ProofComplete

Proof of the First and Second Variation Formulas

The variation formulas compute how the energy of a curve changes under deformation, establishing geodesics as critical points and providing the tools for analyzing their stability.


Setup

Definition8.9Variation of a curve

A variation of a curve γ:[a,b]M\gamma: [a,b] \to M is a smooth map Γ:[a,b]×(ϵ,ϵ)M\Gamma: [a,b] \times (-\epsilon, \epsilon) \to M with Γ(t,0)=γ(t)\Gamma(t, 0) = \gamma(t). The variation vector field is V(t)=Γs(t,0)V(t) = \frac{\partial \Gamma}{\partial s}(t, 0). The variation is proper (fixed-endpoint) if Γ(a,s)=γ(a)\Gamma(a, s) = \gamma(a) and Γ(b,s)=γ(b)\Gamma(b, s) = \gamma(b) for all ss.


First Variation Formula

Theorem8.8First variation of energy

Let γ:[a,b]M\gamma: [a,b] \to M be a smooth curve and Γ\Gamma a proper variation with variation vector field VV. Then

dEdss=0=abg(γγ,V)dt.\frac{dE}{ds}\Big|_{s=0} = -\int_a^b g(\nabla_{\gamma'}\gamma', V) \, dt.
Proof

Write T=ΓtT = \frac{\partial \Gamma}{\partial t} and V=ΓsV = \frac{\partial \Gamma}{\partial s}. Then:

dEds=12ddsabg(T,T)dt=abg(VT,T)dt.\frac{dE}{ds} = \frac{1}{2}\frac{d}{ds}\int_a^b g(T, T) \, dt = \int_a^b g(\nabla_V T, T) \, dt.

By the symmetry lemma: VT=TV\nabla_V T = \nabla_T V (since [T,V]=[t,s]=0[T, V] = [\partial_t, \partial_s] = 0). So:

dEds=abg(TV,T)dt=ab(ddtg(V,T)g(V,TT))dt.\frac{dE}{ds} = \int_a^b g(\nabla_T V, T) \, dt = \int_a^b \left(\frac{d}{dt}g(V, T) - g(V, \nabla_T T)\right) dt.

The first term integrates to g(V,T)ab=0g(V, T)\Big|_a^b = 0 (proper variation: V(a)=V(b)=0V(a) = V(b) = 0). Thus:

dEdss=0=abg(V,γγ)dt.\frac{dE}{ds}\Big|_{s=0} = -\int_a^b g(V, \nabla_{\gamma'}\gamma') \, dt. \quad \blacksquare
RemarkGeodesics as critical points

The first variation vanishes for all proper variations if and only if γγ=0\nabla_{\gamma'}\gamma' = 0, i.e., γ\gamma is a geodesic. This is the Euler-Lagrange equation for the energy functional.


Second Variation Formula

Theorem8.9Second variation of energy

Let γ:[a,b]M\gamma: [a,b] \to M be a geodesic and Γ\Gamma a proper variation with variation vector field VV. Then

d2Eds2s=0=ab(g(γV,γV)g(R(V,γ)γ,V))dt=ab(V2K(V,γ)Vγ2)dt,\frac{d^2 E}{ds^2}\Big|_{s=0} = \int_a^b \left(g(\nabla_{\gamma'} V, \nabla_{\gamma'} V) - g(R(V, \gamma')\gamma', V)\right) dt = \int_a^b \left(|V'|^2 - K(V, \gamma')|V \wedge \gamma'|^2\right) dt,

where V=γVV' = \nabla_{\gamma'} V and KK is the sectional curvature.

Proof

Differentiating dEds=g(TV,T)dt\frac{dE}{ds} = \int g(\nabla_T V, T) \, dt once more:

d2Eds2=ab(g(VTV,T)+g(TV,VT))dt.\frac{d^2E}{ds^2} = \int_a^b \left(g(\nabla_V \nabla_T V, T) + g(\nabla_T V, \nabla_V T)\right) dt.

For the first term, use VTV=TVV+R(V,T)V\nabla_V \nabla_T V = \nabla_T \nabla_V V + R(V, T)V (definition of curvature and [T,V]=0[T,V] = 0). At s=0s = 0, TT=0\nabla_T T = 0 (geodesic), so integrating by parts:

g(TVV,T)dt=g(VV,T)abg(VV,TT)dt=0.\int g(\nabla_T \nabla_V V, T) \, dt = g(\nabla_V V, T)\Big|_a^b - \int g(\nabla_V V, \nabla_T T) \, dt = 0.

The curvature term gives g(R(V,T)V,T)dt=g(R(V,T)T,V)dt\int g(R(V, T)V, T) \, dt = -\int g(R(V, T)T, V) \, dt. The second term in the original sum is TV2dt\int |\nabla_T V|^2 \, dt. Combining at s=0s = 0:

d2Eds2s=0=ab(γV2g(R(V,γ)γ,V))dt.\frac{d^2E}{ds^2}\Big|_{s=0} = \int_a^b \left(|\nabla_{\gamma'} V|^2 - g(R(V, \gamma')\gamma', V)\right) dt. \quad \blacksquare

The Index Form

Definition8.10Index form

The index form along a geodesic γ\gamma is the bilinear form on vector fields along γ\gamma vanishing at endpoints:

I(V,W)=ab(g(γV,γW)g(R(V,γ)γ,W))dt.I(V, W) = \int_a^b \left(g(\nabla_{\gamma'} V, \nabla_{\gamma'} W) - g(R(V, \gamma')\gamma', W)\right) dt.

The second variation equals I(V,V)I(V, V). The geodesic is a local minimum if and only if II is positive definite, which holds precisely when there are no conjugate points in (a,b)(a, b).