整体规范变换是指不依赖时空坐标的变换。
在标量场的规范变换
ϕ ( θ ) = e − i q θ ϕ 0 , ϕ ˉ ( θ ) = e i q θ ϕ ˉ 0 \phi(\theta) = \mathrm e^{-iq\theta}\phi_0, ~ \bar\phi(\theta) = \mathrm e^{iq\theta}\bar\phi_0 ϕ(θ)=e−iqθϕ0, ϕˉ(θ)=eiqθϕˉ0
下,拉格朗日密度不变。也就是说,对于
L = − [ ( ∂ a ϕ ˉ ) ∂ a ϕ + m 2 ϕ ϕ ˉ ] \mathscr L = -[(\partial^a\bar\phi)\partial_a\phi + m^2\phi\bar\phi] L=−[(∂aϕˉ)∂aϕ+m2ϕϕˉ]
有
0 = d L d θ = ∂ L ∂ ϕ d ϕ d θ + ∂ L ∂ ( ∂ a ϕ ) d ( ∂ a ϕ ) d θ + ∂ L ∂ ϕ ˉ d ϕ ˉ d θ + ∂ L ∂ ( ∂ a ϕ ˉ ) d ( ∂ a ϕ ˉ ) d θ 0 = \frac{d\mathscr L}{d\theta} = \frac{\partial \mathscr L}{\partial \phi}\frac{d\phi}{d\theta} + \frac{\partial \mathscr L}{\partial(\partial_a\phi)}\frac{d(\partial_a\phi)}{d\theta} + \frac{\partial \mathscr L}{\partial \bar\phi}\frac{d\bar\phi}{d\theta} + \frac{\partial \mathscr L}{\partial(\partial_a\bar\phi)}\frac{d(\partial_a\bar\phi)}{d\theta} 0=dθdL=∂ϕ∂Ldθdϕ+∂(∂aϕ)∂Ldθd(∂aϕ)+∂ϕˉ∂Ldθdϕˉ+∂(∂aϕˉ)∂Ldθd(∂aϕˉ)
容易验证其中 d ϕ d θ = − i q ϕ , d ( ∂ a ϕ ) d θ = − i q ∂ a ϕ \frac{d\phi}{d\theta} = -iq\phi, ~ \frac{d(\partial_a\phi)}{d\theta} = -iq\partial_a\phi dθdϕ=−iqϕ, dθd(∂aϕ)=−iq∂aϕ,再代入拉格朗日方程就得到
d L d θ = − i q ∂ a [ ϕ ∂ L ∂ ( ∂ a ϕ ) − ϕ ˉ ∂ L ∂ ( ∂ a ϕ ˉ ) ] = i q ∂ a ( ϕ ∂ a ϕ ˉ − ϕ ˉ ∂ a ϕ ) \frac{d\mathscr L}{d\theta} = -iq\partial_a[\phi\frac{\partial \mathscr L}{\partial(\partial_a\phi)} - \bar\phi\frac{\partial \mathscr L}{\partial(\partial_a\bar\phi)}] = iq\partial_a(\phi\partial^a\bar\phi - \bar\phi\partial^a\phi) dθdL=−iq∂a[ϕ∂(∂aϕ)∂L−ϕˉ∂(∂aϕˉ)∂L]=iq∂a(ϕ∂aϕˉ−ϕˉ∂aϕ)
记
J a = i e q ( ϕ ∂ a ϕ ˉ − ϕ ˉ ∂ a ϕ ) J^a = ieq(\phi\partial^a\bar\phi - \bar\phi\partial^a\phi) Ja=ieq(ϕ∂aϕˉ−ϕˉ∂aϕ)
则有 0 = ∂ a J a 0 = \partial_aJ^a 0=∂aJa,可见 J a J^a Ja就是对应的守恒流。
所有这样的变换的集合 { e − i q θ ∣ θ ∈ R } \{\mathrm e^{-iq\theta} | \theta \in \R \} {e−iqθ∣θ∈R}是酉群 U ( 1 ) U(1) U(1),显然是阿贝尔群。
考虑同位旋变换
[ ϕ 1 ′ ϕ 2 ′ ] = U [ ϕ 1 ϕ 2 ] \begin{bmatrix} \phi'_1 \\ \phi'_2 \end{bmatrix} = U \begin{bmatrix} \phi_1 \\ \phi_2 \end{bmatrix} [ϕ1′ϕ2′]=U[ϕ1ϕ2]
其中 U U U为一复矩阵, ϕ = [ ϕ 1 ϕ 2 ] \phi = \begin{bmatrix} \phi_1 \\ \phi_2 \end{bmatrix} ϕ=[ϕ1ϕ2]为核子的同位旋态波函数。在改变换下,系统的拉格朗日量仍是不变的。由归一化条件 ϕ † ϕ = 1 = ϕ ′ † ϕ ′ = ( U ϕ ) † ( U ϕ ) = ϕ † U † U ϕ \phi^\dagger\phi = 1 = \phi'^\dagger\phi' = (U\phi)^\dagger(U\phi) = \phi^\dagger U^\dagger U\phi ϕ†ϕ=1=ϕ′†ϕ′=(Uϕ)†(Uϕ)=ϕ†U†Uϕ,这就要求 U † U = I U^\dagger U = I U†U=I,所以 U ∈ U ( 2 ) U \in U(2) U∈U(2)是酉群,但 U ( 2 ) U(2) U(2)是非阿贝尔的。又 1 = det ( I ) = det ( U † U ) = ∣ det U ∣ 2 1 = \det(I) = \det(U^\dagger U) = |\det U|^2 1=det(I)=det(U†U)=∣detU∣2,所以 det U = e i α , α ∈ R \det U = \mathrm e^{i\alpha}, ~ \alpha \in \R detU=eiα, α∈R. 以下为讨论方便起见,我们假定 det U = 1 \det U = 1 detU=1,于是 U ∈ S U ( 2 ) U \in SU(2) U∈SU(2),这不会影响问题的实质。注意到 S U ( 2 ) SU(2) SU(2)的群元可由其李代数 s u ( 2 ) \mathfrak{su}(2) su(2)的指数映射给出,即
U ( θ ⃗ ) = e − i 2 τ ⃗ ⋅ θ ⃗ U(\vec\theta) = \mathrm e^{-\frac{i}{2}\vec\tau\cdot\vec\theta} U(θ)=e−2iτ⋅θ
其中 τ ⃗ \vec\tau τ为三个泡利矩阵, θ ⃗ ∈ R 3 \vec\theta \in \R^3 θ∈R3为三个实系数。
一般来说,规范场论总是涉及一个李群(内部变换群) G G G和一个(或多个)矩阵李群 G ^ \hat G G^,且 ρ : G → G ^ \rho \colon G \to \hat G ρ:G→G^是 G G G的表示。群元 g ∈ G g \in G g∈G总可以借指数映射表示为
g = exp ( θ μ e μ ) g = \exp(\theta^\mu e_\mu) g=exp(θμeμ)
其中 e μ e_\mu eμ是李代数 g \mathfrak g g的一组基矢, θ μ \theta^\mu θμ为对应的实系数。以 V V V代表 ρ \rho ρ的表示空间, dim V = N \dim V = N dimV=N,则 G ^ \hat G G^中的元素就可以表示为 N × N N \times N N×N的矩阵。以 g ^ \mathfrak{\hat g} g^代表 G ^ \hat G G^的李代数,则同态 ρ \rho ρ在恒等元 e ∈ G e \in G e∈G处的推前映射 ρ ∗ : g → g ^ \rho_* \colon \mathfrak g \to \mathfrak{\hat g} ρ∗:g→g^是李代数同态。记 U ( θ ⃗ ) = ρ ( g ) ∈ G ^ U(\vec\theta) = \rho(g) \in \hat G U(θ)=ρ(g)∈G^,则有
U ( θ ⃗ ) = exp [ ρ ∗ ( θ μ e μ ) ] = exp ( θ μ ρ ∗ e μ ) U(\vec\theta) = \exp[\rho_*(\theta^\mu e_\mu)] = \exp(\theta^\mu\rho_*e_\mu) U(θ)=exp[ρ∗(θμeμ)]=exp(θμρ∗eμ)
我们引入记号 − i L μ ≡ ρ ∗ e μ ∈ g ^ -iL_\mu \equiv \rho_*e_\mu \in \mathfrak{\hat g} −iLμ≡ρ∗eμ∈g^,就有
U ( θ ⃗ ) = e − i L μ θ μ U(\vec\theta) = \mathrm e^{-iL_\mu\theta^\mu} U(θ)=e−iLμθμ
相应的规范变换就推广为
ϕ = U ( θ ⃗ ) ϕ 0 , ϕ ˉ = ϕ ˉ 0 U ( θ ⃗ ) − 1 \phi = U(\vec\theta)\phi_0, ~ \bar\phi = \bar\phi_0U(\vec\theta)^{-1} ϕ=U(θ)ϕ0, ϕˉ=ϕˉ0U(θ)−1
注意阿贝尔群的情况是现在情况的特例,只要令 L ⃗ = q \vec L = q L=q就回到阿贝尔群的情况。
规范变换的例子有
李群 | 表示空间维度 | 物理含义 |
---|---|---|
U ( 1 ) U(1) U(1) | 1 | 电荷守恒 |
S U ( 2 ) SU(2) SU(2) | 2 | 核子同位旋守恒 |
S U ( 2 ) SU(2) SU(2) | 3 | π \pi π介子( π + , π − , π 0 \pi^+, \pi^-, \pi^0 π+,π−,π0) |
S U ( 3 ) SU(3) SU(3) | 3 | 夸克的味(u, d, s) |
局域性原理认为一个特定物体只能与其周围的物质发生相互作用,这导致我们考虑变换的参数 θ \theta θ依赖时空坐标的情况。
如果变换参数 θ \theta θ依赖时空坐标 x x x,那么容易看出此时的拉格朗日密度 L 0 = − [ ( ∂ μ ϕ ˉ ) ∂ μ ϕ + m 2 ϕ ϕ ˉ ] \mathscr L_0 = -[(\partial^\mu\bar\phi)\partial_\mu\phi + m^2\phi\bar\phi] L0=−[(∂μϕˉ)∂μϕ+m2ϕϕˉ]在如下的局域规范变换下
ϕ ( x ) = e − i q θ ( x ) ϕ 0 ( x ) , ϕ ˉ ( x ) = e i q θ ( x ) ϕ ˉ 0 ( x ) \phi(x) = \mathrm e^{-iq\theta(x)}\phi_0(x), ~ \bar\phi(x) = \mathrm e^{iq\theta(x)}\bar\phi_0(x) ϕ(x)=e−iqθ(x)ϕ0(x), ϕˉ(x)=eiqθ(x)ϕˉ0(x)
无法再保持不变。实际上
∂ μ ϕ ( x ) = e − i q θ ( x ) ∂ μ ϕ 0 ( x ) − i q ϕ ( x ) ∂ μ θ ( x ) ∂ μ ϕ ˉ ( x ) = e i q θ ( x ) ∂ μ ϕ ˉ 0 ( x ) + i q ϕ ˉ ( x ) ∂ μ θ ( x ) \partial_\mu\phi(x) = \mathrm e^{-iq\theta(x)}\partial_\mu\phi_0(x) - iq\phi(x)\partial_\mu\theta(x) \\ \partial^\mu\bar\phi(x) = \mathrm e^{iq\theta(x)}\partial^\mu\bar\phi_0(x) + iq\bar\phi(x)\partial^\mu\theta(x) ∂μϕ(x)=e−iqθ(x)∂μϕ0(x)−iqϕ(x)∂μθ(x)∂μϕˉ(x)=eiqθ(x)∂μϕˉ0(x)+iqϕˉ(x)∂μθ(x)
解决此问题的办法是修改拉格朗日密度为
L 1 = − [ ( D μ ϕ ‾ ) D μ ϕ + m 2 ϕ ϕ ˉ ] \mathscr L_1 = -[(\overline{D^\mu\phi})D_\mu\phi + m^2\phi\bar\phi] L1=−[(Dμϕ)Dμϕ+m2ϕϕˉ]
其中
D a = ∂ a − i e q A a D_a = \partial_a - ieqA_a Da=∂a−ieqAa
称为协变导数算符,基本电荷 e e e在这里作为耦合常数。在容易看出,为了保证拉格朗日密度在局域规范变换下的不变性,我们需要有
D μ ′ ϕ ( x ) = ( ∂ a − i e q A μ ′ ) ϕ ( x ) = e − i q θ ( x ) D μ ϕ 0 ( x ) D_\mu'\phi(x) = (\partial_a - ieqA'_\mu)\phi(x) = \mathrm e^{-iq\theta(x)}D_\mu\phi_0(x) Dμ′ϕ(x)=(∂a−ieqAμ′)ϕ(x)=e−iqθ(x)Dμϕ0(x)
这就要求
A μ ′ = A μ − 1 e ∂ μ θ ( x ) A'_\mu = A_\mu - \frac{1}{e}\partial_\mu\theta(x) Aμ′=Aμ−e1∂μθ(x)
而这不过就是电磁4势的规范变换式。从物理角度考虑,既然复标量粒子带电,我们就需要考虑其与电磁场耦合后的总拉格朗日密度
L = L 0 + L i n t + L E M = L 1 + L E M \mathscr L = \mathscr L_0 + \mathscr L_{int} + \mathscr L_{EM} = \mathscr L_1 + \mathscr L_{EM} L=L0+Lint+LEM=L1+LEM
其中 L E M = − 1 16 π F μ ν F μ ν \mathscr L_{EM} = -\frac{1}{16\pi}F^{\mu\nu}F_{\mu\nu} LEM=−16π1FμνFμν是电磁场, L i n t = J μ A μ − e 2 q 2 ϕ ( x ) ϕ ˉ ( x ) A μ A μ \mathscr L_{int} = J^\mu A_\mu - e^2q^2\phi(x)\bar\phi(x)A^\mu A_\mu Lint=JμAμ−e2q2ϕ(x)ϕˉ(x)AμAμ表示 ϕ \phi ϕ与电磁场的相互作用对整体拉格朗日密度的贡献。因为 ∂ L ∂ ϕ ˉ = − ( i e q A μ ∂ μ ϕ + e 2 q 2 A μ A μ ϕ + m 2 ϕ ) , ∂ L ∂ ( ∂ μ ϕ ˉ ) = i e q A μ ϕ − ∂ μ ϕ \frac{\partial\mathscr L}{\partial \bar\phi} = -(ieqA_\mu\partial^\mu\phi + e^2q^2A_\mu A^\mu\phi + m^2\phi), ~ \frac{\partial\mathscr L}{\partial(\partial_\mu\bar\phi)} = ieqA^\mu\phi - \partial^\mu\phi ∂ϕˉ∂L=−(ieqAμ∂μϕ+e2q2AμAμϕ+m2ϕ), ∂(∂μϕˉ)∂L=ieqAμϕ−∂μϕ, L \mathscr L L对 ϕ ˉ \bar\phi ϕˉ的拉格朗日方程就给出
[ ( D μ ϕ ) D μ ϕ − m 2 ] ϕ = 0 [(D^\mu\phi)D_\mu\phi - m^2]\phi = 0 [(Dμϕ)Dμϕ−m2]ϕ=0
当 A μ = 0 A^\mu = 0 Aμ=0时这就回到Klein-Gordon方程。又 ∂ L ∂ A μ = i e q ϕ ∂ μ ϕ ˉ − i e q ϕ ˉ ∂ μ ϕ − 2 e 2 q 2 A μ ϕ ϕ ˉ , ∂ L ∂ ( ∂ ν A μ ) = − 1 4 π ( ∂ ν A μ − ∂ μ A ν ) \frac{\partial\mathscr L}{\partial A_\mu} = ieq\phi\partial^\mu\bar\phi - ieq\bar\phi\partial^\mu\phi - 2e^2q^2A^\mu\phi\bar\phi, ~ \frac{\partial\mathscr L}{\partial(\partial_\nu A_\mu)} = -\frac{1}{4\pi}(\partial^\nu A^\mu - \partial^\mu A^\nu) ∂Aμ∂L=ieqϕ∂μϕˉ−ieqϕˉ∂μϕ−2e2q2Aμϕϕˉ, ∂(∂νAμ)∂L=−4π1(∂νAμ−∂μAν),并注意到 ∂ ν A ν = 0 \partial_\nu A^\nu = 0 ∂νAν=0. [如果 ∂ ν A ν ≠ 0 \partial_\nu A^\nu \ne 0 ∂νAν=0,我们可以选择这样的参数 θ \theta θ,使得 ∂ ν ( A ν − ∂ ν θ ) = 0 \partial_\nu(A^\nu - \partial^\nu\theta) = 0 ∂ν(Aν−∂νθ)=0]。 L \mathscr L L对 A μ A_\mu Aμ的拉格朗日方程就给出
∂ ν ∂ ν A μ = 4 π [ − i e q ( ϕ ∂ μ ϕ ˉ − ϕ ˉ ∂ μ ϕ ) + 2 e 2 q 2 ϕ ϕ ˉ A μ ] = − 4 π i e q ( ϕ D μ ϕ ‾ − ϕ ˉ D μ ϕ ) \partial_\nu\partial^\nu A^\mu = 4\pi[-ieq(\phi\partial^\mu\bar\phi - \bar\phi\partial^\mu\phi) + 2e^2q^2\phi\bar\phi A^\mu] = -4\pi ieq(\phi\overline{D^\mu\phi} - \bar\phi D^\mu\phi) ∂ν∂νAμ=4π[−ieq(ϕ∂μϕˉ−ϕˉ∂μϕ)+2e2q2ϕϕˉAμ]=−4πieq(ϕDμϕ−ϕˉDμϕ)
记
J A μ = i e q ( ϕ D μ ϕ ‾ − ϕ ˉ D μ ϕ ) J_A^\mu = ieq(\phi\overline{D^\mu\phi} - \bar\phi D^\mu\phi) JAμ=ieq(ϕDμϕ−ϕˉDμϕ)
就容易看出 ∂ μ J A μ = − 1 4 π ∂ ν ∂ ν ∂ μ A μ = 0 \partial_\mu J_A^\mu = -\frac{1}{4\pi}\partial_\nu\partial^\nu\partial_\mu A^\mu = 0 ∂μJAμ=−4π1∂ν∂ν∂μAμ=0,因此 J A μ J_A^\mu JAμ是现在的新的守恒流。
以下将把上述讨论推广到非阿贝尔群的情况。此时局域规范变换是
ϕ ( x ) = U ( θ ⃗ ( x ) ) ϕ 0 ( x ) , ϕ ˉ ( x ) = ϕ ˉ 0 ( x ) U ( θ ⃗ ( x ) ) − 1 \phi(x) = U(\vec\theta(x))\phi_0(x), ~ \bar\phi(x) = \bar\phi_0(x)U(\vec\theta(x))^{-1} ϕ(x)=U(θ(x))ϕ0(x), ϕˉ(x)=ϕˉ0(x)U(θ(x))−1
前述讨论的推广概述有二:
我们当然希望 D μ D_\mu Dμ仍能够按照如下方式变换,即
D μ ′ ϕ ( x ) = U ( θ ⃗ ( x ) ) D μ ϕ 0 ( x ) D_\mu'\phi(x) = U(\vec\theta(x))D_\mu\phi_0(x) Dμ′ϕ(x)=U(θ(x))Dμϕ0(x)
这将保证 L 1 \mathscr L_1 L1的规范不变性,展开上式左边得到
D μ ′ ϕ ( x ) = ∂ μ [ U ( θ ⃗ ( x ) ) ϕ 0 ( x ) ] − i k L r A μ ′ r ( x ) U ( θ ⃗ ( x ) ) ϕ 0 ( x ) = ϕ 0 ( x ) ∂ μ U ( θ ⃗ ( x ) ) + U ( θ ⃗ ( x ) ) ∂ μ ϕ 0 ( x ) − i k L r A μ ′ r ( x ) U ( θ ⃗ ( x ) ) ϕ 0 ( x ) \begin{aligned} D_\mu'\phi(x) &= \partial_\mu[U(\vec\theta(x))\phi_0(x)] - ikL_rA'^r_\mu(x)U(\vec\theta(x))\phi_0(x) \\ &= \phi_0(x)\partial_\mu U(\vec\theta(x)) + U(\vec\theta(x))\partial_\mu\phi_0(x) - ikL_rA'^r_\mu(x)U(\vec\theta(x))\phi_0(x) \end{aligned} Dμ′ϕ(x)=∂μ[U(θ(x))ϕ0(x)]−ikLrAμ′r(x)U(θ(x))ϕ0(x)=ϕ0(x)∂μU(θ(x))+U(θ(x))∂μϕ0(x)−ikLrAμ′r(x)U(θ(x))ϕ0(x)
而上式右边是
U ( θ ⃗ ( x ) ) D μ ϕ 0 ( x ) = U ( θ ⃗ ( x ) ) ∂ μ ϕ 0 ( x ) − i k U ( θ ⃗ ( x ) ) L r A μ r ( x ) ϕ 0 ( x ) U(\vec\theta(x))D_\mu\phi_0(x) = U(\vec\theta(x))\partial_\mu\phi_0(x) - ikU(\vec\theta(x))L_rA^r_\mu(x)\phi_0(x) U(θ(x))Dμϕ0(x)=U(θ(x))∂μϕ0(x)−ikU(θ(x))LrAμr(x)ϕ0(x)
于是只要下式成立
ϕ 0 ( x ) ∂ μ U ( θ ⃗ ( x ) ) − i k L r A μ ′ r ( x ) U ( θ ⃗ ( x ) ) ϕ 0 ( x ) = − i k U ( θ ⃗ ( x ) ) L r A μ r ( x ) ϕ 0 ( x ) \phi_0(x)\partial_\mu U(\vec\theta(x)) - ikL_rA'^r_\mu(x)U(\vec\theta(x))\phi_0(x) = - ikU(\vec\theta(x))L_rA^r_\mu(x)\phi_0(x) ϕ0(x)∂μU(θ(x))−ikLrAμ′r(x)U(θ(x))ϕ0(x)=−ikU(θ(x))LrAμr(x)ϕ0(x)
即
− i L r A μ ′ r ( x ) = − i U ( θ ⃗ ( x ) ) L r A μ r ( x ) U − 1 ( θ ⃗ ( x ) ) − 1 k [ ∂ μ U ( θ ⃗ ( x ) ) ] U − 1 ( θ ⃗ ( x ) ) - iL_rA'^r_\mu(x) = - iU(\vec\theta(x))L_rA^r_\mu(x)U^{-1}(\vec\theta(x)) - \frac{1}{k}[\partial_\mu U(\vec\theta(x))]U^{-1}(\vec\theta(x)) −iLrAμ′r(x)=−iU(θ(x))LrAμr(x)U−1(θ(x))−k1[∂μU(θ(x))]U−1(θ(x))
记 A ^ μ ( x ) = − i L r A μ r ( x ) \hat A_\mu(x) = - iL_rA^r_\mu(x) A^μ(x)=−iLrAμr(x),上式还可以简化为
A ^ μ ′ ( x ) = U ( θ ⃗ ( x ) ) A ^ μ ( x ) U − 1 ( θ ⃗ ( x ) ) − 1 k [ ∂ μ U ( θ ⃗ ( x ) ) ] U − 1 ( θ ⃗ ( x ) ) \hat A'_\mu(x) = U(\vec\theta(x))\hat A_\mu(x)U^{-1}(\vec\theta(x)) - \frac{1}{k}[\partial_\mu U(\vec\theta(x))]U^{-1}(\vec\theta(x)) A^μ′(x)=U(θ(x))A^μ(x)U−1(θ(x))−k1[∂μU(θ(x))]U−1(θ(x))
从而当 U ( θ ⃗ ( x ) ) = e − i q θ ( x ) U(\vec\theta(x)) = \mathrm e^{-iq\theta(x)} U(θ(x))=e−iqθ(x)时,上式就回到 A μ ′ = A μ − 1 e ∂ μ θ ( x ) A'_\mu = A_\mu - \frac{1}{e}\partial_\mu\theta(x) Aμ′=Aμ−e1∂μθ(x)
我们还需要给出 L Y M \mathscr L_{YM} LYM的定义,为此可以仿照 A ^ μ ( x ) \hat A_\mu(x) A^μ(x),记 F ^ μ ν ( x ) = − i L r F μ ν r ( x ) \hat F_{\mu\nu}(x) = -iL_rF^r_{\mu\nu}(x) F^μν(x)=−iLrFμνr(x),并定义
F ^ μ ν ( x ) = ∂ μ A ^ ν ( x ) − ∂ ν A ^ μ ( x ) + k [ A ^ μ ( x ) , A ^ ν ( x ) ] \hat F_{\mu\nu}(x) = \partial_\mu\hat A_\nu(x) - \partial_\nu\hat A_\mu(x) + k[\hat A_\mu(x), \hat A_\nu(x)] F^μν(x)=∂μA^ν(x)−∂νA^μ(x)+k[A^μ(x),A^ν(x)]
其中 [ A ^ μ ( x ) , A ^ ν ( x ) ] [\hat A_\mu(x), \hat A_\nu(x)] [A^μ(x),A^ν(x)]是李代数元 A ^ μ ( x ) , A ^ ν ( x ) \hat A_\mu(x), ~ \hat A_\nu(x) A^μ(x), A^ν(x)的李括号。这等价于是说
F μ ν r ( x ) = ∂ μ A ν r ( x ) − ∂ ν A μ r ( x ) + k C s t r A μ s ( x ) A ν t ( x ) F^r_{\mu\nu}(x) = \partial_\mu A^r_\nu(x) - \partial_\nu A^r_\mu(x) + kC^r_{st}A^s_\mu(x)A^t_\nu(x) Fμνr(x)=∂μAνr(x)−∂νAμr(x)+kCstrAμs(x)Aνt(x)
其中 C s t r C^r_{st} Cstr是 G G G的李代数 g \mathfrak g g在基底 { e r } \{e_r\} {er}下的结构常数, r , s , t = 1 , … , R r, s, t = 1, \dots, R r,s,t=1,…,R. 可以证明, F ^ μ ν ( x ) \hat F_{\mu\nu}(x) F^μν(x)的变换关系为
F ^ μ ν ′ ( x ) = U ( θ ⃗ ( x ) ) F ^ μ ( x ) U − 1 ( θ ⃗ ( x ) ) \hat F'_{\mu\nu}(x) = U(\vec\theta(x))\hat F_\mu(x)U^{-1}(\vec\theta(x)) F^μν′(x)=U(θ(x))F^μ(x)U−1(θ(x))
现在 L Y M \mathscr L_{YM} LYM就定义为
L Y M = − 1 16 π ∑ r = 1 R F μ ν r ( x ) F r μ ν ( x ) \mathscr L_{YM} = -\frac{1}{16\pi}\sum^R_{r = 1}F^r_{\mu\nu}(x)F^{r\mu\nu}(x) LYM=−16π1r=1∑RFμνr(x)Frμν(x)
可以证明该拉格朗日量是规范不变的。