-2

A four-divergence term $\partial_\mu K^\mu$ when added to a Lagrangian, the action changes as $$S\to S^\prime=S+\int_R d^4x \partial_\mu K^\mu\tag{1}$$ where $R$ is a region of spacetime. Using Gauss' theorem, the term $\int_R d^4x \partial_\mu K^\mu$ can be converted to a surface integral $$\int_{\partial R} d\sigma_\mu K^\mu$$ where $\partial R$ represents the boundary of $R$. Now consider a pure Yang-Mills action $$S=-\int d^4x~ \frac{1}{4}G_{\mu\nu}^a G^{\mu\nu a}.$$ We add to it a term $\int_R d^4x \partial_\mu K^\mu$ where $K^\mu$ has the form $$K^\mu=\frac{1}{16\pi^2}\epsilon^{\mu\nu\lambda\rho}A_\nu^a\Big(G^a_{\lambda\rho}+\frac{g}{3}f^{bca}A_\lambda^b A_\rho^c\Big).$$ Here, $G_{\mu\nu}^a=\partial_\mu A_\nu^a-\partial_\nu A_\mu^a+gf^{abc}A_\mu^b A_\nu^c$ denotes the gluon field strength tensor, $A_\mu^a$ are the gluon gauge fields and $a,b,c$ denote the color index. Therefore, the action changes to \begin{eqnarray}S\to S^\prime &=& S+\int_R d^4x \partial_\mu K^\mu\\ &=& S+\int_{\partial R} d\sigma_\mu K^\mu\\ &=& S+\frac{1}{16\pi^2}\epsilon^{\mu\nu\lambda\rho}\int_{\partial T}d\sigma_\mu A_\nu^a\Big(G^a_{\lambda\rho}+\frac{g}{3}f^{bca}A_\lambda^b A_\rho^c\Big).\end{eqnarray} Now, if we consider pure gauge i.e., a boundary condition of the form $A_\mu^a\neq 0$ but $G_{\mu\nu}^a=0$ at $\partial R$, we see that the action changes by a nonzero amount $$S^\prime-S=\frac{g}{48\pi^2}\epsilon^{\mu\nu\lambda\rho}f^{bca}\int_{\partial R}d\sigma_\mu A_\nu^a A_\lambda^b A_\rho^c\neq 0.$$

Question Does this mean that the action can change even if a four-divergence is added to the Lagrangian?

SRS
  • 27,790
  • 13
  • 115
  • 365

1 Answers1

1
  1. For an action principle to be mathematically well-posed, the functional/variational derivatives $\delta S/\delta A^a_{\mu}$ should exist. Therefore it is necessary to impose appropriate boundary conditions (BCs).

  2. In OP's specific case of YM theory with boundary terms (BTs), we leave it as an exercise to work out all possible consistent BCs. Note in particular that BTs in the action may alter the set of consistent BCs. (This last fact is presumably the answer to OP's real question.) One BC that always works mathematically is the Dirichlet BC.

  3. If 2 action principles [both with (not necessarily the same) consistent BCs] differ by a BT, then the field equations of motion are the same, cf. e.g. this Phys.SE post.

  4. Finally, we should stress that a BC (besides being mathematically consistent) is often also physically motivated.

Qmechanic
  • 220,844