physics2u
PhysicsMathsChemistryBiology
Tier
⌕ Search ⌘K
Derivation

Noether's First Theorem

D-367 Home PU-402 Threads symmetry · energy · fields Depends on Euler-Lagrange Equations for Fields
Statement

For a classical field theory with action \( S = \int d^4x\, \mathcal{L}(\phi_a, \partial_\mu\phi_a, x^\mu) \), suppose an infinitesimal transformation of the fields and coordinates leaves the action invariant up to a boundary term (i.e. maps solutions of the Euler–Lagrange equations to solutions). Then Noether's first theorem asserts that for each independent continuous parameter of the symmetry there exists a current \( j^\mu \) that is conserved on-shell, \( \partial_\mu j^\mu = 0 \), and hence a charge \( Q = \int d^3x\, j^0 \) that is constant in time. We derive the explicit form of \( j^\mu \) and prove its conservation using only the Euler–Lagrange field equations.

Why it matters

Noether's theorem is the deepest organising principle in physics: it converts a geometric statement (the action has a symmetry) into a dynamical statement (something is conserved). Energy, momentum, angular momentum, and electric charge are not independent postulates but consequences of invariance under time translation, space translation, rotation, and internal phase rotation respectively. The theorem tells us which quantity is conserved and gives its precise microscopic expression in terms of the Lagrangian.

It also draws the sharp modern distinction between a mere constant of motion and a locally conserved current. Local conservation, \( \partial_\mu j^\mu = 0 \), is far stronger than global constancy: it forbids charge from disappearing in one place and instantly reappearing elsewhere, and it is the structural input that makes gauge theory and the Standard Model consistent.

Assumptions
The dynamics follow from a local actionWithout a variational principle there is no "invariance of the action" to exploit; the theorem in this form does not apply to systems defined only by non-variational equations of motion (e.g. generic dissipative dynamics).
The Euler–Lagrange field equations hold (the fields are on-shell)The current is conserved only on solutions. Off-shell, \( \partial_\mu j^\mu \) is generally nonzero, so conservation is a property of physical field configurations, not of the Lagrangian alone.
The symmetry is continuous, parametrised by \( \epsilon^a \) that can be taken infinitesimalDiscrete symmetries (parity, time reversal) carry no infinitesimal generator and yield no Noether current; they give multiplicative quantum numbers instead.
The variation of \( \mathcal{L} \) is a total divergence, \( \delta\mathcal{L} = \partial_\mu K^\mu \), for some \( K^\mu \) (quasi-invariance)If \( \delta\mathcal{L} \) is not a total divergence the action is not invariant and no conserved current follows. Allowing \( K^\mu \neq 0 \) is essential: many symmetries (e.g. Galilean boosts) leave \( S \) invariant only up to a boundary term.
The Lagrangian depends on at most first derivatives of the fieldsHigher-derivative theories still admit a Noether current but the Euler–Lagrange operator and the current acquire extra terms; the compact formula below must be extended.
Derivation
1
\[ \delta\phi_a = \phi_a'(x) - \phi_a(x), \qquad \delta\mathcal{L} = \frac{\partial\mathcal{L}}{\partial\phi_a}\,\delta\phi_a + \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta(\partial_\mu\phi_a) \]
Definition of the field variation and the chain rule applied to \( \mathcal{L} \) as a function of \( \phi_a \) and \( \partial_\mu\phi_a \); sum over the field index \( a \) is implied. A
2
\[ \delta(\partial_\mu\phi_a) = \partial_\mu(\delta\phi_a) \]
The variation \( \delta \) (comparison of two field configurations at the same point) commutes with the coordinate derivative \( \partial_\mu \), since both are ordinary partial derivatives acting on smooth functions. A
3
\[ \delta\mathcal{L} = \frac{\partial\mathcal{L}}{\partial\phi_a}\,\delta\phi_a + \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial_\mu(\delta\phi_a) \]
Substitute step 2 into step 1. No dynamics used yet; this is an identity for any smooth variation. A
4
\[ \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial_\mu(\delta\phi_a) = \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a \right) - \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)} \right)\delta\phi_a \]
Reverse product rule (integration by parts at the level of the integrand) to move the derivative off \( \delta\phi_a \). This is an algebraic identity, valid off-shell. A
5
\[ \delta\mathcal{L} = \left[ \frac{\partial\mathcal{L}}{\partial\phi_a} - \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)} \right) \right]\delta\phi_a + \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a \right) \]
Insert step 4 into step 3 and group. The bracket is exactly the Euler–Lagrange operator \( E_a[\mathcal{L}] \). B
6
\[ \frac{\partial\mathcal{L}}{\partial\phi_a} - \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)} \right) = 0 \]
Impose the Euler–Lagrange field equations (prior result: euler-lagrange-field-equations). This is the single dynamical input; the bracket in step 5 vanishes on-shell. B
7
\[ \delta\mathcal{L} \;\overset{\text{on-shell}}{=}\; \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a \right) \]
On-shell, the variation of the Lagrangian is a pure divergence. This holds for any variation on solutions; symmetry has not yet been used. B
8
\[ \text{Symmetry:}\quad \delta\mathcal{L} = \partial_\mu K^\mu \]
Now invoke the assumed continuous symmetry. Invariance of the action up to a boundary term means the variation of \( \mathcal{L} \) under the symmetry equals a total divergence \( \partial_\mu K^\mu \) (with \( K^\mu = 0 \) for a strict symmetry). B
9
\[ \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a \right) = \partial_\mu K^\mu \quad\Longrightarrow\quad \partial_\mu\!\left( \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a - K^\mu \right) = 0 \]
Equate step 7 (dynamics) with step 8 (symmetry) and bring both divergences together. Legal because \( \partial_\mu \) is linear. A
10
\[ \boxed{\,j^\mu = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a - K^\mu\,}, \qquad \partial_\mu j^\mu = 0 \]
Identify the conserved Noether current. For a symmetry with independent parameters \( \epsilon^\alpha \), write \( \delta\phi_a = \epsilon^\alpha (\delta_\alpha\phi_a) \) and \( K^\mu = \epsilon^\alpha K_\alpha^\mu \); stripping the arbitrary constant \( \epsilon^\alpha \) gives one conserved current \( j_\alpha^\mu \) per parameter. C
11
\[ Q = \int_{\mathbb{R}^3} d^3x\, j^0, \qquad \frac{dQ}{dt} = \int d^3x\, \partial_0 j^0 = -\int d^3x\, \partial_i j^i = -\oint_{\partial V} j^i\, dS_i = 0 \]
Integrate \( \partial_\mu j^\mu = \partial_0 j^0 + \partial_i j^i = 0 \) over all space and apply the divergence theorem; the surface term vanishes for fields (and currents) decaying at spatial infinity. C
Result
\[ j^\mu = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a - K^\mu, \qquad \partial_\mu j^\mu \overset{\text{on-shell}}{=} 0, \qquad \frac{dQ}{dt}=0 \]

Reading. Each independent continuous symmetry of the action supplies one four-current whose four-divergence vanishes on physical field configurations. The current is built from the canonical momentum density \( \pi^\mu_a \equiv \partial\mathcal{L}/\partial(\partial_\mu\phi_a) \) contracted with the shape of the field variation \( \delta\phi_a \), minus the boundary term \( K^\mu \) that quantifies how much the Lagrangian fails to be strictly invariant. The vanishing divergence is a genuine continuity equation, so the spatial integral of \( j^0 \) is a time-independent charge. The theorem is constructive: give it the symmetry, it hands you the conserved quantity.

Units check. In natural units \( [\mathcal{L}] = E^4 \) (energy density in 4D, since \( \int d^4x\,\mathcal{L} \) is dimensionless with \( [d^4x]=E^{-4} \)). A canonical scalar has \( [\phi]=E \), so \( [\partial_\mu\phi]=E^2 \) and \( [\partial\mathcal{L}/\partial(\partial_\mu\phi)] = E^4/E^2 = E^2 \). With a dimensionless internal variation \( [\delta\phi/\phi]\sim 1 \) giving \( [\delta\phi]=E \), we get \( [j^\mu] = E^2\cdot E = E^3 \), the correct dimension of a current density in four dimensions. Then \( [Q] = [d^3x][j^0] = E^{-3}\cdot E^3 = E^0 \), a dimensionless (in the case of \( \hbar=1 \) charge) conserved number, as required.

Limiting cases
  • Strict symmetry (\( K^\mu = 0 \)). The current reduces to \( j^\mu = \pi^\mu_a\,\delta\phi_a \); the boundary term is only needed for quasi-symmetries.
  • Point mechanics (0+1 dimensions). Fields become coordinates \( q(t) \), \( \partial_\mu\to d/dt \); the current collapses to a single conserved quantity \( j^0 = (\partial L/\partial\dot q)\,\delta q - K \), recovering the familiar Noether constant of motion in classical mechanics.
  • Internal symmetry (coordinates unchanged, \( \delta x^\mu=0 \)). Only \( \delta\phi_a \) contributes; for a global phase \( \phi\to e^{i\epsilon}\phi \) this gives the electric-charge current \( j^\mu = i(\phi^*\partial^\mu\phi - \phi\,\partial^\mu\phi^*) \).
  • Spacetime translation (\( \delta x^\nu=\epsilon^\nu \)). The four currents assemble into the canonical stress–energy tensor \( T^\mu{}_\nu \) with \( \partial_\mu T^\mu{}_\nu=0 \); \( \nu=0 \) gives energy, \( \nu=i \) gives momentum.
Breaks when
  • The symmetry is anomalous. A symmetry of the classical action can be broken by quantum effects (the path-integral measure fails to be invariant). Then \( \partial_\mu j^\mu \neq 0 \) even on-shell — the axial current \( \partial_\mu j^\mu_5 \propto F\tilde F \) is the canonical example. The classical theorem gives no hint of this.
  • The symmetry is local (gauge) rather than global. When the parameter \( \epsilon^\alpha(x) \) is spacetime-dependent, the first theorem degenerates: the "current" becomes identically conserved off-shell and \( Q \) reduces to a surface integral of a field strength. The correct statement is Noether's second theorem, which yields off-shell identities among the field equations, not new conserved charges.
  • The action is not stationary / no variational principle exists. Dissipative or explicitly time-dependent non-Lagrangian systems have no action to be invariant, so the construction never starts.
  • Boundary terms do not vanish. If fields do not decay at spatial infinity (e.g. solitons, non-trivial vacua, or a finite box with the wrong boundary conditions) the surface term in step 11 survives and \( Q \) is no longer conserved despite \( \partial_\mu j^\mu=0 \).
Failure modes
  • Forgetting on-shell. Students write \( \partial_\mu j^\mu=0 \) as an identity; it holds only when the Euler–Lagrange equations are imposed (step 6). Off-shell the divergence is nonzero.
  • Dropping \( K^\mu \). Using \( j^\mu = \pi^\mu_a\,\delta\phi_a \) for a symmetry that is only invariant up to a boundary term (boosts, some internal symmetries with a shift) gives a current that is not conserved.
  • Confusing \( \delta\phi \) with the full transformation for spacetime symmetries. For translations one must include the transport term \( -\epsilon^\nu\partial_\nu\phi \) in \( \delta\phi \); using only the coordinate shift or only the field shift produces a wrong stress tensor.
  • Using the first theorem for a gauge symmetry. Applying the global construction to a local symmetry yields a trivial, identically conserved current and the false conclusion that "charge is conserved because of gauge invariance"; the physical charge comes from the global subgroup.
  • Index errors in the canonical momentum. Writing \( \partial\mathcal{L}/\partial(\partial_\mu\phi) \) with the wrong contraction (summing \( a \) but not raising \( \mu \)) breaks Lorentz covariance of \( j^\mu \).
  • Assuming the Noether charge is unique. \( j^\mu \) is defined only up to an identically conserved "improvement" term \( \partial_\nu B^{\nu\mu} \) with \( B^{\nu\mu}=-B^{\mu\nu} \); students treat the canonical form as the only answer and are surprised the stress tensor can be symmetrised.
Discussion

The structural heart of the derivation is the split in step 5: the variation of the Lagrangian always separates into an Euler–Lagrange bulk term plus a total divergence. This is true off-shell and for any variation. Two independent facts then collapse the divergence into a conservation law. First, the equations of motion kill the bulk term (dynamics). Second, the symmetry makes \( \delta\mathcal{L} \) itself a divergence (kinematics of the action). Noether's insight was that these two divergences, being equal, define a current whose divergence vanishes. Nothing about the specific interactions matters — only that the action has the symmetry.

The conserved charge \( Q \) generates the symmetry it comes from. In the Hamiltonian/quantum picture \( Q \) acts on fields through Poisson brackets (or commutators) as \( \delta\phi = \{\phi, Q\} \) (or \( i[Q,\phi] \)), closing the circle: the charge that is conserved because of the symmetry is also the object that implements the symmetry. This is why momentum generates translations and angular momentum generates rotations — the same operator appears on both sides of the story.

The theorem also clarifies what "energy" means. Energy is not a substance but the Noether charge of time-translation invariance. In general relativity, where time-translation invariance is not a global symmetry of a generic spacetime, global energy conservation becomes subtle or ill-defined precisely because the required symmetry is absent — a direct corollary of Noether's framework rather than a paradox.

At the deepest level the first theorem is the statement that for a rigid symmetry the Noether current is conserved on-shell but not identically, whereas for a local (gauge) symmetry the current is conserved identically (off-shell) and the charge is a boundary integral. This dichotomy — first theorem for global, second theorem for local — is the algebraic origin of the difference between physical conserved charges and gauge constraints, and it underlies why the Standard Model's global \( U(1) \) gives baryon/lepton numbers while its local \( SU(3)\times SU(2)\times U(1) \) gives interactions rather than charges. The improvement ambiguity \( j^\mu \to j^\mu + \partial_\nu B^{\nu\mu} \) with antisymmetric \( B \) reflects that only the charge, not the current density, is physically unambiguous, and is exploited (via Belinfante–Rosenfeld) to build the symmetric stress tensor that couples to gravity.

Common misconceptions. "Noether's theorem says every conserved quantity comes from a symmetry" — false in this direction; the theorem runs symmetry \( \Rightarrow \) conservation, and there exist conserved quantities (e.g. in integrable systems) with no manifest point symmetry. "A symmetry of the equations of motion is enough" — not quite; one needs a symmetry of the action (up to a boundary term). Some symmetries of the equations of motion scale the action and yield no conserved current.

Worked examples
1
\[ \mathcal{L} = \partial_\mu\phi^*\,\partial^\mu\phi - m^2\phi^*\phi, \qquad \phi \to e^{i\epsilon}\phi,\quad \delta\phi = i\epsilon\,\phi,\ \delta\phi^* = -i\epsilon\,\phi^* \]
Complex Klein–Gordon field with a global \( U(1) \) phase symmetry. \( \mathcal{L} \) is strictly invariant, so \( K^\mu=0 \). A
2
\[ \pi^\mu \equiv \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi)} = \partial^\mu\phi^*, \qquad \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi^*)} = \partial^\mu\phi \]
Compute the canonical momenta for both \( \phi \) and \( \phi^* \) (treated as independent fields). A
3
\[ j^\mu = \partial^\mu\phi^*\,(i\epsilon\phi) + \partial^\mu\phi\,(-i\epsilon\phi^*) = i\epsilon\left(\phi\,\partial^\mu\phi^* - \phi^*\partial^\mu\phi\right) \]
Assemble \( j^\mu = \sum_a \pi^\mu_a\,\delta\phi_a \). Strip the constant \( \epsilon \) (and conventionally flip sign) to define the physical current. B
4
\[ j^\mu = i\left(\phi^*\partial^\mu\phi - \phi\,\partial^\mu\phi^*\right), \qquad \partial_\mu j^\mu = i\left(\phi^*\Box\phi - \phi\,\Box\phi^*\right) = i\left(\phi^*(-m^2\phi) - \phi(-m^2\phi^*)\right)=0 \]
Verify conservation using the equations of motion \( \Box\phi = -m^2\phi \); the mass terms cancel, confirming \( \partial_\mu j^\mu=0 \) on-shell. B
\[ j^\mu = i\left(\phi^*\partial^\mu\phi - \phi\,\partial^\mu\phi^*\right),\qquad Q = i\int d^3x\left(\phi^*\dot\phi - \phi\,\dot\phi^*\right) \]

Reading. The conserved charge of the complex scalar is its total particle-minus-antiparticle number; after coupling to electromagnetism this is exactly the electric charge, up to the coupling constant \( e \).

1
\[ \mathcal{L} = \tfrac12\,\partial_\mu\phi\,\partial^\mu\phi - V(\phi), \qquad x^\nu \to x^\nu + \epsilon^\nu,\quad \delta\phi = -\epsilon^\nu\partial_\nu\phi \]
Real scalar with an arbitrary potential; consider spacetime translation by constant \( \epsilon^\nu \). The field is dragged along, giving \( \delta\phi = -\epsilon^\nu\partial_\nu\phi \). A
2
\[ \delta\mathcal{L} = -\epsilon^\nu\partial_\nu\mathcal{L} = \partial_\mu\!\left(-\epsilon^\nu\,\delta^\mu_\nu\,\mathcal{L}\right) \;\Rightarrow\; K^\mu = -\epsilon^\nu\,\delta^\mu_\nu\,\mathcal{L} \]
Under a rigid translation \( \mathcal{L} \) shifts as a scalar, so \( \delta\mathcal{L} \) is a total divergence — a quasi-symmetry with nonzero \( K^\mu \). This is the term students forget. B
3
\[ j^\mu{}_{(\nu)} = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi)}\,(-\partial_\nu\phi) - (-\delta^\mu_\nu\mathcal{L}) = -\partial^\mu\phi\,\partial_\nu\phi + \delta^\mu_\nu\mathcal{L} \]
Insert \( \pi^\mu = \partial^\mu\phi \), \( \delta\phi = -\epsilon^\nu\partial_\nu\phi \), and subtract \( K^\mu \); strip \( \epsilon^\nu \) to obtain one current per translation direction \( \nu \). B
4
\[ T^\mu{}_\nu \equiv -j^\mu{}_{(\nu)} = \partial^\mu\phi\,\partial_\nu\phi - \delta^\mu_\nu\,\mathcal{L}, \qquad T^{00} = \tfrac12\dot\phi^2 + \tfrac12(\nabla\phi)^2 + V(\phi) \]
Define the canonical stress–energy tensor (sign chosen so \( T^{00} \) is the positive energy density). The \( \nu=0 \) component is the Hamiltonian density. C
\[ T^\mu{}_\nu = \partial^\mu\phi\,\partial_\nu\phi - \delta^\mu_\nu\mathcal{L}, \qquad \partial_\mu T^\mu{}_\nu = 0, \qquad P_\nu = \int d^3x\, T^0{}_\nu \]

Reading. Time-translation invariance gives conserved energy \( E=\int d^3x\,T^{00} \); space-translation invariance gives conserved momentum \( P^i \). A single symmetry principle produces the entire four-momentum, with units \( [T^{\mu}{}_\nu]=E^4 \) (energy density) and \( [P_\nu]=E \).

Problems
  1. Global \( U(1) \) charge normalisation. For \( \mathcal{L}=\partial_\mu\phi^*\partial^\mu\phi - m^2\phi^*\phi \) with a plane-wave solution \( \phi = A\,e^{-ik\cdot x} \) (\( k^0=\omega>0 \)) in a box of volume \( V \), compute the Noether charge \( Q \).
    Solution With \( j^\mu = i(\phi^*\partial^\mu\phi - \phi\,\partial^\mu\phi^*) \) and \( \partial^\mu\phi = -ik^\mu\phi \): \( j^\mu = i(\phi^*(-ik^\mu)\phi - \phi(+ik^\mu)\phi^*) = 2k^\mu|A|^2 \). Then \( j^0 = 2\omega|A|^2 \) and \( Q = \int_V d^3x\, j^0 = 2\omega|A|^2 V \). Positive, as expected for a particle mode; the antiparticle mode \( e^{+ik\cdot x} \) gives \( Q=-2\omega|A|^2V \).
  2. Energy of a static field. A real scalar sits in a static kink profile \( \phi(x) \) (time-independent) in 1+1 dimensions with \( \mathcal{L}=\tfrac12(\partial_t\phi)^2-\tfrac12(\partial_x\phi)^2-V(\phi) \). Write \( T^{00} \) and argue the total energy is finite for \( V\ge0 \) with \( V\to0 \) at the vacua.
    Solution \( T^{00}=\tfrac12\dot\phi^2+\tfrac12(\partial_x\phi)^2+V(\phi) \). Static \( \Rightarrow \dot\phi=0 \), so \( T^{00}=\tfrac12(\partial_x\phi)^2+V(\phi) \). The energy \( E=\int_{-\infty}^{\infty}dx\,[\tfrac12(\partial_x\phi)^2+V(\phi)] \). As \( x\to\pm\infty \), \( \phi\to \) vacua where \( V=0 \) and \( \partial_x\phi\to0 \); if the approach is exponential (typical for massive vacua) both terms are integrable and \( E \) is finite. Each term is non-negative, so \( E\ge0 \), vanishing only for the constant vacuum.
  3. Non-conservation without EOM. Show explicitly that for the complex scalar the current \( j^\mu=i(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*) \) has \( \partial_\mu j^\mu\neq0 \) off-shell, and identify what must hold to restore conservation.
    Solution \( \partial_\mu j^\mu = i(\partial_\mu\phi^*\partial^\mu\phi + \phi^*\Box\phi - \partial_\mu\phi\,\partial^\mu\phi^* - \phi\,\Box\phi^*) = i(\phi^*\Box\phi - \phi\,\Box\phi^*) \) (the first-derivative cross terms cancel). This is generally nonzero. It vanishes only when \( \Box\phi=-m^2\phi \) and \( \Box\phi^*=-m^2\phi^* \): substituting gives \( i(\phi^*(-m^2\phi)-\phi(-m^2\phi^*))=0 \). Hence conservation is equivalent to the equations of motion — the theorem is on-shell.
  4. Boost is a quasi-symmetry. For a non-relativistic Lagrangian \( L=\tfrac12 m\dot{\mathbf r}^2 \), the Galilean boost \( \mathbf r\to\mathbf r+\mathbf v t \), \( \delta\mathbf r = \mathbf v\,t \) changes \( L \). Find \( K \) and the conserved Noether quantity.
    Solution \( \delta L = m\dot{\mathbf r}\cdot\delta\dot{\mathbf r} = m\dot{\mathbf r}\cdot\mathbf v = \frac{d}{dt}(m\,\mathbf r\cdot\mathbf v) \), so \( K=m\,\mathbf r\cdot\mathbf v \) — a total time derivative, confirming a quasi-symmetry. The Noether quantity is \( j^0=\frac{\partial L}{\partial\dot{\mathbf r}}\cdot\delta\mathbf r - K = m\dot{\mathbf r}\cdot(\mathbf v t) - m\mathbf r\cdot\mathbf v = \mathbf v\cdot(m\dot{\mathbf r}\,t - m\mathbf r) \). Since \( \mathbf v \) is arbitrary, \( \mathbf G = m\dot{\mathbf r}\,t - m\mathbf r = \mathbf p\,t - m\mathbf r \) is conserved. Indeed \( \dot{\mathbf G} = \mathbf p + \dot{\mathbf p}t - m\dot{\mathbf r} = \dot{\mathbf p}\,t = 0 \) for a free particle. This is the center-of-mass theorem; dropping \( K \) would have given the wrong, non-conserved result.
  5. Improvement ambiguity. Given a conserved current \( j^\mu \), show that \( \tilde j^\mu = j^\mu + \partial_\nu B^{\nu\mu} \) with \( B^{\nu\mu}=-B^{\mu\nu} \) is also conserved and carries the same charge (assuming fields decay at infinity).
    Solution Conservation: \( \partial_\mu\tilde j^\mu = \partial_\mu j^\mu + \partial_\mu\partial_\nu B^{\nu\mu} \). The second term contracts the symmetric \( \partial_\mu\partial_\nu \) with the antisymmetric \( B^{\nu\mu} \), giving zero; with \( \partial_\mu j^\mu=0 \), \( \partial_\mu\tilde j^\mu=0 \). Charge: \( \tilde Q - Q = \int d^3x\,\partial_\nu B^{\nu 0} = \int d^3x\,\partial_i B^{i0} \) (the \( \nu=0 \) term is \( \partial_0 B^{00}=0 \) by antisymmetry). By the divergence theorem this is a surface integral \( \oint B^{i0}dS_i \), which vanishes for \( B\to0 \) at spatial infinity. Hence \( \tilde Q = Q \): the improvement term changes the current density but not the physical charge, and can be used to symmetrise \( T^{\mu\nu} \).