Mechanics II: Symmetry and Conservation Laws

Mechanics II: Symmetry and Conservation Laws

In the previous article, we developed the basic framework of classical mechanics from the principle of least action. By varying the action, we obtained the Euler–Lagrange equations. We then used the properties of inertial frames to determine the Lagrangian of a free particle. This led us to momentum, potential energy, and force, and we derived Newton’s three laws in turn from the equations of motion. Along the way, we began to encounter conservation laws. The Lagrangian of a free particle has no explicit dependence on its position coordinates, so its momentum is conserved. Going from a single particle to a closed system, we found that the total momentum of the system is also conserved. These results bring us back to the question left open in the previous article. To see where conservation laws come from, we need to look beyond particular equations of motion and turn to symmetry. From that perspective, conservation laws that once seemed unrelated reveal a common origin.

Symmetry and Conservation Laws

However broadly or narrowly we define symmetry, it “is one idea by which man through the ages has tried to comprehend and create order, beauty, and perfection.”

— Symmetry [1]

In everyday language, “symmetry” often suggests balance, harmony, and beauty. It is widespread in nature and has influenced art and architecture as an aesthetic principle. In physics, we can describe symmetry in terms of transformations. For example, rotate a square by (90^\circ) about its centre or reflect it across one of its axes of symmetry, and the resulting figure coincides with the original. In other words, symmetry is the property of remaining unchanged under a transformation. The laws of nature can also exhibit this kind of invariance. In an inertial frame, for example, a free particle moves in a straight line at constant velocity. If we translate its entire trajectory in space or change the origin of time, the equations describing its motion retain the same form.

Mathematically, we define symmetry in terms of a transformation. Consider the field φ(x) under the coordinate transformation xa→xa−δxa .

We use a minus sign in the coordinate transformation so that the resulting expression for δφ has a plus sign. This discussion draws on Tong’s lecture notes [2].

The field value originally at x moves to x′ under the translation, so φ′(x′)=φ(x) . To evaluate the transformed field at the coordinate (x), we return to the original field at x+δx .

φ′(x)=φ(x+δx)

We can use this to calculate the change in the field.

δφ(x)=φ(x+δx)−φ(x)=δxa∂aφ+O(δx2)

To first order, this gives

δφ(x)=δxa∂aφ

To establish whether a transformation is a physical symmetry, we ask whether the physical laws remain unchanged under it. As discussed in the previous article, if a transformation adds only a total time derivative to the Lagrangian, it adds only an endpoint term to the action. With the endpoints held fixed, this term does not affect the variation of the action, so the equations of motion remain unchanged.

L(q˙+ϵ)−L(q˙)=ddtf(q,t)

This is the infinitesimal change δL in the Lagrangian. We can generalise the total time derivative to a total divergence. If the first-order change in the Lagrangian density under the transformation satisfies

δL=∂afa

then the action changes by at most a boundary term. The symmetries discussed below satisfy this condition.

Let the action over the spacetime region M be

S[φ]:=∫MdnxL[φ(x),∂aφ(x),x]

The symmetry condition gives the first-order change in the action.

δ∫MdnxL=∫MdnxδL=∫Mdnx∂afa

We can also vary the action directly.

δ∫MdnxL=∫MdnxδL=∫Mdnx[∂L∂φδφ+∂L∂(∂aφ)∂aδφ]=∫Mdnx[(∂L∂φ−∂a∂L∂(∂aφ))δφ+∂a(∂L∂(∂aφ)δφ)]

When the field satisfies the Euler–Lagrange equations, the first term vanishes, leaving the second term.

δ∫MdnxL=∫Mdnx∂a(∂L∂(∂aφ)δφ)

Comparing the two expressions pointwise gives

∫Mdnx∂a(∂L∂(∂aφ)δφ−fa)=0

We can then define the Noether current.

∂aJa=0Ja:=∂L∂(∂aφ)δφ−fa

Once a coordinate system is chosen, we can write the equations in components.

This is Noether’s theorem. It states that every continuous global variational symmetry of the action has an associated Noether current whose divergence vanishes when the field satisfies its equations of motion. Noether’s theorem is a central result of theoretical physics. It tells us that every continuous symmetry has a corresponding conservation law [3].

Emmy Noether was a German mathematician. Several prominent scholars have described her as the greatest woman mathematician in history. In physics, the theorem she proved revealed the close connection between symmetries and conservation laws. It is one of the most important theorems in the development of modern physics. [4]

Noether’s theorem gives the continuity equation ∂μJμ=0 . Integrating it over the spacetime region M gives

∫Mdnx∂μJμ=0

Let M=[t1,t2]×Σ , where the spatial region Σ does not change with time. We can therefore separate the time and space integrals. Taking the time derivative outside the spatial integral and applying the divergence theorem to the spatial divergence gives

∫t1t2dt∫Σdn−1x(∂0J0+∂iJi)=∫t1t2dt[∂0∫Σdn−1xJ0+∫Σdn−1x∂iJi]=∫t1t2dt[ddt∫Σdn−1xJ0+∫∂ΣdSiJi]

Here dSi is oriented along the outward normal. Since t1 and t2 can be chosen arbitrarily, the expression in square brackets vanishes at every instant.

ddt∫Σdn−1xJ0+∫∂ΣdSiJi=0

If the net flux through the boundary ∂Σ vanishes, then

dQdt=ddt∫Σdn−1xJ0=0Q:=∫Σdn−1xJ0

Q is the conserved charge associated with the symmetry.

A simple example can help us understand conserved currents and charges. Imagine a house with three rooms, A , B , and C , each connected to the other two. People can move between the rooms, and no one enters or leaves the house. Over a given interval, count the crossings from A to B and subtract the crossings from B to A . Call the net count Jμ . Similarly, let Jν and Jσ be the net crossing counts from B to C and from C to A , respectively. The change in the number of people in each room equals the number of inward crossings minus the number of outward crossings. Dividing by Δt gives the average rate of change in each room’s population.

ΔNAΔt=Jσ−JμΔtΔNBΔt=Jμ−JνΔtΔNCΔt=Jν−JσΔt

Adding the three equations, each net crossing count appears as an outflow for one room and an inflow for another, so the terms cancel. If N is the total number of people in the house, we obtain

ΔNΔt=0.

In this analogy, the time component J0(i) of the Noether current corresponds to the number of people in room i at a given instant.

J0(i)={NAi=ANBi=BNCi=C

The spatial components of the Noether current describe flow across a boundary. Here Jμ , Jν , and Jσ correspond to the accumulated net flux through the respective doorways over the interval of observation. The spatial integral in the continuous case becomes a sum over the three rooms. The resulting conserved charge Q=NA+NB+NC is the total number of people in the house.

Conservation Laws

Examples from Classical Mechanics

Consider a few examples from classical mechanics. We begin with spatial translations. In the chosen inertial frame, set x0=t and write the translation as xμ→xμ+aμ , with aμ constant. Since this is a purely spatial translation, a0=0 , so δxμ=aμ and δt=0 . A spatial translation leaves the velocity unchanged, so δx˙i=0 . Only the position term contributes to the first-order change in the Lagrangian, which satisfies the symmetry condition.

δL=aν∂νL=∂ν(aνL)fμ:=aνδ νμL

In classical mechanics, the coordinate x takes the place of the field φ , and the current has only the component μ=0 . To construct the conserved current, let i=1,2,3…;i≠0 .

J0=∂L∂x˙iai−aiδ i0L=∂L∂x˙iai

Since ai is constant, conservation of J0 means that the component of momentum along the direction of translation is conserved. A free particle has translational symmetry in all three independent spatial directions, so each momentum component satisfies

ddtpi=ddt∂L∂x˙i=0

We can use the same method to examine spatial rotations. If we choose the angle θ about a fixed axis as a coordinate, a rotation acts as a translation of θ . As with a spatial translation, the corresponding conserved quantity takes the form of a conjugate momentum, pθ=∂L/∂θ˙ . This is the component of angular momentum along the rotation axis. To see how angular momentum depends on position and momentum, we now switch to Cartesian coordinates. In these coordinates, a rotation of the position vector is written as

x′=R(θ)x

For an infinitesimal rotation, write the rotation angle as a vector δθ pointing along the rotation axis. The first-order change in position can be written using the Levi-Civita symbol.

δxμ=ϵμνσδθνxσ

Strictly speaking, μ should be an upper index. In Euclidean space, we can use the metric to raise or lower it, giving ϵ νσμ=ϵμνσ .

Suppose this spatial rotation satisfies the symmetry condition. Substituting into the expression for the conserved current gives

J0=∂L∂x˙iδxi−δxiδ i0L=∂L∂x˙iδxi

Equivalently, we have

J0=piϵiνσδθνxσ:=Lνδθν

Since δθν is constant, the continuity equation gives

ddtLν=0

Writing out the individual components gives

L1=x2p3−x3p2L2=x3p1−x1p3L3=x1p2−x2p1

Putting the components together in vector form gives conservation of the total angular momentum.

L=x×p=const.

We can examine time-translation symmetry in much the same way. As in the spatial case, we now consider the translation t0→t0+τ0 .

J0=∂L∂x˙ix˙iτ0−Lτ0ddt[∂L∂x˙ix˙i−L]=0

The conserved quantity associated with time translations is energy. As an example in one dimension, suppose a particle moves in a potential V(x) , with Lagrangian

L=12mx˙2−V(x)

Substituting into the expression for energy gives

E=∂L∂x˙ix˙i−L=mx˙2−12mx˙2+V(x)=12mx˙2+V(x)

The particle’s kinetic and potential energies add up to a constant. This is the familiar law of conservation of mechanical energy.

Energy–Momentum Tensor

Canonical Energy–Momentum Tensor

In the previous section, we considered spatial and temporal translations separately in classical mechanics. We now apply the same idea to field theory, treating both as spacetime translations. Following the earlier convention, take x′a=xa−ξa , where ξa is constant. When the Lagrangian density has no explicit dependence on the spacetime coordinates, the first-order changes in the field and its Lagrangian density are

δL=ξa∂aL=∂a(ξaL)δφ=ξa∂aφ

The first-order change in the Lagrangian density is a total divergence. Substituting the variation of the field and the corresponding boundary term into the earlier expression for the Noether current gives

Ja=[∂L∂(∂aφ)∂bφ−δ baL]ξb

Each choice of translation direction gives an associated Noether current. Collecting the conserved currents for translations in all directions gives the canonical energy–momentum tensor.

Ja=Tabξb=TabηbcξcTab:=∂L∂(∂aφ)∂bφ−ηabL

Since ξa is constant and can be chosen independently along each spacetime direction, we have

∂a(Tabξb)=ξb∂aTab=0∂aTab=0

This is the conservation equation for the canonical energy–momentum tensor. We now examine its components in four-dimensional Minkowski spacetime, with ημν=diag(1,−1,−1,−1) .

The convention ημν=diag(−1,1,1,1) is also possible. We use the former metric signature to raise and lower indices throughout the discussion below.

Consider the component T00 . Since ∂0=∂0 , we have

T00=∂L∂(∂0φ)∂0φ−L=∂L∂φ˙φ˙−L

This expression gives the energy density of the field. Integrating over all of space Σ at a fixed time gives the total energy.

E=∫Σd3xT00

Next, consider T0i . Here ∂i=−∂i , so

T0i=∂L∂(∂0φ)∂iφ=−∂L∂φ˙∂iφ

Its physical meaning can be seen from the time component of the conserved current.

J0=T0μξμ

A pure time translation selects T00 , which is the energy density. A pure spatial translation along the i th coordinate direction selects T0i , the i th component of the momentum density. Integrating over all of space gives the corresponding component of the total momentum.

Pi=∫Σd3xT0i

Next, consider Ti0 .

Ti0=∂L∂(∂iφ)φ˙

For a pure time translation, it gives the spatial components of the corresponding conserved current.

Ji=Ti0ξ0

Expanding the time component of the conservation equation for the canonical energy–momentum tensor gives

∂tT00+∂iTi0=0

Since T00 is the energy density, Ti0 is the energy flux density along the i th coordinate direction. We denote its volume integral over the spatial region Σ by

ei=∫Σd3xTi0

Finally, the purely spatial components of the energy–momentum tensor are

Tij=−∂L∂(∂iφ)∂jφ−(−δij)L=δijL−∂L∂(∂iφ)∂jφ

For a pure spatial translation, these components determine the spatial part of the corresponding conserved current.

Ji=Tijξj

Keeping the second index fixed, we can write the conservation equation as

∂tT0j+∂iTij=0

Here T0j is the density of the j th component of momentum. Thus Tij is the flux density of the j th momentum component in the i th direction. For fixed j , we denote the volume integral of this flux density over the spatial region Σ by

pi=∫Σd3xTij

There is another way to interpret Tij . Integrating the conservation equation gives

∫Σd3x(∂tT0j+∂iTij)=0∫Σd3x∂tT0j=−∫Σd3x∂iTijddtPj=−∫Σd3x∂iTij=−∫∂ΣdSiTij

The rate of change of the j th momentum component within the region is the negative of the net outward flux through the boundary. This gives the connection between momentum flux density and stress. If traction is defined in terms of the force exerted by the boundary on the region inside, the stress components take the corresponding sign in the equation above. For a surface whose normal lies along the i th direction, i=j corresponds to normal stress and i≠j corresponds to shear stress.

Component Symmetry Expression Physical meaning
T00 Time translation ∂L∂φ˙φ˙−L Energy density
T0i Spatial translation −∂L∂φ˙∂iφ Momentum density
Ti0 Time translation ∂L∂(∂iφ)φ˙ Energy flux density
Tij Spatial translation δijL−∂L∂(∂iφ)∂jφ Momentum flux density

This section draws on [2] [5].

Energy–Momentum Tensor

We can extend this idea to a manifold (M,gab) equipped with a metric and describe spacetime symmetries in terms of isometries of the metric. Suppose there is a vector field ξa along whose flow the metric remains unchanged, so that

Lξagab=0

We then call it a Killing vector field [6]. The Minkowski spacetime translations considered above provide an example. Given a tensor Tab and a Killing vector field ξa , we can construct

Ja=Tabξb

Taking its covariant divergence gives

∇a(Tabξb)=(∇aTab)ξb+Tab∇aξb

The Killing equation ∇(aξb)=0 tells us that the symmetric part of ∇aξb vanishes. Decomposing the second term into its symmetric and antisymmetric contributions gives

(∇aTab)ξb+Tab∇aξb=(∇aTab)ξb+T(ab)∇(aξb)+T[ab]∇[aξb]=(∇aTab)ξb+T[ab]∇[aξb]

Let us return to the case of flat spacetime. The canonical energy–momentum tensor obtained from spacetime translation symmetry obeys the conservation equation, and the covariant derivative of the constant vector field generating the translation vanishes. Thus

∇aTab=0∇aξb=0

We therefore have a conserved current.

∇aJa=0

In the spacetime translations discussed above, the generating vector field ξa was constant. In inertial coordinates in Minkowski spacetime, ∇aξb=0 , so the conservation equation ∇aTab=0 is enough to give the conserved current Ja=Tabξb . For more general isometries, the generating Killing vector field may have ∇aξb≠0 . As the preceding calculation shows, even when ∇aTab=0 , the divergence of the current may still contain the term

∇aJa=T[ab]∇[aξb]

If this term is nonzero, the current is not conserved. If Tab is symmetric, its antisymmetric part T[ab] vanishes. We therefore seek a new energy–momentum tensor that is symmetric and obeys the original conservation equation. One way to do this is to add a divergence term to the canonical energy–momentum tensor [6]. Introduce a tensor Bcab satisfying

{Bcab=−Bacb∇cBc[ab]=−T[ab]∇a∇cBcab=0

We define the modified canonical tensor and call it the energy–momentum tensor.

Tnewab:=Tab+∇cBcab

First, let us see how the correction makes the tensor symmetric. Taking the antisymmetric part in (a,b) gives

Tnew[ab]=T[ab]+∇cBc[ab]=T[ab]−T[ab]=0

Next, consider the conservation equation. Taking the covariant divergence of the new tensor gives

∇aTnewab=∇aTab+∇a∇cBcab=∇aTab

If the original tensor satisfies ∇aTab=0 , the new tensor does as well. We can then construct a new conserved current.

Jnewa=Tnewabξb

It satisfies

∇aJnewa=0

In this way, for every continuous isometry of the metric generated by a Killing vector field, we can construct the corresponding conserved current from the same improved energy–momentum tensor.

This section draws on [6].

The Significance of Noether’s Theorem

“Pure mathematics is, in its way, the poetry of logical ideas.” Einstein goes on to describe the search for general operations that gather as many formal relationships as possible into a simple, logical whole. In pursuing this logical beauty, mathematicians find conceptual tools that let us probe the laws of nature more deeply.

— Einstein’s tribute to Noether [7]

Noether’s theorem changed how we understand physical laws. It did more than add another result to theoretical physics. Before it, conserved quantities were often treated as empirical facts. The theorem brought them together as consequences of continuous symmetries. Time-translation symmetry corresponds to conservation of energy, spatial translation symmetry to conservation of momentum, and spatial rotation symmetry to conservation of angular momentum. This is how symmetry enters the laws governing conserved quantities in the motion of matter.

Noether’s theorem also changed how physicists build theories. Given a theory, we can start from its symmetries and identify conserved quantities before solving the equations of motion. In developing a new theory, we can reason in the other direction. We first determine what symmetries the theory has, then use them to decide which terms can appear in the action. From explaining conservation laws in classical mechanics to constraining theories of fundamental interactions, symmetry has become a starting point for building physical theories. As a matter of method, Noether’s theorem changed the way physicists work by making the search for symmetries a central part of constructing theories.

References

[1]
H. Weyl, Symmetry. Princeton University Press, 1952. doi: 10.1515/9781400874347.
[2]
D. Tong, “Quantum field theory,” University of Cambridge, Lecture notes, 2006–2007. Available: https://www.damtp.cam.ac.uk/user/tong/qft/one.pdf
[3]
E. Noether, “Invariante Variationsprobleme,” Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-Physikalische Klasse, pp. 235–257, 1918, Available: https://gdz.sub.uni-goettingen.de/id/PPN252457811_1918?tify=%7B%22view%22%3A%22info%22%2C%22pages%22%3A%5B241%5D%7D
[4]
“埃米·诺特.” Accessed: Nov. 14, 2026. [Online]. Available: https://zh.wikipedia.org/wiki/%E5%9F%83%E7%B1%B3%C2%B7%E8%AF%BA%E7%89%B9
[5]
L. D. Landau and E. M. Lifshitz, The classical theory of fields, Fourth revised English edition., vol. 2. in Course of theoretical physics, vol. 2. Oxford; New York: Pergamon Press, 1975. Available: https://www.sciencedirect.com/book/monograph/9780080250724/the-classical-theory-of-fields
[6]
梁灿彬, 周彬., 梁灿彬., and 周彬., 微分几何入门与广义相对论, Di 2 ban. Beijing Shi: 科学出版社, 2006.
[7]
A. Einstein, “The Late Emmy Noether: Professor Einstein Writes in Appreciation of a Fellow-Mathematician,” The New York Times, p. 12, May 04, 1935. Available: https://mathshistory.st-andrews.ac.uk/Obituaries/Noether_Emmy_Einstein/

Mechanics II: Symmetry and Conservation Laws
https://l-lllll-06.github.io/en/mechanics-02/
Author
-L_LLLLL-
Posted on
November 16, 2026
Licensed under