Exact Green’s function for rectangular potentials and its application to quasi-bound states

Fabiano M. Andradea
a Departamento de Matemática e Estatı́stica, Universidade Estadual de Ponta Grossa, 84030-900 Ponta Grossa-PR, Brazil

Abstract
In this work we calculate the exact Green’s function for arbitrary rectangular potentials. Specifically we focus on Green’s function
for rectangular quantum wells enlarging the knowledge of exact solutions for Green’s functions and also generalizing and resuming
arXiv:1403.5964v1 [quant-ph] 24 Mar 2014

results in the literature. The exact formula has the form of a sum over paths and always can be cast into a closed analytic expression.
From the poles and residues of the Green’s function the bound states eigenenergies and eigenfunctions with the correct normaliza-
tion constant are obtained. In order to show the versatility of the method, an application of the Green’s function approach to extract
information of quasi-bound states in rectangular barriers, where the standard analysis of quantum amplitudes fail, is presented.
Keywords: rectangular potential, Green’s function, bound state, resonant scattering

1. Introduction simple way avoiding complicated calculations, for arbitrary rect-

angular potentials enlarging the knowledge of exact solutions
One-dimensional systems are used frequently in quantum for Greens functions and also generalizing and resuming results
mechanics to approximate real situations because they are rel- in the literature. The procedure outlined here can be thought of
atively easy to treat mathematically, allowing to get a deeper as exact version of the approximation in [18] and provides sim-
insight on the physics involved. They are useful in a num- ple and direct way to construct recurrence relations for quantum
ber of applications in contemporary physics and their simplic- amplitudes for one-dimensional scattering discussed in [19].
ity has made them valuable as academic and research tools. Specifically, we focus on Green’s function for single quantum
For instance, it turns up in the description of resonant tunnel- wells to avoid extra and unnecessary mathematical complica-
ing diode devices, disordered one-dimensional lattices, realis- tions, but the method is general and could be applied to multi-
tic one-dimensional solid-state system such as quantum wells, ple quantum wells as well [17]. From the Green’s functions ob-
junctions and superlattices [1, 2], and time analysis of one- tained, we thus describe how to extract the bound states eigenen-
dimensional tunneling processes [3]. In particular, one-dimen- ergies and eigenfunctions. Also, we discuss an application of
sional rectangular potentials can be used to model isolated tran- the Green’s function approach to extract information of quasi-
sitions, observed in semiconductors, from a bound state within bound states in rectangular potentials. We should observe that
a quantum well to a bound state at an energy greater than the the Green’s function approach used in the present work, has
barrier height [4]. Limiting cases of rectangular potentials can been successfully used in the calculation of exact Green’s func-
be used to describe point interactions [5], as well as regularized tions for segmented potentials [15], in calculation of asymptotic
singular interactions [6]. Also, we should point out that the use Green’s functions [16], in the determination and discussion of
of square-barrier and square-well potentials as simple models bound states in multiple quantum wells [17]. Also, in the cal-
for more realistic physical problems has a long history in the culation of exact Green’s function for quantum graphs [20],
theory of heterostructures in solid-state physics [7]. general point interactions [21], and scattering quantum walks
However, the propagator or its Fourier transform, the Green’s [22, 23].
function, for rectangular potentials, although these are very sim- This paper is organized as follows. In Sec. 2, we give a
ple systems, are difficult to obtain and involve lengthy and te- briefly review on the definition and properties of the Green’s
dious calculations. Indeed, the exact Green’s function and prop- functions. In Sec. 3, the Green’s functions approach for the
agators for step potential and square barrier was obtained only rectangular potential is presented. In Sec. 4, the construction of
in the late 80s until the early 90s [8–12]. Moreover, it is well the exact Green’s functions for a general rectangular asymmet-
known that semiclassical approach, i.e., the Van Vleck-Gutzwiller ric well potential is presented and how the sough eigenenergies
formula [13], or the WKB approach [14], give poor results and eigenfunctions are extracted. In Sec. 5, the case of the
when applied to the class of rectangular potentials due to their square well potential is discussed and the definite parity eigen-
discontinuities. functions are determined. In Sec. 6 the case of the infinite well
In the present work, we show how to obtain the exact Green’s potential is discussed. In Sec. 7 a Green’s function approach
function [15–17], given as a general sum over paths, in a very to extract information from system with quasi-bound states are
presented. Finally, Sec. 8 contains our conclusions.
Email address: fmandrade@uepg.br (Fabiano M. Andrade)

Preprint submitted to Physics Letters A March 25, 2014

2. Green’s function: Definition and properties V (x)

The Green’s functions are an important tool in quantum me- + +

chanics for calculating eigenenergies, eigenfunctions and den-
u4
sity of states [24]. It can be defined as solutions of the inhomo-
geneous differential equation u2

[E − Ĥ(x f )]G(x f , xi ; E) = δ(x f − xi ), (1) u3

subject to certain boundary conditions. Ĥ(x) is the Hamiltonian u1

operator
~2 d 2
Ĥ(x) = − + V(x). (2)
2m dx2 x1 x2 x3 x
The Green’s function G(x f , xi ; E) for a quantum system can
be interpreted as a probability amplitude for a particle that ini- Figure 1: A rectangular potential with three points of discontinuities. The po-
tially at the point xi moves to the point x f for a fixed energy tential can be considered as a sum of the three potential steps.
[25]
G(x f , xi ; E) = hx f |Ĝ(E)|xi i, (3)
potentials. As depicted in Fig. 1, a rectangular potential can be
where Ĝ(E) ≡ limη→0+ [E − Ĥ(x)+iη]−1 is the resolvent operator treated as the sum of n step potentials. The potential function
[26]. for a rectangular potential can be written as
Suppose we have a complete set of normalized energy eigen- 
states ψn (x), n = 0, 1, ... with u for x < x1 , (Region 1),
 1



V(x) =  u j for x j ≤ x < x j+1 = x j + ℓ j , (Region j),
Ĥ(x)ψn = En ψn , (4)  u for x ≥ x , (Region n),

n n
(7)
so that the solution of Eq. (1) can be written as where u j , ℓ j and x j are the height of the potential, the length of
X ψn (x f )ψ∗ (xi ) the region j and the localization of the jth discontinuous point,
n
G(x f , xi ; E) = . (5) respectively. The wave number in each different region is k j =
n
E − En p
2m(E − u j )/~ = iκ j . In what follow, we do not distinguish the
For a discrete spectrum, we thus identify the poles of Green’s cases whether k j is a real or a purely imaginary number (iκ j ),
function with bound states eigenenergies and the residues at except otherwise mentioned.
these points give a tensorial product of the bound states eigen- The Green’s function for this system, for a fixed energy E
functions. The continuous part of the spectrum corresponds to and the end points xi and x f , is given by
a branch cut of G(x f , xi ; E) [26, 27]. The following limit
m X i
G f,i = p Wsp exp [ S sp (x f , xi ; E)], (8)
lim (E − En )G(x f , xi ; E) = ψn (x f ) ψ∗n (xi ) (6) 2
i~ k f ki sp ~
E→En

is used to extract the discrete bound states from the Green’s
function.
There are basically three methods for calculating the Green’s
functions [24]: solving the differential equation in (1); sum-
ming up the spectral representation in (5); or performing the
Feynman path integral expansion for the Green’s function [28].
In this paper, we present a Feynman-like procedure. As we are
going to see, the approach requires one to determine the quan-
tum amplitudes of each individual potential, but in general, this
is a much easier task than to calculate numerically propagation
along the whole V(x).
3. The Green’s function approach for rectangular poten-
tials
According to Refs. [15–17] the Green’s function G(x f , xi ; E)
(in what follow, we will denote it by G f,i for short) for a general
1D potential is obtained by writing the potential V(x) as a sum
of n individual potentials u j , where each u j vanish asymptoti-
cally. In the present work, we focus our attention on rectangular
2
where Wsp is the amplitude (or weight) and S sp is the classical
action. The above sum is performed over all scattering paths
(sp) starting in xi and ending in x f . For each sp, the classical
action is obtained fromRthe propagation over action of potential
xb
u j , S sp (xb , xa ; E) = ~ x k j dx. A few comments concerning
a
the Eq. (8) are necessary. The expression in Eq. (8) is obtained
from a recursive procedure, i.e., G f,i for n potentials are derived
from G f,i for n − 1 potentials. The Wsp are related to local quan-
tum effects, so they depend on quantum amplitudes (R(±) j and
T (±)
j ) of individual u j (cf. Appendix A). In fact, for the present
case of rectangular potentials, R(±) (±)
j and T j are the usual re-
flection and transmission amplitudes of individual u j [19] (up
to a phase [15]). (The superscript (+) ((−)) and subscript j de-
note the physical quantities of a quantum particle incident from
left (right) at the point x j , respectively.) In the general case
of smooth potentials, they are related, but we need to calculate
classical actions for each potential. In this later case, the inter-
ested reader can consult the Refs. [16, 17] for derivations.
 V (x) (a) (b)

u1
+
u3
x1 xi xf x2 x1 xi xf x2
(c) (d)
Region 1 Region 2 Region 3

u2

x1 xi xf x2 x1 xi xf x2
x1 x2 x (f )
(e)
Figure 2: An asymmetric well potential written as a sum of a left and right
potential steps. P1 P2

4. Asymmetric well potential x1 xi xf x2 x1 xi xf x2

In this section we will consider the case of a quantum parti-

Figure 3: In (a)-(e) schematic examples of scattering paths and in (f) the fami-
cle confined in the asymmetric well potential defined by lies of paths P1 and P2 .

u for x < x1 , (Region 1),
 1



V(x) =  u 2 for x 1 ≤ x < x 2 = x 1 + ℓ 2 , (Region 2), The Green’s function is then given by the sum of all contri-
 u for x ≥ x ,

(Region 3),
3 2 butions of n multiple reflections of all possibles scattering paths
(9) and can be written as (ℓ2 = x2 − x1 )
and depicted in Fig. 2. We seek the Green’s function for the
m n ik2 (x f −xi )
situation where E < u3 . The eigenenergies will be obtained G(2)
f,i = ik ~2 e
from the poles of the Green’s function and the eigenfunctions 2
∞ 
from their respective residues (cf. Sec. 2). We can write nine X
(+) 2ik2 ℓ2 n −ik2 (x1 −xi ) (−)
 h
+ R(−)
1 R2 e e R1
different Green’s functions for the problem, but for our pur-
n=0
poses, we analyze only three relevant possibilities, namely: (1)  
x1 < xi < x2 and x f < x1 ; (2) x1 < xi , x f < x2 ; and (3) × eik2 (x f −x1 ) + R(+)
2 e
ik2 ℓ2 −ik2 (x f −x2 )
e
x1 < xi < x2 and x f > x2 . The others six Green’s functions can + eik2 (x2 −xi ) R(+)
2
be obtained from symmetry considerations of the three above   io
cases. × e −ik2 (x f −x2 )
+ R(−)
1 e
ik2 ℓ2 ik2 (x f −x1 )
e . (10)
First, we want to exemplify how to use the Eq. (8) and how
In general, the reflections (transmissions) amplitudes have the
the geometric series appear. So, let us consider the possibility
property |R(±) 2 (±) 2
j | ≤ 1 (|T j | ≤ 1), in such way as the above sum
(2). To obtain G(2)f,i in a closed form, we need to sum up all the always converge. In fact, it is a geometric series. So, after a
possibles scattering paths for a quantum particle starting from
straightforward algebra the final form for the Green’s function
xi and arriving at x f , but it can always be done because the sum
is
in (8) forms a geometric series. Without loss of generality, let
m  
us set x f > xi . G(2)
f,i = 2
eik2 (x f −x1 ) + R(+)2 e
ik2 ℓ2 ik2 (x2 −x f )
e
Thus, in Fig. 3 is depicted five examples of scattering paths. i~ k2 faw
 
The first path in Fig. 3(a) is the direct propagation from xi to x f , × e−ik2 (xi −x1 ) + R(−) 1 e
ik2 ℓ2 −ik2 (x2 −xi )
e , (11)
which contributes with eik2 (x f −xi ) for G(2)
f,i . In Fig. 3(b), a quan-
tum particle leaves the point xi , goes to the point x2 , where it where faw = 1 − R(−) (+) 2ik2 ℓ2
1 R2 e .
hit the potential and is reflected, and so, goes to left to arrives at The method utilized above consists in sum up all scattering
the point x f . This path contributes with eik2 (x2 −xi ) R(+)
2 e
−ik2 (x f −x2 )
, paths, but this could be very tedious and cumbersome. How-
(+) ever, this can be done by a simple diagrammatic classification
where R2 is the reflection amplitude for a quantum particle in-
cident from the left of the potential u3 at the point x2 . Following in families of paths, which is a practical way to identify and
the same reasoning, the contributions for the other examples in perform the geometric series mentioned above. Consider again
Fig. 3 are: the confined particle in the asymmetric well potential. The par-
ticle starting from xi may (a) go to the right arriving x f ; or (b)
(c) e−ik2 (x1 −xi ) R(−)
1 e
ik2 (x f −x1 )
; go to the left, hit the potential u1 at x1 being reflected. There
(d) eik2 (x2 −xi ) R(+) ik2 ℓ2 (−) ik2 (x f −x1 ) is an infinite family of paths for the particle at x1 , we call this
2 e R1 e ;
family P1 ; and (c) the particle go to the right, hit the potential
(e) e−ik2 (x1 −xi ) R(−)
1 e
ik2 ℓ2 (+) −ik2 (x f −x2 )
R2 e . u3 at x2 being reflected. Like before, there is infinite family of
3
 ′
paths for the particle at x2 , we call this family P2 . These two where faw(n) = (d faw /dk)|k=kn , we obtain for the residues of the
infinite family of paths are schematically depicted in Fig. 3(f). Green’s function in Eqs. (11), (15) and Eq. (16)
Thus, by using the above prescription, the Green’s function can
~2
be written as ψ(2) (2)
n (x f )[ψn (xi )] =
∗
lim (k2 − kn2 )G(2)
f,i
2m k→kn
m  ik2 (x f −xi )
G(2) = e (n) (n)
 
f,i i~2 k2 1  ik(n) (x f −x1 ) k2 − k3 ik(n) (ℓ2 +x2 −x f ) 
= ′ (n) e 2 + (n) e 2
k2 + k3(n)

i faw

+ eik2 (x2 −xi ) R(+)
2 P2 + e
−ik2 (x1 −xi ) (−)
R1 P1 . (12)
 −ik(n) (x −x ) k2(n) − k1(n) ik(n) (ℓ −x +x ) 
 
The family P1 (P2 ) posses two contributions: (a) go to the right × e  2 i 1
+ (n) e 2 2 2 i 
 ,
(left), arriving at x f ; and (b) go to the right (left), hit the poten- k1 + k2(n)
tial u3 (u1 ) at x2 (x1 ) being reflected followed by the family P2
(P1 ). Thus, ~2
ψ(1) (2)
n (x f )[ψn (xi )] =
∗
lim (k2 − kn2 )G(2) f,i
2m k→kn
P1 = eik2 (x f −x1 ) + eik2 ℓ2 R(+)
2 P2 , (13a) vt (n)

1  k1
 2k2(n) 
−ik1(n) (x f −x1 ) 
P2 = e ik2 (x f −x2 )
+e ik2 ℓ2
R(−)
1 P1 . (13b) = ′ (n)  e

(n) (n) (n) 
i faw  k2 k1 + k2
Solving for P1 and P2 one obtains  −ik(n) (x −x ) k2(n) − k1(n) ik(n) (ℓ −x +x ) 
 
× e 2 i 1 + (n) e 2 2 2 i  ,
1 k1 + k2(n)
 
P1 = eik2 (x f −x1 ) + R(+)
2 e
ik2 ℓ2 ik2 (x2 −x f )
e , (14a)
faw
1  
~2
P2 = eik2 (x2 −x f ) + R(−)
1 e
ik2 ℓ2 ik2 (x f −x1 )
e . (14b) ψ(3) (x )[ψ (2)
(x )] ∗
= lim (k2 − kn2 )G(3)
faw n f n i
2m k→kn f,i
v t 
(n)
By substitution of Eq. (14) into Eq. (12), one obtains the 1  k3
 2k2(n) 
ik3(n) (x f −x2 ) 
Green’s function in Eq. (11). The other two cases can be ob- = ′ (n)  e 
i faw  k2(n) k2(n) + k3(n)

tained in a similar way. For instance, for case (1) we have
 −ik(n) (x −x ) k2(n) − k1(n) ik(n) (ℓ −x +x ) 
 
m  (−) 
G(1) × e 2 + (n) e 2  .
i 1 2 2 i
f,i = √ T 1 e−ik1 (x f −x1 )
i~2 k1 k2 faw k1 + k2(n)
 
× e−ik2 (xi −x1 ) + R(−)
1 e
ik2 ℓ2 −ik2 (x2 −xi )
e , (15) The correct normalized eigenfunctions corresponding to the bound
states of the system are thus given by:
and for case (3), we obtain Region 1: x < x1
m  (+)  v
G(3) T 2 eik3 (x f −x2 )
t (n)
f,i = √ 1 k1 2k2(n) (n)
i~2 k3 k2 faw ψ(1)
n (x) = ′ (n) e−ik1 (x−x1 ) , (19)
(n) (n) (n)
  i faw k2 k1 + k2
× e−ik2 (xi −x1 ) + R(−)
1 e
ik2 ℓ2 −ik2 (x2 −xi )
e . (16)
Region 2: x1 < x < x2
4.1. Calculation of the bound states s
The bound states are calculated from the residues of the 1  ik(n) (x−x1 )
ψ(2)
n (x) = ′ (n) e
2

Green’s functions in Eqs. (11), (15) and (16). Its poles En = i faw
~2 kn2 /2m are all contained in the term 1/ faw . They are calcu- k2(n) − k3(n) (n) 
lated from faw = 0, which leads to the following transcendental ++ eik2 (ℓ2 +x2 −x)
, (20)
equation k2(n) + k3(n)
 (n) Region 3: x > x2
 k2 − k1(n)   k2(n) − k3(n)  2ik(n) ℓ
 
    e 2 2 = 1, (17) v
t (n)
k1(n) + k2(n) k2(n) + k3(n) 1 k3 2k2(n) (n)
ψ(3)
n (x) = ′ (n) eik3 (x−x2 )
. (21)
i faw k2(n) k2(n) + k3(n)
k(n)
p
where j = 2m(En − u j )/~. This result agree with the one
found by the solution of the Schrödinger equation [29]. Using It should be observed that k2(n) (k1(n) and k3(n) ) is (are) real (imag-
the formula inary) number(s), and consequently ψ(2) (1) (3)
n (x) (ψn (x) and ψn (x))
E − En ~2 k2 − kn2 ~2 kn is (are) an oscillatory complex (decreasing real) exponential
lim = lim = ′ (n) , (18) function(s) of x. Also, the Green’s function used here yields
E→En faw 2m k→kn faw m faw
the correct normalization constant which often involves a diffi-
cult integral in the other methods. In what follow, we apply the
results obtained in this section in other well-known rectangular
potentials.
4
 V (x) Region 2: −a < x < a
u0 s
1  ikn (x+a)
ψ(2)
n (x) = ′ (n) e
i fsw
kn − k0(n) 
+ e2ikn a eikn (a−x) , (27)
kn + k0(n)

Region 3: x > a
v
t (n)
a x 1 k0 2kn (n)
−a ψ(3)
n (x) = ′ (n) (n)
eik0 (x−a) . (28)
k
i fsw n kn + k 0
Figure 4: A square well potential.
From Eq. (25), we can see that there are two equations de-
scribing the eigenenergies. These eigenenergies are determined
5. Square well potential by solving
The Green’s function for a square well potential (depicted in
(kn − k0(n) )e2ikn a = +(kn + k0(n) ), (even parity),
Fig. 4) can be obtained from the results of the previous section
by setting √u1 = u3 = u0 and u2 = 0, thus k1 = k3 = k0 and (kn − k0(n) )e2ikn a = −(kn + k0(n) ), (odd parity),
k2 = k = 2mE/~, also x1 = −a and x2 = a. From Eqs. (11),
(15) and (16) we thus find and the respective eigenfunctions are

m
s 
eikn a cos [kn x],
 (−) 
G(1)
f,i = √ T (−a) e−ik0 (x f +a) (2)
ψn (x) = 2
1   (even parity),
(29)
2
i~ k0 k fsw ′ (n) 
i fsw ieikn a sin [kn x], (odd parity),

 
× e−ik(xi +a) + R(−)
(−a) e
2ika −ik(a−xi )
e , (22)
which are the well-known even and odd solutions for the square
well potential [30].
m  
G(2)
f,i = eik(x f +a) + R(+)(a) e
2ika ik(a−x f )
e
i~2 k fsw
  6. Infinite well potential
× e−ik(xi +a) + R(−)(−a) e
2ika −ik(a−xi )
e , (23)
In this section, we consider the infinite well potential. Sim-
m  (+)  ilarly as in previous sections, the Green’s function for the infi-
G(3)
f,i = √ T (a) eik0 (x f −a) nite well potential can be obtained by letting u1 = u3 = ∞ and
i~2 k0 k fsw x1 = 0 and x2 = L, so that the potential function is given by
 
× e−ik(xi +a) + R(−)
(−a) e
2ika −ik(a−xi )
e , (24) 
0 for 0 < x < L,


V(x) =  (30)
where fsw = 1 − R(−) (+) 4ika
(−a) R(a) e . ∞ otherwise.


5.1. Calculation of the bound states As the potentials are infinite in x1 and x2 , the reflection am-
The eigenenergies are calculated from the condition fsw = plitude assume the value R(−) (+)
(0) = R(L) = −1 and a vanishing
0. By considering the symmetry of the potential, we can infer transmission amplitude. Thus, the Green’s function in Regions
that the reflections amplitudes are equal (up to a phase), i.e., 1 and 3 vanishes, so that ψ(1) (3)
n (x) = ψn (x) = 0. The Green’s
R(−) (+) function in Region 2 take the form
(−a) = R(a) = (k − k0 )/(k + k0 ), leading to the following tran-
scendental equation m  
G(2)
f,i = eikx f − eikL eik(L−x f )
2 i~2 k fiw
 kn − k0(n)  4ik a

 
  e n = 1. (25) × e−ikxi
− eikL e−ik(L−xi ) , (31)
kn + k0(n)
with fiw = 1 − e2ikL . After a straightforward algebra, the Eq.
Using Eq. (18) the residues are determined and leads to the
(31) can be written as
following bound state eigenfunctions:
Region 1: x < −a 2m
G(2)
f,i = sin [k(x f − L)] sin [kxi ]. (32)
v
t ~2 k sin [kL]
(n)
(1) 1 k0 2kn (n)
ψn (x) = ′ (n) (n)
e−ik0 (x+a) , (26) This result is indeed the exact one obtained through the spectral
i fsw kn k + kn 0 expansion of the Green’s function in Eq. (5) (cf. Appendix B).

5
 V (x)
u1
u3
E
u4
u2
x1 x2 x3
Figure 5: A rectangular potential able to support quasi-bound states as dis-
cussed in the text.
6.1. Calculation of the bound states
The poles are all contained in sin [kL], and eigenenergies of
the bound states are thus obtained from sin [kn L] = 0, i.e.,
nπ
kn = , n = 1, 2, 3, . . . .
L forever, even if E < u3 , as a consequence of the tunnel effect.
From the residues of the Green’s function, the correctly normal-
ized eigenfunctions corresponding for the bound states are thus
given by r quasi-energies. In the scattering of particles by such potential,
(2) 2  nπx 
ψn (x) = sin . (34)
L L
Indeed, these are the exact results for the bound states for the
infinite well potential.
7. Application on quasi-bound states
In this section, we apply the Green’s function approach of
previous Sections to extract information of systems with quasi-
bound states. A quasi-bound state occurs when a particle move
inside a system for a considerable period of time, leaving it
when a fairly long time interval τ has elapsed [29], where τ is
called lifetime of the quasi-bound state. The concept of quasi-
bound states is a fundamental one, and has been applied in all
areas of physics. They have been used to calculate tunneling
ionization rates [31], to understand the phenomenon of diffrac-
tion in time [32], to describe the decay of cold atoms in quasi-
one-dimensional traps [33], and are directly relevant to recent
condensed-matter experiments [34].
Let us consider the potential depicted in Fig. 5. Suppose
the potentials u1 and u3 were infinitely high. It would then be
possible for particles to be trapped inside the region for x1 <
x < x2 , i.e., the system would have genuine bound states, with
well defined energy E > 0. They are genuine bound states in the
sense that they are eigenstates of the Hamiltonian with a infinite
lifetime. From the Heisenberg uncertainty principle, ∆E∆t ≈ ~,
if the energy has null uncertainty its state’s lifetime is infinite.
In the situation of a finite barrier as in Fig. 5 (this is rough
rectangular approximation for the effective potential in a central
Transmission
G1
G2
x HqbL HqbL
E1 E2
Energy
Figure 6: Typical behavior for the transmission coefficient for a potential which
(qb) (qb)
has two quasi-bound states with energies E1 and E2 and widths Γ1 and Γ2 ,
respectively.
force problem V(r) plus the centrifugal barrier (~2 /2m)[l(l +
1)/r2 ] [30]), the particle can be trapped, but it cannot be trapped
(33)
The energy spectrum of these particle will be quasi-discrete,
and it consists of a series of broadened levels, whose width in
represented by Γ = ~/τ [14], and the energy values are called
the situation becomes very interesting when the incident energy
is close to the quasi-energy
E (inc) ≈ E (qb) . (35)
In this energy interval, the module square of the transmission
amplitude exhibits pronounced peaks, and this is called reso-
nant scattering [30]. In Fig. 6 it is depicted a typical transmis-
sion probability as a function of incident energy for a scattering
of a potential which supports quasi-bound states.
Now, let us consider the case of a finite square barrier at
x2 with an infinite barrier at x1 . In this situation, the system
can also has quasi-bound states, due the tunneling through the
right square barrier. The scattering eigenfunction for a particle
incident from the right is given by
1
ψ(x) ≈ √ (e−ik4 x + R(−)
pot e
ik4 x
), (36)
2π
where R(−)pot is the reflection amplitude of whole potential. By
analogy with the previous case, we would try to extract the
information from quasi-bound states from the reflection coef-
ficient R(−)
pot . Unfortunately, due to the potential to be infinite
at left, the reflection coefficient has the value |R(−) 2
pot | = 1 for
all range of energies. Thus, we cannot extract information of
quasi-bound states for this kind of potential by the above method.
So, we propose a Green’s function approach to extract informa-
tion of quasi-bound states for such kind of potential, as we ex-
plain below. Following the same steps described in the Section
4, the Green’s function for xi > x3 and x1 < x f < x2 is readily
6
 only interested in the position of quasi-energies and the width
of quasi-states.
1.0 For a numerical example, in the Fig. 7 is shown a graph
8 of |A|2 as function of energy for the potential in Fig. 5. For
Transmission
0.8

Reflection
0.6 the parameters we choose typical values for heterostructures in
6 0.4 GaAs [35]. The particle’s mass is m = 0.07me , where me is the
electron mass, u1 = ∞, u2 = u4 = 0, u3 = 0.23 eV, ℓ3 = 80 Å
 A¤2

0.2
4 0.0 and ℓ2 = 10.05ℓ3. As we can see, it is evident in the graph the
0 0.5 1.0 1.5 2.0 2.5 3.0
existence of quasi-bound states.
Energyu3
2 The approach presented here for rectangular potentials, can
be generalized for smooth potentials, but in this case it is nec-
essary the calculation of the classical action for the quantum
0
particle under the action of the potential. Specifically, we write
0 0.5 1.0 1.5 2.0 2.5 3.0
the amplitude as
Energyu3 Tb
A= , (42)
f
Figure 7: Behavior of |A|2 = |T b(−) / f¯qb |2 as function of the energy displaying the
presence of quasi-bound states. The parameters are u1 = ∞, u2 = u4 = 0, u3 = where T b is the generalized transmission amplitude for the smooth
0.23 eV, ℓ3 = 80 Å and ℓ2 = 10.05ℓ3 . In the inset is shown the transmission barrier between x2 and x3 ,
(solid line) and reflection (dashed line) coefficients for the potential barrier.
i
f = 1 − RR∞ exp [ S (x2 , x3 ; E)], (43)
~
obtained and is written as
where S (x2 , x3 , k) is the classical action, R is the generalized
m T b(−) ik4 (xi −x3 ) reflection amplitude for the barrier, and R∞ is the generalized
G f,i = √ e
i~2 k2 k4 fqb reflection amplitude for the infinity barrier at x1 . Since |R∞ |2 =
  1, we can write R∞ = exp [−iφ(E)], in such way that
× eik2 (x2 −x f ) + R(−)
1 e
ik2 (x f +x2 −2x1 )
, (37)
i
f = 1 − R exp [ S (x2 , x3 ; E) − iφ(E)]. (44)
with fqb = 1 − R(−) (+) ik2 ℓ2
1 Rb e . R(+) (−)
b and T b are the reflection and ~
transmission amplitudes for the potential barrier given, respec-
tively, by All those generalized amplitudes above are obtained by the pro-
cedure outlined in [16, 17].
(+) (+) T 2(−) T 2(+) R(+)
3 e
2ik3 ℓ3
Rb = R2 + , (38)
1 − R(−) 2 R3 e
(+) 2ik3 ℓ3
8. Conclusion
T 2(−) T 3(−) eik3 ℓ3
T b(−) = . (39) In this work, the exact Green’s functions for rectangular sin-
1 − R(−) (+) 2ik3 ℓ3
2 R3 e gle wells are obtained in a rather general way and by a sim-
For an infinite barrier at x1 , we have R(−) ple method. Our results are the exact ones and, although of
1 = −1, and the Eq.
(37) simplifies to the simplicity of the systems considered, Green’s functions for
such system are not so easy to obtain by standard procedures
(−) (for example, solving the inhomogeneous differential equation
2m T b ik4 (xi −x3 )
G f,i = √ e sin[k2 x f ], (40) in Eq. (1)). The procedures allows one to discuss complete ar-
i2 ~2 k2 k4 f¯qb
bitrary rectangular single wells and barriers, generalizing and
resuming results in the literature. For instance, by withdraw-
where f¯qb = 1+R(+)
b e
ik2 ℓ2
. From the interpretation of the Green’s
ing of the potential step at x1 by setting R(−) = 0, from Eq.
function as a probability amplitude (cf. Section 2), we can thus 1
(11) the Green’s function for square barrier of Ref. [11, 12]
interpret the term
is obtained. The method can be applied for general potentials,
T (−)
A= b , (41) including those multidimensional with radial symmetry, but in
f¯qb this case is necessary the calculation of the classical action for
in Eq. (40) as such amplitude. If the potential has quasi-bound the particle under the action of the potential [16, 17].
states, an incident wave with energy close to the quasi-energy, From the poles and residues of the Green’s function the
will have a high probability of tunneling, entering in the trap- bound state eigenenergies and eigenfunctions were obtained with
ping region. Consequently, a graph of |A|2 as a function of the the correct normalization constant. The determination of the
energy, will have peaks at each energy value close to E (qb) . So, later often involves a difficult integral in the other methods.
we can extract information of quasi-energies and its respective Finally, we also have discussed an application of the Green’s
widths from A. It is worthwhile to observe that the amplitude function approach to extract information from quasi-bound states
A is not normalized, but this is not a problem, because we are in systems which standard analysis of the quantum amplitudes
are not possible. The method could be generalized for smooth
7
potentials and applied to the well-known alpha decaying and Substituting the eigenfunctions in (B.1) into the spectral expan-
determination of the dwell times and will be subject of a future sion of the Green’s function, Eq. (5), we arrive at
work.  
∞
4mL X sin nX f sin (nXi )
G f,i = − 2 2 , (B.3)
Acknowledgments ~ π n=1 n2 + α2

The author would like to thank E. O. Silva, C. F. Woellner √

where X j = πx j /L, α2 = −L2 k2 /π2 and where k = 2mE/~.
and J. A. O. Freire, for critical reading the manuscript and help- Using a trigonometric identity for the product of sines, we have
ful discussions. This work was partially supported by the Fun-
dação Araucária (Brazil) under grant number No. 205/2013. ∞
2mL  X cos [n(X f − Xi )]
G f,i = −
~2 π2 n=1 n2 + α2
Appendix A. The quantum amplitudes for a step potential ∞
X cos [n(X f + Xi )] 
+ . (B.4)
The step potential is used as building block of our construc- n=1
n2 + α2
tion. So, in this appendix we just outline the derivation of quan-
tum amplitudes for the step potential. The potential function for The infinite sum above can be evaluated by using the identity
the potential step is given by 1.445-2 of Ref. [36], and after a straightforward algebra we
 achieve at
u j

 for x < 0,
V(x) =  (A.1) 2mL
u j+1 for x > 0.
 G f,i = − sinh [α(π − X f )] sinh [αXi ]. (B.5)
~2 πα sinh [απ]
The reflection and transmission amplitudes are obtained from
Now, by substitution of X j and α and using sinh [iθ] = i sin [θ],
solution of Schrödinger equation. The scattering solutions for we finally have the Green’s function for the infinite well poten-
the step potential with the incident beam coming from x = −∞
tial
are 
ik x (+) −ik x
e j + R j e j for x < 0,

 2m
ψ(x) =  (+) (A.2) G f,i = sin [k(x f − L)] sin [kxi ]. (B.6)
T e−ik j+1 x

j for x > 0, ~2 k sin [kL]
p
with k j = 2m(E − u j )/~. From the matching conditions at
References
the origin, i.e., ψ(0− ) = ψ(0+ ) and ψ′ (0− ) = ψ′ (0+ ), we find the
sought reflection and transmission amplitudes [1] D. S. Saraga, T. S. Monteiro, Phys. Rev. Lett. 81 (1998) 5796.
s doi:10.1103/PhysRevLett.81.5796.
k j − k j+1 k j+1 2k j [2] C. J. Bolton-Heaton, C. J. Lambert, V. I. Fal’ko, V. Prigodin, A. J. Epstein,
(+) (+) Phys. Rev. B 60 (1999) 10569. doi:10.1103/PhysRevB.60.10569.
Rj = , Tj = . (A.3)
k j + k j+1 k j k j + k j+1 [3] E. H. Hauge, J. A. Støvneng, Rev. Mod. Phys. 61 (1989) 917.
doi:10.1103/RevModPhys.61.917.
In the same way, for the case with the incident beam coming [4] F. Capasso, C. Sirtori, J. Faist, D. L. Sivco, S.-N. G. Chu, A. Y. Cho,
from x = ∞, we have Nature 358 (1992) 565–567. doi:10.1038/358565a0.
[5] S. Flügge, Practical Quantum Mechanics, Springer-Verlag, Berlin, 1999.
s [6] F. L. Scarf, Phys. Rev. 112 (1958) 1137–1140.
(−) (+) (−) k j 2k j+1 doi:10.1103/PhysRev.112.1137.
R j = −R j , Tj = . (A.4) [7] C. Weisbuch, B. Vinter, Quantum Semiconductor Structures: Fundamen-
k j+1 k j + k j+1
tals and Applications, Elsevier, 1991.
[8] R. Baltin, Z. Naturforsch 40a (1985) 379–382.
For the case E > u j+1 (E < u j+1 ), k j+1 is a real (imaginary) [9] P. Moretti, Phys. Rev. A 46 (1992) 1233–.
number and T (±)
j represents the transmission (penetration) am- doi:10.1103/PhysRevA.46.1233.
plitude. [10] T. O. de Carvalho, Phys. Rev. A 47 (1993) 2562–.
doi:10.1103/PhysRevA.47.2562.
[11] M. A. M. de Aguiar, Phys. Rev. A 48 (1993) 2567–.
Appendix B. Green’s function for the infinite well potential doi:10.1103/PhysRevA.48.2567.
[12] M. A. M. de Aguiar, Phys. Rev. A 51 (1995) 2654–.
from spectral expansion doi:10.1103/PhysRevA.51.2654.
[13] M. C. Gutzwiller, Chaos in Classical and Quantum Mechanics, Springer-
In this appendix, we will calculate the Green’s function for Verlag, New York, 1990.
the infinite well potential from the spectral expansion in Eq. [14] E. Merzbacher, Quantum Mechanics, third ed., Wiley, New York, 1998.
(5). The eigenfunctions and energies for a particle in the infinite [15] M. G. E. da Luz, E. J. Heller, B. K. Cheng, J. Phys. A 31 (1998) 2975.
well potential are given by doi:10.1088/0305-4470/31/13/007.
[16] M. G. E. da Luz, B. K. Cheng, M. W. Beims, J. Phys. A 34 (2001) 5041.
r doi:10.1088/0305-4470/34/24/303.
2  nπx 
[17] F. M. Andrade, B. K. Cheng, M. W. Beims, M. G. E. da Luz, J. Phys. A
ψn = sin , n = 1, 2, 3, . . . , (B.1)
L L 36 (2003) 227. doi:10.1088/0305-4470/36/1/315.
[18] B. R. Holstein, Am. J. Phys. 51 (1983) 897. doi:10.1119/1.13374.
n 2 π 2 ~2 [19] M. G. Rozman, P. Reineker, R. Tehver, Phys. Rev. A 49 (1994) 3310–
En = . (B.2)
2mL2 3321. doi:10.1103/PhysRevA.49.3310.

8
[20] A. G. M. Schmidt, B. K. Cheng, M. G. E. da Luz, J. Phys. A 36 (2003)
L545. doi:10.1088/0305-4470/36/42/L01.
[21] A. G. M. Schmidt, B. K. Cheng, M. G. E. da Luz, Phys. Rev. A 66 (2002)
062712. doi:10.1103/PhysRevA.66.062712.
[22] F. M. Andrade, M. G. E. da Luz, Phys. Rev. A 84 (2011) 042343–.
doi:10.1103/PhysRevA.84.042343.
[23] F. M. Andrade, M. G. E. da Luz, Phys. Rev. A 86 (2012) 042309–.
doi:10.1103/PhysRevA.86.042309.
[24] E. N. Economou, Green’s Functions In Quantum Physics, third ed.,
Springer-Verlag, German, 2006.
[25] L. S. Schulman, Techniques and Applications of Path Integrals, Dover
Publications, Mineola, New York, 2005.
[26] H. Kleinert, Path Integrals in Quantum Mechanics, Statistics, Polymer
Physics, and Financial Markets, fourth ed., World Scientific Publishing
Company, Singapore, 2006.
[27] H. Kleinert, I. Mustapic, J. Math. Phys. 33 (1992) 643–.
doi:10.1063/1.529800.
[28] A. R. H. Richard P. Feynman, Quantum Mechanics and Path Integrals:
Emended Edition, Dover Publications, 2010.
[29] L. D. Landau, E. M. Lifschitz, Quantum Mechanics, Pergamon, Oxford,
1981.
[30] J. J. Sakurai, J. Napolitano, Modern Quantum Mechanics, 2nd ed. ed.,
Addison-Wesley, 2011.
[31] G. N. Gibson, G. Dunne, K. J. Bergquist, Phys. Rev. Lett. 81 (1998) 2663.
doi:10.1103/PhysRevLett.81.2663.
[32] E. Torrontegui, J. Muñoz, Y. Ban, J. G. Muga, Phys. Rev. A 83 (2011)
043608. doi:10.1103/PhysRevA.83.043608.
[33] A. del Campo, J. G. Muga, Europhys. Lett. 74 (2006) 965.
doi:10.1209/epl/i2006-10061-5.
[34] M. Sato, H. Aikawa, K. Kobayashi, S. Katsumoto, Y. Iye, Phys. Rev. Lett.
95 (2005) 066801. doi:10.1103/PhysRevLett.95.066801.
[35] A. P. Jauho, M. Jonson, J. Phys.: Condens. Matter 1 (1989) 9027.
doi:10.1088/0953-8984/1/45/025.
[36] I. S. Gradshteyn, I. M. Ryzhik, Table of Integrals, Series, and Products,
seventh edition ed., Academic Press, 2007.