Extensions of D ⁡ ( 4 ) D(4) -pairs { a , k ​ a } \{a,ka\} with k ∈ { 7 , 8 , 10 , 11 , 12 , 13 } k\in\{7,8,10,11,12,13\} Marija Bliznac Trebješanin and Pavao Radić Abstract . We study the extensibility of D ⁡ ( 4 ) D(4) -pairs { a , b } \{a,b\} , where b = k ​ a b=ka and k ∈ { 7 , 8 , 10 , 11 , 12 , 13 } k\in\{7,8,10,11,12,13\} . Firstly, we show that it can be extended to a D ⁡ ( 4 ) D(4) -triple with an element c, which is a member of a family of positive integers depending on a. Then, we prove that such a triple has a unique extension to a D ⁡ ( 4 ) D(4) -quadruple. 2020 Mathematics Subject Classification: 11D09, 11B37, 11J68, 11J86 Keywords : Diophantine m -tuples, Pellian equations, Linear forms in logarithms, Reduction method. 1. Introduction Definition 1.1 . Let n\neq 0 be an integer. We call a set of m distinct positive integers a D(n) - m -tuple, or m -tuple with the property D(n) , if the product of any two of its distinct elements increased by n is a perfect square. We research the n=4 case, which has many similarities to the classical n=1 case. First author and Filipin proved in [ 5 ] the nonexistence of D(4) -quintuples. For a D(4) -triple \{a,b,c\},a10^{5} . It is straightforward to use this bound to determine the lower bound for a in an irregular D(4) -quadruple \{a,ka,c,d\} . This is presented in Table 1 . Table 1. Lower bounds for the element a k 7 8 10 11 12 13 p 4 6 3 4 8 5 a_{p}^{(k)} 24384 13860 17316 47760 10864 42837 Let \{a,b\} be a D(4) -pair. Then there exists a positive integer r such that ab+4=r^{2} . Extending this pair to a D(4) -triple with an element c means finding s,t\in\mathbb{N} such that ac+4=s^{2},\quad bc+4=t^{2}. These two equations yield a Pellian equation (2.10) at^{2}-bs^{2}=4(a-b). Its solutions (t,s) are given by (2.11) (t_{\nu}\sqrt{a}+s_{\nu}\sqrt{b})=(t_{0}\sqrt{a}+s_{0}\sqrt{b})\left(\frac{r+\sqrt{ab}}{2}\right)^{\nu},\quad\nu\geq 0, where (t_{0},s_{0}) is a fundamental solution of the equation ( 2.10 ) and \nu is a nonnegative integer. It has been proved in [ 4 , Lemma 6.1] that for b\leq 6.85a the only fundamental solutions of this equation from which we obtain the third element c , which is an integer, are (t_{0},s_{0})=(\pm 2,2) . From [ 3 ] we know that all solutions (t_{\nu},s_{\nu}) generated with (t_{0},s_{0})=(\pm 2,2) can be represented by recursively defined sequences: (2.12) \displaystyle t_{0}=\pm 2,\ t_{1}=b\pm r,\ t_{\nu+2}=rt_{\nu+1}-t_{\nu}, (2.13) \displaystyle s_{0}=2,\ s_{1}=r\pm a,\ s_{\nu+2}=rs_{\nu+1}-s_{\nu},\ \nu\geq 0. Since c=\frac{s^{2}-4}{a} , an explicit expression for the third element c in terms of a and b is given by (2.14) \displaystyle c=c_{\nu}^{\pm}=\frac{4}{ab} \displaystyle\left\{\left(\frac{\sqrt{b}\pm\sqrt{a}}{2}\right)^{2}\left(\frac{r+\sqrt{ab}}{2}\right)^{2\nu}\right. \displaystyle\left.+\left(\frac{\sqrt{b}\mp\sqrt{a}}{2}\right)^{2}\left(\frac{r-\sqrt{ab}}{2}\right)^{2\nu}-\frac{a+b}{2}\right\}, where \nu\geq 0 is an integer. The first few elements of this sequence are: \displaystyle c_{1}^{\pm} \displaystyle=a+b\pm 2r, \displaystyle c_{2}^{\pm} \displaystyle=(ab+4)(a+b\pm 2r)\mp 4r, \displaystyle c_{3}^{\pm} \displaystyle=(a^{2}b^{2}+6ab+9)(a+b\pm 2r)\mp 4r(ab+3), \displaystyle c_{4}^{\pm} \displaystyle=(a^{3}b^{3}+8a^{2}b^{2}+20ab+16)(a+b\pm 2r)\mp 4r(a^{2}b^{2}+5ab+6). Lemma 2.2 . Let \{a,b\} be a D(4) -pair. Then d_{+}(a,b,c_{\nu}^{\pm})=c_{\nu+1}^{\pm} and d_{-}(a,b,c_{\nu}^{\pm})=c_{\nu-1}^{\pm} . Proof. From ( 2.11 ) we get s_{\nu+1}=\frac{r}{2}s_{\nu}+\frac{a}{2}t_{\nu} . Since c_{\nu}^{\pm}=\frac{(s_{\nu})^{2}-4}{a}=\frac{(t_{\nu})^{2}-4}{b} it follows that c_{\nu+1}^{\pm}=\frac{(s_{\nu+1})^{2}-4}{a}=a+b+c_{\nu}^{\pm}+\frac{1}{2}(abc_{\nu}^{\pm}+rs_{\nu}t_{\nu})=d_{+}(a,b,c_{\nu}^{\pm}) . ∎ The main goal of this section is to improve this upper bound for b in terms of a under an additional restriction. Lemma 2.3 . Let \{a,b,c\} be a D(4) -triple and ab,c^{\prime}=b and c^{\prime}=0 are the same as in [ 3 , Lemma 1] and yield c=c_{\nu}^{\pm}. It is only left to consider the case 00 and b^{\prime}c^{\prime}+4 is a square, we consider the following cases for b^{\prime} , depending on the term c^{\prime} : c^{\prime} 1 2 3 4 5 6 12 b^{\prime} 5,12 6 4 3 1 2 1 Cases c^{\prime}\in\{7,8,9,10,11\} and c^{\prime}\geq 13 imply there are no b^{\prime}>0 that satisfy our conditions. So, we obtain that a and b extend pairs \{1,5\},\{3,4\},\{2,6\},\{1,12\} . As our assumption is that \{1,5,a,b\} is not a D(4) -quadruple, we only need to show that the three remaining cases also cannot be D(4) -quadruples when b\leq 13.92a . Let’s suppose that \{3,4,a,b\} is a D(4) -quadruple. From [ 4 , Lemma 6.1] , it follows that a=a_{\nu}^{\pm}=\frac{1}{3}\left\{\left(\frac{2\pm\sqrt{3}}{2}\right)^{2}\left(\frac{4+\sqrt{12}}{2}\right)^{2\nu}+\left(\frac{\sqrt{2}\mp\sqrt{3}}{2}\right)^{2}\left(\frac{4-\sqrt{12}}{2}\right)^{2\nu}-\frac{7}{2}\right\}, and from Lemma 2.1 it follows that b=d_{+}(3,4,a) . Lemma 2.2 implies d_{+}(3,4,a)=a_{\nu+1}^{\pm} for the same choice of \pm . Define k:=\frac{b}{a}=\frac{a_{\nu+1}^{\pm}}{a_{\nu}^{\pm}} . It is easy to see that k\leq 15.24 and that it is decreasing as \nu increases, and \lim_{\nu\rightarrow\infty}\frac{a_{\nu+1}^{\pm}}{a_{\nu}^{\pm}}=\left(\frac{4+\sqrt{12}}{2}\right)^{2}>13.92, , which gives us a contradiction to the assumption that b\leq 13.92a . We use the same approach with two other cases and arrive at the same conclusion, the only difference being in case \{1,12,a,b\} because here we cannot use [ 4 , Lemma 6.1] to show that a=a_{\nu}^{\pm} . Instead, we use bounds on the fundamental solutions of the corresponding Pellian equation from [ 6 , Theorem 10.21] . ∎ Corollary 2.4 . Let \{a,ka,c\} be a D(4) -triple, k\in\{10,11,12,13\} . Then c=c_{\nu}^{\pm} . Proof. Let’s show that \{1,5,a,ka\} is not a D(4) -quadruple for k\in\{10,11,12,13\} . If \{1,5,a,ka\} is a D(4) -quadruple, by using [ 4 , Lemma 6.1] , Lemma 2.1 , and Lemma 2.2 , we obtain a=a_{\nu}^{\pm} , ka=d_{+}(1,5,a) , and d_{+}(1,5,a)=a_{\nu+1}^{\pm} for the same choice of \pm . We divide both sides of (2.15) ka=d_{+}(1,5,a)=1+5+a+\frac{1}{2}(5a+3\sqrt{5a+4}\sqrt{a+4}) by a and, using the fact that a\geq 12 , we obtain k\leq 8 . ∎ Lemma 2.5 . i) If \{1,5,a,7a\} is a D(4) -quadruple, then a=96 . ii) If \{1,5,a,8a\} is a D(4) -quadruple, then a=12 . Proof. From ( 2.15 ), when k=7 we get a=96 and when k=8 we get a=12 . ∎ Using the theory of Pellian equations, it is easy to see that the only fundamental solutions for the corresponding Pellian equation when extending \{96,672\} are (\pm 2,2) and (\pm 26,10) , and when extending \{12,96\} are (\pm 2,2) and (\pm 10,4) . So we have two pairs of sequences for each of these pairs that extend them to triples. Since the second element in both of these pairs is less than 10^{5} , by Lemma 2.1 we know that all triples from those sequences extend only to regular quadruples. Lemma 2.3 allows us to further investigate the regularity of D(4) -quadruples \{a,ka,c,d\} , i.e., we enhance [ 1 , Theorem 1.4] . In the following Sections, we show that D(4) -triples \{a,ka,c\},\ k\in\{7,8,10, 11,12,13\} , extend only to regular D(4) -quadruples. Since we separately observed the exceptions, it remains to observe extensions of triples of the form \{a,ka,c\} , where c=c_{\nu}^{\pm} . It is easy to see that c_{4}^{-}>a^{3}b^{3} and since ka>10^{5},\ k\in\{7,8,10,11,12,13\} , we get from [ 4 , Theorem 1.6] that c\in\{c_{1}^{\pm},c_{2}^{\pm},c_{3}^{\pm}\} . Also, it is easy to see that inequalities ab hold in all our cases. 3. Extensions of triples and linear forms in three logarithms In this section, we observe a system of Pellian equations which corresponds to the extension of a D(4) -triple to a D(4) -quadruple. Then, we search for the intersection of linear recurrent sequences that describe solutions to these equations. To help us find these intersections, at the end of this section we use the theory of linear forms in logarithms to obtain some useful lemmas and results. Proofs that differ only in calculations from [ 1 ] and [ 2 ] will be omitted. 3.1. System of simultaneous Pellian equations Let us observe an extension of a D(4) -triple \{a,b,c\} to a D(4) -quadruple \{a,b,c,d\} . We need to find x,y,z\in\mathbb{N} such that ad+4=x^{2},\quad bd+4=y^{2},\quad cd+4=z^{2}. By eliminating d from these equations, we obtain a system of generalized Pellian equations (3.1) \displaystyle az^{2}-cx^{2} \displaystyle=4(a-c), (3.2) \displaystyle bz^{2}-cy^{2} \displaystyle=4(b-c), (3.3) \displaystyle ay^{2}-bx^{2} \displaystyle=4(a-b). Its solutions (z,x) , (z,y) , and (y,x) satisfy (3.4) \displaystyle z\sqrt{a}+x\sqrt{c} \displaystyle=(z_{0}\sqrt{a}+x_{0}\sqrt{c})\left(\frac{s+\sqrt{ac}}{2}\right)^{m}, (3.5) \displaystyle z\sqrt{b}+y\sqrt{c} \displaystyle=(z_{1}\sqrt{a}+y_{1}\sqrt{c})\left(\frac{t+\sqrt{bc}}{2}\right)^{n}, (3.6) \displaystyle y\sqrt{a}+x\sqrt{b} \displaystyle=(y_{2}\sqrt{a}+x_{2}\sqrt{b})\left(\frac{r+\sqrt{ab}}{2}\right)^{l}, where m,n,l are nonnegative integers and (z_{0},x_{0}) , (z_{1},y_{1}) , and (y_{2},x_{2}) are fundamental solutions of ( 3.1 )–( 3.3 ). Firstly, we observe the solutions of the system of equations ( 3.4 ) and ( 3.5 ) and determine the intersections z=v_{m}=w_{n} of sequences (v_{m})_{m} and (w_{n})_{n} defined by \displaystyle v_{0}=z_{0},\ v_{1}=\frac{1}{2}\left(sz_{0}+cx_{0}\right),\ v_{m+2}=sv_{m+1}-v_{m}, \displaystyle w_{0}=z_{1},\ w_{1}=\frac{1}{2}\left(tz_{1}+cy_{1}\right),\ w_{n+2}=tw_{n+1}-w_{n}. The initial terms of these sequences are described in the following theorem. Theorem 3.1 . [ 4 , Theorem 1.3] Suppose that \{a,b,c,d\} is a D(4) -quadruple with a0 . Moreover, if d>d_{+} , case ii) cannot occur. Under the assumption that some special D(4) -quadruples do not exist, we have the following lemma, which further reduces the number of cases for fundamental solutions we need to examine. Lemma 3.2 . [ 1 , Lemma 2.2] Assume that \{a,b,c,c^{\prime}\} is not a D(4) -quadruple for any c^{\prime} with 00 . Remark 3.3 . If c=c_{1}^{\pm}=a+b\pm 2r, then it is enough to observe the case v_{2m}=w_{2n} . Secondly, we observe the solutions of the system of equations ( 3.5 ) and ( 3.6 ) and determine the intersections y=A_{n}=B_{l} of sequences (A_{n})_{n} and (B_{l})_{l} defined by (3.7) \displaystyle A_{0}=y_{1},\ A_{1}=\frac{1}{2}(ty_{1}+bz_{1}),\ A_{n+2}=tA_{n+1}-A_{n}, (3.8) \displaystyle B_{0}=y_{2},\ B_{1}=\frac{1}{2}(ry_{2}+bx_{2}),\ B_{l+2}=rB_{l+1}-B_{l},\ n,l\geq 0. The initial terms of these sequences are described in the next lemma, whose proof follows [ 1 , Lemma 2.5] . Lemma 3.4 . Assume that \{a,b,c^{\prime},c\} is not a D(4) -quadruple for any c^{\prime} with 0b, and (3.15) \Lambda^{\prime}=m\log\beta-l\log\alpha+\log\gamma^{\prime}\quad\text{for}\quad c1 for l,m\geq 1 . We can rewrite the equation P_{l}=Q_{m} to get E+4\left(\frac{c-a}{c}\right)E^{-1}=F+4\left(\frac{b-a}{b}\right)F^{-1}\mathrel{\overset{b>c}{>}}F+4\left(\frac{c-a}{c}\right)F^{-1} Then (E-F)\left(EF-4\frac{c-a}{c}\right)>0 . Since EF>4\left(\frac{c-a}{c}\right) we get E>F . It follows that \Lambda^{\prime}>0 and \Lambda^{\prime}=\log(1+\frac{E-F}{F})<\frac{E-F}{F}<4\left(\frac{c-a}{c}\right)F^{-2}<4F^{-2}. Now, since we observe only the solutions in Type 1 for c=c_{1}^{-} , we have \Lambda^{\prime}<4\frac{b}{(\pm 2\sqrt{a}+2\sqrt{b})^{2}}\alpha^{-2l}<\frac{k}{(\sqrt{k}-1)^{2}}\alpha^{-2l}<2.6\alpha^{-2l}. ∎ Lemma 3.6 . 1) Assume that b=ka,\ k\in\{7,8,10,11,12,13\} and c=c_{1}^{-} . If P_{l}=Q_{m} has a solution (l,m) with l\geq 1 , then l\leq m . 2) Assume that b=ka,\ k\in\{7,8,10,11,12,13\} and c\in\{c_{1}^{+},c_{2}^{\pm}\} . If P_{l}=Q_{m} has a solution (l,m) with m\geq 3 , then m\leq l . Proof. Proof of 2) follows [ 2 , Lemma 8] and here we will demonstrate the proof of 1) . Since we proved in the previous Lemma that \Lambda^{\prime}>0 , we have \frac{m}{l}>\frac{\log\alpha}{\log\beta}-\frac{\log\gamma^{\prime}}{l\log\beta}. To prove our statement, we need to show that the right-hand side is greater than 1-\frac{1}{l} , which is equivalent to proving that \left(\frac{\alpha}{\beta}\right)^{l}>\frac{\gamma^{\prime}}{\beta}. Since ac>a^{2}>10864^{2} , we have \frac{\alpha}{\beta}=\frac{r+\sqrt{ab}}{s+\sqrt{ac}}=\frac{\sqrt{1+\frac{4}{ab}}+1}{\sqrt{1+\frac{4}{ac}}+1}\sqrt{\frac{b}{c}}>\frac{2}{2.1}\sqrt{\frac{b}{c}}. Also, \displaystyle\frac{b}{c} \displaystyle=\frac{ka}{c_{1}^{-}}=\frac{ka}{(k+1)a-2\sqrt{ka^{2}+4}}=\frac{ka^{2}\!\left(k+1+2\sqrt{k+\frac{4}{a^{2}}}\right)}{(k-1)^{2}-\frac{16}{a^{2}}} \displaystyle>\frac{k}{(k-1)^{2}}\!\left(k+1+2\sqrt{k}\right)=\frac{k}{(\sqrt{k}-1)^{2}}. Since l\geq 1 we get \left(\frac{\alpha}{\beta}\right)^{l}>1.31. On the other hand, since for c=c_{1}^{-} we only observe fundamental solutions of Type 1 we get \displaystyle\frac{\gamma^{\prime}}{\beta} \displaystyle=\frac{2\sqrt{b}(\pm 2\sqrt{a}+2\sqrt{c})}{(s+\sqrt{ac})\sqrt{c}(2\sqrt{a}+2\sqrt{b})}<\frac{\sqrt{b}(\sqrt{a}+\sqrt{c})}{c\sqrt{a}(-\sqrt{a}+\sqrt{b})} \displaystyle=\frac{\sqrt{k}(\sqrt{a}+\sqrt{c})}{c(\sqrt{k}-1)\sqrt{a}}\leq\frac{2\sqrt{k}}{(\sqrt{k}-1)a}<1. Finally, we have \left(\frac{\alpha}{\beta}\right)^{l}>1.31>1>\frac{\gamma^{\prime}}{\beta}. ∎ For any nonzero algebraic number \alpha of degree d over \mathbb{Q} whose minimal polynomial over \mathbb{Z} is a_{0}\prod_{j=1}^{d}(X-\alpha^{(j)}), we denote by h(\alpha)=\frac{1}{d}\left(\log|a_{0}|+\sum_{j=1}^{d}\log\max\left(1,\left|\alpha^{(j)}\right|\right)\right) its absolute logarithmic height. We recall the following result due to Matveev [ 11 ] . Lemma 3.7 . Denote by \alpha_{1},\ldots,\alpha_{j} algebraic numbers, not 0 or 1, by \log\alpha_{1},\ldots, \log\alpha_{j} determinations of their logarithms, by D the degree over \mathbb{Q} of the number field \mathbb{K}=\mathbb{Q}(\alpha_{1},\ldots,\alpha_{j}), and by b_{1},\ldots,b_{j} integers. Define B=\max\{|b_{1}|,\ldots,|b_{j}|\} and A_{i}=\max\{Dh(\alpha_{i}),|\log\alpha_{i}|,0.16\}\;\;(1\leq i\leq j), where h(\alpha) denotes the absolute logarithmic Weil height of \alpha. Assume that the number \Lambda=b_{1}\log\alpha_{1}+\cdots+b_{n}\log\alpha_{j} does not vanish. Then |\Lambda|\geq\exp\{-C(j,\chi)D^{2}A_{1}\cdots A_{j}\log(eD)\log(eB)\}, where \chi=1 if \mathbb{K}\subset\mathbb{R} and \chi=2 otherwise and C(j,\chi)=\min\left\{\frac{1}{\chi}\left(\frac{1}{2}ej\right)^{\chi}30^{j+3}j^{3.5},2^{6j+20}\right\}. By applying this result we obtain upper bound for l and m in terms of a . Proposition 3.8 . Assume that c\in\{c_{1}^{+},c_{2}^{\pm}\}. If P_{l}=Q_{m}, with m>1 then \displaystyle\frac{l}{\log(el)} \displaystyle<3.34\cdot 10^{13}\cdot\log^{2}(8.09c^{2}),\quad\text{with solutions of Type $1$},\; \displaystyle\frac{l}{\log(el)} \displaystyle<6.63\cdot 10^{13}\cdot\log^{2}(8.09c^{2}),\quad\text{with solutions of Type $2$}. If Q_{m}=P_{l},\ l\geq 1 with c=c_{1}^{-}, then we get \displaystyle\frac{m}{\log(em)}<13.36\cdot 10^{13}\cdot\log^{2}(21.3a),\quad\text{for}\;k=7,8,10,11,12,13. Proof. We will demonstrate the proof for the case c=c_{1}^{-} since the proof for the other cases closely follows [ 1 , Proposition 2.9] . We apply Lemma 3.8 with j=3 and \chi=1 to the linear form 3.15 and take D=4,\ b_{1}=m,\ b_{2}=-l,\ b_{3}=1,\ \alpha_{1}=\beta,\ \alpha_{2}=\alpha,\ \alpha_{3}=\gamma^{\prime}. Since l\leq m , we can take B=m . Also, we have h(\alpha_{1})=\frac{1}{2}\log\beta,\ h(\alpha_{2})=\frac{1}{2}\log\alpha. Since \gamma^{\prime}=\gamma^{-1} , then h(\gamma)=h(\gamma^{\prime}) and from [ 1 , Proposition 2.9] we have h(\gamma)<\frac{1}{4}\log\left[\frac{2^{4}r^{4}c^{4}(1+\sqrt{k})^{4}}{(c-a)^{2}}\right]. We have \displaystyle(c_{1}^{-}-a)^{2} \displaystyle=a^{2}(2\sqrt{k+\frac{4}{a^{2}}}-k)^{2}>a^{2}(2\sqrt{k}-k)^{2}, \displaystyle r \displaystyle-1.3901 \displaystyle\cdot 10^{11}\cdot 16\cdot 2\cdot\log\beta \displaystyle\cdot 2\log\alpha\cdot 6\log(21.3a)\cdot\log(4e)\cdot\log(em). From 1) of Lemma 3.5 and using l\geq 1 , it is easy to conclude that m\log\beta\begin{cases}(14-5\sqrt{7.1})a/4,&\ k=7,\\ (6\sqrt{8}-16)a/4,&\ k=8,\\ (20-6\sqrt{10.1})a/4,&\ k=10,\\ (7\sqrt{11}-22)a/4,&\ k=11,\\ (7\sqrt{12}-24)a/4,&\ k=12,\\ (26-7\sqrt{13.1})a/4,&\ k=13.\end{cases} 5. Proof of Theorem 1.2 In this section, we complete the proof of Theorem 1.2 in two subsections according to the values of c. 5.1. Proof of Theorem 1.2 for c=c_{1}^{\pm},c_{2}^{\pm} Combining ( 2.4 )-( 2.9 ) with Proposition 3.8 , Lemmas 4.2 and 4.3 we obtain the following result. Lemma 5.1 . 1) For the D(4) -triples \{a,ka,c_{1}^{-}\} with a=a_{p}^{(k)} (p\geq 1) defined in ( 2.4 )-( 2.9 ), if the equation P_{l}=Q_{m} has a solution (l,m) with l\geq 1, then p\leq E_{k} and m\leq 2.48\cdot 10^{20} where E_{k}\in\{70,110,53,65,147,81\} for k\in\{7,8,10,11,12,13\} respectively. 2) For the D(4) -triples \{a,ka,c_{1}^{+}\} with a=a_{p}^{(k)} (p\geq 1) defined in ( 2.4 )-( 2.9 ), if the equation P_{l}=Q_{m} has a solution (l,m) with m\geq 3, then p\leq E_{k}^{\prime} and l\leq 2.56\cdot 10^{20} where E_{k}^{\prime}\in\{71,111,54,65,149,82\} for k\in\{7,8,10,11,12,13\} respectively. 3) For the D(4) -triples \{a,ka,c_{2}^{\pm}\} with a=a_{p}^{(k)} (p\geq 1) defined in ( 2.4 )-( 2.9 ), if the equation P_{l}=Q_{m} has a solution (l,m) in Type 1 with m\geq 3, then p\leq E_{k}^{\prime\prime} and l\leq 2.6\cdot 10^{21} where E_{k}^{\prime\prime}\in\{74,116,56,68, 156,86\} for k\in\{7,8,10,11,12,13\} respectively. If the equation P_{l}=Q_{m} has a solution (l,m) in Type 2 with m\geq 3 then p\leq E_{k}^{\prime\prime\prime} and l\leq 3.13\cdot 10^{20} where E_{k}^{\prime\prime\prime}\in\{18,28,13,16,38,21\} for k\in\{7,8,10,11,12,13\} respectively. For the remaining cases, we will use the following lemma which is a slight modification of the original version of Baker-Davenport reduction method (see [ 7 , Lemma 5a] ). Lemma 5.2 . Assume that M is a positive integer. Let p/q be a convergent of the continued fraction expansion of \kappa such that q>6M and let \eta=\parallel\mu q\parallel-M\cdot\parallel\kappa q\parallel, where \parallel\cdot\parallel denotes the distance from the nearest integer. If \eta>0, , then there is no solution to the inequality 010^{5},\ k=7,8,10,11,12,13 . Since the relation m\leq l only holds when m\geq 3 , the final cases m\in\{0,1,2\} are observed in the same way as in [ 1 ] . We conclude that the only possible intersection P_{l}=Q_{m} (besides the trivial (l,m)=(0,0) ) is the one that corresponds to the regular extension of a triple to a quadruple. 5.2. Proof of Theorem 1.2 with c=c_{3}^{\pm} In this case, we examine the equation z=v_{m}=w_{n} using Lemma 3.2 . By [ 9 , Lemma 5] , we know that if this equation has a solution (m,n) , then n-1\leq m\leq 2n+1 . We now examine the solutions for 22, then m>0.495b^{-0.5}c^{0.5}. ii) If the equation z=v_{2m+1}=w_{2n+1} has a solution (m,n) with n>2, then m^{2}>0.0625b^{-1}c^{0.5}. Filipin proved in [ 8 ] that z=v_{m}=w_{n}, for n>2, implies (5.3) \displaystyle\dfrac{m}{\log(m+1)}<6.543\cdot 10^{15}\log^{2}c. Combining this with Lemma 5.3 , in the case of even indices, we get (5.4) \frac{2\cdot 0.495b^{-0.5}c^{0.5}}{\log(2\cdot 0.495b^{-0.5}c^{0.5}+1)}<6.543\cdot 10^{15}\log^{2}c, and in the case of odd indices, we get the inequality (5.5) \frac{2\cdot 0.0625^{0.5}b^{-0.5}c^{0.25}+1}{\log(2\cdot 0.0625^{0.5}b^{-0.5}c^{0.25}+2)}<6.543\cdot 10^{15}\log^{2}c. The solutions obtained from these inequalities are summarized in the following lemma. Lemma 5.4 . For the D(4) -triples \{a,ka,c_{3}^{\pm}\} with a=a_{p}^{(k)} (p\geq 1) defined in ( 2.4 )-( 2.9 ), if z=v_{2m}=w_{2n} has a solution (m,n) , then p\leq E_{k} and m\leq 5.2\cdot 10^{21} where E_{k}\in\{9,14,6,8,19,10\} for k\in\{7,8,10,11,12,13\} respectively. However, if z=v_{2m+1}=w_{2n+1} has a solution (m,n) , then p\leq F_{k} and m\leq 4.3\cdot 10^{22}, where F_{k}\in\{25,40,19,23,53,29\} for k\in\{7,8,10,11,12,13\} respectively. Now, for the remaining small values of p , by using Lemma 5.2 and Lemma 5.4 in the same way as in [ 1 ] we get that z=v_{m}=w_{n} implies n\leq m\leq 2 . In these small ranges, we verify that all solutions of z=v_{m}=w_{n} give the extension of a D(4) -triple \{a,b,c\} to a regular quadruple. This completes the proof of Theorem 1.2 . University of Split, Faculty of Science, Ruđera Boškovića 33, 21000 Split, Croatia Email: marbli@pmfst.hr University of Split, Faculty of Science, Ruđera Boškovića 33, 21000 Split, Croatia Email: pradic@pmfst.hr