Physlib

Physlib.ClassicalMechanics.Lagrangian.TotalDerivativeEquivalence

Equivalent Lagrangians under Total Derivatives

i. Overview

Two Lagrangians are physically equivalent if they differ by a total time derivative d/dt F(q, t). This is because the Euler-Lagrange equations depend only on extremizing the action integral, and total derivatives don't affect which paths are extremal.

This module defines the key concept of a function being a total time derivative, which is essential for analyzing symmetries like Galilean invariance.

Note: Some authors call this "gauge equivalence" by analogy with gauge transformations in field theory, but we avoid that terminology here since no gauge fields are involved.

ii. Key insight

A general function δL(r, v, t) is a total time derivative if there exists a function F(r, t) (independent of velocity) such that: δL(r, v, t) = d/dt F(r, t) = fderiv ℝ F (r, t) (v, 1)

By the chain rule, this expands to: δL(r, v, t) = ∂F/∂t + ⟨∇ᵣF, v⟩

For the special case where δL depends only on velocity v (not position or time), this implies a strong constraint: δL(v) = ⟨g, v⟩ for some constant vector g

This is because: 1. d/dt F(r, t) = ∂F/∂t + ⟨∇F, v⟩ 2. For δL to be r-independent, ∇F must be r-independent 3. For δL to be t-independent, the time-dependent part must vanish 4. The result is δL = ⟨g, v⟩ for constant g

iii. Key definitions

  • `IsTotalTimeDerivative`: General case for δL(r, v, t)
  • `IsTotalTimeDerivativeVelocity`: Velocity-only case, equivalent to δL(v) = ⟨g, v⟩

iv. References

  • Landau & Lifshitz, "Mechanics", §2 (The principle of least action)
  • Landau & Lifshitz, "Mechanics", §4 (The Lagrangian for a free particle)

A. General Total Time Derivative

B. Velocity-Only Total Time Derivative

When δL depends only on velocity (the free particle case), the condition simplifies.

8 declarations

definition

δL\delta L is a total time derivative ddtF(r,t)\frac{d}{dt}F(r, t)

A function δL:Rn×Rn×RR\delta L: \mathbb{R}^n \times \mathbb{R}^n \times \mathbb{R} \to \mathbb{R} (representing a change in a Lagrangian) is a total time derivative if there exists a differentiable function F:Rn×RRF: \mathbb{R}^n \times \mathbb{R} \to \mathbb{R}, depending only on position rr and time tt, such that for all r,v,tr, v, t: δL(r,v,t)=DF(r,t)(v,1)\delta L(r, v, t) = \text{D}F|_{(r, t)}(v, 1) By the chain rule, this condition is equivalent to: δL(r,v,t)=Ft(r,t)+rF(r,t)v\delta L(r, v, t) = \frac{\partial F}{\partial t}(r, t) + \nabla_r F(r, t) \cdot v where vv is the velocity vector.

theorem

Velocity-Only Total Time Derivatives are Linear: δL(v)=g,v\delta L(v) = \langle g, v \rangle

Let δL:RnR\delta L : \mathbb{R}^n \to \mathbb{R} be a function of velocity vv such that δL(0)=0\delta L(0) = 0. Suppose that δL\delta L is a total time derivative, meaning the function (r,v,t)δL(v)(r, v, t) \mapsto \delta L(v) satisfies the condition that there exists a differentiable function F(r,t)F(r, t) depending only on position and time such that: δL(v)=ddtF(r,t)=Ft(r,t)+rF(r,t)v\delta L(v) = \frac{d}{dt} F(r, t) = \frac{\partial F}{\partial t}(r, t) + \nabla_r F(r, t) \cdot v Then, δL\delta L must be linear in velocity. That is, there exists a constant vector gRng \in \mathbb{R}^n such that for all vRnv \in \mathbb{R}^n: δL(v)=g,v\delta L(v) = \langle g, v \rangle where ,\langle \cdot, \cdot \rangle denotes the standard Euclidean inner product. This reflects the physical requirement that for δL\delta L to be independent of position rr and time tt, the gradient rF\nabla_r F must be a constant vector gg and the partial derivative Ft\frac{\partial F}{\partial t} must vanish.

theorem

If δL\delta L is a total time derivative, then δL-\delta L is a total time derivative

If a function δL(t,q,v)\delta L(t, q, v) is a total time derivative, meaning there exists a differentiable function F(q,t)F(q, t) such that δL(t,q,v)=ddtF(q,t)\delta L(t, q, v) = \frac{d}{dt} F(q, t), then its negation δL-\delta L is also a total time derivative.

theorem

Total Time Derivatives are CC^\infty Smooth

Let δL(t,q,v)\delta L(t, q, v) be a function representing a change in a Lagrangian, where tTimet \in \text{Time} and q,vXq, v \in X. If δL\delta L is a total time derivative—meaning there exists a smooth function F(q,t)F(q, t) such that δL(t,q,v)=ddtF(q,t)\delta L(t, q, v) = \frac{d}{dt} F(q, t)—then δL\delta L is infinitely differentiable (CC^\infty).

theorem

The variational derivative of a total time derivative is zero

Let XX be a complete space. Suppose δL:TimeXXR\delta L: \text{Time} \to X \to X \to \mathbb{R} is a total time derivative (meaning there exists a function F(q,t)F(q, t) such that δL(t,q,v)=ddtF(q,t)\delta L(t, q, v) = \frac{d}{dt} F(q, t)). For any infinitely differentiable path q:TimeXq: \text{Time} \to X, the variational gradient of the functional mapping qq' to the density δL(t,q(t),tq(t))\delta L(t, q'(t), \partial_t q'(t)) at the path qq is the zero function. That is, the functional derivative of the action corresponding to δL\delta L is zero: δδqδL(t,q(t),q˙(t))dt=0\frac{\delta}{\delta q} \int \delta L(t, q(t), \dot{q}(t)) \, dt = 0

theorem

Addition of a Total Time Derivative Preserves the Variational Gradient of a Lagrangian

Let XX be a complete space. Let LL and δL\delta L be Lagrangians mapping TimeXXR\text{Time} \to X \to X \to \mathbb{R}, where L(t,q,q˙)L(t, q, \dot{q}) and δL(t,q,q˙)\delta L(t, q, \dot{q}) represent the Lagrangian densities. Suppose δL\delta L is a total time derivative (i.e., there exists a function F(q,t)F(q, t) such that δL=ddtF\delta L = \frac{d}{dt}F). For any CC^\infty trajectory q:TimeXq: \text{Time} \to X, if the functional S[q]=L(t,q(t),q˙(t))dtS[q] = \int L(t, q(t), \dot{q}(t)) \, dt has a variational gradient grad\text{grad} at qq, then the functional S[q]=(L+δL)(t,q(t),q˙(t))dtS'[q] = \int (L + \delta L)(t, q(t), \dot{q}(t)) \, dt also has the same variational gradient grad\text{grad} at qq.

theorem

δSδq=δSδq\frac{\delta S'}{\delta q} = \frac{\delta S}{\delta q} for Lagrangians Differing by a Total Time Derivative

Let XX be a complete space. Let L,L:Time×X×XRL, L' : \text{Time} \times X \times X \to \mathbb{R} be two Lagrangians such that their difference LLL' - L is a total time derivative. For any infinitely differentiable trajectory q:TimeXq : \text{Time} \to X, the variational gradient of the action functional S[q]=L(t,q(t),q˙(t))dtS'[q] = \int L'(t, q(t), \dot{q}(t)) \, dt is equal to the variational gradient of the action functional S[q]=L(t,q(t),q˙(t))dtS[q] = \int L(t, q(t), \dot{q}(t)) \, dt at qq: δSδq=δSδq \frac{\delta S'}{\delta q} = \frac{\delta S}{\delta q} where q˙(t)\dot{q}(t) denotes the time derivative of qq.

theorem

Equality of Euler-Lagrange Operators for Lagrangians Differing by a Total Time Derivative

Let XX be a complete space and let L,L:TimeXXRL, L' : \text{Time} \to X \to X \to \mathbb{R} be two Lagrangians. If LLL' - L is a total time derivative (that is, there exists a function F(q,t)F(q, t) such that (LL)(t,q,v)=Ft+qFv(L' - L)(t, q, v) = \frac{\partial F}{\partial t} + \nabla_q F \cdot v), and assuming that either LL or LL' is infinitely differentiable (CC^\infty) and the trajectory q:TimeXq: \text{Time} \to X is infinitely differentiable, then the Euler-Lagrange operators for LL and LL' are identical when evaluated along qq: LqddtLq˙=LqddtLq˙ \frac{\partial L}{\partial q} - \frac{d}{dt} \frac{\partial L}{\partial \dot{q}} = \frac{\partial L'}{\partial q} - \frac{d}{dt} \frac{\partial L'}{\partial \dot{q}}