Ma-Xu quantization rule and exact WKB condition for translationally shape invariant potentials Y. Grandati and A. B´erard Institut de Physique,, ICPMB, IF CNRS 2843, Universit´e Paul Verlaine, 1 Bd Arago, 57078 Metz, Cedex 3, France For translationally shape invariant potentials, the exact quantization rule proposed by Ma and Xu is a direct consequence of exactness of the modified WKB quantization condition proved by Barclay. We propose here a very direct alternative way to calculate the appropriate correction for the whole class of translationally shape invariant potentials. PACSnumbers: 0 1 INTRODUCTION 0 2 n In2005,MaandXu[1,2],onthebasisofapreviousworkofCao[3,4]andal.,proposedanewimprovedquantization a J rule permitting to retrieve the exact spectrum of some exactly solvable quantum systems. Since, numerous papers have been published on the possible applications of the Ma-Xu formula to different systems [5–14]. Until now, the 4 link between the exactness of the Ma-Xu formula and the solvability of the considered system seems to be stayed ] unclear. In fact, the translationally shape invariance of the potential suffies to ensure the exactness of the Ma-Xu h formula. This is a direct consequence of a result obtained first by Barclay [19] in 1993 and largely overlooked. He p - established that for these specific potentials the higher order terms in the WKB series can be resummed, yielding to h an energy independent correction and he showed that this Maslov index can be obtained in a closed analytical form. t a More than ten years later, Bhaduri and al. [20] proposed another interesting derivation of this result which rests on m periodic orbit theory (POT) [21]. [ In the present article, after having established the connection between the Ma-Xu formula and Barclay’s result, we propose an alternative way to calculate the Maslov index for every translationally shape invariant potential (TSIP). 1 Starting from a classification of TSIP which uses new criterions [22], equivalent to the Barclay-Maxwell ones [18], we v 5 show how to obtain this index using simple complex analysis tools. 9 4 0 MA-XU QUANTIZATION FORMULA . 1 0 Consider the stationary one dimensional Schr¨odinger equation (~=1, m=1/2): 0 1 v: ψ′′(x)+p2(x)ψ(x)=0, (1) i X where p(x)= E−V(x) is the classical momentum function for an energy E. If x1 and x2 are the classical turning ar points then p2(x)≥0 for x∈[x1,x2]. p Defining: ′ ψ (x) w(x)=− , (2) ψ(x) we have: ′′ ψ (x) w′(x)=− +w2(x). ψ(x) In other words, if ψ(x) satisfies Eq.(1), w(x) is a solution of the following Riccati equation: w′(x)=p2(x)+w2(x)=E−V(x)+w2(x). (3) 2 Every node of ψ(x) (necessarily a simple zero) corresponds to a simple pole of w(x) which decreases in the interval [x1,x2]. Consequently,whenxrunsthrough[x1,x2],ateachnodeofψ(x)thefunctionw(x)issubjecttoadiscontinuity from −∞ to +∞. If we define the phase θ(x) via: p(x) tanθ(x)=− , (4) w(x) we can write: p(x) θ(x)=arctan − +nπ, (5) w(x) (cid:18) (cid:19) where arctany ∈ [−π/2,π/2] is the principal determination of the reciprocal of the tangent function and where n increases of 1 at each node of w(x). Then, if x1 and x2 are not nodes of ψ(x): x2 ′ θ (x)dx=Nπ, (6) Zx1 N being the total number of nodes of w(x) on [x1,x2]. But using Eq.(3) we also have: ′ p′(x)w(x) θ (x)=p(x)− . (7) w′(x) Considering the nth bound state at energy E for which N =n+1, we obtain the following identity: n n x2,n p (x)dx=nπ+γ(E ), (8) n n Zx1,n where the correction γ(E ) to the lowest order WKB condition is given by the following integral: n x2,n w (x)p′ (x) γ(E )=π+ n n dx. (9) n w′(x) Zx1,n n Ma and Xu [1, 2] have observed that que for a large class of exactly solvable potentials, this integral correction is in fact independent of the considered energy level, and can be calculated from the ground state: γ(En)=γ(E0)=π+ x2,0 w0w(x′)(px′0)(x)dx. (10) Zx1,0 0 Inserting Eq.(10) in Eq.(8) gives the Ma-Xu formula: x2,n pn(x)dx=nπ+γ(E0). (11) Zx1,n The knowledge of the ground state characteristics permits then to calculate the γ(E0) correction and reaches to an implicit formula for the other energy levels E . The explicit calculations have been performed for many exactly n solvable examples [5–14]. Nevertheless the question of a direct link between the exact solvability of the potential and the validity of the above formula is still stayed open. As much as we know, all the closed form analytically solvable quantum mechanical systems belongs to the set of translationallyshapeinvariantpotentials(TSIP)inthesenseofSUSYquantummechanics[15]. Infactforthisset,the validity of the Ma-Xu formula follows from a result established in 1993 by Barclay [19]. In this article, he established 3 that for these potentials, the WKB series can be resummed beyond the lowest order giving an energy independent correction which can be absorbed in the Maslov index and written in a closed analytical form. He showed equally that this result is directly correlated to the exactness of the lowest order SWKB quantization condition [15, 16]. The starting point is the definition of two classes of potentials, each characterized by a specific change of variable which brings the potential into a quadratic form. This two classes are shown to coincide to the Barclay-Maxwell classes [18] which are based on a functional characterization of superpotentials and which cover the whole set of TSIP. In both cases, the action variable for an energy E is of the form: n x2,n I = p (x)dx=2 p (x)dx=2(nπ+γ), (12) n n n I En Zx1,n where γ is an energy-independent correction characteristic of the considered potential. As shown later by Bhaduri and al. [20], this results can be retrieved in an elegant way using POT [21]. The translational shape invariance of the potential is then a sufficient condition for the validity of the Ma-Xu prescription which reaches to Eq.(11). In the following, we propose an alternative way to recover Barclay’s result. It lies on a different characterization of the Barclay-Maxwell classes that we recently presented [22]. In both cases the action variable can be rewritten as a complex integral on an uniformization domain including only one branch-cut. Using standard tools of complex analysis, it can be readily calculated to recover the explicit form of γ for the whole set of TSIP. Note that analogous complex analysis techniques have been employed by Bhalla and al [24–26] in the frame of the quantum Hamilton-Jacobi (QHJ) formalism [28, 29] where, starting from the exact QHJ quantization condition, they have computed the spectrum of numerous potentials for which they have also showed the exactness of the lowest order SWKB quantization condition. Moreover, Cherqui et al [27] have recently shown that for algebraic and hyperbolic translationally shape invariant potentials, the shape invariance condition provides sufficient information on the singularity structure of the quantum momentum function to determine directly the energy spectrum of the system from the exact QHJ quantization condition. FIRST CATEGORY POTENTIALS Wesaythataonedimensionalpotentialisoffirstcategory[22]ifthereexistsachangeofvariablex→utransforming the potential into an harmonic one V(x) → V(u) = λ2u2 +λ1u+λ0, such that u(x) satisfy a constant coefficient Riccati equation: e e e du(x) =A0+A1u(x)+A2u2(x), (13) dx du/dx being of constant sign in all the range of values of x and u. The one dimensional harmonic oscillator correspond to the special case A1 = A2 = 0 and the Morse potential is generically associated to the case A1 6=0, A2 =0. Harmonic oscillator ItisperfectlywellknownthattheEBKquantizationconditionisexactfortheharmonicoscillatorimplyingthenthe exactnessoftheMa-XuformulawithaMaslovindexequalto1/2. Nevertheless,wewillexaminethiscasecompletely as a first example of use of the complex variable integration technique. The harmonic oscillator potential with zero ground state energy is: V(x,ω)=(ω/2)2x2−ω/2. (14) The classical turning points ±x0,n at energy En =nω satisfy: 4 1 x2 = n+ . (15) 0,n ω 2 (cid:18) (cid:19) 4 The half action variable for a classical periodic orbit of energy E is then: n x2,n ω x0,n I = p (x)dx= dx x2 −x2. (16) n Zx1,n n 2 Z−x0,n q 0,n The complex function fH(z) = x20,n−z2, defined on C−]−x0,n,x0,n[, admits only one isolated singularity at infinity. Then: q x0,n dxf (x)=πiRes(f (z),∞). (17) H H Z−x0,n TheasymptoticbehaviouratinfinityontheRiemannsheetforwhich u2 −z2 = x2 +y2 >0onthepositive 0,n 0,n half imaginary axe (z =iy, y >0) is given by: q q 1 x2 −z2 ∼ −iz 1− x20,n 2 ∼ −iz+ ix20,n1 +O( 1 ). (18) 0,n z→∞ z2 ! z→∞ 2 z z2 q Consequently: ix2 0,n Res(f (z),∞)=− (19) H 2 and: x2,n 1 p (x)dx= n+ π. (20) n 2 Zx1,n (cid:18) (cid:19) which is the expected result. Morse potential Asfortheharmonicoscillator,theEBKformulaisexactfortheMorsepotential. Thissecondexampleisnevertheless very instructive. The Morse potential with zero energy ground state is [15]: α V(x)=A2+B2e−2αx−2B A+ e−αx, α>0. (21) 2 (cid:16) (cid:17) Using the change of variable y =exp(−αx), dy =−αydx, it becomes: V(y)=B2(y−y0)2+V0 (22) with y0 =(A+α/2)/B and: α 2 V0 =A2−B2y02 = A+ −A2. (23) 2 (cid:16) (cid:17) This is a translationally shape invariant potential with a spectrum given by [22]: E =a2−a2. (24) n n 5 with a=A and a =A−kα. k As for the classical turning points, they are given by: y2,n =y0+u0,n (25) y1,n =y0−u0,n (cid:26) with: a2 u0,n = y02− Bn2. (26) r The half action variable for a classical periodic orbit of energy E is then: n x2,n 1 y2,n dy p (x)dx = − E −V(y,a) (27) n n α y Zx1,n Zy1,n B u0,n p = f (u)du, M α Z−u0,n where: u2 −u2 0,n f (u)= . (28) M qu+y0 The complex extended function fM(z) = u20,n−z2/(z+y0)T, defined on C−]−u0,n,u0,n[, admits two isolated singularities at infinity and in −y0. Then: q u0,n dufM(u)=πi(Res(fM(z),−y0)+Res(fM(z),∞)). (29) Z−u0,n With the same choice as before for the square root determination we have: y0 1 f (z) ∼ i −1+ +O( ) (30) M z→∞ z z2 (cid:18) (cid:19) which implies: a+α/2 Res(f (z),∞)=−i . (31) M B The residue in −y0 (which is a simple pole) is readily obtained as: a−nα Res(fM(z),−y0)=i . (32) B Then: x2,n 1 p (x)dx= n+ π. (33) n 2 Zx1,n (cid:18) (cid:19) which is the expected result. 6 Effective radial potential for the Kepler-Coulomb problem Consider the effective radial potential for the Kepler-Coulomb. It can be written as [22]: γ2 V(y)=l(l+1)y2−γy+ , k >0, (34) 4(l+1)2 where y =1/x and dy =−y2dx. We can rather use the equivalent form: 2 V(y)=l(l+1)(y−y0) +V0, (35) with y0 =γ/2l(l+1) and γ2 γ2 V0 = 4(l+1)2 − 4l(l+1). (36) . This is a translationally shape invariant potential characterized by the following energy spectrum [22]: γ2 γ2 E (a)= − . (37) n 4a2 4a2 n where a=l+1 and a =l+1+k,. k The classical turning points are given by: y2,n =y0+u0,n (38) y1,n =y0−u0,n (cid:26) with: 1 γ2 γ2 u0,n = l(l+1)s4l(l+1) − 4a2n. (39) The half action variable for a classical periodipc orbit of energy E is then: n x2,n y2,n dy p (x)dx = − E −V(y,a) (40) n n y2 Zx1,n Zy1,n p u0,n = − l(l+1) duf (u), K Z−u0,n p where: u2 −u2 0,n f (u)= . (41) K q(u+y0)2 Again, the complex function fK(z) = u20,n−z2/(z+y0)2, defined on C−]−u0,n,u0,n[, admits two isolated singularities at infinity and in −y0. Then: q u0,n dufK(u)=πi(Res(fK(z),−y0)+Res(fK(z),∞)). (42) Z−u0,n 7 With the same choice as before for the square root determination, we have: 1 −i u20,n 2 y0 −2 −i 1 f (z) ∼ 1− 1+ ∼ +O( ) (43) K z→∞ z z2 ! z z→∞ z z2 (cid:16) (cid:17) which implies: Res(f (z),∞)=i. (44) K The residue in −y0 which is a double pole is readily obtained as: a n Res(fK(z),−y0)=i . (45) l(l+1) Then: p x2,n p (x)dx=nπ+γ, (46) n Zx1,n where: γ =π l+1+ l(l+1) (cid:16) p (cid:17) which is the expected result. Other first category potentials If we except the two preceding examples, all the other first category potentials are such that there exists a change of variable x→y transforming the potential into an harmonic one: V(x)→V(y)=a(a∓α)y2+λ1y+λ0(a), withλ0(a)=λ21/4a2−αa(inordertohaveazeroenergygroundstate)andwherey(x)satisfiesaconstantcoefficient Riccati equation of the form [22]: dy =α±αy2(x)>0. (47) dx The potential can still be written [22]: 2 V±(y,a)=a(a∓α)(y−y0) +V0, (48) where: y0 =−λ1/a(a∓α), V0 =λ0(a)−a(a∓α)y02 (49) . V is a translationally shape invariant potential and its energy spectrum is then given by [22]: E (a)=φ (a)−φ (a ), (50) n 1,± 1,± n 8 where a =a±kα and: k 2 λ φ (a)=∓a2+ 1 . (51) 1,± 4a2 The classical turning points y for an energy E are determined by the condition: i,n n En =V(yi,n)⇔En =a(a∓α)u20,n+V0, (52) where: y2,n =y0+u0,n , (53) y1,n =y0−u0,n (cid:26) and: φ (a ) u0,n =s∓1+y02− a1(,a±∓αn). (54) Consider first the case where dy/dx=α+αy2(x). The half action variable for a classical periodic orbit of energy E is given by: n x2,n y2,n dy p (x)dx = E −V(y,a) (55) n n α(y−i)(y+i) Zx1,n Zy1,n p a(a−α) u0,n = f1,+(u)du, p α Z−u0,n where: u2 −z2 0,n f1,+(z)= (zq−z0)(z−z0) (56) with z0 =−y0+i. The complex extended integrand f1,+(z), defined on C−]−u0,n,u0,n[, admits three isolated singularities at infinity, z0 and z0. Then: u0,n duf1,+(u)=πi(Res(f1,+(z),z0)+Res(f1,+(z),z0)+Res(f1,+(z),∞)). (57) Z−u0,n When z →∞: i 1 f1,+(z)z→∼∞−z +O(z2), (58) which implies: Res(f1,+(z),∞)=i. (59) As for the residues in z0 and z0, using Eq.(52), Eq.(53), Eq.(54) and Eq.(50) we deduce: 1 λ1 Res(f1,+(z),z0)=− ian+ (60) 2 a(a−α)(cid:18) 2an(cid:19) p 9 and: 1 λ1 Res(f1,+(z),z0)=− −ian+ . (61) 2 a(a−α)(cid:18) 2an(cid:19) Then: p x2,n p (x)dx=nπ+γ (62) n Zx1,n with: πa α γ = 1− 1− . (63) α a (cid:18) r (cid:19) Consider now the case dy/dx=α−αy2(x). The half action variable for a classical periodic orbit of energy E is: n x2,n y2,n dy p (x)dx = E −V(y,a) (64) n n α(1−y)(1+y) Zx1,n Zy1,n p a(a−α) u0,n = − f1,−(u)du, p α Z−u0,n where: u2 −u2 0,n f1,−(u)= (uq−z1)(u−z2) (65) with z1 =−1−y0 and z2 =1−y0. The complex extended integrand f1,+(z), defined on C−]−u0,n,u0,n[, admits three isolated singularities at infinity, z1 and z2. Then: u0,n duf1,−(u)=πi(Res(f1,−(z),z1)+Res(f1,−(z),z2)+Res(f1,−(z),∞)). (66) Z−u0,n When z →∞: i 1 f1,+(z)z→∼∞−z +O(z2), (67) which implies: Res(f1,+(z),∞)=i. (68) As for the residues in z0 and z0, from Eq.(52), Eq.(91), Eq.(92) and Eq.(50) we deduce: i λ1 Res(f1,−(z),z2)= an+ (69) 2 a(a−α)(cid:18) 2an(cid:19) and: p i λ1 Res(f1,−(z),z1)= an− . (70) 2 a(a−α)(cid:18) 2an(cid:19) p 10 Note that in this case, we had to change the square root determination when we pass from z1 to z2, since we have either−1<y1,n <y2,n <1(wheny =tanh(αx+ϕ)),thatis−1−y0 <−u0,n <u0,n <1−y0,ory1,n <−1<1<y2,n (when y =coth(αx+ϕ)), that is −u0,n <−1−y0 <1−y0 <u0,n. Then: x2,n p (x)dx=nπ+γ (71) n Zx1,n with: πa α γ =− 1+ 1− . (72) α a (cid:18) r (cid:19) We recover the results obtained in [20] for the first class potentials. SECOND CATEGORY POTENTIALS Wesaythataonedimensionalpotentialisofsecondcategoryifthereexistsachangeofvariablex→utransforming the potential into an isotonic one V(x) → V(u) = λ2u2 +λ0 + µue22, such that u(x) satisfies a constant coefficients Riccati equation of the form: e e du(x) =A0+A2u2(x), (73) dx du(x)/dx being of constant sign in all the range of values of x and u. Isotonic oscillator The isotonic potential with a zero energy ground state (E0 =0) is [22, 30]: ω2 l(l+1) 3 V−(x)= x2+ −ω l+ , l>0. (74) 4 x2 2 (cid:18) (cid:19) Its energy spectrum is given by [22]: E (a)=2nω, (75) n where a= ω,l+1 and a = ω,l+1+k . 2 k 2 The classical turning points x for an energy E are determined by the condition: i,n n (cid:0) (cid:1) (cid:0) (cid:1) ω2 3 E =V(x )⇔ x4 −ω 2n+l+ x2 +l(l+1)=0, (76) n i,n 4 i,n 2 i,n (cid:18) (cid:19) that is: x22,n =u0,n+δn (77) (cid:26)x21,n =u0,n−δn with: 2n+l+ 3 2 3 2 u0,n =2 2, δn = 2n+l+ −l(l+1). (78) ω ωs 2 (cid:18) (cid:19)