ON CONJUGATE POINTS AND GEODESIC LOOPS IN A COMPLETE RIEMANNIAN MANIFOLD SHICHENGXU Abstract. Awell-knownLemmainRiemanniangeometrybyKlingen- 4 berg says that if x0 is a minimum point of the distance function d(p,·) 1 to p in the cut locus Cp of p, then either there is a minimal geodesic 0 from ptox0 along whichtheyareconjugate, orthereisageodesic loop 2 at p that smoothly goes through x0. In this paper, we prove that: for n anypointqandanylocalminimumpointx0ofFq(·)=d(p,·)+d(q,·)in a Cp, either x0 is conjugate to p along each minimal geodesic connecting J them, or there is a geodesic from p to q passing through x0. In partic- 3 ular, for any local minimum point x0 of d(p,·) in Cp, either p and x0 2 are conjugate along every minimal geodesic from p to x0, or there is a geodesic loop atpthatsmoothly goesthroughx0. Earlier resultsbased ] on injective radius estimate would hold underweaker conditions. G D . h t a 1. Introduction m Let M be a complete Riemannian manifold. For a point p ∈ M, let T M [ p be the tangent space at p and exp : T M → M be the exponential map. p p 2 For any unit vector v ∈ T M, let σ(v) be the supremum of l such that the v p 9 geodesic exptv : [0,l] → M is minimizing, κ(v) be the supremum of s such 4 that there is no conjugate point of p along exp tv : [0,s) → M. Let p 5 5 C˜ = {σ(v)v : for all unit vector v ∈T M} p p . 1 0 be the tangential cut locus of p, 4 1 J˜p = {κ(v)v : for all unit vector v ∈ TpM} : v be the tangential conjugate locus, and C = expC˜ ⊂ M be the cut locus of i p p X p. Let r D˜ = {tv|0 ≤ t < σ(v), for all unit vector v ∈T M} a p p be the maximal open domain of the origin in T M such that the restriction p exp | of exp on D˜ is injective. Any geodesic throughout the paper is p D˜p p p assumed to be parametrized by arclength. A geodesic γ : [0,l] → M is called a geodesic loop at p if γ(0) = γ(l) = p. Let d(·,·) be the Riemannian distance function on M. In [8] Klingenberg proved a lemma that is well- known now in Riemannian geometry and particularly useful in injectivity radius estimate. Date: January 24, 2014. Keywords: geodesic, cut point,conjugate point, injectivity radius. 2000 MSC: Primary 53C22; Secondary 53C20. Project 11171143 supported byNational Natural Science Foundation of China. 1 CONJUGATE POINT AND GEODESIC LOOP 2 Lemma 1.1 (Klingenberg [8, 9]). If x ∈ C satisfies that d(p,x ) = 0 p 0 d(p,C ), then either there is a minimal geodesic from p to x along which p 0 they are conjugate, or there exist exactly two minimal geodesics from p to x that form a geodesic loop at p smoothly passing through x . 0 0 Recently this lemma was generalized to the case of two points ([7]). Let p,q be two points in a complete Riemannian manifold M such that C 6= ∅ p and q 6∈ C . Let F : C → R be a function defined on C by F (x) = p p;q p p p;q d(p,x)+d(x,q). Innami, Shiohama and Soga proved in [7] that Lemma 1.2 ([7]). If J˜ ∩ C˜ = ∅, then for any minimum point x ∈ C p p 0 p of F (x) = d(p,x)+d(x,q), there exist a geodesic (and at most two) α : p;q [0,F (x )] → M from p to q such that α(d(p,x )) = x . p;q 0 0 0 Notethatifp = q,thenLemma1.2isreducedtothecaseofKlingenberg’s Lemma. Inthis paper,weimprove bothresultsin theabove tothefollowing theorems whose constraints are sharp in general. Theorem A (Generalized Klingenberg’s Lemma). Let M be a complete Riemannian manifold and p,q ∈ M such that C 6= ∅ and q 6∈ C . Let x ∈ p p 0 C such that F (x ) = d(p,x )+d(q,x ) is local minimum of F in C . p p;q 0 0 0 p;q p Then either p and x are conjugate along every minimal geodesic connecting 0 them, or there is a geodesic (and at most two) α :[0,F (x )]→ M from p p;q 0 to q such that α(d(p,x )) = x . 0 0 Theorem B (Improved Klingenberg’s Lemma). Let M be a Riemannian manifold and p be a point in M. Let x ∈ C such that d(p,x ) is a local 0 p 0 minimum of d(p,·) in C . If there is a minimal geodesic from p to x along p 0 which p is not conjugate to x , then there are exactly two minimal geodesics 0 from p to x that form a whole geodesic smoothly passing through x . 0 0 Moreover, if d(p,x ) is also a local minimum of d(x ,·) in C , then the 0 0 x0 two minimal geodesics form a closed geodesic. Motivated by Theorem B, we will call a cut point q of p an essential conjugate point of p if p is conjugate to q along every minimal geodesic con- necting them. Geodesic loops may not exist if there are essential conjugate points. For example, a complete noncompact Riemannian manifold of pos- itive sectional curvature always contains a simple point p (i.e. there is no geodesic loop at p) whose cut locus is nonempty [6]. By Theorem B, every local minimum point of d(p,·) in C must be essentially conjugate to p. p The conjugate radius had been involved in the injectivity radius estimate besides the length of geodesic loops (see [8, Lemma 4] or [1, Lemma 1.8]). Recall that the conjugate radius at p is defined by conj(p) = min{κ(v)| for all unit vector v ∈ T M} p and the conjugate radius of M, conj(M) = inf conj(p). The injectivity p M ∈ radius of p is defined by injrad(p) = min{σ(v)| for all unit vector v ∈ T M} p and injectivity radius of M, injrad(M) = inf injrad(p). It follows from p M ∈ Theorem B that the conjugate radius in the injectivity radius estimate can be replaced by the distance from p to its essential conjugate points. Let CONJUGATE POINT AND GEODESIC LOOP 3 Jε = C \exp (C˜ \J˜)) be the the set consisting of all essential conjugate p p p p p points of p. Let d(p,Jε) if Jε 6= ∅, conjε(p) = p p (+∞ otherwise. and conjε(M) = min conjε(p). Then the injectivity radius can be ex- p M ∈ pressed in the following way, where non-essential conjugate points are cov- ered by geodesic loops. Theorem 1.3. conjε(p), injrad(p)= min ; half length of the shortest geodesic loop at p (cid:26) (cid:27) conjε(M), injrad(M) = min . half length of the shortest closed geodesic in M (cid:26) (cid:27) By Theorem 1.3, in this paper we call conjε(p) the essential conjugate radius of p and conjε(M) the essential conjugate radius of M. There have been rich results in Riemannian geometry where an upper sectional curvature bound K shows up to offer a lower bound π of the √K conjugate radius, which now could be weakened to the essential conjugate radius. Forinstance,oneisabletogeneralizeCheeger’sfinitenesstheoremof diffeomorphism classes to Riemannian manifolds whose essential conjugate radius has a lower bound but sectional curvature has no upper bound. Theorem 1.4. For any positive integer n, real numbers D,v,r > 0 and k, there are only finite C -diffeomorphism classes in the set consisting of ∞ n-dimensional Riemannian manifolds whose sectional curvature is bounded below by k, diameter ≤ D, volume ≥ v and essential conjugate radius ≥ r. In general conjε(p) 6= conj(p). RPn with the canonical metric is a trivial example. An immediate question is, when conjε(p) = conj(p)? Next two applications of Theorem B offers examples where conjε(p) coincides with conj(p). Theorem 1.5. Let a point p be the soul of a complete noncompact Rie- mannian manifold of nonnegative sectional curvature in sense of [4]. Then either the nearest point to p in C is an essential conjugate point, or exp : p p T M → M is a diffeomorphism. p By Theorem B, Theorem 1.5 directly follows from the fact that the soul is totally convex ([4]). For a closed Riemannian manifold M, the radius of M is defined by rad(M) = min max{d(p,x)| for any x ∈ M}. p M ∈ Theorem 1.6. Let M be a complete Riemannian manifold whose sectional curvature ≥ 1 and radius rad(M) > π. Then either there are two points 2 p,q ∈ M of distance d(p,q) = d(p,C ) = d(q,C ) < rad(M) and p,q are p q essentially conjugate to each other, or M is isometric to a sphere of constant curvature π2 . rad2(M) Xiaproved([11])thatifthemanifoldM inTheorem1.6satisfiesconj(M) ≥ rad(M) > π, then M is isometric to a sphere of constant curvature. Theo- 2 rem1.6follows hisproofafter replacingKlingenberg’slemmatoTheoremB. CONJUGATE POINT AND GEODESIC LOOP 4 In particular, if rad(p) = max d(p,x) > π, then injrad(p) = conjε(p) = x M 2 ∈ conj(p). Weaker versions of Theorem 1.6 can also be found in [10]. Theorem A provides a new characterization on general Riemannian man- ifolds, which is an improvement of Theorem 1 in [7]. Theorem 1.7. Let M be a complete Riemannian manifold and p be a point in M such that C 6= ∅. Then p (1.7.1) either p has an essential conjugate point; (1.7.2) or there exist at least two geodesics connecting p and every point q ∈ M (regarding the single point p as a geodesic when p = q). We now explain the idea of our proof of Theorem A. Let p,q ∈ M such that C 6= ∅, q 6∈ C and x ∈ C is a minimum point of F in C . Let p p 0 p p;q p us consider the special case that there is a unique minimal geodesic [qx ] 0 connecting q and x . A key observation from [7] is that the level set of 0 {F ≤ C} is star-shaped at both p and q. In particular, if two minimal p;q geodesics [px ] and [x q] from p to x and from x to q are broken at x , 0 0 0 0 0 then for any point x 6= x in [px ] and any minimal geodesic [qx] connecting 0 0 q and x, we have [qx]∩ C = ∅. Thus [qx] admit a unique lifting [qx] in p D˜ . If two minimal geodesics, say [px ] and [px ] , do not form a whole p 0 1 0 2 geodesicwith[x q]atthesametime,thenbymovingxtox along[pxg] and 0 0 0 1 [px ] respectively, one may expect two liftings of [qx ] in D˜ with different 0 2 0 p endpoints when partial limits of [qx] exist. Under the condition that J˜ ∩C˜ = ∅ as in Lemma 1.2 and [7], all such p p minimal geodesics [px] are clearlygdefinite away from the tangential conju- gate locus. So we are able to take limit to meet a contradiction, for the lift ] [qx ] of [qx ] is unique in the closure of D˜ . 0 0 p Inthe general case onedoes notknow whether[qx ] admits a lifting at its 0 endpoint in the closure of D˜ , nor the liftings [qx] have a partial limit as x p approaches x along [px ]. We will prove that, if there is a minimal geodesic 0 0 α from p to x along which they are not conjuggate to each other, and the 0 union of α and [qx ] is broken at x , then [qx ] always has a lifting in the 0 0 0 closure of D˜ , which share a common endpoint with α˜ (see Lemma 2.2). p The uniqueness of the minimal geodesic [qx ] is not an essential problem, 0 becauseby(2.2.4)and(2.2.5)wearealways abletomove q along[qx ]while 0 keeping F minimal at x in C . A local minimum of F can be reduced p;q 0 p p;q to the minimum case similarly. The detailed proof of Theorem A will be given in the next section. 2. Proof of the generalized Klingenberg’s lemma InthissectionwewillproveTheoremA.LetM beacompleteRiemannian manifold and p,q be two points in M such that C 6= ∅ and q 6∈ C . Let us p p consider the function F :C → R, F (x) = d(x,q)+d(x,p) p;q p p;q and assume that F takes its minimum at x ∈ C . Because q is not a p;q 0 p cut point of p, q 6= x . Let γ : [0,d(q,x )] → M be a minimal geodesic 0 0 CONJUGATE POINT AND GEODESIC LOOP 5 from q = γ(0) to x = γ(d(q,x )). Because x is a cut point of p, for any 0 0 0 0 ≤ t < d(q,x ) we have 0 F (γ(t)) = d(q,γ(t))+d(γ(t),p) q < d(q,γ(t))+d(γ(t),x )+d(x ,p) (2.1) 0 0 = d(q,x )+d(x ,p) 0 0 = F (x ) q 0 = minF . q Therefore γ(t) (0 ≤ t < d(q,x )) is not a cut point of p, and we are able to 0 lift γ| to (exp | ) 1◦γ uniquely in the tangential segment domain [0,d(q,x0)) p D˜p − D˜ ⊂ T M, where D˜ is the maximal open domain of the origin in T M p p p p such that the restriction exp | of exp on D˜ is injective. p D˜p p p Wenowprovethat,ifx ∈ exp (C˜(p)\J˜(p)),theneitherγ canbeliftedon 0 p the whole interval [0,d(q,x )] such that the endpoint of the lift is a regular 0 point of exp , or γ can be extended to a geodesic from q to p that goes p through x . 0 Lemma 2.2. Assume that there is a minimal geodesic α :[0,d(p,x )]→ M 0 from p = α(0) to x = α(d(p,x )) along which p is not conjugate to x . Let 0 0 0 w = d(p,x )α(0) ∈ T M, where α(0) is the unit tangent vector of α at p. 0 ′ p ′ Then (2.2.1) either γ and α form a whole geodesic at x ; 0 (2.2.2) or there is a unique smooth lift γ˜ : [0,d(q,x )] → T M of γ : 0 p [0,d(q,x )] → M in the tangential segment domain D˜ ⊂ T M such 0 p p that γ˜(0) = (exp | ) 1γ(0) and γ˜(d(q,x )) = w. p D˜p − 0 Proof. We first prove the case that γ is the unique minimal geodesic from q to x . Assumethat γ and αdonot forma wholegeodesic at x . Let {α(s )} 0 0 i (0 < s < d(p,x )) be a sequence of interior points of α that converges to i 0 x as i → ∞, and let γ : [0,d(q,α(s ))] → M be a minimal geodesic from 0 si i q to α(s ). Then γ converges to γ as i → ∞. i si Because γ and α are broken at x , by the triangle inequality and similar 0 calculation in (2.1), for each s , we have i F (α(s )) < F (x ) = minF , p;q i p;q 0 p;q and thus for any 0≤ t ≤ d(q,α(s )), i F (γ (t)) ≤ F (α(s )) < minF , (2.2.3) p;q si p;q i p;q which implies that none of points in γ is a cut point of p. Therefore there si is a unique lift curve γ˜ of γ starting at (exp | ) 1γ(0) in the tangential si si p D˜p − segment domain D˜ ⊂ T M. p p If {γ˜ } is uniformly Lipschitz, then by the Arzel`a-Ascoli theorem, a sub- si sequence of {γ˜ } converges to a continuous curve γ˜ :[0,d(q,x )]→ T M, si 0 p ∞ which satisfies that γ˜ (0) = (exp | ) 1γ(0), γ˜ (d(q,x )) = w, ∞ p D˜p − ∞ 0 and exp (γ˜ (t)) = lim γ (t) = γ(t), for all 0 ≤ t ≤ d(q,x ). p ∞ i si 0 →∞ CONJUGATE POINT AND GEODESIC LOOP 6 That is, γ has a unique lifting in the segment domain that satisfies (2.2.2). Toprove that{γ˜ } is uniformly Lipschitz, it suffices to show that there is si N > 0 such that the distance between γ˜ ([0,d(q,α(s ))]) and tangen- i N si i tial conjugate locus J˜ ⊂ T M is positive≥. Indeed, because γ˜ (d(q,α(s ))) p p S si i converges to w = d(p,x )α(0), at which the differential d(exp ) is non- 0 ′ p singular, there is a small δ > 0 and some ǫ >0 such that d(γ˜ (t),J˜)> ǫ, for all d(q,α(s ))−δ ≤ t ≤ d(q,α(s )) and large i. si p i i On the other hand, because the restriction γ | converges to si [0,d(q,α(si)) δ] γ| , which lies in the segment domain D = exp−D˜ ⊂ M, there [0,d(q,x0) δ] p p p is some ǫ−> 0 such that 1 d(γ˜ (t),J˜) > ǫ , for all 0 ≤ t ≤ d(q,α(s ))−δ and large i. si p 1 i Nowwhatremainsistoprovethecasethattheminimalgeodesicfromq to x isnotunique. Letusfixsomeinterior pointq = γ(t )(0 < t < d(q,x )) 0 1 1 1 0 of γ and consider the function F :C → R instead. Because p;q1 p F (x) = d(q ,x)+d(x,p) p;q1 1 ≥ d(x,q)−d(q,q )+d(x,p) 1 = F (x)−d(q,q ) (2.2.4) p;q 1 and F (x ) = d(x ,q)−d(q,q )+d(x ,p) p;q1 0 0 1 0 = F (x )−d(q,q ), (2.2.5) p;q 0 1 we see that F (x) also takes minimum at x . Now the minimal geodesic p;q1 0 from q to x is unique. By the same argument as above, either γ| 1 0 [t1,d(q,x0)] from a whole geodesic with α at x , or it has a unique lift in the tangential 0 segment domain. So does γ. (cid:3) A local minimum point of F can be reduced to the case of global mini- p;q mum by the following lemma. Lemma 2.3. Let x be a local minimum point of F : C → R in C , 0 p;q p p and γ : [0,d(q,x )] → M be a minimal geodesic from q = γ(0) to x = 0 0 γ(d(q,x )). Then for any interior point q = γ(t) of γ that is sufficient 0 t close to x , the function F : C → R takes its minimum at x . 0 p;qt p 0 Proof. By (2.2.4) and (2.2.5), x is also a local minimum point of F 0 p;qt for any t ∈ [0,d(q,x )). Therefore, it suffices to show that F takes its 0 p;qt minimum near x as q sufficient close to x . 0 t 0 Let us argue by contradiction. Assuming the contrary, one is able to find a sequence of points {q = γ(t )} that converges to x such that F takes i i 0 p;qi its minimum at some point z ∈ C outside an open ball of x , i p 0 B (x )= {x ∈ M|d(x,x ) < ǫ}. ǫ 0 0 By passing to a subsequence, we assume that z → z ∈ C . Then for any i 0 p y ∈ C , p d(p,z )+d(z ,q )= F (z )≤ F (y) = d(p,y)+d(y,q ). i i i p;qi i p;qi i CONJUGATE POINT AND GEODESIC LOOP 7 Taking limit of the above inequality, we get F (z )≤ F (y), for any y ∈ C . p;x0 0 p;x0 p Let y = x , then 0 d(p,z )+d(z ,x )≤ F (x )= d(p,x ). 0 0 0 p;x0 0 0 Because d(z ,x ) ≥ ǫ, this implies that z is an interior point of a minimal 0 0 0 geodesic from p to x , and it contradicts to the fact that z ∈ C . (cid:3) 0 0 p We now are ready to prove Theorem A. Proof of Theorem A. Let x ∈C be a local minimum point of the function 0 p F :C → R, F (x) = d(x,q)+d(x,p). p;q p p;q Let α : [0,d(p,x )] → M be a minimal geodesic from p = α(0) to x = 0 0 α(d(p,x )),alongwhichpisnotconjugatetox . Letγ : [0,d(q,x )] → M be 0 0 0 aminimalgeodesicfromq = γ(0)tox = γ(d(q,x )), andw = d(p,x )α(0). 0 0 0 ′ First let us prove that there is a minimal geodesic from p to x that 0 forms a whole geodesic with γ. According to Lemma 2.3, by moving q to an interior point in γ that is sufficient close to x and denoted also by q, it can 0 be reduced to the case that x is a minimal point of F and γ is a unique 0 p;q minimal geodesic connecting q and x . 0 By Lemma 2.2, if α and γ are broken at x , then γ has a unique lift 0 γ˜ : [0,d(q,x )] → T M in the tangential segment domain D˜ ⊂ T M, whose 0 p p p endpoint satisfies γ˜(d(q,x )) = w. 0 Because x is not conjugate to p along α, there is another minimal geodesic 0 β : [0,d(p,x )] → M from p to x . We now prove that β must form a whole 0 0 geodesic with γ. Assuming the contrary, that is, β does not form a whole geodesic with γ neither. Foranyinteriorpointβ(s)(0 < s < d(p,x ))inβ,letl = d(q,β(s)) 0 s and γ : [0,l ] → M be a minimal geodesic from q to β(s). Then by the s s same argument as (2.2.3), for any 0 ≤ t ≤ l we have s F (γ (t)) ≤ F (β(s)) < F (x ) = minF , p;q s p;q p;q 0 p;q which implies that γ has a unique lift γ˜ in the tangential segment domain s s D˜ ⊂ T M. p p We point it out that, because the endpoint γ˜(l ) of γ˜ may approach s s J˜ ⊂ T M,onecannotdirectlyconcludethatthefamilyofcurvesγ˜ contains p p s any convergent subsequenceas s → d(p,x ). Instead, let us consider a small 0 metric ball B (w) at w = d(p,x )α(0) in T M on which the restriction of ǫ 0 ′ p exp p exp :B (w) → M p Bǫ(w) ǫ is an diffeomorphism onto its (cid:12)image. Then d(p,x0)β′(0) 6∈ Bǫ(w). Because (cid:12) γ converges to γ as s → d(p,x ) and the restriction exp of exp on s 0 p D˜p p the tangential segment domain D˜p is a diffeomorphism, γ˜s|[0,l(cid:12)(cid:12)s) converges to γ˜ | pointwisely, that is, for 0 < t < 1, s [0,d(q,x0)) γ˜ (l ·t)→ γ˜(d(q,x )·t) as s→ d(p,x ). s s 0 0 CONJUGATE POINT AND GEODESIC LOOP 8 Then for any 0 < t < 1 that is sufficient close to 1, there exists 0< s(t ) < 1 1 d(p,x ) such that for all s(t ) < s < d(p,x ), 0 1 0 γ˜s(lst1) ∈ Bǫ(w), γ˜s(ls)∈ Bǫ(d(p,x0)β′(0)), 4 2 and γs|[lst1,ls] lies in the open neighborhood exppB4ǫ(w) of x0. Because the lift of γs|[lst1,ls] is unique in D˜p and γ˜s(lst1) ∈ B4ǫ(w), we conclude that γ˜s|[lst1,ls] lies in B4ǫ(w), which contradicts to the fact that Bǫ(w)∩Bǫ(d(p,x0)β′(0)) = ∅. 4 2 Secondly, we prove that there are at most two geodesics from q to p passing through x . Indeed, if there are two distinct minimal geodesics γ , 0 1 γ connecting q and x , then one of them, say γ , does not form a whole 2 0 1 geodesic with α. By the proof above, any other minimal geodesic except α connecting p and x from a whole geodesic with γ . By Lemma 2.2, γ form 0 1 2 awholegeodesic withα. Therefore, thereareexactly two minimal geodesics α and β from p to x , the union of β and γ form a whole geodesic going 0 1 through x , and α and γ form another whole geodesic. In particular, there 0 2 won’t be a third geodesic from q to x . 0 (cid:3) TheoremBandTheorem 1.7 areimmediate corollaries of TheoremA.We give a proof of Theorem 1.4 to end the paper. Proof of Theorem 1.4. Let M(n,k,D,v,r) be the set consisting of all com- plete n-dimensional Riemannian manifolds whose sectional curvature ≥ k, diameter ≤ D, volume ≥ v and essential conjugate radius ≥ r. Let M ∈ M(n,k,D,v,r). By Cheeger’s lemma ([2], see Theorem 5.8 in [3]), there is an universal constant c (D,v,k) > 0 depending only on n,D,v,k such n that every smooth closed geodesic on M has length > c (D,v,k). There- n fore, by Theorem 1.3 the injectivity radius of M is bounded below by min{r, 1c (D,v,k)}. The finiteness of diffeomorphism classes in the set 2 n M(n,k,D,v,r)follows fromthestandardargument(see[5,12]forexample) on the construction of a diffeomorphism between two Riemannian manifolds with small Gromov-Hausdorff distance . (cid:3) References [1] U.AbreschandW.T.Meyer.Injectivityradiusestimatesandspheretheorems.Com- parison geometry, 30:1–47, 1997. [2] J.Cheeger.Finitenesstheoremsforriemannianmanifolds.Am.J.Math.,92(1):61–74, 1970. [3] J. Cheeger and D.G. Ebin. Comparison theorems in Riemannian geometry, volume 365. AMS Bookstore, 1975. [4] J. Cheeger and D. Gromoll. On the structure of complete manifolds of nonnegative curvature.Ann. of Math., 96(3):413–443, 1972. [5] K. Fukaya. Collapsing riemannian manifolds to ones of lower dimensions. J. Differ- ential Geom., 25(1):139–156, 1987. [6] D. Gromoll and W. Meyer. On complete open manifolds of positive curvature. Ann. of Math., 90(1):75–90, 1969. [7] N. Innami, K. Shiohama, and T. Soga. The cut loci, conjugate loci and poles in a complete riemannian manifold. Geom. Funct. Anal., 22(5):1400–1406, 2012. [8] W. Klingenberg. Contributions to riemannian geometry in the large. Ann. of Math., 69(3):654–666, 1959. CONJUGATE POINT AND GEODESIC LOOP 9 [9] W. Klingenberg. Riemannian geometry. Walter de Gruyter,1995. [10] Q. Wang. On the geometry of positively curved manifolds with large radius. Illinois J. Math., 48(1):89–96, 01 2004. [11] Changyu Xia. A round sphere theorem for positive sectional curvature. Compositio Math., 142(5):1327–1331, 2006. [12] T. Yamaguchi. Collapsing and pinching under a lower curvature bound. Ann. of Math., 133(2):317–357, 1991. School of Mathematics, Capital Normal University, Beijing, China E-mail address: [email protected]