Vieta jumping and small norms in quadratic number fields
Franz Lemmermeyer · 2026
Read Vieta jumping and small norms in quadratic number fields by Franz Lemmermeyer (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-2601-15229.txt.
Introduction
Franz Lemmermeyer links a famous contest problem to the theory of number fields. His 2026 arXiv paper deals with Vieta jumping, a proof trick named for the way solutions leap from one to another. The trick is best known from a problem set at the 1988 olympiad.
That problem is widely seen as one of the hardest ever set at the event. Lemmermeyer explains its tie to elements of small norm in quadratic number fields. Parametrized units in those fields supply the hidden structure. The article lays the connection out step by step.
Catalog record
| CREATOR | Franz Lemmermeyer |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-15229
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Vieta jumping and small norms in quadratic number fields Franz Lemmermeyer Abstract . In this article we explain the connection between the famous problem from the IMO 1988 and elements of small norms in quadratic number fields with parametrized units. Introduction The following problem 1 1 1 See [ 3 ] . was posed at the International Mathematical Olympiad in 1988. Let a and b be positive integers such that ab+1 divides a^{2}+b^{2} . Prove that \frac{a^{2}+b^{2}}{ab+1} is a perfect square. If we set \frac{a^{2}+b^{2}}{ab+1}=k , then the statement is equivalent to the claim that if (a,b) is an integral point on the conic (1) {\mathcal{C}}_{k}:a^{2}-kab+b^{2}=k, then k is a perfect square. In this article we will give several proofs of this result, point out connections with elements of small numbers in quadratic number fields of Richaud-Degert type, and present various generalizations. It is our intention to show that Vieta jumping is a technique that may be used outside of olympiad problems. 1. Vieta Jumping and an IMO Problem Before we look at the solution of this IMO problem we determine all solutions in integers and rational numbers when the quotient is a perfect square. If \frac{a^{2}+b^{2}}{ab+1}=m^{2} , then (a,b) is an integral point on the curve (2) {\mathcal{C}}_{m^{2}}:a^{2}-m^{2}ab+b^{2}=m^{2}. For finding all rational points on {\mathcal{C}}_{m^{2}} we compute the points of intersection of {\mathcal{C}}_{m^{2}} and the lines \ell_{t}:y=t(x-m) through the point (m,0) and find (x,y) with (3) x=\frac{m(t^{2}-1)}{t^{2}-m^{2}t+1},\quad y=\frac{m(m^{2}t^{2}-2t)}{t^{2}-m^{2}t+1} as the second point of intersection (x,y) of \ell_{t} with {\mathcal{C}}_{m^{2}} . Proposition 1 . The rational points (x,y)\neq(m,0) on the conic {\mathcal{C}}_{m^{2}} are parametrized by ( 3 ); the point (m,0) corresponds to t=\infty . Finding all integral solutions of ( 2 ) is more difficult. The obvious integral solutions are (a,b)=(\pm m,0) and (0,\pm m) . The idea is constructing new integral points from known ones using “Vieta jumping”, and then reversing this procedure and showing that each integral point is generated by the obvious points above. Proposition 2 . The integral points on the conic {\mathcal{C}}_{m^{2}} are exactly the points in the sequences (4) \displaystyle\ldots,\ (-m,-m^{3}),\ (-m,0),\ \displaystyle(m,0),\ (m,m^{3}),\ (m^{5}-m,m^{3}),\ \ldots (5) \displaystyle\ldots,\ (-m^{3},-m),\ (0,-m),\ \displaystyle(0,m),\ (m^{3},m),(m^{3},m^{5}-m),\ \ldots. These points have the form (a_{n},a_{n+1}) and (a_{n+2},a_{n+1}) , where the sequence (a_{n}) satisfies the recurrence relation a_{n+2}=m^{2}a_{n+1}-a_{n} and has the property a_{n}a_{n+2}=a_{n+1}^{2}-m^{2} . 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 -2 -2 Figure 1. Vieta jumping on {\mathcal{C}}_{4}:x^{2}-4xy+y^{2}=4 . If we begin with the point (m,0) and set a=m in ( 2 ), then we obtain the quadratic equation m^{2}-m^{3}b+b^{2}=m^{2} , i.e., 0=b^{2}-m^{3}b=b(b-m^{3}) with the known solution b_{1}=0 and the second solution b_{2}=m^{3} , giving rise to the point (m,m^{3}) . If we now set b=m^{3} , we obtain a quadratic equation in a with one integral solution a_{1}=m ; the second solution is necessarily integral: Lemma 3 . If the equation a^{2}-m^{2}ab+b^{2}=m^{2} has a solution a_{1}\in{\mathbb{Z}} when considered as a quadratic equation in a , then the second solution a_{2} is also integral. Proof. Write a^{2}-m^{2}ab+b^{2}-m^{2}=(a-a_{1})(a-a_{2}) ; then a_{1}+a_{2}=m^{2} implies that a_{2} is an integer whenever a_{1} is. ∎ Continuing in this way we find an infinite sequence of integral points on the conic {\mathcal{C}}_{m} . Assume that we have an integral point (a,b) on {\mathcal{C}}_{m^{2}} . For computing a second point with the same y -coordinate b we have to solve x^{2}-m^{2}xb+b^{2}=m^{2} knowing that a^{2}-m^{2}ab+b^{2}=m^{2}. Subtracting these equations from each other we find 0=x^{2}-a^{2}-m^{2}b(x-a)=(x-a)(x+a-m^{2}b), hence the solutions of the quadratic equation are the known solution x_{1}=a as well as x_{2}=m^{2}b-a . Thus Vieta jumping produces the new solution \flat(a,b)=(m^{2}b-a,b). Similarly, if we keep the x -coordinate fixed we have to solve a^{2}-m^{2}ay+y^{2}=m^{2} , and this produces the new point \sharp(a,b)=(a,m^{2}a-b). Thus we obtain the following operations (see Fig. 2 ) for finding new integral points on {\mathcal{C}}_{k} ; points (a,b) with a<b are on the upper branch of the curve, those with a>b lie on the lower branch. Starting with the obvious points (\pm m,0) and (0,\pm m) we now can construct the two sequences of integral points ( 4 ) and ( 5 ) on {\mathcal{C}}_{m^{2}} . \sharp \flat (a,m^{2}a-b) (a,b) (m^{2}b-a,b) \sharp \flat (a,b) (a,m^{2}a-b) (m^{2}b-a,b) Figure 2. Vieta jumping on {\mathcal{C}}_{m}:a^{2}-m^{2}ab+b^{2}=m^{2} for points on the lower (left) and upper branch (right). Next we will show that each integral point on {\mathcal{C}}_{m^{2}} is among those in ( 4 ) or ( 5 ). The basic idea is using descent: Start with an arbitrary integral point (A,B) and use Vieta jumping to find an integral point close to the origin; the only such points with coordinates whose absolute values are \leq m are the points (\pm m,0) and (0,\pm m) , and reversing the procedure we see that (A,B) belongs to one of the two sequences. The tool for proving this claim is the following lemma: Lemma 4 . If (a,b) is an integral point with positive coordinates a\neq b on {\mathcal{C}}_{k} , then there is another integral point (a^{\prime},b^{\prime}) on {\mathcal{C}}_{k} with a^{\prime},b^{\prime}\geq 0 and a^{\prime}+b^{\prime}<a+b . Proof. We set (a^{\prime},b^{\prime})=\begin{cases}(a,ka-b)&\text{ if }b>a,\\ (kb-a,b)&\text{ if }b<a.\\ \end{cases} From a^{2}-kab+b^{2}=k we deduce that ka-b=\frac{a^{2}-k}{b} and kb-a=\frac{b^{2}-k}{a} . • a^{\prime}+b^{\prime}<a+b : In the first case, b^{\prime}=ka-b=\frac{a^{2}-k}{b}<b since a^{2}-k<a^{2}<b^{2} ; in the second case, a^{\prime}=kb-a=\frac{b^{2}-k}{a}<a since b^{2}-k<b^{2}<a^{2} . Thus in both cases a^{\prime}+b^{\prime}<a+b . • a^{\prime}\geq 0 , b^{\prime}\geq 0 : In fact, if b^{\prime}=ka-b<0 in the first case, then ab^{\prime}\leq-1 , hence 0={b^{\prime}}^{2}-kab^{\prime}+a^{2}-k\geq{b^{\prime}}^{2}+k+a^{2}-k=a^{2}+{b^{\prime 2}}>0, which is a contradiction; thus b^{\prime}\geq 0 . ∎ Next we show Lemma 5 . For all integers k\geq 1 , the conic {\mathcal{C}}_{k} does not contain any integral points (a,b) with a=b except when k=1 . Proof. If a=b , then the quotient k=\frac{a^{2}+b^{2}}{ab+1}=\frac{2a^{2}}{a^{2}+1} must be an integer; since \gcd(2a^{2},a^{2}+1)=\gcd(2,a^{2}+1) is either 1 or 2 , this is only possible if a=0 or a=1 . But (0,0) is a point on {\mathcal{C}}_{k} if and only if k=0 , which we have excluded, and (1,1) is a point on {\mathcal{C}}_{k} if and only if k=1 . ∎ Now we can finish the proof of Prop. 2 : Given an integral point (a,b) in the first quadrant with a>0 and b>0 we can find another integral point (a^{\prime},b^{\prime}) with a^{\prime}\geq 0 , b^{\prime}\geq 0 and a^{\prime}+b^{\prime}<a+b . We can repeat this step until we end up with an integral point (r,s) with r=0 or s=0 . But the only such points are (\pm m,0) and (0,\pm m) . First Solution of the IMO problem Assume that a^{2}+b^{2}=k(ab+1) , which means that P(a,b) is an integral point on the conic {\mathcal{C}}_{k}:x^{2}-kxy+y^{2}=k ; also assume that k is not a square. Then k\geq 2 (and as a matter of fact k\geq 3 , since k=2 implies 2=(x-y)^{2} , which is impossible in integers). We know from Lemma 4 that as long as a>0 and b>0 we can find an integral point (a^{\prime},b^{\prime}) on {\mathcal{C}}_{k} lying in the first quadrant with a^{\prime}+b^{\prime}<a+b . Applying this step sufficiently often we obtain an integral point of the form (A,0) or (0,A) ; but then k=A^{2} is a square. 2. A family of equations Our next result strengthens an observation given without proof by Milchev [ 5 ] : Theorem 6