Sharp eigenvalue estimates on degenerating surfaces Nadine Große and Melanie Rupflin Abstract Weconsiderthefirstnon-zeroeigenvalueλ1 oftheLaplacianonhyperbolicsurfacesfor 7 which one disconnecting collar degenerates and prove that 8π∇log(λ1) essentially agrees 1 withthedualofthedifferentialofthedegeneratingFenchel-Nielsonlengthcoordinate. Asa 0 corollaryofouranalysis,whichisbasedinparticularonthefinepropertiesofholomorphic 2 quadratic differentials from the joint works [15, 16] of Topping and the second author, n we can improve previous results of Schoen, Wolpert, Yau [20] and Burger [4] to obtain a estimates with optimal error rates. J 0 3 1 Introduction and results ] G D Let M be a closed oriented surface of genus γ 2 (always assumed to be connected) and let h. g be a hyperbolic (i.e. Gauss curvature K ≥1) metric on M. Let σ1 be a simple closed ≡ − t geodesic in (M,g) which decomposes M into two connected components M+ and M−. We a m considersurfacesforwhichthe lengthℓ1 =Lg(σ1)issmallcomparedto thelengthofanyother simple closed geodesic in (M,g). In this case the first eigenvalue of the Laplacian on (M,g) [ turns out to be small and to essentially only depend on ℓ and the genera of M±. 1 1 v The asymptotic behaviour of small eigenvalues of the Laplacian on degenerating surfaces was 1 first considered by Schoen, Wolpert and Yau in [20]. They studied surfaces with bounded 9 negative curvature c K c˜<0 and proved in particular that if the collapsing geodesics 4 − ≤ g ≤− 8 decompose M into n+1 connectedcomponents then precisely n eigenvalues0<λ1 ... λn ≤ ≤ 0 tend to zero,with the rate ofconvergencebeing linearwith respectto the (sumof) the lengths . of the corresponding geodesics. Their results apply in particular to the setting we described 1 0 above and in this case yield that 7 cℓ λ Cℓ (1.1) 1 1 1 1 ≤ ≤ : while v λ c˜>0, (1.2) i 2 ≥ X for constantsc,c˜>0 andC < that depend, apartfromthe genus,only onalowerbound on ar the lengths of the simple closed∞geodesics different from σ1, or equivalently on a lower bound δˆ> 0 for the injectivity radius on M C(σ1). Here and in the following C(σ1) denotes the so \ called collar neighbourhood around σ1 whose precise definition we recall in Lemma A.1 in the appendix. Remark 1.1. We note that (1.1) and (1.2) imply in particular that λ is simple provided 1 ℓ =L (σ1) ℓ for a suitably small constant ℓ =ℓ (δˆ,γ) ℓ (γ). 1 g 0 0 0 0 ≤ ≤ A refined picture of the behaviour of small eigenvalues on degenerating hyperbolic surfaces was then given by Burger in [3] and [4], who compared the small eigenvalues of ∆ on M g − with the eigenvalues λ of the Laplacian of a weighted graph that is associated to the set of j b 1 collapsing geodesics. In [3] he established that λj 1 , 1 j n, as the surface collapses λbj → 2π2 ≤ ≤ and subsequently refined this convergence result in [4] by giving both a lower bound (of order O(√ℓ)) and an upper bound (of order O(ℓlogℓ)) on the resulting errors. We note that in the setting we consider here his result from [4] yields that λ 1 C C ℓ C +Cℓ log(ℓ ) (1.3) top 1 top 1 1 − ≤ ℓ ≤ | | 1 p where C is given in terms of the genera γ± of the connected components M± of M σ1 top \ χ(M) 2(γ 1) C = − = − . (1.4) top 2π2χ(M−)χ(M+) 2π2(1 2γ+) (1 2γ−) − · − Weremarkthattheupperboundin(1.3)canbeobtaineddirectlyfromcomparingwithafunc- tion that is linear onthe collar C(σ1)(or alternativelya function that solvesthe corresponding ODE on the collar) and constant on the rest of the surface while the proof of the lower bound is far more involved and does not yield the same order of the error. We note that (1.3) implies in particular that if g and g˜ are two metrics which satisfy the assumptions above for geodesics σ1 and σ˜1 of the same length ℓ and connected components 1 M± and M˜± of the same genera γ± then ℓ−1 λ (M,g) λ (M,g˜) Cℓ1/2. (1.5) 1 | 1 − 1 |≤ 1 It is naturalto ask whether the lower bound in (1.3) and hence also the above estimate can be improvedtoO(ℓ log(ℓ ))and,moreimportantly,whethersuchanestimatewouldbeoptimal, 1 1 | | respectively whether one can derive an estimate of the form (1.5) with optimal error rates. In the present work we will give positive answers to both of these questions and indeed derive bothC0-andC1-estimateswithsharperrorbounds. Mostofouranalysisisquitedifferentfrom the methods in [4] as we use a dynamic approachand consider the variation of the eigenvalues inducedbyachangeofthegeometryof(M,g),ortobemoreprecisebyachangeoftheFenchel- Nielson coordinates. We then obtain C0-bounds, such as refinements of (1.5) and (1.3), only as corollary of our C1-bounds. Weremarkthatboundsonsomederivativesofsmalleigenvalueshavebeenobtainedpreviously by Batchelor[2]whoconsideredthe changeofthe smalleigenvaluesinduced by achangeofthe length of the collapsing geodesics,so our case ∂λ1, though his errorestimates are only of order ∂ℓ1 O( 1 ) and would thus in particular not allow for any improvement of (1.3). |log(ℓ)| To state our first main result, we recall that we may extend any given simple closed geodesic σ1 inaclosedorientedhyperbolicsurface(M,g)to acollectionE= σ1,...,σ3(γ−1) ofsimple { } closed geodesics in (M,g) that decompose the surface into pairs of pants. We also recall that we can and will choose this collection so that the length of all geodesics σj is bounded from abovebyaconstantL¯ thatdependsonlyonthegenusandanupperboundonL (σ1),compare g also Lemma A.4, so in the situation of Remark 1.1, by some L¯ =L¯(γ). Thegeometryof(M,g)isdeterminedbythecorrespondingFenchel-Nielsoncoordinates,i.e.the set of length parameters ℓ = L (σi) together with a set of twist parameters ψ that describe i g i the way in which the pairs of pants are glued together. We refer the reader to Appendix A.2 for a summary of the results on hyperbolic surfaces that we use in the present paper and in particular to Remark A.3 for a precise definition of the twist coordinates. The idea that λ essentially only depends on the length of σ1 if ℓ is small compared to the 1 1 lengths of the other simple closed geodesics of (M,g) can be quantified as follows: 2 Theorem 1.2. Let (M,g) be a closed oriented hyperbolic surface of genus γ 2 and let σ1 be ≥ a simple closed geodesic which disconnects M into two connected components. We let δˆ>0 be a lower bound on the injectivity radius infM\C(σ1)injg(p) away from the collar around σ1 and suppose that ℓ ℓ , for ℓ =ℓ (δˆ,γ)>0 as in Remark 1.1. 1 0 0 0 ≤ Let σ2,...,σ3(γ−1) be simple closed geodesics so that E= σ1,...,σ3(γ−1) decomposes (M,g) into pairs of pants, which we can furthermore assume to b{e chosen so that}L (σj) L¯ =L¯(γ) g ≤ for every j. Then the first non-zero eigenvalue λ=λ of ∆ has the following dependence on 1 g − the corresponding Fenchel-Nielson length and twist coordinates ℓ and ψ : i i There exists a constant C depending only on δˆand the genus of M so that ∂λ λ Cℓ log(ℓ ) 1 1 ∂ℓ − ℓ ≤ | | (cid:12) 1 1(cid:12) (cid:12) (cid:12) and (cid:12) (cid:12) (cid:12) (cid:12) ∂λ Cℓ2 for j =1 ∂ℓ ≤ 1 6 (cid:12) j(cid:12) (cid:12) (cid:12) while a change of the twist coordin(cid:12)ates(cid:12)can only change the first eigenvalue by (cid:12) (cid:12) ∂λ ∂λ Cℓ4, respectively Cℓ2 for j =2,...,3(γ 1). (1.6) ∂ψ ≤ 1 ∂ψ ≤ 1 − (cid:12) 1(cid:12) (cid:12) j(cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) Hereandinthe followingC(σ1)denotesthe collarneighbourhoodofσ1 describedbythe Collar lemma A.1 which we recall in the appendix and inj (p) stands for the injectivity radius in g p (M,g). We also note that the facts that λ is simple and invariant under pull-back by ∈ diffeomorphisms guarantee that the above derivatives are well defined, compare also Lemma 2.1. As a consequence of the C1-bounds on λ stated in the above result we immediately obtain the following refinement of the result of Burger [4] which not only gives the same type of lower bound on the dependence of λ on ℓ but furthermore establishes a quantitative version of the 1 idea that λ essentially only depends onℓ whichis sharp as we shallsee in Theorem1.4 below. 1 Corollary 1.3. Let M be a closed oriented surface of genus γ 2, let σ¯ be a simple closed ≥ curve that disconnects M into two connected components of genera γ± and let C be given top by (1.4). Then there exists a function f: (0,2arsinh(1)) R+ that depends only on γ± and → satisfies f(ℓ) C Cℓ log(ℓ) top ℓ − ≤ | | (cid:12) (cid:12) (cid:12) (cid:12) and for any δˆ>0 there exists a co(cid:12)nstant C =(cid:12)C(δˆ,γ) such that the following holds true. (cid:12) (cid:12) Let g be any hyperbolic metric on M for which inj (x) δˆon M C(σ1), σ1 theunique geodesic g ≥ \ in (M,g) that is homotopic to σ¯, and for which ℓ := L (σ1) < 2arsinh(1). Then the first g eigenvalue λ(M,g)=λ (M,g) of ∆ satisfies 1 g − λ(M,g) f(ℓ) Cℓ2. (1.7) | − |≤ In particular, for metrics g,g˜ for which the lengths of the corresponding geodesics σ1 and σ˜1 agree, we have that λ(M,g) λ(M,g˜) Cℓ2. (1.8) | − |≤ For surfaces of genus at least 3 this result is sharp as we shall prove 3 Theorem 1.4. For every genus γ 3 and every number δˆ>0 there exist constants c¯>0 and ≥ ℓ > 0 so that the following holds true. Let M be a closed oriented surface of genus γ and let 0 σ¯ be a simple closed curve that disconnects M into two connected components of genera γ±. Then there exist families of hyperbolic metrics (g ) and (g˜ ) satisfying the as- ℓ ℓ∈(0,ℓ0) ℓ ℓ∈(0,ℓ0) sumptions of Corollary 1.3, for the fixed constant δˆ> 0 and with L (σ1 )= ℓ = L (σ1 ), for g˜ℓ g˜ℓ gℓ gℓ which λ(M,g ) λ(M,g˜ ) c¯ ℓ2. ℓ ℓ | − |≥ · We willobtainthe aboveresults basedonanessentiallyexplicitcharacterisationof λ thatwe ∇ state in Theorem1.5 below andona carefulanalysisof the tensorsthat induce a change ofthe Fenchel-Nielson length respectively twist coordinates. In order to state these results in detail we need to introduce some more notation and recall some well-known properties of hyperbolic surfaces as well as results on holomorphic quadratic differentials obtainedinthe jointworks[16] ofTopping andthe secondauthor respectively[17] of Topping, Zhu and the second author. Given a closed orientable surface M of genus γ 2 we denote by M the set of smooth −1 hyperbolic metrics on M and recall that the tange≥nt space to T M splits orthogonally g −1 T M = L g,X Γ(TM) H(g) g −1 X { ∈ }⊕ intothedirectionsgeneratedbythepull-backbydiffeomorphismsandthehorizontalspace H(g) which can be characterised as space of symmetric (0,2) tensors that are both divergence- and trace-free. An equivalent characterisation of the horizontal space which will be more suitable for our analysis is to view H(g)=Re(H(M,g)) as the real part of the complex vector space H(M,g):= Ψ: holomorphic quadratic differentials on (M,g) { } whose (complex) dimension is 3(γ 1). − GivenacollectionE= σi 3(γ−1) ofsimpleclosedgeodesicswhichdecomposes(M,g)intopairs of pants we consider C-{line}ai=r1maps ∂ℓ : H(M,g) C j → whichcanbe seenasC-linear derivativesofthe lengthcoordinatesℓ =L (σj)andwhichwere j g introduced in [16, Remark 4.1] as follows: We view H(M,g) as real vector space with complex structure J and identify H(M,g) with a subspace of T M via the isomorphism Φ Re(Φ). g −1 7→ We then define ∂ℓ (Φ):= 1 dℓ (Φ) idℓ (JΦ) (1.9) j 2 j − j where dℓ : T M R is given by dℓ (k)(cid:0)= d L (σj(ε)(cid:1)) where g is a smooth curve of metrics inj Mg −so1t→hat ∂ g =k anjd σj(ε)dεis|εt=h0e ugεnique simple closεed geodesic in (M,g ) −1 ε ε=0 ε ε | homotopic to σj. Inthejointwork[17]ofTopping,Zhuandthesecondauthoritwasshownthatthereexistssome η (0,arsinh(1)) (depending at most on the genus of M) so that if σ1,...,σk are the simple 1 ∈ closed geodesics of length no more than 2η then the codimension of ker(∂ℓ ,...∂ℓ ) is k. We 1 1 k denote by Θ˜1,...,Θ˜k the basis of (ker(∂ℓ ,...,∂ℓ ))⊥ which is dual to ∂ℓ k , compare { } 1 k { j}j=1 (2.22), and set Ω˜j = Θ˜j . The fine properties of this renormalised dual basis Ω˜1,...,Ω˜k −kΘ˜jkL2 of (ker(∂ℓ ,...,∂ℓ ))⊥ were analysed in the joint work [16] of Topping and the second author. 1 k The resultsof [16], in particular[16, Lemma 4.5]whichwe recallin Section2.2, guaranteethat upto error terms of order O(ℓ3/2) each of these elements is concentrated on the corresponding j 4 collar C(σj) and there given essentially as a constant multiple of dz2. As explained in [16] the error rate of ℓ3/2 is optimal. j While the theoryin [16, 17]wasinitially developedasa toolforthe analysisofthe Teichmu¨ller harmonic map flow, the obtained estimates will play a crucial role in the present paper: Our lastmain resultaboutthe properties ofthe firsteigenvalue, whichatthe same time will be the basis of the proofs of both Theorems 1.2 and 1.4, establishes that the gradient of λ = λ is 1 essentially givenin terms of the element Θ˜1 ker(∂ℓ )⊥ which is dual to the differential ∂ℓ of 1 1 ∈ the degenerating length coordinate. Theorem 1.5. Let(M,g)beaclosed orientedhyperbolic surface, let σ1 beadisconnectingsim- ple closed geodesic and let ∂ℓ be the complex differential of the corresponding length coordinate 1 introduced in (1.9). Then,codim(ker(∂ℓ ))=1andifL (σ1)issmallcomparedtotheinjectivityradiusonM C(σ1) 1 g then the L2-gradient of the first eigenvalue λ : M R of ∆ is essentially determi\ned by 1 −1 g → − the element Θ˜1 of ker(∂ℓ )⊥ for which ∂ℓ (Θ˜1)=1, namely 1 1 log(λ) 1 Re(Θ˜1) ∇ ∼ 8π holds true in the following sense: Suppose that ℓ =L (σ1) ℓ for ℓ =ℓ (δˆ,γ) chosen as in Remark 1.1 and δˆas usual a lower 1 g 0 0 0 bound on inj on M C(σ≤1). Then there exists a number α>0 with g \ α 1 Cℓ log(ℓ ) (1.10) | − 8π|≤ 1| 1 | such that log(λ) αRe(Θ˜1) Cℓ , (1.11) L∞(M,g) 1 k∇ − k ≤ where C depends only on the genus of M and on δˆ. Wewillobtaintheproofofthistheorembycombiningtwodifferenttypesofresults: Ontheone hand we derive energy estimates for the first eigenfunction, and this part is similar to existing approaches, including [4]. These results will be stated in Section 2.1 and proved later on in Section 3. On the other hand we crucially use the fine properties of holomorphic quadratic differentials on degenerating surfaces as developed in [16] and [17], as well as the uniform Poincar´e estimate for holomorphic quadratic differentials that was proven in [15] by Topping andthe secondauthor. We recallthe relevantresults from[15, 16, 17]in Section2.2. Basedon these results we will then be able to give the proof of the above Theorem 1.5 in Section 2.3. While Theorem 1.5 forms the basis of the proofs of our other main results on the first eigen- value, we will need further results on the fine properties of holomorphic quadratic differentials in order to prove Theorems 1.2 and 1.4, and as a consequence also Corollary 1.3. In par- ticular, we need to study bases of H(M,g) that are associated to a full set of simple closed geodesics σj whichdecomposesM intopairsofpantsrespectivelythecorrespond- j=1,...,3(γ−1) { } ing Fenchel-Nielson coordinates. Tobemoreprecise,weshallneedandprovethatthereisadualbasis Θj 3(γ−1) to ∂ℓ 3(γ−1) of the whole space H(M,g) for any such decomposing set of geode{sics}(jw=1ithout {requj}irji=n1g a smallness assumption on the ℓ ) and that these elements still satisfy estimates with optimal j error rates as obtained in [16] for the elements Θ˜j described above. We will furthermore need to consider elements Ψj,Λj of H(M,g) that are dual to the (real differentials of the) Fenchel- Nielson coordinates ψ ,ℓ in the sense that for every i,j j j dℓ (Re(Λi))=δi =dψ (Re(Ψi)) and dψ (Re(Λi))=0=dℓ (Re(Ψi)). (1.12) j j j j j 5 As we shall see, also these bases can be well controlled and, for small ℓ , can be characterised j essentially explicitly in terms of the Θj, respectively the Ω˜j from [16]. These results, which will be stated in detail in Section 2.4 and proven in Section 4, can be summarised by the following proposition which may be of independent interest. Proposition 1.6. Let (M,g) be any closed oriented hyperbolic surface of genus γ and let E = σ1,...,σ3(γ−1) be any set of simple closed geodesics which decomposes M into pairs of pants{. Let ℓ ,ψ be}the corresponding Fenchel-Nielson coordinates and let ∂ℓ be the C-linear i i i { } differentials introduced in (1.9). Then (I) There exists a dual basis Θj 3(γ−1) to (∂ℓ ,...,∂ℓ ). These Θj and the correspond- { }j=1 1 3(γ−1) ing renormalised elements Ωj = Θj satisfy the same sharp error estimates as −kΘjkL2(M,g) obtained in [16] (there for Ω˜j), see Lemma 2.13 and Corollary 2.14 for details. (II) The element Ψj which generates a Dehn-twist around σj, compare (1.12), satisfies Ψj (ker(∂ℓ ))⊥ and ∂ℓ (Ψj) iR 0 j j ∈ ∈ \{ } and, for indices for which ℓ is small, is such that j Ψj iΩj Ψj ∼− L2 k k as is made precise in Lemma 2.19. (III) The duals Λj of the (real) differentials of the Fenchel-Nielson lengths coordinates intro- duced in (1.12) are so that Λj 1Θj ∼ 2 upto error terms whose L∞-norm is of order O(ℓ ), compare (2.18), which agrees with j the sharp error rates on the Θj from part (I). We remark that bases of the space of holomorphic quadratic differentials which are related to Fenchel-Nielsonandother choicesofcoordinatesonTeichmu¨llerspacehavebeen consideredby many authors and we refer in particular to the works of Masur [12], Yamada [25, 26], Wolpert [22, 23, 24] and the references therein for an overview of existing results. While the bases previouslyconsideredwereoften characterisedin terms of the gradientofthe coordinates,here we follow the approach of [16] and consider bases that are obtained as dual bases as this is needed to consider the dependence of a function on the coordinates as done in Theorem 1.2. Wealsostressthattheestimateswederivehavesharperrorrates,correspondingtoL∞-bounds oforderO(ℓ3/2)forelementswithunitL2-norm,asobtainedin[16]forthe elementsΩ˜j,rather j than bounds of order O(ℓ1/2) as obtained in previous works. This feature is essentialto obtain j sharp bounds on the dependence of the first eigenvalue on the Fenchel-Nielson coordinates. Remark 1.7. The results of Schoen, Wolpert and Yau [20], Burger [4] and Batchelor [2] apply to more general settings of several degenerating collars and so does our our analysis of holomorphic quadratic differentials and Fenchel-Nielson coordinates carried out in Section 4. Therefinedanalysisofsmalleigenvaluesinthismoregeneralsettingwillbeaddressedinfuture work. It would furthermore be of interest to know whether (1.8) is sharp also for surfaces of genus 2 and whether the error rate of ℓ logℓ for the dependence (1.7) of the first eigenvalue 1 1 | | on the degenerating length coordinate ℓ is optimal. 1 We note that the asymptotic behaviour of small eigenvalues has been considered also by Gro- towski, Huntley and Jorgenson in [9]. We also remark that related questions about small 6 eigenvalues on hyperbolic surfaces have attracted a lot of attention over the past decades and remark in particular that Colbois and Colin de Verdi`ere used the study of eigenvalues on weighted graphs to obtain multiplicity results for eigenvalues on hyperbolic surfaces [8] and that the question of how many eigenvalues of ∆ on a hyperbolic surface of genus γ can be g − smaller than 1 has been addressed in particular by [6], [19] and [13]. 4 2 Proofs of the main results Inthe firsttwopartsofthis sectionwe collectpropertiesofthe firsteigenfunction, provedlater on in Section 3, and results from [15, 16, 17] on the fine properties of holomorphic quadratic differentials, both of which are needed to give the proof of our first main result Theorem 1.5 in the subsequent Section 2.3. In Section 2.4 we then state in detail the properties of the dual bases Θj, Λj and Ψj of the space of holomorphic quadratic differentials that we outlined in Proposition 1.6 and that we prove in Section 4. These results then allow us to give the proof of our other main results in the subsequent sections: we prove Theorem 1.2 in Section 2.5, Corollary 1.3 in Section 2.6 and finally Theorem 1.4 in Section 2.7. 2.1 Properties of the first eigenfunction We first recall that the gradient of an eigenvalue, considered as a function on the set M of −1 all smooth hyperbolic metrics, is determined by the holomorphic part of the Hopf-differential of the corresponding eigenfunction, namely Lemma2.1. Let(M,g)beahyperbolicsurfaceforwhichthek-theigenvalueλ issimple, kany k element of N. Let u be the corresponding eigenfunction, normalised to have u =1. k k L2(M,g) Then the L2-gradient of λ : M R is given by k k k −1 → λ (g)= 1Re(PH(Φ(u ,g)) ∇ k −2 g k where Φ(u,g) denotes the Hopf-differential of u, given in local isothermal coordinates (x,y) of (M,g) as Φ(u,g)=(u 2 u 2 2iu u )dz2, z =x+iy, (2.1) x y x y | | −| | − and PH is the L2(M,g)-orthogonal projection from the space of L2-quadratic differentials onto g the space of holomorphic quadratic differentials H(M,g). A proof of this lemma is provided later on in Section 3.1 while more details about the space of holomorphic quadratic differentials will be provided in Section 2.2. We recallthatthe (realpartofthe)Hopf-differentialdescribesthe L2-gradientoftheDirichlet- energy with respect to variations of the metric and remark that for real-valued functions as considered here Φ(v,g) =2 dv 2 for every K M, (2.2) k kL1(K,g) k kL2(K,g) ⊂ see Remark 3.2 for more details. As we shall from now on always assume that eigenfunctions are normalised to have u =1, we thus have that L2(M,g) k k Φ(u,g) =2λ. (2.3) L1(M,g) k k WerecallthatRemark1.1ensuresthatthefirsteigenvalueissimpleinthesituationsconsidered in our main results, allowing us to apply the above lemma for λ = λ and the corresponding 1 (normalised) eigenfunction u=u . 1 7 We also recall that u is characterisedas minimiser of the Rayleigh-quotient dv 2 λ =min k kL2(M,g) :v H1(M,g) so that vdv =0 1 ( kvk2L2(M,g) ∈ ˆM g ) and will use that u satisfies the following energy estimates which are proven in Section 3.2. Lemma 2.2. There exist constants C and δ (0,arsinh(1)) depending only on the genus 0 2 ∈ γ 2 so that the following holds true for any closed oriented hyperbolic surface (M,g) of genus γ ≥and any number δ¯ (0,δ ]. 2 ∈ Suppose that inj(M,g) δ¯ and that all simple closed geodesics σ1,...,σk of length no more then 2δ¯ are so that M ≤σj is disconnected. Then the first eigenfunction u of ∆ (as always g \ − normalised by u =1) satisfies the estimate L2(M,g) k k C du 2 0λ2 for every 0<δ δ¯. k kL2(δ-thick(M,g)) ≤ δ 1 ≤ Here and in the following we denote by δ-thick(M,g) := p M : inj (p) δ while { ∈ g ≥ } δ-thin(M,g):=M δ-thick(M,g). \ Weshallfurthermoreseethatoncollarsaroundshortsimpleclosedgeodesicstheangular energy decaysrapidlytowardsthe centreofthe collar,allowingustoobtaininparticularthe following bounds on weighted angular energies. Lemma2.3. ThereexistuniversalconstantsC andδ >0sothatthefollowingholdstruefor 1,2 3 any closed oriented hyperbolic surface (M,g) and any eigenfunction u of ∆ to an eigenvalue g λ R. − ∈ Let σ be a simple closed geodesic of length ℓ < 2arsinh(1) and let (s,θ) ( X(ℓ),X(ℓ)) S1 ∈ − × be the corresponding collar coordinates in which the metric takes the form g = ρ2(ds2 +dθ2), compare Lemma A.1. Then X(ℓ) u 2ρ−4dsdθ C du 2 +C λ2X(ℓ) ˆ ˆ | θ| ≤ 1k kL2(δ3-thick(C(σ)) 1 −X(ℓ) S1 and X(ℓ) u 2ρ−2dsdθ C du 2 +C λ2 u 2 . (2.4) ˆ ˆ | θ| ≤ 2k kL2(δ3-thick(C(σ)) 2 k kL∞(M,g) −X(ℓ) S1 In particular, if (M,g) satisfies the assumptions of Lemma 2.2 for some δ¯ (0,arsinh(1)) and ∈ geodesics σj k and if u is the first eigenfunction of ∆ then the angular energies on the { }j=1 − g corresponding collars C(σj) are bounded by X(ℓj) u 2ρ−4dsdθ C(δ¯,γ) λ2+C λ2 ℓ−1 (2.5) ˆ ˆ | θ| ≤ · 1 1 1· j −X(ℓj) S1 and X(ℓj) u 2ρ−2dsdθ C(δ¯,γ) λ2. (2.6) ˆ ˆ | θ| ≤ · 1 −X(ℓj) S1 We furthermore recall, compare Appendix A.4 Remark 2.4. There exists a constant C depending at most on the genus of M so that the 3 following holds true: Let (M,g) be a closed hyperbolic surface whose shortest simple closed geodesic σ is such that M σ is disconnected. Then the (normalised) first eigenfunction u of \ ∆ is bounded by u C . g L∞(M,g) 3 − k k ≤ 8 2.2 Results on holomorphic quadratic differentials from [15, 16, 17] Here we recall some general properties of the space of holomorphic quadratic differentials H(M,g) as well as results on H(M,g) from the joint works [15, 16] and [17] of Topping (re- spectively Topping, Zhu) and the second author that will be used in the proof of Theorem 1.5. We recall that a quadratic differential is a complex tensor Ψ which is given in local isothermal coordinates (x,y) as Ψ=ψ dz2, z =x+iy. Here ψ is a complex function which, for elements of H(M,g), is furthermore a·sked to be holomorphic. Usingthe normalisationthat dz2 =2ρ−2 forg =ρ2(dx2+dy2)wemaywritethe (hermitian) g | | L2-inner product on the space of quadratic differentials locally as Ψ,Φ = ψ φ¯dz2 2dv =4 ψ φ¯ρ−2dxdy. h iL2 ˆ · | |g g ˆ · In particular Re(Ψ),Re(Φ) = 1Re Ψ,Φ , (2.7) h iL2(M,g) 2 h iL2(M,g) andwerecallthatthisrelationimpliesthattheprojectionPH fromthespaceofsymmetricreal g (0,2)-tensorsontoH(g)=Re(H(M,g)) andthe projectionPH fromthespaceofL2-quadratic g differentials onto H(M,g) are related by PH(Re(Φ))=Re(PH(Φ)). g g We furthermore recall from [16, Proposition 4.10] that for any quadratic differential Υ H P (Υ) C Υ (2.8) k g kL1(M,g) ≤ k kL1(M,g) for a constant C that depends only on the genus. Let now C(σ) be a collar around a simple closed geodesic σ in (M,g) described by the Collar lemma A.1 of Keen-Randol that we recall in the appendix. We we will often use that on C(σ) we may represent any Υ H(M,g) by its Fourier series in collar coordinates (s,θ) ∈ ∞ Υ= b (Υ)en(s+iθ)dz2, b (Υ)=b (Υ,C(σ)) C n n n ∈ n=−∞ X and that on C(σ) we may split Υ orthogonally into its principal part b (Υ)dz2 and its collar 0 decay part Υ b (Υ)dz2. Hence, for any Υ,Ψ H(M,g) 0 − ∈ hΥ,ΨiL2(C(σ)) =b0(Υ)·b0(Ψ)kdz2k2L2(C(σ))+hΥ−b0(Υ)dz2,Ψ−b0(Ψ)dz2iL2(C(σ)) (2.9) =b0(Υ)·b0(Ψ)kdz2k2L2(C(σ))+hΥ,Ψ−b0(Ψ)dz2iL2(C(σ)), where here and in the following we sometimes abbreviate b (Ψ) = b (Ψ,C(σ)) respectively 0 0 bi(Υ) = b (Υ,C(σi)) if it is clear from the context that we work on a fixed collar respectively 0 0 on collars around a fixed collection σi of simple closed geodesics. We will also use the convention that norms over C(σ) are{alw}ays computed with respect to the hyperbolic metric g =ρ2(ds2+dθ2). We recall from [17, Lemma 2.2] that the collar decay part of any holomorphic quadratic dif- ferential Υ H(M,g) on a collar C(σ) around a simple closed geodesics of length L (σ) g ∈ ≤ 2arsinh(1) decays rapidly along that collar in the sense that there exist universal numbers δ (0,arsinh(1)) and C so that 4 ∈ kΥ−b0(Υ)dz2kL∞(δ-thin(C(σ)) ≤Cδ−2e−π/δkΥkL2(δ4-thick(C(σ))) for every 0<δ ≤δ4. (2.10) 9 Conversely, for every δ >0 we may bound an arbitrary element Υ H(M,g) by ∈ Υ C Υ , (2.11) k kL∞(δ-thick(M,g)) ≤ δk kL1(δ-thick(M,g)) 2 in particular Υ C Υ , (2.12) L∞(δ-thick(M,g)) δ L2(M,g) k k ≤ k k where C depends on δ and the genus. Indeed, [18, Lemma 2.6] ensures that (2.12) holds true δ for C = Cδ−1/2, C depending only on the genus, and indeed also with the L∞-norm on the δ left hand side replaced by the Ck-norm (then with C depending additionally on k). We also recall that the collar regions around disjoint geodesics are disjoint, that the arsinh(1) thin part of a hyperbolic surface is always contained in the union of the collars around the simple closed geodesics of length less than 2arsinh(1), that such geodesics are always disjoint and that their number is no more than 3(γ 1). We refer to Appendix A.2 for an overview of − relevant results about hyperbolic surfaces. If σ1,...,σk is the set of all simple closed geodesics of (M,g) of length no more than some { } constant 2η <2arsinh(1) we hence have that, as observed in [17, Lemma 2.4], w C w (2.13) L∞(M,g) η L1(M,g) k k ≤ k k for all elements w W := Υ H(M,g) : b (Υ,C(σj)) = 0, 1 j k . Here and in η 0 ∈ { ∈ ≤ ≤ } the following all constants are allowed to depend on the genus in addition to the indicated dependences unless explicitly said otherwise. Wealsorecallthewell-knownfactthatalongacurve(g(t)) ofhyperbolicmetricswithg(0)=g t and∂ g(0)=ReΥforΥ H(M,g)theevolutionofthelengthℓ(t)ofthesimpleclosedgeodesic t ∈ σ (M,g(t)) homotopic to σ is given by t 0 ⊂ d ℓ= 2π2Re(b (Υ,C(σ ))) at t=0, (2.14) dt − ℓ 0 0 seee.g. [16,Remark4.12]. So,asobservedin[16,Remark4.1],ifweselectanyk disjointsimple closed geodesics σj in (M,g) we have ker(∂ℓ ,...,∂ℓ )= Υ H(M,g):bj(Υ)=b (Υ,C(σj))=0 for j =1,...,k , (2.15) 1 k { ∈ 0 0 } where ∂ℓ is defined as in (1.9) and thus given by j ∂ℓ (Υ)= 1( 2π2Re(bj(Υ))+i2π2Re(bj(iΥ)))= π2bj(Υ). (2.16) j 2 − ℓj 0 ℓj 0 −ℓj 0 In particular for σ1,...,σk chosen as above as the set of geodesics of length 2η we have { } ≤ that W = ker(∂ℓ ,...,∂ℓ ) and it was shown in the joint work [17] of Topping, Zhu and the η 1 k second author that at least for η sufficiently small codim(W )=codim(ker(∂ℓ ,...,∂ℓ ))=k. η 1 k The fine propertiesofthe elements ofW⊥, η small,were thenanalysedin[16] andwe shalluse η in particular the following version of [16, Lemma 4.5]. Lemma 2.5 (Contents of [16, Lemma 4.5]). For any genus γ 2 there exists a number ≥ η (0,arsinh(1)) so that for every η¯ (0,η ] the following holds true for a constant C that 1 1 ∈ ∈ depends only on η¯ and the genus: Let (M,g) be a closed oriented hyperbolic surface of genus γ and let σ1,...,σk be the set of all simple closed geodesics in (M,g) of length no more than { } 2η¯. Define W =W := Υ H(M,g) ∂ℓ (Υ)=0, j =1,...,k , η¯ j { ∈ | } ∂ℓ the differentials of the length coordinates associated to σj, compare (1.9). j 10