We consider in Lorentz spacetime, $(x^0,x^1,x^2,x^3)=(t,x,y,z)$, choose the unit of time such that $c=1$.

Given a four vector $A_\mu$, and let the Lagrangian $$L(x^i,\dot x^i,t)=-m\sqrt{1-\dot x_i\dot x^i}+qA_0+qA_i\dot x^i,\tag{1}$$ where we use Einstein's convention for summation. (See video at 35:15 with electron charge $q=-e$.)

By the Euler-Lagrange equation $$\frac{d}{dt}\frac{\partial L}{\partial\dot x^i}-\frac{\partial L}{\partial x^i}=0\tag{2}$$ we can get three equations, and if we use the fact that $$\frac{d}{d\tau}=\frac{1}{\sqrt{1-\dot x_i\dot x^i}}\frac{d}{dt}\tag{3}$$ then these equations are, for $i=1,2,3$ $$m\frac{d^2x^i}{d\tau^2}=q(\frac{\partial A_\mu}{\partial x^i}-\frac{\partial A_i}{\partial x^\mu})\frac{dx^\mu}{d\tau}\tag{4}$$

My question follows, can we obtain a fourth equation which is just letting $i=0$?

The video at 59:30 says that this fourth equation is from the property that the action is invariant under Lorentz transformation. But I think this is not very convincing. I tried to view the first three equations as two four vectors having three equation space components, but I do not think this fact can lead to the equality for the last time component.


  1. L. Susskind, Special Relativity, video lecture 7, May 21, 2012.
  • $\begingroup$ Note that the Euler-Lagrange equation is nonsense for $i=0$. What would $\frac{\partial}{\partial\dot t}$ even mean? $\endgroup$
    – Chris
    Commented Dec 31, 2017 at 9:07
  • 1
    $\begingroup$ @Chris: Dot could in principle be differentiation wrt. a world-line parameter, which may be different from coordinate time $x^0=ct$. $\endgroup$
    – Qmechanic
    Commented Jan 6, 2018 at 17:07

2 Answers 2


The action you start with is $$ S = \int d\tau L $$ and $$ L = - m \sqrt{ - \eta_{\mu\nu} {\dot x}^\mu {\dot x}^\nu } +q {\dot x}^\mu A_\mu \, . $$ This action has a gauge symmetry, which is reparameterization invariance, $\tau \to \tau'(\tau)$. In order to write down your Lagrangian you choose the gauge $\tau = x^0 = t$. Once in this gauge, you can derive equations of motion for $x^i$ but not $x^0$ because you have gauge-fixed it. These are the ones you have written down.

However, since you are fixing a gauge, there's going to be a gauge constraint which is essentially the equation of motion derived w.r.t. $x^0$ in the first action. This is the equation you are missing.

  1. Susskind's statement is the following.

    If three components of a Lorentz covariant 4-vector equation $V^{\mu}=W^{\mu}$ are equal in any inertial frame, then the fourth components must be equal as well.

    Proof: By linearity it is enough to consider the case where $$ W^{\mu}~=~\begin{bmatrix} 0 \cr 0 \cr 0 \cr 0 \end{bmatrix}. $$ Now we perform an arbitrary Lorentz transformation on a 4-vector $$ V^{\mu}~=~\begin{bmatrix} V^{0} \cr 0 \cr 0 \cr 0 \end{bmatrix}. $$ Since this does not change the three 0s by assumption, it must be because $V^0=0$ is zero to begin with. $\Box$

  2. Another idea to get 4 Euler-Lagrange (EL) eqs. is to have 4 independent dynamical variables $x^{\mu}$, cf. Prahar's answer. The main point is that OP's original action $S[x^1,x^2,x^3]$ can be viewed as a gauge theory $S[x^0,x^1,x^2,x^3]$ with a certain gauge-fixing condition $$\chi~\approx~0. \tag{A}$$ Now strictly speaking, we should introduce a Lagrange multiplier $\color{red}{B}$ to implement the gauge-fixing condition (A), i.e. include a gauge-fixing term $$ L_{\rm gf}~:=~\color{red}{B} ~\chi \tag{B} $$ in the Lagrangian. But this raises a question.

    How do we know that the presence of the Lagrange multiplier $\color{red}{B}$ does not affect the EL eqs.?

    Formal Answer: Because of the (second) Noether identity, cf. e.g. Ref. 1.

    Here we want to explicitly verify this fact for OP's model (1).

  3. We prefer to work with a non-square root$^1$ formulation$^2$ $$\begin{align} S~:=~&\int\!d\tau~L , \cr L~:=~&L_0 - U +L_{\rm gf}, \end{align}\tag{C}$$ with $$\begin{align} L_0~:=~& \frac{\dot{x}^2}{2e}-\frac{e m^2}{2}, \cr \dot {x}^2~:=~& g_{\mu\nu} \dot {x}^{\mu} \dot {x}^{\nu}, \cr \dot {x}^{\mu} ~:=~&\frac{dx^{\mu}}{d\tau},\end{align}\tag{D}$$ and with an einbein $e>0$. Here $\tau$ is the world-line (WL) parameter (which is not necessarily proper time). The velocity-dependent Lorentz potential is $$ U~:=~ - q{\dot x}^{\mu} A_{\mu}, \tag{E} $$ with corresponding generalized Lorentz 4-force$^3$ $$\begin{align} F_{\mu}~:=~&\frac{d}{d\tau} \frac{\partial U}{\partial \dot{x}^{\mu}} - \frac{\partial U}{\partial x^{\mu}}~=~qF_{\mu\nu}\dot {x}^{\nu}, \cr F_{\mu\nu}~:=~&\partial_{\mu}A_{\nu}-\partial_{\nu}A_{\mu}. \end{align}\tag{F}$$

  4. The canonical/conjugate 4-momentum$^3$ reads $$ p_{\mu}~:=~\frac{\partial L}{\partial \dot{x}^{\mu}}~=~ p^{\rm kin}_{\mu}+ q A_{\mu}, \tag{G}$$ where the kinetic 4-momentum$^3$ reads $$ p^{\rm kin}_{\mu} ~:=~ \frac{g_{\mu\nu}\dot{x}^{\nu}}{e}. \tag{H}$$

  5. Now the EL eqs. for the action (B) read $$\begin{align} 0~\approx~&\frac{\delta S}{\delta x^{\mu}} ~=~ -\dot{p}^{\rm kin}_{\mu} +\frac{\partial_{\mu}g_{\nu\lambda}}{2e}\dot {x}^{\nu} \dot {x}^{\lambda} +F_{\mu} + \color{red}{B}\frac{\partial \chi}{\partial x^{\mu}} \cr ~=~& -g_{\mu\nu}\frac{d}{d\tau}\left(\frac{\dot{x}^{\nu}}{e}\right) -\frac{\Gamma_{\mu,\nu\lambda}}{e}\dot {x}^{\nu} \dot {x}^{\lambda} +F_{\mu} + \color{red}{B}\frac{\partial \chi}{\partial x^{\mu}}, \end{align} \tag{I}$$ $$ 0~\approx~\frac{\delta S}{\delta e} ~=~-\frac{1}{2}\left(\frac{\dot{x}^2}{e^2}+m^2 \right) + \color{red}{B}\frac{\partial \chi}{\partial e}, \tag{J} $$ $$ 0~\approx~\frac{\delta S}{\delta B}~=~\chi. \tag{K}$$

  6. The question we would like to address is the following.

    How do we conclude that the Lagrange multiplier $\color{red}{B}~\approx~0$ vanishes in EL eqs. (I) & (J)?

  7. In anticipation of what to come, let us first differentiate eq. (J) wrt. $\tau$: $$\begin{align} \frac{d}{d\tau}\left( \color{red}{B}\frac{\partial \chi}{\partial e}\right) ~\stackrel{(J)}{\approx}~&\frac{\dot{x}^{\mu}}{e} \left(g_{\mu\nu}\frac{d}{d\tau}\left(\frac{\dot{x}^{\nu}}{e}\right) +\frac{\Gamma_{\mu,\nu\lambda}}{e}\dot {x}^{\nu} \dot {x}^{\nu} \right)\cr ~\stackrel{(I)}{\approx}~&\frac{\dot{x}^{\mu}}{e}\color{red}{B}\frac{\partial \chi}{\partial x^{\mu}}. \end{align}\tag{M}$$

  8. Static gauge $\tau=x^0$. This is OP's gauge. In the static gauge $$ \chi~:=~x^0-\tau . \tag{N}$$ Then eq. (M) becomes $$ 0~\stackrel{(M)+(N)}{\approx}~ \frac{\color{red}{B}}{e}, \tag{O}$$ i.e. the Lagrange multiplier $\color{red}{B}$ vanishes! $\Box$

  9. Affine parametrization gauge $e=1/m$.
    $$ \chi~:=~e-1/m. \tag{P}$$ Then eq. (M) becomes $$ \dot{\color{red}{B}}~\stackrel{(M)+(P)}{\approx}~0.\tag{Q} $$ We can only conclude that the Lagrange multiplier $\color{red}{B}$ is a constant. This is related to the fact that this is only a partial gauge fixing. In this gauge the world-line (WL) parameter $\tau$ is an affine function of proper time.


  1. H. Motohashi, T. Suyama & K. Takahashi, Phys. Rev. D 94 (2016) 124021, arXiv:1608.00071.


$^1$ To achieve OP's square root formulation (1), simply integrate out the $e$ field, cf. e.g. this Phys.SE post. The gauge symmetry is e.g. discussed in this Phys.SE post and links therein.

$^2$ Notation & conventions. We use signature convention $(-,+,+,+)$ and $c=1$. The $\approx$ symbol means equality modulo eoms.

$^3$ The usual notions of 4-momentum & 4-force correspond to the gauge where the world-line (WL) parameter $\tau$ is proper time.

  • $\begingroup$ Notes for later: What happens if we have quadratic gauge-fixing instead of a delta function? $\endgroup$
    – Qmechanic
    Commented Oct 30, 2018 at 9:20
  • $\begingroup$ In the first point, you say that for $V^\mu=[V^0, 0, 0, 0]^T$, an arbitrary Lorentz transformation (which I know as any matrix $\Lambda$ satisfying $\Lambda^T\eta\Lambda=\eta$) does not change the three 0's. However I doubt that. Can you please prove? $\endgroup$
    – Atom
    Commented Jul 13, 2020 at 14:51
  • $\begingroup$ It is not true in general. That's merely Susskind's assumption. $\endgroup$
    – Qmechanic
    Commented Jul 13, 2020 at 15:04
  • $\begingroup$ So Susskind is wrong? And the only way to get to the 4th equation is to use the gauge invariance argument? $\endgroup$
    – Atom
    Commented Jul 13, 2020 at 15:19
  • $\begingroup$ Susskind's statement in the video is correct. $\endgroup$
    – Qmechanic
    Commented Jul 13, 2020 at 15:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.