Book 9A

© 2026 NeckPinch · www.neckpinch.com · All rights reserved.

Course 9Book 9A: Metrics, Connections and CurvatureChapter 7

Jacobi Fields and Curvature versus Topology

Variations of length, gravitational lensing, Myers and Cartan–Hadamard.

26 min read · Updated Oct 3, 2026

Read with Lee, Introduction to Riemannian Manifolds (2nd edition), chapter 6 (the first variation formula), chapter 10 (Jacobi fields, conjugate points and the second variation formula) and chapters 11–12 (comparison theory, Myers' and the Cartan–Hadamard theorems). Petersen's Riemannian Geometry, chapter 6, is the second voice.

In this chapter · 7 sections
  1. 7.1Gravitational lensing
  2. 7.2Variations of length
  3. 7.3Jacobi fields
  4. 7.4Positive curvature: Bonnet–Myers
  5. 7.5Nonpositive curvature: Cartan–Hadamard
  6. 7.6History
  7. 7.7Exercises

Curvature is local: it is computed at a point from second derivatives of the metric. Topology is global. This chapter shows how to get from one to the other, and the bridge is the behaviour of families of geodesics. Nearby geodesics leaving a point separate at a rate governed by curvature, through the Jacobi equation, a linear ODE along each geodesic. Positive curvature makes them reconverge, so geodesics stop minimising after a bounded length. Negative curvature keeps them apart forever.

Two theorems turn this into topology. Bonnet–Myers: a complete manifold with Ricci curvature bounded below by a positive constant is compact, with finite fundamental group. Cartan–Hadamard: a complete, simply connected manifold with nonpositive sectional curvature is diffeomorphic to Rn\mathbb{R}^n. The Ricci flow's whole strategy is to make curvature positive, or to cut the manifold into pieces where it is, so that theorems of this kind identify the topology. Hamilton's 1982 theorem starts where Myers stops.

By the end of this chapter you will be able to:

  • derive the first and second variation formulas for length, and write the index form;
  • define Jacobi fields, solve the Jacobi equation in constant curvature, and relate Jacobi fields to the exponential map;
  • define conjugate points and explain why geodesics stop minimising past them;
  • prove the Bonnet–Myers theorem, and state Synge's and the Cartan–Hadamard theorems;
  • derive the normal-coordinate expansion of the metric from Jacobi fields.

Gravitational lensing

In the world Data Light refocused by a galaxy

Light rays are geodesics of spacetime (9A.3 Geodesics and the Exponential Map). A large mass between us and a distant source curves spacetime so that rays from the source that pass on different sides of the mass are bent towards each other and reconverge at the observer. We then see the source more than once. The first such gravitational lens was found in 1979, when Dennis Walsh, Robert Carswell and Ray Weymann noticed that two quasars a few arcseconds apart, QSO 0957+561, had identical spectra: two images of one quasar, lensed by a galaxy in a cluster in front of it. In the Einstein Cross, Q2237+0305, discovered in 1985, a quasar at redshift about 1.71.7 appears as four images around the core of a galaxy at redshift about 0.040.04. When source, lens and observer are aligned the images merge into an Einstein ring, and the Hubble and James Webb space telescopes have photographed many.

Geometrically, the observer sits at a conjugate point of the source: a family of light rays leaving the source refocuses there. The separation between neighbouring rays is a Jacobi field, and the lens works because the matter in the galaxy makes the Ricci curvature positive along the rays, which focuses them, as in 9A.4 Curvature and What It Means.

Figure 7.1. A point mass lensing a source slightly off the axis, in the thin-lens approximation (angles exaggerated). Two rays, deflected at the lens by angles inversely proportional to their distance from it, reach the observer; traced back in straight lines (dashed) they give two images on opposite sides of the lens. The image positions are the two solutions of the lens equation β=θ−θE2θ\beta = \theta - \frac{\theta_E^2}{\theta}, computed in the figure script.

Variations of length

A variation of a curve γ:[0,L]→M\gamma : [0, L] \to M is a smooth family Γ(s,t)\Gamma(s, t) with Γ(0,t)=γ(t)\Gamma(0, t) = \gamma(t); its variation field is V(t)=∂sΓ(0,t)V(t) = \partial_s\Gamma(0, t). It is proper if the endpoints are fixed. Write T=∂tΓT = \partial_t\Gamma, S=∂sΓS = \partial_s\Gamma, and recall the symmetry lemma DsT=DtSD_sT = D_tS (from torsion-freeness, used in the Gauss lemma of 9A.3 Geodesics and the Exponential Map).

Theorem 7.1 First variation

If γ\gamma has unit speed, then for any variation

dds∣s=0L(Γ(s,⋅))=⟨V,γ′⟩∣0L−∫0L⟨V,Dtγ′⟩ dt.\frac{d}{ds}\Big|_{s = 0}L(\Gamma(s, \cdot)) = \langle V, \gamma'\rangle\Big|_0^L - \int_0^L\langle V, D_t\gamma'\rangle\,dt.

Proof. dds∣T∣=⟨DsT,T⟩∣T∣=⟨DtS,T⟩\frac{d}{ds}|T| = \frac{\langle D_sT, T\rangle}{|T|} = \langle D_tS, T\rangle at s=0s = 0, where ∣T∣=1|T| = 1. Then ⟨DtS,T⟩=∂t⟨S,T⟩−⟨S,DtT⟩\langle D_tS, T\rangle = \partial_t\langle S, T\rangle - \langle S, D_tT\rangle; integrate in tt.

So a unit-speed curve is a critical point of length for all proper variations exactly when Dtγ′=0D_t\gamma' = 0: geodesics are the critical points of length. Whether a geodesic minimises is a second-order question.

Theorem 7.2 Second variation

If γ\gamma is a unit-speed geodesic and Γ\Gamma is a proper variation whose variation field VV is orthogonal to γ′\gamma', then

d2ds2∣s=0L(Γ(s,⋅))=I(V,V)=∫0L(∣DtV∣2−Rm⁡(V,γ′,γ′,V)) dt,\frac{d^2}{ds^2}\Big|_{s = 0}L(\Gamma(s, \cdot)) = I(V, V) = \int_0^L\Big(|D_tV|^2 - \operatorname{Rm}(V, \gamma', \gamma', V)\Big)\,dt,

where Rm⁡(V,γ′,γ′,V)=K(V,γ′)∣V∣2\operatorname{Rm}(V, \gamma', \gamma', V) = K(V, \gamma')|V|^2. The symmetric bilinear form II on such fields is the index form.

Proof. Differentiate ddsL=∫⟨DtS,T⟩∣T∣dt\frac{d}{ds}L = \int\frac{\langle D_tS, T\rangle}{|T|}dt once more in ss, at s=0s = 0 where ∣T∣=1|T| = 1 and DtT=0D_tT = 0:

d2Lds2=∫(⟨DsDtS,T⟩+⟨DtS,DsT⟩−⟨DtS,T⟩2)dt.\frac{d^2L}{ds^2} = \int\Big(\langle D_sD_tS, T\rangle + \langle D_tS, D_sT\rangle - \langle D_tS, T\rangle^2\Big)dt.

Use DsT=DtSD_sT = D_tS in the second term, and in the first, DsDtS=DtDsS+R(S,T)SD_sD_tS = D_tD_sS + R(S, T)S (the definition of curvature, with [∂s,∂t]=0[\partial_s, \partial_t] = 0). The term ⟨DtDsS,T⟩=∂t⟨DsS,T⟩\langle D_tD_sS, T\rangle = \partial_t\langle D_sS, T\rangle integrates to zero for a proper variation, and ⟨R(S,T)S,T⟩=Rm⁡(S,T,S,T)=−Rm⁡(V,γ′,γ′,V)\langle R(S, T)S, T\rangle = \operatorname{Rm}(S, T, S, T) = -\operatorname{Rm}(V, \gamma', \gamma', V). Finally ⟨DtS,T⟩=∂t⟨V,γ′⟩=0\langle D_tS, T\rangle = \partial_t\langle V, \gamma'\rangle = 0 since V⊥γ′V \perp \gamma'.

Curvature enters with a minus sign. Positive sectional curvature along a geodesic makes some variations shorten it, and a long enough geodesic on a positively curved manifold can always be shortened.

Jacobi fields

A Jacobi field along a geodesic γ\gamma is a vector field JJ along it satisfying the Jacobi equation

Dt2J+R(J,γ′)γ′=0.D_t^2J + R(J, \gamma')\gamma' = 0.

It is a linear second-order ODE, so a Jacobi field is determined by J(0)J(0) and DtJ(0)D_tJ(0), and they form a 2n2n-dimensional space. The variation field of any variation through geodesics is a Jacobi field (Exercise 7.6), and conversely. In particular, for a geodesic γ(t)=exp⁡p(tv)\gamma(t) = \exp_p(tv) and w∈TpMw \in T_pM,

J(t)=d(exp⁡p)tv(tw)J(t) = d(\exp_p)_{tv}(tw)

is the Jacobi field with J(0)=0J(0) = 0 and DtJ(0)=wD_tJ(0) = w, the variation field of the fan of geodesics s↦exp⁡p(t(v+sw))s \mapsto \exp_p(t(v + sw)). Jacobi fields measure how fast nearby geodesics from pp spread apart.

Constant curvature. If K≡kK \equiv k, then R(J,γ′)γ′=k JR(J, \gamma')\gamma' = k\,J for unit-speed γ\gamma and J⊥γ′J \perp \gamma', and the Jacobi fields vanishing at 00 and orthogonal to γ\gamma are

J(t)=sn⁡k(t) E(t),sn⁡k(t)={sin⁡(kt)k,k>0,t,k=0,sinh⁡(−kt)−k,k<0,J(t) = \operatorname{sn}_k(t)\,E(t), \qquad \operatorname{sn}_k(t) = \begin{cases}\frac{\sin(\sqrt kt)}{\sqrt k}, & k > 0,\\ t, & k = 0,\\ \frac{\sinh(\sqrt{-k}t)}{\sqrt{-k}}, & k < 0,\end{cases}

with EE parallel (Exercise 7.7, Figure 7.2). On the sphere they vanish again at t=πkt = \frac{\pi}{\sqrt k}; in the plane they grow linearly; in hyperbolic space exponentially. These are the functions sn⁡k\operatorname{sn}_k of the polar metrics of 9A.1 Riemannian Metrics and Model Spaces, dr2+sn⁡k(r)2gSn−1dr^2 + \operatorname{sn}_k(r)^2g_{S^{n-1}}, for the same reason.

Figure 7.2. The length of a Jacobi field vanishing at t=0t = 0 with ∣DtJ(0)∣=1|D_tJ(0)| = 1, in curvature 11, 00 and −1-1: sin⁡t\sin t, tt and sinh⁡t\sinh t. On the sphere it returns to zero at t=πt = \pi, a conjugate point.

Conjugate points. γ(t0)\gamma(t_0) is conjugate to γ(0)\gamma(0) along γ\gamma if a nonzero Jacobi field vanishes at both. Equivalently, d(exp⁡p)d(\exp_p) is singular at t0vt_0v: the fan of geodesics from pp focuses at γ(t0)\gamma(t_0). A geodesic does not minimise length past its first conjugate point: the index form then has a negative direction (Jacobi's theorem, Lee, chapter 10). Lensing is this: the source and the observer are conjugate along the rays.

Proposition 7.3 The metric in normal coordinates

In normal coordinates centred at pp,

gij(x)=δij−13Riklj(p)xkxl+O(∣x∣3).g_{ij}(x) = \delta_{ij} - \tfrac13R_{iklj}(p)x^kx^l + O(|x|^3).

Proof. Fix a unit vv and any ww, and let J(t)=d(exp⁡p)tv(tw)J(t) = d(\exp_p)_{tv}(tw), the Jacobi field with J(0)=0J(0) = 0, DtJ(0)=wD_tJ(0) = w. In normal coordinates exp⁡p\exp_p is the identity map of coordinates, so J(t)J(t) is the coordinate vector twtw at the point tvtv, and gtv(w,w)=∣J(t)∣2/t2g_{tv}(w, w) = |J(t)|^2/t^2. In a parallel frame along γ\gamma, DtD_t is the ordinary derivative, and the Jacobi equation gives Dt2J(0)=−R(J(0),v)v=0D_t^2J(0) = -R(J(0), v)v = 0 and Dt3J(0)=−R(DtJ(0),v)v=−R(w,v)vD_t^3J(0) = -R(D_tJ(0), v)v = -R(w, v)v (the term with the derivative of RR is multiplied by J(0)=0J(0) = 0). So J(t)=tw−t36R(w,v)v+O(t4)J(t) = tw - \frac{t^3}{6}R(w, v)v + O(t^4) and

∣J(t)∣2=t2∣w∣2−t43Rm⁡(w,v,v,w)+O(t5).|J(t)|^2 = t^2|w|^2 - \frac{t^4}{3}\operatorname{Rm}(w, v, v, w) + O(t^5).

Dividing by t2t^2 and writing x=tvx = tv: gx(w,w)=∣w∣2−13Rm⁡(w,x,x,w)+O(∣x∣3)g_x(w, w) = |w|^2 - \frac13\operatorname{Rm}(w, x, x, w) + O(|x|^3), which is the stated formula, both sides being quadratic forms in ww.

This is the expansion used in 9A.3 Geodesics and the Exponential Map and 9A.4 Curvature and What It Means; for unit w⊥vw \perp v it gives circumference 2πr(1−K6r2+… )2\pi r(1 - \frac K6r^2 + \dots) of small geodesic circles (Exercise 7.11).

Positive curvature: Bonnet–Myers

Theorem 7.4 Bonnet–Myers

Let MM be complete and connected with Ric⁡≥(n−1)k g\operatorname{Ric} \geq (n - 1)k\,g for a constant k>0k > 0. Then diam⁡(M)≤πk\operatorname{diam}(M) \leq \frac{\pi}{\sqrt k}. Consequently MM is compact and π1(M)\pi_1(M) is finite.

Proof. Let p,q∈Mp, q \in M with d(p,q)=Ld(p, q) = L, and let γ:[0,L]→M\gamma : [0, L] \to M be a unit-speed minimising geodesic between them (Hopf–Rinow, 9A.3 Geodesics and the Exponential Map). Then I(V,V)≥0I(V, V) \geq 0 for every proper V⊥γ′V \perp \gamma', by Theorem 7.2. Take parallel orthonormal fields E1,…,En−1E_1, \dots, E_{n-1} orthogonal to γ′\gamma', and Vi=sin⁡πtLEiV_i = \sin\frac{\pi t}{L}E_i. Then ∣DtVi∣2=π2L2cos⁡2πtL|D_tV_i|^2 = \frac{\pi^2}{L^2}\cos^2\frac{\pi t}{L} and ∑iRm⁡(Vi,γ′,γ′,Vi)=sin⁡2πtLRic⁡(γ′,γ′)\sum_i\operatorname{Rm}(V_i, \gamma', \gamma', V_i) = \sin^2\frac{\pi t}{L}\operatorname{Ric}(\gamma', \gamma'), so

0≤∑iI(Vi,Vi)≤(n−1)∫0L(π2L2cos⁡2πtL−ksin⁡2πtL)dt=(n−1)L2(π2L2−k),0 \leq \sum_iI(V_i, V_i) \leq (n - 1)\int_0^L\Big(\frac{\pi^2}{L^2}\cos^2\frac{\pi t}{L} - k\sin^2\frac{\pi t}{L}\Big)dt = (n - 1)\frac L2\Big(\frac{\pi^2}{L^2} - k\Big),

which forces L≤πkL \leq \frac{\pi}{\sqrt k}. A complete manifold of bounded diameter is compact (Hopf–Rinow). The universal cover, with the pulled-back metric, is complete with the same Ricci bound, so it is compact too; hence the covering has finitely many sheets and π1(M)\pi_1(M) is finite.

The unit sphere shows the diameter bound is sharp. Some consequences:

  • S2×S1S^2\times S^1, T3T^3 and every manifold with infinite fundamental group carry no metric with Ric⁡>0\operatorname{Ric} > 0, although S2×S1S^2\times S^1 carries one with Ric⁡≥0\operatorname{Ric} \geq 0 (9A.5 Computing Curvature).
  • A closed 3-manifold with Ric⁡>0\operatorname{Ric} > 0 has finite fundamental group. Hamilton's theorem (11A.6 Hamilton’s 1982 Theorem) goes much further: such a manifold is diffeomorphic to a quotient S3/ΓS^3/\Gamma of the round sphere, because the normalised Ricci flow deforms the metric to one of constant positive curvature.

Synge's theorem (1936) uses the same second variation idea along a shortest closed geodesic: a compact, even-dimensional manifold with positive sectional curvature is simply connected if it is orientable, and has fundamental group Z/2\mathbb{Z}/2 if it is not. So RP2×RP2\mathbb{RP}^2\times\mathbb{RP}^2, compact and four-dimensional with fundamental group of order 44, has no metric of positive sectional curvature.

Nonpositive curvature: Cartan–Hadamard

Theorem 7.5 Cartan–Hadamard

If MM is complete and connected with sectional curvature K≤0K \leq 0, then for every pp the map exp⁡p:TpM→M\exp_p : T_pM \to M is a covering map. In particular the universal cover of MM is diffeomorphic to Rn\mathbb{R}^n, and if MM is simply connected, exp⁡p\exp_p is a diffeomorphism.

The key step is that there are no conjugate points: if K≤0K \leq 0 and J(0)=0J(0) = 0, then d2dt2∣J∣2=2∣DtJ∣2−2Rm⁡(J,γ′,γ′,J)≥0\frac{d^2}{dt^2}|J|^2 = 2|D_tJ|^2 - 2\operatorname{Rm}(J, \gamma', \gamma', J) \geq 0, so ∣J∣2|J|^2 is convex and a nonzero Jacobi field cannot vanish twice (Exercise 7.8). So exp⁡p\exp_p is a local diffeomorphism everywhere, and completeness makes it a covering map (Lee, chapter 12). Consequences:

  • Manifolds with K≤0K \leq 0 are aspherical: all higher homotopy groups vanish. Flat tori, hyperbolic surfaces and closed hyperbolic 3-manifolds are examples, and the hyperbolic pieces of the geometrization theorem are of this kind (10A.6 Hyperbolic Three-Manifolds).
  • SnS^n, n≥2n \geq 2, has no metric with K≤0K \leq 0: it is compact and simply connected, so it would be diffeomorphic to Rn\mathbb{R}^n.
In the world Model Satellites in neighbouring orbits

Spacecraft rendezvous and formation flying rely on the equations for the motion of one satellite relative to another in a nearby circular orbit, the Hill–Clohessy–Wiltshire equations (W. H. Clohessy and R. S. Wiltshire, 1960, after G. W. Hill's 1878 lunar theory). They are the Newtonian analogue of the Jacobi equation: the linearised equation for the separation of two neighbouring free-fall paths. In the direction perpendicular to the orbital plane the equation is z¨+n2z=0\ddot z + n^2z = 0, where nn is the orbital angular rate, with solutions sin⁡nt\sin nt, exactly the Jacobi fields of a positively curved space. Two satellites in slightly tilted circular orbits of the same radius therefore cross each other's orbital plane twice per orbit, half an orbit apart: the out-of-plane motion has a "conjugate point" every half revolution, just as the meridians from a pole meet again after length π\pi.

Where this goes Comparison geometry

The arguments of this chapter compare a manifold with a model space through its Jacobi fields. 9B.1 Laplacian Comparison and 9B.2 Volume Comparison turn them into comparison theorems for the Laplacian of the distance function and for volumes (Bishop–Gromov), the tools behind Perelman's noncollapsing and every compactness argument of the Ricci flow. The second variation formula returns for surfaces in 9A.8 Submanifolds and Minimal Surfaces, with the Ricci curvature of the ambient space in place of the sectional curvature.

History

Carl Jacobi studied the second variation of geodesics on surfaces in 1837, and Pierre Ossian Bonnet proved the diameter bound for surfaces with Gauss curvature bounded below in 1855. Sumner Byron Myers extended it to Ricci curvature in 1941. John Lighton Synge proved his theorem in 1936. Jacques Hadamard showed in 1898 that surfaces of negative curvature have no conjugate points, and Élie Cartan proved the general covering theorem in 1928. Marston Morse developed the index theory of geodesics in the 1930s. The first gravitational lens was found in 1979.

Recall Where we stand

Geodesics are critical points of length; the second variation along a geodesic is the index form I(V,V)=∫(∣V′∣2−Rm⁡(V,γ′,γ′,V))I(V, V) = \int(|V'|^2 - \operatorname{Rm}(V, \gamma', \gamma', V)), in which curvature appears with a minus sign. Jacobi fields Dt2J+R(J,γ′)γ′=0D_t^2J + R(J, \gamma')\gamma' = 0 are variation fields of geodesic families, and d(exp⁡p)tv(tw)d(\exp_p)_{tv}(tw) is one; in constant curvature they are sn⁡k(t)E(t)\operatorname{sn}_k(t)E(t), and on the sphere they vanish again at conjugate points. Jacobi fields give the normal-coordinate expansion gij=δij−13Rikljxkxl+…g_{ij} = \delta_{ij} - \frac13R_{iklj}x^kx^l + \dots. Bonnet–Myers: Ric⁡≥(n−1)k>0\operatorname{Ric} \geq (n - 1)k > 0 forces diameter ≤π/k\leq \pi/\sqrt k, compactness and finite π1\pi_1. Cartan–Hadamard: K≤0K \leq 0 and complete makes exp⁡p\exp_p a covering, so the universal cover is Rn\mathbb{R}^n. 9A.8 Submanifolds and Minimal Surfaces moves from curves to surfaces: submanifolds, mean curvature and minimal surfaces.

Exercises

Exercise 7.6 Variations through geodesics give Jacobi fields

If every curve t↦Γ(s,t)t \mapsto \Gamma(s, t) is a geodesic, show that V=∂sΓ∣s=0V = \partial_s\Gamma|_{s = 0} satisfies the Jacobi equation. (Differentiate DtT=0D_tT = 0 in ss, use DsDtT−DtDsT=R(S,T)TD_sD_tT - D_tD_sT = R(S, T)T and the symmetry lemma.)

Solution

0=DsDtT=DtDsT+R(S,T)T=DtDtS+R(S,T)T0 = D_sD_tT = D_tD_sT + R(S, T)T = D_tD_tS + R(S, T)T. At s=0s = 0: Dt2V+R(V,γ′)γ′=0D_t^2V + R(V, \gamma')\gamma' = 0.

Exercise 7.7 Jacobi fields in constant curvature

Using R(X,Y)Z=k(⟨Y,Z⟩X−⟨X,Z⟩Y)R(X, Y)Z = k(\langle Y, Z\rangle X - \langle X, Z\rangle Y) for constant curvature kk (from Rijkl=k(gilgjk−gikgjl)R_{ijkl} = k(g_{il}g_{jk} - g_{ik}g_{jl})), show that for unit-speed γ\gamma and J⊥γ′J \perp \gamma', R(J,γ′)γ′=kJR(J, \gamma')\gamma' = kJ. Deduce that J=f(t)E(t)J = f(t)E(t), with EE parallel, is a Jacobi field exactly when f′′+kf=0f'' + kf = 0, and find the first conjugate point.

Solution

R(J,γ′)γ′=k(⟨γ′,γ′⟩J−⟨J,γ′⟩γ′)=kJR(J, \gamma')\gamma' = k(\langle\gamma', \gamma'\rangle J - \langle J, \gamma'\rangle\gamma') = kJ. For J=fEJ = fE, Dt2J=f′′ED_t^2J = f''E, so the equation is (f′′+kf)E=0(f'' + kf)E = 0. With f(0)=0f(0) = 0, f=csn⁡kf = c\operatorname{sn}_k; for k>0k > 0 it vanishes again first at t=π/kt = \pi/\sqrt k, the first conjugate point; for k≤0k \leq 0 never.

Exercise 7.8 No conjugate points in nonpositive curvature

If K≤0K \leq 0 along γ\gamma and JJ is a Jacobi field, show d2dt2∣J∣2=2∣DtJ∣2−2Rm⁡(J,γ′,γ′,J)≥0\frac{d^2}{dt^2}|J|^2 = 2|D_tJ|^2 - 2\operatorname{Rm}(J, \gamma', \gamma', J) \geq 0. Deduce that a Jacobi field with J(0)=0J(0) = 0 and DtJ(0)≠0D_tJ(0) \neq 0 never vanishes again.

Solution

ddt∣J∣2=2⟨DtJ,J⟩\frac{d}{dt}|J|^2 = 2\langle D_tJ, J\rangle and d2dt2∣J∣2=2∣DtJ∣2+2⟨Dt2J,J⟩=2∣DtJ∣2−2⟨R(J,γ′)γ′,J⟩\frac{d^2}{dt^2}|J|^2 = 2|D_tJ|^2 + 2\langle D_t^2J, J\rangle = 2|D_tJ|^2 - 2\langle R(J, \gamma')\gamma', J\rangle, and ⟨R(J,γ′)γ′,J⟩=Rm⁡(J,γ′,γ′,J)≤0\langle R(J, \gamma')\gamma', J\rangle = \operatorname{Rm}(J, \gamma', \gamma', J) \leq 0. So ∣J∣2|J|^2 is convex, with ∣J∣2(0)=0|J|^2(0) = 0 and ∣J∣2=t2∣DtJ(0)∣2+O(t3)>0|J|^2 = t^2|D_tJ(0)|^2 + O(t^3) > 0 for small t>0t > 0. A convex function that is zero at 00 and positive just after cannot return to zero.

Exercise 7.9 Using Myers

(a) Show the diameter bound is attained by the unit sphere. (b) Explain why the paraboloid z=x2+y2z = x^2 + y^2, which has K>0K > 0, does not contradict Myers. (c) Show that a compact manifold with Ric⁡>0\operatorname{Ric} > 0 (not bounded below by a constant a priori) also has finite π1\pi_1.

Solution

(a) Ric⁡=(n−1)g\operatorname{Ric} = (n - 1)g, k=1k = 1, and antipodal points are at distance π\pi. (b) Its curvature tends to 00 at infinity, so there is no positive lower bound kk. (c) On a compact manifold the continuous function min⁡∣v∣=1Ric⁡(v,v)\min_{|v| = 1}\operatorname{Ric}(v, v) attains a positive minimum, which gives a k>0k > 0.

Exercise 7.10 Out-of-plane motion of satellites

Two satellites move in circular orbits of the same radius around the Earth, in planes tilted by a small angle δ\delta. Show that their separation perpendicular to one orbital plane is approximately z(t)=aδsin⁡ntz(t) = a\delta\sin nt, where aa is the orbit radius, and that it vanishes twice per orbit. For an orbit of period 9090 minutes, how often do they pass through each other's plane?

Solution

Take the line of intersection of the two planes as the reference. The tilted satellite's height above the first plane is asin⁡δsin⁡(nt)≈aδsin⁡nta\sin\delta\sin(nt) \approx a\delta\sin nt, which solves z¨+n2z=0\ddot z + n^2z = 0. It vanishes when ntnt is a multiple of π\pi, so every half orbit: every 4545 minutes.

Exercise 7.11 Rehearsal: circumference from Jacobi fields

For a plane Π⊂TpM\Pi \subset T_pM with orthonormal basis v,wv, w, the circle of radius rr in Π\Pi maps under exp⁡p\exp_p to the curve Cr(θ)=exp⁡p(r(cos⁡θ v+sin⁡θ w))C_r(\theta) = \exp_p(r(\cos\theta\,v + \sin\theta\,w)). Show ∣Cr′(θ)∣=∣Jθ(r)∣|C_r'(\theta)| = |J_\theta(r)|, where JθJ_\theta is the Jacobi field along the geodesic in direction uθ=cos⁡θ v+sin⁡θ wu_\theta = \cos\theta\,v + \sin\theta\,w with J(0)=0J(0) = 0 and DtJ(0)=uθ′D_tJ(0) = u_\theta', a unit vector orthogonal to uθu_\theta. Use the expansion in Proposition 7.3 to show ∣Jθ(r)∣=r−K(Π)6r3+O(r4)|J_\theta(r)| = r - \frac{K(\Pi)}{6}r^3 + O(r^4), and integrate in θ\theta to obtain the circumference formula of 9A.4 Curvature and What It Means.

Solution

∂θCr=d(exp⁡p)ruθ(ruθ′)\partial_\theta C_r = d(\exp_p)_{ru_\theta}(ru_\theta'), which is Jθ(r)J_\theta(r) for the Jacobi field d(exp⁡p)tuθ(tuθ′)d(\exp_p)_{tu_\theta}(tu_\theta'). From the proof of the proposition, ∣J(t)∣2=t2−t43Rm⁡(uθ′,uθ,uθ,uθ′)+O(t5)=t2−t43K(Π)+O(t5)|J(t)|^2 = t^2 - \frac{t^4}{3}\operatorname{Rm}(u_\theta', u_\theta, u_\theta, u_\theta') + O(t^5) = t^2 - \frac{t^4}{3}K(\Pi) + O(t^5), since uθ,uθ′u_\theta, u_\theta' is an orthonormal basis of Π\Pi. So ∣J(r)∣=r(1−K6r2+O(r3))|J(r)| = r(1 - \frac{K}{6}r^2 + O(r^3)), and integrating over θ∈[0,2π]\theta \in [0, 2\pi] gives L(Cr)=2πr(1−K(Π)6r2+O(r3))L(C_r) = 2\pi r(1 - \frac{K(\Pi)}{6}r^2 + O(r^3)).

© 2026 NeckPinch (www.neckpinch.com). All content in the guidebook (text, mathematics, figures and exercises) is protected by copyright. All rights reserved. No part may be copied, republished or redistributed without written permission.