Revista Brasileira de Ensino de Física, vol.

46, e20230316 (2024) Articles

www.scielo.br/rbef cb
DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Licença Creative Commons

Did you really miss a Delta function here?

Você esqueceu mesmo uma função Delta?

M. Amaku1,2 , F.A.B. Coutinho*1,3 , P.C.C. dos Santos1 , E. Massad1,3

1
Universidade de São Paulo, Faculdade de Medicina, São Paulo, SP, Brasil.
2
Universidade de São Paulo, Faculdade de Medicina Veterinária e Zootecnia, São Paulo, SP, Brasil.
3
Fundação Getúlio Vargas, Escola de Matemática Aplicada, Rio de Janeiro, RJ, Brasil.

Received on October 23, 2023. Revised on December 20, 2023. Accepted on December 22, 2023.
This is the third article in a series dedicated to examining the use of the delta function in the physics literature.
Here we discuss the appearance of a delta function in the second derivative of the wave function if its derivative
is discontinuous. When this happens it is often claimed in the literature that the proposed wave function must
be eigenfunction of an operator that contains a delta function and, in some cases, it is rejected. We study in this
paper what we believe are alternatives ways to deal with this situation.
Keywords: Dirac delta function, time independent Schrodinger equation.

Este é o terceiro artigo de uma série dedicada a examinar o uso da função delta na literatura física. Aqui
discutimos o aparecimento de uma função delta na segunda derivada da função de onda se a sua derivada for
descontínua. Quando isto acontece, é frequentemente afirmado na literatura que a função de onda proposta deve
ser uma função própria de um operador que contém uma função delta e, em alguns casos, é rejeitada. Neste artigo
estudamos o que consideramos serem formas alternativas de lidar com esta situação.
Palavras-chave: Função delta de Dirac, equação de Schrodinger independente do tempo.

1. Introduction of functions that is the domain of the operator. From

this point of view the boundary conditions are included
The use of the Dirac delta function in an intuitive way automatically in the definition of the operator domain
is found very often in the literature, more often in the and so the boundary conditions are not arbitrary. As
literature dedicated to the teaching of Physics. But the we shall see the domain of the operator must be such
intuitive use of the Dirac delta function may lead to that makes it self- adjoint. We shall expand on this on
some wrong statements as we have been discussing in section 4 of this paper.
this series of papers [1, 2]. In this paper we present yet Suppose someone, following the first point of view,
another interesting problem that can lead to errors. solves the time independent Schrodinger (TISE) equa-
There are two ways to view the time independent tion and finds a solution that has, for example, the first
Schrodinger (TISE). Take for example the TISE for a derivative discontinuous. Then one can hear (awfully
particle moving on the line under the influence of some frequently) that since “as is well known the second
potential V (x) derivative will have a delta function at the point of
discontinuity and so the TISE equation is not satisfied”.
Hψ = Eψ (1) There is an immediate problem with this statement.
or more explicitly As we will see, if the potential in the TISE has an infinite
discontinuity, then the derivative of the solution of the
ℏ2 d2
 
TISE equation may have a discontinuity. But potentials
− + V (x) ψ(x) = Eψ(x). (2)
2m dx2 with discontinuities are very common in the examples
presented to students. We shall see some examples in
The most common point of view found in the literature this paper but to warm, from the top of our heads,
is to consider that the TISE is a differential equation that up we can say that the infinite square well potential
must be solved satisfying some mysterious boundary is an example (see below). In section 2 we present
conditions and finding out the values of E for which the an overview of the papers dealing with discontinuities
solution satisfy these boundary hconditions. i in the potential and its effect on the wave function.
ℏ2 d2
Another one is to consider the − 2m dx 2 + V (x) as an Then in section 3 we present some examples where
operator acting on functions that belong to some class the discontinuity of the first derivative is not obvious.
In section 4 we briefly review operator theory needed
* Correspondence email address: coutinho@dim.fm.usp.br to understand the arguments presented in section 5.

Copyright by Sociedade Brasileira de Física. Printed in Brazil.

e20230316-2 Did you really miss a Delta function here?

We show that the sentence in the title of this paper require only that the operator H in its domain to be
should be answered “Really, you got to be kidding. Not self-adjoint.
so fast because maybe you are the one that is wrong.” The paper by T. Cheon and T. Shigehara [6] discuss
In fact, from the second point of view, as we shall see in in simple language how to produce wave functions that
detail, the domain of the operator may include functions are discontinuous without violating self-adjointness.
or derivatives of functions with the discontinuity and
there is no need to talk about delta function. 3. Examples of Questionable Statements
The reader can easily work out the case of the infinite
square well potential defined 3.1. The infinite spherical square well [7]

∞ for − ∞ < x < −L The radial part of the TISE for the infinite square well,
V (x) = 0 for − L ≤ x ≤ L . (3) of radius a, in polar coordinates is
∞ for L < x < ∞

d2 Rℓ (r) 2 dRℓ (r)
 
2 ℓ(ℓ + 1)
+ + k − Rℓ (r) = 0, (4)
This potential has discontinuities at x = −L and d2 r r dr r2
x = L and the first derivative of the wave function q
has discontinuities at those points. So naïve use of the where k = 2m E h , 0 < r < a, and its general solution
Dirac function concludes that the usual solution, found is [7]
in almost every book on quantum mechanics, is not
a solution of the TISE because the second derivative Rℓ (r) = Ajℓ (kr) + Bnℓ (kr), (5)
of the wave functions has delta functions at −L and
L. (If the reader is intrigued by this he/she can see a where jℓ (kr) and nℓ (kr) are the spherical Bessel and
solution, different from the considerations carried out in spherical Neumann functions of order ℓ. For ℓ ̸= 0
this paper, in [2]). the Neumann solution is unacceptable because it is not
Other more complicated case will be discussed later, square integrable, but for ℓ = 0 the Neumann solution
but first let’s discuss some other results found in the is integrable.
literature about the wave function at discontinuities of In the literature there are innumerous arguments for
the potential. abandoning the ℓ = 0 Neumann solution. Most of these
arguments rest upon the claim that the ℓ = 0 Neumann
2. Some Results Found in the Literature wave function produces a delta function in the origin.
We shall examine other arguments at the end of this
About the Behavior of the Wave section.
Function and its Derivative at The arguments involving the delta function were put
Singular Points forward by Mohammad Khorrami [8], Antonio Prados
and Calos A. Plata [9] and by Jorge Munzenmayer
The first result we want to discuss is given in a beautiful and Derek Frydel [10]. The argument is that when we
paper by David Branson [3]. He discusses the continuity calculate the Laplacian of the Neumann function
of the wave function and of its derivative at points where  
the potential is discontinuous and concludes the wave 2 2 cos(kr)
− ∇ ((n0 (kr)) = −∇
functions must be continuous. His complete result is the kr
justification of the two results found in the literature: 2
−∇ (cos(kr))
= − 2(∇ cos(kr))
a) If the discontinuity is finite then both the wave  kr   
function and its derivative are continuous 1 2 1
· ∇ − cos(kr)∇
b) If the discontinuity is infinite the wave function kr kr
must vanish at this point k cos(kr) 4π 4π
= + δ(r) = k 2 n0 (kr) + δ(r) (6)
r k k
These results are discussed by M. Andrews [4] from
a more physical point view, but we consider Branson’s and hence the appearance of the 4π k δ(r) shows that
paper [3] more general. n0 (kr) is not a solution of the Schrodinger equation.
The second result is by D. Home and S. Sengupta [5]. In fact, equation (4) is the result of transforming to
They examine the discontinuity of the first derivative of spherical polar coordinates. Since this transformation is
the wave function in a few cases by imposing that the singular at the origin, we can define the domain of the
momentum operator must be self-adjoint. He exam- operator to be functions on the open interval (0, ∞)
ines the cases of the infinite square well potential, the that are square integrable in this interval. Note that
Dirac Delta function potential and the one-dimensional to include zero do not alter the value of the integral
Coulomb Potential. that proves that this solution is square integrable, tech-
These results however are not general enough and not nically L2 (0, ∞) = L2 [0, ∞). Taking the open interval
totally correct. The principles of quantum mechanics eliminates the delta function. If, however we include

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-3

the zero for the functions that are the domain of the essential based on the fact that the equation in rectan-
operator then self-adjointness demands R(r = 0) = 0, gular coordinates do include this point, equation (12)
but if we exclude the origin other self-adjoint extensions becomes
are possible. This is a rather technical and mysterious  2 
d 2 d 1
point that will be discussed in the section 5 of this paper. + = −4πδ 3 (r), (14)
dr2 r dr r
3.2. Singular behavior of the Laplace operator in  2
d 2 d

because the operator dr 2 + r dr is the radial part of
Polar spherical coordinates
the Laplacian and as is well known ∇2 1r = −4πδ 3 (r).


In a series of papers [11–13], the authors claim that not Then equation (11) becomes
only the solutions of the TISE equation may have unno-
1 d2
 
ticed delta functions, but the differential expressions of 2m u
the equation may have “unnoticed” delta functions. u(r) − u(r)4πδ 3 (r) + 2 (E − V (r)) = 0.
r dr2 ℏ r
They consider the transformation of the Laplacian (15)
in cartesian coordinates to polar spherical coordinates Note that at this point we cannot multiply equa-
and conclude that unless you impose an extra-condition tion (15) by r to take advantage of the fact that
to the radial part of the wave function you get a non- rδ 3 (r) = 0 cannot be used because we would get 0 = 0.
solution. They notice that the solution to the TISE in One is then forced, according to the authors, to conclude
polar coordinates can be written in two forms, viz. that u(r = 0) = 0 to eliminate the spurious delta.
Again, the problem here is that the transformation to
Ψ(r) = Rℓ (r)Υm
ℓ (θ, φ) (7) polar coordinates is singular. As show in the section 4
of this paper, the use of polar coordinates essentially
or we can write
excludes the origin and then we can or cannot put the
uℓ (r) m delta function and if we decide to put it, we eliminate it
Ψ(r) = Υℓ (θ, φ). (8) by requiring the wave function vanishes at the origin. In
r
fact, we must require that limr→0 u(r) = 0.
In the above expressions Υm ℓ (θ, φ) are spherical har-
monics. The problem, according to the authors appears 3.3. The infinite spherical square well again
when you try to write an equation for uℓ (r). The
equation for Rℓ=0 (r) = R(r) is A paper by Jorge Munzenmayer and Derek Frydel [10]
also consider the Neumann solution of the TISE and find
d2 R 2 dR 2m equation (6) and conclude that the Neumann solution is
+ + 2 (E − V (r))R = 0. (9)
dr2 r dr ℏ not a solution of the TISE. But then they ask themselves
what is the equation that has the Neumann function as
uℓ=0 (r) u(r)
Making the substitution R(r) = r = r , we get solution. To answer this, they manipulate equation (6)
as follows
d2
   2 
1 2 d d 2 d 1
+ u(r) + u(r) + 
4π δ(r)

r dr2 r dr dr2 r dr r −∇2 − n0 (kr) = k 2 n0 (kr). (16)
du 1 2m u k n0 (kr = 0)
+2 + 2 (E − V (r)) = 0. (10)
dr r ℏ r Since as r approaches r = 0 the Neumann function
1
The first derivatives disappear, and we are left with n0 (kr) behaves as kr they write equation (16) as

−∇2 − 4πrδ(r) n0 (kr) = k 2 n0 (kr).

 
1 d2
   2
d

2 d 1 (17)
u(r) + u(r) +
r dr2 dr2 r dr r To conclude that the origin of the divergence in n0 (kr)
2m u as r approaches zero is that there is a hidden potential
+ 2 (E − V (r)) = 0. (11)
ℏ r for r = 0, which is
The term
V (r) = −4πrδ(r). (18)
d2 1 d2
   
2 d 1 1
2
+ = r = 0. (12) They argue that since rδ(r) = 0 this potential should
dr r dr r r dr2 r
be considered invisible. But, since the matrix element of
So that we get this V (r) is infinite, viz.
Z ∞
d2 u(r) 2m 2 1
+ 2 (E − V (r))u(r) = 0. (13) ⟨|V |⟩ = (−4πrδ(r)) (n0 (kr)) r2 dr = −2πC lim ,
dr2 ℏ 0 r→0 r
(19)
But the relation (12) is only valid for r ̸= 0. If we where C is a normalization constant, we can argue that
include the point r = 0, that the authors claim to be this matrix element diverges, and that this divergence

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024
e20230316-4 Did you really miss a Delta function here?

exactly cancel the divergence of the kinetic energy, which function. But to be an operator we must specify the
they consider a sign that the potential (18) should be domain, that is the set of function where it is allowed to
taken seriously. Nevertheless, they consider the (18) act. In quantum mechanics the space of functions of a
potential as unphysical and therefore the Neumann solu- system is a Hilbert space.
tion should be rejected because, according to them, it is Consider a set of complex valued functions ψ(x)
this unphysical potential that produces its divergence. defined in an interval [a, b]. Later we are going to
As can be seen the argument in paper [10] is a hybrid consider cases where a or b can be ∞ or both a = −∞
of the two arguments discussed above. It rejects the and b = +∞. The “scalar product” of two such a
Neumann solution because it diverges at the origin and function in this set, φ(x) and ψ(x), is defined as
because it is produced by a potential that they discuss Z b
and conclude is not physical.
(φ, ψ) = φ∗ (x)ψ(x)dx. (21)
A more sophisticated argument to eliminate the singu- a
lar solution is to say as noted by [10] that the expectation
value of the kinetic energy with such a function (n0 (kr)) The set of function for which (21) is finite (with some
is infinite. This is a very serious argument and can be others technical conditions not important for us now) is
found early in the classical book by S. Flugge [14]. called square integrable and is denoted L2 (a, b).
However, we can circumvent this argument by using Consider now the linear, second order, “differential
other self-adjoint extensions that are different from zero expression”, already mentioned in (20) and repeated
at the origin. The physical significance of these self- here.
adjoint extensions is that we renormalize the kinetic d2 d
energy by adding a point interaction (also known as a2 (x) + a1 (x) + a0 (x). (22)
dx2 dx
zero range interaction, but not a delta function) to the
Hamiltonian [15]. But as show in the section 4 of this This “differential expression” when acting on a func-
paper this is equivalent to finding self-adjoint extensions tion, say f (x) produces another function, say g(x),
of the Laplacian when we exclude the origin, so that the that is
two solutions are acceptable.
The example of the biharmonic oscillator is laborious d2 f (x) df (x)
L(f (x)) = a2 (x) 2
+a1 (x) +a0 (x)f (x) = g(x).
and so it is presented later in the paper as a final dx dx
(23)
example. It is better to read this final example after
The “differential expression” (23) is the action of
reading the two sections that follow.
an operator which is defined by this action and by a
set of functions, the domain of the operator so that
4. Finding the Domain of the Operators if f (x) belongs to the domain the set of functions
Associated with a Differential g(x) constitute the range of the operator. The domain
Expression is usually given by boundary conditions that the
functions and its first derivatives satisfy at the end
In this section we quickly recapitulate some operator points, x = a, and x = b, plus some continuity conditions
theory that we need to explain in more detail the on f (x) and dfdx
(x)
.
second point of view regarding the time independent The adjunct of this operator, annotated as L+ (f1 (x)),
Schrodinger equation presented in the introduction. is given by another “differential expression” that acts on
According to this point of view the TISE is just a function f1 (x) as follows
a search for the eigenvalues and eigenfunctions of an
operator which is something that acts on function and d2
L+ (f1 (x)) = (a2 (x)f1 (x))
produces another function, as explained below. dx2
An operator consists of an action (what it does d
+ (a1 (x)f1 (x)) + a0 (x)f1 (x) = g1 (x)
to function where it acts) and a domain that is the dx
(24)
specification of a set of function where it acts.
A differential expression of order two in one dimension
and the set of functions f1 (x) that constitute its domain.
acting on a function ϕ(x) produces another function
This set of function is not arbitrary, as we show below.
ψ(x). It is an object like
In fact, we have that, given the scalar products
d2
 
d Z b
O{ϕ(x)} = a2 (x) 2 + a1 (x) + a0 (x) ϕ(x) = ψ(x).
∗
(L+ u, v) = L+ u(x), v(x) = L+ u(x) v(x)dx


dx dx
a
(20) (25)
It is clearly linear, because O{a ϕ1 (x) + b ϕ2 (x)} = and
a O{ϕ1 (x)} + b O{ϕ2 (x)} where a and b are numbers.
A differential expression is the action of a differ-
Z b

ential operator, that is, what it does when acting on (u, Lv) = (u(x), L(v(x))) = u∗ (x)L(v(x))dx. (26)
a

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-5

We have that Lagrange formula regardless of the values of ϕ2 (x) at x = a and x = b.

So, the domain of the adjunct is larger than the domain
(L+ u, v) − (u, Lv) = J(u, v) (27) of the operator and hence O1 is not self-adjoint, that is
O1 ̸= O1+ and the operator is just symmetric. Of course,
where
since we used integration by parts, we must impose some
du(x) d(a2 (x)) conditions on both functions.
J(u, v) = a2 (x) − u(x) + a1 (x)u(x)v(x)
dx dx Example (2) The second operator has as domain
(28)
functions ϕ1 (x) that vanish at a and b. This operator
is obeyed. This follows from integration by parts, so
is self-adjoint. In fact, let ϕ2 (x) be the functions in
we must restrict ourselves to functions u(x) so that
the domain of the adjunct. Then integration by parts
integration by parts can be carried out.
(Lagrange formula) show that
Now we insist that the domain of L+ are all the
functions u(x) that make (O2+ ϕ2 , ϕ1 ) − (ϕ2 , O2 ϕ1 )
Z b ∗
J(u(b), v(b)) − J(u(a), v(a)) = 0 (29) ℏ2 d2 ϕ2 (x)
= − ϕ1 (x)dx
a 2m dx2
for all the function v(x) of the domain of L. This is the Z b
ℏ2 d2 ϕ1 (x)
 
definition of the domain of L+ . Note that the action −
∗
(ϕ2 (x)) − dx
and the domain of L+ are usually different from the a 2m dx2
  ∗ 
action and the domain of L. ∗ dϕ1 (a) dϕ2 (a)
But the action of L+ and L can be the same. For = (ϕ2 (a)) − ϕ1 (a)
dx dx
ℏ2
example, if we take a2 (x) = − 2m and a1 (x) = a0 (x) = 0   ∗ 
∗ dϕ1 (b) dϕ2 (b)
we have that both L+ and L have the same action, − (ϕ2 (b)) − ϕ1 (b) = 0,
ℏ2 d2 dx dx
namely − 2m dx2 . But we must examine the domains. If (31)
the domain of L+ is not the same domain of L (but by
definition obeys (29)) then the operator is called Her- when ϕ2 (x) vanishes at x = a and x = b regardless
mitian by Physicists (symmetric by mathematicians). of its derivative. So, both O2+ and O2 has the same
If the domains are the same the operators are called action and the same domain, and hence O2 = O2+ and
self-adjoint. By the same domain we mean that if we the operator is self-adjoint. To carry the integration by
change the functions of the domain of the operator, we parts and some other technicalities we must also demand
must make a different change on the functions in the that the functions and the first derivatives are absolutely
domain of the adjunct. Then the domains of L and L+ continuous (see below) so that the second derivative
become different because the change is not the same. See belong to L2 (a, b).
the examples below.
Remark 1 Crudely, a function is absolutely continuous
4.1. Examples in an interval (a, b) if we can write it as
2 2 Z x
ℏ d
Consider two operators with the same action is − 2m dx2 df (x)
f (x) = f (a) + dx. (32)
and whose domain are specified in the examples. We call a dx
O1 and O2 the first and second operators respectively.
Technically dfdx
(x)
may exist only almost everywhere,
Example (1) The first operator O1 has as domain
meaning that it may not exist in some points.
functions ϕ1 (x) that vanish together with its derivative
at x = a and x = b. This operator is symmetric but
not self-adjoint. In fact, let ϕ2 (x) be the functions in Example (3) Consider the operator whose action
d
the domain of the adjunct. Then integration by parts is −iℏ dx . Impose that the domain are differentiable
(Lagrange formula, see above) show that functions ψ(x) defined in an interval [a, b] such that
Rb
a
|ψ(x)|2 dx is finite, that is, the functions ψ(x) belongs
(O1+ ϕ2 , ϕ1 ) − (ϕ2 , O1 ϕ1 ) to L2 (|a, b|). Now assume that in addition the domain
Z b ∗
ℏ2 d2 ϕ2 (x) are functions with ψ(a) = ψ(b) = 0. In this domain the
= − ϕ1 (x)dx operator is symmetric but not self-adjoint. To see this
a 2m dx2
Z b calculate
ℏ2 d2 ϕ1 (x)
 
∗
− (ϕ2 (x)) − dx Z b  
2m dx2 d
a
∗ ψ2∗ (x) −iℏ ψ1 (x) dx

∗ dϕ1 (a)

dϕ2 (a)

a dx
= (ϕ2 (a)) − ϕ1 (a)
dx dx = ψ2∗ (b)ψ1 (b) − ψ2∗ (a)ψ1 (a)
  ∗  Z b 
∗ dϕ1 (b) dϕ2 (b) d ∗
− (ϕ2 (b)) − ϕ1 (b) = 0 (30) + −iℏ ψ2 (x) ψ1 (x)dx. (33)
dx dx a dx

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024
e20230316-6 Did you really miss a Delta function here?

Now if ψ1 (x) is in the imposed domain we have Example: The operator with the same action,
d
Z b   −iℏ dx , operating on functions defined on the
∗ d interval [a, b] such that ϕ(a) = ϕ(b) = 0 have
ψ2 (x) −iℏ ψ1 (x) dx
a dx deficiency indices (1, 1) and modification of original
Z b
d
 boundary conditions, ϕ(a) = ϕ(b) = 0, are needed.
= −iℏ ψ2∗ (x) ψ1 (x)dx, (34) New boundary conditions, which we will see latter
a dx
how to obtain, are ϕ(a) = eiθ ϕ(b), they produce
regardless of the values of ψ2 (x) at the points x = a an infinite number of operators one for each dif-
and x = b. Therefore, the operator is symmetric but not ferent values of θ. They are all different from the
self-adjoint. But, if we take the domain of the operator original which had the domain fixed by boundary
to be functions such that ψ(a) = eiθ ψ(b) for arbitrary conditions ϕ(a) = ϕ(b) = 0.
real θ, 0 < θ < 2π, then it is easy to see that within 3) If n+ ̸= n− the operator cannot be made self-
d adjoint.
this domain, the operator whose action is −iℏ dx is self-
adjoint. 4) Finally to get the boundary conditions we must
The examples show that we need some criteria to find find a unitary transformation connecting the solu-
if an operator which is symmetric, is in fact self- adjoint tions of equations (35) and (36). This procedure is
or, more importantly, if it can be modified so it becomes illustrated in the examples that follow.
self-adjoint. This is given by the following theorem due
to von Neumann. This first example illustrate how we can use the
Let H be a symmetric differential operator such that formalism of self-adjoint extension to interpret the quan-
tum mechanics of a particle moving in the real axis form
its domain, D(H), are functions ϕ(x) and dϕ(x)
dx are such
2 which the zero was singled out or removed. A few more
the d dx
ϕ(x)
2 exists, that is ϕ(x) and dϕ(x)
dx are absolutely complicated examples can be found in [16].
continuous.
To see if H is self-adjoint in this domain we search Example (4) The delta function potential as a self-
for the independent square integrable solutions of the adjoint extension.
d2
differential equation Consider the operator − dx 2 in the following domain:
2
f (x) and dfdx(x)
are continuous and d dx f (x)
belongs to
H + ϕ+ (x) = iκϕ+ (x)
2
(35)
L2 (−∞, +∞) and f (0) = 0. This last condition is
and essential to this example because the point x = 0 was
singled out. (In the next example we consider the case
H + ϕ− (x) = −iκϕ− (x), (36) where we create a hole in the line by saying that at x = 0
the value of the wave function is not specified, that is we
where H + is the action of the adjoint of H, which for remove the point x = 0 from the real line.)
simplicity we are assuming here are the have the same This above operator is symmetric. To calculate the
action as H. The operator H is assumed to be symmetric domain of its adjoint we integrate its action between a
because the adjunct may have different domains. The function φ∗2 (x) of is adjoint and a function φ1 (x) of its
constant κ was introduced to maintain the dimension of domain. We have, integrating by parts
the equation.
∞ Z ∞
d2 φ∗2 (x)
Z    2 
Now let n+ and n− (called deficiency indexes) be ∗ d φ1 (x)
− φ1 (x)dx − φ2 (x) − dx
the number of linearly independent solutions of (35) −∞ dx2 −∞ dx2
and (36), respectively. Then dφ1 (0− ) dφ1 (0+ )
= φ∗2 (0− ) − φ∗2 (0+ )
1) If n+ = 0 and n− = 0 then the operator H in dx dx
its original domain is essentially self-adjoint. No dφ∗2 (0− ) − dφ∗2 (0+ )
− φ1 (0 ) + φ1 (0+ ). (37)
boundary conditions are needed. dx dx
d
Example: Take the operator −iℏ dx acting on func- Setting the second member equal to zero and remem-
tions ϕ(x) defined in the interval [−∞, +∞] and bering that φ1 (0− ) = φ1 (0+ ) = φ1 (0) = 0 and that
such that ϕ(−∞) = ϕ(+∞) = 0 and ϕ(x) is square dφ1 (0− ) (0+ )
d = dφ1dx we have that the domain of the adjoint
integrable. The equations −iℏ dx χ(x) = +iκχ(x) dx
d is given by functions φ2 (x) whose derivatives satisfies
and −iℏ dx χ(x) = −iκχ(x) have no solution in
L2 (−∞, +∞) and so the operator is essentially dφ∗2 (0+ ) dφ∗2 (0− )
self-adjoint and no extra conditions are needed. − = δ, (38)
dx dx
2) If n+ = n− ̸= 0 the operator is not self-adjoint
but we can construct self-adjoint operators with where δ is a real number and the functions themselves
it, usually by specifying the boundary conditions. satisfy
The resulting operator is called a self-adjunct
extension of the original symmetric operator. φ∗2 (0− ) = φ∗2 (0+ ) = φ∗2 (0). (39)

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-7

2
d
In words, the domain of the adjoint are functions that are namely − dx 2 + gδ(x), and by doing standard manip-

continuous at x = 0 but with derivatives discontinuous ulations.

at x = 0. The domain of the adjunct is different from the This is our main result: The problem can be solved by
domain of the original operator. Therefore, the original saying that we have an operator with an action and a
operator is not self-adjoint. So let us use von Neumann domain, or you can say, no, you have an action plus a
theory and see if it has extensions that are self-adjoint. delta function. This second interpretation requires that
According to the above recipe we must ask if the you promote the wave functions and everything else to
equations generalized functions if you want to be mathematically
rigorous (see [2]). Furthermore, in some cases it takes
d2 Ψ± (x) a lot of work to identify the delta function interactions
− = ±iχΨ± (x) (40)
dx2 that reproduces the boundary conditions, as we shall
has solution belonging to L2 (−∞, ∞), that is square see in the next example, and in more complicate case
integrable. the interaction to be added to the action is not a delta
Each equation has a solution each, viz function as we shall see further in this paper.
  π 1  Example (5) A free particle in the real line with the
−i
 e 4 χ2 x
 point x = 0 removed.
 e for − ∞ < x < 0
Ψ+ (x) =  π 1  (41) In this example we examine the operator whose action
−i d2
 − e 4 χ2 x

 is again − dx 2 whose domain are function uabcd (x) which
e for 0 < x < +∞ vanishes from −∞ to x = a < 0, are continuously and
and infinitely differentiable between x = a and x = b with
  π 1  uabcd (a) = uabcd (b) = 0, vanish between x = b < 0 and
i
x = c > 0 (containing the origin) and is continuously
e e χ x
 4 2
for − ∞ < x < 0

Ψ− (x) =   (42) differentiable from x = c to x = d with uabcd (c) =
 − ei π4 χ 12 x
 uabcd (d) = 0 and finally vanished for x = d to x = ∞.
e for 0 < x < +∞ This domain in called by mathematicians C0∞ (R/0). The

points x = b and x = c can be arbitrarily close to zero
Since each equation has one solution, we have that so that these functions are zero in an arbitrarily small
the deficiency indices are (1, 1). So, the original operator interval [b, c]. The point x = 0 was thus removed from
has a one parameter, let’s call it α, family of self-adjoint the domain of the operator.
operators that are extension of it. To find this family The operator is clearly symmetric because if φ1 (x) is
we use the following prescription: We begin by finding a in its domain then
unitary transformation that connect the solutions with Z ∞ 2 ∗  Z ∞  2 
iχ with the solutions with −iχ. Since we have just one d φ2 (x) ∗ d φ1 (x)
− φ1 (x)dx − φ 2 (x) − dx
solution for each of the equations (4.13) the unitary −∞ dx2 −∞ dx2
transformation connecting then is just a phase eiα . dφ1 (0− ) dφ1 (0+ )
Let φ(x) be a function on the domain we are seeking. = φ∗2 (0− ) − φ∗2 (0+ )
dx dx
We impose dφ∗2 (0− ) dφ ∗ +
(0 )
− φ1 (0− ) + 2
φ1 (0+ ) (46)
Z +∞  2
d (Ψ+ (x) + eiα Ψ− (x))∗
 dx dx
− φ(x)dx φ1 (0− ) = φ1 (0+ ) = 0 (47)
−∞ dx2
Z +∞  2 
d φ(x) and
= (Ψ+ (x) + eiα Ψ− (x))∗ − dx
−∞ dx2 dφ1 (0− ) dφ1 (0+ )
(43) = =0 (48)
dx dx
and using equation
(37) with φ∗2 (x) replaced with because φ1 (x) is continuous and infinitely differentiable.
∗
Ψ+ (x) + eiα Ψ− (x) and φ1 (x) by φ(x) we have Therefore, the right-hand side of the above equation
(see [16, p. 208]) vanishes independently of φ2 (x) and so, the domain of
the adjunct is larger than the domain of the operator.
φ(0− ) = φ(0+ ) = φ(0) (44) In fact, the domain of the adjunct are functions φ2 (x)
that vanishes when x → ±∞, are not defined at x = 0,
and
are square integrable from −∞ to +∞ and have square
1 π α


dφ(0+ ) dφ(0− ) 2α 2 cos 4
+ 2
integrable second derivatives.
− =− α φ(0) = gφ(0), According to the von Neumann theory to see if this
dx dx cos 2
(45) operator has self-adjoint extensions we investigate the
where g is a real arbitrary number. number of solutions of the equations
The boundary conditions (44) and (45) can be d2 Ψ± (x)
obtained by adding to the operator a delta function, − = ±iχΨ± . (49)
dx2

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024
e20230316-8 Did you really miss a Delta function here?

But now, because we excluded the point x = 0, there are Roman letters a, b, c, d and to the constant χ by
four solutions. "
#
√ 1/2 cos a + π4 − cos b sin c − π4


 β = 2χ
0 for − ∞ < x < 0 sin b
Ψ1+ (x) = (50)
 π 1

e−i 4 √
 
 − x χ2 sin a − cos b cos c
e for 0 < x < +∞ α = 2χ1/2
  π 1  sin b
 − e−i 4 χ 2 x √ 1

cos a + cos b cos c

Ψ2+ (x) = e for − ∞ < x < 0 (51) δ = 2 1/2
χ sin b
0 for 0 < x < +∞

"
#
√ 1 cos a + π4 + cos b sin c + π4



0 for − ∞ < x < 0 γ = 2 1/2
χ sin b
Ψ1− (x) = (52)
 π 1 
+i
 − e 4 χ2 x
e for 0 < x < +∞ θ = d.
  π 1 
 − e+i 4 χ 2 x The physics of the Hamiltonian defined by the action
Ψ2− (x) = e for − ∞ < x < 0 (53) d2
− dx whose domain are functions in the holed real
2
0 for 0 < x < +∞

line with boundary conditions given by (57) was given
in [17, 18]. Note that until this point, we didn’t speak
Therefore, the deficiency indices are (2, 2) and so there of zero range interactions or more specifically, in this
is a four parameter self-adjoint extensions of the original case, of delta functions. But if the reader wishes he/she
operator. can see interpret the above result as a Hamiltonian
d2
The unitary operator connecting the two solutions with the action − dx 2 plus a contact interaction (or

is now a 4 × 4 unitary matrix Depending on four zero range interactions) that includes not only the delta
parameters viz. function but also derivatives of the delta function. The
complete result can be found in a beautiful paper by S.

u11 u12
 
cos bei(a+b)
i sin bei(d−a)
 De Vincenzo and C. Sánchez [19] and will be reproduced
U= = . latter in this paper.
u21 u22 i sin bei(d+a)
cos bei(c−a)
(54) The above procedures are only prescriptions, that
The boundary conditions are obtained by enforcing are described in more detail in [16]. The reader can
that if φ(x) is a function in the domain of the operator find detailed treatments in the following reference in
we seek then ascending order of mathematical complexity: [20], [21]
and [22].
∞
"
∗ # Now we can proceed and examine the problems raised
d2 Ψ1+ (x) + u11 Ψ1− (x) + u12 Ψ2− (x)
Z
− φ(x)dx in the literature and described above in section 3 and in
−∞ dx2 section 6.
Z ∞  2 
d φ(x)
= (Ψ1+ (x) + u11 Ψ1− (x) + u12 Ψ2− (x))∗ − dx
−∞ dx2 5. Alternative Solutions to the
(55) Questionable Statements Presented
Before in Section 3
and
Now we are ready to decipher as promised the mystery
of how transforming the Laplacian operator, −∇2 , from
∞
"
2

∗ #
d Ψ1+ (x) + u21 Ψ1− (x) + u22 Ψ2− (x)
Z
− φ(x)dx cartesian coordinates to polar coordinates allow us to
−∞ dx2
modify the problem tremendously.
Z ∞  2 
d φ(x) In cartesian coordinates the action of operator is
= (Ψ1+ (x) + u21 Ψ1− (x) + u22 Ψ2− (x))∗ − dx
−∞ dx2  2
∂2 ∂2

2 ∂
(56) −∇ = − + 2+ 2 . (58)
∂x2 ∂y ∂z
After some algebra (see [16]) we find finally that Now, when we transform to spherical polar coordi-
nates (or to cylindrical polar coordinates) the point
 dφ(0+ )  "
dφ(0− )
# (x = 0, y = 0, z = 0) becomes singular because the Jaco-
iθ α β
dx =e dx
−
, (57) bian of the transformation tends to zero as we approach
φ(0+ ) γ δ φ(0 ) the origin. Therefore, we must remove the point
(x = 0, y = 0, z = 0) and the plane becomes a holed
where αγ − βδ = 1, and where the parameters in Greek plane which is different from our original space. In fact,
letters α, β, δ, γ and θ are related to the parameters in translation invariance is lost.

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-9

Therefore, we must first study the self-adjointness that square integrable, that is L2 (R3 ), is essentially self-
of the operator (58) in the space of function ψ(r) = adjoint.
ψ(x, y, z) defined by Now, when we remove the origin by going to spherical
Z coordinates, we must consider the equations
ψ ∗ (r)ψ(r)d3 r = finite, (59) d2 R0 (r) 2 dR0 (r)
+ ∓ ik 2 R0 (r) = 0. (64)
d2 r r dr
that is the space L2 (R3 ), that has no holes in it.
Each of the two above equations admits one linearly
We shall examine the operator in the space with a
independent solution and so, we have that the deficiency
hole in the point r = 0 after examining the action of
indices are (1, 1), and hence the operator admits a one
the Laplacian in polar coordinates. In spherical polar
parameter family of self-adjoint extensions, whose wave
coordinates we have
functions are given by
 2
2 d d 1 d2
−∇ = − + 2r + Rℓ=0 (r) = j0 (kr) + αn0 (kr), r > 0, (65)
dr2 dr r2 sin2 φ dθ2
1 d

d
 each member of the family being characterized by a value
+ 2 sin φ . (60) of the parameter α.
r sin φ dφ dφ
Now we see that the Neumann function appears
After separation of variables, we recovered the opera- naturally. It is part of the eigenfunction of an extension
tor given by equation (4) which is the radial part of the of the Laplacian characterized by the parameter α.
action of the differential expression given by (60), viz. If we take α = 0 we have the eigenfunctions of the
usual operator Hamiltonian, used in the literature.
d2 Rℓ (r) 2 dRℓ (r)
 
2 ℓ(ℓ + 1) To find the eigenvalues and eigenfunction of the
+ + k − Rℓ (r)(r). (61) extended operator it is easier to calculate the extension
d2 r r dr r2
of the operator obtained by making the transformation
Then we must work with functions defined in the space
L2 (R3 \(0, 0, 0)) which means that the origin (0, 0, 0) was u(r)
R(r) = , valid for r > 0 (66)
removed. r
We shall now show that the operator given by the in equation (45), for ℓ = 0.
action (58) and domain (59) is essentially self-adjoint The resulting operator is
whereas the operator given by the action (61) (for ℓ = 0)
in the domain L2 (r2 dr\0)) that is functions of r such d2 u(r)
∓ ik 2 u(r) = 0, (67)
that dr2
Z ∞ 2
d
which again show us that the operator − dr 2 in the space
R02 (r)r2 dr = finite (62) 2
of function L (0, ∞), note that the origin was removed,
0
have deficiency indexes (1, 1) and so has a one parameter
is not. For ℓ ̸= 0 the operator is essentially self-adjoint family of self-adjoint extensions.
because one of the equations of the most general solution We have now three options
is not square integrable.
Consider the following equations 1. Seek for a contact interaction (with zero range)
that cancels the infinity of the kinetic energy that
ℏ2
 2
∂2 ∂2 appears if we calculate the expectation value of

∂
− + 2 + 2 ψ(x, y, z) = ∓iEψ(x, y, z). ℏ2 d 2 u(r)
2m ∂x2 ∂y ∂z 2m dr 2 with functions of the form r .
(63) This contact interaction is constructed for example
This equation separates if we write ψ(x, y.z) as by adding to the kinetic energy a square well
ψ(x, y.z) = X(x)Y (y)Z(z) and gives potential with length ϵ and whose depth is fine
tuned.
d2 X(x)
= ±il2 X(x) This contact interaction will depend on a parame-
dx2
ter that we denote by α and the new Hamiltonian
d2 Y (y)
= ±im2 Y (y) is
dy 2
ℏ2 d 2 π2 2α 4α2
 
d2 Z(z) H(α) = − lim − + + + α 2
= ±i(k 2 − l2 − m2 )Z(z) ϵ→0 2m dr 2 4ϵ2 ϵ π2
dz 2
for r < ϵ
where k 2 = 2mE
ℏ2 .
The equations have no square integrable solution, so and
the deficiency indices are (0, 0) the operator given whose ℏ2 d2
action is given by (58) and the domain are function H(α) = − for r > ϵ.
2m dr2

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024
e20230316-10 Did you really miss a Delta function here?

For more details, see references [15] for three The Hamiltonian of this problem is
dimensions, and [23] and [24] for two dimensions.
2. To extend the operator domain by imposing that ℏ2 d 2 1
+ mx2 ω12 θ(x) + ω22 θ(−x) .


H=− (68)
du(0)
2m dx2 2
u(0) = α. This is a general result. When we
dr

restrict the range of the functions where the action As explained already the above should be interpreted
of your operator acts you must put boundary as a differential expression of an operator. To be an
conditions. These boundary conditions are easily operator we must specify its domain. For the above
deduced if you can solve the equations that give the differential expression when ω1 = ω2 we don’t need to
deficiency indexes. In this case they can be derived add any boundary condition, the domain are functions
by integration by parts. Assume that ϕ1 (r) and that are square integrable on the line. To require that the
ϕ2 (r) are in the domain we search. Then we have eigenfunctions can be normalized is sufficient to guaran-
tee the self-adjointness of the operator. See V.S. Araujo
∞
d2 ϕ∗2 (x)
Z  
et al. [26]. In fact, the operator whose action is (68) in
− ϕ1 (x)dx
0 dx2 the domain L2 (−∞, +∞) is essentially self-adjoint.
∞ To find the eigenvalues and eigenfunctions of the above
d2 ϕ∗1 (x)
Z  
− ϕ∗2 (x) − dx operator before specifying its domain we solve the two
0 dx2 differential equations
" dϕ dϕ2 (0)
#
1 (0)

= ϕ∗2 (0)ϕ1 (0) dx

− dx ℏ2 d2 1
ϕ1 (0) ϕ2 (0) − Ψ(x) + mω12 x2 Ψ(x) = EΨ(x),
2m dx2 2
dϕ1 (0)
for − ∞ < x < 0 (69a)
and so ϕ1 (0) = α real (see [16] for details).
dx 2
ℏ d 2
1
3. By giving up arguing and to declare that there is − 2
Ψ(x) + mω22 x2 Ψ(x) = EΨ(x),
2m dx 2
a delta function in the Hamiltonian that forces us for 0 < x < ∞. (69b)
to take u(0) = 0. This last option is not entirely
correct because as is well know the delta function However, let’s first consider the usual solution of
is too strong in three dimensions and it is more the simple harmonic oscillator. This is relevant for
restrictive than the one described above. the problem, because to solve the problem described
above one must solve the Schrödinger equation for the
Remark 2 The same problem as the one discussed harmonic oscillator in some range. Let’s therefore briefly
above occurs in two dimensions and the alternative recall the solution of the harmonic oscillator.
solutions to problems that occur is the same as the ones
given above. Consider the Laplacian in two dimensions. 6.1. The standard harmonic oscillator
When we transform to polar coordinates, translation
invariance is lost, because the point (x = 0 and y = 0) The equation to be solved in standard notation is:
becomes singular. Therefore, the Kinetic energy operator
ℏ2 d 2 1
must be studied carefully. The solutions obtained to − 2
Ψ(x) + mω 2 x2 Ψ(x) = EΨ(x),
the problem described above are the same as the ones 2m dx 2
described above for three dimensions, namely for − ∞ < x < ∞. (70)

1) Seek for a contact interaction that cancels the The point x = 0 is an ordinary point of the equation (70)
infinity of the kinetic energy if we want to use the and so have two linear independent solutions that can
singular solution. This was done in [23]. be obtained as a series expansion [27].
2) Look for self-adjoint extensions of the operator By making different, but related, transformations
extensions This was done in [16, 24]. we get the usual versions of the harmonic oscillator
3) To give up arguing. But in this case, we cannot say equation. We are going to need all of them for clarity,
that there is a delta function at the origin, because because they are all used in different textbooks and
as explained in [23] the delta function is too strong because as have seen these transformations may change
in two dimensions. the domain of the operators.
1. The Weber equation
6. A Final Example Change to ξ = 2mω

12
x and ε = ℏωE
to get the Weber
ℏ
equation
In this final example we discuss a more complicated case
using the machinery developed before. d2 ψ ξ2
 
The problem is the asymmetric, or biharmonic, har- + ε − ψ = 0. (71)
dξ 2 4
monic oscillator in one dimension and was presented in
the article by W. Edward Gettys and H.H. Graben [25]. This form is used in [28].

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-11

2. The Hermite equation must examine the derivative of ψ(ξ), dψ(ξ)

dξ , that is

12 ξ2
Change to ξ = mω x and ψ(x) = e− 2 f (ξ) to get
  
dψ(ξ) ξ2 ν 1
= − 2Cξe− 2 U − ; ; z
ℏ
dξ 2 2
d2 f
 
df 2E
− 2ξ + − 1 f = 0. (72) 
ν 1

dξ 2 dξ ℏω + νU 1 − ; ; z . (76)
2 2
The point x = 0 is an ordinary point of the equa-
When z → 0, we find
tion 1 [27] and so have two linear independent solutions √
that can be obtained as a series expansion. This equation dψ(ξ) πCν ξ
∼ 1 ν ν |ξ| , (77)
was used in the book by Eugen Merzbacher [29] to solve



dξ Γ 2 − 2 Γ 1− 2
the so-called double oscillator.
which is discontinuous at ξ = 0!
3. The confluent hypergeometric equation Then since we are examining equation (77), the
Change in the Hermite equation z = ξ 2 and u(z) = f (ξ) demonstration on the paper by D. Branson [3] requires
to get that this discontinuity should be eliminated by requiring
that
d2 u
 
1 du ν ν
z 2 + −z + u = 0, (73)
= 0. (78)
Γ 12 − ν2 Γ 1 − ν2


dz 2 dz 2

where E = ν + 12 ℏω, which is the confluent hypergeo- This will hold when ν is a negative integer and we get

metric equation [30, chapter VI]. the usual harmonic oscillator solution. In more detail,
Each of these forms are equations that have been when ν is an integer only the first term of equation (74)
studied very thoroughly in the 19th century, and each remains, and the function U becomes an even function
of these forms have been used to solve similar problems of z. When −ν is odd only the second term survives, and
to the bi-harmonic oscillator. For example, as men- U becomes an odd function of z. Furthermore when ν is
tioned before, the Hermite equation was used by Eugen an integer the F s in equation (74) turns out to be the
Merzbacher [29] to solve the so-called double oscillator. Hermite polynomials and the equation (75) becomes the
We prefer to use equation (73). It has two independent usual solution of the harmonic oscillator.
ν 1


solutions [30, Chapter 6]. They are 1 F 1 − 2 ; 2 ; z and 3.1. The bi harmonic oscillator
ν 1 ν 1



U − 2 ; 2 ; z where 1 F1 −
2 ; 2 ; z is the hypergeometric The reader must by now becoming a bit uncomfortable,
function, and U − ν2 ; 12 ; z is defined by because we can write the potential of the bi harmonic
( oscillator as
ν 1


√ 1 F1 − 2 ; 2 ; z
 
ν 1 1
U − ; ;z = π V (x) = mx2 (ω1 θ(x) + ω2 θ(−x)) , (79)
Γ 21 − ν2


2 2 2

) where θ(x) is the Heaviside step function whose deriva-
1 1 F1
1
2 − ν2 ; 32 ; z
− z2 , (74) tive, at x = 0, is discontinuous. In fact, the solution of
Γ − ν2


equation (69a) is given by equation (73) with ω replaced
by ω1 and the solution of equation (69b) is given by the
where Γ stands for the gamma
function. same equation (73) with ω replaced by ω2 and x replaced
When z
→ ∞, 1 F1 − ν2 ; 12 ; z diverges like ez , while by −x.
U − ν2 ; 12 ; z → 0; hence, since

ψ(x) must be normalized We have now three options:
we are left with U − ν2 ; 12 ; z . This can be checked in [30,
p. 289]. 1) The eigenvalues of the bi harmonic oscillator can
Hence be obtained by requiring that the wave functions
  and their derivatives are continuous at x = 0.
−ξ2 ν 1 This is equivalent to retain the point x = 0 and
ψ(ξ) = Ce 2 U − ; ; z , (75)
2 2 therefore the operator whose action is (79) belong
to L2 (−∞, +∞). This is done in the paper by W.
where C is a constant and so equation (75) is the solution Edward Gettys and H.H. Graben [25].
of the problem. (See in this connection [25]) 2) To remove the point x = 0. This means that we
Note that ν, and hence E, are still arbitrary. These are in mathematical parlance putting a barrier
are not determined by the normalizability of ψ. We or frontier at x = 0. By doing this we can
emphasize this because the usual treatment in the construct self-adjoint operators by imposing the
literature gives the impression (see for example [31, boundary conditions (57) to the wave functions at
Chapter 4]) that the convergence of ψ for x → ∞ the removed origin
determines the eigenvalue. " #
dφ(0− )
 dφ(0+ ) 
When the potential is continuous the derivative of the iθ α β
dx =e dx . (80)
wave function should not have a discontinuity, hence we φ(0+ ) γ δ φ(0− )

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024
e20230316-12 Did you really miss a Delta function here?

Let’s take the boundary condition given by θ = 0, Acknowledgments

α = 1, and γ = δ = 0. Then we have φ(0+ ) =
+ −
φ(0− ) = φ(0) and dφ(0dx
)
− dφ(0
dx
)
= βφ(0), and The authors acknowledge partial support from CNPq
we have construct a self-adjoint operator, that (Conselho Nacional de Desenvolvimento Científico e
corresponds to a delta function at the origin. In Tecnológico).
fact, some people would claim that the solution we
got is wrong because we missed a delta function References
at x = 0. To circumvent, or accept this sentence
[1] M. Amaku, F.A.B. Coutinho, O.J.P. Eboli and E.
we may declare that the Hamiltonian have a delta
Massad, Rev. Bras. Ens. Fis. 43, e20210132 (2021).
function at x = 0, with strength α. This option
[2] M. Amaku, F.A.B. Coutinho, O.J.P. Eboli and E.
requires that we declare that the wave functions Massad, Braz. J. Phys. 51, 1324 (2021).
are generalized function (or distribution as they [3] D. Branson, Am. J. Phys. 47, 1000 (1979).
are also called). However, we can ignore this point. [4] M. Andrews, Am. J. Phys. 49, 291 (1981).
See J. Viana-Gomes and N.M.R. Peres [32] or S.H. [5] D. Home and S. Sengupta, Am. J. Phys. 50, 552 (1982).
Patil [33]. [6] T. Cheon and T. Shigehara, Phys. Lett. A 243, 111
We do not want to argue with them but in fact (1998).
we have construct a five-parameter family of self- [7] Y.S. Huang and H.R. Thomann, Europhys. Lett. 115,
adjoint operators that can indeed be interpreted as 60001 (2016).
contact interactions as shown below. [8] M. Khorrami, Europhys. Lett. 116, 60010 (2016).
3) These operators can be interpreted as an operator [9] A. Prados and C.A. Plata, Europhys. Lett. 116, 60011
d2 (2016).
whose action is − dx 2 plus contact interactions (or

zero range interactions) that includes not only the [10] J. Munzenmayer and D. Frydel, arxiv:2012.00166v1
delta function but also derivatives of the delta (2020).
function. These contact interactions given by [19] [11] A.A. Khelashvili and T.P. Nadareshvili, Am. J. Phys.
79, 668 (2011).
are
[12] A.A. Khelashvili and T.P. Nadareshvili, Bull. Georgian
d Natl. Acad. Sci. 6, 68 (2012).
Vb (x) = g1 δ(x) − (g2 − ig3 )δ(x) [13] A.A. Khelashvili and T.P. Nadareshvili, Phys. Part.
dx   Nucl. Lett. 12, 11 (2015).
d d d
+ (g2 + ig3 ) δ(x) − g4 δ(x) [14] S. Flugge, Practical Quantum Mechanics (Springer, New
dx dx dx York, 1974), v. 1.
(81) [15] F.A.B. Coutinho and M. Amaku, Eur. J. Phys. 30, 1015
(2009).
where the four parameters, gi (i = 1 to 4), are [16] V.S. Araujo, F.A.B. Coutinho and J.F. Perez, Am. J.
related to the parameters given by equation (42), Phys. 72, 203 (2004).
for example, g1 = β. [17] F.A.B. Coutinho, Y. Nogami and J.F. Perez, J. Phys A
The energy levels of potential (81) are given in ref- 30, 3973 (1997).
erence [18]. The complete analysis is complicated [18] F.A.B. Coutinho, Y. Nogami and J.F. Perez, J. Phys. A
and a recent reference is given by [34]. 32, L133 (1999).
[19] S. De Vincenzo and C. Sanchez, Can. J. Phys. 88, 809
(2010).
7. Conclusions [20] A. Cintio and A. Michelangeli, Quantum Stud.: Math.
Found. 8, 271 (2021).
In this paper we examined the claim that some solutions [21] D.M. Gitman, I.V. Tyutin and B.L. Voronov, Self-
of the TISE are not solutions because they contain adjoint Extensions in Quantum Mechanics (Birkhauser,
New York, 2012).
a delta function that was forgotten. We showed that
[22] C.R. de Oliveira, Intermediate Spectral Theory and
these solutions are eigenfunctions of operators that are
Quantum Dynamics (Birkhauser, Bassel, 2009).
self-adjoint extensions of an appropriate operator. If
[23] J.F. Perez and F.A.B. Coutinho, Am. J. Phys. 59, 52
you admit that self-adjointness is the only criteria to
(1991).
accept operators and their eigenfunctions in quantum
[24] R. Kowalski, K. Podlaski and J. Rembielinski, Phys.
mechanics these solutions are perfectly acceptable. Rev. A 66, 032118 (2002).
Another point of view is to accept that these eigen- [25] W.E. Gettys and H.H. Graben, Am. J. Phys. 43, 626
functions are eigenfunctions of operators with the same (1975).
action but with added zero-range or contact iterations. [26] V.S. Araujo, F.A.B. Coutinho and F.M. Toyama, Braz.
In the case of one-dimensional problems (or problems J. Phys. 38, 178 (2005).
that can be reduced to one dimensional problems but [27] E.T. Copson, An Introduction to the Theory of Functions
with barries) it is better to interpret the wave function of a Complex Variable (Oxford University Press, Oxford,
as generalized functions. 1970).

Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024 DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316
Amaku et al. e20230316-13

[28] F.D. Mazzitelli, M.D. Mazzitelli and P. Soubelet, Am.

J. Phys. 85, 750 (2017).
[29] E. Merzbacher, Quantum Mechanics (John Wiley and
Sons, Hoboken, 1998), 3 ed.
[30] W. Magnus, F. Oberhettinger and R.P. Soni, Formulas
and Theorems for the Special Functions of Mathematical
Physics (Springer-Verlag Inc, New York, 1966).
[31] L.I. Schiff, Quantum Mechanics (McGraw-Hill, New
York, 1995), 3 ed.
[32] J. Viana-Gomes and N.M.R. Perez, Eur. J. Phys. 32,
1377 (2011).
[33] S.H. Patil, Eur. J. Phys. 27, 899 (2006).
[34] R.-J. Lange, J. Math. Phys. 56, 122105 (2015).

DOI: https://doi.org/10.1590/1806-9126-RBEF-2023-0316 Revista Brasileira de Ensino de Física, vol. 46, e20230316, 2024