no concrete example had been found. Most recently, we have finally succeeded in giving the
p-Frobenius number as a closed-form expression of the triangular number triplet [
5]. Results on
p-Frobenius numbers for repunits [
6] and Fibonacci triples [
16] will also be available.
In this paper, we find closed forms of
p-Frobenius number of the linear Equation (
1). In detail, we give complete expressions of
p-Frobenius numbers about
p-Frobenius numbers for
, and for general
p when
.
The Sylvester number (or genus) is as important as the Frobenius number. That is, the number of positive integers
N such that the linear Equation (
2) has at most
p non-negative integer solutions is called the
p-Sylvester number (or
p-genus), and is denoted by
. When
, 0-Sylvester number is the original Sylvester number. The concept of genus is very important in studying algebraic geometry, in particular, algebraic curves. Similarly to Frobenius numbers, Sylvester obtained the explicit form
, but for three (or more) variables, no explicit form of
has been found. If
, the problem becomes even harder too.
In this paper, we also find closed forms of
p-genus of the linear Equation (
1). In order to compute both
p-Frobenius numbers and
p-genus, we analyze the geometrical structure of
p-Apéry set. This framework is partly similar to that in Fibonacci triples [
16], but complete descriptions require delicate case-by-case comparisons.
Throughout the paper, some known identities on Jacobsthal numbers are used. Two of the most important ones are
(exercise 12 [
1], [Chapter 44]) and
(exercise 26 in [
1], [Chapter 44]).
3. Preliminaries
Without loss of generality, we assume that
. For each
, we introduce the positive integer
congruent to
i modulo
, such that the number of representations of
is bigger than or equal to
and that of
is less than or equal to
p. Note that
is defined to be 0. The set
is called the
p-Apéry set of
for a non-negative integer
p (see, [
17]), which is congruent, modulo
, to the set
When , 0-Apéry set is the original Apéry set.
If one can know the structure of the elements of
p-Apéry set, we can obtain the
p-Frobenius numbers and
p-gaps (
p-Sylvester numbers) by the following formulas. A more general form can be seen in [
7].
Lemma 1. Let k and p be integers with and .
Assume that . We have By using GAP package
numericalsgps [
18] (the authors thank Professor Pedro A. García-Sánchez for his kind instruction),
p-Frobenius numbers of Jacobsthal triples can be obtained systematically.
gap> jacob:= function(n)
> if n=0 then
> return 0;
> elif n in [1, 2] then
> return 1;
> else
> return jacob(n-1) + 2*jacob(n-2);
> fi;
> end;
function( n ) … end
gap> jacob(15);
10923
gap> Set([3..10],i->FrobeniusNumber(DenumerantIdeal
(NumericalSemigroup (jacob(i),jacob(i+2),jacob(2*i-1)),3)));
[118, 145, 718, 2960, 12430, 50064, 202638, 811920 ]
For example,
(It is quite immediate to obtain the values until 50064. However, even by using GAP, it takes several hours to obtain 811920.)
We can have the 1-Apéry set and the representations of each element. For example, for , we can obtain the 1-Apéry set , and each element has exactly two representations. For instance, .
gap> s:=NumericalSemigroup(11,43,171);
gap> s1:=DenumerantIdeal(s,1);
gap> AperyList(s1,11);
[385, 386, 387, 344, 301, 258, 215, 513, 514, 471, 428 ]
gap> FactorizationsElementListWRTNumericalSemigroup(AperyList(s1,11)
,s);
[[[0, 1, 2 ], [35, 0, 0 ]], [[0, 5, 1 ], [4, 0, 2 ]], [[0, 9, 0 ], [4, 4, 1 ]],
[[0, 8, 0 ], [4, 3, 1 ]], [[0, 7, 0 ], [4, 2, 1 ]], [[0, 6, 0 ], [4, 1, 1 ]],
[[0, 5, 0 ], [4, 0, 1 ]], [[0, 0, 3 ], [31, 4, 0 ]], [[0, 4, 2 ], [35, 3, 0 ]],
[[0, 3, 2 ], [35, 2, 0 ]], [[0, 2, 2 ], [35, 1, 0 ]]]
4. Proof of the Main Results
Put the linear representation
Given two Jacobsthal numbers
and
, integers
r and
ℓ are determined uniquely by
Proof of Theorem 1. Since
(
), for
and
or
and
, we have each element of
belongs to
. See
Chart 1. Namely,
implies that
and
. When
, the largest possible element is
. When
, the largest possible element is
or
, and
is equivalent to
. Hence, by (
17), we obtain the desired result. □
Proof of Theorem 2. First, assume that
. Since there are three congruence relations
we can obtain the 1-Apéry set
from the 0-Apéry set
, where
. By
, we can see that each element of
has exactly two different expressions in terms of
:
Concerning the largest value in the set
, as seen in
Chart 2, there are four candidates:
Since
(
), we can know that
Therefore, it is enough to compare
and
. Since
is equivalent to
, we have the identity (
7).
The identities (
4), (5) and (6) are yielded from the case
. In this case, there are two different patterns for the arrangement of the elements of 1-Apéry set:
or
.
Case 1. When
, that is,
, all the elements of 0-Apéry set and 1-Apéry set are in a unique line, as shown in
Chart 3. There is the relation between each element in 0-Apéry set and 1-Apéry set as
. Each element in 1-Apéry set has exactly two representations because for
Hence, the largest element of 1-Apéry set is
Notice that the condition
is valid when
or
and
i is even. By (
17), we obtain the identity (
4), and the identity (5) when
i is even.
Case 2. When
, that is,
, some elements of 1-Apéry set come to the second line, as shown in
Chart 4. There are two corresponding relations between each element in 0-Apéry set and 1-Apéry set:
(
) and
(
). Each element in 1-Apéry set has exactly two representations because for
and for
Hence, concerning the largest element of 1-Apéry set, there are two possibilities: and . Since implies that , the condition is valid only when or and i is odd. When , only possibility is , yielding the identity (6). When (and i is odd), by , we obtain the identity (5) when i is odd. □
Proof of Theorem 3. When
, the elements of 2-Apéry set are determined from those of 1-Apéry set, by using the third block too (see
Chart 5). Except for the elements in the top row, each element in the 1-Apéry set corresponds to the element in the 2-Apéry set with the same residue modulo
so that it shifts up one step and moves to the right block. Only the elements in the 1-Apéry in the top row correspond to the elements of the 2-Apéry set with the same residue so that they are arranged in order so as to fill the gap at the bottom of the leftmost block (first block).
Case 1. When
, the elements of 2-Apéry set are placed below the first block, below the second block, and throughout the third block, as shown in
Chart 6. Note that when
, the third block consists of only one line with
or less elements.
Each element in the lower two lines of the first block (or just one line consisting of elements) can be represented in only three ways in terms of , and , respectively, and is congruent to an element in the top row of the second block modulo . Each element in the lower two lines of the second block (or just one line consisting of elements) can be represented in only three ways in terms of , and , respectively, and is congruent to an element in the second row from the bottom of the first block modulo . Every element of the third block has also only three ways of being represented in terms of , and , and there is a congruence relation with each element (except the first row) of the second block shifted up by one step.
Concerning the largest element of
, there are six candidates:
Since
(
),
. Hence, we obtain
Therefore, the largest value must be
or
. If
, the largest value in
is
. If
, the largest value in
is
. By (
17), we obtain the identity (
12).
Case 2. When , there are two patterns depending on or .
If then , so .
However, by
,
. So, this case does not exist. If
, by
, the only possibility is
and
i is even. As shown in
Chart 7, there are four candidates for the largest value of
:
Since
when
i is even, by
we have
By (
17), we can obtain the identity (11).
Case 3. When
, there are three patterns, depending on
,
or
. If
, we see that
. As shown in
Chart 8, the largest value of
is
. By (
17), we obtain the identity (
8).
If
, that is,
, together with
, only the possibility is
and
i is even. As shown in
Chart 9, the largest value of
is
or
. Since
for even
i, we have
. By (
17), we obtain the identity (9) when
i is even.
If
, by
, we see that
or
and
i is odd. As shown in
Chart 10, the largest value of
is
or
. When
(and
i is odd), we have
. When
, we have
. By (
17), we obtain the identity (9) when
i is odd, and the identity (10), respectively. □
Proof of Theorem 4. When , the structure will be systematically analyzed if . On the contrary, in the case where , the description is not simple at all because the delicate discussion of each case is required. Here, we outline the case of for generalization in a later section.
In
Chart 11, ⓝ denotes the area of elements in
. Here, each
, satisfying
(
), can be expressed in at least
ways but
in at most
n ways. Each element of
existing in the second block to the fourth block corresponds to each element having the same residue of
existing in the block immediately to the left thereof in a form of shifting up one step. The
elements of
existing over two rows (or one row) at the bottom of the first block correspond to the
elements with the same residue of
at the top of the third block. Therefore, since all the elements in
form a complete remainder system, so is
. We omit the details, but it can be seen that all the elements of
have exactly four ways of expressing in terms of
,
, and
. Within each region of
, the rightmost (lower right) element is the largest, so by comparing those elements, the maximum element of
can be found.
Eventually, the largest element of is from the first block, that is, or . Hence, if , then . If , then . Notice that implies that with . □
5. General p Case
Let p be a given non-negative integer. When , by repeating the same process as those in previous theorems, we can obtain an explicit formula of p-Frobenius number.
Theorem 5. Let and p be a non-negative integer. When with , we have However, when
, the situation is not simple at all. We do need the case-by-case discussion. To see this depends on how regularly the elements of the
p-Apéry set continues to appear. The elements of the 0-Apéry set represented by
Chart 2 are continually arranged regularly to those of the 1-Apéry set, the 2-Apéry set, and until the
r-Apéry set. However, when it reaches the time of the (
p)-Apéry set when
, the elements arranged at the lower left are insufficient, and the regularity shifts. That is, the positional relationship of the elements giving the largest value suddenly changes. Hereinafter, this deviation increases as
,
, ⋯, and the regularity is greatly impaired.
Chart 12 shows the arrangement of the
p-Apéry sets (
) when
. One can see that there will be a deficiency in the arrangement of the lower left for 6-Apéry set.