Exactly solvable hydrogen-like potentials and

factorization method
J. Oscar Rosas-Ortiz∗
arXiv:quant-ph/9806020v3 28 Oct 1998

Departamento de Fı́sica Teórica†

Universidad de Valladolid, 47011 Valladolid, Spain

Abstract
A set of factorization energies is introduced, giving rise to a generalization of the
Schrödinger (or Infeld and Hull) factorization for the radial hydrogen-like Hamil-
tonian. An algebraic intertwining technique involving such factorization energies
leads to derive n-parametric families of potentials in general almost-isospectral to
the hydrogen-like radial Hamiltonians. The construction of SUSY partner Hamilto-
nians with ground state energies greater than the corresponding ground state energy
of the initial Hamiltonian is also explicitly performed.

Key-Words: Factorization, Hydrogen atom

PACS: 03.65.Ge, 03.65.Fd, 03.65.Ca

∗
On leave of absence from Departamento de Fı́sica, CINVESTAV-IPN, A.P. 14-740, 07000 México
D.F., Mexico. E-mail orosas@fis.cinvestav.mx
†
Electronic address: orosas@klander.fam.cie.uva.es
1 Introduction
The factorization method has been historically atributed to Schrödinger [1]. This ap-
proach (originated from the well known treatment of the harmonic oscillator in non-
relativistic Quantum Mechanics) avoids the use of cumbersome mathematical tools. Since
Schrödinger the factorization method has been successfully applied to solve essentially any
problem for which there exists exact solution [2]. Considering the narrow set of exactly
solvable potentials, it is interesting to look for new potentials for which the corresponding
Schrödinger equation becomes analitically solvable. There are a lot of results on the mat-
ter, in particular it is worthwhile mentionning a work of Mielnik [3], where a variant of the
standard factorization was introduced in order to get a family of Hamiltonians isospectral
to the harmonic oscillator. The first application of Mielnik’s factorization was performed
for the hydrogen-like radial potentials [4], the resulting family of Hamiltonians became
isospectral to the standard radial hydrogen-like Hamiltonians. An additional interesting
application arose from the arena of the coherent states. Concerning this subject, it is
well known that the usual first order factorization operators for the harmonic oscillator
Hamiltonian are also its creation and anihilation operators. This is not the case for the
isospectral oscillator Hamiltonians derived by Mielnik because their first order factoriza-
tion operators are different from their creation and anihilation operators, which turn out
to be of order equal to (or greater than) three [3, 5]. These last operators are the genera-
tors of non-linear algebras, and allow one to get immediately the corresponding coherent
states as the eigenstates of the annihilation operator [5]. Recently, these algebras have
been rederived (using different methods) and reinterpreted by other authors [6, 7]. It has
been shown as well that such non-linear algebras can be associated to other systems [7].
The factorization techniques [1, 2, 3], the Darboux transformation [8], the super-
symmetric Quantum Mechanics (SUSY QM) [9], and other procedures have a common
point: they can be embraced in the algebraic scheme known as first order intertwining
transformation which makes use of a first order differential operator to intertwine (the
eigenfunctions of) two different Hamiltonians [10]. The generalization of this technique
involves higher order differential operators and leads in a natural way to the higher order
SUSY QM [11-15]. One of the advantages of the intertwining technique is that the com-
position of n first order intertwining transformations allows one to express the nth-order
intertwiner operators as the product of n first order intertwining operators [11, 13, 15].
The experimental study of spectra (for systems as nuclei, atoms, molecules, etc) makes
important per se the theoretical determination of the energy levels of potentials capable
of being used as models for describing different physical situations. However, for a quan-
tum system with a given Hamiltonian it is seldom possible to get exact solutions to the
corresponding eigenproblem.
In this paper we are going to show that, given an exactly solvable potential as an
input, the intertwining technique provides, in general, new exactly solvable potentials
which, from the traditional viewpoint, would be solved by perturbative or approximate
methods. Hence, working with Hamiltonians whose spectrum is known, one has at hand
the tools to test the validity and convergence of some perturbative or approximate meth-
ods [13, 16]. In Section 2 we shall present the nth-order intertwining technique, discussing

1
in more detail those aspects which have not been previously reported in the literature.
In particular, in Section 2.1 we shall show that the standard Infeld-Hull factorization as
well as the modified Mielnik’s method can be recovered by the first order intertwining
treatment. We will introduce a generalization of the Mielnik’s approach by using solu-
tions to the Schrödinger equation corresponding to factorization energies not belonging
to the physical spectrum of the initial problem. In Section 2.2, by using the second order
intertwining approach, we will introduce explicit expressions allowing the construction of
two-parametric families of potentials whose spectrum is equal to the initial spectrum plus
two new levels at predetermined positions. In Section 2.3 we shall prove that the second
order, as well as the higher order cases, can be recovered by the iterative application of
our first order intertwining procedures. Section 3 contains the derivation of a wide set
of new n-parametric families of potentials whose spectrum is the same (isospectral), or
almost the same (almost isospectral), as the corresponding hydrogen-like spectrum. We
shall introduce a set of factorization energies generalizing the choice made for the Infeld-
Hull and Mielnik factorizations of the hydrogen-like potentials. As a consequence, the
first n energy levels of the n-parametric families can be placed at any previously fixed
spectral positions chosen from the above mentionated set of factorization energies. This
result opens the possibility of placing holes between these levels or between them and the
other ones. Some particular cases related to results previously derived by other authors
are also mentioned.

2 The Intertwining Method

Let H and H f be the one dimensional Hamiltonians
d2 2
f ≡ − d + Ve (x).
H ≡ − 2 + V (x), H (1)
dx dx2
The nth-order intertwining method aims to find a nth-order differential operator A such
that
f = AH.
HA (2)
Let H be a Hamiltonian with known solutions to the time-independent Schrödinger
equation Hψ = Eψ. We are looking for solutions to the corresponding equation for
the intertwined Hamiltonian H, f H fψe = E e The transformation (2) preserves many
e ψ.
f
spectral properties of the initial Hamiltonian, e.g. H(Aψ) = E(Aψ), ∀ Aψ 6= 0. We are
interested only in those operators A whose action on L (R) produces functions in L2 (R);
2

hence we can choose ψe ≡ c0 Aψ, with c0 a normalization constant, and we will have
Ee = E, ∀ ψe = c0 Aψ 6= 0, ψ ∈ L2 (R). Therefore, the intertwining operator A transforms
solutions ψ of the initial Schrödinger equation into solutions ψe of a new Schrödinger
equation, both equations sharing the initial eigenvalues. However, the new Hamiltonian
could have a finite number of additional eigenvalues; if this is the case, the corresponding
eigenfunctions have to be elements of the kernel of A† , as we shall see below. We will
show now how the intertwining method does work for the simplest case where A is a first
order differential operator. This case is specially interesting because it provides a general
framework in which various apparently different methods are included.

2
2.1 The first order intertwining method
Let the operator involved in the intertwining relationship be the following first order
differential operator
d
A≡a= + β(x). (3)
dx
By introducing (1) and (3) in (2) we get
− β ′(x) + β 2 (x) = V (x) − ǫ, (4)
Ve (x) = V (x) + 2β ′(x), (5)
where ǫ is an arbitrary integration constant, and the prime denotes derivative with respect
to x. Notice the implicit dependence of β(x) and Ve (x) on ǫ.
From equations (1) and (5) it becomes apparent that H f − H = 2β ′ (x), from (3) we
get [a, a† ] = 2β ′(x), therefore (4) and (5) lead to:
H = a† a + ǫ, f = aa† + ǫ.
H (6)
Equation (6) shows that the first order intertwining relationship (2), with A given by (3),
leads in a natural way to the factorization of the intertwined Hamiltonians for each value
of ǫ ∈ R, provided that β(x) satisfies (4) with a given V (x). The traditional approach
to the factorization method involves just a particular solution to the Riccati differential
equation (4) for a specific value of ǫ [2]. However, it is well known that a particular
solution to a Riccati equation allows one to get the corresponding general solution by
means of two quadratures [18]. Mielnik’s approach takes this into account [13] and gives
a term in β(x) unnoticed by the traditional factorization method. With the exception
of the harmonic oscillator and the hydrogen-like potentials, the general solutions to the
corresponding Riccati equation (4) remain unexplored for almost all the Infeld and Hull
cases [19, 20]. An interesting discussion on the relationship between Mielnik’s method
and Darboux transformation was given by Andrianov et al in 1985 [11] (see also [20]). On
the other hand, equations (4) and (5) are basic in the construction of the SUSY partner
Hamiltonians H + ≡ H − ǫ = a† a, and H − ≡ H f − ǫ = aa† in supersymmetric quantum
mechanics. In this language, it is said that equation (5) relates the two SUSY partner
potentials Ve (x) and V (x), and the function β(x) is named the SUSY potential [9]. Notice
that the two Hamiltonians H ± are positive semi-definite, i.e. hH ± i ≥ 0; hence we have
E ≥ ǫ and Ee ≥ ǫ.
If the eigenfunctions {ψ} and eigenvalues {E} of H are known, the corresponding
eigenfunctions of Hf are given by ψ e = (E − ǫ)−1/2 aψ with eigenvalues E
e = E. They form
an orthonormal set
s
E′ − ǫ ′
hψe′ , ψi
e ′
= [(E − ǫ)(E − ǫ)]−1/2 ′
haψ , aψi = hψ , ψi = 0, ∀ψ ′ 6= ψ.
E−ǫ
However, they do not span yet the whole of L2 (R) because the solution to a† ψeǫ = 0 is
orthogonal to all the members of the set {ψ}, e hψ e , ψi
e = (E − ǫ)−1/2 ha† ψe , ψi = 0. That
ǫ ǫ
first order linear differential equation is immediately solved:
R
ψe ǫ ∝e β(x)dx
. (7)

3
From (6), it is easy to see that H fψe = ǫψe . Hence, ψ e is an eigenfunction of Hf with
ǫ ǫ ǫ
eigenvalue ǫ. In the case when ψeǫ is a square integrable function (i.e. ψeǫ has physical
sense as a wavefunction), it has to be added to the set {ψe = (E − ǫ)−1/2 aψ} in order to
complete a new basis on L2 (R). The set of eigenvalues of H f is given by {E}
e = {ǫ, E}, so
that the Hamiltonian H f has the same spectrum as H plus a new level at ǫ. Indeed, since
the general solution β(x) of the Riccati equation (4) has a ‘free parameter’ λ, labeling
also Ve (x) (see equation (5)), the Mielnik’s method allows one to construct, for a specific
value of ǫ, a one-parametric family of potentials Ve (x) which in general will be almost
isospectral to V (x).
We are going to disscus now how the Mielnik’s method can be further generalized.
First, let us remark that the factorization energy considered by Infeld-Hull and Mielnik
is not the only posibility leading to a solvable Riccati equation (4). The key point arises
when the non-linear first order differential equation (4) is replaced into a homogeneous
linear second order differential equation by means of the transformation1 :
d
β(x) = − ln u(x), (8)
dx
which leads to the second order differential equation:

d2
u(x) − [V (x) − ǫ]u(x) = 0. (9)
dx2
If the factorization energy ǫ belongs to {E}, then the solutions u(x) to (9) are the same
as the previously known eigenfunctions ψ(x) of the initial Hamiltonian H. Following [3],
we shall consider just the cases where the factorization energy ǫ does not belong to the
spectrum of H. Therefore, the solutions to (9) do not have direct physical meaning, but
we have just seen that they naturally lead to the factorization of the Hamiltonians H
and Hf in the spirit of Mielnik’s approach, providing as well the explicit form for the new
potentials Ve (x).
Let us remark that n different factorization energies ǫ1 , ǫ2 , ..., ǫn , lead to n different
Riccati equations (4) and to n one-parametric families of potentials Vei (x), i = 1, 2, ..., n,
almost isospectral to V (x). It must be clear that there will be n different factorizations
of the initial Hamiltonian H (see equation (6)). With the aim of iterating later the first
order intertwining technique let us introduce a slightly different notation. Let us write
ǫ = ǫ(k) , where k labels the specific real number ǫ 6∈ {E}, which we will use as factorization
energy. Moreover, we shall write explicitly the dependence of β(x) and Ve (x) on ǫ(k) (see
equations (4) and (5)) by making β(x) = β (k) (x) and Ve (x) = V (k) (x). For the sake
of simplicity the other functions and operators labeled by a tilde will be represented as
f = H (k) , E
H e = E (k) , etc. Finally, the number of indices will denote the order of the
intertwining transformation (2), as well as the number of ‘free parameters’ λ which will
label the intertwined potential. For example, H (k) represents a first order intertwined
Hamiltonian whose corresponding one-parametric family of potentials V (k) (x) is almost
isospectral to V (x), while H (km) represents a second order intertwined Hamiltonian with
1
R
The inverse transformation of (8) is u(x) = exp(− β(x)dx).

4
the corresponding two-parametric family of potentials V (km) (x) almost isospectral to V (x).
The benefits of such notation will be clear in the next sections.

2.2 Second order intertwining method

Let us consider the second order differential operator
d2 d
B (km) = 2
+ η(x) + γ(x), (10)
dx dx
where η(x) and γ(x) depend implicitly on ǫ(k) and ǫ(m) . Introducing H (km) = −d2 /dx2 +
V (km) (x) and (10) in the second order intertwining relationship H (km) B (km) = B (km) H,
with H given in (1), and by ordering at different powers of the operator d/dx, we have
for the corresponding coefficients

V (km) (x) = V (x) + 2η ′ (x), (11)

2γ(x) = η 2 (x) − η ′ (x) − d − 2V (x), (12)

′′ ′ ′ ′′
V (x) + η(x)V (x) = 2γ(x)η (x) − γ (x), (13)
where d is an integration constant. Following [12], we can express the functions V (x),
V (km) (x) and γ(x) in terms of η(x). The substitution of (12) into (13) produces a third
order differential equation which, after multiplying by η(x), becomes reduced (by a first
integration) to the second order differential equation
!
(η ′ )2 η2 2
ηη ′′ − + 2γ − η ′ − η + 2c = 0, (14)
2 2

where we have used once again the equation (12) and c is a new integration constant.
In order to solve the non-linear second order differential equation (14), let us try the
ansatz [21]
η ′ (x) = η 2 (x) + z(x)η(x) + b, (15)
where the constant
√ b and the function z(x) are to be determined. Introducing (15) in (14)
we get b± = ±2 c, c > 0, besides two Riccati differential equations
2
′ z± (x)
z± (x) + − 2V (x) − d − b± = 0. (16)
2
The identification z+ (x) = −2 βk (x), z− (x) = −2 βm (x), and
!2
(m) (k) ǫ(m) − ǫ(k)
d = −ǫ −ǫ , c= , ǫ(m) 6= ǫ(k) , (17)
2

allows one to write the solutions of (14) in the form

!
ǫ(m) − ǫ(k)
η(x) = − , ǫ(m) 6= ǫ(k) , (18)
β (m) (x) − β (k) (x)

5
where β (m) (x) and β (k) (x) are both solutions to (4) with factorization energies ǫ(m) and ǫ(k)
respectively. Equations (18) and (12) determine the second order intertwining operator
B (km) . Notice that, by means of this process, we can derive solutions to the non-linear
second order differential equation (14) by just solving the easier non-linear first order
differential equation (4) for two different values of the factorization energy ǫ.
The action of B (km) on the eigenfunctions {ψ} of H gives the basic set of eigenfunc-
tions {Ψ(km) ∝ B (km) ψ} of the intertwined Hamiltonian H (km) . On the other hand, the
† (km)
solutions to the second order differential equation B (km) ΨB = 0 are orthogonal to all
†
the Ψ(km) ∝ B (km) ψ. Now, because the kernel of B (km) is a 2-dimensional subspace, the
maximum number of linearly independent elements is two. Therefore, we are looking for
† (km)
two square integrable linearly independent solutions to B (km) ΨB = 0 such that they
are simultaneously eigenfunctions of H (km) , with eigenvalues to be determined.
(km) R †
Let us now write ΨB = C0 exp( f (km) (x)dx) as a generic kernel element of B (km) ;
† (km)
then the equation B (km) ΨB = 0 can be rewritten as
d (km)
f (x) − η(x)f (km) (x) + (f (km) (x))2 − η ′ (x) + γ(x) = 0. (19)
dx
Using equations (4), (12) and (15), the general solutions to (19) can be obtained:

(km) u(k) (x) u(m) (x)

ΨB (x) = Ck + Cm , (20)
W (k, m) W (k, m)

where u(k) (x) and u(m) (x) are the two unphysical solutions to (9) with eigenvalues ǫ(k) and
ǫ(m) , respectively. The function W (k, m) = u(k) (u(m) )′ − (u(k) )′ u(m) is the wronskian of
u(k) (x) and u(m) (x).
By using (8) and (9) we now rewrite (18) as

u(k) (x)u(m) (x) d

η(x) = (ǫ(m) − ǫ(k) ) = − ln W (k, m). (21)
W (k, m) dx

Notice that the right hand side of (21) corresponds to the Crum’s determinant [17].
Hence, the coefficient of the second term of B (km) in (10) can be constructed either by
using (21) or simply by (18). Using now (16), (8) and (9), with η(x) as given above, it
is easy to show that the function u(k) /W (k, m) is eigenfunction of H (km) with eigenvalue
ǫ(m) . A similar procedure shows that u(m) /W (k, m) is also an eigenfunction of H (km) with
eigenvalue ǫ(k) . Taking this into account we write

u(k) (x) u(m) (x)

Ψǫ(km) (x) ∝ ; Ψǫ(km) (x) ∝ , (22)
m
W (k, m) k
W (k, m)

where the subindex ǫm (ǫk ) indicates the corresponding eigenvalue ǫ(m) (ǫ(k) ). Therefore,
the eigenfunctions of H (km) are {Ψǫ(km)
m
, Ψǫ(km)
k
} ∪ {Ψ(km) ∝ B (km) ψ | B (km) ψ 6= 0}, and the
corresponding eigenvalues are {E (km) } = {ǫ(m) , ǫ(k) , E}. Thus, H (km) has the same spec-
trum as H plus two new levels at ǫ(m) and ǫ(k) , provided that Ψǫ(km) m
and Ψǫ(km)
k
are square

6
integrable functions. As (18) involves two general solutions to the Riccati equation (4),
thus V (km) (x) represents a set of two-parametric families of potentials almost isospectral
to V (x).
Notice that we have not placed any ordering to the levels ǫ(m) and ǫ(k) so that it
is possible either that ǫ(m) > ǫ(k) or that ǫ(m) < ǫ(k) . This is implicit in the fact that
η(x) in (18) is invariant under the change k ↔ m. This will be important in the next
section, where we are going to show that the results found by means of the second order
intertwining technique can be found also by the iteration of two successive first order
intertwining transformations.

2.3 Iterative factorization

Let us come back to the results of Section 2.1 and suppose that by means of the first
order intertwining technique we have obtained a Hamiltonian H (k) from H using a certain
factorization energy ǫ(k) . We look now for a new Hamiltonian h(km) = −d2 /dx2 + v (km) (x)
attainable from H (k) by means of the first order intertwining method, i.e., h(km) and H (k)
satisfy the following relationship

h(km) a(km) = a(km) H (k) ,

where a(km) is the first order differential operator a(km) ≡ d/dx + β (km) (x). As usual, we
arrive at the standard equations linking v (km) (x), V (k) (x) and β (km) (x):

− (β (km) )′ + (β (km) )2 = V (k) (x) − εm = 2(β (k) )′ + V (x) − εm , (23)

v (km) (x) = V (k) (x) + 2(β (km) )′ (x) = V (x) + 2[β (km) (x) + β (k) (x)]′ , (24)
where εm is a new integration constant (factorization energy) and we have used equation
(5). The intertwined Hamiltonians once again become factorized
† †
H (k) = a(km) a(km) + εm , h(km) = a(km) a(km) + εm .

The eigenvalues of h(km) are given by {E (km) } = {εm , E (k) }, and the eigenfunctions by
ψε(km)
m R
∪{ψ (km) = (E (k) −εm )−1/2 a(km) ψ (k) | a(km) ψ (k) 6= 0}, where the missing state ψε(km)
m
∝
exp( β (km) (x)dx) corresponds to the eigenvalue εm .
The key point here becomes to find the general solution to the new Riccati’s differential
equation (23) for some εm ∈ R. The obvious solution (β (km) )′ (x) = −(β (k) )′ (x), for which
εm = ǫ(k) , corresponds to a Hamiltonian h(km) equal to H because v (km) (x) = V (x).
Avoiding this trivial solution we shall consider εm 6= ǫ(k) . Non-trivial solutions to (23)
can be found by inspecting the following intertwining relationship:

h(km) a(km) a(k) = a(km) H (k) a(k) = a(km) a(k) H. (25)

Equation (25) shows that h(km) and H are related by the product of two first order
intertwining operators, i.e., by the second order differential operator C (km) ≡ a(km) a(k) .
An interesting question immediately arises: is the Hamiltonian h(km) the same as the

7
H (km) of Section 2.2? If C (km) = B (km) the answer is affirmative, and in this case the
SUSY potential β (k) (x), the iterated new SUSY potential β (km) (x), and the functions γ(x)
and η(x) are related by the equations
d (km)
η(x) = β (km) (x) + β (k) (x), γ(x) = β (x) + β (km) (x)β (k) (x). (26)
dx
Comparing (18) and (26) we get the solutions of (23) (with εm = ǫ(m) ):
!
ǫ(m) − ǫ(k)
β (km) (x) = −β (k) (x) − , (27)
β (m) (x) − β (k) (x)

where β (k) (x) (β (m) (x)) is a solution of (4) associated to the factorization energy ǫ(k) (ǫ(m) ).
Equation (27) is nothing but the theorem of Fernández et al [15]. The eigenfunctions of
h(km) = H (km) are given by the basic set

Ψǫ(km)
k
= (ǫ(k) − ǫ(m) )−1/2 a(km) ψǫ(k)
k
, Ψ(km) = (E − ǫ(m) )−1/2 (E − ǫ(k) )−1/2 a(km) a(k) ψ,

plus the new missing state R

β (km) (x)dx
Ψǫ(km)
m
∝e ,
which, by definition, is such that a(km)† Ψǫ(km)
m
= 0. Hence, by construction B (km)† Ψǫ(km)
m
=
(k)† (km)† (km) (k)† (k)
a a Ψǫm = 0. On the other hand, remembering that a ψǫk = 0, we get

B (km)† Ψǫ(km)
k
= (ǫ(k) − ǫ(m) )−1/2 a(k)† [H (km) − ǫ(m) ]ψǫ(k)
k
= 0.

Now, by using (26), (8) and after (21), we can write

u(m) (x)
Ψǫ(km) (x) (k)
= (ǫ −ǫ (m) 1/2
) ,
k
W (k, m)

which corresponds to the definition of Ψǫ(km) k

(x) given in (22). A similar procedure but
now using (27), (8) and (21), allows one to recover the corresponding expression for
Ψǫ(km)
m
(x). Finally, the set of eigenvalues of h(km) = H (km) is given by {E (km) = E (km) } =
{ǫ(m) , ǫ(k) , E}, just as we have shown in Section 2.2.
It is worth to notice that the operator B (km) = a(km) a(k) does not factorize the inter-
twined Hamiltonians H and H (km) , but some of their quadratic forms:

B (km)† B (km) = a(k)† a(km)† a(km) a(k) = a(k)† [H (k) − ǫ(m) ]a(k) = a(k)† a(k) [H − ǫ(m) ]
= (H − ǫ(k) )(H − ǫ(m) ) = (H − ǫ(m) )(H − ǫ(k) ), (28)

and
B (km) B (km)† = (H (km) − ǫ(m) )(H (km) − ǫ(k) ). (29)
Up to now, from the general solution to (4) for each factorization energy ǫ(k) , we
have derived a one-parametric family of potentials V (k) (x) isospectral to V (x) but by a
new level at ǫ(k) . From two of these solutions with ǫ(k) 6= ǫ(m) we have derived as well

8
two-parametric families isospectral to V (x) but by two new levels at ǫ(m) and ǫ(k) . It is
possible to iterate further the first order method by considering now the general solutions
to (4), associated to three different factorization energies ǫ(k) , ǫ(m) , ǫ(l) . The key Riccati
equation to be solved becomes now:

−(β (kml) )′ (x) + (β (kml) )2 (x) = V (km) (x) − ǫ(l) .

The algorithm (27) allows one to write β (kml) (x) in terms of two different solutions of
(23), !
(kml) (km) ǫ(l) − ǫ(m)
β (x) = −β − ,
β (kl) (x) − β (km) (x)
which leads to the following 3-parametric family of potentials V (kml) (x):

V (kml) (x) = V (km) (x) + 2(β (kml) )′ (x) = V (x) + 2[β (kml) (x) + β (km) (x) + β (k) (x)]′ , (30)

having the same spectrum as V (x) but by three new energy levels at ǫ(l) , ǫ(m) and ǫ(k) [15].
The corresponding third order intertwining operator is given by C (kml) = a(kml) a(km) a(k) .
This leads to the factorization of some six order differential operators, related to the two
intertwinned Hamiltonians by means of:
†
C (kml) C (kml) = (H − ǫ(l) )(H − ǫ(m) )(H − ǫ(k) )
†
C (kml) C (kml) = (H (kml) − ǫ(l) )(H (kml) − ǫ(m) )(H (kml) − ǫ(k) ).
This process can be continued at will, and the corresponding formulae for the key functions
can be algebraically obtained from the solutions β (k) (x) to (4). It is interesting to see now
how the process does work explicitly for a physically meaningful system.

3 The hydrogen-like potentials

Let us consider a single electron in the field produced by a nucleus with charge Ze, where
Z is the number of protons in the nucleus2 . The time-independent Schrödinger equation
is given by:
2
− ∇2 Ψ(~r) − Ψ(~r) = EΨ(~r), (31)
r
where we are using the units of rB = h̄2 /Ze2 m, and E = Z/2rB for the coordinate
and energy respectively. Due to the spherical symmetry, the standard choice Ψ(~r) =
R(r)Y (θ, φ) separates (31) into a radial and an angular equation. The angular solutions
are given by the spherical harmonics Ylm (θ, φ) and the energy spectrum can be finally
obtained by solving the radial equation:
" #
d2 l(l + 1) 2
− 2+ − ψ(r) = Eψ(r), (32)
dr r2 r
2
Z = 1 describes the Hydrogen atom, while ions or atoms with only one electron in the outermost
shell can be considered as hydrogen-like systems with Z > 1.

9
where, l = 0, 1, 2, ..., is the azimutal quantum number and, for the sake of simplicity, we
will work with ψ(r) ≡ rR(r), 0 ≤ r < +∞. The set of solutions of (32) span the whole
R
of L2 (R+ ), with an inner product defined by hψ, ψ ′i = 4π 0+∞ ψ(r)ψ ′ (r)dr < ∞. The
differential operator of the left hand side of (32) will be refered as the radial Hamiltonian,
and it will be denoted by Hl , where
d2 l(l + 1) 2
Hl = − + Vl (x); Vl (x) ≡ − . (33)
dr 2 r2 r
The eigenvalues, for a fixed l, are given by the well known formula
1
En ≡ ElK = − , K = 1, 2, 3, ... (34)
(l + K)2

3.1 The new one-parametric families

(k) (k) (k)
Let us consider the first order intertwining relationship Hl−1 al = al Hl where, as a result
of the further calculations, we are labeling from the beginning the unknown Hamiltonian
(k)
Hl−1 with the subindex l − 1 because it will arise a centrifugal term for the new potential
with exactly that index. In order to solve the corresponding Riccati equation (4) we shall
use its equivalent second order differential equation (9) rewritten as
" #
d2 l(l + 1) 2 (k) (k) (k)
− 2+ 2
− ul (r) = ǫl ul (r). (35)
dr r r

Comparing (35) with (32) and (34), it is natural to propose the factorization energies as:

(k) 1
ǫl ≡− , k 6= K, l > 0, (36)
(l + k)2
where, in general, k could be either a discrete as well as a continuous real number. On
the other hand, note that the case k = K reproduces the well known solutions to the
physical eigenproblem (32-34); hence, following Mielnik we have taken k 6= K. It is clear
(k)
that, for a fixed value of l, we have ǫl 6= ElK , ∀ k 6= K.
In order to solve (35) we make the transformation
(k) (k)
ul (r) = r −l er/(l+k) Φl (r), (37)
(k)
leading to a confluent hypergeometric equation for Φl (r), whose general solution for the
discrete values k = 0, −1, −2, ..., −(l − 1) is given by the linear combination of confluent
hypergeometric functions [22, 23]:
(k)
Φl (r) = 1 F1 [k, −2l, −2r/(l+k)]−νlk [2r/(l+k)]1+2l 1 F1 [1+k+2l, 2+2l, −2r/(l+k)], (38)

with
(k)
Γ(1 + |k|) λl
νlk ≡ , l > 0, (39)
Γ(2l + 2) (−2l)|k|

10
 (k)
where λl is a constant to be determined and

Γ(−2l + |k|)
(−2l)|k| ≡ = (−2l + |k| − 1)(−2l + |k| − 2) · · · (−2l).
Γ(−2l)
(k)
Note that Φl (r = 0) = 1, while its asymptotic behaviour is given by
!|k| (k)
(k) 2r 1 − λl
Φl (r) ∼ .
l − |k| (−2l)|k|

Hence, the solutions (37) are divergent at the origin as well as in the limit r → ∞, and
(k)
it is clear that ul (r) 6∈ L2 (R+ ). This is the reason why we consider them as unphysical
solutions to the Schrödinger equation (32) with the atypical eigenvalues (36). Let us
remark that the discreteness of k in (37-39) leads to the most transparent case available
with the mathematical tools at hand [22]. The general solution to the corresponding
Riccati equation (4) arises after introducing (37) in (8), which gives:

(k) l 1 d (k)
βl (r) = − − ln Φl (r), l > 0. (40)
r (l + k) dr

Using (5) and (40), we get the first order intertwined potentials

(k) d2 (k)
Vl−1 (r) = Vl−1 (r) − 2 2
ln Φl (r), l > 0. (41)
dr
The second term of (41) is free of singularities if

 (−∞, 1),
 for |k| even;
(k)
λl ∈ (42)


(1, ∞), for |k| odd.
(k)
Hence, the new potentials Vl−1 (r) have the same singularity at r = 0 as Vl−1 (r),
(k) (k)
provided that λl takes only values in the domain (42). Since the second term of Vl−1 (r)
(k)
tends to zero when r → 0 and r → ∞, we conclude that Vl−1 (r) behaves as Vl−1 (r) at the
ends of [0, ∞).
(k) (k)
Now, the eigenfunctions of Hl−1 are obtained from the action of al on the eigenfunc-
(k) (k) (k)
tions {ψnl (r)} of Hl , ψn,l−1 = (En − ǫl )−1/2 al ψnl (r), plus the isolated ground state:

(k) r l e−r/(l+k)
ψl−1,ǫk = Clk (k)
, (43)
Φl (r)
(k)
where Φl (r) is given by (38), and
 !2l+1   1/2
(k)
2 1 − λ l  |k|! 
Clk =   , k = 0, −1, −2, ..., −(l − 1). (44)
l − |k| (−2l)|k| (2l)!

11
It is important to notice that, despites the resemblance between the missing states (43)
(k) (k)
and the unphysical functions (37), ψl−1,ǫk ∼ 1/ul , just the first ones become square
(k)
integrable functions. The eigenvalues of Hl−1 (for fixed l and k) are:

(k) 1 1
El−1,K = − , − , K = 1, 2, ... (45)
(l + k)2 (l + K)2
By comparing with (34) it is apparent that our families of potentials (41) have, in general,
the same spectrum as the corresponding spectrum of the hydrogen-like potentials Vl−1 (r),
(k)
plus a new level at ǫl .

3.2 Particular one-parametric families

Notice that, for fixed l, k can take l different values k = 0, −1, −2, ..., −(l−1). Hence, there
(k)
are l different factorization energies ǫl generating l non-equivalent families of solvable
(k) (0) (k)
potentials Vl−1 (r). In particular, let us notice that ǫl = −1/l2 ≡ L(l), with λl = 0,
lead to the standard factorization of the hydrogen-like potentials as presented by Infeld-
Hull (see equation 8.1.3, and page 68 of [2]). Thus, our technique can be considered as
a generalization of the Infeld-Hull factorization where, instead of looking for particular
(0) (k)
solutions to (4) with ǫl = L(l), we look for the general solutions with ǫl = L(l + k). In
a similar way, our results can be seen as a generalization of those derived by Fernández,
who found a general solution to (4) but just in the case with k = 0 [4]. Hence, for a fixed
(0) (0)
l and taking k = 0, the equation (45) leads to El−1,K = El−1,K , i.e., Vl−1, (r) is a family of
potentials strictly isospectral to Vl−1 (r). On the other hand, for k 6= 0, we have not only
(k) (k) (k6=0)
ǫl−1 6= El−1,K , but ǫl−1 < El−1,K=1 = −1/(l + 1)2 . In this case the spectrum of Hl−1
is almost the same as the one of Hl−1 , the difference resting in the ground state energy
level. It is clear now that the functions (41) represent a set of l − 1 families of potentials
almost isospectral to Vl−1 (r), plus a family strictly isospectral to Vl−1 (r).

3.2.1 The case k = 0

The member of the family (41), for k = 0, is given by

(0) d2 (0)
Vl−1 (r) = Vl−1 (r) − 2 2
ln Φl (r), l > 0, (46)
dr
where
 2l+1
λl Z r 2l −2x/l
(0)
(0) 2
Φl (r) = 1 − x e dx. (47)
l (2l)! 0
(0) (0)
From (42) we see that Vl−1 (r) does not have more singularities than Vl−1 (r) if λl ∈
(0) (0) (0)
(−∞, 1). In particular, for λl = 0, we have Φl (r) = 1 and Vl−1 (r) = Vl−1 (r), which
means that Vl−1 (r) itself is a member of the family (46). On the other hand, we get also the
(0)
standard SUSY potential W (r) ≡ βl (r) = l/r −1/l, besides the corresponding first order
(0)
operator al = d/dr+l/r−1/l, both of them used to solve the hydrogen-like Hamiltonians

12
 (0) (0)
[1, 2, 9]. Let us remark once again that Vl−1 (r) with λl = (2l)!(l/2)2l+1 γl−1 leads to the
one-parametric family of isospectral hydrogen-like potentials derived by Fernández [4]. If
(0)
l = 1 and λ1 → 1, the potential (46) tends also to the one found by Abraham and Moses
[24].

3.2.2 The case k = −1

Let us take now k = −1. Thus, the potential (41) becomes

(−1) d2 (−1)
Vl−1 (r) = Vl−1 (r) − 2 2
ln Φl (r), l > 1, (48)
dr
where
" #  
 (−1)   Z 
(−1) r λl 2 2l−1 r x2l e−2x/(l−1)
Φl (r) = 1− 1+ dx . (49)
l(l − 1)  (2l − 1)! l − 1 0 [l(l − 1) − x]2 

(−1)
In this case λl ∈ (1, ∞), and although Vl−1 (r) governs the behaviour at the origin
(−1)
and at infinity of Vl−1 (r), it is not a member of the family (48). The corresponding
eigenvalues are given by
(−1) 1 1 1
El−1,K = − 2
,− 2
,− , ..., l > 1. (50)
(l − 1) (l + 1) (l + 2)2
(−1)
Notice the ‘jump’ between the first two energy levels of Hl−1 ; this produces a hole in the
spectrum (50) because the level at −1/l2 is absent. Comparing (50) with (34) it turns
(−1)
out that Vl−1 (r) represents a family of potentials almost isospectral to Vl−1 (r). As a
(−1)
particular example let us consider l = 2: in this case V1 (r) has the same levels, −1/n2 ,
(−1)
as V1 (r) for n ≥ 3. However, the ground state energy level of V1 (r) is at −1, which
is forbidden for V1 (r) whose ground state is at −1/4. The potentials (48-49) have been
recently reported by the author (see equations (10) and (22) of [14]) and they are plotted
(−1)
in Figure 1 for l = 2 and different values of λ2 .

3.2.3 The case |k| > 1

(k)
For |k| > 1, there is a gap between the two first energy levels of Hl−1 , which grows up as
|k| increases. For instance, when |k| = 2, and l = 3, the levels labeled by n = 2 and n = 3
(−2)
are not present in the spectrum of H2 . A similar result, but with a different method
and Hamiltonian, has been derived by Samsonov [11].
Recently, the SUSY potential ω(x) = a/γ − γ/x, with a > 0 and γ > 0, has been used
to construct a ‘conditionally exactly solvable (CES) potential’ V+ (x) for the Hydrogen
atom problem [25, 26]. In such a paper a is a coupling constant while γ is a parameter
taking some strange values as γ = 2.8 (see Fig. 7 in [26]). Although there is a close
(k)
relationship between that potential V+ (x) and our first order intertwined potential Vl−1 (r),
it is important some caution:

13
 i) If γ > 0 is not an integer, the SUSY potential ω(x) of [26] is not related to the
physically relevant SUSY potential W (r) for the Hydrogen atom, and in this case
V+ (x) is the ‘CES potential’ for some mathematical problem. If the so called CES
potentials derived in [26] are to be related with a “physical potential”, the condition
γ > 0 is not sufficient.

ii) If γ is an integer greater than zero, the SUSY potential ω(x) can be related to
the Hydrogen atom problem. In this case the function (5.14) of [26] has a term
1 F1 (−γ − a/ρ, −2γ, 2ρx), which is meaningless at 2ρx = 0 unless γ + a/ρ is an
integer greater than or equal to zero [23]. Indeed, since γ is a positive integer, the
fraction a/ρ is an integer such that γ ≥ −a/ρ. Hence, the condition a > 0 is neither
sufficient to ensure the generality of the potential V+ (x).

Taking this into account, a general ‘CES potential’ V+ (x) really related to the Hydro-
gen atom problem would take for instance −a/ρ = γ + k, where k is an integer such that
k ≤ 0. In particular, for γ = l, and x = r/a, the CES potential V+ (x) derived in [26]
(k)
would be equal to our first order intertwined potential Vl−1 (r), provided that νlk = β/α,
where α and β are constants.
We are going to derive next some two-parametric families by means of the technique
discussed in sections 2.2 and 2.3 for the particular case we are dealing with.

3.3 The new two-parametric families

(km) (km)
In Section 2.2 we have seen that the second order intertwining relationship Hl−2 Bl−2 =
(km)
Bl−2 Hl provides new solvable potentials if we are able to find some solutions η(x) to
equation (11). Then we also have shown that the general form of that solutions is given
by:   
(k) −1
 d ul (r) 
(k) (m) (k) (m)
η(r) = (ǫl − ǫl ) ln  (m) , ǫl 6= ǫl . (51)
 dr ul (r) 
(k)
For our particular system we take the functions ul (r) as given by (37). The behaviour
of our η(r) at the ends of [0, ∞) is given by

2l + k + m (1 − 2l)
η(r)r→∞ ∼ − , η(r)r→0 ∼ − .
(l + k)(l + m) r
(km)
Hence, it is natural to write the new potential Vl−2 (r) as (see equation (11)):
(km)
Vl−2 (r) = Vl−2 (r) + 2α′(r), (52)

where α(r) ≡ η(r) + (1 − 2l)/r is an appropriate function making evident the limit
(km) (k)
Vl−2 (r) → Vl−2 (r) when r → +∞, or r → 0. The two-parametric domain of λl and
(m)
λl , for which α′ (r) is free of singularities, is determined by the parities of k and m. In
order to make transparent the choice of that domain we take the following convention: n

14
 (n) (n) (k) (m)
is the index labelling the factorization energy ǫl defined by ǫl ≡ max(ǫl , ǫl ), and s
(s) (k) (m)
is such that ǫl ≡ min(ǫl , ǫl ). Hence, we have:
 (n) (s)

 λl ∈ (−∞, 1), λl ∈ (−∞, 1) ; |n| even, |s| odd
(53)

 (n) (s)
λl ∈ (1, ∞), λl ∈ (1, ∞) ; |n| odd, |s| even.

 (n) (s)

 λl ∈ (−∞, 1), λl ∈ (1, ∞) ; |n| even, |s| even
(54)

 (n) (s)
λl ∈ (1, ∞), λl ∈ (−∞, 1) ; |n| odd, |s| odd
(km)
The eigenfunctions of Hl−2 are given by
(km)
(km) Bl−2 ψnl (r)
Ψn,l−2(r) =q , (55)
(m) (k)
(En − ǫl )(En − ǫl )

and
q (m) q (k)
(km) (k) (m) ul (r) (km) (m) (k) ul (r)
Ψǫk ,l−2 (r) = Clk ǫl − ǫl , Ψl−2,ǫm (r) = Clm ǫl − ǫl , (56)
W (k, m) W (m, k)

where Clk is given by (44). The corresponding eigenvalues are:

( ) ( )
(km) 1 1 1
El−2,K = − , − ∪ − , K > 0. (57)
(l + m)2 (l + k)2 (l + K)2
(k) (m)
Let us remark that the only restriction for the factorization energies, ǫl 6= ǫl ,
(km)
leads to the fact that η(r) and Vl−2 (r) are symmetric under the change k → m and
(km) (mk)
vice versa, Vl−2 (r) = Vl−2 (r). However, this symmetry is broken for the intermediate
(km)
one-parametric potentials arising when Vl−2 (r) is obtained after two iterated first order
(k) (m)
intertwining procedures. In order to see that, let us suppose that ǫl > ǫl . We have
(k)
two different ways to get the potentials (52): we can add first the level ǫl , and then the
(m)
level ǫl ; the other option is to perform the two operations in the opposite order.
(k)
(I) In the first case, after the first step we shall get the Hamiltonian Hl−1 , with
(k)
a spectrum equal to the spectrum of Hl plus a new level at ǫl ≤ El,K=1, where
El,K=1 = −1/(l + 1)2 is the ground state energy level of Hl . After the second step,
(km) (k)
the final Hamiltonian Hl−2 has the same spectrum as Hl−1 plus a new level at
(m) (k)
ǫl < ǫl . In both of these procedures the factorization energy (the ground state
energy of the new Hamiltonian) is less than the ground state energy of the initial
Hamiltonian, just as it happens in the SUSY approach.

15
 (II) Taking now the two operations in the opposite order, the first intertwined
(m)
Hamiltonian Hl−1 has a spectrum equal to the one of Hl plus a new level at
(m)
ǫl ≤ El,K=1, in agreement once again with the SUSY approach. However, the
(mk) (k) (m)
next intertwined Hamiltonian Hl−2 is derived by adding a new level at ǫl > ǫl ,
(m)
i.e., above the ground state energy level of Hl−1 , which is against the SUSY doctrine
but it is valid in our treatment.
(km) (mk)
A final point is that the equality of the two Hamiltonians Hl−2 and Hl−2 , gotten of
two iterations of the first order transformations, leads to two different factorizations of
the global second order intertwining operator
(km) (km) (k) (mk) (m) (mk)
Bl−2 = al−1 al = al−1 al = Bl−2

3.4 Particular two-parametric families

The simplest two-parametric family of potentials (52) corresponds to k = −1 and m = 0.
In this case, by means of (51) we get the solution to (14):
  −1
(−1)
1 − 2l  d 
er/l(l−1) Φl (r) 
η(r) = 2 ln (0)
, l > 1, (58)
l (l − 1)2  dr Φl (r) 

(0) (−1)
which has been recently reported in [14], where Φl (r), and Φl (r) are given by (47) and
(49), respectively. The appropriate two-parametric domain (see (53) and (54)) is given by
(−1) (0) (−1,0)
λl , λl ∈ (−∞, 1). From (57) and (34) it is clear that Vl−2 (r) is strictly isospectral
(−1) (0) (−1,0)
to Vl−2 (r). Notice that λl = λl = 0 leads to Vl−2 (r) = Vl−2 (r). Moreover, for
(0) (−1)
λl = 0 and λl = [(l − 1)/2]2l−1 (2l − 2)!/γl−1 , the potentials (52) become the family
(46) derived by Fernández, provided l is changed by l + 1. We make concrete now the
discussion of the cases (I) and (II) of Section 3.3.
(I) Departing from the Hamiltonian Hl , by means of a first order intertwining op-
(0)
erator we get the Hamiltonian Hl−1 , whose potential is given by (46), where we
(0)
have added the new level ǫl = −1/l2 below the ground state energy level of Hl ,
(0,−1)
El,K=1 = −1/(l + 1)2 . The iteration of this procedure leads to Hl−2 , whose po-
tential is given by (52) and (58). In this second step we have added the new level
(−1) (0)
ǫl = −1/(l − 1)2 below the previous ground state energy level ǫl = −1/l2 , which
agrees with the usual SUSY asumption.
(−1)
(II) We depart now from Hl to get Hl−1 , whose potential is given by (48). In
(−1)
this step, we have added the new level ǫl = −1/(l − 1)2 below the ground state
energy level of Hl but we have left a hole to be filled during the second step (see
(−1,0)
Section 3.2.2). The next step provides now Hl−2 , whose potential is given again
(−1,0) (0,−1) (0) (−1)
by (52) and (58). Notice that βl−1 (r) = βl−1 (r) + βl (r) − βl (r). The
(0)
new energy level ǫl = −1/l2 is now added above the ground state energy level

16
 (−1) (−1)
ǫl = −1/(l − 1)2 of Hl−1 but below the ground state energy level −1/(l + 1)2
of Hl . This last procedure fills the hole generated during the first step of the whole
procedure which is not typical in the usual SUSY asumption.

A different choice of k and m produces families whose spectra are almost equal to the
corresponding hydrogen-like spectra, with the two first energy levels different from the
levels of the corresponding hydrogen-like potential. For instance, if k = −3, m = 0, and
(−3,0)
l = 4, the potential V2 (r) has a spectrum of the form −1/n2 but without the levels
at n = 2 and n = 3, similarly as it happens in the first order case mentioned in Section
3.2.3. This is illustrated in Figure 2. A more exotic case is obtained if k = −4, m = −1,
and l = 5, where the levels with n = 2, n = 3, and n = 5 are not present for the potential
(−4,−1)
V3 (r).

4 Concluding remarks
The nth-order intertwining technique developed in this paper is intended to derive n-
parametric families of potentials which can be isospectral, or almost isospectral to some
well known potential. In particular, we have successfully derived new potentials from the
hydrogen-like atom for the one parametric and two parametric cases. We have shown also
that the derivation of multi-parametric families of almost isospectral potentials becomes
mathematically easy by the iteration of the first order intertwining technique. Although
the only trouble in our approach could be to find the solutions to a set of n Riccati type
equations, characterized by n different factorization energies, we have shown that the
eigenvalues problem for the initial potential allows one to get that solutions by means of
unphysical eigenfunctions. Our one-parametric results for the hydrogen-like atom show
how the technique effectively works, moreover, our resulting families of potentials are
more general than other results previously derived [1, 2, 4, 9]. On the other hand, our
two-parametric potentials represent a further generalization of the one-parametric cases
presented in this paper. Finally, we have shown that the construction of a SUSY partner
Hamiltonian using a factorization energy ǫ less than the ground-state energy of the depar-
ture Hamiltonian is not a general rule in SUSY QM. In particular, the generation of holes
in the spectra of the n-parametric families of almost isospectral hydrogen-like potentials
becomes the key-stone in the construction of the atypical n-SUSY partners of the Hydro-
gen potential derived in this paper. These techniques can be applied to other physically
interesting situations, e.g., for systems with continuous spectra as the free particle. A
detailed study in this direction will be given elsewhere [27].

Acknowledgements
This work has been supported by CONACyT, Mexico (Programa de Estancias Posdoctor-
ales en Instituciones del extranjero 1997-1998). The author would like to thank to Dr. D.
J. Fernández for fruitful discussions and suggestions. This work has been benefited from
the comments of Dr. L. M. Nieto and has been partially supported by Junta de Castilla

17
y León, Spain (project C02/97). The kind hospitality at Departamento de Fı́sica Teórica
and the suggestions of the referees are also aknowledged.

18
References
[1] E. Schrödinger, Proc. Roy. Irish Acad. 46 A, 9 (1940).

[2] L. Infeld and T.E. Hull, Rev. Mod. Phys. 23, 21 (1951).

[3] B. Mielnik, J. Math. Phys.25, 3387 (1984).

[4] D.J. Fernández C., Lett. Math. Phys. 8, 337 (1984).

[5] D.J. Fernández C., V. Hussin and L.M. Nieto, J. Phys A. 27, 3547 (1994);
D.J. Fernández C., L.M. Nieto and O. Rosas-Ortiz, J. Phys A. 28, 2693 (1995);
J.O. Rosas-Ortiz, J. Phys A. 29, 3281 (1996).

[6] S. Seshadri, V. Balakrishnan and S. Lakshmibala, J. Math. Phys. 39, 838 (1998).

[7] G. Junker and P. Roy, Supersymmetric construction of exactly solvable potentials and
non-linear algebras, PREPRINT quant-ph/9709021.

[8] G. Darboux, Compt. Rend. Acad. Sci. 94, 1456 (1882).

[9] E. Witten, Nucl. Phys B 188, 513 (1981); B 202, 253 (1982); M. M. Nieto, Phys.
Lett. B 145, 208 (1984); N. A. Alves and E. Drigo Filho, J. Phys. A. 21, 3215 (1988);
F. Cooper, A. Khare, U. Sukhatme, Phys. Rep. 251, 267 (1995).

[10] R.W. Carrol, Transmutation and Operator Differential Equations. North-Holland

Mathematics Studies 37, The Netherlands (1979).

[11] A. Andrianov and N.V. Borisov, Teor. Math. Phys. 61, 1078 (1985); A.A. Andrianov,
M.V. Ioffe and V.P. Spiridonov, Phys. Letts A, 174, 273 (1993); B.F. Samsonov, J.
Phys. A 28, 6989 (1995); V.M. Eleonsky and V.G. Korolev, Phys. Rev. A 55, 2580
(1997).

[12] D.J. Fernández C., Int. J. Mod. Phys. A 12, 171 (1997); D.J. Fernández, M.L. Glasser
and L.M. Nieto, Phys. Letts A 240, 15 (1998).

[13] J.O. Rosas-Ortiz, Ph.D. Thesis, CINVESTAV-IPN, México (1997).

[14] J.O. Rosas-Ortiz, J. Phys. A 31, L507 (1998).

[15] D.J. Fernández C., V. Hussin and B. Mielnik, Phys. Lett. A 244, 309 (1998).

[16] See for example the discussion given by Mielnik in [3] and by Fernández in [4].

[17] M. M. Crum, Quart. J. Math. 6, 121 (1955).

[18] E.L. Ince, Ordinary Differential Equations, Dover Pub. Inc., N.Y. (1926).

19
[19] An interesting application of Mielnik’s approach to the isotropic oscillator can be
found in Section 6 of D.J. Fernández C., J. Negro and Mariano A. del Olmo, Ann.
Phys. 252, 386 (1996).

[20] Some solutions to the Riccati equation for a particular case of the Hypergeometric
type classification, as given by Infeld-Hull in [2], as well as for the Dutra potential,
can be found in eqs (47) and (57) of N. Nag and R. Roychoudhury, J. Phys. A.
28, 3525 (1995). See also the discussion about exactly solvable models and Darboux
transformations in W. A. Schnizer and H. Leeb, J. Phys. A. 26, 5145 (1993).

[21] The author aknowledges to Dr. Luis M. Nieto by suggesting this ansatz.

[22] The case when k is different from a negative integer or zero will be discussed else-
where.

[23] Z.X. Wang and D.R. Guo, Special Functions, World Scientific, Singapore (1989).

[24] P.B. Abraham and H.E. Moses, Phys. Rev. A22, 1333 (1980).

[25] For the sake of clarity in the further discussions we have not used the original notation
for ω(x) as it has been given in [26], where it is W (x).

[26] G. Junker and P. Roy, Conditionally exactly solvable potentials. A Supersmmetric

construction method, PREPRINT quant-ph/9803024.

[27] B. Mielnik, L.M. Nieto and O. Rosas-Ortiz, in preparation.

20
 4

2 V 1 (r )

-2

V1( 1)
r
( ) (2 1)
=8
-4

2( 1)
= 15
-6
2( 1)
= 40

-8

r
1 2 3 4 5

Figure 1: Some members of the one-parametric family of potentials (48) for l = 2 and
(−1)
the indicated values of λ2 . The broken curve represents the hydrogen-like potential
(−1)
Vl−1 (r). Notice the arising of global minimum for V1 (r), which produces a well which
(−1)
can be modified by changing the values of the parameter λ2 .

21
 . .
. .
. .
= 2
1 5 = 2
1 5

1=4 1=4
2 2

=
1 3
2
=
1 3
2

=
1 2
2

=
1 1
2

( 3;0)
H2 H2

(−3,0)
Figure 2: Diagram showing the first energy levels of H2 and H2 . Notice the gap
(−3) (0)
between the ground state energy level ǫ4 = −1 and the first excited state ǫ4 = −1/42
(−3,0)
of the two-parametric family of Hamiltonians H2 .

22