Rational Torsion Points

The rational points on an elliptic curve—points whose coordinates are simple fractions—harbor a deep and elegant arithmetic structure. While these points might seem randomly scattered across the curve, they obey a consistent set of rules, forming an algebraic group. However, a fundamental question arises: how can we classify the behavior of these points? How do we distinguish the points that generate an infinite sequence of new points from those that follow a finite, rhythmic pattern, always returning to their starting point? This article addresses this gap by focusing on the latter group: the rational torsion points.

This exploration is divided into two main chapters. In "Principles and Mechanisms," we will uncover the secret arithmetic of elliptic curves, known as the group law. We will define torsion points and introduce the powerful Nagell-Lutz theorem, a critical tool for identifying them, while also exploring alternative methods like reduction modulo a prime. The journey continues in "Applications and Interdisciplinary Connections," where we reveal how this seemingly abstract theory provides concrete solutions to ancient mathematical puzzles like the Congruent Number Problem, underpins modern cryptographic methods, and serves as a cornerstone for understanding one of the deepest questions in modern mathematics, the Birch and Swinnerton-Dyer conjecture.

Imagine you're standing on a vast, rolling landscape defined by an equation like $y^2 = x^3 + ax + b$. This landscape is an elliptic curve. The "points" we are interested in are not just any points, but those with rational number coordinates—the landmarks you could precisely map. Now, what if I told you these points aren't just scattered randomly? They follow a secret set of rules, an elegant arithmetic that allows them to be added and subtracted. This chapter is a journey into that secret world, a world where we uncover the principles that govern these points and the mechanisms we can use to understand their structure.

The first surprising thing about the rational points on an elliptic curve is that they form a group. What does that mean? It means there's a consistent way to "add" any two points to get a third, and this addition follows familiar rules, like associativity. This isn't your everyday addition; it's a beautiful geometric game called the ​​chord-and-tangent law​​.

Here’s how you play: Pick two rational points, $P$ and $Q$. Draw a straight line through them. Because our curve is a cubic (it involves $x^3$), this line is guaranteed to intersect the curve at exactly one other point, which we'll call $R*$. To find the sum $P+Q$, you simply reflect $R*$ across the x-axis to a point $R$. That’s it. You've added two points!

What if you want to add a point $P$ to itself? You draw the tangent line at $P$. This line touches the curve at $P$ (counting for two intersections) and will cross it at one other point, $R*$. Reflect $R*$ across the x-axis, and you have $2P$. The inverse of a point $(x,y)$ is simply its reflection $(x,-y)$, which we call $-P$.

But where is the "zero" in this system? Every group needs an identity element, a point that, when added to any other point, changes nothing. For elliptic curves, this special point is not on the affine plane we can easily visualize. We call it the ​​point at infinity​​, denoted $\mathcal{O}$. Think of it as a point infinitely far up (and down) the y-axis, where all vertical lines meet. It acts as the anchor for our entire group structure.

Now that we can add points, a fascinating question arises. What happens if we take a point $P$ and keep adding it to itself: $P, 2P, 3P, 4P, \dots$? Two things can happen.

Some points, which we call ​​points of infinite order​​, will wander across the curve forever, never repeating a location. The sequence of their multiples never returns to the starting identity element $\mathcal{O}$. For the curve $E: y^2 = x^3 - 2$, the point $(3,5)$ is one such wanderer. If you keep adding it to itself, you generate an infinite sequence of distinct points on the curve.

Other points, however, perform a rhythmic dance. After a certain number of steps, they land right back on $\mathcal{O}$. These are the ​​torsion points​​. A point $P$ is a torsion point if there exists a positive integer $n$ such that $nP = \mathcal{O}$. The smallest such $n$ is the order of the point. The set of all rational torsion points on a curve forms a finite subgroup called the ​​torsion subgroup​​, denoted $E(\mathbb{Q})_{\text{tors}}$. These points are the heart of the curve's arithmetic structure.

The simplest torsion points, besides the identity $\mathcal{O}$ itself (which has order $1$), are the points of order $2$. A point $P$ has order $2$ if $2P=\mathcal{O}$ and $P \neq \mathcal{O}$. In our group law, this means $P = -P$. Since the inverse of $(x,y)$ is $(x,-y)$, this condition is met only when $y = -y$, which forces $y=0$.

This is a wonderful moment of clarity! The non-identity points of order two are simply the points where the curve intersects the x-axis. To find them, we just set $y=0$ in the curve's equation and find the rational roots of the resulting cubic in $x$.

Let's take the curve $E: y^2=x^3-x$. Setting $y=0$ gives $x^3-x = x(x-1)(x+1) = 0$. The rational roots are $x=-1, 0, 1$. This gives us three points of order $2$: $(-1,0)$, $(0,0)$, and $(1,0)$. These three points, together with the identity $\mathcal{O}$, form the complete $2$-torsion subgroup $E[2](\mathbb{Q})$, a tidy little group of four elements isomorphic to $\mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z}$.

Finding the points of order $2$ was easy. But what about points of order $3, 4, 5, \dots, 17$? How can we possibly find them all? It seems like an infinite search. This is where a truly remarkable tool comes to our rescue: the ​​Nagell-Lutz Theorem​​.

For an elliptic curve given by $y^2 = x^3 + ax + b$ with ​​integer coefficients​​ $a$ and $b$, the theorem provides two powerful rules that slash the infinite search down to a finite, manageable checklist:

1. ​​Integrality Condition:​​ Any rational torsion point (other than $\mathcal{O}$) must have ​​integer coordinates​​. That is, if $(x,y)$ is a torsion point, both $x$ and $y$ must be integers.

2. ​​Divisibility Condition:​​ For an integer torsion point $(x,y)$, either $y=0$ (our old friends of order $2$) or $y^2$ must be a divisor of the curve's ​​discriminant​​, $\Delta = -16(4a^3 + 27b^2)$.

The discriminant $\Delta$ is just a number calculated from the curve's coefficients, $a$ and $b$. The fact that $y^2$ for any torsion point must divide this specific, fixed number is nothing short of magical. It means we don't have to search all infinite integers for $y$; we only have to check the handful of integers whose squares divide $\Delta$. This turns an impossible task into a weekend puzzle.

Let's see this wrench in action on $E: y^2 = x^3 - x$. The coefficients are integers ($a=-1, b=0$), so we can use the theorem. We calculate the discriminant $\Delta = -16(4(-1)^3 + 27(0)^2) = 64$. The theorem tells us any torsion point $(x,y)$ must have $x,y \in \mathbb{Z}$.

• Case 1: $y=0$. We already found the points $(-1,0), (0,0), (1,0)$.
• Case 2: $y \neq 0$. Here, $y^2$ must divide $\Delta=64$. The square divisors of $64$ are $1, 4, 16, 64$. So the possible integer values for $y$ are only $\{\pm 1, \pm 2, \pm 4, \pm 8\}$. We can now test each possibility. For example, if $y^2=1$, we need to find an integer $x$ such that $x^3-x=1$. A quick check shows no such integer exists. It turns out that none of these $y$-values yield an integer $x$. The conclusion? The only affine torsion points are those with $y=0$. The entire torsion subgroup is $E(\mathbb{Q})_{\text{tors}} = \{\mathcal{O}, (-1,0), (0,0), (1,0)\}$. The search is complete.

The Nagell-Lutz theorem is powerful, but it comes with fine print: the coefficients $a$ and $b$ must be integers. What happens if they are not? Consider the curve $E': Y^2 = X^3 - \frac{1}{16}X$. Its coefficient $a = -\frac{1}{16}$ is not an integer, so the theorem does not apply. And indeed, we can check that the point $P = (\frac{1}{4}, 0)$ lies on this curve. This point has order $2$ (since its $Y$-coordinate is $0$), so it is a rational torsion point. But its $X$-coordinate is $\frac{1}{4}$, which is not an integer!

This example beautifully illustrates that the integrality of torsion points is not a property of the abstract curve itself, but a property of the specific integral model (equation with integer coefficients) we choose to represent it. It's a crucial lesson: in mathematics, the assumptions behind a theorem are just as important as its conclusion.

The Nagell-Lutz theorem states that on an integral model, a rational torsion point must have integer coordinates. It's tempting to flip this around and assume that any point with integer coordinates must be a torsion point. This is a classic logical error.

To see why, let's look at the curve $E: y^2 = x^3 - 7x + 10$. This is a perfectly fine model with integer coefficients. The point $P=(5,10)$ has integer coordinates and lies on the curve, as $10^2 = 100$ and $5^3 - 7(5) + 10 = 125 - 35 + 10 = 100$. Is it a torsion point? Let's check the Nagell-Lutz condition. The discriminant is $\Delta = -21248$. If $P$ were torsion, then $y^2 = 100$ would have to divide $\Delta$. But $100$ does not divide $-21248$. Therefore, despite having integer coordinates, $(5,10)$ is a point of infinite order. It's an "integral point," but it's not a torsion point. Remember: All rational torsion points are integral (on the right model), but not all integral points are torsion.

The Nagell-Lutz theorem is a fantastic tool, but number theory is full of different paths to the same truth. Another powerful technique is to look at the curve's "shadow" in the world of finite arithmetic. This is the idea of ​​reduction modulo a prime​​.

Imagine taking the equation $y^2 = x^3 - x + 1$ and considering it not over the rational numbers, but over the finite field $\mathbb{F}_5$, the world of arithmetic modulo $5$ (where $3+3=1$). The curve's equation still makes sense, and we can count the number of points on this "shadow curve," $\tilde{E}(\mathbb{F}_5)$. Let's say we find there are $8$ points.

Here's the key: the group of rational torsion points $E(\mathbb{Q})_{\text{tors}}$ injects into the group of points on the shadow curve $\tilde{E}(\mathbb{F}_p)$ for any "good" prime $p$. This means the size of the rational torsion group must divide the number of points in its shadow.

So, if we find that $\#\tilde{E}(\mathbb{F}_5) = 8$, we know that $\#E(\mathbb{Q})_{\text{tors}}$ must be a divisor of $8$. That's useful, but we can do better. Let's cast more shadows. For the same curve, we can compute:

• $\#\tilde{E}(\mathbb{F}_7) = 12$
• $\#\tilde{E}(\mathbb{F}_{11}) = 10$
• $\#\tilde{E}(\mathbb{F}_{13}) = 19$

The size of our rational torsion group must divide $8$, and $12$, and $10$, and $19$. The only positive integer that divides all these numbers is $1$. Therefore, we can conclude with certainty that $\#E(\mathbb{Q})_{\text{tors}} = 1$. The only rational torsion point is the identity $\mathcal{O}$. This method, combining information from different prime worlds, showcases the profound unity and interconnectedness of number theory, allowing us to solve a problem about rational numbers by looking at their reflections in finite fields.

Now that we have acquainted ourselves with the principles of rational torsion points—the elegant group law, the crisp criteria for identifying these special points—we might be tempted to see them as a beautiful, but perhaps isolated, piece of mathematical machinery. Nothing could be further from the truth. The study of torsion points on elliptic curves is not an end in itself; it is a gateway. It is a powerful lens through which we can explore, and in some cases solve, problems across the vast landscape of mathematics and science, from questions posed by the ancient Greeks to the security of our modern digital world. In this chapter, we will embark on a journey to see what this machinery does, revealing the surprising and profound connections it forges.

At its most fundamental level, the theory of torsion points is a practical tool for number theorists, a "rational point sieve." When faced with a Diophantine equation in the form of an elliptic curve, the question of finding all its rational solutions can seem impossibly vast. The Nagell-Lutz theorem, however, acts as a remarkably fine sieve. It tells us that any rational point of finite order must have integer coordinates. Furthermore, it places a strict constraint on these integer coordinates: the square of the $y$-coordinate must divide the curve's discriminant.

This transforms an infinite search into a finite, manageable task. Consider the curve $E: y^2 = x^3 - 4x$. The discriminant is a tidy power of two, $\Delta = 4096$. The Nagell-Lutz theorem directs us to check for integer points where $y=0$ or $y^2$ divides $4096$. The case $y=0$ immediately yields three points: $(0,0)$, $(2,0)$, and $(-2,0)$. A careful check reveals that no other integer points exist on the curve. These three, along with the point at infinity, form the complete torsion subgroup, a group isomorphic to the Klein four-group, $\mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z}$. On another curve, like $y^2 = x^3 - 2x + 1$, the same process uncovers a point of order 4, revealing a cyclic torsion subgroup $\mathbb{Z}/4\mathbb{Z}$. The sieve gives us a complete, precise classification.

But the true power of the number theorist's toolkit comes from combining different perspectives. We can also view the curve through an arithmetic lens by reducing its equation modulo a prime number, say $p=3$ or $p=5$. When we do this, our elegant curve over the infinite expanse of rational numbers becomes a finite collection of points over a finite field $\mathbb{F}_p$. A key insight is that the group of rational torsion points injects into this finite group. This means the number of points on the reduced curve, $\#E(\mathbb{F}_p)$, must be a multiple of the size of the rational torsion subgroup. For our curve $y^2 = x^3 - 4x$, counting points modulo $3$ and modulo $5$ both reveal that the number of rational torsion points must divide $4$. Since we've already found four such points (the three of order 2 and the identity), we know with certainty that our search is over. The algebraic sieve and the arithmetic lens provide a powerful cross-check, a testament to the unified nature of number theory.

So, the Nagell-Lutz theorem is a wonderful tool for finding and classifying points of finite order. But what happens when we find a rational point that fails the test? What if we discover a point whose coordinates are integers, but its $y$-coordinate squared does not divide the discriminant? Is our sieve broken?

On the contrary! This failure is not a bug; it is a feature—and perhaps the most exciting outcome of all. A failure of the Nagell-Lutz test is a rigorous proof that the point in question is not a torsion point. It must, therefore, be a point of infinite order.

Let’s look at the curve $E: y^2 = x^3 - 2$. A quick search for small integer solutions reveals the point $P = (3,5)$. Its coordinates are integers. The discriminant of this curve is $\Delta = -1728$. The $y$-coordinate squared is $y^2=25$. Does $25$ divide $-1728$? No. The test fails. The conclusion is immediate and profound: the point $(3,5)$ has infinite order.

This simple observation has monumental consequences. The celebrated Mordell-Weil theorem tells us that the group of all rational points on an elliptic curve, $E(\mathbb{Q})$, is finitely generated. It is structured as a direct sum of the finite torsion subgroup and a number of copies of the integers, $\mathbb{Z}^r$. This integer $r$ is called the rank of the curve. By finding a single point of infinite order, we have proven that the rank of the curve $y^2 = x^3 - 2$ is at least one. We have used a tool designed to understand finiteness to open a door into the infinite. The quest to understand the rank of elliptic curves is one of the central driving forces of modern number theory, and the ability to certify that a point has infinite order is the first step on that quest.

The distinction between points of finite and infinite order is not merely an abstract structural curiosity. This single concept provides the key to unlocking problems that are centuries old and, simultaneously, to building the technologies of the future.

​​The Congruent Number Problem​​

Let us travel back in time. An ancient problem, first studied by the Arabs and Greeks, asks: which whole numbers $n$ can be the area of a right-angled triangle whose sides are all rational numbers? Such a number is called a "congruent number." For example, $6$ is a congruent number because it's the area of the classic $(3,4,5)$ right triangle. $5$ is also congruent, being the area of the triangle with sides $(\frac{3}{2}, \frac{20}{3}, \frac{41}{6})$. But what about $1$, $2$, or $3$? Are they congruent numbers? For centuries, this simple geometric question remained bafflingly difficult.

The astonishing breakthrough came when mathematicians realized the problem was not about triangles at all. It was secretly a problem about elliptic curves. A number $n$ is a congruent number if and only if the elliptic curve $E_n: y^2 = x^3 - n^2x$ has a rational point of infinite order—that is, if its rank is greater than zero.

Suddenly, our entire toolkit becomes relevant. To prove that $1$ is not a congruent number, we must show that the curve $E_1: y^2 = x^3 - x$ has rank zero. Using our sieve, we can establish that the only rational points on this curve are the four torsion points $\mathcal{O}, (0,0), (1,0), (-1,0)$, all of which have $y=0$ and correspond to degenerate triangles. Proving that no other points exist requires a deeper tool known as "descent," but the fundamental goal is clear: show that every rational point is a torsion point. The machinery of torsion points provides the framework for solving a riddle that predates the Roman Empire.

​​Cryptography and the Digital World​​

Now, let's leap forward to the 21st century. The security of much of our digital information relies on the difficulty of factoring very large integers. One of the most powerful algorithms for this task is the Elliptic Curve Method (ECM). The strategy is beautifully counterintuitive. To factor an integer $N$, we don't study it directly; instead, we pick a random elliptic curve and a point on it, and we perform the group law computations modulo N.

Here's the trick: if $p$ is an unknown prime factor of $N$, our calculations are secretly taking place on an elliptic curve over the finite field $\mathbb{F}_p$. The number of points on this curve, $\#E(\mathbb{F}_p)$, is some integer near $p+1$. If we are lucky, this number is "smooth"—meaning it is composed only of small prime factors. When this happens, our group computations get stuck in a predictable way that, through a simple calculation, reveals the factor $p$.

This raises a fascinating question: Does Mazur's theorem, which tells us that rational torsion is very rare and limited to small orders, imply that smooth group orders are also rare, thus making ECM inefficient? The answer is a decisive no. The rules of the game change completely when we move from the rational numbers $\mathbb{Q}$ to a finite field $\mathbb{F}_p$. Over a finite field, the group structures are far wilder and more varied than over $\mathbb{Q}$. Mazur's theorem is a statement about the rigid arithmetic of a single global field, whereas ECM thrives on the rich statistical landscape of curves over many different local fields. The distinction between torsion over $\mathbb{Q}$ and the group structure over $\mathbb{F}_p$ is a beautiful illustration of a deep principle: in number theory, the world you are in matters profoundly.

We have seen that the rank—the number of independent, infinite-order points—is a concept of central importance. This naturally leads to the million-dollar question (literally, as it's a Millennium Prize Problem): how can we compute the rank?

This question leads us to the frontier of modern mathematics and the breathtaking Birch and Swinnerton-Dyer (BSD) conjecture. The conjecture posits a deep and mysterious connection between the algebraic world of rational points and the analytic world of complex functions. For every elliptic curve $E$, one can construct a special function called its Hasse-Weil $L$-function, $L(E,s)$. The BSD conjecture predicts that the algebraic rank of the curve is equal to the order of vanishing of its $L$-function at the central point $s=1$. In simpler terms, the number of fundamental infinite-order solutions is encoded in how "flat" this special function is at a single point.

What does this have to do with torsion? The simplest case of the conjecture is for curves of rank zero. These are precisely the curves where every rational point is a torsion point. For these curves, the BSD conjecture predicts that the $L$-function should not be zero at $s=1$. Verifying the torsion structure of a curve is therefore the first step toward understanding the simplest case of one of the deepest and most far-reaching conjectures in all of mathematics.

From a practical computational tool to a key for ancient puzzles, a component of modern technology, and a stepping stone to the highest peaks of number theory, the study of rational torsion points is a perfect microcosm of the mathematical experience. It begins with a simple question and, in pursuit of an answer, leads us on a grand tour of the intellectual world, revealing unexpected connections and a profound, underlying unity.