Irregular Primes

In the vast landscape of prime numbers, most behave predictably, following elegant rules. However, a special class known as ​​irregular primes​​ defies simple classification, creating fascinating complexities within number theory. For centuries, these primes were seen as mere curiosities or, more frustratingly, as stubborn obstacles that thwarted attempts to solve profound problems like Fermat's Last Theorem. The challenge was to understand what made them "irregular" and what deeper mathematical truths their existence might reveal.

This article demystifies the world of irregular primes. The following chapters will uncover their fundamental definition, explore the surprising connection to Bernoulli numbers through Kummer's criterion, and trace their impact from the 19th-century pursuit of Fermat's Last Theorem to their significance in contemporary research. You will learn how these primes are not just renegade numbers, but rather guides to a richer, more unified understanding of mathematics, bridging classical analysis with modern algebra.

So, we've met these peculiar characters called ​​irregular primes​​. They are the renegades, the primes that spoil a pristine picture of number theory and, for a time, stood in the way of proving one of mathematics' greatest challenges: Fermat's Last Theorem. An odd prime $p$ is irregular if it divides the class number of the $p$-th cyclotomic field $\mathbb{Q}(\zeta_p)$. This seems like a rather esoteric definition, locked away in the high towers of algebraic number theory. But the story of why this definition matters and how we can get our hands on it is a beautiful journey, one that reveals the surprising and profound unity of mathematics. It’s a detective story where the clues come from completely different worlds.

Imagine you are a detective trying to identify these "irregular" primes. Checking the definition directly—calculating the class number $h(\mathbb{Q}(\zeta_p))$ and seeing if $p$ divides it—is monstrously difficult. It's like trying to count every grain of sand on a vast, intricate beach. We need a better way, a tell-tale sign, a fingerprint that these primes leave behind. In the mid-19th century, Ernst Kummer found one, and it was a complete shock. The clue wasn't in the field of numbers itself, but in an unassuming sequence of rational numbers that pop up in calculus, of all places: the ​​Bernoulli numbers​​, $B_k$.

These numbers, defined by a generating function $\frac{t}{e^t - 1} = \sum_{n=0}^{\infty} B_n \frac{t^n}{n!}$, seem to live in the world of analysis. They appear when you sum powers of integers ($1^k + 2^k + \dots + N^k$) or write down the Taylor series for trigonometric functions. What could they possibly have to do with the structure of number fields?

And yet, Kummer proved a stunning equivalence, a result so deep it feels like magic.

​​Kummer's Criterion​​: An odd prime $p$ is irregular if and only if $p$ divides the numerator of one of the Bernoulli numbers $B_{2k}$, for an even index $2k$ in the range $2 \le 2k \le p-3$.

Suddenly, the impossible task becomes a finite, computational one. We don't need to understand the whole beach; we just need to test a small, specific set of "sand grains". For a given prime $p$, we can calculate $B_2, B_4, \dots, B_{p-3}$ and check their numerators. For example, for $p=5$, we only need to check $B_2 = 1/6$. The numerator is 1, which 5 does not divide. So 5 is regular. For $p=37$, one has to check $B_2, \dots, B_{34}$. It turns out that the numerator of $B_{32}$ is divisible by 37. Bingo! We've found our first irregular prime. This criterion is not just a curious coincidence; it's a bridge between two distant continents of mathematics. To understand why this bridge exists, we have to look under the hood at the engine driving it all.

The first step in demystifying Kummer’s criterion is to realize that the class group of $\mathbb{Q}(\zeta_p)$ has a hidden symmetry. The field $\mathbb{Q}(\zeta_p)$ contains complex numbers, so we can act on it with ​​complex conjugation​​—the simple act of switching $i$ with $-i$. In our field, this corresponds to sending the root of unity $\zeta_p$ to its inverse, $\zeta_p^{-1}$. This automorphism acts like a mirror.

Just as an object can be separated from its mirror image, the class group can be "split" into two pieces based on this symmetry. There's a "plus" part, which is left unchanged by complex conjugation (like a symmetric object), and a "minus" part, which is inverted (like a chiral molecule). This splits the class number itself into two integer factors: $h_p = h_p^+ h_p^-$.

• The ​​plus part​​, $h_p^+$, is the class number of the maximal real subfield $\mathbb{Q}(\zeta_p + \zeta_p^{-1})$. This part is notoriously difficult to understand. In fact, one of the most famous open problems in the field, ​​Vandiver's Conjecture​​, asserts that $p$ never divides $h_p^+$. While still unproven, it has been verified for millions of primes and gives us a sense that the "plus" part is somehow "tame" with respect to divisibility by $p$.

• The ​​minus part​​, $h_p^-$, is called the relative class number. And it turns out, this is where all the action happens! A deep theorem by Herbrand and Ribet tells us that a prime $p$ is irregular if and only if $p$ divides the minus part, $h_p^-$. The mystery of irregularity is entirely contained within this "anti-symmetric" part of the class group.

So, the original problem of $p$ dividing $h_p$ is equivalent to the more focused problem of $p$ dividing $h_p^-$. And you might have guessed it: the Bernoulli numbers are the key to the minus part.

To make the final connection, we need to bring in the powerful language of modern algebra. Think of the $p$-part of the class group—the part whose size is a power of $p$—as a musical instrument, say, a complex drum. The symmetries of our number field, the ​​Galois group​​, act like a musician striking this drum. Just as a real drum has specific resonant frequencies at which it vibrates most purely, the class group vibrates in specific ways under the action of the Galois group. These "resonant modes" are called ​​eigenspaces​​.

The Herbrand-Ribet theorem makes this analogy precise. It states that for each even index $2k$ (in the range $2 \le 2k \le p-3$), the corresponding eigenspace in the "minus" part of the class group is non-trivial (it "resonates") if and only if $p$ divides the numerator of $B_{2k}$.

This is the heart of the mechanism! Each potentially problematic Bernoulli number corresponds to a specific vibrational mode of the class group. If a Bernoulli number's numerator is divisible by $p$, the corresponding mode is "activated," making the class group non-trivial in a way that contributes to irregularity. The ​​irregular index​​—the number of such Bernoulli numbers—is literally counting the number of these activated resonant modes. This beautiful result replaces Kummer’s "magic" with the sublime music of Galois theory.

The story doesn't end there. The connection between special numbers and number fields has been one of the most fruitful themes in modern mathematics. The Bernoulli numbers themselves are not just random; they are special values of the famous ​​Riemann zeta function​​. For an integer $k \ge 2$, we have the identity $\zeta(1-k) = -\frac{B_k}{k}$.

This allows us to rephrase Kummer's criterion in a new language. The condition that $p$ divides the numerator of $B_k$ (for $k$ in the right range, where $p$ can't divide the denominator) is equivalent to the congruence $\zeta(1-k) \equiv 0 \pmod p$. Why is this better? It hints that there's a deeper analytic object at play.

This object is the ​​Kubota-Leopoldt $p$-adic L-function​​, denoted $L_p(s, \chi)$. Think of it as a strange function that lives not in the familiar world of real or complex numbers, but in the "p-adic numbers"—a number system where closeness is measured by divisibility by powers of $p$. This function is engineered to have values at negative integers that match the values of the classical zeta function (up to a correction factor). Kummer's criterion then gets a breathtakingly elegant restatement: a prime $p$ is irregular if and only if one of these special $p$-adic L-functions has a value at a certain point that is divisible by $p$. This reframes the entire theory, connecting the discrete arithmetic of prime numbers to the continuous world of $p$-adic analysis.

Throughout this story, we've carefully specified "odd primes". Why? What happens if we try to apply this magnificent machinery to the first prime, $p=2$? The whole thing collapses, and understanding why is a fantastic way to appreciate the machinery itself.

If $p=2$, the "primitive 2nd root of unity" is just $-1$. The great cyclotomic field $\mathbb{Q}(\zeta_2)$ is just... $\mathbb{Q}$, the rational numbers themselves.

• The class number of $\mathbb{Q}$ is $1$. The question of whether $p=2$ divides it is trivially false. The "problem" of irregularity doesn't exist.
• The range for the Bernoulli number indices, $2 \le 2k \le p-3$, becomes $2 \le 2k \le -1$. This range is empty. There are no Bernoulli numbers to check.
• The Galois group is trivial. There is no musician to strike the drum. The complex conjugation mirror is just a plain window—it does nothing. The split into "plus" and "minus" parts degenerates completely.

The theory of irregular primes is built on a rich structure that simply does not exist for $p=2$. Like a grand clockwork mechanism, it requires all its gears to be in place. For odd primes, those gears mesh perfectly, creating a beautiful symphony of connections between analysis, algebra, and number theory. But for $p=2$, the gears are missing, and there is only silence. This elegant failure at the boundary is a testament to the precise and intricate nature of the mathematical universe.

You might be thinking that the distinction between regular and irregular primes is a rather arcane piece of mathematical trivia. A prime $p$ is irregular if it happens to divide the numerator of some Bernoulli number $B_k$—so what? It seems like an accident of arithmetic, a curiosity for the specialists. But nothing in mathematics is ever just an accident. These "irregular" primes are not blemishes; they are beacons. They are signposts, planted by the universe, that point toward some of the deepest and most beautiful structures in number theory. To follow these signs is to embark on a journey that leads from a centuries-old puzzle to the frontiers of modern research, revealing the astonishing unity of mathematics along the way.

Our journey begins with the most famous problem in the history of mathematics: Fermat's Last Theorem. For over 350 years, the assertion that there are no positive integer solutions to the equation $a^n + b^n = c^n$ for any integer $n > 2$ tantalized and tormented mathematicians. The first major breakthrough came in the 19th century from Ernst Kummer. He realized he could prove the theorem for a large class of prime exponents $p$—precisely the ones he called "regular." The irregular primes were the stubborn obstacles, the cases where his methods failed. For a century and a half, they stood as monuments to the difficulty of the problem.

The final conquest of Fermat's Last Theorem, completed by Andrew Wiles in 1994, took a completely different, and breathtakingly modern, route. The strategy was one of profound elegance: a proof by contradiction. Let's suppose, for a moment, that a solution for a prime $p \ge 5$ actually exists. From the three numbers $(a, b, c)$ of this hypothetical solution, one can construct a strange new mathematical object: an elliptic curve, now called a Frey curve. For our purposes, you can picture this as a specific, doughnut-shaped geometric surface defined by the equation $y^2 = x(x-a^p)(x+b^p)$.

This Frey curve would be a very peculiar creature. It would have to exist in the "ecosystem" of all possible elliptic curves. But does it fit? The groundbreaking Modularity Theorem tells us that every elliptic curve over the rational numbers must be "modular"—that is, it must be associated with another kind of mathematical object, a special kind of wave-like function called a modular form. It’s as if we have a theorem stating every animal species must have a corresponding entry in a grand genomic database.

So, our Frey curve must have a corresponding modular form. But what kind? Here is the masterstroke. A deep result by Ken Ribet, known as the level-lowering theorem, allows us to predict the precise "level," a key frequency-like parameter, of the modular form that our Frey curve must match. The analysis shows that this level must be exceptionally simple: it must be $N=2$. So, if a solution to Fermat's equation exists, there must be a modular form of a specific type (a weight 2 newform) at level 2.

And here is the beautiful contradiction, the punchline to the longest-running joke in mathematics. It is a long-established, checkable fact of the theory of modular forms that the space of such forms at level 2 is empty. There are none. It's like a biologist proving that a certain type of animal must have a genetic sequence that is known to be impossible. Our hypothetical Frey curve is an impossible fossil. It cannot exist. And since the curve's existence is a direct consequence of the solution to Fermat's equation, that solution cannot exist either. The ghost of Fermat was finally laid to rest. The irregular primes, which had been the original stumbling blocks, were sidestepped by a glorious arc of reasoning that connected integers, geometry, and complex analysis.

While the modern proof of Fermat's Last Theorem bypassed the original difficulty of irregular primes, Kummer's initial insight remains profoundly important. What do these primes actually tell us? Their original definition connects them to the ideal class group, a concept that lies at the heart of algebraic number theory.

In the familiar world of integers, we enjoy unique prime factorization. The number 12 is, and always will be, $2^2 \times 3$. But in more exotic number systems, like the cyclotomic fields $\mathbb{Q}(\zeta_p)$ that Kummer studied, this property can fail. The class group of such a field is, in essence, a measure of how badly unique factorization fails. If the class group is trivial (of size 1), unique factorization holds. If it's larger, factorization becomes more complex.

Kummer's spectacular discovery was this: a prime $p$ is irregular if and only if $p$ divides the size of the ideal class group of the field $\mathbb{Q}(\zeta_p)$. The irregular primes are precisely those that flag a more complicated arithmetic in their corresponding cyclotomic worlds. The first such prime is $p=37$, a fact we can verify with a direct, though lengthy, computation of Bernoulli numbers modulo 37.

But the story gets even better. The connection is not just a correlation; it is a surgical tool. The Herbrand-Ribet theorem refines this connection with stunning precision. The class group, as an algebraic object, can be broken down into different "frequency components," or eigenspaces, indexed by powers of a special character called the Teichmüller character, $\omega$. The theorem tells us that if $p$ divides the numerator of a specific Bernoulli number, $B_k$, then a specific eigenspace of the class group, the one corresponding to the character $\omega^{p-k}$ (or $\omega^{1-k}$ by another convention), must be non-trivial.

Let's make this concrete. As mentioned, the smallest irregular prime is $p=37$. The only Bernoulli number $B_k$ (for $2 \le k \le 34$) whose numerator is divisible by 37 is $B_{32}$. The Herbrand-Ribet theorem then predicts that the eigenspace of the class group of $\mathbb{Q}(\zeta_{37})$ corresponding to the character $\omega^{37-32} = \omega^5$ must be non-trivial. Computational algebra has verified that the 37-part of this class group is a group of order 37. Combining this with Vandiver's Conjecture (which holds for $p=37$) that the "plus part" of the class group is trivial, we can deduce with mathematical certainty that the entire 37-part of the class group lives exactly in this $\omega^5$ eigenspace. An innocent-looking divisibility property of a single rational number, $B_{32}$, has revealed the precise location of the algebraic "trouble" inside a vast, abstract structure. It's like hearing a single off-key note and knowing instantly which violin string in a huge orchestra is out of tune.

The reach of irregular primes extends far beyond cyclotomic fields. The Bernoulli numbers themselves are famous for appearing in an equation discovered by Leonhard Euler, which connects them to the values of the Riemann zeta function $\zeta(s) = \sum_{n=1}^\infty \frac{1}{n^s}$ at positive even integers: $\zeta(2n) = (-1)^{n+1} \frac{(2\pi)^{2n} B_{2n}}{2(2n)!}$

This formula tells us that the numerators of the Bernoulli numbers are, up to some factors, the interesting parts of the special values of the zeta function. For example, the fact that the prime $691$ is the numerator of $|B_{12}|$ is a direct reflection of the value $\zeta(12)$. Thus, an irregular prime $p$ is one that is intimately tied to the arithmetic nature of a special value of the most celebrated function in number theory.

This connection has become a gateway to one of the most active areas of modern mathematics: Iwasawa theory. In the 20th century, mathematicians constructed $p$-adic $L$-functions, which are analogues of the Riemann zeta function that live in the world of $p$-adic numbers. The celebrated Main Conjecture of Iwasawa Theory (now a theorem) states, in essence, that the algebraic information contained in the class groups of a tower of number fields is perfectly mirrored by the analytic information contained in the zeros of a corresponding $p$-adic $L$-function.

What does this have to do with irregular primes? The number of "irregular indices" for a prime $p$, denoted $i(p)$, turns out to be equal to the number of zeros of the relevant $p$-adic $L$-function (counted with multiplicity) inside the $p$-adic unit disk. This is a profound dictionary, translating an algebraic counting problem into an analytic one.

This dictionary has led to fascinating statistical questions. If you compute the irregularity index $i(p)$ for thousands of primes, a remarkable pattern emerges. The distribution looks eerily like a Poisson distribution with mean $\lambda = 1/2$. About $60.6\%$ of primes appear to be regular ($i(p)=0$, close to $e^{-0.5}$), about $30.3\%$ have $i(p)=1$ (close to $0.5e^{-0.5}$), and so on. This statistical observation provides strong evidence for conjectures about how the zeros of $p$-adic $L$-functions are distributed. It suggests they behave like rare, independent events, a signature of randomness in a world governed by deterministic rules.

From Kummer's struggle with Fermat's Last Theorem, we have journeyed to the structure of factorization in number fields, the special values of the Riemann zeta function, and the statistical behavior of zeros of $p$-adic functions. The humble irregular prime has been our guide, revealing a web of connections that unifies disparate branches of mathematics. It is a perfect testament to the nature of mathematical discovery: follow a curious pattern, and you may just find yourself staring at the secrets of the universe.