Seven squares from three numbers
Andrej Dujella, Matija Kazalicki, Vinko Petričević · 2026
Read Seven squares from three numbers by Andrej Dujella, Matija Kazalicki, Vinko Petričević (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-08729.txt.
Introduction
Dujella, Kazalicki, and Petričević study triples of distinct nonzero rational numbers. For each triple, a plus 1, b plus 1, c plus 1, ab plus 1, ac plus 1, bc plus 1, and abc plus 1 are all perfect squares. They prove infinitely many such triples exist. In contrast, they show no triple of positive integers has the property, drawing a sharp line between rational and integer solutions for this seven squares problem.
Catalog record
| CREATOR | Andrej Dujella, Matija Kazalicki, Vinko Petričević |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 10330 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.08729
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Seven squares from three numbers Andrej Dujella, Matija Kazalicki, Vinko Petričević Abstract . We study triples { a , b , c } \{a,b,c\} of distinct nonzero rational numbers such that a + 1 , b + 1 , c + 1 , a b + 1 , a c + 1 , b c + 1 , a b c + 1 a+1,\quad b+1,\quad c+1,\quad ab+1,\quad ac+1,\quad bc+1,\quad abc+1 are all perfect squares. We prove that there exist infinitely many such triples. In contrast, we show that no triple of positive integers has this property. 2020 Mathematics Subject Classification Primary 11D09; Secondary 11G05 1. Introduction A Diophantine m -tuple is a set of m distinct positive integers with the property that the product of any two distinct elements, increased by 1 , is a perfect square. The first known example is Fermat’s quadruple \{1,3,8,120\}. More generally, one may consider rational Diophantine m -tuples, defined in the same way for sets of nonzero rational numbers. An example already known to Diophantus is \left\{\frac{1}{16},\,\frac{33}{16},\,\frac{17}{4},\,\frac{105}{16}\right\}. Euler proved that there exist infinitely many rational Diophantine quintuples [ 15 ] ; in particular, he extended Fermat’s quadruple to \left\{1,3,8,120,\frac{777480}{8288641}\right\}. Stoll later proved that this extension is unique [ 17 ] . In the integral case, Baker and Davenport proved that if \{1,3,8,d\} is a Diophantine quadruple, then d=120 [ 1 ] . This led to the conjecture that there are no Diophantine quintuples in integers, which was recently proved by He, Togbé, and Ziegler [ 14 ] ; see also [ 4 ] . For rational Diophantine tuples, the maximal size is still unknown. Gibbs found the first rational Diophantine sextuple [ 13 ] , \left\{\frac{11}{192},\,\frac{35}{192},\,\frac{155}{27},\,\frac{512}{27},\,\frac{1235}{48},\,\frac{180873}{16}\right\}. Since then, several constructions of infinitely many rational Diophantine sextuples have been found. In particular, Dujella, Kazalicki, Mikić, and Szikszai [ 8 ] , Dujella and Kazalicki [ 7 ] , and Dujella, Kazalicki, and Petričević [ 10 , 9 ] constructed different infinite families of rational Diophantine sextuples, with [ 7 ] also building on ideas of Piezas [ 16 ] . No example of a rational Diophantine septuple is currently known. Lang’s conjecture implies that the size of a rational Diophantine tuple is bounded by an absolute constant; see the introduction of [ 8 ] . For more background, see [ 5 , 6 ] . A closely related problem was considered by Dujella and Szalay [ 12 ] . They studied triples of positive integers a,b,c>1 such that ab+1,\quad ac+1,\quad bc+1,\quad abc+1 are all perfect squares, and proved that there exist infinitely many such triples. This problem, attributed to John Gowland, appeared in the sci.math newsgroup in 1998 and was later featured in Brown’s book [ 2 ] . Their result also resolves a conjecture from Kenta Takahashi’s 2010 thesis, which states that there exist infinitely many integer quadruples with the property that the product of any three of their elements, increased by one, is a perfect square. In this paper, we study triples \{a,b,c\} of distinct nonzero rational numbers such that (1) a+1,\quad b+1,\quad c+1,\quad ab+1,\quad ac+1,\quad bc+1,\quad abc+1 are all perfect squares. We call such triples exotic Diophantine triples. Equivalently, \{1,a,b,c\} is a rational Diophantine quadruple for which abc+1 is also a perfect square. Our main results are the following. Theorem 1 . There exist infinitely many exotic rational Diophantine triples. For integers, the situation is different. Theorem 2 . There are no exotic Diophantine triples in positive integers. 2. Experiment and proof of Theorem 1 We started our investigation with a numerical search of exotic Diophantine triples \{a,b,c\} . The following two examples turned out to be particularly relevant: \{8,312/529,495/529\},\{312/529,-152880/165649,-78374557/87628321\}. To understand these examples, we introduce the concept of regularity (see [ 9 , 11 ] ). Definition 1 . The triple (a,b,c)\in\mathbb{Q}^{3} is called regular if r_{3}(a,b,c)=0 where r_{3}(a,b,c)=(a+b-c)^{2}-4(ab+1). Similarly, the quadruple (a,b,c,d)\in\mathbb{Q}^{4} is called regular if r_{4}(a,b,c,d)=0 where r_{4}(a,b,c,d)=(a+b-c-d)^{2}-4(ab+1)(cd+1). Note that polynomials r_{3} and r_{4} are symmetric. Our first observation was that for almost half of the exotic triples we found (for 24 out of 56 ), the quadruple \{1,a,b,c\} is a regular Diophantine quadruple, so we decided to focus on this case. The next difficulty was to find an elegant algebraic formulation of the condition that abc+1 is a perfect square. The key observation is that, for every exotic triple, the set \{1,ab,c\} is a Diophantine triple. Moreover, if this triple is regular, as in the two examples above, then r_{3}(1,ab,c)=0 , giving an algebraic condition that forces abc+1 to be a perfect square. Conversely, the symmetry of the polynomial r_{3} implies that whenever a triple \{a,b,c\} satisfies r_{3}(1,ab,c)=0 , the quantities ab+1 , abc+1 , and c+1 are automatically perfect squares. It is also easy to see from the symmetries of the polynomial r_{4} that if the same triple of distinct nonzero rationals satisfies r_{4}(1,a,b,c)=0 , then the only conditions missing for \{a,b,c\} to be an exotic Diophantine triple are that a+1 and b+1 are perfect squares. This suggests setting a=r^{2}-1,\qquad b=s^{2}-1, and hence studying the system of equations r_{3}(1,(r^{2}-1)(s^{2}-1),c)=0,\qquad r_{4}(1,r^{2}-1,s^{2}-1,c)=0. The difference of these two polynomials factors as (rs-1)(rs+1)\bigl(r^{2}s^{2}-2s^{2}-2r^{2}+5+2c\bigr)=0. If (rs-1)(rs+1)=0 , then substituting r=\pm 1/s into r_{4}(1,r^{2}-1,s^{2}-1,c)=0 gives the genus one curve s^{8}+2s^{6}c-2s^{6}+s^{4}c^{2}-6s^{4}c-s^{4}+2s^{2}c-2s^{2}+1=0, which is birationally equivalent to the elliptic curve E_{1}:\ y^{2}=x^{3}-7x-6. Since the Mordell–Weil rank of E_{1} over \mathbb{Q} is zero, this case does not yield infinitely many solutions. In the second case, substituting c=\frac{-r^{2}s^{2}+2s^{2}+2r^{2}-5}{2} into r_{4}(1,r^{2}-1,s^{2}-1,c)=0 leads to the two genus one curves r^{2}s^{2}-\frac{4}{3}r^{2}-\frac{2}{3}rs-\frac{4}{3}s^{2}+\frac{7}{3}=0 and r^{2}s^{2}-\frac{4}{3}r^{2}+\frac{2}{3}rs-\frac{4}{3}s^{2}+\frac{7}{3}=0. These curves are birationally equivalent to the elliptic curve E_{2}:\ y^{2}=x^{3}-111x+450, whose Mordell–Weil group over \mathbb{Q} is generated by the point [6,0] of order two and the point [3,-12] of infinite order. Thus, every point (x,y) on E_{2} gives an exotic Diophantine triple (r^{2}-1,s^{2}-1,c) (provided the resulting triple is nondegenerate) where u=\frac{2y-6x+42}{x^{2}-6x-3}, s=\frac{x^{2}-12x+39+2y}{x^{2}-6x-3}, r=\frac{(x-3)u^{2}+7u-1}{3u^{2}+6u-1}, and c=\frac{-r^{2}s^{2}+2s^{2}+2r^{2}-5}{2}. Note that points P and P+[6,0] on E_{2}(\mathbb{Q}) give the same exotic triple. The two examples at the beginning of this section correspond to the points [7,4]=[6,0]-[3,-12] and [-6,-30]=[6,0]-2\cdot[3,-12] on E_{2}(\mathbb{Q}) . For more examples, one can check that the points \left(\frac{223}{9},-\frac{3068}{27}\right),\qquad\left(\frac{13206}{2401},\frac{285060}{117649}\right),\qquad\left(\frac{2746063}{444889},\frac{45138812}{296740963}\right) on E_{2}(\mathbb{Q}) correspond to the triples \left\{\frac{724255280}{736742449},-\frac{152880}{165649},-\frac{63009087694401}{122040649934401}\right\}, \left\{\frac{24490915482072}{12448992625969},\frac{724255280}{736742449},-\frac{4510665894525110607837}{9171701314839342058081}\right\}, and \left\{-\frac{2539564321528123032}{5054545907282329441},\frac{24490915482072}{12448992625969},\frac{14261842404349331345950974819695}{62924004727379507987985949853329}\right\}, respectively. 3. Proof of Theorem 2 Proof of Theorem 2 . Assume for contradiction that such integers exist. Since their increments by 1 are perfect squares and they are nonzero, we may order them without loss of generality as 3\leq a<b<c (the lower bound 3 follows from a+1 being a square). Let ab+1=r^{2} , ac+1=s^{2} , bc+1=t^{2} , c+1=z^{2} , and abc+1=u^{2} for positive integers r,s,t,z,u . The set \{1,a,b,c\} forms an integer Diophantine quadruple, so Lemma 14 from [ 3 ] implies that c>4ab . Define the integer M:=2r(zu-st) . We first establish the identity (zu)^{2}-(st)^{2}=(c+1)(abc+1)-(ac+1)(bc+1)=c(a-1)(b-1), which yields zu-st=\frac{c(a-1)(b-1)}{zu+st} , and consequently M=\frac{2rc(a-1)(b-1)}{zu+st} . Next, we bound zu+st . For the lower bound, zu+st>\sqrt{c\cdot abc}+\sqrt{ac\cdot bc}=2c\sqrt{ab} . For the upper bound, observe that since c>4ab\geq 12 , we have (rc)^{2}=abc^{2}+c^{2}>\max((zu)^{2},(st)^{2}) , which implies zu<rc and st<rc , yielding zu+st<2rc . Substituting these bounds into the expression for M , we obtain (a-1)(b-1)<M<\frac{2rc(a-1)(b-1)}{2c\sqrt{ab}}=\frac{r}{\sqrt{ab}}(a-1)(b-1). Since \frac{r}{\sqrt{ab}}=\sqrt{1+\frac{1}{ab}}<1+\frac{1}{2ab} , we find (a-1)(b-1)<M<\left(1+\frac{1}{2ab}\right)(a-1)(b-1)<(a-1)(b-1)+\frac{1}{2}. This places the integer M strictly between (a-1)(b-1) and (a-1)(b-1)+1 , which is a contradiction.