ebook img

On generalized Bohr-Sommerfeld quantization rules for operators with PT symmetry PDF

0.09 MB·
by  A. Ifa
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview On generalized Bohr-Sommerfeld quantization rules for operators with PT symmetry

ON GENERALIZED BOHR-SOMMERFELD QUANTIZATION RULES FOR OPERATORS WITH PT 6 SYMMETRY 1 0 2 b A.IFA1, N.M’HADHBI1,2 & M.ROULEUX3 e F 1 February 2, 2016 ] P S . h 1 Universite´ deTunisEl-Manar,De´partement deMathe´matiques,1091Tunis,Tunisia t a m email: [email protected] [ 2 v 3 2 2 DepartmentofMathematics,CollegeofSciences and Arts 9 4 KingAbdulazizUniversity,Rabigh Campus,P.O. Box 344,Rabigh 21911,Saudi Arabia 0 . email: [email protected] 1 0 6 1 3 AixMarseilleUniversite´,Centre dePhysiqueThe´orique, UMR7332, 13288Marseille,France : v i & Universite´ deToulon,CNRS, CPT, UMR7332,83957LaGarde, France X r email: [email protected] a Abstract We give Bohr-Sommerfeld rules corresponding to quasi-eigenvalues in the pseudo-spectrum for a 1-Dh-Pseudodifferential operatorverifying PTsymmetry. 1 Introduction and statement of the result Let p(x,x ;h)beasmooth (possibly complexvalued) Hamiltonian onT∗R,withtheformalexpansion p(x,x ;h)∼ p (x,x )+hp (x,x )+h2p (x,x )+.... Assumethatforsomeorderfunctionm, pbelongs 0 1 2 1 tothespaceofsymbolsS0(m),with SN(m)={p∈C¥ (T∗R):∀a ∈N2,∃Ca >0,∀(x,x )∈T∗R; |¶ (ax,x )p(x,x ;h)|≤Ca hNm(x,x )} (1.1) [for instance m(x,x )=(1+|x |2)M], and that p+i is elliptic. This allows to take Weyl quantization P= pw(x,hD ;h)of p x x+y Pu(x;h)=(2p h)−1 ehi(x−y)h p( ,h ;h)u(y)dydh (1.2) Z Z 2 whichwedenotealsobyP=Opw(p),or p=s w(P). Wecallasusual p theprincipalsymbol, p the 0 1 sub-principal symbol, andassumethroughout that p isreal. 0 Fix some compact interval I = [E ,E ],E < E and assume, following [6] that there exists a − + − + topological ring A ⊂T∗R such that ¶ A =A ∪A with A a connected component of p−1(E ). + − ± 0 ± Assumealsothat p hasnocriticalpontinA,andA isincludedinthediscbounded byA (ifthisis 0 − + notthecase,wecanalwayschange pto−p). WedefinethemicrolocalwellW asthediscboundedby par A . This includes the case of the standard Hamiltonian p (x,x )=x 2+V(x), but allows also for + 0 moregeneral geometries. ForE ∈I, let g ⊂W be acompact embedded Lagrangian manifold (periodic orbit) in theenergy E surface {p (x,x )=E}, and for N =1,2,···, let KN(E) denote the microlocal kernel of P−E up to 0 h orderN,i.e. thesetoflocalsolutionsof(P−E)u=O(hN+1)inthedistributionalsense,microlocalized on g . This is a smooth complex vector bundle over p (g ), where p :T∗R→R. Finding the set of E x E x E =E(h;N)suchthatKN(E)contains aglobalsection, amountstoconstruct normalized quasi-modes h (QM) (u(h;N),E(h;N)) up to order N. In other words, the condition that determines the sequence of quasi-eigenvalues E(h;N)=E (h;N)isthatthecorresponding quasi-eigenfunction u(h;N)=u (h;N) n n be single-valued. It is known as Bohr-Sommerfeld condition (BS for short). In the sequel, we drop indexN whenunnecessary. AssumethatPisself-adjoint, andE <E =liminf p (x,x ). ThenBSdetermines asymptotically + 0 0 |x,x |→¥ alleigenvaluesofPinI,bytheequationS (E (h))=2p nh,toanyorderN. Thesemi-classicalaction h n S (E)hasasymptoticsS (E)∼S (E)+hS (E)+h2S (E)+.... S istheclassicalaction x dx= h h 0 1 2 0 Ig E dx ∧dx,S (E)=p − p ((x(t),x (t))dt includesMaslovcorrectionandthesubprincipal Z{p0≤E}∩W 1 ZgE 1 1-form p dt, where t is the parameter in Hamilton equations. Terms S and S are computed using 1 0 1 Maslovcanonical operator, ormorespecifically inthepresent 1-Dcase, themonodromy operator (see [11] and references therein). A more systematic method (still in the 1-D case) is based on functional calculus for h-PDO’s, in particular Moyal’s product, and uses a general formula due to [16] (see also 2 [2]forearlierwork). Thus,with ¶ 2p ¶ 2p ¶ 2p D = 0 0 −( 0 )2 (1.3) ¶ x2 ¶x 2 ¶ x¶x wehave 1 d 1 d S (E)= D dt− p dt− p2dt (1.4) 2 24 dE Zg Zg 2 2 dE Zg 1 E E E Thismethod wasimplemented indifferent waysby[14],[6]andgivenlater adiagrammatic approach by [5] and [10], providing an algorithm to compute all higher order terms, in particular the 4th order term can be computed in a closed form without too much work. Note that all S (E) with j ≥3 odd j vanish. Itisshownfurtherin[6],usingtraceformulas,thatBSgivesactuallyalleigenvaluesinI. Notethat this approach, in contrast withthe method ofthe monodromy operator, assumes already the existence ofBS,andtheproblem isaboutthemostefficientwayofcomputing theS ’s. j Intherealanalyticcase,whenP=−h2D +V(x)isSchro¨dingeroperator,BScanbeobtainedusing the exact complex WKB method (see [9], [7] and references therein); it consists first in transforming theeigenvalue equation −h2u′′(x)+V(x)u(x)=Eu(x)intoaRicattiequation, andthencompute Jost function whosezeroesareprecisely theeigenvalues ofP. Consider now a h-PDO P (not necessarily self-adjoint) that satisfies PT symmetry i.e. PPT = PT P, where PT =X I, X is the parity operator Xu(x)=u(−x) and I the complex conjugation. At the level of Weyl symbol, this symmetry takes the form p(−x,x ;h)= p(x,x ;h). Such a property issometimesconsidered inPhysicsasanaturalsubstitute forself-adjointness. Itisknownthatfinding quasi-modes is in no ways sufficient to get information about the spectrum of P, but only about its pseudo-spectrum (see [9], [8] and [15] for more recent results). The pseudo-spectrum is symmetric withrespecttotherealaxis,andoneexpectsgenerallytorecoversomerealeigenvalues. Wespecialize furtherinthecasewherePhasarealprincipal symbol. Ourmainresultisthefollowing: Theorem1.1. LetPasaboveenjoyPTsymmetry,and p bereal. Then,foratleastN=4,thereexists 0 b∈ S0(m) defined microlocally in W, such that Q= BPB−1,b =s w(B), is formally self-adjoint (at least modulo an operator with symbol in SN+1(m)). In particular, there is asequence of quasi-modes (u (h),E (h)) such that (P−E (h))u (h)=O(hN+1), with E (h)∈I, satisfying S (E (h))=2p nh, n n n n n h n N for an asymptotic series S (E)= (cid:229) S (E)hj+O(hN+1) where S ∈R are real. In particular, the h j j j=1 pseudo-spectrum of Plies within adistance O(hN+1)ofI. Thecoefficients S (E)canbecomputed as j in[10]fromthesymbolofQ;thus S (E)= x (x)dx= dx ∧dx 0 IgE Z Z{p0≤E}∩W 3 istheactionintegral, S (E)=p − Re(p (x(t),x (t))dt, 1 Zg 1 E and 1 d 1 1 d S (E)= D dt− Re(p )− {{b ,p },b } dt− (Re(p ))2dt 2 24 dE Zg Zg 2 2 0 0 0 2 dE Zg 1 E E(cid:0) (cid:1) E withD asin(1.3),and s b (x,x )= (1− )Im(p )◦expsH (x,x )ds 0 Ig T(E) 1 p0 E Denoting by T(E) the period of the flow on g . Again, S = 0, and S (E) can be computed using E 3 4 ([5],Formula(7.3))andtheformulagivings w(BPB−1)modO(h5). Ofcourse, weconjecture thatTheorem1.1holdsforallN. Example 1.1. Consider the operator Q(x,hD )=(hD )2+p(x)hD +q(x) with smooth, real coeffi- x x x cients. Then Q can be mapped into P(x,hD )=(hD )2+q(x)− 1(p(x))2+ih p′(x) by the unitary x x 4 2 transformation Q=BPB∗, Bv(x;h) =exp(− i xp(t)dt)v(x). Assume Q verifies PT symmetry, i.e. 2h R p and q are even on R, then the same holds for P. The microlocal wellW ={(x,x )∈T∗(R);x 2+ E q(x)− 1(p(x))2 ≤E} for P projects onto the potential wellU ={x∈R;q(x)− 1(p(x))2 ≤E}, so 4 E 4 Theorem1.1holdsprovidedV(x)=q(x)−1(p(x))2 hasnocriticalpointinI. 4 If pandqanalytic, thenthespectrumofPisinfactrealinI andgivenby(exact)BS.Infact,using alsosomeofthetechnicselaboratedin[12],[4]showedthatifP=−h2D +V(x)+ie W(x)isasmall perturbation oftheself-adjoint Schro¨dingeroperatorP =−h2D +V(x),thenthesemiclassicalaction 0 is a real analytic function and the roots of BS are real eigenvalues of P. This implies Theorem 1.1 by choosing e =h,buttheargumentof[4]heavilyreliesuponthatparticular formofP. 2 Proof of the Theorem Since we know S (E) for 1 self-adjoint operator ([14],[16]), it suffices to conjugate P by an elliptic h (but non-unitary) h-PDO so that the resulting operator becomes formally self-adjoint up to order N. Weproceedinseveralsteps. 4 2.1 Birkhoff normal form (BNF) LetP˜beself-adjointasin(1.3)with(real)Weylsymbol p˜∈S0(m),andassumethatitsprincipalsymbol p˜ = p has a periodic orbit g at non critical energy E =0. Then there exists a smooth canonical 0 0 0 transformation (s,t ) 7→ k (s,t ) = (x,x ), s ∈ [0,2p ], defined in a neighborhood of g and a smooth 0 function t 7→ f (t ), f (0)=0, f′(0)6=0 such that p ◦k (s,t )= f (t ). Energy E and momentum t 0 0 0 0 0 arerelatedbythe1-to-1transformationE= f (t ). ThistransformationcanbequantizedtotakeP˜inits 0 semi-classicalBNF.Namely,thereisamicrolocallyunitaryh-FIOoperatorU associatedwithk ,anda classical symbol f(t ;h)= f (t )+hf (t )+h2 f (t )+···, such thatU∗P˜U = f(hD ;h). Seee.g. [3] 0 1 2 s attheleveloftheprincipalsymbol,and[13]forthefullsymbol;BNFturnsouttobeconvergent inthe 1-Dcase. Inthecanonical(action-angle) variables(s,t ),s∈[0,2p ],theparityoperatorX:x7→−xon thereallinetakestheformX:s7→p −sonthecircle. Moreover,wecanchooseU sothatitcommutes withPTsymmetry:UPT =PTU. 2.2 The homological equation Westartwiththefollowingelementary result(seee.g. ([15],p.93)): Lemma2.1. Letqand pberealHamiltonians. Thentheequation q+{b ,p}=0 (2.1) hasa(global) realsolutionb alongg iff E q◦exptH (r )dt =0 (2.2) Ig p E foranyr ∈g . Itisgivenby E t b (r )=− (1− )q◦exptH (r )dt (2.3) Ig T(E) p E Lemma2.2. Assume p= p asaboveandqisoddwithrespecttoPT ;then(2.2)holds. 0 ¶ Proof. Using action-angle coordinates (s,t ) we have p (s,t ) = f (t ), hence H (t,t ) = f′(t ) , 0 0 p0 0 ¶ t where f′(t )=w (E)istheenergydependent frequency. 0 Forr =(s,t ),exptH (r )=f (r )=(s+w (E)t,t ). Then,usingtheperiodicity ofq p0 t T(E) 1 s+2p q◦exptH (r )dt = q(s+w (E)t,t )dt = q(s′,t )ds′ IgE p Z0 w (E)Zs whichis0sinceq(.,t )isoddasafunctiononthecircle. 5 2.3 Reducing to a formally self-adjoint operator Proposition 2.1. Let p(x,x ;h)∼ p (x,x )+hp (x,x )+h2p (x,x )+···∈S0(m)satisfy PTsymmetry 0 1 2 withreal p . Then at least for N =4, there exists b∈S0(m)elliptic such that Weyl symbol of BPB−1, 0 s W(B),isrealmodO(hN+1). Moreover,BPB−1isagainPT-symmetric(uptothatorder). Proof. To shorten the exposition, we content to the lower order accuracy O(h4). First we carry BNF P+P∗ for the self-adjoint part P˜ = of P, which has real Weyl symbol and verifies PT symmetry. 2 P−P∗ SinceU commutes with PT , the anti-self adjoint part also satisfies PTsymmetry (but isnot 2 necessarily inBNF). Check firstthe Proposition for N =1. LetB have Weyl symbol b , which wewrite as s w(B )= 0 0 0 b0. Let b0 = eb0, with real b 0. By h-Pseudodifferential Calculus (i.e. Moyal product), B0PB0−1 = [B ,P]B−1+PhasWeylsymbol 0 0 h2 h3 h4 s W B PB−1 = p−ih{b ,p}+ {b ,p},b +i R (b ,a (p))+ R (b ,a (p))+O(h5) 0 0 0 2 0 0 8 5 0 48 8 0 (cid:0) (cid:1) (cid:8) (cid:9) (2.4) with a (p)={b ,p} (2.5) 0 and ¶ 2b ¶ 2b ¶ 2a (p) R5(b 0,a (p))= (¶ x b 0)2− ¶x 20 × 2¶ xa (p)¶ xb 0+a (p) ¶ x20 + ¶ x2 +a (p)(¶ xb 0)2 (cid:0) (cid:1) (cid:0) (cid:1) ¶ 2b ¶ 2b ¶ 2a (p) + (¶ xb 0)2− ¶ x20 × 2¶ x a (p)¶ x b 0+a (p) ¶x 20 + ¶x 2 +a (p)(¶ x b 0)2 (cid:0) (cid:1) (cid:0) (cid:1) ¶ 2b ¶ 2a (p) ¶ 2b −2 ¶ xb 0¶ x b 0−¶ x¶x 0 × ¶ x¶x +¶ x a (p)¶ xb 0+¶ xa (p)¶ x b 0+a (p)¶ xb 0¶ x b 0+a (p)¶ x¶x 0 (cid:0) (cid:1) (cid:0) (cid:1) ¶ 2b ¶ 3b R (b ,a (p))=F (b ,a (p)) 3¶ b 0 − 0 −(¶ b )3 8 0 5 0 x 0 ¶ x2 ¶ x3 x 0 (cid:0) (cid:1) ¶ 2b ¶ 3b −F˜5(b 0,a (p)) 3¶ x b 0 ¶x 20 − ¶x 30 −(¶ x b 0)3 (cid:0) (cid:1) ¶ 2b ¶ 3b ¶ 2b +3G5(b 0,a (p)) 2¶ x b 0¶ x¶x 0 −¶ x¶x 02 −¶ xb 0(¶ x b 0)2+¶ xb 0 ¶x 20 (cid:0) (cid:1) ¶ 2b ¶ 3b ¶ 2b −3G˜5(b 0,a (p)) 2¶ xb 0¶ x¶x 0 −¶ x2¶x0 −¶ x b 0(¶ xb 0)2+¶ x b 0 ¶ x20 (cid:0) (cid:1) where ¶ 2b ¶ 3b ¶ 2a ¶ 3a ¶ 2b F5(b 0,a (p))=3¶ x a ¶x 20+a ¶x 30+3¶ x b 0 ¶x 2 +¶x 3 +3¶ x a (¶ x b 0)2+3a¶ x b 0 ¶x 20+a (¶ x b 0)2 6 ¶ 2b ¶ 3b ¶ 2a ¶ 3a ¶ 2b F˜ (b ,a (p))=3¶ a 0+a 0+3¶ b + +3¶ a (¶ b )2+3a¶ b 0+a (¶ b )2 5 0 x ¶ x2 ¶ x3 x 0 ¶ x2 ¶ x3 x x 0 x 0 ¶ x2 x 0 ¶ 2a ¶ 2b ¶ 2b ¶ 3b ¶ 3a G5(b 0,a (p))=2¶ xb 0¶ x¶x +2¶ xa ¶ x¶x 0 +¶ x a ¶ x20 +a ¶ x2¶x0 +¶ x2¶x ¶ 2b ¶ 2b ¶ 2a +¶ x a (¶ xb 0)2+2a¶ xb 0¶ x¶x 0 + 2¶ xa¶ xb 0+a ¶ x20 + ¶ x2 +a (¶ xb 0)2 ¶ x b 0 (cid:0) (cid:1) ¶ 2a ¶ 2b ¶ 2b ¶ 3b ¶ 3a G˜5(b 0,a (p))=2¶ x b 0¶ x¶x +2¶ x a ¶ x¶x 0 +¶ xa ¶x 20 +a ¶ x¶x 02 +¶ x¶x 2 +¶ xa (¶ x b 0)2 ¶ 2b ¶ 2b ¶ 2a +2a¶ x b 0¶ x¶x 0 + 2¶ x a¶ x b 0+a ¶x 20 + ¶x 2 +a (¶ x b 0)2 ¶ xb 0 (cid:0) (cid:1) HereR (b ,a (p ))isaHamilton-Jacobipolynomialwithintegercoefficients,polynomialinthederiva- 5 0 0 tivesofb 0 upto order 2, homogeneous ofdegree 5(total degree in(¶ x,¶ x ))whencounting altogether products andderivatives; andsimilarly for R (b ,a (p )). Notethatthese Hamilton-Jacobi polynomi- 8 0 0 alsdependlinearlyona (p). Thefirstordertermofthesymbolisrealiff {b ,p }=Im(p ) (2.6) 0 0 1 andbyLemmas2.1and2.2thisequation canbesolvedong ,and E T(E) t b (s,t )= (1− )Im(p )(s+w (E)t,t )dt (2.7) 0 Z T(E) 1 0 with w (E)T(E)=2p (2.8) Wenoticethatb (.,t )isanevenfunctiononthecircle. Soin(2.4)weareleftwith 0 1 s W(B PB−1)= p +hRe(p )+h2 Re(p )− {b ,p },b +O(h3) 0 0 0 1 2 2 0 0 0 (cid:0) (cid:8) (cid:9)(cid:1) Chek now the proposition for N = 2. Let B1 have Weyl symbol s W(B1) = ehb1 with b 1 real, and computeWeylsymbolofB B PB−1B−1. AgainbyMoyalproduct, wegetmodO(h3) 1 0 0 1 1 s W B B PB−1B−1 ≡ p +h p −i{b ,p })+h2 p −i{b ,p }−i{b ,p }+ {{b ,p },b } 1 0 0 1 0 1 0 0 2 0 1 1 0 2 0 0 0 (cid:0) (cid:1) (cid:0) (cid:0) (cid:1) (2.9) Theequation forb reads 1 {b ,p }=Im(p )−{b ,Re(p )} (2.10) 1 0 2 0 1 and we need to check the solvability condition (2.2). It is fulfilled when q = Im(p ), since this is 2 an odd function on the circle; consider now q = {b ,Re(p )}, in action-angle coordinates we have 0 1 ¶b Re(p )(t,t )= f (t ),so{b ,Re(p )}=−f′(t ) 0. Sinceb is2p -periodic 1 1 0 1 1 ¶ t 0 T(E) {b ,Re(p )}(s+w (E)t,t )dt =−f′(t )(b (s+2p ,t )−b (s,t ))=0 Z 0 1 1 0 0 0 7 soagainthecompatibilityconditionholdsforq={b ,Re(p )},(2.10)canbesolved,and(2.9)reduces 0 1 to 1 s w(B B PB−1B−1)= p +hRe(p )+h2 Re(p )− {b ,p },b +O(h3) (2.11) 1 0 0 1 0 1 2 2 0 0 0 (cid:0) (cid:8) (cid:9)(cid:1) Wenoticethatb (.,t )isanevenfunctiononthecircle. 1 NextwelookforB2=OpW(eh2b2),b 2 realsothatB2B1B0PB−01B−11B−21becomesself-adjointupto O(h3);theequationforb reads 2 {b ,p }=Im(p )−(cid:229) 1 {b ,Re(p )}+1{{b ,Im(p )},b }+1R (b ,a (p )) (2.12) 2 0 3 k 2−k 0 1 0 5 0 0 2 8 k=0 We need to check the compatibility condition for solving (2.12), by the previous work it suffices to considerq={{b ,Im(p )},b },andq=R (b ,a (p )). Usingagainaction-angle variables, wehave: 0 1 0 5 0 0 ¶b ¶ 2b ¶b ¶ 2b ¶b ¶ 3b ¶b {{b ,Im(p )},b }=3f′′(t )( 0)2 0 + f′(t ) 0( 0)2+ f′(t )( 0)2 0 + f′′′(t )( 0)3 0 1 0 0 ¶ t ¶t¶ t 0 ¶ t ¶t¶ t 0 ¶ t ¶t 2¶ t 0 ¶ t ¶ 2b ¶b ¶ 2b ¶b ¶b ¶ 3b ¶b ¶b ¶ 2b ¶b ¶ 3b − f′(t ) 0 0 0 −2f′(t ) 0 0 0 −3f′′(t ) 0 0 0 + f′(t )( 0)2 0 0 ¶t 2 ¶ t ¶ t2 0 ¶t ¶ t ¶t¶ t2 0 ¶t ¶ t ¶ t2 0 ¶t ¶ t3 andthecompatibilityconditionisfulfilledforthattermsinceallfunctionstobeintegratedonthecircle areodd. For R (b ,a (p )) we proceed similarly. We check again that b is an even function on the circle 5 0 0 2 (i.e. underthetransformationt 7→p −t). Onceweknowb ,wecomputes w(B B B PB−1B−1B−1)modO(h4),thisgives: 2 2 1 0 0 1 2 1 s W(B B B PB−1B−1B−1)≡ p +hRe(p )+h2 Re(p )− {b ,p },b 2 1 0 0 1 2 0 1 2 2 0 0 0 (cid:0) (cid:8) (cid:9)(cid:1) 1 +h3 Re(p )− {b ,p },b − {b ,Re(p )},b } 3 1 0 0 0 1 0 2 (cid:0) (cid:8) (cid:9) (cid:8) (cid:1) LetnowB3haveWeylsymbols W(B3)=eh3b3 withb 3real,andcomputeWeylsymbolofB3B2B1B0PB−01B−11B−21B−31. AgainbyMoyalproduct, wegetmodO(h5) s w B B B B PB−1B−1B−1B−1 ≡ p−ih(cid:229) 3 hj{b ,p}−h3 (cid:229) 1 hj b ,{b ,p} +h2 (cid:229) 1 h2j {b ,p},b 3 2 1 0 0 1 2 3 j j+1 0 2 j j (cid:0) (cid:1) j=0 j=0 (cid:8) (cid:9) j=0 (cid:8) (cid:9) h3 h4 h4 +i R (b ,a (p))+ R (b ,a (p))−i b , {b ,p},b 5 0 8 0 1 0 0 8 48 2 (cid:8) (cid:8) (cid:9)(cid:9) Theequation forb nowreads 3 {b ,p }=Im(p )−(cid:229) 2 {b ,Re(p )}+1{{b ,Im(p )},b }−1 {b ,p },b ,b +1R b ,a Re(p ) 3 0 4 k 3−k 0 2 0 0 0 0 1 5 0 1 2 2 8 k=0 (cid:8)(cid:8) (cid:9) (cid:9) (cid:0) (cid:0) (cid:1)(cid:1) 8 ThecompatibilityconditionisfulfilledforIm(p )and{b ,Re(p )},0≤k≤2asbefore,forthedou- 4 k 3−k blePoissonbracket{{b ,Im(p )},b },thetriplePoissonbracket {b ,p },b ,b andR b ,a Re(p ) 0 2 0 0 0 0 1 5 0 1 (cid:8)(cid:8) (cid:9) (cid:9) (cid:0) (cid:0) (cid:1)(cid:1) weproceedsimilarly. WededuceS (E)=0. 3 Onceweknowb ,wecomputes w(B B B B PB−1B−1B−1B−1)modO(h5)thisgives: 3 3 2 1 0 0 1 2 3 1 s w(B B B B PB−1B−1B−1B−1)= p +hRe(p )+h2 Re(p )− {{b ,p },b } 3 2 1 0 0 1 2 3 0 1 2 2 0 0 0 (cid:0) (cid:1) 1 +h3 Re(p )− {b ,p },b − {b ,Re(p )},b 3 1 0 0 0 1 0 2 (cid:0) (cid:8) (cid:9) (cid:8) (cid:9)(cid:1) +h4 Re(p )+(cid:229) 1 {b ,Im(p )}+1{{b ,Re(p )},b }+1{{b ,p },b } 4 k 3−k 0 2 0 1 0 1 2 2 (cid:0) k=0 (cid:1) 1 1 +h4 {b ,Re(p )},b − R b ,a Re(p ) + R b ,a (p ) 0 1 1 5 0 1 8 0 0 8 48 (cid:0)(cid:8) (cid:9) (cid:0) (cid:0) (cid:1)(cid:1) (cid:0) (cid:1)(cid:1) To prove the Theorem, we observe eventually that the knowledge of the symbol of BPB−1 mod O(h5) ([10],Formula(7.3)) is sufficient to compute S (E) (although this formula was derived in the 4 particular casewherethesymbolofPcontains only p ). 0 3 Extension to operators with periodic coefficients Asin[6],wereplaceT∗RbyT∗S1,andthehypothesis onPbythefollowing: • thereisatopologicalringA,homotopictothezerosectionofT∗S1,suchthat¶ A =A ∪A with − + A aconnected component of p−1(E ). ± 0 ± • p hasnocriticalpointsinA. 0 • A is”below”A . + Then Theorem 1.1 holds; the only change is that S (E)=0. Again we reduce P to f(hD ;h) as an 1 t operator onS1. Acknowledgments: Thesecondauthor(N.Mhadhbi)acknowledges withthankstheDeanshipofScien- tificResearchDSR,KingAbdulazizUniversity(Jeddah)foritssupport. Thethirdauthor(M.Rouleux) cheerfully thanks S.Dobrokhotov, for his kind hospitality at Ishlinskiy Institute for Problems of Me- chanics(Moscow). 9 References [1] O.Rouby. Bohr-Sommerfeld quantization conditions for non-selfadjoint perturbations of selfad- jointoperators indimension one.http://arxiv.org/abs/1511.06237 [2] P.Argyres. TheBohr-Sommerfeld quantization rule and Weylcorrespondence, Physics 2, p.131- 199(1995) [3] V.Arnold,V.Koslov,A.Neishtadt. Mathematicalaspects ofclassical andcelestial mechanics. En- cyclopedia ofMath.Sci.DynamicsSystemsIII,Springer, 2006. [4] N.Boussekine, N.Mecherout. PT-syme´trieetpuitsdepotentiel. arXiv:1310.7335v1[math.SP]. [5] M.Cargo, A.Gracia-Saz, R.Littlejohn, M.Reinsch & P.de Rios, Moyal star product approach to theBohr-Sommerfeldapproximation, J.Phys.A:MathandGen.38,1977-2004 (2005). [6] Y.ColindeVerdie`re.Bohr-SommerfeldRulestoAllOrders.Ann.H.Poincare´,6,p.925-936,2005 [7] E.Delabaere, H.Dillinger, F.Pham.Exact semi-classical expansions for 1-D quantum oscillators. J.Math.Phys.Vol.38(12)p.6126-6184 (1997) [8] A.I.Esina,A.I.Shafarevich.Quantizationconditionsonriemanniansurfacesandthesemiclassical spectrumoftheSchro¨dingeroperatorwithcomplexpotential.MathematicalNotes,2010,Vol.88, No.2,pp.6179. [9] M.V.Fedoriouk. Me´thodes asymptotiques pourlesEquations Diffe´rentielles Ordinaires Line´aire. Ed.MIR,Moscou,1987.(=Asymptotic Analysis. Springer,1993) [10] A.Gracia-Saz. The symbol of a function of a pseudo-differential operator. Ann. Inst. Fourier, 55(7),p.2257-2284 (2005) [11] B.Helffer,D.Robert.Puitsdepotentiel generalise´s etasymptotique semi-classique. AnnalesInst. H.Poincare´ (PhysiqueThe´orique), Vol.41,No3,p.291-331 (1984) [12] B.Helffer, J.Sjo¨strand.3. Semi-classical analysis for Harper’s equation III. Me´moire No39, Soc. Math.deFrance,117(4)(1988) [13] M.Hitrik,J.Sjo¨strand, S.Vu-Ngoc.Diophantine toriandspectral asymptotics fornon-self adjoint operators. Amer.J.Math.129(1), p.105-182, 2007. [14] R.Littlejohn,LieAlgebraicApproachtoHigher-OrderTerms,Preprint17p.(2003).(June2003). [15] J.Sjo¨strand. Weyl law for semi-classical resonances with rondomly perturbed potentials. Mem- oiresSo.Math.France,NouvelleSe´rie,No136,2014. 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.