a

r
X
i
v
:
g
r
-
q
c
/
9
9
0
2
0
2
8
v
2


1
1

F
e
b

1
9
9
9
On the spectrum of relativistic Schroedinger
equation in nite dierences.
V.A.Berezin, A.M.Boyarsky
Institute for Nuclear Research
of the Russian Academy of Sciences,
60th October Anniversary Prospect, 7a,
117312, Moscow, Russia
e-mail: [email protected],
[email protected]
A.Yu.Neronov
Theoretische Physik,
Universitat M unchen,
D-80333, M unchen, Germany
e-mail: [email protected]
February 10, 2010
Abstract
We develop a method for constructing asymptotic solutions of -
nite - dierence equations and implement it to a relativistic Schroedinger
equation which describes motion of a selfgravitating spherically sym-
metric dust shell. Exact mass spectrum of black hole formed due to
the collapse of the shell is determined from the analysis of asymptotic
solutions of the equation.
1
1 Relativistic Schroedinger equation in nite
dierences
In this note we develop a method for determining the energy spectrum of
relativistic Schroedinger equation in nite dierences
(m, , S +i) + (m, , S i) =
(F
in
F
out
)
1/2
_
F
in
+F
out
M
2
/4m
2
S
_
(m, , S) (1)
which appear to describe a quantum black hole model with selfgravitating
dust shell introduced in [1]. Here parameter M is the bare mass of the shell,
m = m
out
is the Schwarzchild mass of the space-time outside the collapsing
shell, = m
in
/m
out
is the ratio of Schwarzchild masses inside and outside
the shell, S = R
2
/4G
2
m
2
(G is gravitational constant) is the dimensionless
area of the shell surface and F
in,out
= 1 2Gm
in,out
/R are functions entering
the Schwarzchild line element.
Parameter = m
pl
/ (2m
2
) where m
pl
=
_
c h/ G is Planck mass ( h is
Planck constant and c is the speed of light). It becomes small in semiclassical
( h 0) limit. In this limit the nite shift in the argument of the wave
function in (1) become small and one could us an approximate expression
(S i) (S) i

(S)

2
2

(S) (2)
Then the equation 1 becomes just a usual Schroedinger of nonrelativistic
quantum mechanics:

+
_
2 (F
in
F
out
)
1/2
_
F
in
+F
out
M
2
/4m
2
S
__
= 0 (3)
This equation was studied in [1, 2] in details. There was an expression for
the energy spectrum of bound states found and quasiclassical behaviour of
the wave function studied.
In this paper we will analyse the solutions of the exact equation (1).
The shift of the argument in the wave function is along the imaginary
line, so the equation is naturally dened over some complex one dimensional
manifold. The function
(F
in
F
out
)
1/2
=

_
_
1
1

__
1

_
=
_
( 1)( )

(4)
2
( =

S) which enter the equation (1) is a branching function of its argu-

ment. So (1) is dened over the Riemannian surface S
F
for the function (4)
which is a two dimensional sphere glued from the two Riemannian spheres
S
+
and S

along the sides of the cut made along the interval (, 1) of the
real line.
There are two points = on the Riemannian surface S
F
in S
+
and S

components, which are singular points of equation (1). They dier by the
dierent chooses of the sign of the function (F
in
F
out
)
1/2
.
The natural requirements on the wave function of a bound state is that
it must decrease with But in our case we have two innities in S
+
and S

components of S
F
. As it was argued in [1, 2, 3] the two possibilities
for have the following physical meaning. If one consider maximally
extended spherically symmetric solution of Einstein equations in vacuum
Kruskal manifold [4], one nds that the radial coordinate take the same
values on dierent sides of Einstein-Rosen bridge (in R+ and R

regions
of Kruskal manifold [1]). So could tend to innity twofold: either in R
+
region or in R

region. This complicated structure of Kruskal manifold is

mirrored in complicated structure of conguration space of quantum system
(the real section of S
F
) which describe a selfgravitating dust shell. So the
two possibilities for in S
+
and S

components of S
F
corresponds to
the possibility for in R
+
and R

regions of Kruskal space-time.

But the requirements for the wave function to vanish at innities are not
all the requirements which a reasonable wave function of a physical state
must satisfy. Besides if must be one-valued function.
This requirement is satised automatically if we consider Schroedinger
equation over a complex plane (or over the real line). But the equation (1) is
dened over (the real section of) the Riemannian surface S
F
. The two inni-
ties in S

components are both singular points of equation. So the equation

(1) is dened in fact over complex manifold which is two-dimensional sphere
minus two points a cylinder. But a cylinder possess a nontrivial cycle. The
monodromy of solution along this nontrivial cycle is not identity in general.
Therefore, the requirement for the wave function to be one-valued function on
the Riemannian surface S
F
(and on its real section) become a nontrivial one.
In what follows we will nd the asymptotics of exact solutions of equation (1)
at both innities. The analysis of analytical properties of asymptotics and
the requirement for one-valued solutions to decrease at innities will lead us
to dening of the mass spectrum for the equation (1)
3
2 Asymptotic solutions
Let us nd the asymptotic solutions of this equation at innities in S
components of S
F
.
It is convenient to rewrite (1) in matrix form.
If we truncate the Tailor expansion
(S i) = +
2N

k=1
(i)
k
k!

(k)
(5)
on 2N-th item and expand asymptotically
_
1

S
_
1/2
_
1
1

S
_
1/2

1 +
1 +
2
1

S
+
3
2
+ 2 + 3
8
1
S
+
5
3
+ 3
2
+ 3 + 5
16
1
S
3/2
(6)
in powers of S, we could rewrite (1) as

2N
=
(2N)!
(i)
2N
N1

k=1
(i)
2k
(2k)!

(2k)
+
(2N)!
(i)
2N
_
1 1
(1 )
2
M
2
/ m
2
8S

(1 +) (2(1 )
2
M
2
/ m
2
)
16s
3/2
_
(7)
where the upper signs correspond to S
+
and the lower ones to S

components.
It is convenient to introduce a vector function
Y
1
= ; Y
2
=

; ...; Y
2N
=
2N1
(8)
Then (7) takes the form
Y

= AY (9)
where
A(S) = A
0
+
1

S
A
1
+
1
S
A
2
+
1
S
3/2
A
3
(10)
4
where A
0
has the following form:
A
0
=
_
_
_
_
_
_
_
_
0 1 0 0 ... 0 0
0 0 1 0 ... 0 0
. . . ... ... . .
0 0 0 ... ... 0 1
a
1
0 a
3
0 ... a
2N1
0
_
_
_
_
_
_
_
_
(11)
with coecients
a
1
=
(2N)!
(i)
2N
(1 1)
a
2k
= 0
a
2k+1
=
(2N)!
(2k!)(i)
2N2k
, k > 0 (12)
Matrices A
1,2,3
are
A
1
= 0 (13)
and
A
2
=
_
_
_
_
_
0 0 ... 0
. . ... .
0 0 ... 0
1 0 ... 0
_
_
_
_
_
; =
(2N)!
(i)
2N
((1 )
2
M
2
/ m
2
)
8
(14)
A
3
=
_
_
_
_
_
0 0 ... 0
. . ... .
0 0 ... 0
1 0 ... 0
_
_
_
_
_
; =
(2N)!
(i)
2N
(1 +) (2(1 )
2
M
2
/ m
2
)
16
(15)
After the change of variable S =
2
equation (9) takes the form
1

=

AY ;

A() =

A
0
+
1

A
2
+
1

A
3
(16)
with matrices

A
i
= 2A
i
(17)
If we let N then matrix equation (9) is equivalent to nite - dier-
ence equation (1) in asymptotic regions. There exist well-developed method
for nding asymptotic solutions ( ) of matrix equations of (9) type [5].
5
First we need to make a transformation

Y = TY

A = T

AT
1

Y (18)
such that matrix

A
0
becomes diagonal:

A
0
= diag(
1
, ...,
2N
) (19)
The eigen values
1
, ...,
2N
could be found by equating the characteristic
polynomial for matrix

A
0
to zero:
P
2N
=

A
0
I

= 0 (20)
As a result we get an equation
P
2N
=
(2N)!
2(i)
2N
_
_
N

k=0
1
(2k)!
_
i
2
_
2k
1
_
_
= 0 (21)
In the limit N this equation is equivalent to the following one
cos
_

2
_
= 1 (22)
So we see that all the eigen values become double degenerate in the limit
N as it is seen from Fig. 1. This means that the matrix

A
0
could not
be transformed to diagonal form in this limit, instead it could be transformed
to Jordanian form:
A
0
= T

A
0
T
1
= J
1
J
2
... J
N
; (23)
J
i
=
i
I +H;
I =
_
1 0
0 1
_
; H =
_
0 1
0 0
_
It is convenient to rewrite all the matrices in the form of blocks of dimension
2 2:
T =
_
_
_
T
11
... T
1N
. ... .
T
N1
... T
NN
_
_
_; T
IK
=
_
t
11
t
12
t
21
t
22
_
(24)
6
1

cos
(

2
)
2

i
2

i+1
Figure 1: Double degeneracy of eigen values in N limit
and in analogous way

A
0
=
_
_
_
A
11
... A
1N
. ... .
A
N1
... A
NN
_
_
_; (25)
A
II
=
_
0 2
0 0
_
; A
I,I+1
=
_
0 0
2 0
_
; I = 1, ..., N 1
A
NK
=
_
0 0
2a
2K1
0
_
; K = 1, ..., N 1; A
NN
=
_
0 2
2a
2N1
0
_
Then the matrix T could be found from the equation (23) which lead to the
following system of recurrent equations:
A
II
T
IJ
+A
I(I+1)
T
(I+1)J
=
I
T
IJ
+T
IJ
H (26)
Solving this system of equations gives T in the following form
T
IJ
=
_
_
_
_
_
_

J
2
_
2I2
t
(J)
11
_

J
2
_
2I2)
t
(J)
12
+ (I 1)
_

J
2
_
2I3
t
(J)
11
_

J
2
_
2I1
t
(J)
11
_

J
2
_
2I1
t
(J)
12
+
2I1
2
_

J
2
_
2(I1)
t
(J)
11
_
_
_
_
_
(27)
the coecients t
(J)
11
and t
(J)
12
are arbitrary and the form of asymptotic solution
of equation (9) do not depend on them. So we will take t
(J)
11
= t
(J)
12
= 1 for
simplicity.
7
Matrices

A
2,3
transform under (18) as follows:

A
2
=
_
_
_
L
11
... L
1N
. ... .
L
N1
... L
NN
_
_
_; (28)
L
IK
=
_

t
I
12

t
I
12

t
I
22

t
I
22
_
where

t
I
ij
are elements of the block (T
1
)
IN
of inverse matrix of transforma-
tion T
1
.
The transformed matrix equation (18) could be reduced to a set of 2 2
matrix equations with the help of one more transformation

Y = P()Z (29)
where
P =
_
_
_
_
_
0 P
12
... P
1N
P
21
0 ... P
2N
. . ... .
P
N1
P
N2
... 0
_
_
_
_
_
(30)
is a matrix of 22 blocks such that after the transformation (29) the system
of equations (9) is rewritten as
1

= BZ (31)
with matrix
B = P
1

AP
1

P
1
P

(32)
of block-diagonal form:
B() =
_
_
_
_
_
B
11
0 ... 0
0 B
22
... 0
. . ... .
0 0 ... B
NN
_
_
_
_
_
(33)
Expanding
P = P
0
+
1

P
1
+
1

2
P
2
+...;
B = B
0
+
1

B
1
+
1

2
B
2
+...; (34)
8
and solving equation (32) in each power of 1/ one obtains for B
ii
B
ii
=
i
I +H +

2
L
ii
+

3
L
ii
(35)
where and are dened in (14) and (15) and
L
ii
=
_
l
i
1
l
i
1
l
i
2
l
i
2
_
(36)
is 2 2 matrix (28).
The 2N 2N matrix system (9) is equivalent to the set of 2 2 matrix
systems of the form
1

= BZ (37)
where B is one of the B
ii
.
Now we will construct solutions of the system of ordinary dierential
equations (37). Let us take
Z = exp
_

2
2
___
1 0
0
1

_
+
_

2
+

3
__
0 0
l
1

1

l
2
__
U (38)
Then dierential equation for U will be
U

= DU (39)
where
D =
_
0 1
_
+

2
_
l
2
_
+ 1 +

_
(l
1
+l
2
)
_
+... (40)
The last system of equations could be transformed to diagonal form because
the determinant of matrix D
0
= D|
=
= 0. The eigen values of matrix
D
0
=
_
0 1
l
2
0
_
are
1,2
=

l
2
= . Therefore taking the transformation matrix
T =
_
1
2
1
2
1
2

1
2
_
9
one nds the form of transformed system
W = T U
E = T
1
DT
W

= EW (41)
with
E =
_
1 0
0 1
_
+
1

_

2
l
2
+
+1
2
(l
1
+l
2
)

2
l
2

+1
2
(l
1
+l
2
)

2
l
2

+1
2
(l
1
+l
2
)

2
l
2
+
+1
2
(l
1
+l
2
)
_
+
(l
1
+l
2
)

2
_
1
2

1
2

1
2
1
2
_
+... (42)
Matrices E
1,2
which stand by 1/ and 1/
2
correspondingly could be trans-
formed to diagonal form as well after one more substitution
W = PQ
Q

= FQ (43)
Expanding P and F in powers of 1/ one could nd the form of matrix F
solving matrix equation
F = P
1
EP P
1
P

(44)
The result is
F = F
0
+
1

F
1
+... (45)
where
F
0
=
_
0
0
_
(46)
F
1
=
_
0

2
l
2
+
+1
2
(l
1
+l
2
)

2
l
2
+
+1
2
(l
1
+l
2
) 0
_
(47)
Finally we are able to write down the asymptotic solution of system (43):
Q
1
= exp
_
+
_

2
l
2
+
+ 1
2
(l
1
+l
2
)
_
ln
_
Q
2
= exp
_
+
_

2
l
2
+
+ 1
2
(l
1
+l
2
)
_
ln
_
(48)
10
where
=

_
l
2
=
(2N)!
4(i)
2N
_
(1 )
2

M
2
m
2
_
=
(2N)!
8(i)
2N
(1 +)
_
(2(1 )
2

M
2
m
2
_
(49)
Using this expression we could restore the asymptotic solution of initial ma-
trix system of dierential equations (9) and therefore the asymptotic solution
of equation (1) could be written in the form
() =
a

1
() +
b

2
() (50)
where
1,2
() are some one-valued functions and a, b are given by the expres-
sions
a =
_

2
l
2
+
+ 1
2
(l
1
+l
2
)
_
b =
_

2
l
2
+
+ 1
2
(l
1
+l
2
)
_
(51)
The coecients l
1
and l
2
(see (28) and (36)) are expressed through the
elements

t
ij
of the matrix inverse to T ((24) and (27)). Therefore in order
to determine the asymptotic expression for solution of (1) we just need to
calculate the inverse matrix T
1
.
Matrix T has the form
T =
_
_
_
_
_
_
_
_
1 1 1 ... 1

1

+
1
2

2
...
N
+
1
2

2
1

2
1
+
1

2
2
...
2
N
+
N
. . . ... .

2N1
1

2N1
1
+
2N1
2

2N2
1

2N1
2
...
2N1
N
+
2N2
2

2N2
N
_
_
_
_
_
_
_
_
(52)
Its determinant will not change if we subtract odd columns from even ones:
det T = det T
1
=

1 0 1 ... 0

1
1
2

2
...
1
2

2
1

1

2
2
...
N
. . . ... .

2N1
1
2N1
2

2N2
1

2N1
2
...
2N1
2

2N2
N

(53)
11
The last expression could be written in the following form
det T =
1
2
N

1
...

N
W|

i
=
i
(54)
where W is Wandermond determinant
W =

1 1 1 ... 1

1

1

2
...
N
. . . ... .

2N1
1

2N1
1

2N1
2
...
2N1
N

(55)
Using the well known rule of calculation of Wandermond determinant we
obtain for det T
det T =
1
2
N

i>j
(
i

j
)
4
(56)
Implementing the same trick for the calculation of corresponding minors M
ij
we obtain the expression for the elements of inverse matrix T
1
:

t
1N
= 2
_
_
1

N
k=2
(
k

1
)
2

j
1

1
N

k=2
1
(
k

1
)
2
_
_

t
2N
= 2
N

k=2
1
(
k

1
)
2
(57)
Then
l
2
= 2
N

k=2
1
(
k

1
)
2
l
1
+l
2
= 2
N

j=2
1

1
N

k=2
1
(
k

1
)
2
(58)
The product and the sum entering last equation could be evaluated using
the characteristic equation (20) for matrix A
0
(11). Indeed, if one write the
characteristic polynomial P
2N
in the form
P
2N
= (
1
)...(
2N
) (59)
which for N large enough become approximately
P
2N
(
1
)
2
...(
N
)
2
(60)
12
because all the eigen values of matrix A
0
become twice degenerate, as it is
explained earlier, then
P

2N
|
=
1
(
1

2
)
2
...(
1

N
)
2
(61)
exactly the required product. Taking into account expression (22) for P
2N
in
the limit N one easily nds
N

k=2
(
k

1
)
2
=
(2N)!
(i)
2N
_

2
4
_
cos
_

1
2
_
(62)
In a similar way
N

k=2
1

1
=

N
k=2

N
m=2,m=k
(
1

m
)

N
k=2
(
1

k
)
=

_
P
(4)
2N
P
(2)
2n

=
1
(63)
and then
N

k=2
1

1
=

2

2
(64)
If one substitute these expressions into (51) one nds that the powers of
which describe the branching of solution at innity become
a

=
2(1 )
2
M
2
/ m
2

2
_
(1 )
2
M
2
/ m
2

1
2

2
_
(1
2
)
2
M
2
_
m
2
_
b

=
2(1 )
2
M
2
/ m
2

2
_
(1 )
2
M
2
/ m
2

1
2

2
_
(1
2
)
2
M
2
_
m
2
_
(65)
The plus or minus signs are taken at innities in S
+
and S

components of
S
F
correspondingly.
3 Mass spectrum
Having determined the form (50) of asymptotic solutions of equation (1) at in-
nities in S
+
and S

components of the Riemannian surface S

F
we can try to
impose the condition of decreasing of the wave function at both innities. But
the asymptotic expression for the solution (50) contains branching functions
13
if powers a and b are not integer. This means that the asymptotic expression
(50) could not be a good approximation for the one valued wave function
. It is known from complex analysis, that if the powers a and b of in
asymptotic expression (50) for the wave function are not integer numbers,
the approximation (50) is valid only in some sector
1
< Arg (1/ ) <
2
in the neighbourhood of innity point. Dierent approximations are valid
in dierent overlaping sectors. This behaviour of the asymptotic approxi-
mation is called Stokes phenomenon [5]. But if both a and b are integers,
the expression (50) is a good approximation for non-branching wave function
everywhere in the punctured neighbourhoods of innities.
Due two the double degeneracy of eigen values of matrix A
0
the innite ma-
trix equation is reduced to the innite set of second order matrix equations.
This second order equations can be analysed in the same way as we do this
in nonrelativistic quantum mechanics. And we conclude therefore that the
wave function with needed boundary conditions on the both sides of the Rie-
mannian surface exists only in the situation described above - when a

and
b

are integers.
The conditions for a, b to be integers are equivalent to
2(1 )
2
M
2
/ m
2

2
_
(1 )
2
M
2
/ m
2
= n
1
2

2
_
(1 )
2
M
2
_
m
2
_
= k (66)
where n and k are integers. These conditions dene the mass spectrum for
the equation (1).
The the motion of dust shell in its own gravitational eld is parameterised
by three parameters the bare mass of the shell M, the Schwarzchild mass
which is measured at right innity m = m
out
and the Schwarzchild mass
inside the shell (or the mass at left innity) m
in
= m. There are two
equations (66) on these three parameters, which we could solve for two of
them. Therefore, the spectrum of Schwarzchild mass m is parameterised by
two discrete parameters n and k and one continuous (for example ). But if
we restrict ourselves with situations when the gravitational eld in which the
shell moves is produced only by the shell itself, that is there is no space-time
singularity inside the shell, then = 0 and the mass spectrum of such states
is discrete.
14
The expression (66) for the mass spectrum is valid only when (1 )
2

M
2
/ m
2
> 0 when the square root expression entering (66) is well dened.
This restriction corresponds to classical restriction on the trajectories of
bound motion of the shell. If it is satised, then there exist some maxi-
mal radius of expansion of the shell (t) <
max
.
Otherwise the shell could escape to innity in its classical motion. For
such shells rst of conditions (66) disappears and the only left is
1
2

2
_
(1 )
2
M
2
_
m
2
_
= k (67)
It is remarkable that for light-like shells (M = 0) in their own gravita-
tional eld ( = 0) this quantization condition become (we recall that
= m
2
pl
_
2m
2
)
m = m
pl

2k (68)
4 Conclusion
In this technical note we developed a method of constructing asymptotical
solutions (50) of relativistic Schroedinger equation in nite dierences (1),
which describe a self-gravitating spherically symmetric dust shell of matter
[1]. This nite dierence equation is dened over Riemannian surface S
F
(4).
The conguration space of the system is the real section of this surface and
has nontrivial geometry. This complicated geometry of conguration space
of quantum system is due to complicated geometric structure of classical
space-time manifold of self-gravitating shell. Analysis of the behaviour of
asymptotic solution on Riemannian surface S
F
enabled us to write down
quantization conditions (66) under which the one valued solution of (1) with
physical boundary conditions exists. These quantization conditions dene
discrete spectrum for Schwarzchild mass m if the mass inside the shell is zero.
The main result of this paper is that this spectrum up to the numerical factors
coincides with the spectrum, found in [1] in large black hole approximation
and in [2] in quasiclassical approximation. This discrete mass spectrum is of
the form
m m
pl

k (69)
for the light-like shells (shells with bare mass M = 0). This form of quantum
black hole spectrum was extensively discussed by many authors [6, 7].
15
References
[1] V.A.Berezin, A.M.Boyarsky, A.Yu.Neronov, Phys. Rev. D 57, 1118,
(1998)
[2] A.Yu.Neronov, Phys. Rev. D 59 044023 (1999)
[3] V.A.Berezin, A.M.Boyarsky, A.Yu.Neronov, Black hole mass spectrum
vs. spectrum of Hawking radiation, gr-qc/9808027
[4] C.W.Misner, K.S.Thorne, J.A.Wheeler, Gravitation, Freeman, (1972)
[5] W.Wasow, Asymptotic expansions for ordinary dierential equations,
John Wiley and Sons, Inc. (1965)
[6] J.D.Bekenstein, Black Holes as Atoms, gr-qc/9710076
[7] J.D.Bekenstein, V.F.Mukhanov, Phys.Lett., B 360, 7 (1995)
16