 Research
 Open Access
 Published:
Producttype system of difference equations with a complex structure of solutions
Advances in Difference Equations volume 2017, Article number: 140 (2017)
Abstract
The solvability of the following system of difference equations
where \(a,b,c,d\in {\mathbb {Z}}\), \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}, z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\), is studied in detail by using several methods. The system has the most complex structure of solutions of all the related systems studied so far, and some of the forms of solutions appear for the first time.
Introduction
Various types of difference equations and systems have been investigated in the last twenty years [1–31]. Many of the systems studied there, such as the ones in [2, 3, 5–9, 11–13, 17–20, 22–26, 28–31], were essentially obtained as some sorts of symmetrization of scalar ones, and were studied first by Papaschinopoulos and Schinas. One of the basic problems investigated in the field is the solvability (see, e.g., [32–36] for widely known methods). For some recent results, see, e.g., [2, 10, 14, 16–24, 26–31]. Recently several papers presenting formulas for solutions to some difference equations and systems, but without mentioning any theory, have appeared. A theoretical explanation of formulas for such an equation given in our note [14] has attracted some attention. The note shows that the equation is closely related to a known solvable one, and the main idea has been used and developed later in many papers (see, for example, [2, 10, 16, 21, 26, 27] and the references therein).
During the study of some equations and systems (for example, those in [15] and [25]), we have noticed the importance of some classes of solvable ones in their investigation. Eliminating the constant addends in the equations and systems in these two papers, producttype ones were obtained. It is known that producttype equations and systems are solvable if initial values are positive. But the standard method for solving them by using the logarithm is not suitable if some of the values are not positive. The corresponding producttype system in [15] is a special case of the following one:
where \(k,l,m,s\in {\mathbb {N}}\). This observation has motivated us to investigate the solvability of producttype systems with nonpositive initial values. However, there are several obstacles in the investigation. An obvious one is that the functions \(f_{a}(z)=z^{a}\), when a is not an integer, are multivalued. The obstacle naturally suggests the choice \(a\in {\mathbb {Z}}\) in dealing with the systems. Our first paper on producttype systems [24] investigated the case \(k=m=2\), \(l=s=1\), and our original intention was to study, as usual, the longterm behavior of their solutions. Not so long after that we realized that a related system with three dependent variables was also solvable [19], and that the case \(k=m=1\), \(l=s=2\), was solvable either [28]. Bearing in mind that calculations in these papers were quite technical, we put aside the problem of investigating the longterm behavior of their solutions, and concentrated our efforts on studying the solvability of related systems. In the next phase of the investigation we realized that adding the constant multipliers kept the solvability of such a system [18], which was also verified for an extension of the system in [24], in our paper [29]. Our further investigation showed that the solutions of solvable producttype systems can have several different forms for different values of parameters \(a, b, c, d\), and that the forms can be obtained by a detailed analysis, which we did for the first time in [23] and [31]. Later we investigated another system in [30] where such analysis was unnecessary. A technical difficulty in studying the solvability problem led us to finding another method for getting solutions in [20]. Paper [22] was the first paper which successfully dealt with an associated polynomial to a producttype system of the fourth order in detail. We would also like to say that the maxtype system in [17] was also solved by using producttype systems, as well as some equations in [21].
An important fact connected to the investigation is that there are only several solvable producttype systems with two dependent variables due to the finite number of combinations of the sums of two delays which cannot exceed four. Our general task is to present all solvable producttype systems with two dependent variables.
Here we show that the following system
where \(a,b,c,d\in {\mathbb {Z}}\), \(\alpha ,\beta \in {\mathbb {C}}\) and \(w_{2}, w_{1}, w_{0}, z_{2}, z_{1}, z_{0}\in {\mathbb {C}}\), is solvable, complementing our previous results in [18, 20, 22–24, 28–31]. In fact, we will consider the case when \(\alpha , \beta , w_{2}, w_{1}, w_{0}, z_{2}, z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\), since otherwise trivial or not welldefined solutions to system (1) are obtained. The system has the most complex structure of solutions to all the related systems studied so far, and some of the forms of solutions appear for the first time. The complexity forced us to use and develop almost all methods and tricks that we have used so far. We have to also mention that we regard that \(\sum_{i=k}^{k1}a_{i}=0\), \(k\in {\mathbb {Z}}\).
Auxiliary results
Here we quote three lemmas which we employ in the section that follows. The first one contains what was proved in [37], formulated in a compact form.
Lemma 1
Let

(a)
If \(\Delta <0\), then two zeros of \(P_{4}\) are real and different, and two are complex conjugate;

(b)
If \(\Delta >0\), then all the zeros of \(P_{4}\) are real or none is. More precisely,
 1^{∘} :

if \(P<0\) and \(D<0\), then all four zeros of \(P_{4}\) are real and different;
 2^{∘} :

if \(P>0\) or \(D>0\), then there are two pairs of complex conjugate zeros of \(P_{4}\).

(c)
If \(\Delta =0\), then and only then \(P_{4}\) has a multiple zero. The following cases can occur:
 1^{∘} :

if \(P<0\), \(D<0\) and \(\Delta _{0}\ne0\), then two zeros of \(P_{4}\) are real and equal and two are real and simple;
 2^{∘} :

if \(D>0\) or (\(P>0\) and (\(D\ne0\) or \(Q\ne0\))), then two zeros of \(P_{4}\) are real and equal and two are complex conjugate;
 3^{∘} :

if \(\Delta _{0}=0\) and \(D\ne0\), there is a triple zero of \(P_{4}\) and one simple, all real;
 4^{∘} :

if \(D=0\), then
 4.1^{∘} :

if \(P<0\) there are two double real zeros of \(P_{4}\);
 4.2^{∘} :

if \(P>0\) and \(Q=0\), there are two double complex conjugate zeros of \(P_{4}\);
 4.3^{∘} :

if \(\Delta _{0}=0\), then all four zeros of \(P_{4}\) are real and equal to \(b/4\).
Two different proofs of the following known lemma can be found, for example, in [28] and [34].
Lemma 2
If \(\lambda _{j}\), \(j=\overline {1,k}\), are mutually different zeros of the polynomial
with \(a_{k}a_{0}\ne0\), then
for \(l=\overline {0,k2}\), and
The following lemma contains some known summation formulas which we employ in this paper (see, e.g., [34, 36]). For a general method for calculating this type of sums, see [23].
Lemma 3
Let \(i\in {\mathbb {N}}_{0}\) and
where \(z\in {\mathbb {C}}\).
Then
for every \(z\in {\mathbb {C}}\setminus\{1\}\) and \(n\in {\mathbb {N}}\).
Main results
This section formulates and proves our main results. The first two results deal with the case when \(c=0\) and one of the parameters b and d is also zero.
Theorem 1
Assume that \(a,d\in {\mathbb {Z}}\), \(b=c=0\), \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}\), \(z_{1}, z_{0}\in {\mathbb {C}}\setminus\{ 0\}\). Then the following statements are true.

(a)
If \(a\ne1\), then the solution to system (1) is given by
$$\begin{aligned} &z_{n}=\alpha ^{\frac{1a^{n}}{1a}}z_{0}^{a^{n}},\quad n\in {\mathbb {N}}, \end{aligned}$$(7)$$\begin{aligned} &w_{n}=\beta \alpha ^{d\frac{1a^{n3}}{1a}}z_{0}^{da^{n3}},\quad n \ge4. \end{aligned}$$(8) 
(b)
If \(a=1\), then the solution to system (1) is given by
$$\begin{aligned} &z_{n}=\alpha ^{n} z_{0}, \quad n\in {\mathbb {N}}, \end{aligned}$$(9)$$\begin{aligned} &w_{n}=\beta \alpha ^{d(n3)}z_{0}^{d},\quad n \ge4. \end{aligned}$$(10)
Proof
Since \(b=c=0\), we have
from which the following is obtained:
Employing (12) in the second equation in (11), we get
From (12) and (13), formulas (7)(10) easily follow. □
Theorem 2
Assume that \(a,b\in {\mathbb {Z}}\), \(c=d=0\), \(\alpha ,\beta , z_{0}, w_{0}\in {\mathbb {C}}\setminus\{ 0\}\). Then the following statements are true.

(a)
If \(a\ne1\), then the solution to system (1) is given by
$$\begin{aligned} &z_{n}=\alpha ^{\frac{1a^{n}}{1a}}\beta ^{b\frac {1a^{n1}}{1a}}z_{0}^{a^{n}}w_{0}^{ba^{n1}},\quad n\ge2, \end{aligned}$$(14)$$\begin{aligned} &w_{n}=\beta ,\quad n\in {\mathbb {N}}. \end{aligned}$$(15) 
(b)
If \(a=1\), then the solution to system (1) is given by (15) and
$$\begin{aligned} z_{n}=\alpha ^{n} \beta ^{b(n1)}z_{0}w_{0}^{b},\quad n\ge2. \end{aligned}$$(16)
Proof
Since \(c=d=0\), we have
Employing the second equation in (17) in the first one, we get
from which, along with the fact \(z_{1}=\alpha z_{0}^{a}w_{0}^{b}\), it follows that
From (18), formulas (14) and (16) easily follow. □
Theorem 3
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\), \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}\), \(z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\). Then system (1) is solvable in closed form.
Proof
The assumption \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}, z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\) along with (1) implies \(z_{n}w_{n}\ne0, n\ge2\). Hence, from (1) we have
and
which together imply
We also have
Now we follow our method presented, for example, in [28], p. 6, and [31], Theorem 3.5. Let \(\delta =\alpha ^{1c}\beta ^{b}\),
then
Further, from (24) we have
for \(n\ge3\), where
Suppose
for \(k\in {\mathbb {N}}\setminus\{1\}\) and every \(n\ge k+1\), and
Employing (24), where n is replaced by \(nk\) in (27), it follows that
for \(n\ge k+2\), where
From (25), (26), (30)(32) and the induction, we conclude that (27)(29) hold for \(k,n\in {\mathbb {N}}\) such that \(2\le k\le n1\).
Setting \(k=n1\) in (27) and employing (22), we get
for \(n\ge2\).
The equalities in (28) show that for \(k\ge5\)
The same equation is also satisfied by sequences \(b_{k}\), \(c_{k}\) and \(d_{k}\), \(k\in {\mathbb {N}}\), due to the relations \(b_{k}=a_{k+1}a_{1}a_{k}\), \(c_{k}=b_{k+1}b_{1}a_{k}\), \(d_{k}=d_{1}a_{k1}\).
The assumption \(d_{1}=bdac\ne0\), along with (29) and (34), shows that it must be
and
(see, for example, the corresponding calculations in [28]).
From (36) and since \(y_{1}=a_{0}\), we obtain
It is clear that closedform formulas for solutions to problem (34)(35) can be easily found, from which along with (37) and Lemma 3 closedform formulas for \(y_{n}\) can also be found, from which along with (33) the solvability of system (21) follows.
Further, we have
and
which together imply
We also have
Following the lines of the above method, it is proved that
for \(k, n\in {\mathbb {N}}\), \(n\ge k1\), where \(\eta=\alpha ^{d}\beta ^{1a}\), \((a_{k})_{k\in {\mathbb {N}}}\), \((b_{k})_{k\in {\mathbb {N}}}\), \((c_{k})_{k\in {\mathbb {N}}}\) and \((d_{k})_{k\in {\mathbb {N}}}\) satisfy (23) and (28), and where \((y_{k})_{k\in {\mathbb {N}}}\) satisfies (29) and (36).
From (42) with \(k=n+1\) and by using (41), we get
for \(n\in {\mathbb {N}}_{0}\).
From this and since the closedform formulas for \(a_{k}\) and \(y_{k}\) can be found as above, the solvability of (40) follows. It is easily checked that (33) and (43) present a solution to system (1). □
Corollary 1
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\), \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}\), \(z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the general solution to system (1) is given by (33) and (43), where \((a_{k})_{k\ge3}\) satisfies (34) and (35), while \((y_{k})_{k\ge3}\) is given by (36) and (37).
Now we conduct a detailed analysis of the form of sequences \(a_{k}\) and \(y_{k}\) appearing in the proof of Theorem 3. The reason why equation (34) is solvable when \(ac\ne bd\) is based on the fact that its characteristic polynomial
is of the forth degree, so, solvable by radicals. The equation \(p_{4}(\lambda )=0\) is equivalent to
Now choose parameter s so that \((as2c)^{2}=(a^{2}+4s)(s^{2}+4bd4ac)\) (see, for example, [38]), that is,
where \(p=4bd3ac\) and \(q=(bdac)a^{2}c^{2}\).
Then (45) becomes
which implies that
or
Recall that solutions to (46) are found in the following form: \(s=u+v\), by posing the condition \(uv=p/3\). Since \(u^{3}+v^{3}=q\) and \(u^{3}v^{3}=p^{3}/27\), we have that \(u^{3}\) and \(v^{3}\) are solutions to the equation \(z^{2}+qzp^{3}/27=0\), from which it follows that
which, by using the change of variables \(p=\Delta_{0}/3\) and \(q=\Delta _{1}/27\), is written in the following form:
For this s quadratic equations (48) and (49) are solved, from which it follows that the zeros of \(p_{4}\) are
where
According to Lemma 1, we see that the nature of \(\lambda _{i}\)’s, \(i=\overline {1,4}\), depends on the signs of
and
All zeros of \(p_{4}\) are different and none of them is equal to 1. By Lemma 1 we see that such a situation appears if \(\Delta\ne0\) and \(p_{4}(1)\ne0\). From (59) we see that Δ is certainly negative if \(\Delta _{0}<0\), that is, if \(3ac<4bd\). For example, if \(a=1\), \(c=2\) and \(bd=3\), then \(\Delta <0\), \(p_{4}(1)=3\ne0\) and
Since due to (60), P cannot be positive, \(p_{4}\) cannot have two pairs of complexconjugate zeros.
It is well known that in the case the general solution to (34) has the form
where \(\gamma _{i}\), \(i=\overline {1,4}\), are constants.
By Lemma 2 we also have
From this and (35), it is obtained
for \(n\ge3\).
Using (64) into (37) and applying (3) four times, we get
since \(\lambda _{i}\ne1\), \(i=\overline {1,4}\). In fact, (63) shows that (65) holds also for \(n=j\), \(j=\overline {0,3}\).
All zeros of \(p_{4}\) are different and one of them is equal to 1. Polynomial \(p_{4}\) has 1 as a zero if \(p_{4}(1)=1ac+acbd=0\), that is, if
Hence
We may assume that \(\lambda _{1}=1\). To calculate the other three zeros of \(p_{4}\), the following equation should be solved:
By using the change of variables \(\lambda =s+\frac{a1}{3}\), equation (68) is transformed in (46) with
Hence
where
ε is a complex zero of the equation \(t^{3}=1\), and p and q are given in (69).
For example, if \(a=3\) and \(c=2\), then \(bd=2\) and \(\Delta \ne0\), so by Lemma 1, \(p_{4}\) has four different zeros such that exactly one of them is equal to 1, and
Formula (64) holds with \(\lambda _{1}=1\). Further, we have
for \(n\in {\mathbb {N}}\). In fact, using (63) it is shown that (73) also holds for \(n=j\), \(j=\overline {0,3}\).
From this and by Corollary 1, we get the following result.
Corollary 2
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\), \(\Delta\ne0\), and \(\alpha , \beta , z_{2}, z_{1}, z_{0}, w_{2}\), \(w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the following statements are true.

(a)
If \((a1)(c1)\ne bd\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (64), \((y_{n})_{n\ge3}\) is given by (65), while \(\lambda _{j}\), \(j=\overline {1,4}\), are given by (52)(55).

(b)
If \((a1)(c1)=bd\) and \(43ac\ne0\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (64) with \(\lambda _{1}=1\), \((y_{n})_{n\ge 3}\) is given by (73), \(\lambda _{1}=1\), while \(\lambda _{j}\), \(j=\overline {2,4}\), are given by (70) and (71).
\(p_{4}\) has exactly one double zero which is different from 1. If \(a=4\), \(c=0\) and \(bd=27\), then
is a polynomial with exactly one double zero \(\lambda _{1,2}=3\ne1\) and two complex conjugate zeros \(\lambda _{3,4}=1\pm i\sqrt{2}\).
In such cases, that is, when \(\lambda _{1}=\lambda _{2}\), \(\lambda _{i}\ne \lambda _{j}\), \(2\le i,j\le4\), we have
where \(\gamma _{i}\), \(i=\overline {1,4}\), are constants, and the solution satisfying (35) can be obtained, for example, by letting \(\lambda _{1}\to \lambda _{2}\) in (64), which yields (see [22])
Combining (37) and (75) and using Lemma 3, we have
\(p_{4}\) has exactly one double zero equal to 1. Polynomial \(p_{3}\) has 1 as a double zero if (66) holds and
that is, \(c=43a\), which implies that
It will be exactly a double zero if \(p_{4}''(1)=126a\ne0\), that is, \(a\ne2\).
We may assume that \(\lambda _{1,2}=1\). Then from (78) it follows that
In this case we have ([22])
and
Corollary 3
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\), \(\Delta=0\) and \(\alpha , \beta , z_{2}, z_{1}, z_{0}, w_{2}\), \(w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the following statements are true.

(a)
If only one of the zeros of characteristic polynomial (44) is double, say \(\lambda _{1}=\lambda _{2}\) and \((a1)(c1)\ne bd\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (75), while \((y_{n})_{n\ge3}\) is given by (76).

(b)
If only a double zero of characteristic polynomial (44) is equal to 1, say \(\lambda _{1}=\lambda _{2}=1\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (80), \((y_{n})_{n\ge3}\) is given by (81), while \(\lambda _{3,4}\) are given by (79).
\(p_{4}\) has two pairs of different double zeros. By Lemma 1 in this case it must be \(D=0\) which is equivalent to \(64bd=48ac3a^{4}\) and \(\Delta =0\) which is equivalent to
that is,
Hence
or
If it were \(a=0\) or \(c=0\), then from (82) or (83) we would get \(a=c=0\), which would imply \(bd=0\), which is impossible. If it were \(a,c\in {\mathbb {Z}}\setminus\{0\}\), from (82) we would get \(a^{3}/c=8(1\pm2i)/5\), which is also not possible.
If \(ac\ne0\), then from (83) it follows that \(c=a^{3}/8\). Let \(\lambda =at\), then
Thus
are two double zeros of \(p_{4}\) for each \(a\ne0\).
Hence, for every \(a\ne0\), polynomial \(p_{4}\) has two pairs of equal zeros. It will have integer coefficients only if \(a=4\hat{a}\), for some \(\hat{a}\in {\mathbb {Z}}\setminus\{0\}\). Note that since \(a(1\pm\sqrt {3})/4\ne1\), for every \(a\in {\mathbb {Z}}\), 1 cannot be a double zero of \(p_{4}\).
In such cases the general solution to (34) has the following form:
where \(\gamma _{i}\), \(i=\overline {1,4}\), are constants, and the solution satisfying (35) is ([22])
from which along with (37) and by Lemma 3, it follows that
Corollary 4
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\), \(\Delta=D=0\), and \(\alpha , \beta , z_{2}, z_{1}, z_{0}\), \(w_{2}, w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the following statements are true.

(a)
If \((a1)(c1)\ne bd\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (86), \((y_{n})_{n\ge3}\) is given by (87), while \(\lambda _{2,4}\) are given by (84).

(b)
Characteristic polynomial (44) cannot have two pairs of double zeros such that one of them is equal to 1.
\(p_{4}\) has a triple zero. In this case it must be \(\Delta =\Delta _{0}=0\) which is equivalent to \(\Delta _{0}=\Delta _{1}=0\), that is, if \(bd=3ac/4\) and
If \(c=0\), then \(bd=0\) and consequently \(ac=bd\), which is impossible. If \(c=a^{3}/4\), then \(bd=3a^{4}/16\). Let \(\lambda =at\), then
Hence, for every \(a\ne0\), \(a/2\) is a triple zero of \(p_{4}\), and \(p_{4}\) cannot have a zero of the fourth order. For \(a=2\hat{a}\), \(\hat{a}\in {\mathbb {Z}}\setminus\{0\}\), polynomial \(p_{4}\) has integer coefficients, and for \(a=2\) it has a triple zero equal to 1, which could be also obtained by further analyzing the polynomial in (78).
Let \(\lambda _{j}\), \(j=\overline {1,4}\), be the zeros of polynomial \(p_{4}\). In the case, we may assume that \(\lambda _{1}=\lambda _{2}=\lambda _{3}\) and \(\lambda _{3}\ne \lambda _{4}\). Then, in the case, the general solution to (34) has the following form:
where \(c_{i}\), \(i=\overline {1,4}\), are arbitrary constants.
Since (35) must hold, the following system also holds:
where \(\lambda =\lambda _{4}/\lambda _{1}\).
By solving system (89), we get
from which along with (88) it follows that
Hence, if \(\lambda _{1}=1\), we have
for \(n\ge3\), whereas if \(\lambda _{1}\ne1\), we have
for \(n\ge3\).
Combining (37) and (90), using Lemma 3 and by some calculation, we get
(recall that \(\lambda _{4}\) cannot be equal to 1 in this case).
Combining (37) and (91), using Lemma 3 and by some calculation, we get
for \(n\in {\mathbb {N}}\).
Corollary 5
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac\ne bd\) \(\Delta=\Delta_{0}=0\), \(D\ne0\), and \(\alpha , \beta , z_{2}, z_{1}, z_{0}, w_{2}, w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\} \). Then the following statements are true.

(a)
If the triple zero of polynomial (44) is different from 1, then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (91), \((y_{n})_{n\ge3}\) is given by (93), \(\lambda _{j}=a/2\), \(j=\overline {1,3}\) and \(\lambda _{4}=a/2\).

(b)
If the triple zero of polynomial (44) is equal to 1, say \(\lambda _{1}=\lambda _{2}=\lambda _{3}=1\), which is equivalent to \(a=2\), \(c=2\) and \(bd=3\), then the general solution to system (1) is given by (33) and (43), where \((a_{n})_{n\ge3}\) is given by (90), \((y_{n})_{n\ge3}\) is given by (92), while \(\lambda _{4}=1\).
Now we deal with the case \(acbd=0\), \(c\ne0\).
Theorem 4
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac=bd\), \(c\ne0\) and \(\alpha ,\beta , z_{2}, z_{1}, z_{0}, w_{2}\), \(w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then system (1) is solvable in closed form.
Proof
As above, the condition \(\alpha ,\beta , w_{2}, w_{1}, w_{0}, z_{2}, z_{1}, z_{0}\in {\mathbb {C}}\setminus\{0\}\) along with (1) implies \(z_{n}w_{n}\ne0\), \(n\ge2\). Hence, (19)(21) hold, which along with the condition \(ac=bd\) yields
Let \(\delta=\alpha ^{1c}\beta ^{b}\),
Then equation (94) can be written as
Further, from (96), we get
for \(n\ge3\), where
Suppose
for \(k\in {\mathbb {N}}\setminus\{1\}\) and every \(n\ge k+1\), and
Then, by using relation (96) with \(n\to nk\) into (99), we obtain
for \(n\ge k+2\), where
From (97), (98), (102)(104) and the induction is obtained that (99)(101) hold for \(k, n\in {\mathbb {N}}\) such that \(2\le k\le n1\).
Setting \(k=n1\) in (99) and employing (22), we obtain
for \(n\ge2\).
From (100), we get
which along with \(b_{k}=a_{k+1}a_{1}a_{k}\) and \(c_{k}=c_{1}a_{k1}\) implies that \(b_{k}\) and \(c_{k}\) also satisfy the equation.
Using the condition \(c_{1}=c\ne0\) along with (101) and (106), we easily get
and
From (101), (108) and since \(y_{1}=a_{0}\), we obtain
It is clear that closedform formulas for solutions to (106) can be easily found, from which along with (109) and Lemma 3 closedform formulas for \(y_{n}\) can be also found, from which along with (105) the solvability of (94) follows.
On the other hand, we also have that (38)(40) hold, from which along with the condition \(ac=bd\) it follows that
for \(n\in {\mathbb {N}}_{0}\).
As above, it can be proved that for all \(k, n\in {\mathbb {N}}\) such that \(1\le k \le n\)
where \(\eta=\alpha ^{d}\beta ^{1a}\), \((a_{k})_{k\in {\mathbb {N}}}\), \((b_{k})_{k\in {\mathbb {N}}}\) and \((c_{k})_{k\in {\mathbb {N}}}\) satisfy (100) and (95), while \((y_{k})_{k\in {\mathbb {N}}}\) satisfies (101) and (108).
From (111) with \(k=n+1\) and by using (41), we get
for \(n\in {\mathbb {N}}_{0}\).
From this and since the closedform formulas for \((a_{k})_{k\ge2}\) and \((y_{k})_{k\ge2}\) can be found as above, the solvability of (110) follows. It is easily checked that (105) and (112) present a solution to system (1) in this case. □
Now we conduct a detailed analysis of the form of sequences \(a_{k}\) and \(y_{k}\) appearing in the proof of Theorem 4. The reason why equation (106) is solvable when \(c\ne0\) is based on the fact that its characteristic polynomial
is of the third degree, so, solvable by radicals.
By using the change of variables \(\lambda =s+\frac{a}{3}\), the equation \(p_{3}(\lambda )=0\) is transformed into the following one:
By using formula (50), we get (51) with
Hence
where ε is a complex third root of the unity, are the zeros of \(p_{3}\).
All zeros of \(p_{3}\) are different and none of them is equal to 1. In this case it must be \(\Delta\ne0\), that is, \(\Delta_{1}^{2}\ne 4\Delta_{0}^{3}\), from which it follows that \(4a^{6}\ne(2a^{3}+27c)^{2}\), which is equivalent to \(c(4a^{3}+27c)\ne0\). Hence, if \(0\ne c\ne4a^{3}/27\), then all the zeros of \(p_{3}\) are different. If, additionally, \(p_{3}(1)\ne 0\), that is, \(a+c\ne1\), then none of them is equal to 1. For example, such a situation appears if \(a=c=k\in {\mathbb {N}}\).
All zeros of \(p_{3}\) are different and one of them is equal to 1. Polynomial \(p_{3}\) will have a zero equal to 1 if \(a+c=1\), so that
Since \(p_{3}'(1)=32a\notin {\mathbb {Z}}\), for every \(a\in {\mathbb {Z}}\), it follows that 1 cannot be a double zero of \(p_{3}\). Solving the equation \(p_{3}(\lambda )=0\), we get \(\lambda _{3}=1\),
The general solution to (106) in this case has the following form:
where \(\alpha _{i}\), \(i=\overline {1,3}\), are constants, which due to \(c_{1}=c\ne0\) can be prolonged for every nonpositive index.
From Lemma 2 with \(p_{3}(t)=\prod_{j=1}^{3}(t\lambda _{j})\), we have
This along with (107) implies that
for \(n\ge2\).
From (109) and (120) and the fact \(a_{0}=1\), we have
for \(n\in {\mathbb {N}}\).
If \(\lambda _{i}\ne1\), \(i=\overline {1,3}\), then from formula (121) it follows that
for \(n\in {\mathbb {N}}\) (in fact, (122) holds for every \(n\ge2\)).
If one of the zeros is equal to one, say \(\lambda _{3}\), then \(1\ne \lambda _{1}\ne \lambda _{2}\ne1\), and we have
for \(n\in {\mathbb {N}}\) (in fact, (123) holds for every \(n\ge2\)).
Corollary 6
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac=bd\), \(c\ne0\), \(\Delta_{1}^{2}\ne4\Delta_{0}^{3}\) and \(\alpha , \beta , z_{2}\), \(z_{1}, z_{0}, w_{2}\), \(w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the following statements are true.

(a)
If \(a+c\ne1\), then the general solution to (1) is given by (105) and (112), where \((a_{n})_{n\ge2}\) is given by (120), \((y_{n})_{n\ge2}\) is given by (122), while \(\lambda _{j}\), \(j=\overline {1,3}\) are given by (116).

(b)
If one of the zeros of characteristic polynomial (113) is equal to 1, say \(\lambda _{3}\), i.e., if \(a+c=1\) and \(2a\ne3\), then the general solution to (1) is given by formulas (105) and (112), where \((a_{n})_{n\ge2}\) is given by (120) with \(\lambda _{3}=1\), \((y_{n})_{n\ge2}\) is given by (123), and \(\lambda _{1,2}\) are given by (117).
One of the zeros is double. In this case it must be \(\Delta _{1}^{2}=4\Delta_{0}^{3}\), that is, \(c(4a^{3}+27c)=0\). Since the case \(c\ne0\) is excluded, it must be \(c=4a^{3}/27\), from which it follows that
Since in this case it must be also \(p_{3}'(\lambda )=0\), it follows that \(\lambda _{1,2}=2a/3\) is a double zero and
In order that \(c\in {\mathbb {Z}}\), it is clear that it must be \(a=3\hat{a}\) for some \(\hat{a}\in {\mathbb {Z}}\). Since \(2a/3\ne1\) when \(a\in {\mathbb {Z}}\), the polynomial cannot have 1 as a double zero, and consequently cannot have 1 as a triple zero.
If \(\lambda _{1}\ne \lambda _{2}=\lambda _{3}\), then the general solution to (106) has the following form:
where \(\hat{\alpha }_{i}\in {\mathbb {R}}\), \(i=\overline {1,3}\). Since, in our case, condition (107) must be satisfied, the solution \((a_{n})_{n\ge 2}\) to (106) can be found by letting \(\lambda _{3}\to \lambda _{2}\) in (120), so that
for \(n\ge2\) (see [31]).
From (109) and (125) and the fact that \(a_{0}=1\), we have
for every \(n\in {\mathbb {N}}\).
From (126) and Lemma 3, it follows that
for \(n\in {\mathbb {N}}\) (in fact, (127) holds also for every \(n\ge2\)).
If we assume that \(\lambda _{1}=1\) and \(\lambda _{2}=\lambda _{3}\ne1\), which is possible if \(a=3\), then from (126) it follows that
for every \(n\in {\mathbb {N}}\) (in fact, (128) holds also for every \(n\ge2\)).
Corollary 7
Assume that \(a,b,c,d\in {\mathbb {Z}}\), \(ac=bd\), \(c\ne0\), \(\Delta_{1}^{2}=4\Delta_{0}^{3}\), and \(\alpha , \beta , z_{2}\), \(z_{1}, z_{0}, w_{2}\), \(w_{1}, w_{0}\in {\mathbb {C}}\setminus\{0\}\). Then the following statements are true.

(a)
If \(a+c\ne1\), then the general solution to (1) is given by (105) and (112), where \((a_{n})_{n\ge2}\) is given by (125), \((y_{n})_{n\ge2}\) is given by (127), \(\lambda _{1}=a/3\) and \(\lambda _{2,3}=2a/3\).

(b)
If only one of the zeros of polynomial (113) is equal to one, say \(\lambda _{1}\), that is, if \(a=3\) and \(c=4\), then the general solution to system (1) is given by (105) and (112), where \((a_{n})_{n\ge2}\) is given by (125) with \(\lambda _{1}=1\), \((y_{n})_{n\ge2}\) is given by (128), \(\lambda _{1}=1\) and \(\lambda _{2,3}=2\).

(c)
It is not possible that two zeros of polynomial (113) are equal to one.
Triple zero case. In this case it must be \(p_{3}(\lambda )=p_{3}'(\lambda )=p_{3}''(\lambda )=0\). From \(p_{3}''(\lambda )=0\) it is obtained that \(\lambda =a/3\). Since \(p_{3}'(\lambda )=3\lambda ^{2}2a\lambda \), we see that \(a/3\) is its root only if \(a=0\), which would imply that \(p_{3}(\lambda )=\lambda ^{3}c\), but this polynomial has a triple zero if and only if \(c=0\), which contradicts the assumption \(c\ne0\). Hence, polynomial (113) cannot have a triple zero.
References
AndruchSobilo, A, Migda, M: On the rational recursive sequence \(x_{n+1}=ax_{n1}/(b+cx_{n}x_{n1})\). Tatra Mt. Math. Publ. 43, 19 (2009)
Berg, L, Stević, S: On some systems of difference equations. Appl. Math. Comput. 218, 17131718 (2011)
Fotiades, N, Papaschinopoulos, G: On a system of difference equations with maximum. Appl. Math. Comput. 221, 684690 (2013)
Karakostas, GL: Asymptotic 2periodic difference equations with diagonally selfinvertible responses. J. Differ. Equ. Appl. 6, 329335 (2000)
Papaschinopoulos, G, Fotiades, N, Schinas, CJ: On a system of difference equations including negative exponential terms. J. Differ. Equ. Appl. 20(56), 717732 (2014)
Papaschinopoulos, G, Psarros, N, Papadopoulos, KB: On a system of m difference equations having exponential terms. Electron. J. Qual. Theory Differ. Equ. 2015, Article ID 5 (2015)
Papaschinopoulos, G, Schinas, CJ: On a system of two nonlinear difference equations. J. Math. Anal. Appl. 219(2), 415426 (1998)
Papaschinopoulos, G, Schinas, CJ: On the behavior of the solutions of a system of two nonlinear difference equations. Commun. Appl. Nonlinear Anal. 5(2), 4759 (1998)
Papaschinopoulos, G, Schinas, CJ: Invariants for systems of two nonlinear difference equations. Differ. Equ. Dyn. Syst. 7(2), 181196 (1999)
Papaschinopoulos, G, Stefanidou, G: Asymptotic behavior of the solutions of a class of rational difference equations. Int. J. Difference Equ. 5(2), 233249 (2010)
Papaschinopoulos, G, Schinas, CJ: On the dynamics of two exponential type systems of difference equations. Comput. Math. Appl. 64(7), 23262334 (2012)
Stefanidou, G, Papaschinopoulos, G, Schinas, C: On a system of max difference equations. Dyn. Contin. Discrete Impuls. Syst., Ser. A 14(6), 885903 (2007)
Stefanidou, G, Papaschinopoulos, G, Schinas, CJ: On a system of two exponential type difference equations. Commun. Appl. Nonlinear Anal. 17(2), 113 (2010)
Stević, S: More on a rational recurrence relation. Appl. Math. ENotes 4, 8085 (2004)
Stević, S: On a generalized maxtype difference equation from automatic control theory. Nonlinear Anal. TMA 72, 18411849 (2010)
Stević, S: On the difference equation \(x_{n}=x_{nk}/(b+cx_{n1}\cdots x_{nk})\). Appl. Math. Comput. 218, 62916296 (2012)
Stević, S: Solutions of a maxtype system of difference equations. Appl. Math. Comput. 218, 98259830 (2012)
Stević, S: Firstorder producttype systems of difference equations solvable in closed form. Electron. J. Differ. Equ. 2015, Article ID 308 (2015)
Stević, S: Producttype system of difference equations of secondorder solvable in closed form. Electron. J. Qual. Theory Differ. Equ. 2015, Article ID 56 (2015)
Stević, S: New solvable class of producttype systems of difference equations on the complex domain and a new method for proving the solvability. Electron. J. Qual. Theory Differ. Equ. 2016, Article ID 120 (2016)
Stević, S: Solvable subclasses of a class of nonlinear secondorder difference equations. Adv. Nonlinear Anal. 5(2), 147165 (2016)
Stević, S: Solvable producttype system of difference equations whose associated polynomial is of the fourth order. Electron. J. Qual. Theory Differ. Equ. 2017, Article ID 13 (2016)
Stević, S: Solvability of a producttype system of difference equations with six parameters. Adv. Nonlinear Anal. (2016). doi:10.1515/anona20160145
Stević, S, Alghamdi, MA, Alotaibi, A, Elsayed, EM: Solvable producttype system of difference equations of second order. Electron. J. Differ. Equ. 2015, Article ID 169 (2015)
Stević, S, Alghamdi, MA, Alotaibi, A, Shahzad, N: Boundedness character of a maxtype system of difference equations of second order. Electron. J. Qual. Theory Differ. Equ. 2014, Article ID 45 (2014)
Stević, S, Diblik, J, Iričanin, B, Šmarda, Z: On a solvable system of rational difference equations. J. Differ. Equ. Appl. 20(56), 811825 (2014)
Stević, S, Diblik, J, Iričanin, B, Šmarda, Z: Solvability of nonlinear difference equations of fourth order. Electron. J. Differ. Equ. 2014, Article ID 264 (2014)
Stević, S, Iričanin, B, Šmarda, Z: On a producttype system of difference equations of second order solvable in closed form. J. Inequal. Appl. 2015, Article ID 327 (2015)
Stević, S, Iričanin, B, Šmarda, Z: Solvability of a close to symmetric system of difference equations. Electron. J. Differ. Equ. 2016, Article ID 159 (2016)
Stević, S, Iričanin, B, Šmarda, Z: Twodimensional producttype system of difference equations solvable in closed form. Adv. Differ. Equ. 2016, Article ID 253 (2016)
Stević, S, Ranković, D: On a practically solvable producttype system of difference equations of second order. Electron. J. Qual. Theory Differ. Equ. 2016, Article ID 56 (2016)
Agarwal, RP: Difference Equations and Inequalities: Theory, Methods, and Applications, 2nd edn. Dekker, New York (2000)
Jordan, C: Calculus of Finite Differences. Chelsea Publishing Company, New York (1956)
Krechmar, VA: A Problem Book in Algebra. Mir Publishers, Moscow (1974)
Levy, H, Lessman, F: Finite Difference Equations. Dover Publications, Inc., New York (1992)
Mitrinović, DS, Kečkić, JD: Metodi Izračunavanja Konačnih Zbirova/Methods for Calculating Finite Sums. Naučna Knjiga, Beograd (1984) (in Serbian)
Rees, EL: Graphical discussion of the roots of a quartic equation. Am. Math. Mon. 29(2), 5155 (1922)
Faddeev, DK: Lektsii po Algebre/Lectures on Algebra. Nauka, Moscow (1984) (in Russian)
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The author declares that he has no competing interests.
Author’s contributions
The author has contributed solely to the writing of this paper. He read and approved the manuscript.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Stević, S. Producttype system of difference equations with a complex structure of solutions. Adv Differ Equ 2017, 140 (2017). https://doi.org/10.1186/s1366201711906
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s1366201711906
MSC
 39A20
 39A45
Keywords
 system of difference equations
 producttype system
 solvable in closed form