See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344229659

Comment on: "Rashba coupling induced by Lorentz symmetry breaking

effects". Ann. Phys. (Berlin) 526 , 187 (2013)

Preprint · November 2020

DOI: 10.13140/RG.2.2.24070.19526

CITATIONS READS

0 10

1 author:

Francisco Marcelo Fernández

Universidad Nacional de La Plata
611 PUBLICATIONS   5,077 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ALGEBRAIC TREATMENT OF QUADRATIC HAMILTONIANS View project

CONDITIONALLY SOLVABLE QUANTUM MECHANICAL MODELS View project

All content following this page was uploaded by Francisco Marcelo Fernández on 13 September 2020.

The user has requested enhancement of the downloaded file.

 Comment on: “Rashba coupling induced by
Lorentz symmetry breaking effects” . Ann. Phys.
(Berlin) 526, 187 (2013)

Francisco M. Fernández∗
INIFTA, DQT, Sucursal 4, C.C 16,
1900 La Plata, Argentina

Abstract

We analyze the results of a paper on “the arising of a Rashba-like

coupling, a Zeeman-like term and a Darwin-like term induced by Lorentz
symmetry breaking effects in the non-relativistic quantum dynamics of a
spin-1/2 neutral particle interacting with external fields”. We show that
the authors did not obtain the spectrum of the eigenvalue equation but
only one eigenvalue for a specific relationship between model parameters.
In particular, the existence of allowed cyclotron frequencies conjectured
by the authors is a mere artifact of the truncation condition used to obtain
exact solutions to the radial eigenvalue equation.

In a paper published in this journal Bakke and Belich [1] study “the arising
of a Rashba-like coupling, a Zeeman-like term and a Darwin-like term induced
by Lorentz symmetry breaking effects in the non-relativistic quantum dynamics
of a spin-1/2 neutral particle interacting with external fields”. They derive an
eigenvalue equation for the radial coordinate and solve it exactly by means of
the Frobenius method. This approach leads to a three-term recurrence relation
∗ fernande@quimica.unlp.edu.ar

1
that enables the authors to truncate the series and obtain eigenfunctions with
polynomial factors. They claim to have obtained the bound-state eigenvalues
ad eigenfunctions of the model. Since the truncation condition requires that a
model parameter depends on the quantum numbers they conclude that not all
the cyclotron frequencies are allowed. In this Comment we analyze the effect of
the truncation condition used by the authors on the physical conclusions that
they derive in their paper.
It is not our purpose to discuss the validity of the model but the way in
which the authors solve the eigenvalue equation. For this reason we do not
show the main equations displayed in their paper and restrict ourselves to what
we consider relevant. We focus in the eigenvalue equation

1 δ2 α τs Rs
Rs′′ + Rs′ − s2 Rs − 3/4
ξRs − ξ 2 Rs − 1/4
+ W Rs = 0,
ξ ξ (2ma2 ) (2ma2 ) ξ
1 gbλ gbλ
δ 2s = γ 2s + 2ma1 , γ s = l + (1 − s), τ s = s γ + , α = gbλm,
2 4m s 8m
ζ
ζ = 2m (E − V0 ) , W = 1/2
, a2 = mω 2 (1)
(2ma2 )

where l = 0, ±1, ±2, . . ., s = ±1, m is a mass, E the energy, a1 , a2 , V0 parameters

of the model potential V (ρ) = a1 ρ−2 + a2 ρ2 + V0 and a, b, and λ are constants
that appear in the interactions included in the model. The authors choose units
such that h̄ = c = 1 although there are rigorous ways of deriving dimensionless
equations, as well as the choice of natural units [2].
The authors’ eigenvalue equation (1) is a particular case of

L̂R = W R,
d2 1 d γ2 a
L̂ ≡ − 2 − + 2 − + bξ + ξ 2 , (2)
dξ ξ dξ ξ ξ

where γ, a and b are arbitrary real numbers that have nothing to do with the
parameters in equation (1). Since the behaviour at origin is determined by the
term γξ −2 and the behaviour at infinity by the harmonic term ξ 2 we conclude
that there are bound states for all −∞ < a, b < ∞.

2
 By means of the ansatz
∞
∑
ξ2
R(ξ) = ξ |γ| e− 2 −
bξ
2 P (ξ), P (ξ) = cj ξ j , (3)
j=0

we derive a three-term recurrence relation for the coefficients cj :

b (2|γ| + 2j + 3) − 2a 4 (2|γ| + 2j − W + 2) − b2
cj+2 = cj+1 + cj ,
2 (j + 2) (2|γ| + j + 2) 4 (j + 2) (2|γ| + j + 2)
j = −1, 0, 1, . . . , c−1 = 0, c0 = 1. (4)

In order to obtain polynomial solutions the authors force the termination

conditions
8 (|γ| + n + 1) − b2
W = Wγ(n) = , cn+1 = 0, n = 1, 2, . . . . (5)
4
Clearly, under such conditions cj = 0 for all j > n and P (ξ) reduces to a
polynomial of degree n. In this way, they obtain analytical expressions for the
(n)
eigenvalues and the radial eigenfunctions Rγ (ξ). For the sake of clarity and
generality we will use γ instead of l as an effective quantum number.
For example, when n = 1 we have
√
8 (|γ| + 2) − b2 (1,1) 2b (|γ| + 1) − b2 + 8 (2|γ| + 1)
Wγ(1) = , aγ = ,
4 √ 2
2b (|γ| + 1) + b2 + 8 (2|γ| + 1)
a(1,2)
γ = , (6)
2
or, alternatively,
[ √ ]
2
2 2a (|γ| + 1) − a2 + 2 (2|γ| + 3) (2|γ| + 1)
b(1,1)
γ = ,
(2|γ| + 1) (2|γ| + 3)
[ √ ]
2 2
2 2a (|γ| + 1) + a + 2 (2|γ| + 3) (2|γ| + 1)
b(1,2)
γ = . (7)
(2|γ| + 1) (2|γ| + 3)
When n = 2 we obtain a cubic equation for either a or b, for example,

8 (|γ| + 3) − b2
Wγ(2) = ,
4
( ( ) )
4a3 − 6a2 b (2|γ| + 3) + a b2 12γ 2 + 36|γ| + 23 − 16 (4|γ| + 3)
( )
b (2|γ| + 1) b2 (2|γ| + 3) (2|γ| + 5) − 16 (4|γ| + 7)
− = 0, (8)
2

3
 (2) (2) (2,1) (2,2)
from which we obtain either aγ (b) or bγ (a); for example, aγ (b), aγ (b),
(2,3) (n,i)
aγ (b). In the general case we will have n + 1 curves of the form aγ (b),
(n,i) (n,i+1)
i = 1, 2, . . . , n + 1, labelled in such a way that aγ (b) < aγ (b) and it
can be proved that all the roots are real [4, 5]. Notice that Bakke and Belich
completely overlooked such multiplicity of roots.
It is obvious to anybody familiar with conditionally solvable (or quasi-
solvable) quantum-mechanical models (see [4–7] and, in particular, the remark-
able review [8] and references therein for more details) that the approach just
described does not produce all the eigenvalues of the operator L̂ for a given set
of values of γ, a and b but only those states with a polynomial factor P (ξ). Each
(n)
of the particular eigenvalues Wγ , n = 1, 2, . . . corresponds to a set of particular
(n,i)
curves aγ (b). On the other hand, if we solve the eigenvalue equation (2) in
a proper way we obtain an infinite set of eigenvalues Wν,γ (a, b), ν = 0, 1, 2, . . .
for each set of real values of a, b and γ. The condition that determines these
allowed values of W is that the corresponding radial eigenfunctions R(ξ) are
square integrable ∫ ∞
2
|R(ξ)| ξ dξ < ∞. (9)
0

Notice that ν is the actual radial quantum number (that labels the eigenvalues
in increasing order of magnitude), whereas n is just a positive integer that labels
some particular solutions with a polynomial factor P (ξ). In other words: n is a
fictitious quantum number given by the truncation condition (5).
It should be obvious to everybody that the eigenvalue equation (2) supports
bound states for all values of a and b and that the truncation condition (5) only
yields some particular solutions. Besides, according to the Hellmann-Feynman
theorem [3] the true eigenvalues Wν,γ (a, b) of equation (2) are decreasing func-
tions of a and increasing functions of b
⟨ ⟩
∂W 1 ∂W
=− , = ⟨ξ⟩ . (10)
∂a ξ ∂b

Therefore, for a given value of b and sufficiently large values of a we expect

negative values of W that the truncation condition fails to predict. It is not

4
difficult to prove, from straightforward scaling [2], that
Wν,γ 1
lim =− 2. (11)
a→∞a2 (2ν + 2|γ| + 1)
[ ]
(n,i) (n)
What is more, we can conjecture that the pairs aγ (b), Wγ , i = 1, 2, . . . , n+
1 are points on the curves Wν,γ (a), ν = 0, 1, . . . , n, respectively, for a given value
of b.
The eigenvalue equation (2) cannot be solved exactly in the general case.
In order to obtain sufficiently accurate eigenvalues of the operator L̂ we resort
to the reliable Rayleigh-Ritz variational method that is well known to yield
increasingly accurate upper bounds to all the eigenvalues [9] (and references
therein). For simplicity we choose the basis set of non-orthogonal Gaussian
{ ξ2
}
functions uj (ξ) = ξ |γ|+j e− 2 , j = 0, 1, . . . and test the accuracy of these re-
sults by means of the powerful Riccati-Padé method [10].
(2)
As a first example, we choose n = 2, γ = 0 and b = 1 so that W0 = 5.75 for
[ ] [ ]
(2,1) (2,2)
the three models a0 = −1.940551663, b = 1 , a0 = 1.190016441, b = 1
[ ]
(2,3)
and a0 = 5.250535221, b = 1 . The first four eigenvalues for each of these
models are





W0,0 = 5.750000000


 W = 9.894040660
(2,1) 1,0
a0 → ,




W2,0 = 14.06831985


 W = 18.24977457
3,0


 W0,0 = −0.1664353619




 W = 5.750000000
(2,2) 1,0
a0 → ,




W2,0 = 10.52307155


 W = 15.06421047
3,0


 W0,0 = −27.32460313




 W = −0.5108147276
(2,3) 1,0
a0 → .




W2,0 = 5.750000000


 W = 10.90599171
3,0

(2)
We appreciate that the eigenvalue W0 = 5.75 coming from the truncation con-

5
dition (5) is the lowest eigenvalue of the first model, the second lowest eigenvalue
of the second model and the third lowest eigenvalue for the third model. The
truncation condition misses all the other eigenvalues for each of those models
and for this reason it cannot provide the spectrum of the physical model for any
set of values of γ, a and b as suggested by Bakke and Belich.
In the results shown above we have chosen model parameters on the curves
(2,i)
a0 (b). In what follows we consider the case a = 2, b = 1 that does not belong
to any of those curves. For this set of model parameters the first five eigen-
values are W0,0 = −3.230518994, W1,0 = 4.510929109, W2,0 = 9.532275968,
W3,0 = 14.19728140 and W4,0 = 18.70978427. As said above: there are square-
integrable solutions (actual bound states) for any set of real values of a, b and
γ. The obvious conclusion is that the dependence of the frequency ω on the
quantum numbers n, l, s (ω n,l,s ) and the consequent allowed cyclotron frequen-
cies conjectured by Bakke and Belich [1] are just artifacts of the truncation
condition (5). Such claims are nonsensical from a physical point of view. To be
clearer, since there are bound states for all a and b then there are bound states
for all ω.
(n)
Figure 1 shows some eigenvalues W0 (b = 1) given by the truncation condi-
tion (red points) and the lowest variational eigenvalues Wν,0 (a, 1) (blue lines).
We clearly appreciate that the truncation condition (5) yields only some partic-
(n)
ular points of the curves Wν,0 (a, 1). Therefore, any conclusion drawn from Wγ
is meaningless unless one is able to organize these eigenvalues properly [5–7].
Bakke and Belich [1] completely overlooked this fact. The reason is that these
authors appear to believe that the only acceptable solutions to the eigenvalue
equation are those with polynomial factors P (ξ). The fact is that this kind of
solutions already satisfy equation (9) but they are not the only ones. Notice
that the variational method also yields the polynomial solutions as shown by
the fact that the blue lines connect the red points in Figure 1. In order to make
(n)
the meaning of the eigenvalues Wγ and the associated multiplicity of roots
i = 1, 2, . . . , n + 1 clearer, Figure 1 shows an horizontal line (green, dashed) at
(8)
W = W0 that intersects the curves Wν,0 (a, 1) exactly at the red points. The

6
most important conclusion of present analysis is that the occurrence of allowed
oscillator frequencies are fabricated by Bakke and Belich by picking out some
(n) (n,i)
isolated eigenvalues Wγ for some particular curve aγ (b). Since there are
eigenvalues Wν,γ (a, b) for all real values of a and b then there are bound states
for every positive value of ω in their equations (1).
Summarizing: The authors make two basic, conceptual errors. The first
one is to believe that the only possible bound states are those with polynomial
factors P (ξ). We have shown above that there are square-integrable solutions
(n,i)
for model parameters a and b outside the curves aγ (b) associated to these
polynomials. The second error is the assumption that the spectrum of the
problem is given by the truncation condition (5). It is clear that this equation
only provides one energy eigenvalue for a particular set of model parameters
given by the curves just mentioned. From these mistakes the authors conjecture
the existence of allowed cyclotron frequencies. Here we have shown that such
allowed cyclotron frequencies are fabricated by Bakke and Belich by means
of the truncation method. Therefore, such conclusion is nonsensical from both
mathematical and physical points of view. It is clear that there are bound states
for all values of ω because the eigenvalues Wν,γ (a, b) are continuous functions
of both a and b.
Dazed and confused by a plethora of pseudo-physical garble the reviewers
and editors of a once respectable journal publish a paper with obvious errors and
nonsensical conclusions.

References
[1] K. Bakke and H. Belich, Ann. Phys. (Berlin) 526, 187 (2013).

[2] F. M. Fernández, Dimensionless equations in non-relativistic quantum me-

chanics, arXiv:2005.05377 [quant-ph].

[3] R. P. Feynman, Phys. Rev. 56, 340 (1939).

[4] M. S. Child, S-H. Dong, and X-G. Wang, J. Phys. A 33, 5653 (2000).

7
 45

40

35
30

25
Wν,0
20
15

10

5
0
-5
-50 -40 -30 -20 -10 0 10 20 30 40 50
a

(n)
Figure 1: Eigenvalues W0 (a, 1) from the truncation condition (red points) and
Wν,0 (a) obtained by means of the variational method (blue lines)

[5] P. Amore and F. M. Fernández, Phys. Scr. 95, 105201 (2020).

arXiv:2007.03448 [quant-ph]

[6] F. M. Fernández, The rotating harmonic oscillator revisited,

arXiv:2007.11695 [quant-ph].

[7] F. M. Fernández, The truncated Coulomb potential revisited,

arXiv:2008.01773 [quant-ph].

[8] A. V. Turbiner, Phys. Rep. 642, 1 (2016). arXiv:1603.02992v2

[9] F. L. Pilar, Elementary Quantum Chemistry (McGraw-Hill, New York,

1968).

[10] F. M. Fernández, Q. Ma, and R. H. Tipping, Phys. Rev. A 39, 1605 (1989).

View publication stats