Sunada's Theorem

The question "Can one hear the shape of a drum?", posed by mathematician Mark Kac, captures a fundamental problem in spectral geometry: does an object's complete set of vibrational frequencies—its spectrum—uniquely determine its shape? While our physical intuition suggests a direct link, the mathematical reality is far more subtle and surprising. This article addresses the gap between this intuitive expectation and geometric fact by exploring the definitive "no" that theory provides, centered on the profound insights of Sunada's theorem. The reader will first journey through the ​​Principles and Mechanisms​​, uncovering what can be "heard" from a spectrum and how the group-theoretic concept of almost conjugacy provides a recipe for building identically sounding yet distinct shapes. Subsequently, the ​​Applications and Interdisciplinary Connections​​ chapter explores the far-reaching impact of this theorem, from resolving Kac's famous question to its astonishing parallels in number theory, revealing a deep and unexpected unity across different mathematical fields.

You may recall from our introduction that the central, poetic question we’re chasing is, “Can one hear the shape of a drum?” It’s a beautiful question, but what does it really mean? A physicist or mathematician hears this and thinks not of a goatskin drum, but of an abstract geometric shape, and its "sound" is not a noise, but a list of numbers—the frequencies at which it can naturally vibrate. This list of vibrational frequencies is called its ​​spectrum​​. So, the question really is: If two shapes have the exact same spectrum, must they have the exact same shape?

It seems intuitive that they should. After all, a tiny violin and a giant cello produce different sets of notes because their shapes are different. But as we’ll see, the world of geometry holds some wonderful surprises.

Let's imagine our "drum" is a flat shape, a domain $\Omega$ on a plane. When we "strike" it, we’re mathematically solving an equation for its vibrations, the eigenvalues $\lambda_k$ of an operator called the ​​Laplacian​​. This infinite list of numbers, $0 \lt \lambda_1 \le \lambda_2 \le \dots$, is the drum's spectrum. Each number corresponds to a "pure tone" the drum can make. The full sound of the drum is a combination of these pure tones, just as a piano chord is made of individual notes. It’s the multiplicity—how many ways the drum can vibrate at the same frequency—that's crucial. The spectrum is not just a set of numbers, but a ​​multiset​​, where frequencies are repeated according to their multiplicity.

Now, suppose you are given only this list of frequencies. What can you deduce about the drum's shape? A powerful way to analyze the spectrum is to cook it into a function called the ​​heat trace​​, $Z(t) = \sum_{k=1}^{\infty} \exp(-t \lambda_{k})$. You can think of this as describing how heat would dissipate from an object of that shape. If two drums have the same spectrum, they must have the exact same heat trace for all "time" $t > 0$.

Here’s the remarkable part. By looking at what happens to this heat trace for very, very small amounts of time ($t \to 0$), we can extract concrete geometric information about the drum. It’s like watching a flash of heat spread out.

• At the very first instant, the heat spreads locally, not yet "feeling" the boundaries of the drum. The rate at which the drum cools in this instant is related purely to its size. From the leading term of the heat trace's expansion, we can directly calculate the ​​Area​​ $A$ of the drum.

• A moment later, the heat begins to reach the edges, and it starts leaking out (if we imagine the boundary is held at zero temperature). The rate of this leakage is, naturally, proportional to the length of the boundary. The next term in the expansion tells us the ​​Perimeter​​ $L$.

• Wait just a little longer, and the heat starts to sense the overall topology of the drum. For example, does it have any holes? It turns out that a later term in the expansion reveals the ​​Euler characteristic​​ $\chi$, a number that, for a simple shape on a plane, is just $1 - h$, where $h$ is the number of holes.

This is a fantastic result! It means that if you can "hear" a drum, you can definitely hear its area, its perimeter, and whether it has any holes. So, the spectrum tells us quite a lot. But does it tell us everything? Does it tell us the exact shape?

To answer that, we must take a detour into what might seem like a completely unrelated field: the mathematics of symmetry, known as ​​group theory​​. This is where the story takes a surprising and elegant turn, thanks to a profound insight by the mathematician Toshikazu Sunada.

The core idea is to build our interestingly shaped drums not from scratch, but by starting with a much larger, highly symmetric "master shape," let’s call it $\tilde{M}$. Think of it like an infinite, perfectly repeating wallpaper pattern. This pattern has a group of symmetries, $G$—a collection of translations, rotations, and reflections that leave the overall pattern unchanged.

Now, from this single master pattern $\tilde{M}$, we can create smaller, finite patterns. How? By picking a smaller set of symmetries—a subgroup, let's call it $H$—and declaring that any two points on the wallpaper that are related by a symmetry in $H$ are now considered to be the same point. This process of "gluing" or "identifying" points is called taking a ​​quotient​​, and the resulting shape is written as $M_H = \tilde{M}/H$.

For example, if our wallpaper is just the grid of lines on graph paper, and our subgroup $H$ consists of all translations by integer amounts horizontally and vertically, the quotient shape is a single square of the graph paper with its opposite edges identified—a torus, or the shape of a donut!

Sunada's brilliant idea was to take two different subgroups, $H_1$ and $H_2$, from the same large group of symmetries $G$, and build two different quotient shapes, $M_1 = \tilde{M}/H_1$ and $M_2 = \tilde{M}/H_2$. The killer question then becomes: under what conditions will these two differently constructed shapes, $M_1$ and $M_2$, "sound" exactly the same?

This is the heart of the mechanism. Sunada found a strange and beautiful condition on the subgroups $H_1$ and $H_2$ that guarantees their quotient shapes will be isospectral. This condition is called ​​almost conjugacy​​, or being a ​​Gassmann-Sunada pair​​.

What does it mean? Let's think about our big group of symmetries $G$. Its elements (the individual symmetry operations) can be sorted into "families" based on their geometric character. For instance, in the group of symmetries of a square, all $90$-degree rotations form one family, while all reflections across the diagonals form another. These families are called ​​conjugacy classes​​.

Two subgroups, $H_1$ and $H_2$, are ​​almost conjugate​​ if they are perfectly balanced with respect to these families. That is, for every single conjugacy class $C$ in the big group $G$, the subgroup $H_1$ must contain the exact same number of symmetries from that class as $H_2$ does. Formally, $|H_1 \cap C| = |H_2 \cap C|$ for all $C$. They might not contain the same elements, but they must have the same "flavor profile" of symmetry types. You can even test this yourself with specific groups like the permutation group $S_4$.

Why does this peculiar condition lead to identical sounds? The reason lies in the language of representation theory, but the intuition is this: the vibrational modes (eigenfunctions) on the master shape $\tilde{M}$ are themselves classified by the full symmetry group $G$. When we form a quotient shape like $M_1 = \tilde{M}/H_1$, the only vibrational modes from $\tilde{M}$ that can survive on $M_1$ are those that are left unchanged (are "invariant") by all the symmetries in $H_1$. The almost-conjugacy condition is the mathematical guarantee that the process of averaging over the symmetries in $H_1$ gives the exact same result as averaging over the symmetries in $H_2$ for every single vibrational mode. Therefore, for every possible vibration on $M_1$ with a certain frequency, there is a corresponding vibration on $M_2$ with the exact same frequency. Their spectra are identical.

So, we have a machine for making pairs of shapes that sound the same. The final, crucial step is to find a pair of subgroups $H_1$ and $H_2$ that are almost conjugate, but not conjugate.

What's the difference? Two subgroups are conjugate if you can turn one into the other using some symmetry from the big group $G$. Geometrically, this means the two quotient shapes $M_1$ and $M_2$ are actually congruent—they are isometric, just oriented differently in space. That's not interesting; of course congruent drums sound the same.

The magic happens when we find almost conjugate subgroups that are not conjugate. This means there is no rigid motion that can make $M_1$ look exactly like $M_2$. They are fundamentally different shapes. And yet, by Sunada's theorem, they are perfectly isospectral.

This provides the definitive and resounding "No!" to Mark Kac's question. Gordon, Webb, and Wolpert used this very method to construct the first examples of such shapes in 1992—a pair of funny-looking polygons made of the same pieces rearranged differently, proving that you can't always hear the shape of a drum.

The consequences can be quite startling. For instance, using this method, one can construct two curved surfaces that are perfectly isospectral, yet the shortest possible closed loop one can walk on the first surface has a different length than the shortest loop on the second. A property as fundamental as the ​​injectivity radius​​ (half the length of the shortest closed geodesic) is not a spectral invariant; you cannot "hear" it!

The story doesn't even end there. A geometric shape can "vibrate" in more complex ways than just its surface moving up and down. There are also vibrations of higher-dimensional structures called differential forms. An object has a whole family of spectra, one for each dimension of form (the $p$-form spectra). Two manifolds are called ​​strongly isospectral​​ if all of these corresponding spectra are identical.

Now, is it possible for two drums to sound the same on the fundamental level (functions, or $0$-forms) but have different "sounds" at these higher levels? In general, yes! This shows just how subtle the relationship between spectrum and geometry is.

But here is a final testament to the power and beauty of Sunada's construction. The condition of almost conjugacy, this simple-looking balancing act of symmetry types, is so robust that it doesn't just guarantee the function spectra are the same. It guarantees that the shapes are strongly isospectral! If $\tilde{M}/H_1$ and $\tilde{M}/H_2$ are built from almost conjugate subgroups, all their corresponding $p$-form spectra will be identical. The same deep, underlying symmetry principle governs the vibrations of all kinds on these remarkable shapes, revealing a hidden unity between the worlds of algebra and geometry.

After a journey through the intricate machinery of group theory and representations, we arrive at a vantage point from which a spectacular landscape unfolds. The principles we have just explored are not mere abstract curiosities; they are powerful tools that have carved new paths through the mountains of geometry, analysis, and even number theory. They answer old questions in surprising ways and reveal a stunning, unexpected unity across disparate fields of mathematics. Our guide on this next leg of the journey is the famous question posed by the mathematician Mark Kac in 1966: "Can one hear the shape of a drum?"

Imagine you are in a room with a collection of drums, but you are not allowed to see them. Your only tool is a perfect-pitch mallet. You can strike each drum and meticulously record all of its resonant frequencies—its "spectrum." The question is, can you, from this list of frequencies alone, reconstruct the exact shape of each drumhead?

Our physical intuition screams "yes!" It seems preposterous that two differently shaped drums could produce the exact same set of sounds. Every nuance of the boundary, every corner and curve, ought to leave its faint signature on the harmonics. For many years, this was the prevailing belief. But mathematics, in its beautiful and often mischievous way, tells us something different. The answer is no.

Sunada's theorem is not just a proof that the answer is no; it is a constructive recipe, a veritable machine for creating pairs of drums that sound identical yet have different shapes. These are known as isospectral, non-isometric domains.

The construction, in principle, is beautifully simple. We start with a large, symmetric tiling of the plane, often by a single polygonal shape, like a triangle. The symmetries of this tiling form a group, let's call it $G$. Sunada's theorem then provides us with a group-theoretic blueprint for selecting a specific number of these tiles and gluing them together in two different ways. The blueprint is a "Sunada triple" $(G, H_1, H_2)$, where $H_1$ and $H_2$ are subgroups that are almost conjugate but not conjugate. This algebraic distinction is the heart of the matter. The "almost conjugate" property guarantees that the two resulting shapes, let's call them $\Omega_1$ and $\Omega_2$, will be isospectral. But the "not conjugate" property ensures they cannot be the same shape—one cannot be rotated, shifted, or flipped to become the other.

How can we be so sure they sound the same? The proof reveals another layer of mathematical magic known as the "transplantation method." For any possible vibration pattern, or eigenfunction, on drum $\Omega_1$, the theorem gives us an explicit procedure to "transplant" it to drum $\Omega_2$. This isn't a simple cut-and-paste job. It's a subtle process of mixing and re-weighting the function's values across the various tiles that make up the drums. This transplantation, dictated by the algebraic relationship between $H_1$ and $H_2$, is guaranteed to produce a valid eigenfunction on $\Omega_2$ with the exact same frequency, or eigenvalue. Since this works for every frequency, the two drums must share an identical spectrum.

The question of whether one can hear a shape is not limited to flat, two-dimensional drums. It can be asked of any geometric object, or Riemannian manifold. Can we hear the shape of a sphere, a doughnut, or more complex, curved spaces that might model our own universe?

Again, Sunada's theorem provides a resounding "no." The same machinery that builds isospectral drums can be used to construct pairs of non-isometric curved surfaces and higher-dimensional spaces that are spectrally identical. For instance, one can construct pairs of beautiful, saddle-shaped hyperbolic surfaces that are perfect spectral twins.

These examples force us to refine our understanding of what the spectrum truly tells us. The eigenvalues of the Laplacian are "spectral invariants," meaning they are the same for any two isospectral manifolds. Through a miraculous connection known as the heat trace expansion, these invariants can be decoded to reveal certain geometric properties. For any pair of isospectral manifolds, we can know for certain that they have the same dimension, the same total volume, and the same total scalar curvature, among other things. The spectrum gives us the global "statistics" of a shape. However, it does not, as the Sunada examples show, determine the shape's local geometry—its specific twists, turns, and overall configuration. The spectrum provides an echo of the geometry, but some details are lost in the reverberation.

Perhaps the most breathtaking application of these ideas lies in a field that, at first glance, seems to have nothing to do with vibrating drums or curved spaces: the theory of numbers. We can ask an analogous question: "Can one hear the shape of a number system?"

A "number system" in this context is a number field—an extension of the rational numbers like $\mathbb{Q}(\sqrt{2})$ or $\mathbb{Q}(\sqrt[3]{5})$. Just as a drum has a spectrum of frequencies, a number field $K$ has a "spectrum" called its Dedekind zeta function, $\zeta_K(s)$. This complex function acts like a unique barcode, encoding profound arithmetic information about the field, such as how prime numbers factorize within it. From the functional equation of the zeta function, we can extract fundamental invariants of the number field, like its degree and its discriminant, a number that measures the "size" of its arithmetic structure.

So, the question becomes: if two number fields have the same Dedekind zeta function, must they be the same (isomorphic) field?

Astonishingly, the answer is again no. And the reason is precisely the same group-theoretic phenomenon of Gassmann equivalence that underpins Sunada's theorem. It is possible to find two number fields, $K_1$ and $K_2$, that are constructed as fixed fields of Gassmann equivalent but non-conjugate subgroups. By Sunada's theorem, these fields will be arithmetically equivalent, meaning $\zeta_{K_1}(s) = \zeta_{K_2}(s)$. As a consequence, they must have the same degree, the same number of real and complex embeddings, and the same discriminant. Yet, because the subgroups are not conjugate, the fields are not isomorphic. They are different number systems that "sound" arithmetically identical.

This is not just a theoretical possibility. The first explicit examples of isospectral, non-isometric manifolds, constructed by Marie-France Vignéras in 1980 even before Sunada's general formulation, came from exactly this deep connection to number theory. Vignéras used the arithmetic of quaternion algebras—a generalization of complex numbers—to build her examples. She found pairs of distinct arithmetic groups, derived from "maximal orders" in a quaternion algebra that were locally indistinguishable but globally different. These groups, when used to construct hyperbolic surfaces, resulted in non-isometric shapes that were proven to be isospectral via the Selberg trace formula and the profound Jacquet-Langlands correspondence, cornerstones of modern number theory,. These examples are not just isospectral for the Laplacian; they are "strongly isospectral," sharing the same spectrum for an entire family of commuting operators called Hecke operators, which probe even deeper into their arithmetic soul,.

From the tangible vibrations of a drum to the esoteric world of number fields, Sunada's theorem acts as a master key, unlocking a hidden unity. It reveals that at the heart of these seemingly unrelated questions lies a single, elegant principle about symmetry and its representations. It teaches us that knowing an object's "spectrum"—be it of frequencies, eigenvalues, or arithmetic invariants—does not always mean we know the object itself. Nature, both physical and mathematical, can produce echoes that are delightfully, and profoundly, deceptive. This is not a failure of our methods, but a discovery of a richer, more subtle reality, where different forms can create the same beautiful symphony.