On theories, symmetries and gauge

See also my preprint.

Abstract

We provide a foundational exposition on the concepts of theory, transformation, and symmetry in mathematical physics. By establishing a working definition of a physical theory as a selection mechanism for a set of fields, we meticulously distinguish between passive and active transformations. Passive transformations, including coordinate changes and gauge frame changes, are presented as mere relabelings of the underlying physical reality, which alter the description but not the gist of the theory. In contrast, active transformations, such as diffeomorphisms and active gauge transformations, are treated as genuine operations on the fields, potentially generating a new theory. A symmetry is then formally defined as an active transformation that leaves the selection criterion, typically a variational principle, invariant. We conclude by highlighting how this framework culminates in the modern gauge principle, where symmetries observed in nature are elevated to a foundational postulate used to construct the very dynamics of a theory, a cornerstone of both General Relativity and the Standard Model.

Introduction

Modern theoretical physics is built upon the twin pillars of General Relativity and the Standard Model of particle physics. At the heart of these frameworks lies a deep and intricate relationship between the dynamics of physical fields and the symmetries they obey. The concept of gauge has evolved from a simple redundancy in description to a powerful principle for dictating the fundamental interactions of nature.

However, the distinction between transformations as passive changes of description versus active operations on the state of a system can be a source of confusion. The Kretschmann objection to General Relativity, for instance, stemmed from a failure to clearly separate the trivial covariance achievable by relabeling from the profound physical principle of general covariance.

The purpose of this expository note is to provide a clear and formal framework for these ideas. We will define what constitutes a theory, analyze the distinct roles of passive and active transformations (for both spacetime and internal gauge spaces), and finally define symmetry as an invariance under active transformations. This will be illustrated with toy examples.

Defining a theory

We begin by establishing a clear operational definition of what we mean by a physical theory.

Definition (Theory). Given a base space $M$ and a fiber bundle $E\to M$ whose sections encompass all field degrees of freedom, the set of all possible field configurations is the total space of sections, $\mathcal{S}=\mathrm{\Gamma }(M,E)$. A theory is a mechanism $\mathcal{E}$ that selects a preferred subset of fields $S\subseteq \mathcal{S}$. The elements of $S$ are the fields of the theory, or its solutions.

In modern physics, the selection mechanism $\mathcal{E}$ is almost always a variational principle, where the fields in $S$ are the extrema of an action functional.

Examples

Example 1. A 0-dimensional spacetime.

• Space: A single-point manifold, $M=\{p\}$.
• Fields: Sections of the trivial bundle $E=M\times {\mathbb{R}}^n\to M$. A field is simply a vector $x\in {\mathbb{R}}^n$.
• Total Field Space: $\mathcal{S}=\mathrm{\Gamma }(M,E)\cong {\mathbb{R}}^n$.
• Selection Mechanism (Theory): A variational principle where the action is a function $\mathcal{E}:{\mathbb{R}}^n\to \mathbb{R}$:$$\mathcal{E}[x]=\frac{1}{2}\parallel x{\parallel }^2-⟨b,x⟩,$$where $b\in {\mathbb{R}}^n$ is a fixed vector (e.g., an external source).
• Selected Fields (Solutions): The set $S$ of fields satisfying $\delta \mathcal{E}[x]=0$. This yields a unique solution:$$S=\{x\in {\mathbb{R}}^n\mid x=b\}.$$

Example 2. The line $x=2$.

• Space: $M={\mathbb{R}}^2$ with standard coordinates $\phi =(x,y)$.
• Fields: Sections of the trivial bundle $E=M\times \mathbb{R}\to M$, i.e., scalar fields $f:M\to \mathbb{R}$.
• Total Field Space: $\mathcal{S}=\{f:M\to \mathbb{R}\mid f\text{ is measurable}\}$.
• Selection Mechanism (Theory): A variational principle with the action:$$\mathcal{E}[f]={\iint }_{{\mathbb{R}}^2}(x-2{)}^2(f_{\phi }(x,y){)}^{2}dxdy+{\iint }_{{\mathbb{R}}^2}\delta (x-2)(f_{\phi }(x,y)-1{)}^{2}dxdy$$where $f_{\phi }=f\circ {\phi }^{-1}$ is the coordinate representation of the field.
• Selected Fields: The variational principle $\delta \mathcal{E}[f]=0$ yields a unique solution, a field $h$ whose coordinate representation is$$h_{\phi }(x,y)=\{\begin{array}{ll}1,& \text{if }x=2\\ 0,& \text{if }x\ne 2\end{array}$$Thus, $S=\{h\}$. The theory selects a single field, which is supported on the vertical line $x=2$.

Example 3. Classical mechanics of particles.

• Space: Let $M=\mathbb{R}$, the one-dimensional manifold of Newtonian (absolute) time.
• Fields: Sections of the trivial bundle $E=M\times {\mathbb{R}}^3\to M$, so that a section $q(t)=(t,\mathbf{x}(t))$ describes the particle's trajectory $\mathbf{x}(t)\in {\mathbb{R}}^3$.
• Total Field Space: $\mathcal{S}=\mathrm{\Gamma }(M,E)=\{q:\mathbb{R}\to {\mathbb{R}}^3\mid q\in C^2(\mathbb{R},{\mathbb{R}}^3)\}$.
• Selection Mechanism (Theory): A variational principle with action$$\mathcal{E}[q]={\int }_{t_1}^{t_2}(\frac{1}{2}m\parallel \dot{\mathbf{x}}(t){\parallel }^2-V(\mathbf{x}(t)))dt,$$where $m$ is the particle mass and $V:{\mathbb{R}}^3\to \mathbb{R}$ a prescribed potential.
• Selected Fields (The Theory's Solutions):$$S=\{q\in \mathcal{S}|m\ddot{\mathbf{x}}(t)+\mathrm{\nabla }V(\mathbf{x}(t))=0\}.$$

Example 4. General Relativity with matter.

• Space: A 4-dimensional smooth manifold $M$ (spacetime).
• Fields: Sections of the bundle $E=\mathrm{Lor}(M)\oplus \mathcal{F}\to M$, where $\mathrm{Lor}(M)$ is the bundle of Lorentzian metrics and $\mathcal{F}$ represents matter fields. A typical element is a pair $(g,\varphi )$.
• Total Field Space: $\mathcal{S}=\mathrm{\Gamma }(M,\mathrm{Lor}(M))\times \mathrm{\Gamma }(M,\mathcal{F})$.
• Selection Mechanism (Theory): The Einstein--Hilbert action coupled to matter:$$\mathcal{E}[g,\varphi ]={\int }_M(R(g)+{\mathcal{L}}_{\text{matter}}(g,\varphi ,\mathrm{\nabla }\varphi )){\mathrm{vol}}_g$$where $R(g)$ is the Ricci scalar and ${\mathrm{vol}}_g$ is the volume form associated with the metric $g$.
• Selected Fields (Solutions): The set $S$ consists of all pairs $(g,\varphi )$ that satisfy the Euler-Lagrange equations, $\delta \mathcal{E}[g,\varphi ]=0$. These are precisely the solutions to the Einstein and matter field equations.

We can particularize the previous example into the following:

Example 4b. General Relativity with Pressureless Dust.

• Space: Let $M$ be a 4-dimensional smooth manifold (spacetime).
• Fields: Sections of the bundle$$E=\mathrm{Lor}(M)\oplus ({\mathrm{\Lambda }}^0(M)\oplus TM)\to M,$$where:
  • $\mathrm{Lor}(M)$: the bundle of Lorentzian metrics on $M$,
  • ${\mathrm{\Lambda }}^0(M)$: scalar fields (the mass density $\rho$),
  • $TM$: tangent bundle (the dust 4-velocity field $u^a$).
• Total Field Space:$$\mathcal{S}=\mathrm{\Gamma }(M,E)=\mathrm{\Gamma }(\mathrm{Lor}(M))\times \mathrm{\Gamma }({\mathrm{\Lambda }}^0(M))\times \mathrm{\Gamma }(TM).$$
• Selection Mechanism (Theory): A variational principle with action$$\mathcal{E}[g,\rho ,u]={\int }_M(R(g)+\rho g_{ab}{u}^a{u}^b)\sqrt{-g}{d}^{4}x,$$with constraints:
  • $g_{ab}{u}^a{u}^b=-1$ (normalization of 4-velocity),
  • ${\mathrm{\nabla }}_a(\rho u^a)=0$ (mass conservation).
• Selected Fields (The Theory's Solutions):$$S=\{(g,\rho ,u)\in \mathcal{S}|\begin{array}{rl}& G_{ab}=8\pi \rho u_a{u}_b,\\ & u^b{\mathrm{\nabla }}_b{u}^a=0,\\ & {\mathrm{\nabla }}_a(\rho u^a)=0\end{array}\}.$$

Example 4c. Mercury perihelion problem.

Example 5. Theories on fixed backgrounds.
Sometimes, theories are formulated on a spacetime with a prescribed structure, typically involving a fixed field such as a metric or a connection. For instance, classical field theories in special relativity, such as electromagnetism or scalar field theory, are often defined on a fixed Minkowski metric. Similarly, certain gauge theories may assume a fixed background connection. These cases can still be accommodated within our framework by extending the relevant bundle $E$ to include the prescribed field, and modifying the corresponding criteria $\mathcal{E}$ by adding terms with very large weights forcing those fields to be in the solution. This adjustment effectively encodes the choice of background structure into the formalism.

Passive transformations: relabelings

A passive transformation is a change in the descriptive language used to model the system. It does not alter the physical content of the theory, only its representation. There are two types: relabeling the spacetime $M$ (coordinate changes), and relabeling the target space of the fields (frame changes, also known as gauge transformations). Of course, these relabelings change the description of the criteria $\mathcal{E}$, and this was a reason for controversy when coordinates were not distinguished from the objects themselves. This was the origin of the Kretschmann objection: any theory can be converted into general passive covariant if we introduce enough mathematical objects codified as sections of a certain bundle.

Coordinate changes

This type of relabeling concerns the base space $M$. In Example 1, since the space consists of a single point, there are no nontrivial coordinates to change, apart from renaming point $p$.

We turn instead to Example 2. Let's introduce a new coordinate system $\psi =(a,b)$ related to the old one by a (passive) rotation:

with $\varphi =\psi \circ {\phi }^{-1}$.

The field $h\in S$ is unchanged, but its description in the new coordinates, $h_{\psi }$, is different:

On the other hand, if we express our selection criterion in the new coordinates, we obtain:

The solution to $\delta \mathcal{E}=0$ in these coordinates is a function that equals 1 if $b=2$ and 0 otherwise. This is precisely the coordinate representation $h_{\psi }$ of the original solution $h$. We have not rotated the line; we have rotated our coordinate grid and found the new description of the original, un-rotated line.

Gauge transformations as frame changes

This relabeling concerns the target space of the fields, i.e., how we trivialize the bundle $E$. This is formalized through the principal bundle $P$ associated to $E$, which plays the role of the frame bundle $FM$ for the tangent bundle $TM$. But we will not get into the details here.

In Example 1, a passive gauge transformation is a change of basis in ${\mathbb{R}}^n$ via an invertible matrix $A\in GL(n,\mathbb{R})$. If we had had more points in $M$, we would have needed an invertible matrix for each of them. A field $x$ is described in the new frame by $\tilde{x}=A^{-1}x$.

Then, the variational principle $\mathcal{E}$ expressed in terms of $\tilde{x}$ becomes:

That is,

Since $\frac{1}{2}\parallel b{\parallel }^2$ is independent of $\tilde{x}$, minimizing $\mathcal{E}$ is equivalent to

The minimizer in this frame is given by

which corresponds, via $x=A\tilde{x}$, to

as expected. So the selected field is the same, merely described in new coordinates.

Similarly, we can relabel the target space in Example 2. Since the bundle $E$ is a $GL(1)$-bundle, we can consider a different trivialization: we can choose in each fibre $E_p$, $p\in M$, the basis $b(p)\ne 0$, instead of the canonical frame $e(p)=1$, $p\in M$ (this corresponds to a gauge transformation in the corresponding principal bundle or, in other words, we are using a different moving frame).

For instance, suppose $b(p)=3,p\in M$. Given a field $f$, described in $\phi$-coordinates by $f_{\phi }$, with the new moving frame it will be described by ${\tilde{f}}_{\phi }=\frac{1}{3}{f}_{\phi }$. So the new description for the criteria $\mathcal{E}$ takes the form:

Obviously, the solution for this functional is

which is the transformed version of the description of the distinguished $h$ in Example 2.

Remark. When the bundle $E$ is a natural bundle (tangent bundle, cotangent bundle, tensor bundle, ...), a single spacetime diffeomorphism $\varphi$ carries out two simultaneous relabelings: it reassigns each point's coordinates on the base manifold, and, via the Jacobian of $\varphi$, it reassigns the local frame (fiber basis) in exactly the way a gauge frame change would. Only in this passive viewpoint, and only for bundles that are naturally tied to the base (tangent, tensor, etc.), does one diffeomorphism deliver both a base-point relabeling and an internal (frame) relabeling.

Active transformations and symmetries

An active transformation is a genuine operation on the space of fields

which sends one field configuration to another (possibly physically distinct) one. Two common sources of active transformations are diffeomorphisms and gauge transformations.

Given a diffeomorphism $F:M\to M$, one obtains a map

by pulling back (or pushing forward) the fields along $F$, together with any required action on the fibres. Concretely:

1. If $E\to M$ is a natural bundle (for example the tangent bundle, tensor bundles, differential forms, etc.), then $F$ canonically lifts to a bundle automorphism $\tilde{F}:E\to E$ covering $F$. The induced action on a field $\varphi \in \mathcal{S}$ is

   $$T_F(\varphi )={\tilde{F}}^{\ast }(\varphi )={\tilde{F}}^{-1}\circ \varphi \circ F.$$

2. If $E\to M$ is not natural (for instance a nontrivial principal $G$-bundle in Yang-Mills theory, or the spinor bundle in fermionic theories), then a diffeomorphism $F$ does not by itself act on sections. One must choose a lift $\tilde{F}$ of $F$ to a bundle automorphism (equivalently, a choice of gauge-transformation-valued map over $F$). Once such a lift is specified, the active transformation is again

   $$T_F(\varphi )={\tilde{F}}^{-1}\circ \varphi \circ F,$$

   where $\tilde{F}$ provides the necessary $G$- or $\mathrm{Spin}(n)$-rotation in the fibre.

In every case, $T_F$ defines a well-defined map $\mathcal{S}\to \mathcal{S}$, carrying any field configuration to its pulled-back (and appropriately rotated) version under the diffeomorphism $F$.

The situation is analogous for gauge transformations. A gauge transformation is typically defined as an automorphism of the principal bundle $P$ associated with the vector bundle $E$. Such a transformation induces a fiberwise action on sections of $E$, i.e., a transformation $T_g:\mathcal{S}\to \mathcal{S}$, where $g$ denotes a section of the associated gauge group bundle $\mathrm{Aut}(P)$. Concretely, if $g:M\to G$ is a gauge transformation valued in the structure group $G$, then for a field $\varphi \in \mathcal{S}$, the transformed field is

where $\cdot$ denotes the representation of $G$ on the fiber. This constitutes an active deformation of the field content: rather than changing the frame of reference, we have changed the physical field itself pointwise across $M$.

Also, observe that given an active transformation $T:\mathcal{S}\to \mathcal{S}$ (either coming from a diffeomorphism or a gauge transformation) we can obtain a new selection mechanism $\tilde{\mathcal{E}}$ by means of the expression

Remark. It is important to clarify that, at the level of points, a general transformation, such as a diffeomorphism $F:M\to M$, can often be interpreted in two ways: as a global passive transformation (a change of coordinates) or as an active transformation (a physical deformation of the fields on the manifold). The same ambiguity applies to gauge transformations: they may represent a mere change in description (passive), or they can act as active transformations on the fields themselves. This distinction is closely analogous to what happens in linear algebra. A change of basis is a passive transformation: it alters the coordinate description of vectors while leaving the vectors themselves unchanged. In contrast, an invertible linear transformation maps vectors to different vectors and is an active transformation. Despite their formal similarities (e.g., both are described by invertible matrices), their interpretations are conceptually distinct. See also here.

Diffeomorphisms as active transformations

We first consider transformations acting on the base space $M$. Again, the case of Example 1 is trivial, since the only possible active transformation of the base space is the identity.

On the other hand, in our toy Example 2, let $F:M\to M$ be, for instance, the active rotation that corresponds to our previous passive coordinate change:

This diffeomorphism induces a new variational principle $\tilde{\mathcal{E}}$:

which takes the form

This is a fundamentally different theory. Its solution set $\tilde{S}=\{\tilde{h}\}$ contains a single field whose coordinate representation is

The original theory selected a vertical line at $x=2$; this new theory selects a horizontal line at $y=2$. The active diffeomorphism has changed the physics.

Active gauge transformations

Next, consider transformations acting on the target space of the fields. An active gauge transformation maps every field $f$ to a new field $f^{\prime }$.

For instance, in Example 1 we can interpret the same change $A\in GL(n,\mathbb{R})$ as a transformation of the field itself, rather than its description. That is, for every field $x$ we define a new field

and construct the new variational principle:

The new minimizer is then

which is clearly different from the original minimizer $x=b$. That is, under the particular active transformation $x\mapsto Ax$, the set of selected fields has changed.

If we now think of Example 2, we can define a transformation that multiplies any field by a constant, $f\mapsto 3f$. Applying this transformation to our toy theory's selection mechanism means we define a new action $\tilde{\mathcal{E}}[f]=\mathcal{E}[3f]$. Expressed in coordinates, this is:

This new action defines a different theory with a different solution. In the original theory we had a single solution $h$ supported on $x=2$ with value 1, but now we have a new solution $\tilde{h}$, whose coordinate representation is

The definition of symmetry

The previous examples show that an active transformation generally produces a new physical theory. This leads to the crucial definition of a symmetry: an active transformation is a symmetry of a theory if it does not change the selection mechanism $\mathcal{E}$ of the theory.

Definition (Symmetry). An active transformation $T:\mathcal{S}\to \mathcal{S}$ is a symmetry of a theory if it leaves the set $S$ of selected fields invariant:

This happens, in particular, if the variational principle $\mathcal{E}$ is invariant under the transformation, i.e.,

since this invariance implies that if $f\in S$ (is a solution), then $T(f)$ must also be in $S$. A symmetry is, then, a permutation of the solution set.

This single definition encompasses both cases:

• A diffeomorphism symmetry is a diffeomorphism $F$ such that the action is invariant under the corresponding transformation of the fields. In Example 2, a translation along the y-axis, $F(x,y)=(x,y+c)$, is a diffeomorphism symmetry. General covariance in General Relativity (Example 4) is the statement that the Einstein-Hilbert action is invariant under any diffeomorphism, meaning all diffeomorphisms are symmetries of the theory. This is linked to the fact that the relevant bundle is a natural bundle.

• A gauge symmetry is an active gauge transformation that leaves the action invariant. These symmetries form the gauge group of the theory. For example, in Example 1, the transformation $x\mapsto Ax$ is a gauge symmetry if $A\in O(n)$ and $Ab=b$, since in that case:

  $$\tilde{\mathcal{E}}[x]=\frac{1}{2}\parallel Ax-b{\parallel }^2-\frac{1}{2}\parallel b{\parallel }^2=\frac{1}{2}\parallel x-b{\parallel }^2-\frac{1}{2}\parallel b{\parallel }^2=\mathcal{E}[x].$$

  In the case of Example 2, the gauge symmetry group consists of the group of active transformations $f\mapsto b\cdot f$ where the function $b(x,y)$ is restricted to be exactly $1$ on the line $x=2$, and can take the values $\pm 1$ everywhere else:

  $$b(x,y)=\{\begin{array}{ll}1,& x=2\\ \pm 1,& x\ne 2\end{array}$$

  This is seen by imposing the symmetry condition $\mathcal{E}[f]=\mathcal{E}[f/b]$ for all fields $f$. The term in the action weighted by $(x-2{)}^2$ immediately forces $b(x,y{)}^2=1$ for all $x\ne 2$. Simultaneously, the Dirac delta term, which constrains the fields on the line $x=2$, requires that $(f/b-1{)}^2=(f-1{)}^2$ on that line. For this to hold for arbitrary field values $f$, we must have $b(x,y)=1$ when $x=2$.

The gauge principle

The previous sections might suggest a one-way process: we are given a theory via an action $\mathcal{E}$, and we then find its symmetry group. However, the paradigm of modern physics often reverses this procedure. The gauge principle is the powerful idea that we can postulate a symmetry group first, based on physical principles or experimental evidence, and then demand that our action $\mathcal{E}$ be invariant under that group.

Historically, understanding a force meant directly observing interactions (e.g., magnets and iron filings, falling balls). But this empirical method fails at the subatomic scale because particles like quarks and gluons can’t be manipulated or observed directly.

So particle physicists reversed the process: they guess the theory first, then test it against nature. The tool for making educated guesses is symmetry—asking what transformations leave physics unchanged. Symmetries constrain theories so tightly that the theory nearly writes itself.

This insight underpins the Standard Model: forces and their interactions are derived from symmetry groups (via group theory), not from direct observation. Each symmetry group corresponds to a force; representations of the group determine which particles experience that force. Group theory thus replaces the classical method of “pushing things together.”

This principle is the foundation of the Standard Model, where demanding local gauge invariance under $SU(3)\times SU(2)\times U(1)$ forces the existence of the gluon, W, Z, and photon fields and dictates the form of their interactions with matter, not direct observation. Similarly, demanding that physics be independent of the choice of local inertial frame (invariance under the diffeomorphism group) leads to General Relativity. In this sense, symmetry is not an accidental property of our theories, but the very principle from which they are constructed.

Conclusion

We have formalized the distinction between passive relabelings and active transformations. Passive transformations merely alter our description, while active transformations alter the physical system. A symmetry is an active transformation that leaves the theory's selection mechanism (its action) invariant, and therefore preserves its set of solutions. This viewpoint clarifies that general covariance and gauge invariance are not descriptive redundancies, but profound physical symmetries. The ultimate expression of this idea is the gauge principle, which uses symmetry as a constructive tool, making it one of the most powerful and successful concepts in all of theoretical physics.

Related: classical field#Transformations of fields.
Maybe related: general covariance and contravariance.
Related: effective field theories.