Representations of groups

Symmetry is a concept we intuitively understand, from the balance of a butterfly's wings to the perfect form of a crystal. But how can we use this abstract idea of symmetry to make concrete, mathematical predictions about the world? This is the central question addressed by the theory of group representations. This powerful mathematical framework provides a formal language to translate the abstract principles of symmetry into tangible insights about physical systems. While symmetry itself is descriptive, representation theory turns it into a predictive engine, revealing hidden rules that govern everything from subatomic particles to macroscopic materials.

This article will guide you through this fascinating field. In the first chapter, "Principles and Mechanisms," we will delve into the core machinery of the theory, discovering the "atomic" building blocks known as irreducible representations and the beautiful rules that dictate their behavior. Following that, in "Applications and Interdisciplinary Connections," we will see this theory in action, exploring how it explains quantum phenomena, dictates the properties of materials, and even provides a framework for analyzing the shapes of living organisms. Prepare to see how the elegant language of symmetry shapes the laws of nature.

Now that we have a taste of what group representations are, let's roll up our sleeves and look under the hood. How does this all work? You might think that representing a group with matrices is a bit of a free-for-all. Can we just pick any matrices that happen to follow the group's multiplication table? The answer, remarkably, is no. There is a deep, beautiful, and surprisingly rigid structure that governs these representations. Understanding this structure is like a physicist discovering that all matter is made of a few fundamental particles governed by a few simple rules.

Let’s think about the vector space $V$ on which our matrices act. We can think of it as a stage, and the vectors are the actors. When we apply a group element, represented by a matrix, it's like a director shouting "Action!"—the vectors move around.

Sometimes, a peculiar thing happens. We might find a smaller group of actors who only ever interact among themselves. No matter what the director shouts (i.e., no matter which group matrix we apply), these actors never leave their corner of the stage. This corner, a subspace of the main stage $V$, is called an ​​invariant subspace​​. If our representation has such a self-contained, non-trivial corner (meaning it's not the entire stage or just the empty set), we call the representation ​​reducible​​. It's like watching a play with two entirely separate subplots happening at once. You could, in principle, just watch one subplot and it would make perfect sense.

But what if a representation cannot be broken down like this? What if every actor, sooner or later, interacts with every other actor? What if there are no private corners on the stage? This is the fundamental, indivisible unit of our theory: the ​​irreducible representation​​, or ​​irrep​​ for short. These are the "atoms" of symmetry. Just as all matter is built from atoms, we will see that all representations are built from irreps.

To get a feel for this, consider the simplest group imaginable: the trivial group $G = \{e\}$, containing only the identity element. Any representation $\rho$ must map $e$ to the identity matrix, $\rho(e) = I$. Now, what does it take for such a representation to be irreducible? The action of the group is just the identity matrix, which leaves every vector exactly where it is. This means that any subspace is an invariant subspace! For the representation to be irreducible, we are forced into a corner: the only way to satisfy the definition (that the only invariant subspaces are the trivial one $\{0\}$ and the whole space $V$) is if there are no other subspaces to begin with. This happens only if the vector space $V$ is one-dimensional. Thus, for the trivial group, the only irreducible representation is a one-dimensional one. It's the simplest atom of them all.

Here is where the magic begins. It turns out that for any finite group, the properties of its irreps are not random. They are strictly controlled by the structure of the group itself. There are two golden rules that are incredibly powerful.

​​Rule 1: The Sum of Squares​​

This is one of the most astonishing results in mathematics. If you find all the non-equivalent irreducible representations of a finite group $G$, and their dimensions are $d_1, d_2, \dots, d_k$, then they must satisfy the equation:

The sum of the squares of the dimensions of the fundamental building blocks equals the order (the number of elements) of the group! This is not an approximation; it is an exact, beautiful law. So, if you have a group with 4 elements, like the $C_4$ rotation group, the squares of the dimensions of its irreps must add up to 4. If a group has 169 elements, they must add up to 169. This is an incredibly powerful constraint. It's like a conservation law for symmetry.

​​Rule 2: The Number of Atoms​​

So, we know something about the dimensions of the irreps. But how many of them are there? Is there one? A dozen? A million? The theory gives us another exact law:

The number of non-equivalent irreducible representations is equal to the number of ​​conjugacy classes​​ of the group.

What's a conjugacy class? Intuitively, it's a subset of group elements that are structurally "alike". For example, in the group of symmetries of a square, all 90-degree rotations (clockwise and counter-clockwise) are in one class, while the 180-degree rotation is in a class by itself. The identity element is always in a class of its own. This rule connects the "atomic number" of our representation theory directly to the internal structure of the group.

Armed with these two rules, we can become detectives and deduce an incredible amount about a group's representations just from its basic properties. Let's see how this plays out for two major families of groups.

​​The Amiable Abelian Groups​​

First, let's consider the friendly, well-behaved ​​abelian groups​​, where the order of operations doesn't matter ($ab=ba$). A simple example is the cyclic group $\mathbb{Z}_{12}$ (the numbers on a clock face with addition). In an abelian group, because everything commutes, conjugating an element $g$ by some other element $h$—calculating $hgh^{-1}$—just gives you back $g$. This means every element sits in its own conjugacy class, all alone.

So, for an abelian group of order $|G|$, there are $|G|$ conjugacy classes. Our second rule tells us there must be $|G|$ irreducible representations. Now, let's bring in the first rule: the sum of the squares of their dimensions must equal $|G|$.

Think about this equation. We are adding up $|G|$ numbers, each of which is the square of a positive integer ($d_i \ge 1$), and the total must be $|G|$. The only possible way to do this is if every single dimension is 1! $1^2 + 1^2 + \dots + 1^2$ ($|G|$ times) $= |G|$.

This leads to a profound conclusion: ​​all irreducible representations of a finite abelian group are one-dimensional.​​ This holds for cyclic groups of prime order $p$, groups of order $p^2$, and any other finite abelian group you can think of.

We can even look at this from another angle using a powerful result called ​​Schur's Lemma​​. It leads to the same place: if a group has an irreducible representation with a dimension greater than 1, that group simply cannot be abelian. The existence of higher-dimensional irreps is a definitive fingerprint of a non-commutative structure.

​​The Rugged Non-Abelian Groups​​

What about the more complex non-abelian groups, where $ab$ is not always the same as $ba$? Our rules still apply, turning what seems like a chaotic situation into a solvable puzzle.

Imagine scientists discover a crystal whose symmetries are described by a point group of order 8. They also find it has 5 conjugacy classes. What can we say about its fundamental modes (its irreps)? Well, we know there must be 5 irreps. Let their dimensions be $d_1, d_2, d_3, d_4, d_5$. Our sum-of-squares rule demands:

Now we just have to find five positive integers whose squares sum to 8. You can quickly convince yourself that any dimension of 3 or more is impossible ($3^2 = 9$, which is already too big). So the dimensions can only be 1 or 2. If we try two 2s, we get $2^2 + 2^2 = 8$, but we need five terms. The only way to make it work is with one 2 and four 1s: $2^2 + 1^2 + 1^2 + 1^2 + 1^2 = 4 + 1 + 1 + 1 + 1 = 8$. Just like that, we've deduced the dimensions of the irreps must be {1, 1, 1, 1, 2}.

Let's try another one: a non-abelian group of order 10. It is known to have 4 conjugacy classes (so 4 irreps) and two 1-dimensional representations. We need to find the dimensions $d_3$ and $d_4$. The puzzle is:

The only way to write 8 as the sum of two squares of integers is $2^2 + 2^2$. So, the dimensions must be {1, 1, 2, 2}. The rigid rules of representation theory reveal the group's inner structure with mathematical certainty.

So we have our "atoms"—the irreps. What about the "molecules"—the reducible representations? Here we find the final piece of this beautiful puzzle. For any finite group, a miraculous theorem by Heinrich Maschke, called ​​Maschke's Theorem​​, comes into play. It guarantees that any reducible representation can be cleanly and completely broken down into a direct sum of irreducible representations. It tells us that our analogy is perfect: every representation is simply a collection of irreps, just as every molecule is a collection of atoms. A representation is said to be ​​completely reducible​​.

One of the most important examples is the ​​left regular representation​​. This is a "universal" representation you can build for any finite group $G$. Its dimension is simply the order of the group, $|G|$. When you decompose this giant representation, you find something amazing: it contains every single irreducible representation of the group. Furthermore, each irrep of dimension $d_i$ appears exactly $d_i$ times in the decomposition. This means the total dimension of the regular representation, $|G|$, must be the sum of the dimensions of all the irreps it contains, which is $\sum_i d_i \times d_i = \sum_i d_i^2$. This provides a deep and intuitive reason for our "sum of squares" rule! Everything fits together.

A word of caution is in order. This elegant picture of complete reducibility, guaranteed by Maschke's Theorem, is a special privilege of finite groups. For infinite groups, like the group of all integers $\mathbb{Z}$, things can get murky. It's possible to construct representations that are reducible—they have an invariant subspace—but that cannot be broken down completely. They are like a faulty zipper that gets stuck halfway. This highlights just how special and well-behaved the world of finite group representations truly is, a world of perfect decomposition into fundamental, atomic parts.

After our journey through the formal structure of group representations, you might be tempted to view it as a beautiful but isolated piece of abstract mathematics. Nothing could be further from the truth. In fact, you have just learned the rules of a language that nature speaks constantly, from the deepest levels of the quantum realm to the intricate forms of life itself. Wherever there is symmetry, the machinery of group representations provides the grammar, allowing us to predict, classify, and understand the behavior of the world. It is not merely a descriptive tool; it is a predictive one, revealing the profound consequences of symmetry on the laws of physics, chemistry, and biology.

Perhaps the most fundamental and immediate application of representation theory is in quantum mechanics. The state of a quantum system, like an electron in a molecule, is described by a wavefunction. If the molecule has some symmetry—say, the threefold rotational symmetry of an ammonia molecule—then the Hamiltonian, the operator for the system's total energy, must also share that symmetry. What does this mean? It means the laws of physics look the same after you perform a symmetry operation.

A staggering consequence follows from this simple fact. The set of wavefunctions that all share the same energy level must form a basis for one of the irreducible representations (irreps) of the molecule's symmetry group. And how many wavefunctions are there at that energy level? In other words, what is the degeneracy of the level? The answer is simply the dimension of the irrep! This dimension, as we've learned, is given by the character of the identity element, $\chi(\hat{E})$. So, if we find from calculation or experiment that a set of states in a molecule with $C_{3v}$ symmetry transforms as the $E$ irrep, we know instantly that the energy level must be doubly degenerate, because for any $E$ representation, $\chi(\hat{E})=2$ by definition. This is not a coincidence; it is a mandate from symmetry. The universe does not permit a single, non-degenerate state in this situation that transforms according to the full $E$ representation.

This insight allows us to draw a crucial distinction between two types of degeneracy. When a set of states forms a basis for a single irrep with a dimension greater than one (like the $p_x, p_y, p_z$ orbitals of a hydrogen atom, which form a basis for a 3D irrep of the rotation group), we call this a ​​symmetry-protected degeneracy​​. It is robust and necessary. However, sometimes energy levels belonging to different irreps just happen to have the same energy. We call this an ​​accidental degeneracy​​. For example, in a molecule with $C_{2v}$ symmetry (like a water molecule), all the irreps are one-dimensional. Therefore, this symmetry group cannot protect any degeneracy. If we ever find two energy levels with the same energy in such a molecule, we know the degeneracy must be accidental. This distinction is profound: an accidental degeneracy is fragile and can be broken by any tiny perturbation that preserves the overall symmetry. A symmetry-protected degeneracy can only be broken if the perturbation itself breaks the symmetry.

This brings us to one of the most powerful predictive uses of group theory: understanding what happens when symmetry is reduced. Consider an atom in the perfect spherical symmetry of empty space. Its orbitals ($s, p, d, \dots$) correspond to the irreps of the full rotation group, $SO(3)$. Now, place this atom into a crystal, for instance at a site with tetrahedral ($T_d$) symmetry. The atom no longer sees the universe as spherically symmetric; it sees the more limited symmetry of the tetrahedron. The original representation spanned by the five d-orbitals is no longer irreducible under this new, smaller symmetry group. It must "break apart," or decompose, into the irreps of the $T_d$ group. Using character theory, we can calculate precisely how this happens. The five-fold degenerate $d$-orbital level splits into a doubly-degenerate level (of type $E$) and a triply-degenerate level (of type $T_2$). This phenomenon, known as crystal field splitting, is the key to understanding the colors of transition metal complexes and the magnetic properties of materials. The abstract process of restricting a representation to a subgroup has a direct, measurable, and colorful physical manifestation.

The influence of group theory is not confined to the invisible quantum world. It provides the operating principles for the macroscopic properties of matter and, in a truly stunning leap of disciplines, for the very shape of living things.

How do we probe the quantum energy levels we just discussed? A primary tool is spectroscopy, which measures how matter absorbs or scatters light. Here again, group theory acts as the ultimate arbiter, establishing ​​selection rules​​. A vibrational mode in a crystal, for example, is active in Raman spectroscopy only if its irrep is contained in the representation of the symmetric polarizability tensor. In simple terms, the symmetry of the vibration must "match" the symmetry of how light interacts with the material. This allows us to predict which vibrations will be "seen" in an experiment and which will be "dark." This principle is so universal that it applies even to the most exotic forms of matter. Quasicrystals, which have long-range order but lack the periodic symmetry of conventional crystals, possess beautiful icosahedral symmetry ($I_h$) forbidden in periodic lattices. Yet, by decomposing the polarizability tensor into the irreps of the $I_h$ group, we can precisely predict the Raman selection rules for these strange materials, finding which of their unique "phason" modes are degenerate and observable.

Symmetry not only dictates what we see, but also constitutes the very substance of a material's physical laws. Consider the elasticity of a crystal—how it deforms under stress. This relationship is described by a rank-4 tensor, $T_{ijkl}$, which naively could have up to $3^4 = 81$ independent components. Measuring all of them would be a nightmare. But for a crystal with, say, full cubic symmetry ($O_h$), the components of this tensor cannot be arbitrary. The tensor itself, as a physical property of the crystal, must be invariant under all 48 symmetry operations of the cube. In the language of representation theory, this means we are looking for the component of the tensor representation that transforms as the trivial irrep ($A_{1g}$). A character-based calculation reveals a stunning simplification: for any rank-4 tensor describing a physical property in a cubic crystal, there can be at most a handful of independent components, not 81. For elasticity, this reduces the problem to just 3 constants, and for more complex phenomena, the principle remains the same. The same logic allows us to determine the 9 constants for an orthotropic material or even the 51 constants in advanced theories of strain-gradient elasticity. Symmetry writes the constitutive law.

Perhaps the most breathtaking application of this way of thinking comes from biology. Look at a butterfly. It is, to a good approximation, bilaterally symmetric. Is there a way to mathematically capture and quantify this symmetry, and the small deviations from it? The answer is a resounding yes, using the very same tools we used for quantum mechanics. The shape of an organism can be digitized as a vector of landmark coordinates. The action of reflection across the body's midline is a linear operator, $R$, acting on this vector. Just as in our quantum examples, we can define projection operators: $P_{sym} = \frac{1}{2}(I+R)$ projects any shape onto its perfectly symmetric component, while $P_{asym} = \frac{1}{2}(I-R)$ projects it onto its purely asymmetric component.

This is not just a mathematical game. It allows biologists to decompose the total shape variation in a population into a part that represents symmetric differences between individuals (e.g., some are larger, some smaller) and a part that represents asymmetry. This asymmetry can then be further analyzed: is there a consistent, population-wide skewing to one side (directional asymmetry)? Or is it random deviation from the symmetric plan (fluctuating asymmetry), a sensitive indicator of developmental stress and environmental pressure? The same logic extends to the radial symmetry of a starfish (governed by the cyclic group $C_n$) or the helical symmetry of a snail shell. The decomposition of shape variation into its irrep components is analogous to a Fourier analysis, breaking down complex biological form into its fundamental symmetry components.

From the degeneracy of an electron's orbit, to the color of a gemstone, to the stiffness of a steel beam, and finally to the delicate asymmetry in a butterfly's wing, the logic of group representations provides a single, unified framework. It teaches us a profound lesson: to understand a system, first understand its symmetry. The consequences will then unfold before you, dictated by the rigorous and beautiful mathematics of representation theory.