discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/283091216
CITATIONS READS
3 22
1 author:
Andrey M. Pupasov
Federal University of Juiz de Fora
30 PUBLICATIONS 166 CITATIONS
SEE PROFILE
All content following this page was uploaded by Andrey M. Pupasov on 23 October 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Annals of Physics 363 (2015) 122135
Annals of Physics
journal homepage: www.elsevier.com/locate/aop
1. Introduction
http://dx.doi.org/10.1016/j.aop.2015.09.021
0003-4916/ 2015 Elsevier Inc. All rights reserved.
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 123
levels = { [[1]], [[2]], . . . , [[1]]}, [[i]] N which are deleted from the spectrum by Dar-
bouxCrum transformations [2].
The simplest rational extension is given by the potential [6]
x2 (x2 1)
V {1,2} [x] = +2 1+2 2 , (1)
4 (x + 1)2
1
which leads to the quasi-equidistant spectrum for the Hamiltonian [3], En = n + 2
, where n
N0 \ {1, 2}.
Another example is the two-well perturbation of the oscillator
x2 x4 9
{2,3} 2
V [x] = + 2 1 + 4x (2)
4 (x4 + 3)2
with the quasi-equidistant spectrum, En = n + 12 , n N0 \ {2, 3}.
Darboux transformations represent a powerful tool to manipulate physical properties of one-
dimensional quantum systems, [6] and to construct (polynomial) supersymmetric extensions of
quantum mechanics (SUSY QM) [7,8]. Exactly solvable models obtained by Darboux transformations
are widely applied in nuclear physics, condensed matter physics, quantum optics, etc. [914].
The possibility to calculate propagators using ideas of supersymmetric quantum mechanics and the
theory of solitons was considered in [15]. Quasiclassical approach to propagators and path integration
in SUSY QM was developed in [16]. In the case of shape-invariant potentials, SUSY relations between
propagators allow to calculate them explicitly [17,18]. We proposed a more general approach to
calculations of propagators in SUSY QM without restricting ourselves by shape-invariant potentials
in [1922]. This approach can be extended and applied to the analysis of quantum tunnelling in multi-
well potentials [23,24]. In the case of generalized Schrdinger equation the SUSY propagators were
calculated in [25]. Green functions of the Dirac equation were studied by means of SUSY QM in [26].
Here we re-examine results of [22], where propagators K {k,k+1} for the V {k,k+1} family of potentials
were defined by means of a generating function S (x, y; t |J ) which contains the error-function. We will
extend this result to arbitrary sequences . Moreover, we will show that propagators K are expressed
by elementary functions only.
In the case of potentials (1) and (2) we will obtain the following propagators
{1,2} 4i sin t xy e i t
K ( x, y ; t ) = e 2i t
Kosc (x, y; t ) 1 , (3)
(1 + x2 )(1 + y2 )
where symmetric polynomials Qk (x, y) = Qk (y, x) can be determined iteratively. This iterative
method proofs to be more efficient than the method based on the generating function [22]. The
polynomials Qk (x, y) allow also to calculate Green functions G (x, y; E ) and establish an analogue
of Mehlers formula for the exceptional Hermite polynomials [27].
The obtained results and possible applications are discussed in Section 4.
x2
12
n (x) = pn Hen (x)e 4 , pn = n! 2 . (8)
Rational extensions are defined by Darboux transformations of the Harmonic oscillator [27]
x2
H = x2 + 2x2 (ln Wr[ (x), x]), (9)
4
where ,
= {k1 , k1 + 1, . . . , kM , kM + 1}, | | = 2M , (10)
1
is a strictly increasing sequence of natural numbers, or a KreinAdler sequence.
Following standard notations of the Mathematica program language we denote by [[i]] and by
[[1]] ith element and the last element of this sequence, respectively. The following agreement is
in order. Consider a sequence of elements (An )n>0 . If a sequence of natural numbers appears as a
subscript, A{n1 ,...,nm } , this means a set of elements A{n1 ,...,nm } = {An1 , . . . , Anm }. For instance, (x),
implies a set of wave-functions,
(x) = {n1 ,n2 ,...,n2M } (x) = {n1 (x), n2 (x), . . . , n2M (x)}.
This agreement allows us to write Wronskians in a compact form
[[1]] (x) [[2]] (x) ... [[1]] (x)
[[1]] (x) [[2]] (x) ... [[1]] (x)
Wr[ (x), x] = ... ... ... ... . (11)
(| |1) (| |1) (| |1)
[[1]] (x) [[2]] (x) ... [[1]] (x)
Hamiltonians Hosc and H can be embedded into a polynomial SUSY algebra [28,29]
2M 2M
LHosc = H L, (H [[j]]),
L+ L = (Hosc [[j]]), LL+ = (12)
j =1 j =1
1 We assume that V is defined for all x R, which restricts possible choice of , in particular, | | = 2M. See for the
details [2].
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 125
Operator L maps the oscillator eigen-functions (8) to the eigen-functions of the rationally extended
oscillator
Using the explicit form of oscillator eigen-functions (8) and the following identities
2M
Mx2
Wr[ (x), x] = e 2 Wr[He (x), x] p [[n]] ,
n=1
2M
Mx2
ln Wr[ (x), x] = + ln p [[n]] + ln Wr[He (x), x]
2 n=1
we can express rationally extended Harmonic oscillators through the Wronskian of probabilistic
Hermite polynomials only
x2
H = x2 + 2x2 (ln Wr[He (x), x]) + 2M . (16)
4
Note (see for instance [2]) that Wr[He (x), x] (for the chosen class of ) is the polynomial of x2 and
2M
degx Wr[He (x), x] = ( [[n]] n + 1).
n=1
Wn (x) = Wr[ \{ [[n]]} (x), x], n (x) = Wr[He \{ [[n]]} (x), x],
W (20)
Wr[He (x) {f }, x]
=
Lf . (21)
Wr[He (x), x]
Note that
2M
Mx2
W (x) = e 2 (x)
W p [[n]] , (22)
n =1
2M
(2M 1)x2
Wn (x) = e 4 n (x)p
W 1
[[n]] p [[j]] , (23)
j =1
x2 x2
Le 4 f = e 4 .
Lf (24)
126 A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135
If two Hamiltonians H0 and HN are related by Nth order Darboux transformation which remove N
levels from the spectrum of H0 , then corresponding propagators K0 and KN are related as follows [22],
Wn (y) b
N
KN (x, y; t ) = Lx (1)n K0 (x, z ; t )un (z )dz . (26)
n =1
W (y) y
2M
y2 n (y)
W
z2
K (x, y; t ) = (1) e n 4 Lx Kosc (x, z ; t )e 4 He [[n]] (z )dz . (27)
n=1
(y)
W y
z2
The occurring integrals y Kosc (x, z , t )e 4 Hen (z )dz can be represented as derivatives of the
generating function with respect to the auxiliary current J
z2
Kosc (x, z , t )e He [[n]] (z )dz = He [[n]] J S (J ) J =0
4
y
[[n]]
k S (J )
= h [[n]],k ,
k=0
Jk
J =0
where hm,k are coefficients of the Hem . The generating function reads
it
i t x2
1 i2t xe 2
S ( J | x, y , t ) = , E [J , x, y , t ]
2 4
e R iJ 2i sin te (28)
2 i 2i sin t
where
it
i2t xe 2
R iJ 2i sin te , = exp (J (i J sin t + x) exp(it )) ,
i 2i sin t
i t i i it
i2t
E [J , x, y, t ] = 1 + erf J i sin te 2 ye 2 xe ,
2 sin t
z
2
erf (z ) = exp(t 2 )dt .
0
Function R is the generating function of rescaled Hermite polynomials (see also Appendix A)
z 2
zn
x
R z , = exp xz = n (x)
He[] .
2 n=0
n!
it 1 1
= 2i sin te i t = 1 e 2i t , e2 i sin t = 1 e 2i t 2 . (29)
2
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 127
1. J-derivatives of R-function
i t x i (km)t
J R J e ,
km
= He[]
km (x) e
J =0
(k m)! i (km)t dm []
= e He (x) . (30)
k! dxm k
2. J-derivatives of E-function
m+1 E [J ] m
x2 y2 i t
= Kosc (x, y; t )e 4 e 2 2i sin t qj,m+1 (x, y)e i tj . (31)
J m+1 J =0 j =0
3. x-derivatives of E0 -function
k+1 E0 [x, y, t ]
k
k
x2 y2 i t 1
= Kosc (x, y; t )e 4 e 2 wj,m+1 (x, y)e i tj . (32)
x k + 1 2i sin t j =0
k m+1 E [J ]
k1 m+k
1 i t x2 y2
= e 2 Kosc (x, y; t )e 4 qj,m+k+1 (x, y)e i tj . (33)
xk J m+1 J =0 2i sin t j =0
In the above expressions qi,j and wk,m are some polynomials of x and y with constant coefficients.
Now we will substitute these derivatives to calculate
He [[j]] J S (J ) J =0
i t x2
[[j]] k
e 2 4 x
Ckm km
J e i t ,
m
[J , x, y, t ] .
= h [[j]],k R J E (34)
2 k=0 m=0
J
J =0
We will change the order of summation (k, m m, k), that is we will fix m and calculate first sum
by k ( [[j]] m, [[j]]).
[[j]]
k
1 i (km)t dm []
(x) Jm E [J , x, y, t ]
h [[j]],k e Hek
k=0 m=0
m! dxm
J =0
[[j]] [[j]] m
1 i (km)t d
m
[J , x, y, t ] []
Hek (x) .
= J E h [[j]],k e
m=0
m! k=m
dxm
J =0
128 A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135
dm []
In the last expression we can change the inferior limit k = m to k = 0 since dxm
Hek (x) = 0 when
k < m,
dm
[[j]] [[j]]
i [[j]]t e i mt [[j]]k
e m
J E [J , x, y, t ] m
h [[j]],k (e 2i t ) 2 Hek (x)
[]
.
m=0
m! dx k=0 J =0
[[j]]k
Afterwards we note that h [[j]],k (e 2i t ) = h[1[[]
2
j]],k (see Appendix A, (55)),
dm
[[j]] [[j]]
i [[j]]t e i mt
J E [J , x, y, t ] [1] []
h [[j]],k Hek (x) .
m
e m
m=0
m! dx k=0 J =0
The sum by k represents the umbral composition (57) for the generalized Hermite polynomials
Hek (x) [30], which yields
[]
[[j]]
e i mt d m
e i [[j]]t Jm E [J , x, y, t ] He [[j]] (x) .
m! m dx
m=0 J =0
We split the last expression in two terms K = KE + KR , where the first term contains error function
in E0 (x, y; t ) whereas the second term contains elementary functions only
i t
e 2
2M
y2 j (y)
W x2
e i [[j]]t E0 [x, y, t ]He [[j]] (x) ,
KE = (1)j e 4 Lx e 4
2 j =1
(y)
W
[[j]] i mt m E [J ]
i t
e 2
2M
j (y)
y2 W x2 i [[j]]t e J J =0 d He [[j]] (x)
m
KR = (1) e 4
j
Lx e 4 e .
2 j =1 (y)
W m=1
m! dxm
e
i t
2
2M
j (y)
y2 x2 W
[[j]] i ( [[j]]m)t
e dm He [[j]] (x)
(1)j e 4 J E [ J ] J =0 .
m
KR = L x
2 j =1 (y) m=1
W m! dxm
Let us consider the first term KE . Though the function E0 contains non-elementary error function erf
it can be seen that KE (x, y; t ) is expressed in terms of elementary functions only. The error function
will be cancelled due to the operator L as follows
i t
e
2M
2 y2 x2 j (y)
W
(1)j e 4 e i [[j]]t
KE = L x E0 [x, y, t ]He [[j]] (x)
2 j =1 (y)
W
i t
e
2M
2 y2 x2 W j (y)
(1)j e 4 e i [[j]]t
=
2 j =1 W (y)
2M
An,j (x)x E0 (x, y; t )
n
n =1
E0 [x, y, t ]L x He [[j]] (x) + ,
W ( x)
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 129
where An,j (x) are some polynomials in x. By the definition (21) of the operator L we have
L x He [[j]] (x) = 0, therefore
i t y2 x2
2M 2M 1
e 2 e 4
(1)j e i [[j]]t W
KE = j (y) An,j (x)xn+1 E0 (x, y; t )
(y)W
2W (x) j=1 n =0
i t y2 x2
e 2 e 4 x2 y2 i t
= Kosc (x, y; t )e 4 e 2
(y)W
2W (x)
An,j (x)
2M 2M 1 n
(1)j e i [[j]]t W
j (y) wl,n+1 (x, y)e i tl
j =1 n=0
(2i sin t ) n
l =0
Thus we proved that KE contains elementary functions only. Denote e i t by . We can further simplify
KE writing it as a product of Kosc and a rational function of x, y and ,
Kosc (x, y; t ) An,j (x)n
2M 2M 1 n
KE = (1)j [[j]] W
j (y) n wl,n+1 (x, y)l
(y)W
(x) 1 2
W j =1 n=0 l=0
4M + [[1]]1
Q n (x, y)n
Kosc (x, y; t ) n =1
= ,
(y)W
W (x) (1 2 )2M 1
where Q n (x, y) are some polynomials.
Following the same line we will consider the second term. It can be seen that structures of KE and
KR coincide, and KR reads
4M + [[1]]1
Q n (x, y)n
Kosc (x, y; t ) n=0
KR = .
(y)W
W ( x) (1 ) 2 2M 1
where some constant multiplier Wr[h (x), x] = C W (x) is absorbed by redefinition of polynomials Q .
Consider the limit t 0, where K (x y) and Kosc (x y). From here it follows that
the rational -depending factor has no pole, therefore
[[1]]+
1+4M 2 2M 1 [[
1]]+1
Q n (x, y)n 1 2 Qn (x, y)n
[[ 1]]+1
Qn (x, y)n
n n=0
2M 1 = =
1 2 1 2 2M 1
n =0
and
[[ 1]]+1
Qk (x, y) = Wr[h (x), x]Wr[h (y), y].
(35)
k=0
130 A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135
Substituting into the rational ansatz (36) expansions of propagators in terms of eigen-functions
x2 +y2 1
Kosc (x, y; t ) = e 4 2 hn (x)hn (y)n , = e i t , (37)
n =0
x2 +y2
e 4 1
hn (x)hn (y)n
K (x, y; t ) = 2 (38)
Wr[h (x), x]Wr[h (y), y] nN\
we obtain a system of equations for the polynomial coefficients Qk (x, y). The solution of this system
is given by a finite recursive procedure
k
1
Qk (x, y) = hk (x)hk (y) Qkj (x, y)hj (x)hj (y) ,
0 k [[1]] + 1. (39)
h0 (x)h0 (y) j =1
[[ 1]]+1
hmk (x)hmk (y)Qk (x, y) = hm (x)hm (y),
(40)
k=0
= m (x)m (y).
k=0
(41)
[[ 1]]+1
Qk (x, y)
k=0
where deg h0 =
| |
j=1 ( [[j]] j + 1) | |.
In Appendix B we present an example of a non-KreinAdler sequence = {1} when the nonlinear
connection lemma also holds. We suppose that (40) holds for an arbitrary .
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 131
The transformation formula for the propagators implies also the following relation for the Green
functions
[[1]]+1
1
G (x, y; E ) = Qk (x, y)Gosc (x, y; E k).
(47)
W [h (x), x]W [h (y), y] k=0
3.3. Examples
x2 (x2 1)
V{1,2} [x] = +2 1+2 2 . (48)
4 (x + 1)2
Connection polynomials read
x2 y2 + x2 + y2 1
{1,2} 1 xy xy
Q{0,1,2,3} = , , , .
2 2 4 2
The propagator for the Schrdinger equation with Hamiltonian K (1,2) = x2 + V {1,2} (x) has the
following compact expression
{1,2} 4i sin t xy e i t
K (x, y; t ) = e 2i t
Kosc (x, y; t ) 1 . (49)
(1 + x2 )(1 + y2 )
We can obtain the next simple expression for the propagator using second and third excited states
of Harmonic oscillator
2 /4 2 /4
2 (x) = p2 (x2 1)e x 3 (x) = p3 x(x2 3)e x .
In this case we obtain two-well perturbed Harmonic oscillator potential (potentials V{k,k+1} have k
shallow minima at their bottom)
x2 x4 9
V{2,3} [x] = + 2 1 + 4x2 4 . (50)
4 (x + 3)2
Connection polynomials read
(x2 + 1)(y2 + 1) xy(3 x2 y2 ) x4 y4 + 3x4 + 3y4 12x2 y2 3
{2,3}
Q{0,1,2} = , , ,
4 6 24
xy(x2 y2 3) (x2 1)(y2 1)
{2,3}
Q{3,4} = , .
6 4
The propagator for the Schrdinger equation with Hamiltonian H {2,3} = x2 + V {2,3} (x) reads
4. Conclusions
Propagators
[[ 1]]+1
Qk (x, y)e i kt
present a new example of Feynman path integrals that can be calculated analytically [32]. The key
formula (39)
k
1
Qkj (x, y)hj (x)hj (y)
Qk (x, y) = hk (x)hk (y) , 0 k [[1]] + 1,
h0 (x)h0 (y) j=1
A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135 133
which defines a non-linear connection between x-Hermite and Hermite polynomials can be easily
realized in any computer algebra system. Since the definition of polynomials Qk (x, y) requires
calculations of first [[1]] + 1 exceptional Hermite polynomials, new recurrence relations [33] can
be useful.
One can use these propagators to test various approximations for the corresponding path integrals.
A typical rationally extended Harmonic oscillator has a multi-well profile, that is may contain a
degenerate classical vacuum (global minimum of the potential energy) and various pseudo-vacua.
In this case it is interesting to compare the exact answer with a multi-instanton approximation [34].
Another possible application of obtained propagators is related with studies of quantum dynamics
in multi-well potentials [3539,24], quantum transients [40]. In particular, a non-linear representa-
tion of the potential (46) involves squares of first [[1]] + 1 wave functions which indicates a pos-
sibility to relate rational extension of Harmonic oscillator with a non-linear Schrdinger equation.
Appendix A. Rescaled Hermite polynomials, Appell sequences and umbral composition [30]
Then the sequence Cn (x) = (An B)(x) also is an Appell sequence. Its generating function, denoted by
U [x, g ] = S [x, g ] R[x, g ], has the following form
gn
U [x, g ] = s(g )r (g )e xg = Cn (x) .
n =0
n!
134 A.M. Pupasov-Maksimov / Annals of Physics 363 (2015) 122135
n (x)
We can apply these facts to the sequences of generalized Hermite polynomials He[]
z 2 z2 ( + )z 2
exp xz exp xz = exp xz .
2 2 2
As a result we get the umbral composition of rescaled Hermite polynomials
n
[]
He[] []
(x) = hn,k Hek (x) = He[+]
[]
(x).
n He n (57)
k=0
Here we consider an example of the polynomial connection when = {1}, which is a non-
KreinAdler sequence [2], and hence, when polynomials h1n do not represent a sequence of orthogonal
polynomials. Nevertheless, by direct calculations we can verify that the Non-linear connection lemma
is valid in this case. Consider formally normalized product of two wronskians
x hm (x) y hm (y)
1
hm (x)hm (y) =
{1} {1}
2 (1 m) 1 hm (x) 1 hm (y)
1
x mhm1 (x) hm (x) y mhm1 (y) hm (y) .
= (58)
2 (1 m)
Using the recurrence relation for the normalized probabilistic Hermite polynomials
x mhm1 (x) = mhm (x) + m(m 1)hm2 (x)
we get
1
h{m1} (x)h{m1} (y) = (hm (x)hm (y) xyhm1 (x)hm1 (y) hm2 (x)hm2 (y)) .
2
{1} {1}
That is, polynomials hm (x)hm (y) satisfy to the following 3-term representation
2
{1}
h{m1} (x)h{m1} (y) = hmk (x)hmk (y)Qk (x, y),
k=0
where
{1} 1
Q0 (x, y) = ,
2
{1} xy
Q1 (x, y) = ,
2
{1} 1
Q2 (x, y) = .
2
We also verified by computer algebra that the nonlinear connection Lemma is valid for an arbitrary
with [[1]] < 5.
References