Given a Lie group $G$, whose most general transform depends on $\rho$ parameters, under the action of which an integral $I$ is invariant, there are $\rho$ linearly independent combinations of the Lagrange expressions which become divergences.

I know what they mean by a Lagrange expression: namely, $\frac{\partial L}{\partial \phi}-\sum \frac{d}{dx_i}\frac{\partial L}{\partial\frac{d\phi}{dx_i}}.$

  1. But what is the significance of the "linearly independent combinations of Lagrange expressions"?

  2. Also, I don't get how that expression is equivalent to $\frac{d\rho^0}{dt}+\nabla \cdot \vec{\rho}$ given that $\rho^{\mu}=[\rho^0,\vec{\rho}]$. Namely, what is $\rho$?

  3. Also, if my system follows Hamilton's principle, then how can that be a divergence? Namely, it should become zero, not infinity.

As you might have noticed, this is beyond my level. But I have to do a presentation on it, so please bear with me.

I've gotten my information from here: http://inside.mines.edu/~tohno/teaching/PH505_2011/Ryan_FinalPaperNoetherThm.pdf


First of all, my opinion is that the paper on your link is full of notational inconsistencies and therefore causes a great amount of confusion for someone who struggles to understand Noether’s theorem. So, allow me to formulate the field-theoretic version of Noether’s theorem in a more, according to my taste, charming way.

Preliminaries 1: Lie Groups and Lie algebras

Each element of an $N$-dimensional lie Group $G$ can be parameterized by a point of a subspace of ${{\mathbf{R}}^{N}}$. In other words, an element $g\in G$ can be considered as a map: $g:{{\mathbf{R}}^{N}}\backepsilon \omega \mapsto {{g}_{\omega }}\in G$. Without loss of generality, we can assume that ${{g}_{0}}=e$ , the identity element of $G$. Now suppose that we have an $n$-dimensional real vector field:

$\varphi :\Omega \backepsilon x\mapsto \varphi \left( x \right)\in {{\mathbf{R}}^{n}}$

where $\Omega \subseteq {{\mathbf{R}}^{\left( 1,3 \right)}}$, with $n$ real components ${{\varphi }_{i}}\ ,\ i=1,..,n$, such that it obeys the boundary condition ${{\left. \varphi \right|}_{\partial \Omega }}=0$. Also, assume that a group element ${{g}_{\omega }}$ is implemented as an $n$-dimensional representation of $G$. We can now consider the linear transformation:

${{g}_{\omega }}:\varphi \to {{g}_{\omega }}\varphi \quad ,\quad {{\left( {{g}_{\omega }}\varphi \right)}_{i}}:=\sum\limits_{j=1}^{n}{{{\left( {{g}_{\omega }} \right)}_{ij}}{{\varphi }_{j}}}$

If $\varepsilon \in {{\mathbf{R}}^{N}}$ is a point infinitesimally close to $0$ then ${{g}_{\varepsilon }}$ will be a group element that is “infinitesimally close” to the identity $e$, in the sense that:

${{g}_{\varepsilon }}\varphi ={{g}_{0}}\varphi +\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{\left[ \tfrac{\partial }{\partial {{\omega }_{I}}}{{g}_{\omega }}\varphi \right]}_{\omega =0}}}+\mathcal{O}\left( {{\varepsilon }^{2}} \right)=\varphi +\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{T}_{I}}\varphi }+\mathcal{O}\left( {{\varepsilon }^{2}} \right)$

where ${{T}_{I}}\varphi :={{\left[ \tfrac{\partial }{\partial {{\omega }_{I}}}{{g}_{\omega }}\varphi \right]}_{\omega =0}}$. The vectors ${{T}_{I}}$ form a basis of the $n$-dimensional representation of the Lie Algebra $\mathrm{}$ of $G$ and they transform the field as:

${{T}_{I}}:\varphi \to {{T}_{I}}\varphi \quad ,\quad {{\left( {{T}_{I}}\varphi \right)}_{i}}=\sum\limits_{j=1}^{n}{{{\left( {{T}_{I}} \right)}_{ij}}{{\varphi }_{j}}}$

Preliminaries 2: Lagrangian formulation and Hamilton’s Principle

The dynamics of the field $\varphi \left( x \right)$ can be encoded in the “Lagrangian density” function $\mathsf{L}\left( \varphi ,{{\partial }_{\mu }}\varphi \right)$, which is a function such that the “Euler-Lagrange equations”:

$\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}-{{\partial }_{\mu }}\left( \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}} \right)=0\quad ,\quad i=1,...,n$

are equivalent to the equations of motion of the field’s components. Having in our disposal $\mathsf{L}\left( \varphi ,{{\partial }_{\mu }}\varphi \right)$ we can restate the equations of motion as variational principle, the “Principle of Least Action” (“Hamilton’s principle”) as follows:

i) Define the “Action functional” as:

$S\left[ \varphi \right]:=\int\limits_{\Omega }{\mathsf{L}\left( \varphi ,{{\partial }_{\mu }}\varphi \right){{d}^{4}}x}$

ii) The functional derivative of $S\left[ \varphi \right]$ w.r.t. the field component ${{\varphi }_{i}}$ is a Lagrange expression, i.e.:

$\frac{\delta S}{\delta {{\varphi }_{i}}}=\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}-{{\partial }_{\mu }}\left( \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}} \right)$

where for the derivation of this result the boundary condition ${{\left. \varphi \right|}_{\partial \Omega }}=0$ was been taken into account.

iii) By the above we deduce that if the field is a stationary point of $S\left[ \varphi \right]$, i.e. if:

$\frac{\delta S}{\delta {{\varphi }_{i}}}=0\quad ,\quad i=1,...,n$

then the Euler-Lagrange equations are satisfied and vice-versa. Concluding, the field’s equations of motion can be derived by a variational principle, the “Principle of Least Action”. This is “Hamilton’s principle”.

Definition of symmetry Group

The Lie group $G$ is a “symmetry group” for the theory of the field $\varphi $ if the action functional $S\left[ \varphi \right]$ is invariant (more generally, if it differences by a boundary term ${{S}_{\partial \Omega }}$) under the action of any ${{g}_{\omega }}\in G$, i.e. if: $S\left[ {{g}_{\omega }}\varphi \right]=S\left[ \varphi \right]$ For sake of simplicity, we will restrict to the case where the Lagrangian density is invariant:

$\mathsf{L}\left( {{g}_{\omega }}\varphi ,{{\partial }_{\mu }}\left( {{g}_{\omega }}\varphi \right) \right)=\mathsf{L}\left( \varphi ,{{\partial }_{\mu }}\varphi \right)$

which is a sufficient condition for the invariance of the action. Let’s investigate the case where ${{g}_{\omega }}$ is an infinitesimal transformation. Then the field components would vary as:

${{\delta }_{\varepsilon }}\varphi :={{g}_{\varepsilon }}\varphi -\varphi =\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{T}_{I}}\varphi }$

and their spatial derivatives as:

${{\delta }_{\varepsilon }}{{\varphi }_{,\mu }}:={{\partial }_{\mu }}\left( {{g}_{\varepsilon }}\varphi \right)-{{\partial }_{\mu }}\varphi ={{\partial }_{\mu }}\left( \varphi +\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{T}_{I}}\varphi } \right)-{{\partial }_{\mu }}\varphi ={{\partial }_{\mu }}\left( {{\delta }_{\varepsilon }}\varphi \right)$

Then the Lagrangian would vary as:

$\begin{align} & {{\delta }_{\varepsilon }}\mathsf{L}:=\mathsf{L}\left( {{g}_{\varepsilon }}\varphi ,{{\partial }_{\mu }}\left( {{g}_{\varepsilon }}\varphi \right) \right)-\mathsf{L}\left( \varphi ,{{\partial }_{\mu }}\varphi \right)=\sum\limits_{i=1}^{n}{\left[ \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}{{\delta }_{\varepsilon }}{{\varphi }_{i}}+\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\delta }_{\varepsilon }}{{\varphi }_{i,\mu }} \right]} \\ & \quad \quad =\sum\limits_{i=1}^{n}{\left[ \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{\left( {{T}_{I}}\varphi \right)}_{i}}}+\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}{{\partial }_{\mu }}{{\left( {{T}_{I}}\varphi \right)}_{i}}} \right]} \\ & \quad \quad =\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}\sum\limits_{i=1}^{n}{\left[ \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}{{\left( {{T}_{I}}\varphi \right)}_{i}}+\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\partial }_{\mu }}{{\left( {{T}_{I}}\varphi \right)}_{i}} \right]}} \\ & \quad \quad =\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}\sum\limits_{i=1}^{n}{\left[ \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i}}}{{\left( {{T}_{I}}\varphi \right)}_{i}}+{{\partial }_{\mu }}\left[ \frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\left( {{T}_{I}}\varphi \right)}_{i}} \right]-{{\partial }_{\mu }}\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\left( {{T}_{I}}\varphi \right)}_{i}} \right]}} \\ & \quad \quad =\sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}\left( \sum\limits_{i=1}^{n}{\frac{\delta S}{\delta {{\varphi }_{i}}}{{\left( {{T}_{I}}\varphi \right)}_{i}}+{{\partial }_{\mu }}J_{I}^{\mu }} \right)} \\ \end{align}$


$J_{I}^{\mu }:=\sum\limits_{i=1}^{n}{\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\left( {{T}_{I}}\varphi \right)}_{i}}}=\sum\limits_{i,j=1}^{n}{\frac{\partial \mathsf{L}}{\partial {{\varphi }_{i,\mu }}}{{\left( {{T}_{I}} \right)}_{ij}}{{\varphi }_{j}}}$

So, the invariance of the Lagrangian density is expressed as:

${{\delta }_{\varepsilon }}\mathsf{L}=0\Rightarrow \sum\limits_{I=1}^{N}{{{\varepsilon }_{I}}\left( \sum\limits_{i=1}^{n}{\frac{\delta S}{\delta {{\varphi }_{i}}}{{\left( {{T}_{I}}\varphi \right)}_{i}}+{{\partial }_{\mu }}J_{I}^{\mu }} \right)=0}$

and since ${{\varepsilon }_{I}}$’s are considered independent parameters, then the above expression holds iff:

$\sum\limits_{i=1}^{n}{\frac{\delta S}{\delta {{\varphi }_{i}}}{{\left( {{T}_{I}}\varphi \right)}_{i}}}+{{\partial }_{\mu }}J_{I}^{\mu }=0\quad ,\quad I=1,..,N$

This is Noether’s first theorem! Now observe that if the fields are such that they obey the equations of motion (are “on-shell” in the physics jargon), i.e. if $\tfrac{\delta S}{\delta {{\varphi }_{i}}}=0$ then:

${{\partial }_{\mu }}J_{I}^{\mu }=0\quad ,\quad I=1,..,N$

i.e. that the currents $J_{I}^{\mu }$ are locally conserved. This is the essence of Noether’s theorem: it states that for any symmetry of a physical theory there is a corresponding conserved current; in my opinion it is one of the most beautiful theorems in physics. The above conservation law, which is in covariant form, can be re-expressed as:

${{\partial }_{\mu }}J_{I}^{\mu }=0\Rightarrow {{\partial }_{t}}J_{I}^{0}=-\nabla \cdot {{\mathbf{J}}_{I}}\Rightarrow {{\partial }_{t}}\int\limits_{{{\Omega }_{t}}}{J_{I}^{0}{{d}^{3}}x}=-\int\limits_{{{\Omega }_{t}}}{\left( \nabla \cdot {{\mathbf{J}}_{I}} \right){{d}^{3}}x}$

where ${{\Omega }_{t}}$ is an iso-temporal sub-surface of $\Omega $. Using Gauss’ divergence theorem we get:

${{\partial }_{t}}{{Q}_{I}}=-\int\limits_{\Sigma \equiv \partial {{\Omega }_{t}}}{{{\mathbf{J}}_{I}}\cdot d\,\mathbf{\Sigma }}$


${{Q}_{I}}:=\int\limits_{{{\Omega }_{t}}}{J_{I}^{0}{{d}^{3}}x}$

Assuming a “Dirichlet boundary condition”:

${{\left. {{\mathbf{J}}_{I}} \right|}_{\partial {{\Omega }_{t}}}}=0$

or a “Neuman boundary condition”:

${{\mathbf{J}}_{I}}\left( x \right)\cdot \mathbf{n}\left( x \right)=0\ ,\ \forall x\in \partial {{\Omega }_{t}}\quad |\quad d\mathbf{\Sigma }=\mathbf{n}d\Sigma $


$\int\limits_{\Sigma \equiv \partial {{\Omega }_{t}}}{{{\mathbf{J}}_{I}}\cdot d\mathbf{\Sigma }}=0$

so the ${{J}_{I}}$’s current conservation implies that the “charge” ${{Q}_{I}}$ is constant in time, i.e. that it is conserved.

  • $\begingroup$ Surprisingly I understood almost everything in this post. I guess I'll look pretty smart today when giving my presentation :P $\endgroup$ – DLV Sep 25 '15 at 17:53
  • $\begingroup$ By the way are each $\phi_i$ different fields, or are they a component of one field? Also what do you mean by $R^{(1,3)}$ being contained in $\Omega$ ? $\endgroup$ – DLV Sep 25 '15 at 17:54
  • $\begingroup$ I am glad I could held. φi's are the different components of the field φ (e.g. the electric field E has three independend components E1, E2 and E3). And Ω is a subpsace of the Minkowski space, where the field is confined; it could as well be the entire Minkowski space $\endgroup$ – user3257624 Sep 25 '15 at 17:58
  • $\begingroup$ Hmm. Why use different subscripts $\mu$ and $i$ then? I mean, shouldnt $\partial_\mu \phi_i = 0$ if $\mu\neq i$? I guess I'll study this post a bit more then. $\endgroup$ – DLV Sep 25 '15 at 18:04
  • 1
    $\begingroup$ No! The index μ=0,1,2,3 refers to the space-time variable of the field while i=1,...,n to the field component. In our example of the electric field, the xμ would refer to the spacetime point where each field component Ei(x0,x1,x2,x3) is evaluated (here i=1,2,3) $\endgroup$ – user3257624 Sep 25 '15 at 18:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.