Internal symmetry of a classical field theory

An internal symmetry changes field values while leaving spacetime coordinates fixed, preserving the action. A constant phase rotation of a complex scalar field is an example. This distinguishes internal charges from the spacetime translation charges in the canonical stress-energy tensor. Localizing the parameter generally requires gauge fields; a global internal symmetry is not automatically a gauge symmetry.

For $N$ Dirac fields with common mass, $\mathcal{L}=\bar{\Psi }(i/\partial -m{1}_N)\Psi$ is invariant under constant complex-linear rotations $\Psi \mapsto U\Psi$ with $U\in U(N)$. A Hermitian flavour generator $T^A$ has Noether current $j_A^{\mu }=\bar{\Psi }{\gamma }^{\mu }{T}^A\Psi$. A nondegenerate fermion mass matrix restricts the symmetry to unitary matrices commuting with it. If the theory is entirely free and rotations into charge-conjugate fields are allowed, its enhanced internal symmetry of free Dirac fields can be larger than this subgroup.

Without fixed gauge charges or interactions distinguishing particles from antiparticles, $N$ free Dirac fields give $2N$ left-handed Weyl fields after charge conjugation of the right-handed fields. Their massless kinetic term has unitary group $U(2N)$. Equal nonzero Dirac masses have equal positive singular values and can be written as $2N$ equal Majorana mass terms; their preserving group is the orthogonal group $O(2N)$. Explicitly, in a basis with symmetric mass matrix $m1$, invariance requires $V^{†}V=V^{T}V=1$, hence real $V$. Restricting to transformations preserving a chosen common fermion number gives the usual $U(N{)}_L\times U(N{)}_R$ at zero mass and $U(N)$ at nonzero mass.

The free kinetic terms of left- and right-handed Weyl fields are separately invariant under unitary group rotations in flavour space. For $N$ massless Dirac fields, the subgroup preserving common fermion number is $U(N{)}_L\times U(N{)}_R$. A common nonzero Dirac mass term couples the two chirality sectors and leaves its diagonal $U(N)$ subgroup. Gauge interactions can obstruct some of the classical axial currents through a chiral anomaly.

Two canonically normalized complex scalar fields with potential $V=m^2{\Phi }^{†}\Phi +\lambda ({\Phi }^{†}\Phi {)}^2$ are four real fields with a radial potential. Consequently the isolated action is invariant under the orthogonal group $O(4)$, not only under complex-linear $U(2)$ transformations. Writing ${\varphi }_1=(q_1+i{q}_2)/\sqrt{2}$ and ${\varphi }_2=(q_3+i{q}_4)/\sqrt{2}$ makes the kinetic term ${\sum }_A(\partial q_A{)}^2/2$ and the potential a function of ${\sum }_A{q}_A^2$. The six real-plane rotations give currents $j_{AB}^{\mu }=q_B{\partial }^{\mu }{q}_A-q_A{\partial }^{\mu }{q}_B$. Gauge or Yukawa couplings can break this larger symmetry, so it should not be asserted for an arbitrary interacting doublet.

For $\mathcal{L}={\varphi }_{1,t}{\varphi }_{2,t}-c^2{\varphi }_{1,x}{\varphi }_{2,x}$, the real linear internal transformations preserving the density are $\mathrm{diag}(r,r^{-1})$ and $\left(\begin{array}{cc}0& r\\ r^{-1}& 0\end{array}\right)$ with $r\ne 0$. There are four connected components; the identity component is the displayed reciprocal rescaling with $r=e^s>0$. In the sum-and-difference basis it preserves a bilinear form of signature $(1,1)$, with the identity component isomorphic to that of the One-dimensional Lorentz group. Its Noether current in coordinates $(t,x)$ is $j^0={\varphi }_1{\varphi }_{2,t}-{\varphi }_2{\varphi }_{1,t}$ and $j^x=c^2({\varphi }_2{\varphi }_{1,x}-{\varphi }_1{\varphi }_{2,x})$. The two wave equations give ${\partial }_t{j}^0+{\partial }_x{j}^x=0$. The energy quadratic form is indefinite: the fields ${\varphi }_{\pm }=({\varphi }_1\pm {\varphi }_2)/\sqrt{2}$ have equal and opposite kinetic signs.