PHYSICAL REVIEW A VOLUME 50, NUMBER 5 NOVEMBER 1994
Expanding the
class
of
conditionally
exactly
solvable
potentials
A.
de Souxa DutraUniuersidade Estadual Paulista Ca-mpus de Guaratingueta-DFQ, Auenida Dr.Ariberto Pereira da Cunha, 333, Codigo de Enderegamento Postal 12500-000, Guaratingueta,
Sio
Paulo, BrazilFrank Uwe Girlich
Fachbereich Physik, Universitat Leipzig, Augustusplatz 10,04109Leipzig, Germany
(Received 30December 1993)
Recently a class ofquantum-mechanical potentials was presented that is characterized by the fact that they are exactly solvable only when some oftheir parameters are fixed to a convenient value, so they
were christened as conditionally exactly solvable potentials. Here we intend toexpand this class by
in-troducing examples in two dimensions. As a byproduct ofour search, we found also another exactly
solvable potential.
PACSnumber{s): 03.65.Ca,03.65.Ge, 02.90.
+
pRecently the existence
of
a whole classof
exactly solv-able potentials in quantum mechanics was discovered[1].
Furthermore it was observed that these potentials obey a
supersymmetric algebra
[2].
In fact the potentials dis-cussed inRef. [1]
were also discussed independently in a previous workof
Stillinger[3].
In that article the author did not perceive that such potentials were only the first representativesof
an entire new classof
potentials. What distinguishes these potentials from the usual exactly solv-able potentials isthe fact that at least oneof
their param-eters must be fixedto
a specific value, becauseif
you donot do that the potentials no longer have exact solutions. On the other hand, they cannot be classified among the so called quasi exactly solvable potentials [4
—10],
because these have only afinite numberof
exactly solvable energy levels, and this happens only when the potential parame-ters obey some constraint equations, sothat only someof
the parameters are free, and the others are fixed in terms
of
them. The continuous interest in obtaining andclassi-fying quantum-mechanical potentials is because
of
their importance as a basis for expanding more realistic cases, and also for useof
their solutions to test numerical calcu-lations. Another possibility is thatof
trying to use them as a kindof
theoretical laboratory, searching for new in-teresting physical features. Someof
the examples that we will study below have, in fact, the interesting characteris-ticof
representing a kindof
semi-infinite charged string, because, ascan beseen inFig.
1,the potential isinfinitelyattractive along the
x
semiaxis. Here in this work we in-tendto
expand the classof
conditionally exactly solvable(CES) potentials. This is going to be done through the presentation
of
two examples in two dimensions. As a byproductof
our search forCES
potentials, we found another exactly solvable potential, whose solution willalso be presented.
The potentials that will be considered here are
and
Vn(x, y)
=
X A
arctan—
2/3
8
[arctan(y/x) ]
+
go+C
[arctan(y
/x)
] (lb)20
V p
where the function
arctan(y/x)
isdefined in the principalbranch.
It
can be observed that the first case above has an exact solution forany valueof
Goif
8
=0.
Otherwisethe case
8%0
has an exact solution forGo=
—
3irt /32@,and only for a restricted region
of
the space (x and y both bigger orboth less than zero).The second potential has exact solutions only when
go=
—
5trt /32@. In this case the solution is valid along all the x-y plane.It
isnot difBcult toverify that these po-tentials are singular along the positive semiaxisx,
as canbe seen from
Fig. 1.
Sothey represent akindof
"chargedstring.
"
V,
(x,y)
=
1x
+y
B
arctan(y
/x
) [arctan(y/x
}]'
2Go
+
+C
[arctan(y
/x)
]+D(x +y
)FIG.
1. Plotofthe potential CESII
with unitary parameters.4370
BRIEF
REPORTSd
+
V,(u)
—
6''4(u)=0
t2p du
where
(C+K —
iri /8p,)with
K
being the separation constant andp
the massof
the particle. The subscript e in the potential stands for
its exact solvability. In the variable vwe have d
&
+
VCEs(u)Eg(v—
)=0,
2p dU
with VCEs(u) having two possibilities:
8
GoVcEs(u)
=
+
+
U
(4a)
The solution for these potentials can be obtained through a suitable change
of
coordinates:x
=u
cos(u),
y=u
sin(u) .After this change
of
variables, and their separation, the Schrodinger equation can be split into two others. In the variable u we have thatlike multivaluedness
of
the wave function. However, onecan see that this transformation is typical
of
a stereo-graphic mapping, in this caseof R
onto5
. This kindof
compactification can be done provided that one has suit-able asymptotic conditions
[11].
In fact, this implies that one will have a finite wave function throughout variationof
the variables, and particularly that at infinity it should vanish. This imposes the condition that the parameter Dbe nonzero, because along the positive
x
semiaxis thewave function will be singular at infinity
if
D=0.
Fur-thermore, there is no periodicity
of
the wave function inthe variables
x
and y, aswe will see later.So we see that the original potentials were, in some
cases, reduced to one-dimensional
CES
potentials, whose solutions were obtained previously. Now we proceed tothe solution
of
the above mentioned possibilities.Case
I
for
8 =0
and Gu arbitrary The fi.rst example isthat
of
an exactly solvable potential. In this case the po-tential V,(x,
y) ismapped into a harmonic oscillator with centrifugal barrier in the variable u, and a Coulomb pluscentrifugal barrier in the variable U. Sowe only need to
identify the corresponding parameters, substitute them in
the equation for the "energy,
"
which in the transformed equation is the parameter K,and solve it for the original eAergpand
VCEs(u)II
=
Av2/3+
+
(4b)8 A D
(,,
'„„(a=0)=
4CD-
A'[2m+
I+a,
]'
(2n
+1),
1/2
The wave function can be obtained by returning
%(u,u)
=u
'4(u)y(u)
tothe original variables.At this point, we shall observe that the transformation used leads us to a mapping into an angle variable
u
=
arctan(y /x ), which can, in principle, create problemsI
where a&
—
=
(I+8pGO/fi
)'~ . It can be seen that theabove spectrum is a kind
of
mixingof
the Coulomb and harmonic oscillator spectra. The normalized wave func-tions areqI,B=0( ) 32pD
f2
1/8
21'(n
+1)
1(a,
+n+1)
'1/2 '1/2
AD
$2 (X
+y2)
01 /2
arctan—
X(2~)+1)/4 1/2
—
SpK
$2
I(m
+1)
I
(a&+m+1)
arctan-y X
j
.1/2
arctan
—
y exppD
2%2 1/2
—
8@K(X2+y
)—
( 1/2
AD
(z+
z)J 1/2
—
2pE
arctan—
$2 X wherep
2AD
1/2—
(2n+1),
(2n
+1)
—
C,
2pA (2m
+
a,
+
I)iri1/2 2
1 ( fi 2D
x=-
(7b)4C
"'p
pand
L„'(x,
y) is the generalized Laguerre polynomial.It
isremarkable that, as can be seen from Eq. (5), the domain
of
validityof
the parameter C is defined through tequation
in order to keep the energies real. So, as the left hand side
of
the above equation has its minimum value whenm
=0
(n arbitrary), we must have C~
2@A/(a
& +1)A' .Apparently
if
C is less than this we should impose the condition thatwhere Int( ) stands for taking the first integer greater
50
BRIEF
REPORTS 4371with
'1/2
2A D
K
1(2n+1)
—
C
.(11)
p
As we are looking for
8'„,
we see that it is necessary tosolve athird order polynomial equation for
E,
and then use this solution inthe expression forthe energy' 1/2
2A D
=+2&D
(K
+C)'
+
(2n+1);
be verified through numerical calculation. Perhaps we
would need some kind
of
analytical continuation for the energy in such cases. Now, the next case.Case
I
for
8%0
andGo=
—
3A'/32p.
In this case wehave our first example
of
a two-dimensionalCES
poten-tial. After repeating the previous procedure we get the following equation forthe energy eigenstates:
pB4
i'
(m+
—,')K
+2pA K
+pAB
K
+
=0,
(10)K
m=(R +D)' +(R
D—
)'
a13 (13a)
with
D—
:
+Q
+R,
Q:—
(3az—
a&)/9,
R:—
(9a&a2—
27a&
—
2a,
)/54, and2@A
pAB
vari (m
+
—
')
fi (m+
—')
B4
8'
(m+
—,')
(13b)
once more we have a limitation on the parameter C. The
solution
of
the equation for the parameterE
isobtained following the procedure appearing inRef.
[1].
This givesus
(12) The wave function is given by
'1/2
('+')
$2
'1/2
'
1/8
tg
(x,
y)=
32p,D
En+1)
2 +
m!I
(a+n+1)
1/2
AD
B
f2
(x
+y
)H
P
&arctan(y/x)—
2E
XL:
B
Xexp
—
z(x
+y
)—
&arctan(y/x)—
2E
2
a/2 '1/4
arctan-x
(14)
where we defined
' 1/2
—
(2n+1),
2A2D
and
2R
D
p
1/2
(2n
+
1)—
C,
(15a)'1/2
1/2
+4v'A
(m+1/2)iri
+
—
3 24
p
8pK
$2
'1/4 with the solution
' 1/2 1/2
2
9iri A
+
98
n, m
2
p
4and
H
(x,
y)isthe Hermite polynomial.It
is easyto
verify that the wave function, being pro-portional toarctan(y/x),
vanishes along the positivex
semiaxis. This should be expected due to the singularity
of
the potential in this region. Moreover, it is easy tosee from the wave function (14) that, as observed previouslywhen we presented the potential, one must restrict the re-gion
of
space such thatx
and yare both positive or both negative. Furthermore, the inverse tangent function is defined in the principal branch, as said previously.Final-ly we treat the second case.
Case
II
(go=
—
5R/72p).
Now we present the last ex-ampleof
a two-dimensionalCES
potential that will be discussed in this work.If
we repeat once more the same procedure we get the following equation for the energy in this last case:1/2 1/2
+4CD
+
p
(2n+
1).
(15b)B~—
(16)in order toinclude m
=0
in the solution. Besides, the pa-rameterC
isalso restricted byC)
4&A
(
+,
)9A' A
3 2p
1/2 1/2
(17) This time we need to do a careful analysis
of
the rangeof
validity for the potential parameters. In the caseof
the parameter
B,
we have the restrictionBRIEF
REPORTSSowe see that the negative solution (positive in the above equation) introduces a limit on the number
of
boundstates. The positive case (negative above) is such that we
only need impose that
C
is bigger than the left hand sideof
Eq. (17) with m=0.
After this brief analysis
of
the regionof
validityof
the potential parameters, we present the corresponding wavefunctions:
] /8
(x,
y)=
32D
tl, m
3r(
+1)
2
'm!I
(a+n
+1)
2pa
$2
(x
+y
)~/2 2/3 ]1/4
z~ arctan
—
P
XL„'
1/2
(x'+y
)8
~Parctan—
2/3
pD
Xexp
2A
][/2
(x+y
)——
2
arctan—
3' X
E
'2A
where
p
2A
D
—
(2n+1)
and
']/4
9@A 2A
(19)
This finishes the presentation
of
examplesof
CES
poten-tials, now in two-dimensional space. The principal feature presented by these potentials, besides belonging to theCES
classof
potentials, isthat astringlike singularityof
the potential appears. This iscorroborated by the fact that the probabilityof
finding the particle over this singu-larity iszero. This accounts for the interest in doingfur-ther investigations
of
the characteristicsof
suchpoten-I
tials.
It
would be interesting to study the behaviorof
their corresponding coherent states, and also see what happens with wave packets around the linear singularity
of
the potential.Furthermore, in analogy with what has been done with
the exact potentials
[12,13],
and also with the quasiexact ones [8,9],
it is our intention to look for the dynamical algebra behind the solvabilityof
theCES
potentials. This study is presently under development and we hope tore-port on itin the near future.
One
of
the authors (A.S.
D.
) thanks CNPq (ConselhoNacional de Desenvolvimento Cientifico e Tecnologico)
of
Brazil for partial financial support andFAP ESP
(Fundal'Ko de Amparo a Pesquisa no Estado de Sao Pau-lo)forproviding computer facilities.
[1]A.de Souza Dutra, Phys. Rev.A 47,R2435(1993).
[2]E.Papp, Phys. Lett. A17$,231(1993). [3]
F.
H.Stillinger,J.
Math. Phys. 20,1891(1979).[4]
G.
P.Flessas, Phys. Lett. 72A, 289(1979);78A, 19(1980); 81A,17(1981);J.
Phys. A14, L209 (1981).[5]P.G.L.Leach,
J.
Math. Phys. 25, 974 (1984); Physica D 17,331(1985).[6] A.deSouza Dutra, Phys. Lett.A 142,319 (1988).
[7]R.S.Kaushal, Phys. Lett. A142, 57(1989).
[8]M.A.Shiffman, Iut.
J.
Mod. Phys. A 4, 2897(1989).[9] A.de Souza Dutra and H.Boschi Filho, Phys. Rev. A 44, 4721(1991).
[10]L. D.Salem and R.Montemayor, Phys. Rev. A 43, 1169 (1991).
[11]
T.
Eguchi, P. B.Gilkey, and A.J.
Hanson, Phys. Rep. 66, 213 (1980).[12] A. O.Barut, A. Inomata, and
R.
Wilson,J.
Phys. A 20, 4075(1987);20,4083(1987}.[13]H.Boschi-Filho and A.N. Vaidya, Ann. Phys. (NY) 212,