Introduction
As a continuation of three previous papers [1 , 2 , 3 ] , where a detailed analysis of the Bohr-Sommerfeld (B-S) quantization condition for one-dimensional power-like potentials
V ( x ) = ∣ x ∣ m , m > 0 , V(x) = |x|^{m},\quad m > 0, V ( x ) = ∣ x ∣ m , m > 0 ,
and for the S S S -states of the d d d -dimensional radial power-like potentials
V ( r ) = m ∣ m ∣ r m , m > − 2 , and V ( r ) = log r , m = 0 , V(r) = \frac{m}{|m|}r^{m},\quad m > -2,\quad\text{and}\quad V(r) = \log r,\quad m = 0, V ( r ) = ∣ m ∣ m r m , m > − 2 , and V ( r ) = log r , m = 0 ,
was carried out, the states with non-zero angular momentum ℓ \ell ℓ in radial power-like potentials are studied in this work.
In the papers [1 , 3 ] , both absolute and relative deviations of the B-S energies E B S E_{BS} E B S from the exact ones E e x a c t E_{exact} E e x a c t were studied for the ground state, low-lying excited and the highly-excited S S S states. It was demonstrated that, even for the lowest-lying excited states, these energies are sufficiently close—not only for the highly excited states, as is usually proclaimed. Naturally, the maximal deviations occurred for the ground state energy case. Following the seminal paper by Kramers [4 ] , we also proposed a simple modification of the r.h.s. of the B-S quantization condition for radial confining potentials V ( r ) V(r) V ( r ) at d > 1 d > 1 d > 1 for the case of S S S -states (ℓ = 0 \ell= 0 ℓ = 0 ) by introducing the specific d d d -dependent Kramers-type correction d / 2 d/2 d /2 (which becomes the familiar 1 / 2 1/2 1/2 in the one-dimensional case [4 ] ), and additionally a certain correction term denoted as γ \gamma γ , defined through
1 ℏ ∫ r A r B 2 μ [ E e x a c t − V ( r ) ] d r = π ( n r + d 4 + γ ( n r , d ) ) , n r = 0 , 1 , 2 , … (1) \frac{1}{\hbar}\int_{r_A}^{r_B}\sqrt{2\mu\left[E_{exact}-V(r)\right]}\,\mathrm{d}r
=
\pi\left(n_r+\frac{d}{4}+\gamma(n_r,d)\right),\qquad n_r=0,1,2,\ldots \tag*{(1)} ℏ 1 ∫ r A r B 2 μ [ E e x a c t − V ( r ) ] d r = π ( n r + 4 d + γ ( n r , d ) ) , n r = 0 , 1 , 2 , … (1)
where r A , B r_{A,B} r A , B are the turning points, μ \mu μ is the mass of particle, d d d is spatial dimension and n r n_r n r is the radial quantum number. By choosing γ \gamma γ in such a way as to reproduce the exact energy spectrum, it turned out that for all studied radial confining potentials γ = γ ( n r , d ) \gamma=\gamma(n_r,d) γ = γ ( n r , d ) is a numerically-small, dimensionless, and easily numerically-calculable function; we called it the WKB correction . It was also checked that for various power-like potentials at d = 1 d=1 d = 1 this correction systematically ∣ γ ∣ ≤ 1 / 2 |\gamma| \le1/2 ∣ γ ∣ ≤ 1/2 ! Furthermore, it was shown in [2 ] also that γ \gamma γ is a function of the square of the shifted quantum number I = n + 1 / 2 I=n+1/2 I = n + 1/2 ; i.e., γ = γ ( I 2 ) \gamma=\gamma\left(I^2\right) γ = γ ( I 2 ) .
Evidently, for the harmonic oscillator in any dimension d d d , the WKB correction vanishes: γ = 0 \gamma= 0 γ = 0 . As for the square well potential with infinite walls it is a constant at d = 1 d = 1 d = 1 : it reaches its maximal value γ = 1 / 2 \gamma= 1/2 γ = 1/2 . It is not well-known but quite evident that for two exactly-solvable problems in d = 1 d = 1 d = 1 , Morse and Pöschl-Teller (V ( x ) = − λ ( λ + 1 ) sech 2 x V(x) = -\lambda(\lambda+ 1)\operatorname{sech}^{2}x V ( x ) = − λ ( λ + 1 ) sech 2 x ) potentials, the integral on the l.h.s. of (1) can be evaluated analytically, both γ \gamma γ are constants, which are equal to 0 0 0 and λ ( λ + 1 ) − λ − 1 / 2 \sqrt{\lambda(\lambda+ 1)} - \lambda- 1/2 λ ( λ + 1 ) − λ − 1/2 , respectively. As for the highly-excited states n r ≫ 1 n_r \gg1 n r ≫ 1 , γ \gamma γ tends to zero for any d d d , γ ( n r , d ) ∼ 1 / n r \gamma(n_r,d) \sim1/n_r γ ( n r , d ) ∼ 1/ n r as n r → ∞ n_r \to\infty n r → ∞ . The goal of the present paper is to generalize the quantization rule (1) to the states with non-zero angular momentum ℓ ≠ 0 \ell\ne0 ℓ = 0 in d d d -dimensional confining radial power-like potentials.
Needless to say, from the early days of quantum mechanics there were numerous attempts to generalize the B-S quantization condition (1) to the d d d -dimensional radial potentials V ( r ) V(r) V ( r ) . In this work, we propose a generalized, B-S quantization type condition of the form
1 ℏ ∫ r A r B E B S − V ( r ) d r = π ( n r + d + 2 ℓ 4 ) , n r , ℓ = 0 , 1 , 2 , … , (2) \frac{1}{\hbar}\int_{r_A}^{r_B}\sqrt{E_{BS}-V(r)}\,\mathrm{d}r = \pi\left(n_r+\frac{d+2\ell}{4}\right),\qquad n_r,\ell=0,1,2,\ldots, \tag*{(2)} ℏ 1 ∫ r A r B E B S − V ( r ) d r = π ( n r + 4 d + 2 ℓ ) , n r , ℓ = 0 , 1 , 2 , … , (2)
where r B > r A ≥ 0 r_B > r_A \ge0 r B > r A ≥ 0 are the radial turning points, V ( r A , r B ) = E B S V(r_A,r_B) = E_{BS} V ( r A , r B ) = E B S . 3 Here, n r n_r n r is the radial quantum number, it is equal to the number of radial nodes in the eigenfunction of the d d d -dimensional radial Schrödinger equation, namely,4
− ℏ 2 2 μ ( ψ ′ ′ ( r ) + d + 2 ℓ − 1 r ψ ′ ( r ) ) + V ( r ) ψ ( r ) = E ψ ( r ) , r ∈ [ 0 , ∞ ) , (3) -\frac{\hbar^2}{2\mu}\left(\psi^{\prime\prime}(r)+\frac{d+2\ell-1}{r}\psi'(r)\right)+V(r)\psi(r)=E\psi(r),\qquad r\in[0,\infty), \tag*{(3)} − 2 μ ℏ 2 ( ψ ′′ ( r ) + r d + 2 ℓ − 1 ψ ′ ( r ) ) + V ( r ) ψ ( r ) = E ψ ( r ) , r ∈ [ 0 , ∞ ) , (3)
where, for the sake of simplicity, we will set ℏ = 2 μ = 1 \hbar= 2\mu= 1 ℏ = 2 μ = 1 from now on. Here, d d d is the space dimension and ℓ \ell ℓ is the angular momentum. Unlike the standard quantization rules, see e.g. [5 ] for discussion, the quantization rule (2) contains the original potential V ( r ) V(r) V ( r ) as proposed in [6 ] , and not an effective potential which includes the centrifugal term. We also exclude the appearance of the Langer-type correction [7 ] to the original potential in the l.h.s. of (2) .
Note that for d = 2 d = 2 d = 2 , the quantization condition (2) was presented in [8 , 9 ] by using the so-called Langer transformation. For arbitrary d d d , a quantization condition somewhat similar to (2) was proposed in [10 ] , see Appendix therein, Eq. (A.3), but with a different potential-dependent r.h.s.: the quantization rule proposed there, was applied to singular at origin radial potentials with behavior V ( r ) ∼ r m V(r) \sim r^{m} V ( r ) ∼ r m at r → 0 r \to0 r → 0 for − 2 < m < 0 -2 < m < 0 − 2 < m < 0 but decaying exponentially at large distances r → ∞ r \to\infty r → ∞ , V ( r ) ∼ exp ( − a r ) V(r) \sim\exp(-ar) V ( r ) ∼ exp ( − a r ) . It was shown that this modification leads to reasonably accurate results. Following these works [6 , 8 , 9 , 10 ] , we propose that the quantization condition (2) holds for general confining, not necessarily singular at origin, radial power-like potentials for any angular momentum and in any dimension d d d .
There are three fundamental properties that the proposed quantization condition (2) fulfills:
(i) At d = 1 d = 1 d = 1 it coincides with the celebrated one-dimensional WKB quantization condition5 , if it is assumed an even potential, V ( x ) = V ( − x ) V(x) = V(-x) V ( x ) = V ( − x ) , with symmetric turning points x A = − x B x_A = -x_B x A = − x B ;
(ii) For the harmonic oscillator potential, V ( r ) = r 2 V(r) = r^{2} V ( r ) = r 2 , the condition (2) leads to the exact energies for any d d d with the correct multiplicity;
(iii) At large n r n_r n r but fixed d d d and ℓ \ell ℓ , the Bohr-Sommerfeld energies E B S E_{BS} E B S from (2) approach the exact ones.
Property (i) is evident, property (ii) can be easily checked by direct calculation, and property (iii) was already verified for ℓ = 0 \ell= 0 ℓ = 0 in [3 ] . Verification of (iii) for non-vanishing ℓ ≠ 0 \ell\ne0 ℓ = 0 requires extensive numerical calculations, which are carried out in this work.
It must be emphasized that for singular potentials the r.h.s. in (2) has to be modified:
1 ℏ ∫ r A r B E B S − V ( r ) d r = π ( n r + d + 2 ℓ − 1 2 ) , n r , ℓ = 0 , 1 , 2 , … , (4) \frac{1}{\hbar}\int_{r_A}^{r_B} \sqrt{E_{BS} - V(r)}\,\mathrm{d}r = \pi\left(n_r + \frac{d + 2\ell- 1}{2}\right), \qquad n_r,\ell= 0,1,2,\ldots, \tag*{(4)} ℏ 1 ∫ r A r B E B S − V ( r ) d r = π ( n r + 2 d + 2 ℓ − 1 ) , n r , ℓ = 0 , 1 , 2 , … , (4)
for d > 1 d > 1 d > 1 . This quantization condition will be applied to the potentials with singular behavior at origin
V → − r m , 0 > m > − 2 at r → 0. V \to-r^{m}, \quad0 > m > -2 \quad\text{at } r \to0. V → − r m , 0 > m > − 2 at r → 0.
Thus, we may replace condition (ii) by (iv).
Naturally, the WKB correction γ \gamma γ , see (1) , can be introduced into the r.h.s. of (2) ,
1 ℏ ∫ r A r B E − V ( r ) d r = π ( n r + d + 2 ℓ 4 + γ + ) (5) \frac{1}{\hbar}\int_{r_A}^{r_B} \sqrt{E-V(r)}\,\mathrm{d}r = \pi\left(n_r+\frac{d+2\ell}{4}+\gamma_+\right) \tag*{(5)} ℏ 1 ∫ r A r B E − V ( r ) d r = π ( n r + 4 d + 2 ℓ + γ + ) (5)
for non-singular potentials, or into the r.h.s. of (4) for singular ones,
1 ℏ ∫ r A r B E − V ( r ) d r = π ( n r + d + 2 ℓ − 1 2 + γ − ) (6) \frac{1}{\hbar}\int_{r_A}^{r_B} \sqrt{E-V(r)}\,\mathrm{d}r = \pi\left(n_r+\frac{d+2\ell-1}{2}+\gamma_-\right) \tag*{(6)} ℏ 1 ∫ r A r B E − V ( r ) d r = π ( n r + 2 d + 2 ℓ − 1 + γ − ) (6)
with the aim to find such a γ ± \gamma_\pm γ ± for which the E B S E_{BS} E B S energies coincide with the exact energies E e x a c t E_{\mathrm{exact}} E exact . It will be checked that in many occasions ∣ γ ± ∣ ≤ 1 / 2 |\gamma_\pm|\leq1/2 ∣ γ ± ∣ ≤ 1/2 . Certainly, the presence of γ ± \gamma_\pm γ ± in (5) or (6) does not break the fundamental properties (i)-(iv). It is worth noting that explicit indications that the quantization conditions can be different for non-singular and singular potentials were already presented in [5 ] .
The formulas (5) and (6) will be the main objects of study in the present paper. We will continue to call these formulas (5) and (6) the exact WKB (radial) quantization conditions . For the potential V = r 4 V=r^4 V = r 4 , it will be checked (and confirmed) that γ + \gamma_+ γ + in (6) depends on the square of the r.h.s.:
γ + = γ + ( ( n r + d + 2 ℓ 4 ) 2 ) . \gamma_+ = \gamma_+\left(\left(n_r+\frac{d+2\ell}{4}\right)^2\right). γ + = γ + ( ( n r + 4 d + 2 ℓ ) 2 ) .
Specifically, the general radial power-like potential of the form
V ( r ) = g m − 2 sign ( m ) r m , m > − 2 , sign ( m ) = m ∣ m ∣ (7) V(r)=g^{m-2}\operatorname{sign}(m)r^m,\quad m>-2,\quad\operatorname{sign}(m)=\frac{m}{|m|} \tag*{(7)} V ( r ) = g m − 2 sign ( m ) r m , m > − 2 , sign ( m ) = ∣ m ∣ m (7)
will be studied, where g g g is a coupling constant. Signum function is defined as
sign ( m ) = 1 for m ≥ 0 and sign ( m ) = − 1 for m < 0 , (8) \operatorname{sign}(m)=1\ \text{for}\ m\geq0\ \text{and}\ \operatorname{sign}(m)=-1\ \text{for}\ m<0, \tag*{(8)} sign ( m ) = 1 for m ≥ 0 and sign ( m ) = − 1 for m < 0 , (8)
which guarantees the existence of the discrete spectra in (7). Let us emphasize that the quantity
λ = ( ℏ 2 2 μ ) 1 / 4 g , (9) \lambda=\left(\frac{\hbar^2}{2\mu}\right)^{1/4}g, \tag*{(9)} λ = ( 2 μ ℏ 2 ) 1/4 g , (9)
is dimensionless and thus the dimensionless WKB correction γ \gamma γ can be a function of λ \lambda λ .6
For the potentials (7) with positive m > 0 m>0 m > 0 , the integral in the l.h.s. of (2) and (5) is reduced to the Euler Beta function B ( a , b ) B(a,b) B ( a , b ) and the energies can be written in closed analytic form. The case m ≤ 0 m \le0 m ≤ 0 requires special attention and it will be discussed in § 1 , in particular, for the cases of logarithmic and Coulomb potentials. We will follow the same program as in our previous studies [1 , 3 ] : first we will calculate and compare the energies emerging from the B-S quantization condition (2) with the ones obtained in highly-accurate numerical calculations of equation (3) , and then we calculate the absolute A.D. and relative R.D. deviations of the B-S from the exact energies. Second, we will calculate the WKB correction γ \gamma γ in (5) for different power-like potentials in different dimensions d d d .
The structure of this paper is the following. In § 0.2 , the quantization rules (2) and (5) are hinted. Section II is devoted to studying the analytical and numerical properties of E B S E_{BS} E B S for the non-singular power-like potentials (7) with m > 0 m > 0 m > 0 . The explicit calculations of the WKB correction γ \gamma γ for two physically relevant dimensions, d = 2 , 3 d = 2, 3 d = 2 , 3 are presented. In § 1 , we apply the quantization rule (2) to two singular potentials (7) with m ≤ 0 m \le0 m ≤ 0 : the Coulomb and logarithmic potentials, and two other singular potentials, ( − r − 1 / 2 ) (-r^{-1/2}) ( − r − 1/2 ) , ( − r − 3 / 2 ) (-r^{-3/2}) ( − r − 3/2 ) , and also funnel-type potential, V = − Z / r + g r V = -Z/r + gr V = − Z / r + g r , are mentioned briefly. We explore the accuracy of E B S E_{BS} E B S and compute the WKB correction γ ± \gamma_{\pm} γ ± . A modification of the quantization rule (2) , leading to the exact energy spectrum of the Coulomb problem in d d d dimension (6) , is proposed. Section IV is dedicated to the comparison of (2) with the formalism, where the original radial potential is modified by the Langer correction; the quartic oscillator potential r 4 r^4 r 4 is taken as the basic example. The results are summarized in Conclusions. Extensive numerical results are presented in the Appendix.
From S-states to =0 states
Let us write the original d d d -dimensional Schrödinger equation in the Euler coordinates ( r , Ω ) (r,\Omega) ( r , Ω ) , where r r r is the radius, r = x 1 2 + x 2 2 + ⋯ + x d 2 r = \sqrt{x_1^2 + x_2^2 + \cdots+ x_d^2} r = x 1 2 + x 2 2 + ⋯ + x d 2 , and { Ω } \{\Omega\} { Ω } is the set of d − 1 d - 1 d − 1 Euler angles. Separating out the angular variables, we arrive at the standard radial Schrödinger equation
− ( ψ ′ ′ ( r ) + d − 1 r ψ ′ ( r ) ) + ℓ ( ℓ + d − 2 ) r 2 ψ ( r ) + V ( r ) ψ ( r ) = E ψ ( r ) (10) -\left(\psi''(r) + \frac{d-1}{r}\psi'(r)\right) + \frac{\ell(\ell+d-2)}{r^2}\psi(r) + V(r)\psi(r) = E\psi(r) \tag*{(10)} − ( ψ ′′ ( r ) + r d − 1 ψ ′ ( r ) ) + r 2 ℓ ( ℓ + d − 2 ) ψ ( r ) + V ( r ) ψ ( r ) = E ψ ( r ) (10)
where ℓ = 0 , 1 , … \ell= 0, 1, \ldots ℓ = 0 , 1 , … . The centrifugal term (singular at r = 0 r = 0 r = 0 ) in the l.h.s. of (10) can be removed by taking the radial wave function in the form
ψ ( r ) = r ℓ R ( r ) , (11) \psi(r) = r^{\ell} R(r), \tag*{(11)} ψ ( r ) = r ℓ R ( r ) , (11)
which is equivalent to making a gauge rotation of the radial Hamiltonian operator7 with r ℓ r^{\ell} r ℓ as the gauge factor. As a result, we find that R ( r ) R(r) R ( r ) satisfies a radial Schrödinger-type equation,
− ( R ′ ′ ( r ) + d + 2 ℓ − 1 r R ′ ( r ) ) + V ( r ) R ( r ) = E R ( r ) . (12) -\left(R^{\prime\prime}(r) + \frac{d + 2\ell- 1}{r} R^{\prime}(r)\right) + V(r) R(r) = E R(r). \tag*{(12)} − ( R ′′ ( r ) + r d + 2 ℓ − 1 R ′ ( r ) ) + V ( r ) R ( r ) = E R ( r ) . (12)
At ℓ = 0 \ell= 0 ℓ = 0 , this equation becomes the one used as the basis for writing the B-S quantization condition (1) , see [3 ] , for the S S S states. This observation suggests that the formal replacement
d → d + 2 ℓ (13) d \to d + 2\ell \tag*{(13)} d → d + 2 ℓ (13)
in (1) should lead to the quantization rule for ℓ \ell ℓ -states (2) ,
∫ r A r B E B S − V ( r ) d r = π ( n r + d + 2 ℓ 4 ) . \int_{r_A}^{r_B} \sqrt{E_{BS} - V(r)}\,\mathrm{d}r = \pi\left(n_r + \frac{d + 2\ell}{4}\right). ∫ r A r B E B S − V ( r ) d r = π ( n r + 4 d + 2 ℓ ) .
One can easily check that the fundamental properties (i)-(ii) are satisfied. In most of the remaining part of the present work, we essentially focus on checking the correctness of the third property (iii) by using the same power-like potentials (7) which were explored in our previous work [3 ] . Naturally, the WKB correction γ \gamma γ can be introduced into (2) ,
∫ r A r B E e x a c t − V ( r ) d r = π ( n r + d + 2 ℓ 4 + γ + ) , \int_{r_A}^{r_B} \sqrt{E_{exact} - V(r)}\,\mathrm{d}r = \pi\left(n_r + \frac{d + 2\ell}{4} + \gamma_+\right), ∫ r A r B E e x a c t − V ( r ) d r = π ( n r + 4 d + 2 ℓ + γ + ) ,
see (5) , where, in general, γ + = γ + ( n r , ℓ , d ) \gamma_+ = \gamma_+(n_r,\ell,d) γ + = γ + ( n r , ℓ , d ) . We should emphasize that by generalizing the argument of H. Kramers [4 ] for one-dimensional potentials to the effective (radial) potential V ( r ) + V c e n t r i f u g a l V(r) + V_{centrifugal} V ( r ) + V ce n t r i f ug a l , one can show that the ( d + 2 ℓ ) / 4 (d + 2\ell)/4 ( d + 2 ℓ ) /4 term in r.h.s. of (2) comes from the linearization of the effective potential near the radial turning points r A , B r_{A,B} r A , B and asymptotic matching, while the corrections to the linear term (quadratic, cubic etc) may lead to the appearance of γ + \gamma_+ γ + .7
Non-Singular Power-Like Potentials
Let us consider the radial power-like potential
V ( r ) = r m , m > 0 , (14) V(r) = r^{m}, \qquad m > 0, \tag*{(14)} V ( r ) = r m , m > 0 , (14)
cf. (7) . Note that this potential (14) coincides with the Green function for ( 2 − m ) (2-m) ( 2 − m ) -dimensional Newtonian gravity, being the potential of a point-like source of unit mass (up to a normalization constant). The quantization condition (2) for potential (14) can be evaluated analytically, leading to B-S energies in closed form
E B S ( m , d ) = ( 2 M B ( 1 2 , M ) ( n r + d + 2 ℓ 4 ) ) 1 M , M = 1 m + 1 2 . (15) E_{BS}^{(m,d)} = \left(2M B\left(\frac{1}{2},M\right)\left(n_r+\frac{d+2\ell}{4}\right)\right)^{\frac{1}{M}}, \qquad M = \frac{1}{m}+\frac{1}{2}. \tag*{(15)} E B S ( m , d ) = ( 2 M B ( 2 1 , M ) ( n r + 4 d + 2 ℓ ) ) M 1 , M = m 1 + 2 1 . (15)
Here B ( a , b ) B(a,b) B ( a , b ) is the Euler beta function, see e.g. [11 ] . This generalizes the analytic expression for the spectrum of one-dimensional power-like potentials obtained in [1 ] . Let us emphasize that the E B S E_{BS} E B S energies may be easily analytically continued: from integer n r , ℓ , d n_r,\ell,d n r , ℓ , d to complex n r , ℓ , d n_r,\ell,d n r , ℓ , d , respectively. It is evident that E B S ( m = 2 , d ) E_{BS}^{(m=2,d)} E B S ( m = 2 , d ) (15) lead to the exact energy spectrum [12 ] of the radial harmonic oscillator m = 2 m=2 m = 2 for any d d d :
E B S ( m = 2 , d ) = 4 n r + d + 2 ℓ = E e x a c t ( m = 2 , d ) . (16) E_{BS}^{(m=2,d)} = 4n_r+d+2\ell= E_{\mathrm{exact}}^{(m=2,d)}. \tag*{(16)} E B S ( m = 2 , d ) = 4 n r + d + 2 ℓ = E exact ( m = 2 , d ) . (16)
As discussed in [1 , 3 ] , the m = ∞ m=\infty m = ∞ case provides an insight into the accuracy of the predicted B-S energies. In this limit, the power-like potential describes a radial square well of unit length defined in r ∈ [ 0 , 1 ] r\in[0,1] r ∈ [ 0 , 1 ] . In this case, the B-S energy displayed in (15) becomes
E B S ( m = ∞ , d ) = π 2 ( n r + d + 2 ℓ 4 ) 2 . (17) E_{BS}^{(m=\infty,d)} = \pi^2\left(n_r+\frac{d+2\ell}{4}\right)^2. \tag*{(17)} E B S ( m = ∞ , d ) = π 2 ( n r + 4 d + 2 ℓ ) 2 . (17)
On the other hand, for d > 1 d>1 d > 1 , the exact energies obey the transcendental equation
J d + 2 ℓ − 2 2 ( E e x a c t ( m = ∞ , d > 1 ) ) = 0 , (18) J_{\frac{d+2\ell-2}{2}}\left(\sqrt{E_{\mathrm{exact}}^{(m=\infty,d>1)}}\right)=0, \tag*{(18)} J 2 d + 2 ℓ − 2 ( E exact ( m = ∞ , d > 1 ) ) = 0 , (18)
where J ν J_\nu J ν is the Bessel function of the first kind of order ν \nu ν , see e.g. [11 ] . In general, equation (18) can be solved numerically in order to find the energy spectra: the only exception appears at d = 3 d = 3 d = 3 where at ℓ = 0 \ell= 0 ℓ = 0 it can be solved analytically [3 ] . Using the asymptotic expansion of the zeros of the Bessel functions; see, e.g., [11 ] , keeping d d d and ℓ \ell ℓ fixed, one can establish the asymptotic behavior of the exact energy at large n r n_r n r ,
E e x a c t ( m = ∞ , d ) = π 2 ( n r + d + 2 ℓ + 1 4 ) 2 + O ( n r − 1 ) . (19) E_{\mathrm{exact}}^{(m=\infty,d)} = \pi^2 \left(n_r + \frac{d + 2\ell+ 1}{4}\right)^2 + \mathcal{O}(n_r^{-1}) . \tag*{(19)} E exact ( m = ∞ , d ) = π 2 ( n r + 4 d + 2 ℓ + 1 ) 2 + O ( n r − 1 ) . (19)
cf. (17) . Consequently, by comparing (17) and (19) , we can immediately check that the condition (iii) is fulfilled for the square-well radial potential. Note that the agreement between (17) and (19) gets better for larger ( d + 2 ℓ ) (d + 2\ell) ( d + 2 ℓ ) . At large n r n_r n r the WKB correction
γ + = 1 / 4. \gamma_{+} = 1/4 . γ + = 1/4.
In order to find the accuracy of the BS energies for finite m m m , we need to compare them with the exact energies obtained via a solution of the radial Schrödinger equation, which nowadays can be solved numerically to any desired accuracy. Based on the experience gained in [1 , 3 ] , we chose the Lagrange Mesh Method for the numerical calculations [13 ] , which provided highly accurate results for the energies with minimal computational effort. In general terms, the Lagrange Mesh Method is an approximate variational approach based on the discretization of the time-independent Schrödinger equation onto a non-uniform lattice. Further details as well as generalities can be found in [14 ] .
Numerical Results
In this section, for angular quantum numbers 1 ≤ ℓ ≤ 5 1 \le\ell\le5 1 ≤ ℓ ≤ 5 , we will compare the E B S E_{BS} E B S energies with the exact ones E e x a c t E_{\mathrm{exact}} E exact obtained using the Lagrange Mesh Method. For the power-like potentials, calculations with 1000 mesh points were fast and sufficient to establish the energies of the first 100 low-lying states, at least, with 4 significant figures for given ℓ \ell ℓ , d d d , and m m m . This comparison is done explicitly for the linear, cubic, quartic, and sextic oscillators (m = 1 , 3 , 4 , 6 m = 1, 3, 4, 6 m = 1 , 3 , 4 , 6 , respectively) and for two physically important dimensions: d = 2 d = 2 d = 2 and d = 3 d = 3 d = 3 . The cases of the logarithmic (m = 0 m = 0 m = 0 ) and the Coulomb-type (m = − 1 m = -1 m = − 1 ) potentials will be considered separately below.
In Table V -Table IX , the results with d = 2 d = 2 d = 2 are presented, while Table X -Table XIV contain the results with d = 3 d = 3 d = 3 ; for both dimensions the orbital quantum numbers ℓ = 1 , 2 , 3 , 4 , 5 \ell= 1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 are studied. Those Tables display the E e x a c t E_{exact} E e x a c t and the E B S E_{BS} E B S energies for some radial quantum numbers in the interval 0 ≤ n r ≤ 20 0 \le n_r \le20 0 ≤ n r ≤ 20 . Additionally, we include the absolute (A.D.) and relative (R.D.) deviations, which are defined as follows
Table V. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 1 \ell= 1 ℓ = 1 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table IX. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 5 \ell= 5 ℓ = 5 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits .
Table X. Table X: Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 3 d = 3 d = 3 and ℓ = 1 \ell= 1 ℓ = 1 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table XIV. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 3 d = 3 d = 3 and ℓ = 5 \ell= 5 ℓ = 5 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_{+} γ + rounded to two significant digits.
A . D . = ∣ E B S − E e x a c t ∣ , R . D . = ∣ E B S − E e x a c t E e x a c t ∣ . (20) \mathrm{A.D.} = \lvert E_{BS} - E_{exact} \rvert,\qquad\mathrm{R.D.} = \left\lvert\frac{E_{BS} - E_{exact}}{E_{exact}} \right\rvert. \tag*{(20)} A.D. = ∣ E B S − E e x a c t ∣ , R.D. = E e x a c t E B S − E e x a c t . (20)
For all potentials we consider, both A.D. and R.D. deviations decrease with growing n r n_r n r , if the dimension d d d and the angular momentum ℓ \ell ℓ are kept fixed. For n r = 20 n_r = 20 n r = 20 , the value of the R.D. is ≲ 10 − 3 \lesssim10^{-3} ≲ 1 0 − 3 . This is a clear indication that condition (iii) is fulfilled. For the low-lying states of those potentials, the A.D. may take large values regardless of d d d , while the R.D. remains sufficiently small. For example, if we consider ℓ ≥ 2 \ell\ge2 ℓ ≥ 2 , then the absolute deviation varies in domain 1 ≲ A . D . ≲ 10 1 \lesssim\mathrm{A.D.} \lesssim10 1 ≲ A.D. ≲ 10 .
In Figure 1 , the relative deviation R.D. as a function of m m m is displayed for the ground state (n r = 0 n_r = 0 n r = 0 ) at d = 2 , 3 d = 2,3 d = 2 , 3 . Surprisingly, in the domain 1 ≤ m ≤ 40 1 \le m \le40 1 ≤ m ≤ 40 , the R.D. may develop two minima (see e.g. yellow curve). However, for sufficiently large m m m and ℓ > 3 \ell> 3 ℓ > 3 , the R.D. decreases monotonically, approaching a finite, nonzero asymptotic limit for both d = 2 d = 2 d = 2 and d = 3 d = 3 d = 3 .
Figure 1. Relative Deviation (R.D.) of the ground state energy, see (20) versus m m m for angular momenta ℓ = 1 , 2 , 3 , 4 , 5 \ell= 1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 in dimensions d = 2 , 3 d = 2,3 d = 2 , 3 . Bullets indicate actual calculations; dashed curves drawn for convenience. At m = 2 m = 2 m = 2 , marked by black bubble, the relative deviation is zero.
Exact Quantization Condition
In the case of radial power-like potentials (14), the exact WKB (radial) quantization condition (5) can be solved with respect to the energy, leading to a closed expression for the spectrum,
E e x a c t ( m > 0 , d ) = ( 2 M B ( 1 2 , M ) ( n r + d + 2 ℓ 4 + γ + ) ) 1 M , M = 1 m + 1 2 , (21) E_{\mathrm{exact}}^{(m>0,d)}=\left(2MB\left(\frac{1}{2},M\right)\left(n_{r}+\frac{d+2\ell}{4}+\gamma_{+}\right)\right)^{\frac{1}{M}},\qquad M=\frac{1}{m}+\frac{1}{2}, \tag*{(21)} E exact ( m > 0 , d ) = ( 2 M B ( 2 1 , M ) ( n r + 4 d + 2 ℓ + γ + ) ) M 1 , M = m 1 + 2 1 , (21)
for non-singular at origin potentials, m > 0 m>0 m > 0 . The practical value of this formula is related to the question of the magnitude of γ + \gamma_{+} γ + : if γ + \gamma_{+} γ + is relatively small with a negligible dependence on its parameters formula (21) can provide information on the dependence of the spectrum on d , n r , ℓ d,n_{r},\ell d , n r , ℓ . It is still unclear how to calculate γ + \gamma_{+} γ + analytically/theoretically. Thus, we will proceed to this question numerically.
WKB Correction
Once the exact energies are known, the WKB correction γ \gamma γ can be unambiguously found according to (5) or (21),
γ + = ( E e x a c t ) M − 1 2 M B ( 1 2 , M ) ( 4 n r + d + 2 ℓ ) 2 M B ( 1 2 , M ) . (22) \gamma_+ = \frac{(E_{\mathrm{exact}})^M-\frac{1}{2}MB\left(\frac{1}{2},M\right)(4n_r+d+2\ell)}{2MB\left(\frac{1}{2},M\right)}. \tag*{(22)} γ + = 2 M B ( 2 1 , M ) ( E exact ) M − 2 1 M B ( 2 1 , M ) ( 4 n r + d + 2 ℓ ) . (22)
The exact energies can be obtained via the Lagrange Mesh Method with practically any desired accuracy: we made accurate calculations of γ \gamma γ for the power-like potentials at m = 1 , 3 , 4 , 6 m=1,3,4,6 m = 1 , 3 , 4 , 6 . Results for d = 2 , 3 d=2,3 d = 2 , 3 at n r = 0 , 5 , 10 , 20 n_r=0,5,10,20 n r = 0 , 5 , 10 , 20 and ℓ = 1 − 5 \ell=1-5 ℓ = 1 − 5 are presented in Table V -Table XIV . In Figure 2 , the ground state WKB correction γ + \gamma_+ γ + is displayed as a function of m m m for ℓ = 0 , 1 , 2 , 3 , 4 , 5 \ell=0,1,2,3,4,5 ℓ = 0 , 1 , 2 , 3 , 4 , 5 , Figure 4 therein. According to the calculations, ∣ γ + ∣ |\gamma_+| ∣ γ + ∣ is always largest for the ground state n r = 0 n_r=0 n r = 0 , and then it begins to decrease at a slow pace as n r n_r n r increases. Again, this is an indication that condition (iii) is fulfilled. As occurred in the one-dimensional case and for the S S S -states of radial potentials, the WKB correction for the excited states γ \gamma γ is, as was guessed earlier, a smooth and small function (at least, in comparison with the quantum number n r n_r n r ). Furthermore, γ + \gamma_+ γ + decays like n r − 1 n_r^{-1} n r − 1 for large n r n_r n r .
Figure 2. WKB Correction γ + \gamma_{+} γ + for the ground state energy, see (22) versus m m m for the angular momenta ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 in dimensions d = 2 , 3 d=2,3 d = 2 , 3 . Bullets indicate actual calculations; dashed curves drawn for convenience. At m = 2 m=2 m = 2 , see black bubble, the WKB correction vanishes.
Figure 4. Logarithmic potential: Relative Deviation (R.D), see (20) , as a function of n r n_r n r for the angular momenta ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 and dimension d = 2 , 3 d=2,3 d = 2 , 3 . Bullets indicate actual calculations; dashed curves drawn for illustration.
In principle, one needs to know the whole spectrum of a given potential to find γ + \gamma_+ γ + for all values of n r n_r n r . Conversely, at present one needs to know γ + \gamma_+ γ + to find E e x a c t E_{\mathrm{exact}} E exact . At first glance, this may look as unattainable. However, in practice one can calculate the first low-lying energies and then use a fitting function to find γ + \gamma_+ γ + up to some accuracy. This program was successfully realized in our previous studies [1 , 3 ] . In particular, it was observed that fitting the numerical values of γ + \gamma_+ γ + is much easier than fitting the energy as a function of n r n_r n r directly. Below, we show that the same observation holds for the ℓ ≠ 0 \ell\ne0 ℓ = 0 states.
As illustration, let us consider the quartic oscillator with potential
V ( r ) = r 4 . (23) V(r)=r^4. \tag*{(23)} V ( r ) = r 4 . (23)
In Figure 3 , the WKB correction γ + \gamma_+ γ + vs n r n_r n r is shown in the domain 0 ≤ n r ≤ 100 0\le n_r\le100 0 ≤ n r ≤ 100 at d = 2 , 3 d=2,3 d = 2 , 3 for all values of ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 . Note that for d d d and ℓ \ell ℓ fixed, the WKB correction can be easily fitted by
Figure 3. WKB Correction γ + \gamma_+ γ + for the quartic potential V ( r ) = r 4 V(r)=r^4 V ( r ) = r 4 vs n r n_r n r for ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 and d = 2 , 3 d=2,3 d = 2 , 3 .
γ f i t ( n r ) = − a b n r + 1 . (24) \gamma_{\mathrm{fit}}(n_r)=-\frac{a}{bn_r+1}. \tag*{(24)} γ fit ( n r ) = − b n r + 1 a . (24)
with sufficiently high accuracy, where a a a and b b b are free parameters, which usually demonstrate a smooth dependence on d d d and ℓ \ell ℓ . They are found via least-squares optimization. The fitting function (24) is a simplified version of the function used in [3 ] to study the S S S -states, see Eqs. (20) –(21) therein. Note that at large n r n_r n r , the fit (24) leads to the correct asymptotic behavior ∼ n r − 1 \sim n_r^{-1} ∼ n r − 1 for γ + \gamma_+ γ + . In Table I , the optimal values of a , b a,b a , b are shown for the angular momentum ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 at d = 2 d=2 d = 2 and d = 3 d=3 d = 3 . With the optimal parameters from Table I , the fitting function γ f i t \gamma_{fit} γ f i t (24) predicts all the energies with around 4 exact significant digits if substituted into (21) . Needless to say that a similar fitting function can be used in studies of the WKB correction for other power-like potentials. If a higher accuracy is desired, the fitting function is modified
Table I. Quartic oscillator (m = 4 m = 4 m = 4 ). Parameters a a a and b b b for the fitting function (24) at d = 2 , 3 d = 2, 3 d = 2 , 3 and ℓ = 1 , … , 5 \ell= 1, \ldots, 5 ℓ = 1 , … , 5 .
γ f i t ( n r ) = P L ( n r ) Q L + 1 ( n r ) (25) \gamma_{fit}(n_r)=\frac{P_L(n_r)}{Q_{L+1}(n_r)} \tag*{(25)} γ f i t ( n r ) = Q L + 1 ( n r ) P L ( n r ) (25)
where P L P_L P L and Q L + 1 Q_{L+1} Q L + 1 are polynomials of degrees L L L and ( L + 1 ) (L+1) ( L + 1 ) , respectively.
Analyzing the data in Table V –Table XIV one can see that the WKB correction γ + \gamma_+ γ + is always a small function for power-like radial potentials (14) with m = 1 , 3 , 4 , 6 m=1,3,4,6 m = 1 , 3 , 4 , 6 ,
∣ γ + ∣ < 1 2 , |\gamma_+|<\frac{1}{2}, ∣ γ + ∣ < 2 1 ,
for d = 2 , 3 d=2,3 d = 2 , 3 and any n r ≥ 0 n_r\ge0 n r ≥ 0 and angular momentum ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 . This is the same behavior as in the one-dimensional case [1 ] for any positive degree m > 0 m>0 m > 0 and for the S S S -states for d = 2 , 3 , 6 d=2,3,6 d = 2 , 3 , 6 of radial power-like potentials [3 ] with m = 1 , 3 , 4 , 6 m=1,3,4,6 m = 1 , 3 , 4 , 6 . One can guess that for any non-singular power-like radial potential the WKB correction γ \gamma γ is small for any d ≥ 1 d \ge1 d ≥ 1 , any n r ≥ 0 n_r \ge0 n r ≥ 0 and any angular momentum ℓ \ell ℓ (for d ≥ 2 d \ge2 d ≥ 2 ), it obeys the inequality
∣ γ + ∣ ≤ 1 2 . (26) \lvert\gamma_+ \rvert\le\frac{1}{2}. \tag*{(26)} ∣ γ + ∣ ≤ 2 1 . (26)
Singular Power-Like Potentials
Coulomb Potential
If m = − 1 m = -1 m = − 1 in (14) we arrive at
V ( r ) = − 1 r , (27) V(r) = -\frac{1}{r}, \tag*{(27)} V ( r ) = − r 1 , (27)
which corresponds to the celebrated Coulomb potential at d = 3 d = 3 d = 3 , where r = x 2 + y 2 + z 2 r = \sqrt{x^2 + y^2 + z^2} r = x 2 + y 2 + z 2 . It is well known that for arbitrary dimension d > 1 d > 1 d > 1 , the energy spectrum for potential (27) can be found explicitly,
E e x a c t ( m = − 1 ) = − 1 ( 2 n r + d + 2 ℓ − 1 ) 2 . (28) E_{\mathrm{exact}}^{(m=-1)} = -\frac{1}{(2n_r + d + 2\ell- 1)^2}. \tag*{(28)} E exact ( m = − 1 ) = − ( 2 n r + d + 2 ℓ − 1 ) 2 1 . (28)
In particular, for d = 3 d = 3 d = 3 , the spectrum reads
E e x a c t ( m = − 1 ) ( d = 3 ) = − 1 4 n 2 , E_{\mathrm{exact}}^{(m=-1)}(d = 3) = -\frac{1}{4n^2}, E exact ( m = − 1 ) ( d = 3 ) = − 4 n 2 1 ,
see e.g. [16 ] , where n = n r + ℓ + 1 n = n_r + \ell+ 1 n = n r + ℓ + 1 , n = 1 , 2 , 3 , … n = 1, 2, 3, \ldots n = 1 , 2 , 3 , … is the principal quantum number. Note that for given d d d the degeneracy of the spectrum is equal to n n n , if the degeneracy with respect to magnetic quantum number(s) is not taken into account.
In turn, the quantization rule (2) results in
E B S = − 1 ( 2 n r + d 2 + ℓ ) 2 , (29) E_{BS} = -\frac{1}{\left(2n_r + \frac{d}{2} + \ell\right)^2}, \tag*{(29)} E B S = − ( 2 n r + 2 d + ℓ ) 2 1 , (29)
while the exact WKB quantization condition (5) leads to
E = − 1 ( 2 n r + d 2 + ℓ + 2 γ + ) 2 . (30) E = -\frac{1}{\left(2n_r + \frac{d}{2} + \ell+ 2\gamma_+\right)^2}. \tag*{(30)} E = − ( 2 n r + 2 d + ℓ + 2 γ + ) 2 1 . (30)
It is evident that the E B S E_{BS} E B S and E e x a c t E_{\mathrm{exact}} E exact are close asymptotically when n r n_r n r is large at fixed d , ℓ d, \ell d , ℓ . This ensures that (iii) is fulfilled. By comparing (28) , (29) and (30) one can find the value of the WKB correction,
γ + = 1 4 ( d + 2 ℓ − 2 ) , (31) \gamma_+ = \frac{1}{4}(d + 2\ell- 2), \tag*{(31)} γ + = 4 1 ( d + 2 ℓ − 2 ) , (31)
which has no n r n_r n r -dependence. Note that for fixed d d d , the WKB correction γ + \gamma_+ γ + grows linearly with the increase of ℓ \ell ℓ . Furthermore, this WKB correction vanishes for d = 2 d = 2 d = 2 and ℓ = 0 \ell= 0 ℓ = 0 only, implying that the energies E B S E_{BS} E B S coincide with the exact ones for the S S S -states in the planar problem [3 ] . The energies E e x a c t E_{\mathrm{exact}} E exact and E B S E_{BS} E B S , their A.D. and R.D. deviations, and the WKB correction γ + \gamma_+ γ + are shown in Table V –Table XIV .
It is worth noting that a modification of the B-S quantization condition (2) ,
1 ℏ ∫ r A r B E B S ( m o d ) − V ( r ) d r = π ( n r + d + 2 ℓ − 1 2 ) , n r = 0 , 1 , 2 , … , (32) \frac{1}{\hbar}\int_{r_A}^{r_B}\sqrt{E_{BS}^{(\mathrm{mod})} - V(r)}\,\mathrm{d}r = \pi\left(n_r + \frac{d + 2\ell- 1}{2}\right), \qquad n_r = 0, 1, 2, \ldots, \tag*{(32)} ℏ 1 ∫ r A r B E B S ( mod ) − V ( r ) d r = π ( n r + 2 d + 2 ℓ − 1 ) , n r = 0 , 1 , 2 , … , (32)
where r A = 0 r_A = 0 r A = 0 , leads to the exact B-S spectrum for the Coulomb potential (27) ,
E B S ( m o d ) = E e x a c t ( m = − 1 ) , E_{BS}^{(\mathrm{mod})} = E_{\mathrm{exact}}^{(m=-1)}, E B S ( mod ) = E exact ( m = − 1 ) ,
for any d d d and ℓ \ell ℓ . It suggests that a certain modification of (5) by introducing the WKB correction γ − \gamma_{-} γ − into (32) ,
1 ℏ ∫ r A r B E − V ( r ) d r = π ( n r + d + 2 ℓ − 1 2 + γ − ) , n r = 0 , 1 , 2 , … , (33) \frac{1}{\hbar}\int_{r_A}^{r_B}\sqrt{E-V(r)}\,\mathrm{d}r=\pi\left(n_r+\frac{d+2\ell-1}{2}+\gamma_{-}\right),\qquad n_r=0,1,2,\ldots, \tag*{(33)} ℏ 1 ∫ r A r B E − V ( r ) d r = π ( n r + 2 d + 2 ℓ − 1 + γ − ) , n r = 0 , 1 , 2 , … , (33)
c.f.(5), might be the right quantization rule to be used for singular at r = 0 r=0 r = 0 power-like potentials for which γ − \gamma_{-} γ − is small and vanishes for the Coulomb problem, γ − = 0 \gamma_{-}=0 γ − = 0 . This ensures that (iv) is fulfilled.
One can conjecture that (33) would lead to small γ − \gamma_{-} γ − for the general, singular at r = 0 r=0 r = 0 potentials. Taking the Yukawa potential,
V = − e − α r r , V=-\frac{e^{-\alpha r}}{r}, V = − r e − α r ,
as an example with α = 0.595306 \alpha=0.595306 α = 0.595306 , see e.g. [15 ] , for which the ground state energy is equal to zero, E = 0 E=0 E = 0 , one finds that
γ + = 0.284117 , γ − = 0.034117. \gamma_{+}=0.284117,\qquad\gamma_{-}=0.034117. γ + = 0.284117 , γ − = 0.034117.
This example supports the conjecture.
Exact Quantization Condition for Singular otentials
In the case of radial power-like potentials (14) for negative m < 0 m<0 m < 0 the exact WKB (radial) quantization condition (33) can be solved with respect to the energy, leading to a closed analytic expression for the spectrum,
E e x a c t ( m < 0 , d ) = − [ 2 π B ( 1 2 , − M ) ( n r + d + 2 ℓ − 1 2 + γ − ) ] 1 / M , M = 1 m + 1 2 , (34) E_{\mathrm{exact}}^{(m<0,d)}=-\left[\frac{2\pi}{B\left(\frac{1}{2},-M\right)}\left(n_r+\frac{d+2\ell-1}{2}+\gamma_{-}\right)\right]^{1/M},\qquad M=\frac{1}{m}+\frac{1}{2}, \tag*{(34)} E exact ( m < 0 , d ) = − [ B ( 2 1 , − M ) 2 π ( n r + 2 d + 2 ℓ − 1 + γ − ) ] 1/ M , M = m 1 + 2 1 , (34)
cf. (21) , for singular at origin potentials, m < 0 m<0 m < 0 . The practical value of this formula is related to the question of the magnitude of γ − \gamma_{-} γ − : assuming that γ − \gamma_{-} γ − is relatively small with a negligible dependence on its parameters formula (34) can provide information on the dependence of the spectrum on d , n r , ℓ d,n_r,\ell d , n r , ℓ . It is still unclear how to calculate γ − \gamma_{-} γ − analytically/theoretically.
Logarithmic Potential
The logarithmic potential appears as a limiting case of the power-like ones with positive degrees m > 0 m > 0 m > 0 by noting that
V ( r ) = log r = lim m → + 0 1 m ( r m − 1 ) . (35) V(r) = \log r = \lim_{m \to+0} \frac{1}{m}\left(r^m - 1\right). \tag*{(35)} V ( r ) = log r = m → + 0 lim m 1 ( r m − 1 ) . (35)
Hence, this potential can be considered formally as a non-singular potential!
For any dimension d d d at any n r n_r n r and ℓ \ell ℓ , the energy spectrum of the potential (35) can be found numerically with the Lagrange Mesh Method to any desired accuracy. On the other hand, the quantization rule (2) leads to a closed analytic expression for the B-S energies, which reads
E B S = log [ 2 π ( n r + d + 2 ℓ 4 ) ] , (36) E_{BS} = \log\left[2\sqrt{\pi}\left(n_r + \frac{d + 2\ell}{4}\right)\right], \tag*{(36)} E B S = log [ 2 π ( n r + 4 d + 2 ℓ ) ] , (36)
while the quantization rule (32) leads to the alternative B-S energies,
E B S ( m o d ) = log [ 2 π ( n r + d + 2 ℓ − 1 2 ) ] , (37) E_{BS}^{(mod)} = \log\left[2\sqrt{\pi}\left(n_r + \frac{d + 2\ell- 1}{2}\right)\right], \tag*{(37)} E B S ( m o d ) = log [ 2 π ( n r + 2 d + 2 ℓ − 1 ) ] , (37)
Hence, the formula for the exact energies has either form
E e x a c t = log [ 2 π ( n r + d + 2 ℓ 4 + γ + ) ] , (38) E_{exact} = \log\left[2\sqrt{\pi}\left(n_r + \frac{d + 2\ell}{4} + \gamma_+\right)\right], \tag*{(38)} E e x a c t = log [ 2 π ( n r + 4 d + 2 ℓ + γ + ) ] , (38)
where γ + \gamma_+ γ + is the WKB correction in (5) , or
E e x a c t = log [ 2 π ( n r + d + 2 ℓ − 1 2 + γ − ) ] , (39) E_{exact} = \log\left[2\sqrt{\pi}\left(n_r + \frac{d + 2\ell- 1}{2} + \gamma_-\right)\right], \tag*{(39)} E e x a c t = log [ 2 π ( n r + 2 d + 2 ℓ − 1 + γ − ) ] , (39)
with γ − \gamma_- γ − as a new WKB correction in (33) .
Before making a detailed study of B-S quantization rule let us make a check what is the form of the r.h.s. is more relevant for the logarithmic potential: the one which appears in (5) or in (33) . As a test we consider the ground state ( n r = 0 , ℓ = 0 ) (n_r = 0, \ell= 0) ( n r = 0 , ℓ = 0 ) in dimension d = 3 d = 3 d = 3 . In the Lagrange Mesh method this ground state energy is easily calculated,
E ( 0 , 0 ) = 1.044332 , E_{(0,0)} = 1.044332, E ( 0 , 0 ) = 1.044332 ,
see [3 ] . It immediately leads to the WKB corrections,
γ + ( 0 , 0 ) = 0.051573 , γ − ( 0 , 0 ) = − 0.198428 , \gamma_{+}(0,0) = 0.051573 \,,\quad\gamma_{-}(0,0) = -0.198428 \,, γ + ( 0 , 0 ) = 0.051573 , γ − ( 0 , 0 ) = − 0.198428 ,
in (5) and (33) , respectively. Smallness of γ + \gamma_{+} γ + confirms the guess that the logarithmic potential can be treated by using the quantization condition (5) relevant for the non-singular potentials.
The quantitative comparison between the exact E e x a c t E_{\mathrm{exact}} E exact and the B-S low-lying E B S E_{\mathrm{BS}} E BS energies is found in Table V -Table XIV for dimensions d = 2 , 3 d=2,3 d = 2 , 3 and angular momenta ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 . According to the calculations, the maximum deviation (∼ 10 % \sim10\% ∼ 10% ) between E e x a c t E_{\mathrm{exact}} E exact and E B S E_{\mathrm{BS}} E BS occurs for the ground state similarly to all other studied power-like potentials. Then, as n r n_r n r increases the R.D. drops rapidly for both d = 2 , 3 d=2,3 d = 2 , 3 cases, see Figure 4 .
The WKB correction γ + \gamma_{+} γ + for d = 2 , 3 d=2,3 d = 2 , 3 and ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 demonstrates a smooth decreasing behavior with the increase of n r n_r n r , while for fixed n r n_r n r it grows with the increase of ℓ \ell ℓ , see Figure 5 . In order to reach four significant digits in energy, the WKB correction γ \gamma γ can be fitted by a rational function of the type (25) ,
Figure 5. Logarithmic potential: WKB Correction γ + \gamma_{+} γ + , see (5) , as a function of n r n_r n r for the angular momenta ℓ = 1 , 2 , 3 , 4 , 5 \ell=1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 and dimension d = 2 , 3 d=2,3 d = 2 , 3 . Bullets indicate actual calculations; dashed curves drawn for illustration, see (40) .
γ f i t log ( r ) ( n r ) = a 1 n r + a 0 b 2 n r 2 + b 1 n r + 1 , (40) \gamma_{\mathrm{fit}}^{\log(r)}(n_r)=\frac{a_1 n_r+a_0}{b_2 n_r^2+b_1 n_r+1}, \tag*{(40)} γ fit l o g ( r ) ( n r ) = b 2 n r 2 + b 1 n r + 1 a 1 n r + a 0 , (40)
cf. (24) , where the optimal parameters { a 0 , a 1 , b 1 , b 2 } \{a_0,a_1,b_1,b_2\} { a 0 , a 1 , b 1 , b 2 } are shown in tbl:2 . This fit is more complicated than (25) , which we used in order to get four significant digits correctly in the energy of the quartic oscillator, since it involves two extra parameters. However, it is less complicated than the one used for the case ℓ = 0 \ell=0 ℓ = 0 in [3 ] . Note that for fixed d d d , all parameters are monotonic functions in ℓ \ell ℓ : a 0 , 1 a_{0,1} a 0 , 1 are growing, while b 1 , 2 b_{1,2} b 1 , 2 are decreasing. It is worth emphasizing that for the fixed ℓ \ell ℓ the maximal value of the γ + \gamma_{+} γ + corresponds to n r = 0 n_r=0 n r = 0 , it is given by a 0 a_0 a 0 . It grows with increasing ℓ \ell ℓ while being always much smaller than ℓ / 2 \ell/2 ℓ /2 , see (4) .8
Two singular potentials: (− r − 1 / 2 -r^{-1/2} − r − 1/2 ) and (− r − 3 / 2 -r^{-3/2} − r − 3/2 )
(i) Let us take the singular potential
V ( r ) = − r − 1 / 2 . (41) V(r)=-r^{-1/2}. \tag*{(41)} V ( r ) = − r − 1/2 . (41)
This potential has an infinite discrete spectrum. In the Lagrange Mesh method without any difficulties one can calculate the ground state energy at d = 3 d=3 d = 3 ,
E ( 0 , 0 ) = − 0.438041 , E_{(0,0)}=-0.438041, E ( 0 , 0 ) = − 0.438041 ,
Table II. Logarithmic Potential. Parameters a a a ’s and b b b ’s for fitting function (40) at d = 2 , 3 d=2,3 d = 2 , 3 and ℓ = 1 , … , 5 \ell=1,\ldots,5 ℓ = 1 , … , 5 .
which corresponds to
γ + ( 0 , 0 ) = 0.112318 , γ − ( 0 , 0 ) = − 0.137682. \gamma_{+}(0,0)=0.112318,\quad\gamma_{-}(0,0)=-0.137682. γ + ( 0 , 0 ) = 0.112318 , γ − ( 0 , 0 ) = − 0.137682.
(ii) The same study as in (i) can be carried out for the singular potential
V ( r ) = − r − 3 / 2 . (42) V(r)=-r^{-3/2}. \tag*{(42)} V ( r ) = − r − 3/2 . (42)
In the Lagrange Mesh method one can calculate the ground state energy at d = 3 d=3 d = 3 ,
E ( 0 , 0 ) = − 0.2986303 , E_{(0,0)}=-0.2986303, E ( 0 , 0 ) = − 0.2986303 ,
which corresponds to
γ + ( 0 , 0 ) = 0.668352 , γ − ( 0 , 0 ) = 0.418352. \gamma_{+}(0,0)=0.668352,\quad\gamma_{-}(0,0)=0.418352. γ + ( 0 , 0 ) = 0.668352 , γ − ( 0 , 0 ) = 0.418352.
Funnel-type potential
As a further test of the proposed quantization conditions, we consider the three-dimensional funnel-type potential
V ( r ) = − Z r + g r , Z , g > 0 , (43) V(r) = -\frac{Z}{r} + gr,\qquad Z,g>0, \tag*{(43)} V ( r ) = − r Z + g r , Z , g > 0 , (43)
which combines a term with Coulomb singularity at the origin with linear, confinement-type term at large distances. Unlike the potentials discussed earlier, this potential involves two competing terms, singular and non-singular at origin, and therefore provides a useful test ground for checking the accuracy for quantization conditions (2) and (32) .9 We use the parameters Z = 0.68812 Z=0.68812 Z = 0.68812 and g = 1 / 2 g=1/2 g = 1/2 , which were found to be suitable for the description of the bound states of charmonium - the quark-antiquark system made from charm quarks c , c ˉ c,\bar{c} c , c ˉ , see e.g. [17 ] .
Numerical results are presented in Table III for selected radial quantum numbers n r n_r n r and angular momenta ℓ = 0 , 5 \ell=0,5 ℓ = 0 , 5 . Note that the quantization rule (2) provides the B-S energies, which systematically underestimate the exact energies being always below them. This contrasts with the modified quantization rule (32) which systematically overestimates the exact energies being above them. Nevertheless, for both angular momenta, its relative deviation decreases with increasing n r n_r n r , with the largest deviation occurring for the ground state as expected.
Table III. Funnel-type potential (43). Comparison of exact energies with Bohr–Sommerfeld (2) and modified Bohr–Sommerfeld (32) predictions for selected radial states with ℓ = 0 \ell= 0 ℓ = 0 and ℓ = 5 \ell= 5 ℓ = 5 . The absolute (A.D.) and relative (R.D.) deviations are calculated with respect to the corresponding Bohr–Sommerfeld predictions.
The accuracy of the two quantization rules can also be assessed in terms of the WKB corrections γ ± \gamma_{\pm} γ ± . For the funnel-type potential (43) , the original quantization rule (2) generally provides more accurate energies, particularly for ℓ = 5 \ell=5 ℓ = 5 , where it also yields substantially small WKB corrections γ + \gamma_{+} γ + in magnitude; see Table III . In spite of the fact, the modified quantization rule (32) performs better for the lowest radial excited states for ℓ = 0 \ell=0 ℓ = 0 , including the ground state, and it leads to very bad behavior of γ − \gamma_{-} γ − for ℓ = 5 \ell=5 ℓ = 5 . For both angular momenta γ − \gamma_{-} γ − grows with the increase of n r n_r n r , which can be considered as an undesirable feature. The overall better performance of the non-modified quantization rule (2) is consistent with the dominance of the non-singular linear term when we find the bound-state energies for the chosen parameters g g g and Z Z Z , see above. However, ∣ γ + ∣ |\gamma_{+}| ∣ γ + ∣ gets above 1 / 2 1/2 1/2 for ℓ = 5 \ell=5 ℓ = 5 which we set as undesirable property for power-like potentials.
Langer Approach vs Exact B-S Quantization Rule
Let us compare the accuracy of the B-S quantization rule (2) with the one emerging from the well-known quantization rule based on the introduction of the Langer correction [7 ] into the original potential in the standard B-S quantization rule. At d = 3 d = 3 d = 3 , this Bohr-Sommerfeld quantization rule with Langer correction inserted reads10
∫ r A r B E B S ( L ) − V ( r ) − ( ℓ + 1 2 ) 2 r 2 d r = π ( n r + 1 2 ) , n r , ℓ = 0 , 1 , 2 , … (44) \int_{r_A}^{r_B} \sqrt{E_{\mathrm{BS}}^{(L)} - V(r) - \frac{\left(\ell+ \frac{1}{2}\right)^2}{r^2}}\,\mathrm{d}r = \pi\left(n_r + \frac{1}{2}\right), \qquad n_r,\ell= 0,1,2,\ldots \tag*{(44)} ∫ r A r B E BS ( L ) − V ( r ) − r 2 ( ℓ + 2 1 ) 2 d r = π ( n r + 2 1 ) , n r , ℓ = 0 , 1 , 2 , … (44)
where we set ℏ = 1 \hbar= 1 ℏ = 1 . The superscript ( L ) (L) ( L ) stands for the energy with the Langer correction is taken into account, E B S ( L ) E_{BS}^{(L)} E B S ( L ) . This was introduced to distinguish from E B S E_{BS} E B S already used in (2) . For both the Coulomb and the harmonic oscillator potentials, the quantization rule (44) leads to the exact energy spectrum.11
We performed a numeric analysis for the quartic oscillator potential V ( r ) = r 4 V(r) = r^4 V ( r ) = r 4 . In Table IV the energies coming from the quantization rules (2) and (44) are compared with the exact spectrum found in the Lagrange Mesh Method for ℓ = 0 , 1 , 5 \ell= 0, 1, 5 ℓ = 0 , 1 , 5 and different n r = 0 , 5 , 10 , 20 , 100 n_r = 0, 5, 10, 20, 100 n r = 0 , 5 , 10 , 20 , 100 . From those results, we observe that the energy E B S ( L ) E_{BS}^{(L)} E B S ( L ) systematically provides a lower bound for the exact energy. As can be seen, in general, E B S ( L ) E_{BS}^{(L)} E B S ( L ) and E B S E_{BS} E B S are sufficiently close to the exact energy E e x a c t E_{exact} E e x a c t . For the S S S -states E B S E_{BS} E B S is closer to E e x a c t E_{exact} E e x a c t than E B S ( L ) E_{BS}^{(L)} E B S ( L ) while for ℓ > 0 \ell> 0 ℓ > 0 the situation is reversed: E B S ( L ) E_{BS}^{(L)} E B S ( L ) is closer to E e x a c t E_{exact} E e x a c t than E B S E_{BS} E B S .
Table 4. Quartic Oscillator: comparison between the quantization rules (44), (2) and the exact energies at ℓ = 0 , 1 , 5 \ell= 0, 1, 5 ℓ = 0 , 1 , 5 and different radial quantum numbers n r n_r n r .
Conclusions
A new Bohr-Sommerfeld type quantization rule for the d d d -dimensional radial potentials (2) is proposed for the case of ℓ \ell ℓ -states - the states with arbitrary angular momentum.
The four basic requirements behind this quantization condition are fulfilled:
(i) it should coincide with the one-dimensional standard WKB quantization condition at d = 1 d = 1 d = 1 (and at ℓ = 0 \ell= 0 ℓ = 0 ), see [1 ] ;
(ii) the B-S spectrum of the d d d -dimensional harmonic oscillator should be exact. Therefore, it should coincide with the d d d -dimensional WKB quantization condition for the S S S -states at ℓ = 0 \ell= 0 ℓ = 0 , proposed in [3 ] ;
(iii) at large radial quantum numbers n r n_r n r the B-S spectrum should approach the exact one.
(iv) For singular potentials, condition (ii) should be replaced by the requirement that the B-S spectrum of the d d d -dimensional Coulomb problem be exact.
Explicit calculations for radial power-like potentials, including the Coulomb and logarithmic potentials, prove the correctness of the proposal. In particular, at d = 2 d = 2 d = 2 , the original B-S quantization rule reproduces the exact Coulomb energies for the S S S states (ℓ = 0 \ell= 0 ℓ = 0 ). In order to obtain the eigenvalues accurately, the radial Schrödinger equation (12) was solved using the Lagrange Mesh method with 100, 500, 1000 mesh points. This allows us to check for convergence and to reach the extremely high accuracies. In particular, for the linear potential with 100 (optimized) mesh points the energies of the first 37 low-lying S S S states can be established with at least 4 exact d.d. while with 500 optimized mesh points, the energies of the first 100 low-lying S S S states can be established with exact 140 d.d., at least.
The notion of the WKB correction, introduced for the d = 1 d = 1 d = 1 case [1 ] and then extended to the S S S states of the d d d -dimensional radial case in [3 ] , is explored. For power-like radial potentials, V ( r ) = r m V(r) = r^m V ( r ) = r m with fixed m > 0 m > 0 m > 0 and quantum number n r n_r n r , the WKB correction γ \gamma γ , see (5) , is small for all ℓ = 0 , 1 , 2 , 3 , 4 , 5 \ell= 0, 1, 2, 3, 4, 5 ℓ = 0 , 1 , 2 , 3 , 4 , 5 . For fixed d d d , m > 0 m > 0 m > 0 and small ℓ \ell ℓ , the magnitude of the WKB correction is a monotonically decreasing function of n r n_r n r . For fixed d d d , n r n_r n r and small ℓ \ell ℓ , the WKB correction varies smoothly as m > 2 m > 2 m > 2 grows. In general, the WKB correction γ \gamma γ is a smooth, dimensionless function of the parameters ( λ ; d , m , n r ) (\lambda; d, m, n_r) ( λ ; d , m , n r ) : for fixed ( λ ; d , m ) (\lambda; d, m) ( λ ; d , m ) , where
λ = ( ℏ 2 2 μ ) 1 / 4 g \lambda= \left(\frac{\hbar^2}{2\mu}\right)^{1/4} g λ = ( 2 μ ℏ 2 ) 1/4 g , see (9) , this function can easily be fitted by the ratio of a polynomial to the square-root of another polynomial 12 of the argument n r n_r n r . In all studied cases at m > 0 m > 0 m > 0 and ℓ = 1 , 2 , 3 , 4 , 5 \ell= 1,2,3,4,5 ℓ = 1 , 2 , 3 , 4 , 5 the WKB correction
∣ γ + ∣ < 1 / 2. |\gamma_+| < 1/2 . ∣ γ + ∣ < 1/2.
Replacing the condition (ii): the B-S spectrum of the d d d -dimensional harmonic oscillator should be exact by different one: the B-S spectrum of the d d d -dimensional Coulomb potential should be exact , it leads to a modified r.h.s. of the B-S quantization condition,
( n r + d + 2 ℓ 4 ) → ( n r + d + 2 ℓ − 1 2 ) \left(n_r+\frac{d+2\ell}{4}\right)\to\left(n_r+\frac{d+2\ell-1}{2}\right) ( n r + 4 d + 2 ℓ ) → ( n r + 2 d + 2 ℓ − 1 )
see (5) , (33) . Both expressions can be unified into a single function,
K = n r + ( 3 d + 6 ℓ − 2 8 ) − sign ( m ) ( d + 2 ℓ − 2 8 ) K=n_r+\left(\frac{3d+6\ell-2}{8}\right)-\operatorname{sign}(m)\left(\frac{d+2\ell-2}{8}\right) K = n r + ( 8 3 d + 6 ℓ − 2 ) − sign ( m ) ( 8 d + 2 ℓ − 2 )
where the signum function sign ( 2 ) = 1 \operatorname{sign}(2)=1 sign ( 2 ) = 1 (the harmonic oscillator case) and sign ( − 1 ) = − 1 \operatorname{sign}(-1)=-1 sign ( − 1 ) = − 1 (the Coulomb potential case). It results into a natural quantization condition
1 ℏ ∫ r A r B E − V ( r ) d r = π ( K + γ ) , n r = 0 , 1 , 2 , … , (45) \frac{1}{\hbar}\int_{r_A}^{r_B}\sqrt{E-V(r)}\,\mathrm{d}r=\pi(K+\gamma),\qquad n_r=0,1,2,\ldots, \tag*{(45)} ℏ 1 ∫ r A r B E − V ( r ) d r = π ( K + γ ) , n r = 0 , 1 , 2 , … , (45)
which coincides with (5) for non-singular at r = 0 r=0 r = 0 potentials, but including the logarithmic potential, with behavior V → r m V\to r^m V → r m at r → 0 r\to0 r → 0 for m ≥ 0 m\geq0 m ≥ 0 , when sign ( m ) = 1 \operatorname{sign}(m)=1 sign ( m ) = 1 . In turn, for singular potentials with m < 0 m<0 m < 0 , it coincides with (33) , when sign ( m ) = − 1 \operatorname{sign}(m)=-1 sign ( m ) = − 1 . We conjecture that for both cases of non-singular and singular potentials, the WKB correction
∣ γ ± ∣ ≤ 1 / 2. |\gamma_{\pm}|\leq1/2 . ∣ γ ± ∣ ≤ 1/2.
In this article the conjecture was confirmed for some selected power-like potentials 13 .
This article concludes the studies of the power-like potentials using the (exact) Bohr-Sommerfeld quantization condition (2) . A natural question to ask is what would happen if the harmonic oscillator potential is added to the power-like potential. It was shown in [3 ] that for the d d d -dimensional quartic radial anharmonic oscillator,
V = r 2 + g 2 r 4 V = r^2 + g^2 r^4 V = r 2 + g 2 r 4
for d = 1 , 2 , 3 , 6 d = 1, 2, 3, 6 d = 1 , 2 , 3 , 6 and S S S states the WKB correction γ g \gamma_g γ g is very small for any g 2 ≥ 0 g^2 \ge0 g 2 ≥ 0 and always smaller than γ 4 \gamma_4 γ 4 for the corresponding quartic potential,
∣ γ g ∣ < ∣ γ 4 ∣ . |\gamma_g| < |\gamma_4| . ∣ γ g ∣ < ∣ γ 4 ∣.
Evidently, this property will remain for ℓ \ell ℓ -states with ℓ > 0 \ell> 0 ℓ > 0 . Taking another example at d = 1 d = 1 d = 1 of sextic SUSY partners,
V ± = x 6 ± 3 x 2 , V_{\pm} = x^6 \pm3x^2 , V ± = x 6 ± 3 x 2 ,
with superpotential w = x 3 w = x^3 w = x 3 , one can see that the WKB correction γ \gamma γ remains very small for both the bosonic potential V + V_+ V + and the fermionic potential V − V_- V − , see Figure 6 . Note that
Figure 6. WKB correction for QES potential x 6 − 3 x 2 x^6 - 3x^2 x 6 − 3 x 2 and its SUSY partner.
V + = x 6 − 3 x 2 V_+ = x^6 - 3x^2 V + = x 6 − 3 x 2 is quasi-exactly-solvable: its ground state (at N = 0 N = 0 N = 0 ) is known exactly Ψ 0 = e − x 4 / 4 \Psi_0 = e^{-x^4/4} Ψ 0 = e − x 4 /4 , E 0 = 0 E_0 = 0 E 0 = 0 . Its WKB correction can be found analytically, γ V + ( N = 0 ) = 1 / 4 \gamma_{V_+}(N = 0) = 1/4 γ V + ( N = 0 ) = 1/4 : it is larger than γ V = x 6 ( N = 0 ) ∼ 0.15 \gamma_{V=x^6}(N = 0) \sim0.15 γ V = x 6 ( N = 0 ) ∼ 0.15 . In turn, γ V − ( N = 0 ) ∼ 0.04 \gamma_{V_-}(N = 0) \sim0.04 γ V − ( N = 0 ) ∼ 0.04 and it is smaller than γ V = x 6 ( N = 0 ) \gamma_{V=x^6}(N = 0) γ V = x 6 ( N = 0 ) .
In a recent development of the theory of potential curves of diatomic molecules (dimers) see e.g. [18 ] a new class of rational potentials was proposed,
V = − 1 r P n ( r ) Q n + 5 ( r ) , (46) V = -\frac{1}{r}\frac{P_n(r)}{Q_{n+5}(r)} , \tag*{(46)} V = − r 1 Q n + 5 ( r ) P n ( r ) , (46)
which are defined in 3-dimensional space, for neutral molecules. Here P n P_n P n and Q n + 5 > 0 Q_{n+5} > 0 Q n + 5 > 0 at r > 0 r > 0 r > 0 are polynomials of degrees n n n and ( n + 5 ) (n + 5) ( n + 5 ) , respectively. The physical meaning of these potentials is related to a modelling of the screening of the Coulomb interaction of nuclei by the electronic cloud. The value of the WKB correction is of great importance. It will be studied elsewhere.
Acknowledgments
J.C.dV. acknowledges partial financial support from the SONATA BIS 10 Grant No. 2019/34/E/ST1/00390 (Poland) during the initial stages of this work and from UNAM through PAPIIT project IA100926. A.V.T. was partially supported by CONACyT Grant A1-S-17364 and DGAPA grants IN113022 (in early stage of work), and IN104125 (Mexico). A.V.T. expresses his gratitude to I.H.E.S. (France), where this work was completed.
References
Numerical Results
In this Appendix selected numerical results supporting the conclusions of the work are presented.
Table VI: Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 2 \ell= 2 ℓ = 2 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table VI. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 2 \ell= 2 ℓ = 2 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table VII: Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 3 \ell= 3 ℓ = 3 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table VII. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 3 \ell= 3 ℓ = 3 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table VIII: Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 4 \ell= 4 ℓ = 4 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table VIII. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 2 d = 2 d = 2 and ℓ = 4 \ell= 4 ℓ = 4 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table XI. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 3 d = 3 d = 3 and ℓ = 2 \ell= 2 ℓ = 2 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_{+} γ + rounded to two significant digits.
Table XII. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 3 d = 3 d = 3 and ℓ = 3 \ell= 3 ℓ = 3 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_+ γ + rounded to two significant digits.
Table 13. Energies for n r = 0 , 5 , 10 , 20 n_r = 0, 5, 10, 20 n r = 0 , 5 , 10 , 20 for Coulomb (m = − 1 m = -1 m = − 1 ), logarithmic (m = 0 m = 0 m = 0 ), linear (m = 1 m = 1 m = 1 ), cubic (m = 3 m = 3 m = 3 ), quartic (m = 4 m = 4 m = 4 ), and sextic (m = 6 m = 6 m = 6 ) potentials at d = 3 d = 3 d = 3 and ℓ = 4 \ell= 4 ℓ = 4 . Except for Coulomb potential, all printed digits in energies rounded to four decimal digits, while in A.D., R.D. and γ + \gamma_{+} γ + rounded to two significant digits. (ℓ = 4 \ell= 4 ℓ = 4 .)