Dedekind Sums

In the vast landscape of mathematics, some objects appear deceptively simple, yet hold the keys to profound and unexpected connections across different fields. The Dedekind sum is one such object—a curious arithmetic construction that, at first glance, seems to be a mere numerical curiosity. This article seeks to unravel the mystery of these sums, revealing the deep structures they embody and the critical roles they play. We will embark on a journey in two parts. The first section, "Principles and Mechanisms," will deconstruct the Dedekind sum, exploring its definition via the sawtooth function, its beautiful symmetries, and the elegant power of the Dedekind reciprocity law. Following this, the section on "Applications and Interdisciplinary Connections" will showcase its astonishing influence, revealing how this number-theoretic tool becomes essential in the advanced realms of modular forms, the theory of partitions, and even the geometric topology of our universe. Prepare to discover how a simple sum over fractions weaves a thread through the very fabric of modern mathematics and physics.

Alright, let's get our hands dirty. Having been introduced to the world of Dedekind sums, you might be wondering what they're actually made of. Before we can appreciate the grand symphony, we need to meet the individual musicians. The star player in this story is a humble but fascinating little function, the ​​sawtooth function​​, denoted as $((x))$.

Imagine a straight line with a slope of 1. It crosses the origin at a slant. Now, let's play a game. Every time our variable $x$ hits an integer, we reset. More precisely, for any number $x$, we look at its fractional part—the part after the decimal point. For instance, the fractional part of 3.14 is 0.14. The formal definition is $((x)) = x - \lfloor x \rfloor - \frac{1}{2}$ for any non-integer $x$, where $\lfloor x \rfloor$ is the 'floor' of $x$, the greatest integer less than or equal to it. If $x$ is an integer, we simply say $((x))=0$.

What does this function look like? It climbs steadily from $-1/2$ to $+1/2$ as $x$ goes from just above 0 to just below 1. Then, at $x=1$, it jumps back down to $-1/2$ and starts its climb all over again for the next interval. If you were to graph it, you'd see a pattern that looks exactly like the teeth of a saw, hence its name. It's a beautifully simple, periodic object. You can think of it as measuring the signed distance from $x$ to the nearest integer, with a little shift.

Now, what is a ​​Dedekind sum​​, $s(h,k)$? It's defined for two whole numbers, $h$ and $k$, that share no common factors (they are coprime), with $k$ being positive. The definition looks like this:

Let's not be intimidated by the symbols. What is this equation really telling us to do? We are marching along the numbers from $n=1$ to $k-1$. For each $n$, we take the fraction $\frac{n}{k}$ and evaluate its sawtooth value. Then, we take a "scrambled" version of this fraction, $\frac{hn}{k}$, and find its sawtooth value. We multiply these two results together. Finally, we sum up all these products.

You can think of it as a kind of "correlation". We have a set of evenly spaced points on the number line, $\frac{1}{k}, \frac{2}{k}, \ldots, \frac{k-1}{k}$. We're comparing the sawtooth values of these points with the sawtooth values of their "shuffled" counterparts, where the shuffling is done by multiplying by $h$ (and taking the result modulo $k$).

Let's try a simple example to see it in action. What is $s(2, 5)$?. Here, $h=2$ and $k=5$. The formula tells us to sum from $n=1$ to $4$. We calculate the pairs of sawtooth values:

• For $n=1$: $((\frac{1}{5}))((\frac{2}{5})) = (\frac{1}{5}-\frac{1}{2})(\frac{2}{5}-\frac{1}{2}) = (-\frac{3}{10})(-\frac{1}{10}) = \frac{3}{100}$
• For $n=2$: $((\frac{2}{5}))((\frac{4}{5})) = (\frac{2}{5}-\frac{1}{2})(\frac{4}{5}-\frac{1}{2}) = (-\frac{1}{10})(\frac{3}{10}) = -\frac{3}{100}$
• For $n=3$: $((\frac{3}{5}))((\frac{6}{5})) = (\frac{3}{5}-\frac{1}{2})((\frac{1}{5})) = (\frac{1}{10})(-\frac{3}{10}) = -\frac{3}{100}$
• For $n=4$: $((\frac{4}{5}))((\frac{8}{5})) = (\frac{4}{5}-\frac{1}{2})((\frac{3}{5})) = (\frac{3}{10})(\frac{1}{10}) = \frac{3}{100}$

Adding them up: $\frac{3 - 3 - 3 + 3}{100} = 0$. The sum is exactly zero! This isn't just a fluke. These sums, which at first glance seem like they could be any messy fraction, often turn out to be simple, beautiful rational numbers, and sometimes, with a surprising amount of cancellation, they vanish completely. This hints that there is a deep, hidden structure waiting to be discovered.

Whenever you find a strange new object in mathematics, a good strategy is to poke at it from different angles to see if you can find any patterns or symmetries. Let's try this with Dedekind sums.

Instead of a single case, let's look at a whole family, $s(1, c)$ for any integer $c \ge 2$. Here $h=1$, so the sum becomes $\sum_{n=1}^{c-1} ((\frac{n}{c}))^2$. This is a much simpler calculation, and with a bit of algebra involving sums of squares, it boils down to a wonderfully clean formula:

This is a beautiful result. It tells us the value for an infinite class of Dedekind sums in one fell swoop.

Now, what if we choose $h$ to be at the "other end" of the spectrum, say $h=c-1$? Remember that in the world of modular arithmetic, $c-1$ is the same as $-1$. So, we are computing $s(c-1, c)$. The calculation is a bit different, but the final answer is strikingly familiar:

This is a gorgeous symmetry! The sum for $h=1$ is the exact negative of the sum for $h=-1 \pmod c$. These are not random numbers; they obey elegant laws. Other symmetries exist too. For example, if $h h' \equiv 1 \pmod k$ (that is, $h'$ is the modular inverse of $h$), then $s(h,k)=s(h',k)$. Even more curious is an identity involving modular square roots of $-1$. For $k=13$, the solutions to $x^2 \equiv -1 \pmod{13}$ are $x=5$ and $x=8$. Incredibly, it turns out that $s(5,13) = s(8,13)$, and in this case, both are zero!

The patterns we've seen are all consequences of a truly remarkable theorem, the crown jewel of the theory: the ​​Dedekind Reciprocity Law​​. This law connects the sum for the pair $(h,k)$ with the sum for the "flipped" pair, $(k,h)$. It states that for any two coprime positive integers $h$ and $k$:

s(7, 18) = \left( \frac{1}{12}(\frac{7}{18} + \frac{18}{7} + \frac{1}{126}) - \frac{1}{4} \right) - s(18, 7)

s(h,k) = \frac{1}{4k} \sum_{m=1}^{k-1} \cot\left(\frac{\pi m}{k}\right) \cot\left(\frac{\pi mh}{k}\right)

\mathbb{E}[s_\infty^2] = \frac{5}{144}

In science, we often find that the most unassuming-looking formulas can turn out to be Rosetta Stones, unlocking secrets in entirely different languages. On the surface, the Dedekind sum is just a peculiar arithmetic construction, a clever sum over fractions. We have explored its definition and its wonderfully symmetric properties, like the famous reciprocity law. But to leave it there would be like describing a master key as just a piece of sculpted metal without ever trying it on a lock.

In this section, we will embark on a journey to see where this key fits. It turns out to have a surprisingly wide reach. We will first find it in its "native" land, as a crucial cog in the intricate clockwork of number theory and modular forms. Then, in a breathtaking leap, we will see it reappear as a fundamental yardstick for measuring the very fabric of space in modern geometry and theoretical physics. Prepare to see the connections, for this is where the true beauty of the Dedekind sum reveals itself.

The natural home of the Dedekind sum is the shimmering world of complex analysis, specifically in the theory of modular forms. These are functions that possess an extraordinary level of symmetry. One of the most fundamental of these is the Dedekind eta function, $\eta(\tau)$. Far from being just a dry product formula, you can think of it as the fundamental frequency of a vibrating string in string theory, or the partition function describing the states of a physical system.

The defining characteristic of $\eta(\tau)$ is its behavior under a special set of transformations called the modular group, $SL(2, \mathbb{Z})$. These transformations chop up the complex plane and rearrange it in a fascinating way, like looking at the world through an elaborate funhouse mirror. A modular form is a function that looks "the same"—up to a simple factor—after you've performed one of these transformations.

Here is the crucial point: when the eta function transforms, it doesn't just get rescaled. It also picks up a complex phase factor—a precise "twist." And what is the little machine that calculates the exact angle of this twist? Our friend, the Dedekind sum. For a transformation $\gamma = \begin{pmatrix} a & b \\ c & d \end{pmatrix}$ the rule is $\eta(\gamma \tau) = \epsilon(\gamma) \sqrt{c\tau+d} \, \eta(\tau)$ where the phase $\epsilon(\gamma)$ is a root of unity computed directly from a Dedekind sum involving the matrix entries. The Dedekind sum lives in the exponent and acts as the perfect, delicate correction factor needed to preserve the grand symmetry of the function.

You might think this is just a happy accident of algebra, but the structure of complex functions is rigid and unforgiving. The connection is much deeper. If you demand that the logarithm of the eta function, $\log \eta(\tau)$, be a well-behaved, self-consistent object as you navigate the landscape of modular transformations, you find that the Dedekind sums are forced upon you by the principle of analytic continuation. They are, in a very real sense, the mathematical glue that holds this beautiful symmetric picture together.

This might seem like a game for pure mathematicians, but it has profound consequences for a problem a child could ask: in how many ways can you write a number as a sum of positive integers? This is the partition function, $p(n)$. For example, $p(4)=5$ because $4$ can be written as $4$, $3+1$, $2+2$, $2+1+1$, and $1+1+1+1$. This number grows explosively, and finding a formula for it seems hopeless.

However, the generating function for $p(n)$—a device that packages all the values of $p(n)$ into a single object—is simply $1/\eta(\tau)$ (up to a small factor). The secrets of partitions are therefore encoded in the symmetries of the eta function. In a stunning achievement of mathematics, Hardy, Ramanujan, and Rademacher used this fact to find an exact formula for $p(n)$. Their method involves a complex integral where the main contributions come from the singularities of the generating function. And what happens near these singularities is governed by the modular transformations of $\eta(\tau)$. The Dedekind sums, dictating the phases of these transformations, combine to form a crucial "arithmetic factor" $A_k(n)$ in the final formula. This factor captures the intricate, oscillating way in which numbers related to a denominator $k$ contribute to the grand total for $p(n)$. In short, to count partitions, you must understand the modular dance of the eta function, and the choreographer of that dance is the Dedekind sum.

If the role of Dedekind sums in number theory is remarkable, their appearance in geometry and physics is nothing short of miraculous. We are about to jump from the abstract realm of numbers to the study of shapes—specifically, the shape of three-dimensional "universes," or what mathematicians call 3-manifolds.

One of the simplest and most important families of these spaces are the lens spaces, $L(p,q)$. You can imagine constructing one by taking the 3-sphere (the surface of a 4-dimensional ball) and "gluing" points together according to a specific twisted rule defined by the integers $p$ and $q$. The result is a finite, closed universe with a subtle twist in its fabric.

How can you tell if two such universes, say $L(7,1)$ and $L(7,2)$, are fundamentally the same or different? You can't just look at them. You need a special kind of fingerprint, a numerical invariant that captures the essence of the space's topology, a number that doesn't change if you bend or stretch the space.

Deep in the heart of modern geometry lies the monumental Atiyah-Patodi-Singer index theorem, a sweeping statement that connects the local geometry of a space (its curvature and structure) to its global topology (its overall shape). One of the key players in this theorem is a spooky quantity called the ​​eta invariant​​, $\eta(M)$. For a given geometric operator on a manifold $M$ (like an abstract version of a wave operator), the eta invariant measures the asymmetry in its spectrum of eigenvalues—in a sense, it's a measure of the imbalance between the manifold's "positive" and "negative" vibrational modes. It is a profound, and notoriously difficult to compute, topological invariant.

And now for the astonishing reveal. For a lens space $L(p,q)$, this purely geometric, spectral invariant is given by... a Dedekind sum. The formula is breathtakingly simple: $\eta(L(p,q)) = C \cdot s(q,p)$ for some rational constant $C$ which depends on the specific operator being studied.

Let that sink in. A quantity derived from the infinite spectrum of a differential operator on a geometric object is precisely calculated by a finite arithmetic sum over fractions. Why? What on Earth does a sum concocted from integer arithmetic have to do with the spectral asymmetry of a twisted sphere? This connection is one of the most beautiful and startling examples of the hidden unity of mathematics. It tells us that the same deep algebraic structures that govern the symmetries of modular forms also underpin the topological invariants of geometric spaces.

This is not an isolated curiosity. Like a recurring character in a grand epic, the Dedekind sum appears again and again as the soul of other topological invariants.

• ​​Chern-Simons Invariant:​​ In quantum field theory, physicists consider placing quantum theories onto these 3-manifolds. The Chern-Simons invariant measures a subtle "topological phase" acquired by the quantum system, a property that depends only on the shape of the spacetime it lives in. Once again, for certain fundamental configurations on a lens space, this physical quantity is computed directly by a Dedekind sum.
• ​​Casson Invariant and Knot Theory:​​ We can also build 3-manifolds by performing "surgery" on knots in space. Imagine taking a knot, like the famous figure-eight knot, cutting out its neighborhood, and gluing it back in with a twist described by two integers $p$ and $q$. The resulting manifold has a powerful invariant called the Casson invariant, which in a sense counts the different ways one can map surfaces into it. The formula for this invariant, for a manifold obtained by surgery on a knot, again explicitly features a Dedekind sum $s(q,p)$ dependent on the surgical twist.

From a phase factor in the transformation of a complex function, to a key ingredient in an exact formula for partitions, to a measure of asymmetry and topology in 3-dimensional spaces, to a component of quantum invariants—the Dedekind sum is a universal thread. It weaves through number theory, complex analysis, topology, and even theoretical physics. Its story is a vivid testament to the fact that these fields are not separate islands of thought, but different mountain ranges on the same intellectual continent, connected by deep and hidden valleys. The simple sum over fractions has proven to be a key that unlocks doors we never suspected were related, reminding us that in the search for knowledge, the simplest-looking objects can lead to the deepest and most unexpected connections.