Variation method applied to the harmonic oscialltor in presence of a delta function

potential
Indrajit Ghose1 and Parongama Sen1
1
Department of Physics, University of Calcutta, 92 Acharya Prafulla Chandra Road, Kolkata 700009, India.
The problem of the harmonic oscillator with a centrally located delta function potential can be
exactly solved in one dimension where the eigenfunctions are expressed as a superposition of the
Hermite polynomials or as a Tricomi function. The eigenfunctions obtained exactly are difficult
to visualise and hence gain more insight. One can attempt to use model wavefunctions which
are explicitly and simply expressed. Here we use a variational principle to verify how close one
can approach the exact values using such trial wavefunctions. Both one and two parameter wave
functions have been used.

I. INTRODUCTION value of ν. Such solutions, on the other hand, are invalid

when the delta function is absent.
The harmonic oscillator with the Hamiltonian A lot of insight is gained in the appraoch adopted in [2].
Changing theq variables to dimensionless variables with
~2 2 1 2
− ∇ + mω 2 r2 (1) y = xa ; a = mω~
, δ(x) = γa δ( xa ) = γa δ(y),  = ma
~2 E, g =
2m 2 maγ
~2 , we get the reduced Schrödinger equation
is one of the most well known Hamiltonians in quantum
mechanics that has been exactly solved in one and higher
dimensions, the solutions being the Hermite polynomials. ∂2ψ
The energy eigenvalues are En = (n + 1/2)~ω with n = + (2 − y 2 )ψ − 2gδ(y)ψ = 0 (3)
∂y 2
1, 2, ..... in one dimension. The solutions are of definite
parity in all dimensions. One can further use the notation  = ν + 1/2 as in [2].
In this article we consider the harmonic oscillator in The solution for g = 0 can be obtained in the form
one dimension in presence of a centrally located delta
y2
function potential such that the Schrödinger equation ψ(y) = e− 2 w(y). (4)
obeyed by the eigenfunctions ψ is
The solutions for the differential equation obeyed by w(y)
~2 ∂ 2 ψ 1
− + mω 2 x2 ψ + γδ(x)ψ = Eψ. (2) are expressed in terms of Tricomi and Kummer functions
2m ∂x2 2 and for integer values of ν, the solutions reduce to the
This problem can also be exactly solved. In presence of well known Hermite polynomials. However, non-integer
the delta function, the odd parity eigenfunctions are not values of ν are in principle allowed but the solutions not
affected. The energy for the even parity eigenfunctions acceptable for g = 0 due to physical inconsistencies. On
can be expressed in terms of a transcendental equation. the other hand, in preesence of the delta function, it is
The delta function potential demands that there is a the Tricomi function with non-integer values of ν which
discontinuity in the first derivative of ψ at the origin. is an appropriate solution with a discontinuity in the first
Problems in quantum mechanics in presence of a delta derivative at the origin. Precisely, the energy eigenvalues
function potential can be solved using a general prescrip- are obtained solving the transcendental equation
tion given in [1] in one or higher dimensions. For the free
particle and the particle in a box in presence of a delta Γ(1 − ν2 )
ν−g = 0. (5)
function potential, the energy eigenvalues are easily ob- Γ( 12 − ν2 )
tained as one can still use the eigenfunctions in absence
of the delta function and obtain the solutions in a fairly This equation has to be solved numerically to obtain the
straightforward way, using the discontinuity condition. energy eigenvalues.
However, for the harmonic oscillator, this is not possible Although the eigenfunctions are precisely identified in
as the first derivative of the hermite polynomial is con- [2], these functions are expressed implicitly in terms of
tinuous at the origin. Various methods have been used integrals and not simple to visualise.
to obtain the eigenvalues exactly [1–3] or approximately An important question is whether one can gain useful
[4]. In [1] a superposition of the hermite polymonials information by considering model wavefunctions which
was used. In [2] it was shown there exist solutions other have explicit expressions and this approach was taken in
than the hermite polynomials of the Schrödinger equa- [4]. The expectation value was calculated using a model
tion (even without the delta function) and which have wavefunction and compared with the exact values. Here
the property of discontinuity of the first derivative at we attempt to construct trial wavefunctions paramet-
the origin. The energy eigenvalues which can still be ex- rically obyeing the basic criteria and use a variational
pressed in the form (ν + 1/2)~ω albeit with a non-integer method to obtain the optimal values of the parameters.
 2

We have considered both one parameter and two param- following criteria
eter trial wavefunctions and obtained the results for the (a) Definite parity
ground state energy which are closer to the exact values (b) Vanish at infinity
compared to those in [4]. (c) The first order derivative must have a discontinuity.

Taking a cue from [4], we consider trial solutions in the

II. VARIATIONAL METHOD FOR form
ATTRACTIVE DELTA
y2
ψ(y) = AeZ|y| e−α 2 (7)
The model wavefunction used in [4] for g > 0 is
for g < 0.
y2
ψ(y) = Aeg|y| e− 2 (6) In our scheme, we have taken three cases
(i) keep α = 1, vary Z
for g < 0. (ii) keep Z = g, vary α
We note that any trial wavefunction has to satisfy the (ii) Vary both Z and α

R∞ 2 2
α2 1 − α4 0 y 2 e2Zy e−α y g2 g
E = hHi = + R∞
2Zy −α 2 y2
+ + R ∞ 2Zy −α2 y2 . (8)
2 2 0
e e dy 2 2 0 e e dy

A. Different cases

R∞ 2 2
1. Z = g, α = 1 α2 1 − α4 0 y 2 e2gy e−α y dy
E= + R∞
2 2 0
e2gy e−α2 y2 dy
This produces the results in [4]. The energy estimates g2 g
are close to the exact solutions. + R ∞ 2gy −α2 y2 . (11)
2 2 0 e e dy

2. Variable Z, α = 1 We call the value of parameter α corresponding to min-

ima of E from eq. (11) as α1 .
α = 1 corresponds to pure harmonic oscillator and
From eq. (8) we have
α = 0 corresponds to pure δ potential. So, 0 < α < 1.

1 Z2 g
E= + + √ Z2 . (9)
2 2 πe erfc(−Z)
4. Variable Z and α

Differentiating eq. (9) with respect to Z we get

Instead of differentiating eq. (8) with respect to Z and
α we create a matrix with values of E for different values
dE 2gZ of Z and α at certain interval. We find out the minimum
= Z − √ Z2 E value
dZ πe erfc(−Z)
g 2 2
− √ Z2 √ e−Z = 0. (10)
πe (erfc(−Z))2 π

B. Summary of results
We call the solution of eq. (10) as Zmin .

As g → 0 it corresponds pure harmonic oscillator po-

tential (α → 1). As g → ∞ i.e. for larger values of g
3. Z = g and variable α the harmonic oscillator potential can be neglected. So,
Z → g and α → 0. The data from table I is consistent
From eq. (8) we have with the above physical conclusions.
 3

60 -1.005
E vs Z E(alpha)

50 -1.01

40
-1.015

E(alpha)
30

E(Z)
-1.02

20
-1.025

10
-1.03
0
-1.035
-10 Z 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-10 -8 -6 -4 -2 0 2 4 6 8 10 alpha

70

60

50

40
Energy

30

20

10

0 -1
0

-8

0
-1
0

-6
1

-4
0

-2
0
-1

FIG. 1: Plot of energy estimate as a function of the parameters. The first figure gives the variation of E(Z) from eq. (9)
against Z for g = −0.5. The second figure gives the variation of E(α) from eq. (11) as a function of α for g = −2.5. Third
gives the variation of E(Z, α) from eq. (8) for g = −5.0. All of the figure confirms the existence of global minima. Variation
of parameter gives us the best estimate.

variable Z variable α two parameter

g ν(Z = g) Zmin ν(Zmin ) α1 ν(α1 ) Z1 α1 ν(Z1 , α1 ) ν(exact)
-5.0 -12.981750 -4.90709 -12.985895 0.219050 -12.989190 -4.998020 0.237299 -12.989191 -12.990313
-3.0 -4.955630 -2.862564 -4.964406 0.362841 -4.972539 -2.991810 0.394745 -4.972557 -4.972771
-2.5 -3.565851 -2.34641 -3.576601 0.426004 -3.586291 -2.498231 0.431228 -3.586302 -3.5865066
-2.0 -2.418161 -1.829619 -2.431137 0.507489 -2.442049 -1.979531 0.548009 -2.442153 -2.442360
-1.5 -1.506601 -1.315595 -1.521389 0.609249 -1.532188 -1.469000 0.651149 -1.532421 -1.532729
-1.0 -0.819484 -0.818582 -0.833668 0.728889 -0.841655 -0.958952 0.767878 -0.842035 -0.842418
-0.5 -0.333176 -0.367230 -0.341016 0.860943 -0.343908 -0.460010 0.889194 -0.344197 -0.344424

TABLE I: Summary of results. The second collumn gives E − 12 where Z = g and α = 1. The fourth collumn gives E − 12 when
Z is variable and α = 1. The sixth collumn gives E − 21 when Z = g and α is variable. The ninth collum gives value of E − 12
when both Z and α are variable. The tenth collumn gives the exact solutions of eq. (5).
 4

0.03
Z=g
Variable Z
0.025 Variable alpha
Variable Z,alpha

0.02
error

0.015

0.01

0.005

0
-5 -4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5
g

2
Z=g
1 Variable Z
0 Variable alpha
Variable Z,alpha
relative percentage error

-1
-2
-3
-4
-5
-6
-7
-8
-9
-5 -4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5
g

FIG. 2: The first figure plots the error = (energy estimate-exact energy eigenvalue) vs strength of delta potential g.
The purple line and the green line corresponds to the error in estimate keeping Z = g, α = 1 and variableZ, α = 1 respectively.
The light blue and yellow lines give the error keeping Z = g, variable α and variable Z, variable α respectively.
The second figure illustrates relative percentage error(rpe)= energy estimate-exact energy eigenvalue
exactenergyeigenvalue
× 100. The log(rpe) is along y axis
and g is along x axis. In the range 0 > g > −3.0 log(rpe) falls according to some power laws with decreasing g i.e. increasing
strength of attractive delta potential.
For Z = g, α = 1 (yellow line,purple dot) it falls as 1.77823g 0.908048
For variable Z, α = 1 it falls as 0.71461g 0.806717
For Z = g, variable α it falls −1.09748g 1.51496
For variable Z, variable α it falls as −2.10996g 1.18538
Hence we see using a one-parameter trial wavefunction starts with a lower error and also falls more rapidly (in case of variable
α) with decreasing g. Also a two-parameter wavefunction gives a much lower error
 5

III. VARIATIONAL PRINCIPLE FOR (Note : For attractive delta the trial wavefunction
REPULSIVE DELTA is of the form (bound state of delta)×(bound state of
harmonic oscillator). But, repulsive delta potential do
For g > 0 the again taking a hint from [4] the trial not have a bound state.)
wavefunction is

α2 y 2
ψ(y) = A(1 + Z|y|)e− 2 . (12) The energy estimate is

√
√ π Z 3Z 2 √
α2 Z2 π 1 − α4 4α 3 + α4 + 8α5 π g
E= + √ √ + √ + √ . (13)
2 2 π 2Z Z 2
2 π Z Z2 √ π 2Z Z2 √
α + α2 + 2α3 π 2α + α2 + 4α3 π α + α2 + 2α3 π

Once again we consider different cases. We call the solutions of eq. (15) Zmin . This is the
value of the parameter Z which gives the minima in E
from eq. (14).
A. Different cases

1. Z = g and α = 1
B. Summary of results
It reproduces the results in [4].
IV. VARIATIONAL PRINCIPLE FOR EXCITED
STATES
2. Variable Z and α = 1
Any odd trial wavefunction will be automatically or-
From eq. (13) we have thogonal to the trial wavefunctions in eq. (??) and eq.
(??). We take the trial wavefunction as
√
α2 Z2 π 1 g
E= + + (14)
2
√ 4α B 2B α2 y 2

π Z Z2 √ ψ(y) = AxeZ|y| e− 2 . (16)

where B = + 2 + 3 π.
2α α 4α
The above wavefunction will do for both g > 0 and
Differentiating eq. (9) with respect to Z we have g < 0. The energy expectation value is

√ √
dE Z π 1 Z2 π g R∞
= − 2
B1 − B1 (15) 3α2
2 2
(1 − α4 ) 0 y 4 e2Zy e−α y dy Z 2
dZ 2α B √ 4αB 2B 2 hHi = + R∞ + . (17)
π Z Z2 √ 2 2 0
y 2 e2Zy e−α2 y2 dy 2
where B = + 2+ 3 π
2α α 4α
1 Z √ Eq. (17) is independent of g which is physically con-
where B1 = 2 + 3 π.
α 2α sistent.

[1] Daniel A. Atkinson, Horace W. Crater, Am. J. Phys. 43 Phys. Lett. A 124 (1987) 233
(1975) 301 [6] David J. Griffiths, Intro. to Q. Mech., Pearson(2nd Ed.)
[2] J. Viana-Gomes and N.M.R. Peres, Eur. J. Phys. 32 [7] Milton Abramowitz and Irene Stegun, Handbook of Math-
(2011) 1377 ematical Functions, Dover, New York, 1965
[3] N. Ferkous, T. Boujedda, Commun. Theor. Phys. 67 [8] R. Loudon, Am. J. Phys. 27 (1959) 649
(2017) 241-249 [9] Patrick Shea, Brandon P. van Zyl, Rajat K. Bhaduri,
[4] S. H. Patil, Eur. J. Phys., 27 (2006) 899 Am. J. Phys. 77 (2009) 511-515
[5] M. P. Avakian, G. S. Pogosyan, and A. N. Sissakian, [10] T. Bush, B. G. Englert, K, RzaSewaski, M. Wilkens,
 6

2 2.8
E(Z) E(\alpha)
1.9 2.6

1.8 2.4

1.7 2.2

E(\alpha)
1.6 2
E(Z)

1.5 1.8

1.4 1.6

1.3 1.4

1.2 1.2

1.1 1
0 1 2 3 4 5 6 7 8 9 10 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Z \alpha

FIG. 3: Plot of energy estimate as a function of the parameters. The first figure gives the variation of E(Z) from eq. (14)
against Z. The second figure gives the variation of E(α) with α for g = 2.5.

Variable α Variable Z
g ν(Z = g) α1 ν(α1 ) Zmin ν(Zmin ) ν(exact)
0.5 0.241000 1.218475 0.259748 0.348137 0.235783 0.233519
1.0 0.404884 1.129297 0.398822 0.744061 0.397374 0.392743
1.5 0.516372 1.044616 0.508068 1.171873 0.509454 0.503881
2.0 0.595116 1.020001 0.590666 1.620326 0.589487 0.583894
2.5 0.652967 1.010763 0.650547 2.082283 0.648552 0.643356
3.0 0.696958 1.006491 0.695555 2.553323 0.693517 0.688831
5.0 0.800388 1.001558 0.800126 4.485386 0.798998 0.796119

TABLE II: Summary of results

Found. Phys. 28 (1998) 549 [15] I. Richard Lapitus, Am. J. Phys. 38 (1970) 905
[11] S. Albeverio ,S. Fasari, F. Rinaldi, J. Phys. A. : Math. [16] I. Richard Lapitus, Am. J. Phys. 50 (1982) 453
Theor. 49 (2016) 025302 [17] M. K. Srivastava, R. K. Bhaduri, Am. J. Phys. 45 (1977)
[12] S. Albevero et al, J. Phys. A : Math. Theor. 46 (2013) 462
385305 [18] M. K. Srivastava, R. K. Bhaduri, Physical Review A 14
[13] V. V. Papoyan, V. A. Zagrebnov, Phys. Lett. A 113 (1961) 1976
(1985) 8-10 [19] C. C. Tannoudji, Rev. Mod. Phys. 70 (1998) 707
[14] Gerald V. Dunne, Christopher S. Gauthier, Phys. Rev. [20] Boris S. Mityagin, Int J Theor Phys 54 (2015) 4068–4085
A 69 (2004) 053409
 7

1.508
first excited
1.507

1.506

1.505
E(alpha)

1.504

1.503

1.502

1.501

1.5

1.499
0.94 0.95 0.96 0.97 0.98 0.99 1 1.01 1.02 1.03 1.04
alpha

2
Z=g
1 Variable Z
0 Variable alpha
Variable Z,alpha
relative percentage error

-1
-2
-3
-4
-5
-6
-7
-8
-9
-5 -4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5
g

3
FIG. 4: Plot of E from (4) against α. We indeed get a minima at E = 2
which is the expected energyvalue of original
unperturbated HO.