Volume 145B, number 3,4 PHYSICS LETTERS 20 September 1984

RELATIONSHIP BETWEEN SUPERSYMMETRY

AND THE INVERSE METHOD IN QUANTUM MECHANICS

Michael Martin NIETO

Theoretical Division, Los Alamos National Laboratory, University of California,
Los Alamos, NM87545, USA

Received 18 June 1984

In supersymmetric QM the bos9nic hamiltonian H+ = AtA yields the fermionic hamiltonian H_ = AAt. However, the
most general B h that satisfies BRB~ =H_ yields an HB+(h ) = BhBx t which in general is not H+. This new hamiltonian can be
understood as a special case of the application of the inverse method to H÷ to obtain new hamiltonians, one of which is
HB÷(h). When h = 0 the new hamiltonian has the original bosonic spectrum but with the ground state removed.

The description of supersymmetric quantum me- supersymmetric partners. This can be seen by going to
chanics [1 ] can start with the standard Schr6dinger a two component wave function and writing [3]
form

H±~P = i~t~ = [ - ~ x2 + V±(x)] q/, (1) ° °01' 70 j"

(5)

with V±(x) given by The supersymmetric hamiltonian is then

V±(x)=(~u)
1 r 2 -Iu"
+~ . (2)
[;/+ o ]
Hs s = Qt Q + QQt = H " (6)
(The prime means d/dx.) Eq. (2), often motivated by
the Fokker-Planck equation [2] * t, automatically The charges Q and Qt have the supersymmetric prop-
guarantees that the ground state of H+ has zero ener- erties 0 2 = ( a t ) 2 = 0 and [ a , H s s ] = [ a t ,Hss ] = 0.
gy, since a solution to eq. (1) using H+ is The two hamiltonians have the same spectra, except
for the ground state eigenvalue. Only the boson
,I,0 = N o e x p [ - - ~ U ] , E0=0. (3) hamiltonian (H+) has a normalizable ground state with
The hamiltonian H+ (and the associated H _ ) can eigenvalue E 0 = 0.
be written as The decomposition of H± into A and A t actually is
a type of factorization that goes back to SchriSdinger
H+ = A t A , H_ = A A t , [4]. However, as Mielnik [5] observed for the harmon-
ic oscillator, the solutions that can be generated from
A = o x + ~1U ~
, At=-~x+5 1Ut
, [ A , A t ] = U" . such a factorization are more general than usually
(4) realized. Such generalizations will allow us to show the
The hamiltonians H÷ and H _ are boson and fermion connection to the inverse method.
Consider a hamiltonian

#t A clear derivation of the Fokker-Planck equation from the HB_ = B B t =H_ , B = ~x + f ( x ) . (7)
more general master equation is given by Kittel [2]. Given
Combining eqs. (4) and (7) yields
the FP equation, one makes a simple transformation of vari-
ables to get the imaginary time Schr~Jdinger equation. See f2 +#_- (_Iv,)2 +-~v". (s)
the footnote on p. 788 of Tomita et al. [2].

208 0370-2693/84/$03.00 © Elsevier Science Publishers B.V.

(North-Holland Physics Publishing Division)
Volume 145B, number 3,4 PHYSICS LETTERS
Eq. (8) is a Ricatti equation [6], with the obvious so-
lution f = ( U ' / 2 ) ' 2 . However, the general solution is
f = ( U ' / 2 ) - ~b, where ~ is to be determined. Writing
as l / y , the differential equation ( 8 ) b e c o m e s
y ' = U~v - 1 , (9)
whose solution is
1/y = ¢ = exp [ - U ( x ) ]
× (X + f dz exp[-U(z)]) -1 , (10)
x ~2i(x, y ) = - D ~ /(x )O,,/(y ) ,
where X is a constant. Since ¢ has the property
~b' = ~(¢ - U ' ) , (11)
one has
HB+(X) =B~B x -_ - a x2 + V+ + 2~' : / : H + . (12)
If liB+ 4:H+, what is it? The answer can be found
b y using the G e l ' f a n d - L e v i t a n inverse method [7], as
was done b y Abraham and Moses [8]. This program
[8] does the following.
Given an unperturbed hamiltonian H 0 = - ~ x2 + V0,
with discrete eigenvalues-eigenvectors (En, X~n) and
continuous eigenvalues-eigenvectors (Ek, tI,k), one can
generate a new perturbed hamiltonian H 1 = H 0 + V 1
with new eigenvectors Xn and Xk for the same eigen-
values En and Ek, except that:
(i) the normalizations o f a finite number o f the dis-
crete X], having the same El, are changed; or
(ii) a finite number o f the discrete Ej are subtracted
from the spectrum; or
(iii) a finite number o f discrete Ej are added to the
spectrum; or
(iv) combinations of (i), (ii), and (iii) above are
done.
Note that the "finite number" can be done repeat-
edly, one step at a time. Also, the combinations o f (iv)
can be done one step at a time. Further, (iii) is the op-
posite operation to (ii) and the continuous spectrum
is irrelevant to what we are discussing. This means we
can concentrate on (i) and (ii) with a single discrete
.2 The definition of A from H÷ in terms of U amounts to tak-
ing the particular solution of a Ricatti equation instead of
a general solution as we discuss for B. See p. 80 of ref. [6].
A general solution forA would not yield the standard H_.
20 September 1984
eigenvector affected. (As an aside, observe that in (i)
H 0 and H 1 are not unitarily equivalent. This is remi-
niscent of phase equivalent potentials in inverse scat-
tering theory [9] .)
I refer to reader to ref. [8] for a derivation o f the
inverse method procedure. For our purposes it is
enough to state the algorithm.
ALGORITHM. Consider an H 0 with an orthonor-
mal complete set (En, ~n)" Let ] be the discrete state
affected and define (all tI,'s real)
(13)
where D is a real constant. Further, take Kj(x,y) as
the solution of the integral equation
X
K/(x,y)=-a/(x,y)-f Ki(x,z)f2j(z,y)dz. (14)
Then there exists an H 1 = H 0 + V1, where V 1 is
VI(X ) = 2d K/(x, x)/dx, (15)
with an ortho-complete set o f eigenvectors
X
Xn(X) = ~n(X) + f K/(x, Y)~n(Y) d y , (16)
-- oo
and associated eigenvalues E n which are the same as
the original eigenvalues except that: (a) if D = 1, the
Xn are orthonormal but there is no E i or X/'; and (b) if
D ¢ 1, all the eigenvalues obtain, but IIx/'lJ2 = 1/
(1 - D ) ¢ 1 , a
Now we can make the connection to supersymme-
try. One can verify that the solution for Kl(x,y) is
given b y
-I
X
A = (1 - D)/D. (17)
,a This unusual normalization condition was also used [8] in
properly resolving the identity. Also, since asD ~ 1 the
] th eigenvalue is removed, one recalls the known phenom-
ena where the spectrum of a hamiltonian with a singular
perturbation going to zero is not the same spectrum as
that of the unperturbed hamiltonian [10]. Indeed, as one
can explicitly verify for the analytical examples of ref.
[ 11 ], V1(A ¢ 0) can be singular, but V1(A = 0) is not sin-
gular.
209
Volume 145B, number 3,4 PHYSICS LETTERS
Table 1.
A flow chart showing the connection between supersymmetry
and the inverse method in quantum mechanics. The SUSY
bosonic hamiltonianH+ is equal to the inverse method unper-
turbed hamiltonian,Ho. H+ yields the fermionic hamiltonian
H_ by the standard Schr~Sdingerfactorization method. Setting
H_ identically equal to a new HB_ yields, upon solving the
Ricatti equation, Bh. However, after refactoring HB- into
bosonic form one obtains a new hamiltonianHB+(X) :~ H+.
Simultaneously, if the inverse method is applied to H0, one
finds a new HI(I, A), with one original (/') eigenvector af-
fected. If one takes the eigenvector ] as being the ground state
and sets A = N~S, one finds HB+(h ) = HI(0, N~X). Setting h
= 0 gives a new normal hamiltonian with the ground state
eigenvalue of H+ removed.
Supersymmetry Inverse method
H+ = A t A = Ho
H_ = A A t HI(/", A)
HB_= BBt =-H_ HI(O , A)
H+ ~ HB+(X) =BtB T (h = 0)
+ H 1(0, hNg)
HB+ = H 1
ground state
eigenvalue of H+
removed
But if we now specialize to/' being the ground state
(and take A = N02X), then V1 = 2~'. Thus, H B + @ )
= H I ( ] = 0, A =N~X) and the connection is made to
supersymmetry.
Examples of this inverse procedure are given in refs.
[8] and [11]. Also, for X = 0, graphs are given in ref.
[ 11 ] showing the forms of V0, V 1 , V0 + V 1 and the
spectra, for V0 = x 2 and V0 = (x - l / x ) 2.
To summarize, the connection between supersym-
metry and the inverse method in q u a n t u m mechanics
is shown in the flow chart of table 1.
210
20 September 1984
Among the many colleagues who have commented
on this work I wish to specially thank David Campbell,
Terry Goldman, Alan Kosteleck~, and David Sharp.
This work was supported by the US Department of
Energy.
References
[1] E. Witten, Nucl. Phys. B185 (1981) 513;
F. Cooper and B. Freedman, Ann. Phys. 146 (1983) 262;
C.M. Bender, F. Cooper and B. Freedman, Nucl. Phys.
B219 (1983) 61.
[2] C. Kittel, Elementary statistical physics (Wiley, New
York, 1958) pp. 157-158;
H. Tomita, A. It5 and H. Kidachi, Prog. Theor. Phys. 56
(1976) 786.
[3] E. Gozzi, Phys. Lett. 129B (1983) 432;
R. Akhoury and A. Comtet, Nucl. Phys. B., to be pub-
lished;
M. Bemstein and L.S. Brown, Phys. Rev. Lett. 52 (1984)
1933.
[4] E. SchrOdinger,Proc. R. It. Acad. A 46 (1940) 9.
[5] B. Mielnik, J. Math. Phys., to be published.
[6] N.G. van Kampen, J. Stat. Phys. 17 (1977) 71.
[7] I.M. Gel'fand and B.M. Levitan, Izv. Akad. Nauk SSSR,
Set. Mat. 15 (1951) 109 [Am. Math. Soc. Transl. 1
(1955) 2531.
[8] P.B. Abraham and H.E. Moses, Phys. Rev. A22 (1980)
1333.
[9] V. Bargmann, Rev. Mod. Phys. 21 (1949) 488.
[10] J.R. Klauder, Phys. Lett. 47B (1973) 523; Acta Phys.
Austriaca Suppl. 11 (1973) 341;
B. DeFacio and C.L. Hammer, J. Math. Phys. 15 (1974)
1071.
[11] M.M. Nieto and V.P. Gutschick, Phys. Rev. D23 (1981)
922;
M.M. Nieto, Phys. Rev. D24 (1981) 1030.