20 May 1999

Physics Letters B 454 Ž1999. 171–175

Crossing and anticrossing of energies and widths

for unbound levels
1
P. von Brentano ) , M. Philipp
¨ Kernphysik, UniÕersitat
Institut fur ¨ zu Koln,
¨ Zulpicher
¨ ¨ Germany
Str. 77, 50937 Koln,
Received 9 November 1998; received in revised form 18 January 1999
Editor: J.-P. Blaizot

Abstract

The crossing and anticrossing properties of the energies and widths of two unbound levels under the influence of a
symmetrical complex interaction are investigated. It is found that a sufficiently large variation of the difference of the
‘‘unperturbed’’ energies or of the widths leads always to a crossing of either the energies or the widths of the ‘‘perturbed’’
system. A particularly interesting result is that for a real off diagonal interaction there is a joint crossing of the
‘‘unperturbed’’ energies and of either the ‘‘perturbed’’ energies or the ‘‘perturbed’’ widths. q 1999 Elsevier Science B.V.
All rights reserved.

PACS: 03.65.-w; 03.80.q r; 32.80.Bx; 21.10.-k

The two level system is a fruitful tool in physics
and has many applications w1–5x. One usually con-
siders, the properties of a system of two bound
states. It is of interest to extend this study from
bound states to unbound states w6–15x. Interesting
examples of unbound two level systems are e.g.:
1. The I p s 2q, T s 1, T s 0 doublet in 8 Be w16,17x.
2. The r – v T s 1, T s 0 doublet of mesons
w9,18,19x.
3. Doublets of resonance’s in cavities w20x.
)
E-mail: brentano@ikp.uni-koeln.de
1
Present address: SAP AG, Neurottstr. 16, 69190 Waldorf,
Germany.
This paper discusses the crossing and anticrossing
of energies and widths of the two level system for
unbound levels. For the system of two bound states it
is known that the difference D E s E1 y E2 of the
energies E1 and E2 can not vanish if the off diago-
nal matrix element of the interaction does not vanish
w1–6x. In short: the energies of two bound states,
anticross for a non vanishing offdiagonal interaction:
n / 0. This statement is a special case of a theorem
of Wigner and von Neumann w1x.
In the present paper the crossing and anticrossing
of unbound levels is studied. The energy ´ of an
unbound level is in general a complex number:
´ s E y ir2 G Ž 1.
Here E s Re ´ is the real energy and G s y2 Im ´
is the width of the unbound state. The complex

0370-2693r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 0 - 2 6 9 3 Ž 9 9 . 0 0 3 7 8 - 0
172 P. Õon Brentano, M. Philippr Physics Letters B 454 (1999) 171–175

energy difference ´ 1 y ´ 2 of two unbound states is From time reversal invariance one obtains further
thus also in general a complex number that the width matrix Gn m s Ž W t W . n m and the en-
ergy matrix h n m are real and symmetric 2 = 2 matri-
´ 1 y ´ 2 s D E y ir2 DG Ž 2. ces. The width matrix G s W t W is a positive
semidefinite matrix. Combining the energy matrix
There are different possible cases of crossing and h n m with the width matrix Gn m one obtains the
of anticrossing in the unbound two level system effective Hamiltonian matrix: H s h y ir2 G .
depending on the vanishing or non vanishing of the One can write the effective symmetric Hamilto-
real part D E or of the imaginary part y1r2D G of nian H in the form of Eq. Ž6..
the complex energy difference ´ 1 y ´ 2 .
We mention that there have been several works in
which the concept of energy repulsion and width H11 H12
attraction was discussed for a system of two un-
bound states w6–12,17x. This problem is related to
Hs
ž H21 H22 /
but different from the crossing and anticrossing rela- E10 y ir2 G 10 n y ir2 v
tion which will be discussed here. Furthermore the
crossing and anticrossing relations are derived here
for the general case of a 2 = 2 matrix with a sym-
s
ž n y ir2 v D E10 y ir2 G 10 / Ž 6.

metrical complex off diagonal interaction whereas
the energy repulsion width attraction relations were
derived previously only for an off diagonal interac-
tion which was either real or purely imaginary.
Before the crossing problem is discussed in detail,
the meaning of the effective Hamiltonian for un-
bound states will be clarified. A proper description
of the unbound system is done in the frame of an
S-matrix. We can define an effective Hamiltonian H
from the propagator of the S-matrix. A convenient
form of an unitary S-matrix which exhibits the prop-
agator has been given by Mahaux and Weidenmuller ¨
in their book w21x and was also used by other authors
e.g.: w13–15,22–25x. Time reversal invariance is as-
sumed and therefore the S-matrix is symmetric. For
a system with two unbound states this leads to the
following representation of the S-matrix w21x:
S Ž E . s U 1 y iW Dy1 Ž E . W t 4 Uq, Ž 3.
It is further assumed for simplicity that the effective
Hamiltonian H is energy independent. This is a
reasonable assumption far from thresholds.
This form allows the standard decomposition of
H into an ‘‘unperturbed’’ effective Hamiltonian H 0
and a complex off diagonal interaction V:
´ 01 0 0 n y ir2 v
H0s
ž 0 ´ 02
s
/ ž
n y ir2 v 0 /
Ž 7.
The special form n y ir2 v of writing the off
diagonal complex interaction matrix element is used
in order to simplify the relations below. The poles ´ 1
and ´ 2 of the S-matrix are identical with the com-
plex eigenenergies of the effective Hamiltonian and
are given by the solutions of Eq. Ž8.:

D Ž E . n m s Ed n m y H n m det Ž ´d n m y Hn m . s 0 Ž 8.
s Ed n m y h n m q 1r2i Ž W t W . n m , Ž 4. from which the well-known expressions for the com-
with plex energies of the two level system are obtained
w1–6x:
h )n m s h n m ,Wcn s Wcn) and n,m g w 1,2 x Ž 5.
´ 1,2 s 1r2 Ž H11 q H22 .
Here Wcn is the M = 2 matrix of the decay ampli-
tudes which couple the M channels to the 2 levels. U 2 1r2
is a unitary matrix, which describes the background. " 1r2 Ž H11 y H22 . q 4 H12 H21 Ž 9.
 P. Õon Brentano, M. Philippr Physics Letters B 454 (1999) 171–175
From Eq. Ž9. one obtains the square of the differ-
ence of the complex energies Ž ´ 1 y ´ 2 . 2 :
2 2
Ž ´ 1y ´ 2 . s Ž ´ 10 y ´ 20 .
2 crossing Ž D E s 0. or width crossing Ž D G s 0. as is
q 4 Ž n y ir2 v . s A y i B Ž 10 .
In order to discuss Eq. Ž10. it is useful to consider
the differences of the energies D E and of the widths
D G of the two ‘‘perturbed’’ states and of D E 0 and
D G 0 of the ‘‘unperturbed’’ states. By decomposing
Eq. Ž10. into its real part: A and its imaginary part:
yB one obtains Eqs. Ž11a. and Ž11b.:
2 2 2 2
Ž D E . y 1r4 Ž DG . s Ž D E 0 . y 1r4 Ž DG 0 .
q 4 Ž n 2 y 1r4v 2 .
sA
0 0
Ž D E DG . s Ž D E DG . q 4nv s B
These equations are the basis of the following dis-
cussions of the crossing and anticrossing of the
levels. One notes that the functional dependence of
the quantities A and B on the parameters of the
Hamiltonian H of Eqs. Ž2. and Ž6. are given by Eqs.
Ž11a. and Ž11b..
To begin the discussion we first consider the full
complex energy crossing i.e. the case: ´ 1 s ´ 2 . One
finds the Eq. Ž12.:
´ 1 s ´ 2 m A s 0 and B s 0
Eq. Ž12. gives the conditions for full complex energy
crossing. Such crossing is possible in the case of
unbound levels also for a nonvanishing interaction
n y ir2 v / 0. This was noted before w8–12x. The
reason is, that a complex symmetrical 2 = 2 matrix
has more parameters than the real symmetrical 2 = 2
matrix and these many parameters make it possible
to fulfil the two relations A s 0 and B s 0 also for a
non vanishing off diagonal interaction. The complex
energy crossing has been discussed in great detail by
Mondragon ´ and Hernandez
´ w11x.
We now discuss the partial crossing which is
particularly interesting for the unbound system.
Namely we consider that either the energy difference
D E, or the width difference D G vanish. One finds
in both cases that the parameter B must vanish. One
obtains Eq. Ž13.:
B s 0 m Ž D E DG . s 0 m D E s 0 or DG s 0 Ž 13 .
Eq. Ž13. contains the logical ‘‘or’’, which includes of
173
course the possibility that D E and D G vanish si-
multaneously as was discussed above. Eq. Ž13. is a
crossing anticrossing relation. One finds further that
the sign of A specifies whether there is energy
shown in Eqs. Ž14a. and Ž14b.:
A ) 0 and B s 0 m D E / 0 and DG s 0 Ž 14a.
A - 0 and B s 0 m D E s 0 and DG / 0 Ž 14b.
Relation Ž14a. is interesting. It states that for B s 0
energy anticrossing implies width crossing. For
bound states the relation is trivial because the widths
vanish everywhere. The relation is nontrivial for
unbound states, however. Eq. Ž11b. implies that by
Ž 11a. varying D E 0 or D G 0 in a sufficiently large range
Ž 11b. while keeping the other parameters of H constant
one can make B s 0. Thus in cases where D E 0 or
D G 0 can be varied in the experiment in a suffi-
ciently large range one finds either energy or width
crossing. The width crossing relation is a rather
general, somewhat surprising and interesting result.
The conditions under which it holds can be realized
in experiments.
Particularly simple and strong results are obtained
for a special off diagonal interaction for which nv s
0. That is for either a real or a purely imaginary off
Ž 12 . diagonal interaction. One finds:
nv s 0 and B s 0 m Ž D E DG . s Ž D E 0DG 0 . s 0
Ž 15 .
Thus for this special interaction the ‘‘perturbed’’
widths or energies will cross at the crossing point of
the ‘‘unperturbed’’ widths or energies. The question
whether the ‘‘perturbed’’ widths or energies cross
depends again on the sign of A as is shown in Eqs.
Ž11a. and Ž11b.. One finds Eqs. Ž16a. and Ž16b.:
<2 n < ) <1r2 DG 0 < and D E 0 s 0 and v s 0
ED E
´ DG s 0 and and D E / 0 Ž 16a.
ED E 0
<2 n < - <1r2 DG 0 < and D E 0 s 0 and v s 0
EDG
´ D E s 0 and and DG / 0 Ž 16b.
ED E 0
174 P. Õon Brentano, M. Philippr Physics Letters B 454 (1999) 171–175
In Eqs. Ž16a. and Ž16b. all quantities of the ‘‘per-
turbed’’ system as e.g. D E, D G , treated as functions
of the parameter D E0 are to be taken at the value
D E 0 s 0. One can also obtain corresponding rela-
tions for a variation of the parameter D G 0 . The
derivatives ED ErED E 0 and EDGrED E 0 in Eqs. Ž16a.
and Ž16b. are obtained from Eqs. Ž11a. and Ž11b. by
keeping all parameters Ž D G 0 , n , v . constant except
D E0 .The relation Ž16a. implies that for a sufficiently
large real interaction with <2 n < ) <1r2 D G 0 < the
three quantities D E 0 , D G and EDGrED E 0 will
vanish jointly wheras D E does not vanish for
D E 0 s 0.
Eq. Ž16b. gives the conditions for a joint vanish-
ing of the three quantities D E0 , D E and EDGrED E 0
whereas D G does not vanish for D E 0 s 0. One
notes that Eq. Ž16a. is well known for boundstates. It
shows that the difference of the energies Ž D E . s Ž E1
y E2 . of the two states has an extremal value at the
crossing point of the ‘‘unperturbed’’ energies Ž D E 0
s 0..
It should be stressed that the joint crossing of the
three quantities is found for purely real or imaginary
interactions. For a general complex interaction the
three quantities will not vanish jointly in general.
It is useful to derive the width crossing relation
directly in a simple model. The electromagnetic de-
cay of a system of two interacting bound states c 1
and c 2 to the groundstate cg is considered in
perturbation theory. The system is described by the
Hamiltonian of Eq. Ž6. with v s 0 and vanishing
‘‘unperturbed’’ widths: G 01 s G 20 s 0. One assumes
that the two states c 1 , c 2 decay by electromagnetic
E2-transitions to the ‘‘unperturbed’’ groundstate
< cg0 : s < cg :. Thus one obtains the transition decay
widths G 1t and G 2t in perturbation theory by Eq.
Ž17.:
G 1tg s <² cg0 < E2 < c 1 :< 2 a2 , G 2tg s <² cg0 < E2 < c 2 :< 2 a2
Ž 17 .
where a is a constant. We use the same relations
also for the unperturbed case. One assumes in this
model for simplicity that in the ‘‘unperturbed’’ sys-
tem H 0 there is only one nonvanishing transition
width G 1tg0 / 0, G 2tg0 s 0. With these assumptions
one can calculate the difference of the widths: D G t s
Ž G1tg y G 2tg .t as a function of the parameters D E 0
and n . D G is directly related to the difference of
the amplitudes a 2 and b 2 of the expansion of the
‘‘perturbed’’ states into the ‘‘unperturbed’’ states
which is given in Eq. Ž18.:
c 1 s ac 10 q bc 20 , c 2 s ybc 10 q ac 20 Ž 18 .
One finds DG t s Ž a 2 y b 2 .. This shows the origin
of width crossing. Namely the amplitudes a 2 and
b 2 cross at the parameter value D E 0 s 0.
This model gives an intuitive explanation for
width crossing in terms of a complete mixing of the
two states. This model is valid of course only for
very small widths, when perturbation theory applies.
We do not know a similarly intuitive explanation for
energy crossing.
Summing up the anticrossing relation for the ener-
gies in a system of two bound states is extended to
crossing–anticrossing relations for the energies and
widths of a system of two unbound states. Since the
energies of the unbound states are complex numbers
there is a rich physical scenario for crossing and
anticrossing of the real or imaginary parts of the
complex energies.
A particularly interesting result is that in a system
of two unbound states with an off diagonal interac-
tion there is a value of the difference of the real
‘‘unperturbed’’ energies D E0 for which the ‘‘per-
turbed’’ energies or the ‘‘perturbed’’ widths cross.
An even stronger result is obtained for a sufficiently
strong real interaction. For such an interaction one
finds that the three quantities D E 0 , D G and ED Er
ED E 0 will vanish jointly.
This is a rather general result because the condi-
tions for which it holds are rather weak and can be
realized in experiments.
Finally we give an experimental example for
which the scenario of joint ‘‘unperturbed’’ energy
crossing: D E 0 s 0 and ‘‘perturbed’’ width crossing:
D G s 0 is at least approximately fullfilled. This ex-
ample is the famous doublet of I p s 2q T s 0 and
T s 1 resonances in 8 Be. These resonances were
studied both in experiment and in analysis in great
detail, e.g. by Hinterberger et.al. w16x. They have
become one of the most completely studied exam-
 P. Õon Brentano, M. Philippr Physics Letters B 454 (1999) 171–175
ples of resonance doublets in nuclear physics Žsee
also references in w16x.. We follow here an analysis
w17x which used the same form of the S-matrix as is
used in the present paper.
The parameters ´ 1 , ´ 2 and ´ 10 , ´ 20 and n ob-
tained for the 8 Be doublet in Ref. w17x are:
´ 1 s Ž 16722y ir2 P 108 . keV
´ 2 s Ž 17010y ir2 P 74 . keV
´ 10 s Ž 16838y ir2 P 182 . keV
´ 20 s 16893 keV
n s 148 keV
From these parameters one can find the parameters
D ´ , D ´ 0 and n . One finds:
D´ 0 s Ž y55 y ir2 P 182 . keV
D´ s Ž y288 y ir2 P 35 . keV
As the interaction n s 148 keV is real, Eq. Ž15.
holds. This is indeed true: ( 288 = 35 s 55 = 182.
We note that although both D E0 and D G both do
not vanish they are both rather small compared to
< D E < and < D G 0 < respectively:
< D E0 <s55 keV < 288 keVs D E
and < DG < s 35 keV < 182 keV s < DG 0 <
Thus although there is no true ‘‘perturbed’’ width
crossing in 8 Be the system is near to such a crossing.
We also note that the interaction < n < is large: <2 n < s
298 keV ) 91 keV s <1r2 D G 0 < and thus Eq. Ž16a.
implies near ‘‘perturbed’’ width crossing as found in
the experiment. The two resonances in 8 Be have also
been used as an example of a near complex energy
crossing w12x.
Other applications may be found in atomic physics
or in microwave cavities. One can consider e.g. that
the energies of an atom are changed by a magnetic
field or one can consider a set of nearly identical
coupled microwave cavities in which the energies
widths and the interaction n of the individual cavi-
ties can be separately varied w20,26x.
175
Acknowledgements
We thank Dr. G. Pascovici for many discussions.
References
w1x J. von Neumann, E.P. Wigner, Z. Phys. 30 Ž1929. 467.
w2x L.D. Landau and E.M. Lifschitz, Quantenmechanik 8. Au-
flage, Akademie-Verlag Berlin, 1988, Ch. 11.
w3x The Feynman lectures, Vol. 3. Quantum mechanics, R.P.
Feynman, R.B. Leighton and M. Sands, Addision-Wesley
Publishing Co., London, 1970, Chs. 9–11.
w4x M.V. Berry, M. Wilkinson, Proc. Roy. Soc. London A 392
Ž1984. 15.
w5x M.V. Berry, Proc. Roy. Soc. London A 239 Ž1983. 45.
w6x C. Cohen-Tannoudji, B. Diu, F. Laloe, Mecanique quantique,
Herman, Paris, 1973, Ch. 4.
w7x H. Friedrich, D. Wintgen, Phys. Rev. A 32 Ž1985. 3231.
w8x P. von Brentano, Phys. Lett. B 238 Ž1990. 1.
w9x P. von Brentano, Physics Report 264 Ž1996. 57.
w10x E. Hernandez,
` ´ Rev. Mex. Fis. 38
A. Jauregui, A. Mondragon,
ŽS2. Ž1992. 128.
w11x A. Mondragon,´ E. Hernandez,
` J. Phys. A 26 Ž1994. 5595.
w12x E. Hernandez,
` A. Mondragon,´ Phys. Lett. B 326 Ž1994. 1.
w13x V.V. Sokolov, V.G. Zelevinski, Phys. Lett. B 202 Ž1988. 10.
w14x V.V. Sokolov, V.G. Zelevinski, Nucl. Phys. A 504 Ž1989.
562.
w15x V.V. Sokolov, V.G. Zelevinski, Ann. Phys. ŽNY. 216 Ž1992.
323.
w16x F. Hinterberger et al., Nucl. Phys. A 299 Ž1978. 397 and
references therein.
w17x P. von Brentano, Phys. Lett. B246 Ž1990. 320.
w18x L.M. Barkov et. al., Nucl. Phys. B 256 Ž1985. 365 and
references therein.
w19x R.P. Feynman, Photon Hadron Interactions, W.A. Benjamin,
1972.
w20x H. Alt et. al., Nucl. Phys. A 560 Ž1993. 293 and references
therein.
w21x C. Mahaux and H.A. Weidenmuller, ¨ Shell-Model approach
to nuclear reactions, North-Holland, Amsterdam, 1969, p. 51.
w22x H. Feshbach, Theoretical Nuclear Physics, nuclear reactions,
Wiley, New York, 1992.
w23x J.J.M. Verbaarschot, H.A. Weidenmuller,¨ M.R. Zirnbauer,
Phys. Rep. 129 Ž1985. 369.
w24x V.D. Kirilyuk, N.N. Nikolaev, L.B. Okun, Yad. Fiz. 10
Ž1969. 1081.
w25x V.D. Kirilyuk, N.N. Nikolaev, L.B. Okun, Sov. J. Nucl.
Phys. 10 Ž1970. 617.
w26x H. Alt et al., Phys. Rev. Lett. 81 Ž1998. 4847.