Variation of the Nazarov-Sodin constant for random plane waves and arithmetic random waves Par Kurlberg (KTH Stockholm) Igor Wigman, KCL Random Waves in Oxford Oxford, June 19, 2018
1. Motivation & Background
General Setup (π, π) β Compact smooth surface (can ο generalize higher dimensions) Ξ Laplace-Beltrami on M ο Eigenfunctions: (boundary condition) ο π π β₯ 0 Ξπ π + π π π π = 0 ο Orthonormal basis of L 2 (M,dVol), π π β β
Nodal components & domains β1 (0) ο Nodal set: π π π = π π ο Nodal components: Connected components β1 (0) . of π π ο Nodal domains: Connected components of β1 (0) smooth π β π π ο Nodal count: How many components (domains)?
Interesting Questions β non-local Nodal count (Nazarov-Sodin `09,`12, `15, ο Kurlberg-W `14, `17) ο T opology, nesting,geometry(Sarnak-W, Beliaev-W) ο Local: πππ π΅βͺπΆ π = πππ π΅ π + πππ πΆ π π΅ β© πΆ = β ο Semilocality: βMostβ of the nodal domains of π diameter π , R>>0. ο Approximate locally
Nodal count (deterministic) ο Nodal Count. Courant: π π π β€ π π π π ο Pleijel: limsup β€ 0.691 β¦ π πββ ο Constant improved by 3 β 10 β9 (Bourgain) ο No lower bound π π π β₯ 2 Nodal picture for the square, arbitrarily high energy. A. Sternβs thesis, Gottingen, 1925. Courtesy of P. Sarnak.
Berryβs Random Wave Model ο M chaotic. As π β β, π π βbehave randomlyβ wavenumber π monochromatic wave β 2 1 πΎ π π( π π¦,π π +π π ) πΎ ππ π=1 π£ π (π¦) = ο Scale invariant, assume π£ = π£ 1 ο Centered Gaussian, covariance E[π£(π¦) β π£(π§)] = πΎ 0 (|π¦ β π§|) ο Spectral measure β arc length on unit circle
2. Random Band Limited Functions
Random Band-Limited Functions ο Fix M β smooth n-manifold, 0 β€ π½ β€ 1 π x = π½πβ€π π β€π π π π π (π¦) , π π - N(0,1) i.i.d. π ( π½ = 1 summation over π β π π β€ π π β€ π ) ο Covariance function π½ π π π¦ π π π§ = π π (π¦)π π (π§) i.e. the spectral projector.
Example 1. Random Spherical harmonics ο π½ = 1, π = π― 2 , 2d sphere. ο π½ π π x β π π y = π π cos(π π¦, π§ ) . ο π π cos(π) β πΎ 0 ππ Legendre fast uniform ο Scales Berryβs RWM RWM Random spherical harmonics A. Barnett
π½ =1 vs π½ =0 (Alex Barnett) π½ =1 π½ =0 Random spherical harmonics βReal Fubini- Studyβ
Example 2. T oral eigenfunctions. β 2 ο π = π = β€ 2 π π¦ = π 2 =π π π β π ο π π¦, π π π standard Gaussian i.i.d. (save to π βπ = π π ) β arithmetic random waves β π 2 = π ο Summation over π β β€ 2 : lattice points on radius π circle
More general: limiting ensembles ο Natural scaling around any point of M. ο Scaling for covariance (values & derivatives) (classical Hormander, Lax) π½ π π x β π π y β πΏ π½ π β π π¦, π§ ο πΏ π½ π₯ = πΏ π½ π₯ = π½β€ π₯ β€1 π π₯, π ππ ο Canzani-Hanin `16 π½ = 1 thin window. ο Define π β on β 2 , βcleanβ covariance π½[π β π¨ β π β π¨β² ] = πΏ π½ π¨ β π¨β²
Limiting ensembles (cont.) ο π½[π β π¨ β π β π¨β² ] = πΏ π½ π¨ β π¨β² ο π β scaling limit π π (everywhere) ο π β depends on π½ , not on M, x (universality) ο Spectral measure π½ 1 ο Relevant: nodal structures of π β restricted on ball πΆ π , π β β. ο E.g. nodal count of domains lying in πΆ π .
3. Nazarov-Sodin Constant
Scale invariant (Euclidean) case ο πΊ: β 2 β β stationary Gaussian field ο π spectral measure of πΊ ο π(πΊ; π) is the number of connected components (domains) of πΊ inside πΆ(π) ο Assuming: 1. πΊ ergodic ( π has no atoms) 2. πΊ smooth. 3. Non-degeneracy ο Nazarov-Sodin (`12,`15): π = π ππ (π) β₯ 0 πΉ[π πΊ; π ] = π β π 2 + π πββ (π 2 ) ο βUsuallyβ π > 0 (support of π )
Nodal count band-limited functions = π β π 2 + π πββ (π 2 ) ο π = π ππ (π) , E π πΊ; π ο Stronger convergence in mean (ergodicity) π πΊ;π βπβπ 2 E β 0 π 2 ο Band-limited functions π = π ππ π π½ > 0 π π½ area measure annulus π )]~π β π (NS `12,`15) ο E[π(π π½ 1 ο Convergence in mean
Random spherical harmonics ο π£ = Plane monochromatic waves (RWM) ο πΉ[π π£; π ]~π πππ β π 2 , universal NS constant 2π > 0 percolation? ππ ο π πππ = π ππ ο πΉ π(π π ) ~π πππ β π 2 (NS `09) ο Convergence in mean 1 ο Exponential probability concentration
4. Variation on Nazarov-Sodin consant
Generalise NS constant ο Restrict to π supported on the unit ball π¬ (spectral moments), includes band limited case ο Proposition 1(Kurlberg-W): c = π ππ (π) , E[π πΊ; π ] = π β π 2 + π(π) , π β π¬ arbitrary, absolute constant (uniform) ο π ππ (π) bounded (e.g. critical points Kac-Rice) ο No convergence in mean. Can construct π πΊ;π βπβπ 2 examples E doesnβt vanish π 2 ο Example: π atomic supported at 0. πΊ β‘ ππππ‘π’ , Gaussian π πΊ; π β‘ 0 , β π ππ π = 0.
Variation of NS constant ο Proposition 2 (Kurlberg-W): π πΊ;π βπβπ 2 exists π ππ π β lim πββ E π 2 (βNS discrepancy functionalβ) non -uniform, discontinous ο Theorem 1 (Kurlberg-W): π ππ π : π¬ β β β₯0 is continuous (weak* topology on π¬ ). ο Corollary: π ππ π attains an interval [0, π 0 ] ( π¬ is essential) ο Q: Is it true that π = π πππ , uniquely ?
5. Toral Eigenfunctions
Example 2. T oral eigenfunctions. β 2 ο π = π = β€ 2 π π¦ = π 2 =π π π β π ο π π¦, π π π standard Gaussian i.i.d. (save to π βπ = π π ) β arithmetic random waves β π 2 = π ο Summation over π β β€ 2 : lattice points on radius π circle π π¦ ) nodal count ο π(π
On the 2 squares problem π, π πβ€ 2 : π 2 + π 2 = π ο π 2 π = # ο On average π 2 π ~c β log(π) (E. Landau) Equidistributed Exceptional β Cilleruelo β generic π 2 (n) β β 1 1 ο Partial classification (P . Kurlberg-IW `15)
Some pics π = 1105 32 directions π = 9676418088513347624474653 256 directions Fragment, domains long and narrow
On the 2 squares problem 1 π 2 (π) π 2 =π π π/ π probability π― 1 ο π π = (spectral measure) ππ ο Equidistributed ( π π π β π ) π π π β 2π ο Angular distribution β Nodal structure Local: Krishnapur-Kurlberg-W `13, Rudnick-W `14, Rossi-W `17 ο Nonlocal: N(π π π¦ ) β total nodal count Nazarov-Sodin `12,`15+Kurlberg-W `14,17
T oral eigenfunctions arithmetic π π¦ = π 2 =π π π β π π¦, π ο π random waves 1 π 2 (π) π 2 =π π π/ π on π 1 ο π π = ο Apply N-S: if π π β π then π = π ππ π (generalised) πΉ π(π π π¦ ) = π β π + π π π π π π¦ βπ πππ ο Generic πΉ β 0 π ο Exponential concentration (Y. Rozenshein `15)
T oral eigenfunctions (cont.) arithmetic π π¦ = π 2 =π π π β π π¦, π ο π random waves 1 π 2 (π) π 2 =π π π/ π on π 1 ο π π = ο Theorem 2 (Kurlberg-W): 1. Uniformly πΉ π(π π π¦ ) = π ππ (π π ) β π + π π 2. π ππ π = 0 iff π is Cilleruelo or its tilt ( π restricted, in particular by symmetries)
T oral eigenfunctions (cont.) ο π ππ π = 0 iff π is Cilleruelo or its tilt (restricted) ο π ππ π attains an interval 0, π 1 . ο Q.: Is it true that π 1 = π 0 = π πππ uniquely? ο Q.: For Cilleruelo: πΉ π(π π π¦ ) - ? π π¦ ) β β - ? πΉ π(π π - ? πΉ π(π π π¦ ) β« ο Meanwhile full classification π ππ π = 0 (Beliaev-McAuley-Muirhead). Same for π ππ π ?
Recommend
More recommend