a

r
X
i
v
:
p
h
y
s
i
c
s
/
0
4
0
6
0
7
2
v
1


[
p
h
y
s
i
c
s
.
a
t
o
m
-
p
h
]


1
6

J
u
n

2
0
0
4
Quantization Rules for Bound States of
the Schrodinger Equation
Zhong-Qi Ma
1),2)
and Bo-Wei Xu
3)
1) CCAST (World Laboratory), P.O.Box 8730, Beijing 100080, China
2) Institute of High Energy Physics, Beijing 100039, China
3) Department of Physics, Shanghai Jiaotong University, Shanghai 200030, China
An exact quantization rule for the bound states of the one-
dimensional Schr odinger equation is presented and is generalized to the
three-dimensional Schr odinger equation with a spherically symmetric po-
tential.
I. INTRODUCTION
In the development of quantum mechanics, the Bohr-Sommerfeld quantization rules
of the old quantum theory [1] occupy a position intermediate between classical and
quantum mechanics. The WKB approximation [24] is a method for the approximate
treatment of the Schrodinger wave function with another quantization rule [1]:
_
x
B
x
A
kdx = (n + 1/2), k =
_
2[E V (x)]/ h, n = 0, 1, 2, . . . , (1)
where x
A
and x
B
are two turning points
E = V (x
A
) = V (x
B
), EV (x), x
A
< x < x
B
, (2)
and n is the number of nodes of the WKB wave function between two turning points.
The half-integer number is the quantum correction to the Bohr-Sommerfeld result,
which comes from the fact that the wave function in WKB analysis may extend into
the classically forbidden region. The WKB method is precise for the harmonic oscil-
lator potential, and is expected to be most useful in the nearly classical limit of large

Electronic address: mazq@sun.ihep.ac.cn

Electronic address: bwxu@sjtu.edu.cn

1
quantum numbers, namely, it is good when n is large in comparison with unity. Various
renements have been developed to improve the accuracy of the WKB method [57]
where the main modication was made on the phase loss at the turning points. Re-
cently, Cao et al. presented a calculation method, called the analytic transfer matrix
method (ATMM) [8,9], for the energy levels of bound states of the one-dimensional
Schrodinger equation, where the phase contribution picked up by the scattered sub-
waves was included in the new quantization rule. The accurate numerical calculation
results for some examples were given there [8,9]. This is a prospective method for
calculating the energy levels of bound states of the Schrodinger equation, but has to be
developed and improved further. In this Letter we apply the matching condition of the
logarithmic derivatives (which is a standard method in quantum mechanics) and the
fundamental concept of the mathematical analysis to the problem of bound states of
one-dimensional Schrodinger equation, and derive an exact quantization rule without
any approximation. The quantization rule is rigorous and general for any Schrodinger
equation with only one variable. It is a fundamental problem in quantum mechanics.
The plan of this Letter is as follows. In section 2 we will develop the method of
the numerical calculation for the energy levels of bound states of the one-dimensional
Schrodinger equation by the matching condition of the logarithm derivatives of the
wave function. In section 3 we present an exact quantization rule for one-dimensional
Schrodinger equation. The quantization rule is proved without any approximation. Its
validity can also be conrmed with some solvable examples. In section 4 the quantiza-
tion rule is generalized to the three-dimensional Schrodinger equation with a spherically
symmetric potential. The conclusion is given in section 5.
II. MATCHING CONDITION OF LOGARITHM DERIVATIVES
Consider the one-dimensional Schrodinger equation
d
2
dx
2
(x) =
2
h
2
[E V (x)] (x), (3)
where is the mass of the particle, and the potential V (x) is a piecewise continuous
real function of x. The logarithm derivative (x) of the wave function (x) is
(x) =
1
(x)
d(x)
dx
. (4)
From the Schrodinger equation (3) we have
d
dx
(x) = k(x)
2
(x)
2
, E V (x), (5)
d
dx
(x) = (x)
2
(x)
2
, E V (x), (6)
where
k(x) =
_
2[E V (x)]/ h, E V (x),
(x) =
_
2[V (x) E]/ h, E V (x),
k(x) = (x) = 0, E = V (x).
(7)
It is obvious that (x) decreases monotonically with respect to x when E V (x), but
not monotonically when E V (x). Note that near a node of the wave function (x) in
the region where E V (x), (x) decreases to , jumps to +, and then, decreases
again.
Arbitrarily choosing an energy E as a parameter, we assume for deniteness that
the potential V (x) satises
V (x) = V
I
E, < x x
I
,
V (x)E, x
I
< x < x
A
or x
B
< x < x
F
,
V (x) = E, x = x
A
or x = x
B
,
V (x) < E, x
A
< x < x
B
,
V (x) = V
F
E x
F
x < .
(8)
x
A
and x
B
, x
A
< x
B
, are called two turning points where V (x
A
) = V (x
B
) = E. This
potential can be easily generalized.
The Schrodinger equation (3) is a linear dierential equation of the second order, for
which there are two independent solutions. In the region < x x
I
, one solution
is divergent exponentially, and the other is physically admissible:
(x) e

I
x
, (x
I
) =
I
=
_
2(V
I
E)/ h0. (9)
Similarly, in the region x
F
x < , the physically admissible solution is
(x) e

F
x
, (x
F
) =
F
=
_
2(V
F
E)/ h < 0. (10)
For the general potential, if E < V (), both (x
I
) and (x
F
) are positive .
By making use of the fundamental concept of the mathematical analysis, we replace
the continuous potential well with a stack of thin lms each of which has a constant
potential. We rst divide the region x
I
x x
A
into n equal lms with width d
n
,
where E V (x) and x
A
= x
I
+ nd
n
. In the jth lm, x
I
+ jd
n
d
n
x x
I
+ jd
n
,
V (x) is replaced with a constant potential V
j
V
j
= V (x
I
+jd
n
d
n
/2),
j
=
_
2[V
j
E]/ h. (11)
Solving the Schrodinger equation at this lm, we obtain

j
(x) = A
j
e

j
x
+B
j
e

j
x
. (12)
On two ends of the lm, the logarithm derivatives
j1
and
j
, which should match
with the logarithm derivatives at the ends of the neighboring lms, are

j
=
1

j
(x)
d
j
(x)
dx

x=x
I
+jdn
=
j
A
j
e

j
(x
I
+jdn)
B
j
e

j
(x
I
+jdn)
A
j
e

j
(x
I
+jdn)
+B
j
e

j
(x
I
+jdn)
,

j1
=
1

j
(x)
d
j
(x)
dx

x=x
I
+jdndn
=
j
A
j
e

j
(x
I
+jdndn)
B
j
e

j
(x
I
+jdndn)
A
j
e

j
(x
I
+jdndn)
+B
j
e

j
(x
I
+jdndn)
.
(13)
From the second formula of Eq. (13) we obtain
A
j
e

j
(x
I
+jdndn)
{
j

j1
} = B
j
e

j
(x
I
+jdndn)
{
j
+
j1
} .
Substituting it into the rst formula of Eq. (13) we have

j
=
j

j
tanh(
j
d
n
) +
j1

j
+
j1
tanh (
j
d
n
)
. (14)
This is a recursive relation. If
j1
is positive,
j
is positive, too. There is no zero
both in the numerator and in the denominator of Eq. (14), so
j
is nite and non-
vanishing. Since (x
I
) is positive and known, one is able to calculate
n
= (x
A
) from

0
= (x
I
) =
I
with the recursive relation (14) as j increases from 1 to n. (x
A
) is
positive, nite and non-vanishing (see Appendix A). The calculated precision depends
on the number n of the lms. In principle, one may obtain a precise (x
A
) if n is large
enough.
Similar calculation can be made in the region x
F
x < . The recursive relation
(14) becomes

j1
=
j

j
tanh (
j
d
n
)

j
tanh (
j
d
n
)
. (15)
If
j
is negative,
j1
is negative, nite and non-vanishing. Since (x
F
) is negative
and known, one is able to calculate
0
= (x
B
+) from
n
= (x
F
) =
F
with
the recursive relation (15) as j decreases from n to 1. (x
B
+) is negative, nite and
non-vanishing. In principle, one may obtain a precise (x
B
+) if n is large enough.
Physically, the fact that there is no zero of (x) in the regions < x x
A
and
x
B
x < implies that in those two regions the wave function (x) is exponential
decay. However, there may be a zero of (x) in the other classically forbidden regions
(see the end of this section).
Now, we divide the region x
A
x x
B
into m equal lms with width d
m
, where
E V (x) and x
B
= x
A
+md
m
. In the jth lm, x
A
+jd
m
d
m
x x
A
+jd
m
, V (x)
is replaced with a constant potential V
j
V
j
= V (x
A
+jd
m
d
m
/2), k
j
=
_
2[E V
j
]/ h. (16)
Solving the Schrodinger equation at this lm, we obtain

j
(x) = C
j
sin (k
j
x +
j
) . (17)
On two ends of the lm, the logarithm derivatives
j1
and
j
, which should match
with the logarithm derivatives at the ends of the neighboring lms, are

j1
=
1

j
(x)
d
j
(x)
dx

x=x
A
+jdmdm
= k
j
cot [k
j
(x
A
+jd
m
d
m
) +
j
] ,

j
=
1

j
(x)
d
j
(x)
dx

x=x
A
+jdm
= k
j
cot [k
j
(x
A
+jd
m
) +
j
] .
(18)
From Eq. (18) we obtain

j
= k
j
cot
_
Arctan
_
k
j

j1
_
+k
j
d
m
_
, (19)
and
k
j
d
m
= Arctan
_
k
j

j1
_
+ Arctan
_
k
j

j
_
+q,
q =
_

_
0 no zero of (x) occurs in x
A
+jd
m
d
m
x < x
A
+jd
m
1 a zero of (x) occurs in x
A
+jd
m
d
m
x < x
A
+jd
m
,
(20)
where Arctan denotes the principle value of the inverse tangent function:
Arctan = , = tan , /2 < /2. (21)
Note that (x) decreases monotonically with respect to x when EV (x) [see Eq. (5)].
If no zero of (x) occurs in the lm x
A
+jd
m
d
m
x < x
A
+jd
m
,
Arctan
_
k
j

j1
_
< Arctan
_
k
j

j
_
/2,
Equation (20) holds with q = 0. If one zero of (x) occurs in the lm x
A
+jd
m
d
m

x < x
A
+jd
m
,
j1
0, and
j
< 0. Thus,
Arctan
_
k
j

j1
_
/2, and Arctan
_
k
j

j
_
/2,
we have to add an additional on the right-hand side of Eq. (20) such that its right-
hand side is positive and equal to k
j
d
m
. Since the width d
m
of the lm is very small,
we do not consider the case where more than one zeroes of (x) occur in the lm.
Equation (19) is a recursive relation, with which one is able to calculate
m
=
(x
B
) from
0
= (x
A
) as j increases from 1 to m. The calculated precision depends
on the number m of the lms. In principle, one may obtain a precise (x
B
) if m is
large enough.
From the Sturm-Liouville theorem (see Appendix B), as E increases, (x
B
) de-
creases monotonically and (x
B
+) increases monotonically. Choosing the parameter E
by dichotomy such that (x
B
) matches with (x
B
+), we obtain a bound state with
the energy E.
Cao et al. [8] derived the recursive relations similar to Eqs. (14), (15) and (19),
and calculated some examples with more precise results than those obtained with the
nonintegral Maslov index and the standard WKB method. Zhou et al. [9] calculated a
problem with a one-dimensional symmetric double-well potential. Since the potential
is symmetric, V (x) = V (x), the solution (x) of the Schrodinger equation is even
or odd in the spatial inversion, and then, its logarithm derivative (x) is odd. If a
nontrivial solution (x) is odd, (0) = 0 and (x) has to be innity, (0+) = +.
If a nontrivial solution (x) is even, (0) = 0 and (x) has to be zero because (x)
is odd, namely, (0+) = 0. One can calculate the energy levels of this system by the
above method in the half space 0 < x < . The dierent boundary conditions of
(0+) will give dierent solutions with dierent energies, as calculated in [9].
III. QUANTIZATION RULE
In the preceding section, we divided the region x
A
x x
B
into m equal lms,
where E V (x), and obtained Eq. (20) for k
j
d
m
. Summing up Eq. (20) from j = 1
to j = m, we obtain
m

j=1
k
j
d
m
= N Arctan
_
k
1

0
_
+ Arctan
_
k
1

1
_
Arctan
_
k
2

1
_
+ Arctan
_
k
2

2
_
+. . . Arctan
_
k
m1

m2
_
+ Arctan
_
k
m1

m1
_
Arctan
_
k
m

m1
_
+ Arctan
_
k
m

m
_
.
(22)
where
0
= (x
A
),
m
= (x
B
), and N is the number of zeroes of the logarithm
derivative (x) in the region x
A
x < x
B
. When m goes to innity, d
m
tends to zero,
and the sum in Eq. (22) becomes an integral. Thus, we obtain a new quantization
rule:
_
x
B
x
A
k(x)dx = N + lim
m
_
Arctan
_
k
1
(x
A
)
_
+ Arctan
_
k
m
(x
B
)
_
+
m1

j=1
_
Arctan
_
k
j

j
_
Arctan
_
k
j+1

j
__
_
_
_
,
(23)
where
j
is calculated recursively with Eq. (19). The rst term of the right-hand
side of Eq. (23) comes from the zeroes of the logarithm derivative (x) in the region
x
A
x < x
B
. Since (x
A
)0 and (x
B
) < 0, the second and the third terms are
vanishing as m goes to innity if the potential is continuous at the turning points. The
last sum denotes the phase contribution devoted by the scattered subwaves.
The formula (23) has another expression. If one changes
j
=

j
+/2 in Eq. (17),
Eq. (18) becomes

j1
= k
j
tan
_
k
j
(x
A
+jd
m
d
m
) +

j
_
,

j
= k
j
tan
_
k
j
(x
A
+jd
m
) +

j
_
.
Then, equations (20) and (23) become
k
j
d
m
= Arctan
_

j1
k
j
_
Arctan
_

j
k
j
_
+q

,
q

=
_

_
0 no zero of (x) occurs in x
A
+jd
m
d
m
< x x
A
+jd
m
,
1 a zero of (x) occurs in x
A
+jd
m
d
m
< x x
A
+jd
m
,
_
x
B
x
A
k(x)dx = N

+ lim
m
_
Arctan
_
(x
A
)
k
1
_
Arctan
_
(x
B
)
k
m
_
+
m1

j=1
_
Arctan
_

j
k
j+1
_
Arctan
_

j
k
j
__
_
_
_
,
(24)
where N

denotes the number of nodes of the wave function (x) in the region x
A
< x
x
B
. If the potential is continuous at the turning points, due to (x
A
)0 and (x
B
) < 0,
the second and the third terms are
lim
m
Arctan
_
(x
A
)
k
1
_
lim
m
Arctan
_
(x
B
)
k
m
_
= /2 (/2) = . (25)
Since (x) decreases monotonically in the region x
A
< x < x
B
, N = N

+ 1.
The sum on the right-hand side of Eq. (23) can be transformed into an integral
expression:
_
x
B
x
A
k(x)dx = N + lim
m
_
Arctan
_
k
1
(x
A
)
_
+ Arctan
_
k
m
(x
B
)
__

_
x
B
x
A
(x) (dk(x)/dx)
(x)
2
+k(x)
2
dx.
(26)
Two terms in the curly brackets are vanishing as m goes to innity if the potential is
continuous at the turning points. The sum in Eq. (24) can also be transformed into
an integral expression.
The quantization rule (26) is proved without any approximation, so that it is exact.
Its validity can also be conrmed by comparing it with the following solvable exam-
ples. Both Eqs. (23) and (24) are the formulas of numerical calculation for Eq. (26).
Cao et al. [8,9] presented an expression similar to Eq. (24) and demonstrated it to
be very eective in numerical calculation through two examples: the one-dimensional
Schrodinger equation with a power-law potential [8] and a symmetric double-well po-
tential [9].
Ex. 1. The harmonic oscillator potential.
The WKB method is precise for the harmonic oscillator potential. Now, we are
going to check our new quantization rule (26) for the harmonic oscillator potential
V (x) =
2
x
2
/2. Let
=
_

h
, = x, (27)
we have

n
(x) = N
n
e

2
/2
H
n
(), E
n
= h (n + 1/2) ,

n
(x) = + 2nH
n1
()/H
n
(),
(28)
where N
n
is the normalization factor, H
n
() denote the nth Hermitian polynomial. In
the region x
A
x x
B
, where x
A
= x
B
=

2n + 1/, we have
k
(n)
(x) =
_
2n + 1
2
,
dk
(n)
(x)
dx
=
4
x/k
(n)
(x). (29)
After calculation for 0 n 10, we know that
n
(x
A
) =
n
(x
B
)0,
n
(x) has
n + 1 zeroes in the region x
B
x < x
B
, and Eq. (26) becomes
_
x
B
x
A
k
(n)
(x)dx = (n + 1) +
n
,

n
=
_
x
B
x
A
(x)
_
dk
(n)
(x)/dx
_
(x)
2
+k
(n)
(x)
2
dx
= 2
_

2n+1
0
2nH
n1
() H
n
()
[2n + 1
2
] H
n
()
2
+ [2nH
n1
() H
n
()]
2
H
n
()d

2n + 1
2
= 2
_

2n+1
0
(2n
2
) d

2n + 1
2
+F
n
.
(30)
At least for 0 n 10 we obtain
n
= /2 by Mathematica. In fact, the rst term
in
n
is (2n 1)/2, and through a variable transformation:
_
2n + 1
2
= t
_
+

2n + 1
_
, d =
4

2n + 1tdt
(1 +t
2
)
2
,
the second integral F
n
is calculated to be n:
F
0
= 0,
F
1
= 2
_

3
0
2d

3
2
(1 +
2
)
= 2
_
1
0
(1 +t
2
)dt
1 t
2
+t
4
= ,
F
2
= 2
_

5
0
4(4
2
+ 5)d

5
2
(4
4
+ 4
2
+ 5)
= 16
_
1
0
(1 +t
2
)(5 6t
2
+ 5t
4
)dt
25 76t
2
+ 118t
4
76t
6
+ 25t
8
= 2,
F
3
= 2
_

7
0
6(4
4
+ 6
2
+ 9)d

7
2
(4
6
+ 9
2
+ 9)
= 24
_
1
0
(1 +t
2
)(247 748t
2
+ 1146t
4
748t
6
+ 247t
8
)dt
1444 8052t
2
+ 20652t
4
27512t
6
+ 20652t
8
8052t
10
+ 1444t
12
= 3.
Thus, we demonstrate that the quantization rule (26) is the same as Eq. (1) for the
harmonic oscillator potential.
Ex. 2. The square well potential.
Discuss a nite square well potential V (x)
V (x) =
_

_
V
A
x ,
V
B
x ,
0 < x < .
(31)
The logarithm derivatives at the turning points x
A
= and x
B
= are
(x
A
) =
I
=
_
2(V
A
E)/ h, (x
B
) =
F
=
_
2(V
B
E)/ h, (32)
when E < V
A
and E < V
B
. The solution to the Schrodinger equation is

n
(x) = sin
_
k
(n)
x +
(n)
_
, (33)
and its logarithm derivative
n
(x) is

n
(x) = k
(n)
cot
_
k
(n)
x +
(n)
_
. (34)
k
(n)
and
(n)
are determined by the matching conditions at the turning points x
A
and
x
B
,
tan
_
k
(n)
+
(n)
_
= k
(n)
/
I
, tan
_
k
(n)
+
(n)
_
= k
(n)
/
F
.
Hence, we obtain
k
(n)
=
_
Arctan
_
k
(n)
/
I
_
+ Arctan
_
k
(n)
/
F
__
/(2) +n/2. (35)

n
(x) has n zeroes in the region x , and k(x) takes the constant value k
(n)
in the region, dk(x)/dx = 0. The energy levels are E
n
=
_
hk
(n)
_
2
/(2). In terms of
Eq. (35), the right-hand side of Eq. (26) is calculated to be
n Arctan
_
k
1
(x
A
)
_
+ Arctan
_
k
m
(x
B
)
_
= n
_
Arctan
_
k
(n)
/
I
_
+ Arctan
_
k
(n)
/
F
__
= 2k
(n)
.
(36)
Due to the constant k(x), the left-hand side of Eq. (26) is equal to the same value:
_
x
B
x
A
k(x)dx = 2k
(n)
. (37)
This quantization rule is dierent from both the Bohr-Sommerfeld one and that given
by the WKB approximation. When V
A
and V
B
tend to innity, due to Eqs. (32)
and (35) k
(n)
goes to n/2, and the quantization rule (37) coincides with the Bohr-
Sommerfeld one.
IV. THREE-DIMENSIONAL SCHR

ODINGER EQUATION
Consider the three-dimensional Schrodinger equation with a spherically symmetric
potential. After separation of the angular part of the wave function,
(r) = r
1
R(r)Y

m
(, ), (38)
the radial equation of the Schrodinger equation is
d
2
R(r)
dr
2
=
2
h
2
{E U(r)} R(r), U(r) =
h
2
( + 1)
2r
2
+V (r). (39)
Since Eq. (39) is similar to Eq. (3), its energy levels can be calculated by the matching
conditions of the logarithm derivatives, where the logarithm derivative is dened as
(r) = R(r)
1
dR(r)
dr
. (40)
As an example, we discuss the problem of the hydrogen atom, where the potential
V (r) is
V (r) =
e
2
r
. (41)
When r , we have
R(r) exp{r
_
2|E|/ h}, (r)
_
2|E|/ h.
When r 0, we have
R(r) r
+1
_
1
e
2
h
2
( + 1)
r
_
, (r) ( + 1)/r.
By the method of matching condition of logarithm derivatives, one is able to calculate
the energy E with Eqs. (14), (15) and (19).
On the other hand, if the solution to Eq. (39) has known, we are able to check
whether the quantization rule (26) holds. For the energy E
n
E
n
=
e
4
2 h
2
n
2
, n = 1, 2, 3, . . . , (42)
we have the solution [1]
R
n
(r) = N
n
e
/2

+1
L
2+1
n+
(), =
2e
2
n h
2
r, (43)
where N
n
is the normalization factor and L
2+1
n+
() is the associated Laguerre polyno-
mials.
When 0, the turning points r
A
and r
B
satisfying U(r
A
) = U(r
B
) = E
n
are
r
A
=
n h
2
2e
2

A
,
A
= 2
_
n
_
n
2
( + 1)
_
1/2
_
,
r
B
=
n h
2
2e
2

B
,
B
= 2
_
n +
_
n
2
( + 1)
_
1/2
_
.
(44)
When = 0, we dene r
A
= 0 with U(r
A
) = e
2
/r
A
. r
B
with U(r
B
) = E
n
still
satises Eq. (37). The momentum k
n
(r) is
k
n
(r) =
1
2r
{(
A
) (
B
)}
1/2
. (45)
In the Schrodinger equation for the hydrogen atom, the quantization rule (26) becomes
_
r
B
r
A
k
n
(r)dr = N + lim
m
_
Arctan
_
k
1

n
(r
A
)
_
+ Arctan
_
k
m

n
(r
B
)
__

_
r
B
r
A

n
(r) (dk
n
(r)/dr)

n
(r)
2
+k
n
(r)
2
dr.
(46)
Calculating the left-hand side of Eq. (51), we obtain with Eq. (50)
_
r
B
r
A
k
n
(r)dr =
_

B

A
d
2
{(
A
) (
B
)}
1/2
=
_
n
_
( + 1)
_
. (47)
Now, we calculate the right-hand side of Eq. (51). The number N of zeroes of
n
(r)
in the region r
A
r < r
B
is (n). The logarithm derivatives at the turning points r
A
and r
B
are non-vanishing. When = 0, r
A
is not the turning point, but
n0
(r
A
) = 1/r.
Thus, we have
lim
m
Arctan
_
k
m

n
(r
B
)
_
= 0, 0,
lim
m
Arctan
_
k
1

n
(r
A
)
_
= 0, 0,
lim
m
Arctan
_
k
1

n0
(r
A
)
_
= lim
r0
Arctan
_
r
_
2(E
n
+e
2
/r)/ h
_
= 0, = 0.
(48)
Calculating the integral in Eq. (51) by Mathematica, at least for n = 1, 2, 3, and
< n, we obtain:

_
r
B
r
A

n
(r) (dk
n
(r)/dr)

n
(r)
2
+k
n
(r)
2
dr =
_

_
( + 1)
_
. (49)
Therefore, the quantization rule (46) holds for the hydrogen atom.
V. CONCLUSIONS
In this Letter, with the matching condition of the logarithm derivatives and the
fundamental concept of the mathematical analysis, we proposed a formula (23) for nu-
merically calculating the energy levels of bound states of the Schrodinger equation in
one dimension. Calculating the integral form of Eq. (23), we obtained an exact quan-
tization rule (26) for the bound states of the one-dimensional Schrodinger equation.
The exact quantization rule was generalized to bound states of the three-dimensional
Schrodinger equation with a spherically symmetric potential. The quantization rule was
conrmed by checking some examples where the solutions of the Schrodinger equation
are known. Two examples of numerical calculation for the one-dimensional Schrodinger
equation with a power-law potential [8] and with a symmetric double-well potential [9]
demonstrated that the exact quantization rule is very eective in numerical calculation.
Cao et al. [8,9] derived a formula similar to Eq. (24). However, their formulation
contained some unclear points. The integer N in Eq. (29) of Ref. [8] is unclaried.
They introduced the exponentially decaying coecients P
j
, but the physical meaning
of P
j
is rather ambiguous. In fact, P
j
are nothing but the logarithm derivatives of the
wave function (in their region of 0 x x
C
) or that multiplied with 1 (in their
region of x
D
x x
S
). This is the reason why the coecients should be matched at
the turning points, which was not explained clearly in Ref. [8]. Finally we would like
to point out that the series form (23) is an approximate formula of the integral form
(26) derived in the present Letter, which is the exact quantization rule.
ACKNOWLEDGMENTS. One of the authors (BWX) would like to thank Pro-
fessor Z. Cao for drawing his attention to this problem. This work was supported by
the National Natural Science Foundation of China.
[1] L. I. Schi, Quantum Mechanics, Third Edition, (McGraw-Hill Book Co., New York,
1968).
[2] G. Wentzel, Z. Physik. 38, 518 (1926).
[3] H. A. Kramers, Z. Physik. 39, 828 (1926).
[4] L. Brillouin, Compt. Rend. 183, 24 (1926).
[5] F. Xiang and G. L. Yip, J. Lightwave Technol. 12, 443 (1994).
[6] V. S. Popov, B. M. Karmakov, and V. D. Mur, Phys. Lett. A 210, 402 (1996).
[7] S. Zivanovic, V. Milanovic, and Z. Ikonic, Phys. Status Solidi B 204, 713 (1997).
[8] Z. Cao, Q. Liu, Q. Shen, X. Dou, Y. Chen, and Y. Ozaki, Phys. Rev. A 63, 054103 (2001).
[9] F. Zhou, Z. Cao, and Q. Shen, Phys. Rev. A 67, 062112 (2003).
Appendix A Property of (x
A
)
From the recursive relation (14) and (x
I
)0, we have proved that (x
A
) is non-
negative and nite. Now, we are going to prove that (x
A
) = 0 by reduction to
absurdity. When x x
A
and x near x
A
, consider the leading term in the power series
in (x
A
x). Due to (x
A
) = 0, we have (x) C(x
A
x)

, where C0 and 0. If
(x
A
) = 0, from Eq. (6) we have (x
A
) C
2
(x
A
x)
2+1
/(2 + 1) < 0. It conicts
to the fact that (x
A
) is non-negative.
Appendix B The Sturm-Liouville Theorem
Denote by (E, x) the solution of the Schrodinger equation (3) with the energy E.
Multiplying Eq. (3) with (E
1
, x) we have
(E
1
, x)

2
x
2
(E, x) =
2
h
2
[E V (x)] (E, x)(E
1
, x), (B1)
Exchanging E
1
and E and subtracting from it by Eq. (B1), we obtain

x
_
(E, x)
(E
1
, x)
x
(E
1
, x)
(E, x)
x
_
=
2
h
2
(E
1
E) (E, x)(E
1
, x).
(B2)
When E < V (), the boundary condition gives that both solutions (E, x) and
(E
1
, x) are vanishing at negative innity. Integrating Eq. (B2) from to x
B
, we
obtain
1
E
1
E
_
(E, x)
(E
1
, x)
x
(E
1
, x)
(E, x)
x
_
x=x
B

=
2
h
2
_
x
B

(E, x)(E
1
, x)dx.
Taking the limit as E
1
goes to E, we have
(E, x)
E

x=x
B

=

E
_
1
(E, x)
(E, x)
x
_
x=x
B

=
2
h
2
(E, x
B
)
2
_
x
B

(E, x)
2
dx < 0.
(B3)
Namely, at a given point x
B
, the logarithm derivative (E, x) of the wave function
(E, x) decreases monotonically as E increases.
Similarly, when E < V (), the boundary condition gives that both solutions
(E, x) and (E
1
, x) are vanishing at positive innity. Integrating Eq. (B2) from
x
B
+ to , we obtain
1
E
1
E
_
(E, x)
(E
1
, x)
x
(E
1
, x)
(E, x)
x
_
x=x
B
+
=
2
h
2
_

x
B
(E, x)(E
1
, x)dx.
Taking the limit as E
1
goes to E, we have
(E, x)
E

x=x
B
+
=
2
h
2
(E, x
B
)
2
_

x
B
(E, x)
2
dx0. (B4)
Namely, at a given point x
B
+, the logarithm derivative (E, x) of the wave function
(E, x) increases monotonically as E increases.