Compton Scattering by Nuclei ∗

M.-Th. Hütt a,1 , A.I. L’vov b,2 , A.I. Milstein c,3 , M. Schumacher a,4
a Zweites Physikalisches Institut, Universität Göttingen, D-37073 Göttingen, Germany
b P.N. Lebedev Physical Institute, Leninsky Prospect 53, Moscow 117924, Russia
c Budker Institute of Nuclear Physics, 630090 Novosibirsk, Russia

Abstract
arXiv:nucl-th/9905026 v1 13 May 1999

The concept of Compton scattering by even-even nuclei from giant-resonance to nucleon-reso-

nance energies and the status of experimental and theoretical researches in this field are outlined.
The description of Compton scattering by nuclei starts from different complementary approaches,
namely from second-order S-matrix and from dispersion theories. Making use of these, it is possible
to incorporate into the predicted nuclear scattering amplitudes all the information available from
other channels, viz. photon-nucleon and photon-meson channels, and to efficiently make use of
models of the nucleon, the nucleus and the nucleon-nucleon interaction. The total photoabsorption
cross section constrains the nuclear scattering amplitude in the forward direction. The specific
information obtained from Compton scattering therefore stems from the angular dependence of the
nuclear scattering amplitude, providing detailed insight into the dynamics of the nuclear and nu-
cleon degrees of freedom and into the interplay between them. Nuclear Compton scattering in the
giant-resonance energy-region provides information on the dynamical properties of the in-medium
mass of the nucleon. Most prominently, the electromagnetic polarizabilities of the nucleon in the
nuclear medium can be extracted from nuclear Compton scattering data obtained in the quasi-
deuteron energy-region. In our description of this latter process special emphasis is laid upon the
exploration of many-body and two-body effects entering into the nuclear dynamics. Recent results
are presented for two-body effects due to the mesonic seagull amplitude and due to the excitation of
nucleon internal degrees of freedom accompanied by meson exchanges. Due to these studies the in-
medium electromagnetic polarizabilities are by now well understood, whereas the understanding of
nuclear Compton scattering in the ∆-resonance range is only at the beginning. Furthermore, phe-
nomenological methods how to include retardation effects in the scattering amplitude are discussed
and compared with model predictions.

∗
Supported by Deutsche Forschungsgemeinschaft (Schu222, SFB201, 436RUS113/510) and DAAD
1
e-mail: huett@bio.tu-darmstadt.de
2
e-mail: lvov@x4u.lebedev.ru
3
e-mail: A.I.Milstein@inp.nsk.su
4
e-mail: Martin.Schumacher@phys.uni-goettingen.de

i
Contents

1 Introduction 1
1.1 Early work on nuclear Compton scattering in the giant-resonance region . . . . . . . . 1
1.2 Nuclear Compton scattering experiments from the mid-1970th to the present . . . . . 1
1.3 Motivations for studying nuclear Compton scattering . . . . . . . . . . . . . . . . . . . 2
1.4 Fields related to nuclear Compton scattering . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Organization of the present paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 The absorption of photons and the degrees of freedom of the nucleus 6

2.1 Photoabsorption by the nucleon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Photoabsorption by the nucleus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Summary on the degrees of freedom of the nucleus . . . . . . . . . . . . . . . . . . . . 10
2.4 The role of Compton scattering in investigations of the degrees of freedom of the nucleus 11

3 Compton scattering below pion threshold: general aspects 12

3.1 The photon-nucleus scattering amplitude in second-order perturbation theory . . . . . 13
3.2 The nonrelativistic photon-nucleus interaction and the concept of currents . . . . . . . 14
3.3 Low-energy behavior of the scattering amplitude and dispersion relation . . . . . . . . 17
3.4 Multipole expansion of the scattering amplitude for giant resonances . . . . . . . . . . 21
3.5 Multipole and isospin decomposition of the seagull amplitude . . . . . . . . . . . . . . 26

4 Compton scattering by giant resonances 28

4.1 The giant resonance amplitude in the Lorentzian representation . . . . . . . . . . . . . 29
4.2 The unretarded resonance amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3 The retardation problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.4 Gerasimov’s argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.5 The effects of enhancement and retardation in electric multipole sum rules . . . . . . . 39
4.6 Scaling of giant resonance parameters via Compton scattering . . . . . . . . . . . . . . 41
4.7 The isovector giant-quadrupole resonance and higher multipoles . . . . . . . . . . . . . 44
4.8 The electromagnetic polarizabilities of the nucleus . . . . . . . . . . . . . . . . . . . . 45

ii
5 Compton scattering in the quasideuteron range 49
5.1 The quasideuteron amplitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 The single-nucleon contribution to Compton scattering up to quadratic order in the
photon energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.3 Status of the free polarizabilities of the nucleon . . . . . . . . . . . . . . . . . . . . . . 52
5.4 Meson exchange currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.5 Experiments on the bound-nucleon electromagnetic polarizabilities . . . . . . . . . . . 54
5.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6 Compton scattering above π meson threshold 58

6.1 Pioneering work on Compton scattering in the resonance range . . . . . . . . . . . . . 58
6.2 Compton scattering by 12 C in the ∆ range . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3 Compton scattering by 4 He in the ∆ range . . . . . . . . . . . . . . . . . . . . . . . . 59
6.4 Calculations of Compton scattering in the ∆ range . . . . . . . . . . . . . . . . . . . . 60

7 Fermi liquid theory and nuclear Compton scattering 65

8 Dispersion relations at fixed momentum transfer for nuclear Compton scattering 74

8.1 General structure of the Compton amplitude . . . . . . . . . . . . . . . . . . . . . . . 74
8.2 Resonance and seagull amplitudes at fixed momentum transfer . . . . . . . . . . . . . 77

9 The mesonic seagull amplitude 82

9.1 Construction of the mesonic seagull amplitude for nuclear matter . . . . . . . . . . . . 82
9.2 Corrections to the correlation function . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
9.3 The mesonic seagull amplitude in finite nuclei . . . . . . . . . . . . . . . . . . . . . . . 88

10 Retardation effects in Compton scattering. A model study with a relativistic

oscillator 92
10.1 Motivation and aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
10.2 Basics of the model. The seagull and resonance amplitudes . . . . . . . . . . . . . . . 93
10.3 Asymptotic behavior and fixed-t dispersion relations . . . . . . . . . . . . . . . . . . . 97
10.4 Retarded and unretarded sum rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
10.5 Comparison with the phenomenological approach . . . . . . . . . . . . . . . . . . . . . 104

A Parameters of the mesonic seagull amplitude for different nuclei 106

B Exchange form factors 107

C Multipole angular distribution functions 109

iii
1 Introduction

The discovery made by Compton in 1922 [1] that photons are scattered by free electrons is one of
the mile-stones of modern physics. In this process an energy shift of the photon was observed which
quantitatively could be explained by assuming an elastic collision between particles. Later on, the term
Compton scattering has been transferred to photon scattering by other objects as there are atoms,
nucleons and nuclei. An essential difference between free electrons and these other objects is that the
latter have an internal structure so that excitation and particle emission processes become possible. In
that case the term Compton scattering is used to denote coherent-elastic photon scattering where the
scattering object coincides in all (internal) quantum numbers in the initial and final states. In case of
atoms the term Compton scattering occasionally is replaced by the term Rayleigh scattering [2]. This
is of advantage in cases where the coherent-elastic scattering process (Rayleigh scattering) is discussed
together with the incoherent scattering process (also called Compton Scattering), where one bound
electron is ejected from the electron cloud into the continuum. This process then resembles the one
discovered by Compton [1] except for the appearance of binding effects [3]. The term Raman scattering
is used in the case where the atom or nucleus does not return to the initial state but no particle emission
is taking place. Another related photon scattering process is coherent-elastic scattering in the Coulomb
field of nuclei termed Delbrück scattering [4,5]. The three coherent-elastic photon scattering processes,
viz. atomic Rayleigh scattering, Delbrück scattering and nuclear Compton scattering are expected to
interfere with each other. This has been observed in several investigations [5].

1.1 Early work on nuclear Compton scattering in the giant-resonance region

Scattering of photons in the region of the “giant resonance” of the nucleus was probably first observed
by Gaerttner and Yeater [6] and Dressel, Goldhaber, and Hanson [7]. These works were motivated by
Goldhaber and Teller [8] who proposed to understand a pronounced structure in the photoabsorption
cross section between 10 and 30 MeV in terms of a collective motion of protons against neutrons.
The experiments were made feasible by the advent of betatrons delivering photons of sufficient energy
but suffered from the lack of efficient photon detectors. More refined experiments became possible
with the advent of NaI(Tl) scintillation counters which were first applied by Stearns [9] and Fuller
and Hayward [10, 11]. Stearns used monochromatic photons of 17.6 MeV from the 7 Li(p,γ) nuclear
reaction whereas Fuller and Hayward used continuous photons produced by a betatron.
A remarkable progress was made by Hayward et al. in 1973 [12] with the first application of
monochromatic linearly polarized photons. These photons were produced by resonance fluorescence
of the well-known 1+ state at 15.1 MeV in 12 C. With these experiments a discrimination was possible
between nuclear Compton scattering and nuclear Raman scattering, the latter being especially strong
for deformed nuclei.

1.2 Nuclear Compton scattering experiments from the mid-1970th to the present

The breakthrough in investigations of nuclear Compton scattering came with the advent of large-
volume NaI(Tl) detectors and the availability of quasimonochromatic photons [13–48]. As a first
step the technique of positron-annihilation-in-flight has been applied to produce quasimonochromatic
photons. In early uses of bremsstrahlung two different techniques have been developed. In one of
these the total spectrum of bremsstrahlung photons was used. By applying very narrow collimators
in front of the NaI(Tl) detectors the low-energy tails of the response functions were largely suppressed
so that large portions of the spectra of scattered photons could be interpreted in terms of Compton
scattering. This method was efficiently used by B. Ziegler et al. [25, 26, 30] and led to insight into

1
general properties of nuclear scattering amplitudes below meson photoproduction threshold, viz.the
importance of formfactors, the possibility to investigate the electromagnetic polarizabilities of nucleons
in the nuclear medium (cf. also [24]), etc. When using wider collimators in front of the NaI(Tl)
detectors the response functions generate a low-energy tail. In this case only a narrow energy range
at the upper end of the bremsstrahlung spectrum can be interpreted in terms of Compton scattering.
The techniques described in the foregoing paragraph have been more and more replaced by tagged
photons either from bremsstrahlung or from backscattered laser light. In this technique the energy
of the secondary electron is measured in a magnetic spectrometer so that the energy of the primary
photon is known through a coincidence condition with the Compton scattered photon or some other
reaction product. This technique was first applied to nuclear Compton scattering experiments at the
University of Illinois [49] and is now standard in all laboratories — viz. Brookhaven [50], Lund [51,52],
Mainz [53, 54], Saskatoon [55] — where experiments on Compton scattering by nuclei are carried out.
Back scattering of laser light [50] and coherent bremsstrahlung from diamond crystals [56–58] are also
efficient sources of linearly polarized photons. At low energies (5–100 MeV) also the method of off-axis
tagging [33, 59] leads to reasonable degrees of linear polarization.
An other type of experiments which has to be mentioned here has achieved very high energy
resolution (≈ 10 keV) due to the use of Ge(Li) detectors and due to the use of photons from nuclear
reactions [60–66]. These experiments were restricted to energies from about 5 MeV [61] to 17.4
MeV [63]. The high energy-resolution was of essential help in cases where Compton scattered photons
had to be discriminated from photons stemming from nuclear resonance fluorescence or nuclear Raman
scattering.

1.3 Motivations for studying nuclear Compton scattering

Several experiments on Compton scattering by nuclei were concerned with multipole decompositions
of giant resonances [13, 15, 16, 18, 19, 21, 25, 26, 30, 33, 64]. The comparatively weak electric quadrupole
amplitude is enhanced through interference with the dominant electric-dipole amplitude. Therefore,
nuclear Compton scattering became a tool to study the properties of the isovector electric quadrupole
resonance. In fact, this is the only model independent method existing for this purpose. The status
of this research will be discussed in section 4.
An interesting feature of nuclear Compton scattering below meson photoproduction threshold
which has been realized very early is its strong sensitivity to meson exchange currents (MEC) between
nucleons [67–83]. In this respect Compton scattering is of essential superiority to electron scattering
at similar momentum transfers which is to a large extent insensitive to this phenomenon. The study
of MEC degrees of freedom and their interplay with nuclear degrees of freedom have attracted strong
interest in many investigations, and different schools have approached these phenomena in different
ways. Meson exchange currents have long been realized to be the origin for the enhancement of the
integrated dipole strength in comparison with the prediction ΣT RK of the usual Thomas-Reiche-Kuhn
(TRK) sum rule [67]. The integrated strength of the giant-dipole resonance (GDR) overshoots the
TRK sum rule prediction by the factor (1 + κGDR ). The quasi deuteron (QD) photoabsorption pro-
cess which can be interpreted as the absorption by correlated proton-neutron pairs inside the nucleus,
provides an additional strength [84] κQD ΣT RK . It was a great progress to show [68] that due to gauge
invariance MEC also have to modify the non-resonant contribution to the scattering amplitude, i.e.
the so-called seagull amplitude, which accompanies resonant scattering through intermediate excited
nuclear-structure states. In the absence of MEC the seagull amplitude would simply be the superpo-
sition of the Thomson scattering amplitudes of the protons in the nucleus forming the “kinetic seagull
amplitude”. Accordingly it was discussed in what way the modification of the total seagull amplitude
through an additional “exchange” or “mesonic” seagull amplitude could possibly be understood as

2
Thomson scattering by pions in nuclei [71–75, 78, 83]. According to Siegert’s theorem [85] there is no
charge-density modification due to mesons in a nucleus (except for a small relativistic correction) but
an additional (two-body) current due to the exchange of charged mesons. Therefore, the interpre-
tation of the MEC contribution to the seagull amplitude requires special care (cf. sections 3.3 and
7). It turns out that the giant-resonance part of the mesonic seagull amplitude is closely related to
the in-medium mass of the nucleon [63, 66, 86–97] which is a special case of the general Brown and
Rho scaling [91–94, 96–98]. One motivation to study nuclear Compton scattering, therefore, is to get
further insight into the properties of the in-medium mass of the nucleon.
Below meson photoproduction threshold the internal excitation of the nucleon enters into the
nuclear Compton scattering amplitude via the electromagnetic polarizabilities of the nucleon [99–101].
The electromagnetic polarizabilities are fundamental constants, characterizing the response of the
nucleon to an external electromagnetic field and, thus, the dynamical structure of the nucleon. The
advantage of the electromagnetic polarizabilities of the nucleon is that they can precisely be measured
for the free and the bound nucleon by carrying out Compton scattering experiments. Because of
this property, the electromagnetic polarizabilities are an ideal observable for exploring possible in-
medium modifications of the nucleon structure. Some fundamental theoretical aspects of the in-
medium electromagnetic polarizabilities have been discussed on the basis of quark models [102, 103]
and pion-cloud models [80, 82, 104]. In recent time essential progress has been made in this field of
research both experimentally and theoretically (cf. sections 5 and 9).
For energies in the resonance range of the nucleon above meson photoproduction threshold only
first steps have been made in the investigation of nuclear Compton scattering. One reason for this
delay is that only a very fragmentary set of data was available, even for the nucleus 4 He which was the
subject of principal interest for experimental researches [20, 23, 29, 35, 40, 44, 45]. On the experimental
side considerable progress has been made very recently by the Göttingen-Mainz collaboration at MAMI
(Mainz) [44, 46] and by the LEGS collaboration at Brookhaven [45] where differential cross sections
for Compton scattering by 4 He have been measured in large angular and energy intervals. These
experiments showed that the widely accepted theoretical description of Compton scattering in terms
of the ∆-hole model [105–109] is in serious disagreement with experimental data [46]. This leads to
the conclusion that the in-medium excitation of the ∆ resonance is not fully understood and possibly
may lead to new insights into the structure of the bound nucleon.

1.4 Fields related to nuclear Compton scattering

In addition to nuclear Compton scattering, nuclear Raman scattering and nuclear resonance fluores-
cence have remained branches of photon scattering researches up to present days. For deformed nuclei
with ground-state spins I ≥ 1 Raman scattering is possible as an incoherent (non-interfering) elastic
scattering process ending up with a nonexcited nucleus, but in a spin-state differing from that of the
initial state. Raman scattering into excited states [49, 110–115] is also possible as an inelastic scat-
tering process. The theoretical framework appropriate to describe Raman scattering is the concept
of generalized polarizabilities [116–122] leading in a natural way to a discrimination between scalar
(Compton) and tensorial (Raman) scattering and — as a third possibility — vector scattering.
Nuclear resonance fluorescence takes place at energies below particle threshold of the nucleus as an
elastic photon scattering process through the excitation of isolated nuclear states [123]. The study of
this process serves as a tool of low-energy nuclear spectroscopy which is not considered here in detail.

3
1.5 Organization of the present paper

The topic of Compton scattering by nuclei cannot be covered completely in one article. Therefore, this
paper is strongly biased by our own preferences and results. Historical aspects and lines of development
going back to previous researches have briefly been covered in the foregoing parts of this section, or
will be mentioned in introductory parts of sections to come. To supplement on this we recommend to
take notice also of previous reports [121, 124–130]. We believe that the present report differs from the
previous ones by the fact that subnuclear and retardation phenomena of nuclear Compton scattering
have received a much better understanding due to recent — partly unpublished — researches. Also,
due to very recent experimental and theoretical results Compton scattering in the ∆ resonance range
has received a wider coverage than before.
The present article is concerned with Compton scattering by spin-saturated nuclei. This restriction
involves that the lightest nucleus to be discussed here is 4 He. Other nuclei of experimental interest are
12 C, 16 O, 40 Ca and 208 Pb. These are the nuclei where a reasonable amount of experimental data on

Compton scattering are available. In comparison to these the investigations of the deuteron — though
being the most fundamental ones — are at the very beginning and we give here only few references
to recent experimental and theoretical works [131–136].
In section 2 we give a definition of the degrees of freedom of nucleons in a nucleus. Our main
goal is to get a clear-cut distinction between the external (nuclear structure) and the internal degrees
of freedom of the bound nucleon. It is pointed out that the “classical subnuclear degrees of freedom”
viz. meson exchange currents (MEC) are not specific for the one or the other part but separately
enter into both.
In section 3 we formulate an outline of the theoretical aspects of Compton scattering below
meson photoproduction threshold. We start from relativistic second-order perturbation theory, point
out how MEC and the “negative energy states” of the nucleus can be taken care of, and show how
dispersion theory can be used to make links to experimental data.
Section 4 is devoted to our present understanding of giant-resonances. Since previous work
has been covered in previous reviews [121, 126, 127] we concentrate on our own regions of researches,
as there are enhancement and retardation effects and the dynamics of the giant-resonance mode.
Furthermore, we give a detailed description of the retardation problem which is sizable in nuclear
Compton scattering because the wavelengths of the photon 1/ω, where ω is the photon energy, and
the nuclear radius R are of the same order of magnitude. The solution of the retardation problem
which we present may appear as a preliminary guess at a first sight. Nevertheless, the results we
obtain are quite precise numerically, as will be shown in section 10 through a model calculation.
Section 5 is devoted to effects which are observed in the quasideuteron range of the nucleus. Since
the quasideuteron process itself does not make large contributions to the nuclear scattering amplitude
this section is mainly concerned with the polarizabilities of the bound nucleon which belong to the
most interesting aspects of nuclear Compton scattering.
Section 6 describes our present knowledge in the ∆-resonance region. Unfortunately, the study
of nuclear Compton scattering above meson photoproduction threshold is only at the beginning and
very interesting properties of nuclear Compton scattering in this energy range have not yet been
investigated quantitatively. Accordingly, this section only provides a brief description of recent points
of discussion.
Section 7 again is concerned with the giant-resonance region but describes interpretations of
subnuclear effects in terms of effective masses and effective currents of nucleons in the framework of
the Fermi liquid theory. This section also refers to modern views, e.g. the general Brown and Rho
scaling [96, 97].

4
 The following three sections 8–10 are devoted to recent theoretical developments. Section 8
describes a dispersion theory at fixed momentum transfer for nuclear Compton scattering. In section
9 a diagrammatic model of meson exchange currents is discussed, which essentially clarifies the most
important properties of the mesonic seagull amplitude and, in addition, leads to quantitative results
for exchange form factors and meson-induced polarizability modifications. In section 10 retardation
effects are studied within a relativistic oscillator model.
We do not include a discussion of nuclear Compton scattering in the asymptotic range beyond
1.5–2.0 GeV into this article, because there has been no recent research and because there are com-
prehensive reviews of this field [137]. Access to literature is given in [128] for diffractive Compton
scattering taking place at small scattering angles. Large-angle Compton scattering takes place via
direct photon-quark coupling which is described in [138–141].
Summary: Our main motivation for studying nuclear Compton scattering was to investigate in-
medium properties of nucleons as there are
(i) the in-medium mass of the nucleon,
(ii) the in-medium electromagnetic polarizabilities,
(iii) the in-medium ∆ excitation.
During these studies it turned out that some theoretical aspects which are essential for the data
analysis were not understood with the desirable clarity and, therefore, had to be investigated along
with the in-medium properties.
These are
(iv) the relation between retardation and higher multipoles,
and
(v) the effects of meson-exchange currents on the in-medium electromagnetic polarizabilities.
Further work has to be done to understand
(vi) the effects of meson exchange currents on Compton scattering in the ∆-range.

5
2 The absorption of photons and the degrees of freedom of the nu-
cleus

Compton scattering is closely related to the process of photoabsorption, since the Compton amplitude
for forward scattering can be expressed via the total absorption cross section with the help of the
optical theorem and a dispersion relation. In this sense the total photoabsorption cross section is a
special case of Compton scattering and, therefore, a detailed understanding of nuclear photoabsorption
is of essential help for the discussion of Compton scattering. Our goal is to investigate the nuclear
and subnuclear degrees of freedom of the nucleus in a consistent way. This means that we first have
to consider the properties of free nucleons and then we have to ask what the modifications are which
enter into the picture as a consequence of nuclear binding. As a result of this discussion we arrive at
a clear-cut distinction between the external (nuclear structure) and internal degrees of freedom of the
bound nucleon.

2.1 Photoabsorption by the nucleon

For illustration Fig. 2.1 shows the photoabsorption cross section of the proton separated into partial
cross sections. The photoabsorption cross section of the nucleon starts at the meson photoproduction
threshold and continues through different regions of photo-excitation mechanisms. In the region of
nucleon resonances which terminates at photon energies of about 1.5 to 2 GeV we find three broad lines
corresponding to the ∆(1232)P33 , N(1520)D13 and N(1680)F15 nucleon resonances located at photon
energies of 320 MeV, 740 MeV and 1020 MeV, respectively. The excitation of these resonances proceeds
through photons of, respectively, M 1, E1 and E2 multipolarities. The properties of these resonances
and also those of the weaker ones have successfully been understood in terms of three constituent
quarks located in a binding potential. In addition we find a nonresonant background corresponding
to one-pion photoproduction up to about 400 MeV and to more complicated mechanisms above.

Figure 2.1: Photoabsorption cross section of the proton schematically separated into partial cross
sections [142] containing nucleon resonances and a nonresonant component.

The range of energies beyond about 2 GeV is frequently called the asymptotic range of the pho-
toabsorption cross section. Here we find vector mesons, i.e. ω, ρ, φ etc. in the intermediate state.
This means that there is no direct interaction of the photon with the constituents of the nucleon.
In addition, the interaction of these vector mesons with the nucleon also is not direct but proceeds
peripheral through t-channel exchanges. The main candidates for these t-channel exchanges are the
isoscalar tensor meson f2 (1270), the isovector tensor meson a2 (1320) and the Pomeron P, where the
Pomeron takes care of the energy independent part of the photoabsorption cross section, while the
two tensor mesons take care of its energy dependent part. The isovector meson a2 (1320) has to be
introduced in addition to the isoscalar meson f2 (1270) because proton and neutron differ slightly in
the energy-dependent part of the cross section [142]. The t-channel exchanges entering into the de-
scription of forward-direction Compton scattering have spin 2 in accordance with general conservation
rules.
In considering the properties of the nucleon photoabsorption cross section in the asymptotic range,
we have to be aware that the total photoabsorption cross section is subject to inclusive reactions
where there is no detection of the particles in the final state. In this case we predominantly observe
those processes which make the largest contributions to the photoabsorption cross section. These
are the processes at small transverse momentum, i.e. the peripheral processes mentioned above. On
the other hand, in exclusive reactions where particles in the final state are detected, processes with

6
large transverse momenta may be separated experimentally. Here we again observe direct couplings of
photons with constituents which now are the individual valence quarks. Some more detailed discussions
of this aspect can be found in articles contained in [143] and in [138, 139].

2.2 Photoabsorption by the nucleus

When combining nucleons to form a nucleus three major changes are observed:
(i) The photoabsorption cross section extends below pion photoproduction threshold to form
two nuclear-structure modes of photo excitation, viz. the giant resonance (GR) mode and the
quasideuteron (QD) mode. In order to understand these two modes of excitation, the internal struc-
ture of the nucleon probably may not be of explicit relevance — although we do not know this for sure.
To start with, we may use a picture where unmodified nucleons in their ground states and mesons
being exchanged between them are the only nuclear constituents to deal with.
Figure 2.2 gives a schematic view of the photoabsorption cross section of 208 Pb between the gi-
ant resonance range and the ∆ range. The giant resonance range is dominated by the giant-dipole
resonance (GDR). The photoabsorption cross section is well described by a Lorentzian line which
extends below the particle threshold of the nucleus in the form of narrow lines [145]. The strengths
contained in these individual narrow lines is given through the interplay between the Lorentzian tail
of the GDR and the density of levels. This leads to the consequence that there is a minimum in the
individual strengths close to the particle threshold as indicated in Fig. 2.2. The quasideuteron cross
section starts at the p-n two-particle threshold and extends into the ∆ range above the pion threshold.

Figure 2.2: Schematic view of the nuclear photoabsorption cross section of 208 Pb from the giant reso-
nance to the ∆ resonance range. GDR: giant-dipole resonance, GQR: isovector giant-quadrupole res-
onance, QD: quasideuteron mode, ∆: Delta resonance of nucleons in the nucleus, Born: nonresonant
photoexcitation of nucleons in the nucleus through the Born terms of the photo-pion amplitudes [144].

(ii) Above pion photoproduction threshold in the resonance range of the nucleon we find a broaden-
ing of the resonances increasing with mass number A. This is shown in Fig. 2.3 where the normalized
photoabsorption cross section for the proton is compared with those of the three lightest nuclei. The
solid line represents the normalized photoabsorption cross section of complex nuclei (A ∼ > 4). In this
universal curve there is no indication left of a D13 (1520) resonance. Irrespective of this, there seems
to be a conservation of the integrated photoabsorption strength when going from free to bound nucle-
ons [147].

Figure 2.3: Total photoabsorption cross sections per nucleon. Data are shown for 1 H (black circles), 2 H
(gray circles), 3 He (open circles) and 4 He (grey squares) [146]. The solid line represents the universal
curve for complex nuclei.

(iii) In the asymptotic range of the nucleon photoabsorption cross section (beyond 1.5 - 2 GeV)
the scaling of the photoabsorption cross section with mass number A observed through the universal
curve of Fig. 2.3 is no longer valid. Instead we find a proportionality of the photoabsorption cross
section with Aα with α lying somewhere between 1 and 2/3, i.e. between a proportionality to the
nuclear volume and the area of the nuclear cross-section. This change in proportionality is due to
the fact that the interaction of the intermediate-state vector mesons with the nucleus is of very short
range. This property of the photoabsorption cross section is called shadowing which is illustrated by
Fig. 2.4.

7
Figure 2.4: Fits to the total nuclear photoabsorption cross sections per nucleon as an example of
shadowing in the asymptotic region [148].

Further properties of giant resonances:

Giant resonances may be viewed as nuclear transitions in the shell model from occupied states below
the Fermi level to unoccupied states above it. For a given multipolarity these transitions occur
coherently, which may be visualized in terms of a collective model [149]. A further aspect is the
strong interaction between nucleons. This leads to the need of replacing particle-hole excitations by
quasiparticle-quasihole excitations, where the latter may be described in terms of the Fermi liquid
theory of finite nuclei. This aspect will be covered in section 7.
As contributions to Compton scattering, the giant resonances can be characterized by their elec-
tromagnetic multipolarity. In this way a definite dependence on the scattering angle is assigned to
each resonance. The most prominent contribution to giant resonances stems from the isovector giant-
dipole resonance (GDR) which is seen in all nuclei including 4 He. In heavy nuclei this giant dipole
resonance has successfully been parameterized in terms of one Lorentzian line in case of spherical
nuclei, or in terms of the superposition of two Lorentzian lines in case of deformed nuclei [150, 151].
Axially deformed means prolate deformation as far as we know. There seems to be only one case of
an investigation of an oblate deformed nucleus [115], viz. 127 I. However, in this case it turned out that
the giant resonance did not behave like a collective motion with respect to two independent axes.
Most of the available information on photoabsorption cross sections of medium-mass to heavy
nuclei has been obtained by photo-neutron experiments [152]. For light to medium-mass nuclei pho-
toabsorption cross sections in the GR and QD ranges have been measured by photon attenuation
experiments [84], because for these nuclei photoneutron experiments do not lead to information on
the total absorption cross section due to sizable contributions from proton channels.
It is interesting to note that the Lorentzian shape of the GDR extends below the particle threshold
where we find individual nuclear levels instead of a continuum. Some details concerning the properties
of strengths of individual nuclear levels has already been discussed in connection with Fig. 2.2. In order
to prove these expected properties, rather sophisticated methods had to be applied [145]. A series
of narrow γ lines (∆E ≈ 10 eV) produced via the neutron capture reaction was used to investigate
nuclear resonance fluorescence due to a random overlap of photons with nuclear levels. Using statistical
methods it was possible to get insight into the properties of E1 excitations in the tail region of giant
resonances. It may be of interest to note that the same statistical ansatz applied to the giant-resonance
tail also explains the observed distributions of electric-dipole strengths in nuclei [153].
Figure 2.5 gives a schematic view of the giant-resonance structure of 208 Pb for the first three
multipoles, viz. E1, M1 and E2 [95]. Each multipole is split up into an isoscalar and an isovector
component. For the E1 multipole the isoscalar component is identical with Thomson scattering of the
whole nucleus and therefore does not show up as a separate excitation mode.

Figure 2.5: Schematic representation of giant resonance multipoles for 208 Pb. (E1)1 : isovector giant
electric-dipole resonance (GDR), (M1)0 : isoscalar giant magnetic-dipole resonance, (M1)1 : isovector
giant magnetic-dipole resonance,(E2)0 : isoscalar giant electric-quadrupole resonance, (E2)1 : isovector
giant electric-quadrupole resonance [95].

To summarize, the GR degree of freedom of the nucleus corresponds to the complete set of one-
quasiparticle one-quasihole states in a shell model. This allows us to introduce a scattering amplitude
TGR for the giant-resonance degree of freedom which separately is gauge invariant, if the non-resonant
(seagull) contributions are treated properly.

8
 Further properties of quasideuterons:
Quasideuterons (QD) are proton-neutron pairs correlated through the exchange of charged mesons.
Therefore, QD are the smallest cluster states in nuclei which may show up in the interaction with
photons. There are several approaches to a description of QD excitations of which we mention the
approaches of Levinger [154], Gottfried [155] and Laget [156]. From the point of view of a proper
treatment of degrees of freedom and sum rules, the Laget model is the most appropriate one , because
it relates the total QD strength to exchange currents via the relation
N Z ex
σQD = L σ , (2.1)
A D
where σQD is the quasideuteron cross section, σD ex the exchange part of the deuteron photoabsorption

cross section, Z and N are the charge number and the neutron number, respectively, of the nucleus
and L is a parameter. This means that the total classical or non-mesonic (Thomas-Reiche-Kuhn)
dipole strength is contained in the giant resonances as is also suggested by shell model considerations.
Then meson exchange currents provide us with an enhancement of the giant resonance strength and
— in addition — with the QD strength. Beyond this favorable property the Laget formula has not
to be taken too literally because considerable differences are expected between deuterons and quasi-
deuterons.
For the analysis of Compton scattering data it is most convenient to use a Lorentzian representation
as proposed by Ziegler [15].
Similar to the giant-resonance (GR) degree of freedom the (QD) degree of freedom may be related
to a complete set of isovector 2-particle 2-hole states, allowing us to introduce a separately gauge
invariant amplitude TQD , again with the prerequisite that this amplitude is properly accompanied by
non-resonant (seagull) contributions.
Further properties of the ∆ range of the nucleus:
Figure 2.6 shows the photoabsorption cross section of a complex nucleus split up into different con-
tributing components. The two nonresonant components are due to the quasideuteron effect and due
to the Born term of the photo-pion amplitude. The ∆ resonance contribution is split up into two
components. The high energy part with its maximum around 350 MeV corresponds to real-meson
photoproduction whereas the low energy part with its maximum around 260 MeV to virtual-meson
photoproduction.

Figure 2.6: Total nuclear photoabsorption cross sections per nucleon. Data are shown for nuclei
between 4 He and 238 U [148]. In addition the result of a model calculation for the universal photoab-
sorption curve is shown [157]. The left curve extending underneath the ∆ resonance is the QD cross
section. The ∆-resonance cross section is partitioned into two curves, into a non-mesonic part (left
resonant curve) and a mesonic part (right resonant curve). The nonresonant curve in the ∆ range
corresponds to the Born term of meson photoproduction.

This latter phenomenon is well known from photodisintegration experiments on the deuteron where
a bump-like structure is observed in the non-mesonic cross section at energies above meson threshold.
This bump-like structure occurs through excitation of the ∆ resonance with subsequent two-nucleon
disintegration of the deuteron. In recent experimental investigations on 4 He [158] with the subsequent
theoretical investigations [157] on complex nuclei it was shown that — after a ∆ excitation — there is
a branching of strength into the non-mesonic channel and into the mesonic channel with a conservation
of strength as compared to the free nucleon. This consideration shows that the resonance-like structure
in the non-mesonic pn channel is not part of the QD degree of freedom but of the nucleon internal
degrees of freedom which will be described by the amplitude TN . Such a decomposition of the nuclear
photoabsorption cross section at ∆-resonance energies into a mesonic and a non-mesonic channel is
very well understood due to its manifestation in nucleon knock-out reactions [159–164].

9
2.3 Summary on the degrees of freedom of the nucleus

The most clear-cut distinction between external and internal degrees of freedom of nucleons in a
nucleus is contained in the Compton scattering amplitudes of the individual nucleon, where the Born
terms T Born [101] correspond to the external degree of freedom and the non-Born terms T non−Born to
the internal ones. The leading Born term is given by the Thomson scattering amplitude of a point-like
particle. This term is independent of the photon energy. There are other contributions to the Born
terms which simultaneously depend on the charge and the anomalous magnetic moment of the nucleon
and on the photon energy ω. These latter terms do not contribute to the nuclear scattering amplitude
in the GR energy region but are in principle observable in the QD energy range, though the effects
are very small (see section 5).
To summarize, we have to discriminate between nuclear degrees of freedom (A), which may be giant
resonance states (GR) or quasideuteron states (QD), and nucleon (N) degrees of freedom, which may
be resonant nucleon states or asymptotic nucleon states. The correspondence between the nucleon-
structure independent and nucleon-structure dependent terms of the nucleon and nuclear Compton
scattering amplitudes is illustrated by Eqs. (2.2) and (2.3)

T nucleon = T Born + T non−Born (2.2)

m m m
Ttot = TA + TN (2.3)

with
TA = TGR + TQD . (2.4)
The first terms in Eqs. (2.2) and (2.3) are similar in the respect that they both show contributions
which are connected to low-energy effective degrees of freedom of the object involved.
As an overview, the different degrees of freedom and their main characteristics are summarized in
Table 2.1. From the discussion given above it is clear that the resonant peak at the high-energy end
of the non-mesonic γ AX → A−2 Y + p + n cross section is not part of the QD degree of freedom but
belongs to the nucleon-internal (N ) degree of freedom of the nucleus.

Table 2.1: Excitation modes and degrees of freedom of the nucleus

excitation mode corresponding amplitude excitation mode nuclear objects
of the nucleus of the nucleon of the nucleon and phenomena
giant Born terms quasi-p quasi-h states
resonances (GR) − in an average potential
quasi deuteron Born terms proton-neutron pairs
excitations (QD) − correlated by charged
meson exchange
scaling range non-Born amplitude meson-isobar baryon resonances
σ∝A in the resonance range structure and mesons in
(N ) of the nucleon the nuclear medium
shadowing range non-Born amplitude vector dominance in-medium vector
σ ∝ Aα , α < 1 in the asymptotic range and t-channel exchanges dominance & direct
(N ) & direct γ-quark couplings γ-quark couplings

10
2.4 The role of Compton scattering in investigations of the degrees of freedom of
the nucleus

At first sight Compton scattering simply appears to be one further reaction channel among many
others which — in addition — is very difficult to measure. This view would lead to the conclusion
that Compton scattering is not capable of providing any new information in addition to that obtained
from other reaction channels. This, of course, is not the case. On the contrary, Compton scattering is
a reaction channel with unique properties as will be outlined in the following.
Differing from photoparticle physics, Compton scattering is concerned with the first step of the
excitation process only, since the same matrix element hn|Hint |0i of the first-step transition from
the ground state |0i to the excited state |ni enters into the absorption and the emission processes.
This favorable feature of Compton scattering makes it possible to study the photoabsorption process
without the disturbing effects of subsequent processes, taking place between the initial photoabsorption
and the final particle emission. For this reason it is possible to discuss the photoabsorption processes
in terms of the initial photoabsorption process only, and forget about all the other processes. This
property leads to a smaller model dependence of Compton scattering than observed in photoparticle
experiments. The next favorable property of Compton scattering is its unambiguous relation to
the total photoabsorption cross section via the optical theorem and the forward-angle dispersion
relations. These relations fix the Compton differential cross section in the forward direction so that
additional properties of the photoexcitation mechanisms and additional genuine two-photon properties
(e.g. seagull amplitudes and electromagnetic polarizabilities) may be studied through measurements
of the angular distribution of Compton scattered photons.

11
3 Compton scattering below pion threshold: general aspects

At present no quantitative consistent description of nuclear Compton scattering below pion threshold
exists which is based on first principles. Over the years a variety of phenomenological models has
been developed, which provide insight into the physical mechanisms responsible for contributions to
the scattering amplitude. Many important results were obtained by applying dispersion relations
to the Compton amplitude (see e.g. [165, 166]). For those parts of the amplitude, which are not
easily accessible by dispersion theory, microscopic calculations have been attempted, e.g. for distinct
multipole excitations [167], for mesonic effects [70,80,82,83,168] and for the contribution from nucleon
excitations [102]. In this way nuclear Compton scattering can contribute to the understanding of such
subnuclear aspects as mesonic currents between the nuclear constituents and virtual excitations of
the nucleons. In the last few years several important pieces of information have been extracted from
the experimental data. At low energies the relative strengths of electromagnetic multipoles were
analyzed [13, 16, 19, 25, 33, 64, 95, 169] for comparison with predictions from multipole sum rules. The
interesting question, whether the electromagnetic polarizabilities of the nucleon inside the nucleus
essentially differ from those of the free nucleon, has been theoretically addressed [80, 82, 104, 166, 168]
and experimentally studied with good accuracy [39, 43, 48, 170].
In the process of nuclear Compton scattering the exchange of mesons between nucleons, which
represents the nucleon-nucleon interaction, leads to specific observable phenomena in the whole energy
region considered here. This can be understood from the fact that mesons, themselves, may be
charged and can, therefore, interact directly with photons. Best known among these phenomena is
the modification of the Thomas-Reiche-Kuhn (TRK) sum rule [67], i.e. the appearance of the so-
called enhancement constant κ. Originally, the quantity κ was calculated [67] to be about 0.4 and
this value was not in an obvious disagreement with the experimental data then available. These
data [150, 151] were mainly obtained for heavy nuclei and extended up to energies of about 30 MeV,
thus including mainly the giant-resonance region. Later on it was shown that a considerable amount
of photoabsorption strength is located above the giant resonance region [84, 152]. This strength is
due to the excitation of quasideuterons, i.e. of clusters consisting of proton-neutron pairs correlated
through the exchange of charged mesons. Due to this, the enhancement of giant-multipole strength
is increased to κ ∼ 1.0 in total. A wide variety of model calculations has been carried out for the
total enhancement constant κ as well as for the different contributions to it, which are theoretically
or experimentally accessible [87, 171, 172], see also [126]. Parts of κ could be related to parameters
of Fermi liquid theory and to the notion of quasi-particle masses [96]. A compilation of results can
be found in [173]. The contributions to κ have also been studied in a diagrammatic form [80, 83],
similar to the approach described in section 9. In addition, mesonic exchange currents can imitate
a modification of the nucleon polarizabilities [168]. This contribution may be subtracted in order to
single out a change of the bound nucleon’s polarizabilities from its free values. The advantage of this
subtraction is related to the fact that meson-exchange contributions are not a local effect and are
accompanied with two-body form factors. The latter ones are distinctly different from the one-body
form factor which presumably accompanies all single-nucleon contributions, including polarizabilities
of the bound nucleon. For a more detail discussion see sections 5 and 9.
The effect of mesonic exchange currents on the different electromagnetic properties of nuclei has
been the subject of a large number of research articles, books and review papers. The reviews by
Riska [174] and Arenhövel [121], as well as the book by Ericson and Weise [173] give an excellent
introduction to this field. The description of mesonic exchange effects strongly depends on the physical
process considered. In this review only the scattering of real photons is discussed. As no complete
underlying theory of mesonic effects in this process exists, most of the results are model-dependent.
Nevertheless, it is possible to obtain a qualitative description of various features of the Compton
amplitude within the approaches, which will be discussed in this section.

12
3.1 The photon-nucleus scattering amplitude in second-order perturbation theory

Similar to the case of atomic physics [2] the discussion of nuclear photon scattering may start from
second-order perturbation theory. But dissimilar to atomic physics this approach does not lead to
predictions which directly may be compared with experiments because of the complexity of the phys-
ical situation. Nevertheless second-order perturbation theory is an essential building block for the
construction of the general phenomenological theory. In principle, there is no limitation of this ap-
proach with respect to the degrees of freedom taken into account. However, in the present section we
restrict the application of this method to the energy region of giant resonances (GR). Other effects,
like the quasideuteron (QD) contribution and the internal excitation of nucleons, will be accounted
for by alternative approaches.
The amplitude for the scattering of a photon γ by a nucleus A in its ground state γ + A → γ ′ +
A′ to the order e2 (we set h̄ = c = 1; e is the charge of the proton and e2 = 1/137 is the fine-structure
constant) reads
Tf i = (2π)4 δ(4) (pi + k1 − pf − k2 )V ǫµ1 ǫν∗
2 (Tµν + Sµν ) (3.1)
where V is the total (infinite) volume, T µν is given by

(2π)4 δ(4) (pi + k1 − pf − k2 )V ǫµ1 ǫν∗

2 Tµν
Z
= −2πδ(E0 + ω1 − Ef − ω2 )ǫµ1 ǫν∗
2 dx dyeik1 ·x e−ik2 ·y
X  hf |jν (y)|nihn|jµ (x)|ii hf |jµ (x)|nihn|jν (y)|ii

× + (3.2)
n E0 − En + ω1 + i0 E0 − En − ω2 + i0

and S µν is the seagull amplitude [68, 126] which provides the gauge invariance for the amplitude Tf i .
Therefore, one has
k1µ (Tµν + Sµν ) = (Tµν + Sµν ) k2ν = 0. (3.3)
Representations of Sµν will be discussed below. In Eq. (3.2) ǫµ1 and ǫν2 denote the polarization vector
of the incoming and outgoing photon, respectively. In the following we will use the Coulomb gauge,
in which the photon polarization vector is given by ǫµi = (0, ǫi ) with ǫi · ki = 0. Furthermore, for the
sake of simplicity we consider the case of linearly polarized photons and, therefore, omit the complex
conjugation for the polarization vector ǫ2 . The four-momenta of the incoming and outgoing photon
are denoted by k1 = (ω1 , k1 ) and k2 = (ω2 , k2 ), respectively, while the corresponding quantities for the
nucleus are pi and pf . The function j(x) is the electromagnetic current associated with the interaction
of the photon with the nucleus, as will be discussed a little later on. The summation in (3.2) is
performed over a complete set of eigenstates |ni of the nuclear Hamiltonian with eigenvalues En . The
state |ii of the energy E0 stands for the nuclear ground state and |f i for the final nuclear excited state
in the laboratory system. In the case of Compton scattering initial and final states are equal to each
other except for the total kinetic energy of the nucleus. In the laboratory frame the photon energies
ω1 and ω2 are related via the Compton formula
ω1
ω2 = (3.4)
ω1
1+ (1 − cos θ)
AM
where M is the nucleon mass, A again is the mass number of the nucleus and θ is the scattering angle.
The seagull amplitude S µν is identically equal to zero in the case of a relativistic description of
the nucleus and the absence of exchange currents (or velocity dependent potentials in more general
terms [68, 126]). This case has extensively been investigated for Compton scattering by atoms where
electrons are described as Dirac particles in a self-consistent central potential [2]. These investigations
have provided us with an insight into the general properties of the amplitude given in Eq. (3.2).

13
 In the following, a non-vanishing seagull amplitude enters into the scattering amplitude for two
reasons. First, we pass to the nonrelativistic form of the second-order perturbation theory. Then the
negative-energy intermediate states in Eq. (3.2) give rise to a seagull amplitude B(ω, θ). The other
reason are exchange currents which are a consequence of the isovector nucleon-nucleon interaction. In
this case gauge invariance may be restored by introducing the mesonic seagull amplitude S(ω, θ) =
′
εµ ε∗ν S µν . The appearance of a seagull-like contribution B in a nonrelativistic description of photon-
nucleus scattering can easily be seen when the photon interaction is introduced in the nonrelativistic
nuclear Hamiltonian via minimal substitution.

3.2 The nonrelativistic photon-nucleus interaction and the concept of currents

In a nonrelativistic description [81, 121, 130, 175, 176] the interaction between a nucleon and an elec-
tromagnetic field is given by the Hamiltonian
A
" !
X 1 (i) e e2
Hint = (1 + τ3 ) eA0 (ri ) − [ pi · A(ri ) + A(ri ) · pi ] + A(ri ) · A(ri )
i=1
2 2M 2M
#
eh̄ 1 1
 
(i) (i)
− µp (1 + τ3 ) + µn (1 − τ3 ) σ (i) · (∇ × A) , (3.5)
2M 2 2
(i)
where σ(i) /2, τ3 /2, pi and ri are the spin operator, the third component of the isospin operator, the
three-momentum and the coordinate vector of the ith nucleon, respectively. The quantities µp and µn
are the magnetic moments of the nucleons.
The term quadratic in A(r) is related to the kinetic seagull amplitude B(ω, θ). All terms linear in
A(r) contribute to the one-nucleon electromagnetic current. The terms depending on the magnetic
moments give rise to spin currents which are responsible for magnetic spin-flip transitions in a nucleus.
With the help of the usual relation
δHint
jµ (r) = (3.6)
δAµ (r) Aν =0
one obtains the following representation of the one-body contribution to the nuclear charge density
ρ(r) and current j(r):

[1] X 1 + τ (i)
ρ[1] (r) ≡ 3
X
ρi (r) = e δ(r − ri )
i i
2
X 1 + τ (i)  pi pi

[1] 3
j (r) = e δ(r − ri ) + δ(r − ri )
i
2 2M 2M
(i) (i)
!
e X 1 + τ3 1 − τ3
− µp + µn σ i × ∇ δ(r − ri ) (3.7)
2M i 2 2

The exchange of charged mesons between proton and neutron leads to an additional contribution
to the electromagnetic current. The total electromagnetic current fulfills a continuity equation, which
can be decomposed into a relation for the one-body part,
h i
∇ · j[1] (r) = −i H0 , ρ[1] (r) , (3.8)

and a similar expression for the two-body current

(ex)
∇ · j[2] (r) = −i[ , ρ[1] (r)].
X
Vij (3.9)
i<j

14
 (ex)
Here H0 stands for the kinetic and central part of the nuclear Hamiltonian H, while Vij is the
two-nucleon exchange potential. In Eq. (3.9) it was taken into account that in the limit of point-like,
static nucleons the two-body charge density ρ[2] vanishes since ρ[2] (r) ∝ 1/M . The vanishing of ρ[2]
in the static limit is known as Siegert’s theorem [85]. From ρ[1] (r) one obtains a representation of the
operator D for the total dipole moment
Z X 1 + τ (i)
D= dr ρ[1] (r) r = e 3
ri . (3.10)
i
2
(ex) (ex)
Using this expression, Eq. (3.9) and representing the exchange potential as Vij = τ (i) · τ (j) vij ,
(ex)
where vij does not depend on the isospin variables, one finds the following form of the total two-body
current Z X 
[2] (ex) (ex)
dr j[2] (r) = i[ τ (i) × τ (j)
X
J = Vij , D] = −e (ri − rj )vij . (3.11)
3
i<j i<j

The largest contribution to the two-body current comes from the π and ρ meson exchange parts
(ex)
of the two-nucleon potential Vij , which in momentum space are given by

(π) f 2 (i) (j) (σ i · q)(σ j · q)

Vij (q) = − (τ · τ ) ,
m2π q2 + m2π
(ρ) fρ2 (σ i × q) · (σ j × q)
Vij (q) = − 2 (τ (i) · τ (j) ) . (3.12)
mρ q2 + m2ρ
Here mπ and mρ is the pion and ρ-meson mass, respectively, f 2 /4π ≃ 0.08 and fρ2 /4π ∼ 4.5 − 5.5.
It is convenient to introduce the two-body current in momentum space via
1X dk1 dk2
Z Z  
[2]
j[2] (r) = exp ik1 · (ri − r) + ik2 · (rj − r) jij (k1 , k2 ). (3.13)
2 i6=j (2π)3 (2π)3

The simplest way to arrive at an explicit form of the contribution of pion exchange to the current
[2]
jij (k1 , k2 ) is a direct calculation of the Feynman diagrams shown in Figure 9.1 (cf. section 9). In these
diagrams the vertices are obtained via a minimal substitution in the corresponding Hamiltonians. As
a result one finds [174]:
[2] f 2  (i) (j)
  σ (σ · k )
i j 2 σj (σi · k1 )
jij (k1 , k2 ) = −ie 2 τ × τ 2 − +
mπ 3 2
mπ + k2 m2π + k21
(σi · k1 ) (σ j · k2 )

(k2 − k1 ) 2 . (3.14)
mπ + k21 m2π + k22
Similar expressions can be obtained for the contribution of ρ-meson exchange [174].
In a nonrelativistic second-order perturbation theory the resonance amplitude R(ω, θ) is given by
the contribution to ǫµ1 ǫν2 Tµν of only positive-energy intermediate states in the r.h.s. of Eq. (3.2). This
corresponds to using a nonrelativistic Hamiltonian as a starting point.
In the nonrelativistic approach gauge invariance is provided by introducing a non-resonant (seagull)
amplitude, which is non-zero even in the absence of mesonic exchange currents. This part of the seagull
amplitude corresponds to the contribution of negative-energy intermediate states in Eq. (3.2) in the
nonrelativistic approximation. This so-called kinetic seagull amplitude can be written as
1 XZ [1] [1]
B(ω, θ) = − ǫ1 · ǫ2 hf | dx dy eik1 ·x e−ik2 ·y ρi (x − ri ) ρi (y − ri ) |ii
M j
(j)
e2 X1+τ
3 Ze2
=− ǫ1 · ǫ2 hf | exp [i(k1 − k2 ) · rj ] |ii = − ǫ1 · ǫ2 F1 (q), (3.15)
M j
2 M

15
where q = k1 − k2 and F1 (q) is the form factor describing the distribution of protons inside the
nucleus.
This representation of the kinetic seagull amplitude is frequently used in predictions of nuclear
photon scattering. Using (3.15) to represent the negative-energy states in (3.2) certainly is only an
approximation. However, below meson photoproduction threshold the possible uncertainty is only of
the order of a few percent as discussed in [95]. Furthermore, the possibility of using a modified form
factor which incorporates binding effects has been discussed in [166]. This modified form is given
by [2]
(mod) 1 X 1 + τ (j) M
3
F1 (q) = hf | exp [iq · rj ] |ii (3.16)
Z j
2 E − V (rj )

where E is the total energy of the nucleon bound in the potential V (r) which is assumed to be the
4th component of a Lorentz vector. The modified form factor has been introduced in connection with
atomic Rayleigh scattering [177] and later was investigated and explained by Brown et al. [178]. The
predictions of Brown et al. [178] concerning the modified form factor were experimentally confirmed
by Schumacher [179], who showed that atomic Rayleigh scattering data require the application of the
modified form factor at least for the helicity non-flip amplitude.
In the hypothetical case of structureless nucleons and the absence of exchange currents the cross
section for nuclear photoabsorption would contain only giant resonances and, hence, go to zero at
about 30 to 40 MeV. The amplitude RGR (ω, θ) + B(ω, θ) would provide a complete description of
nuclear photon scattering. Furthermore, the integral over the total photoabsorption cross section
would fulfill the usual Thomas-Reiche-Kuhn (TRK) sum rule without any enhancement. Relativistic
corrections, which could in principle destroy this relation, are negligibly small in the giant resonance
region. If we keep the nucleons structureless but take into account meson exchange currents the giant
resonance amplitude RGR is modified. Furthermore, an additional contribution RQD to the resonance
amplitude appears, the so-called quasi-deuteron scattering, which will be discussed in section 5. This
new amplitude RGR + RQD + B is no longer gauge invariant by itself and a counter term S(ω, θ),
which is usually called the mesonic seagull amplitude, is needed to restore gauge invariance. The total
amplitude TA corresponding to all nuclear degrees of freedom is then given by

TA = RGR + RQD + B + S = RGR + RQD + Stot , (3.17)

where Stot is the sum of the kinetic and mesonic seagull amplitudes. The contribution TN of nucleon-
internal degrees of freedom to the nuclear Compton amplitude, which is not included in (3.17), will
be discussed in sections 5 and 6. As will be seen, below pion threshold the main contribution to this
amplitude is due to the electric and magnetic polarizabilities of the bound nucleon.
The microscopic consideration of meson exchange has lead to explicit calculations of the amplitude
S. Such calculations have been carried out by Alberico and Molinari [75] based on expressions derived
by Christillin [68, 71, 72] and Friar [69], as well as by Hütt and Milstein [168, 193]. In the static limit
the mesonic seagull amplitude at ω=0 is of the following form:
in o
S̃(ω, θ) = hf | [ǫ1 · D , ǫ2 · J[2] ] + [ǫ2 · D , ǫ1 · J[2] ] |ii
ω=0 2
2e2 X  (j) (k) (j) (k)

(ex)
= ǫ1 · ǫ2 hf | τ+ τ− + τ− τ+ (rj − rk )2 vjk |ii , (3.18)
3 j<k

(j) (j) (j)

where τ± = 1/2 (τ1 ± i τ2 ) are the isospin raising and lowering operators. Here and in the following
we use the symbol S̃(ω, θ) to denote a static approximation of S(ω, θ). Due to the isospin structure
of the matrix elements in Eq. (3.18) only proton-neutron pairs contribute to the mesonic seagull

16
amplitude. Taking into account the effect of retardation, but still neglecting any other dependence on
the photon energy ω, which is connected with the ω- and k-dependence of the mesonic propagators,
one has
2e2 X  (j) (k) (j) (k)

(ex)
S̃(ω, θ) = ǫ1 · ǫ2 hf | τ+ τ− + τ− τ+ (rj − rk )2 vjk exp [−iq · (rj + rk )/2] |ii (3.19)
3 j<k

It is convenient to write this limit of the mesonic seagull amplitude in the form
N Z e2
S̃(ω, θ) = −ǫ1 · ǫ2 κF2 (q) (3.20)
A M
where κ is the enhancement constant and F2 (q) is the two-body form factor, which characterizes the
distribution of correlated proton-neutron pairs inside the nucleus. The function F2 is normalized in
such a way that F2 (0)=1. A first attempt to qualitatively discuss two-body or exchange form factors
within a microscopic model has been made by Alberico and Molinari [75] for the case of 208 Pb. In
section 9 the construction of such form factors is described on more general grounds within a modified
Fermi gas. To this end the full mesonic seagull amplitude can be represented as a convolution of a
spin-isospin correlation function with matrix elements arising from amputated diagrams for meson
exchange.

3.3 Low-energy behavior of the scattering amplitude and dispersion relation

In this section we discuss another view towards a decomposition of the Compton scattering amplitude
into a sum of resonant and non-resonant parts. While in the previous section the consideration was
based on second-order perturbation theory and the requirement of gauge invariance, here our starting
point is a dispersion relation for the scattering amplitude in forward direction. At the end of the
last section the nuclear Compton amplitude TA presented itself as a sum of three contributions (3.17)
provided by different physical mechanisms (see also [127, 166, 170]): the collective nuclear excitations
(giant resonances), RGR (ω, θ), the scattering by quasi-deuteron clusters, RQD (ω, θ), and the so-called
seagull amplitude, Stot (ω, θ). The physical background of this separation is the following: Via optical
theorem and a subtracted dispersion relation the scattering amplitude in the forward direction is up to
an additive constant determined by the total photoabsorption cross section. This cross section contains
resonant structures, which at different energies correspond to the two different excitation mechanisms
(cf. Fig. 2.2) which we call giant-resonance (GR) and quasideuteron (QD), as was discussed in section
2.
An experimental separation of GR and QD excitations is relatively easy for heavy nuclei where the
giant-dipole resonance (GDR) cross section containing by far most of the total GR strength is given
by a Lorentzian line in case of spherical nuclei as discussed in section 2. In case of the validity of a
Lorentzian shape rather precise information on the integrated strength in the giant resonance may be
obtained because this Lorentzian shape may be used to separate the GDR cross section from the QD
cross section.
On the other hand also the shape of the QD cross section [154–156] is known either in the form of the
Levinger parameterization [154], the Laget parameterization [156] or a (QD) Lorentzian representation
(cf. sections 2 and 5). Then the shape of the QD cross section extrapolated to above the pion threshold
may be used to separate the QD excitation from the nucleon internal excitation. In practice this is
achieved with sufficient precision by adjusting a Lorentzian curve to the QD cross section below pion
threshold.
In this sense the nuclear photoabsorption cross section σA (ω) below pion threshold and slightly
above can experimentally be separated into two parts,
σA (ω) = σGR (ω) + σQD (ω).

17
This serves as a means to construct the resonance parts of the scattering amplitude RGR and RQD as
Z∞
ω2 σGR (ω ′ ) dω ′
Re (RGR (ω, 0) − RGR (0, 0)) = 2 P (3.21)
2π ω ′2 − ω 2
0

and
ω
σGR (ω).
ImRGR (ω, 0) = (3.22)
4π
The same is valid for RQD . This widely used procedure corresponds to the application to the once-
subtracted forward-angle dispersion relation and of the optical theorem to partial cross sections. This
procedure is legitimate because the GR and QD excitations correspond to separate degrees of freedom.
The application of dispersion relations to the total nuclear Compton cross section dates back to
the early work of Gell-Mann, Goldberger and Thirring [165]. There it was shown that for the total
scattering amplitude Ttot (ω, θ), which — in our language — includes all degrees of freedom of the
nucleus a subtracted dispersion relation may be formulated:

ω2 ∞ σtot (ω ′ ) ′
Z
Re (Ttot (ω, 0) − Ttot (0, 0)) = P dω , (3.23)
2π 2 0 ω′2 − ω2
where Ttot (0, 0) = −Z 2 e2 /AM . For convenience in the further discussion we introduce a “free-nucleon
photoabsorption cross section” via
f ree
AσN (ω) = Zσp (ω) + N σn (ω), (3.24)

where σp (ω) and σn (ω) are the free proton and neutron photoabsorption cross sections, respectively.
The cross section of Eq. (3.24) defines a “free-nucleon scattering amplitude” for the ensemble of
nucleons in a nucleus through
f ree ′
ω2 ∞ AσN (ω ) ′
  Z
Re TNf ree (ω, 0) − TNf ree (0, 0) = P 2 dω (3.25)
2π 2 0 ω − ω2
′

with TNf ree (0, 0) = −Ze2 /M . Similarly, one may think of the difference between the total photoab-
sorption cross section σtot and the one σA , where only nuclear degrees of freedom contribute, as a
bound-nucleon cross section:
bound 1
σN (ω) = (σtot (ω) − σA (ω)). (3.26)
A
From (3.23), (3.25) and (3.26) we obtain

N Z e2 ω2 ∞ σA (ω ′ )
  Z
Re Ttot (ω, 0) − TNf ree (ω, 0) = + 2P dω ′
A M 2π
 0 ω′2 − ω2 
bound (ω ′ ) f ree ′
ω2
Z ∞ A σN − σN (ω )
+ P ′2
dω ′ (3.27)
2π 2 0 ω − ω2
which is the basis of our further conclusions.
For ω → ∞ Eq. (3.27) provides us with the Gell-Mann-Goldberger-Thirring (GGT) sum rule1 , viz.
∞ N Z e2
Z  
σA (ω)dω = 2π 2 + 2π 2 Re Ttot (∞, 0) − TNf ree (∞, 0)
0 A M
Z ∞
f ree bound
+ A(σN (ω) − σN (ω))dω. (3.28)
0
1
In the original work the nuclear (A) and nucleon internal (N) d.o.f. were simply separated through a cut at pion
threshold.

18
In the original paper [165] it was assumed that the asymptotic term in Eq. (3.28) is equal to zero. Then
Eq. (3.28) has a very attractive interpretation by stating that the enhancement of strength contained in
the integral over the nuclear photoabsorption cross section σA (ω) is equal to the reduction of strength
in the nucleon photoabsorption cross section due to nuclear binding. Later on it was realized [180,181]
that the basic assumption of GGT about the properties of the asymptotic term in (3.28) is not correct
because of vector dominance phenomena. Due to this fact the GGT sum rule needs refinements. The
refinement discussed by Ahrens et al. [182] fixes the dispersion integral in the second line of (3.27)
at ω = 314 MeV instead of ω → ∞ thus making use of zero crossings of the real parts of scattering
amplitudes. Although in this way reasonable numbers are obtained for the enhancement constant κ
the interesting physical content of the original GGT sum rule [165] is not preserved. Furthermore,
there seem to be no indications for a reduction of photoabsorption strength above pion threshold
which might show up as an enhancement of photoabsorption strength below pion threshold [147].
Some new aspects of the GGT-sum rule may be obtained by showing that this sum rule is not a
property of the photoabsorption cross section but rather a property of the mesonic seagull-amplitude
S(ω, θ). The arguments leading to this conclusion are as follows. The total seagull amplitude Stot ,
which is the sum of the kinetic and mesonic seagull amplitudes, has an imaginary part only above
pion threshold. From Eq. (3.27), together with (3.21) one obtains a dispersion representation of the
mesonic seagull amplitude in the form [80, 166]

ω2 ∞ bound (ω ′ )
AσN
Z
ReS(ω, 0) − S(0, 0) + ReTN (ω, 0) = 2 P dω ′ , (3.29)
2π mπ ω ′2 − ω 2
where on the l.h.s. the energy is restricted to the range below meson threshold where the amplitudes are
real. This relation has the following interpretation. The difference S(ω, 0) − S(0, 0) is the “dynamic”
part of the mesonic seagull amplitude at zero angle, i.e. that part which is energy dependent in the
forward direction, corresponding to meson exchanges beyond the static limit of the meson exchange
potential. In the order ω 2 the l.h.s. of Eq. (3.29) corresponds to the electromagnetic polarizabilities of
the nucleon and their modification due to binding. From (3.29) we are led to the supposition that the
quantity σNbound (ω) should consist of two different parts one, σ ′ bound (ω), related to T and an other,
N N
(S)
∆σ (ω), related to S(ω, 0)− S(0, 0). Below meson photoproduction threshold the amplitude TN may
be expressed in terms of the in-medium electromagnetic polarizabilities α̃N and β̃N (see section 5.2
for more details) via
 
TN (ω, θ) = A α̃N gE1 (θ) + β̃N gM 1 (θ) ω 2 F1 (q) + O(ω 4 ). (3.30)

Then the supposition implies that only the part σN′ bound (ω) is responsible for the genuine in-medium

electromagnetic polarizabilities α̃N and β̃N whereas the rest, ∆σ (S) (ω), leads to the relations

ω2 ∞ −∆σ (S) (ω ′ )
Z
ReS(ω, 0) − S(0, 0) = P dω ′
2π 2 mπ ω′2 − ω2
ω
ImS(ω, 0) = − ∆σ (S) (ω). (3.31)
4π
The sign of ∆σ (S) (ω) in (3.31) has been chosen negative in order to make this quantity itself positive.
The proof of this statement is given by Eq. (3.33) where the positive right side of the equation is
consistent with a negative signs in (3.31). In the case that

lim S(ω, 0) = 0 (3.32)

ω→∞

which appears to us plausible, we are lead to a sum rule in the following form
∞ N Z e2
Z
∆σ (S) (ω ′ )dω ′ = 2π 2 κ (3.33)
mπ A M

19
where use has been made of Eq. (3.20). In a formal sense the sum rule (3.33) is exactly identical
to what Gell-Mann Goldberger and Thirring [165] had in mind, but the physical interpretation is
different in two essential points:
(i) The sum rule (3.33) does not describe a property of the photoabsorption cross section but rather
a property of the mesonic seagull amplitude S(ω, θ). This is an important difference, because due
to this, ∆σ (S) (ω) may not be interpreted in terms of a reduction of photoabsorption strength of the
individual nucleons through a transfer to energies below π-threshold by some mechanism. Instead,
∆σ (S) (ω) has to be understood as being due to an additional two-body effect.
(ii) The condition for the validity of the sum rule (3.33) is not the equality of absorption cross-sections
for the free and bound nucleon in the high-energy limit, but instead is given by the property (3.32) of
the mesonic seagull amplitude which should already be fulfilled in the resonance region of the nucleon
photoabsorption cross-section. This point has been cleared up by Schumacher et al. [166]. A favorable
consequence of this finding is that the sum rule (3.33) does not interfere with the shadowing effect as
suspected in [180–182].
In addition to the fulfillment of a dispersion relation, the main constraint on the nuclear Compton
amplitude is the low-energy theorem [69, 183, 184]. It states that the Compton scattering amplitude
at ω = 0 is equal to the (coherent) Thomson limit

Z 2 e2
TA (0, θ) = − ǫ1 · ǫ2 . (3.34)
AM
Let us represent the low-energy limit of the resonance part of the Compton amplitude in a form which
is convenient for the further discussion:
e2 ZN
RGR (0, θ) = ǫ1 · ǫ2 (1 + κGR ), (3.35)
M A
where κGR is a constant which, as it will become clear later, is the essential parameter characterizing
all phenomena connected with the enhancement of sum rules due to mesonic effects in the giant
resonance region. Taking into account the fact that RGR (ω, 0) vanishes as ω tends to infinity, with
the help of the dispersion relation (3.21) one has
Z∞
1 N Z e2
σGR (ω)dω = RGR (0, 0) = (1 + κGR ). (3.36)
2π 2 A M
0

In the limit of ω = 0, one could expect that only the electric dipole operator contributes to
scattering and that only unretarded quantities should appear. Therefore, one might think that the
unretarded E1 resonance amplitude
◦ e2 ZN ◦
RGDR (0, θ) = ǫ1 · ǫ2 (1 + κGDR ), (3.37)
M A
◦
which introduces the unretarded enhancement parameter κGDR given by the unretarded electric-dipole
◦
photoabsorption cross section σ GDR (again GDR stands for “giant dipole resonance”),
Z∞
1 ◦ ◦ N Z e2 ◦
σ GDR (ω)dω = RGDR (0, 0) = (1 + κGDR ), (3.38)
2π 2 A M
0

coincides with the dispersion amplitude (3.35). The appropriate equation

◦
RGR (0, θ) = RGDR (0, θ) (3.39)

20
and its consequence
◦
κGR = κGDR (3.40)
make the essence of Gerasimov’s argument [185], which is one of the most frequently cited (and
critisized [186–190]) properties of the multipole contributions to giant resonance excitations. It states
that the corrections due to retardation and the effects of multipoles higher than electric dipole cancel
in the integrated total photoabsorption cross section. Gerasimov’s argument is not precisely valid in
nonrelativistic models because they do not maintain the strict validity of dispersion relations [186–188].
Even in relativistic models which obey the constraints of dispersion relations (see [186, 188, 189] and
section 10), Eq. (3.39) does not hold as an identity (see section 10.4). Nevertheless, expected deviations
are rather small, as will be discussed in some more detail in section 4.4. Data analyses given in section
◦
4 supports the assumption that the difference between κGR and κGDR can be neglected in practical
applications. In the following discussion in section 3 we disregard the difference between κGR and
◦
κGDR .
Returning to the remaining low-energy contributions to the amplitude TA we now define the
quasideuteron amplitude at zero energies, RQD (0, θ), in a similar way to Eq. (3.36) as

e2 ZN QD
RQD (0, θ) = ǫ1 · ǫ2 κ . (3.41)
M A
Therefore, one has
Z∞
1 N Z e2 QD
σQD (ω)dω = κ . (3.42)
2π 2 A M
0

From comparing (3.34) with (3.35) and (3.41) one obtains

Ze2 N
 
Stot (0, θ) = −ǫ1 · ǫ2 1+ κ (3.43)
M A

for the low-energy limit of the seagull amplitude, where κ = κGR + κQD .
Early attempts [68,71,72,74, 75, 83] to understand the term proportional to κ in (3.43) were based
on the idea that this term may be interpreted as Thomson scattering by charged pions in the nucleus.
This effect cannot be seen in electron scattering experiments, because this is a one-photon process
which averages over the positive and negative charges of the pions. Thomson scattering on the other
hand is related to the squares of pion charges and, therefore, there is no cancellation. In consequence
it should be possible give a number for the pion excess per nucleon which was derived [83] to be
nπ ≈ 0.07. However, this idea dating back to a work of Christillin and Rosa-Clot [72] implies a special
interpretation of Siegert’s theorem [85] in the sense that there is a number density of pions without
a charge density. In some sense this may be in line with Siegert’s theorem because there are positive
and negative pions so that the two charges may cancel to zero. By following Siegert’s arguments
literally also number densities should be excluded. This latter point of view has been substantiated
by Weyrauch and Arenhövel [78] who realized that pion charges only occur as “transition charges”.
In other words, the term proportional to κ does not appear due to a modification of the number of
particles in the nucleus but due to a modification of the electromagnetic current.

3.4 Multipole expansion of the scattering amplitude for giant resonances

Via a dispersion relation and the optical theorem the resonance structure of the Compton scattering
amplitude in forward direction is completely determined by nuclear photoabsorption. However, in
order to obtain the angular dependence of the scattering amplitude, it is necessary to consider the

21
multipolarity of the resonances. The multipole-dependent transition operators are obtained by ex-
panding the photon field A into multipole components. For this purpose let us consider a circularly
polarized photon with a polarization vector ξ µ equal to
1
ξ ±1 = ∓ √ (ex ± iey ) (3.44)
2
where ex , ey are two orthogonal unit vectors in the plane perpendicular to the photon momentum
k, which is directed along the axis z, k = ωez , and µ = ±1 denotes the helicity of the photon. The
multipole decomposition of the photon wave function for a fixed helicity µ is then given by
∞ q
ξ µ ei k·r = iL
X
2π(2L + 1) {µALµ (r, M ) + ALµ (r, E)}. (3.45)
L=1

Here, at any angular momentum projection µ, |µ| ≤ L,

ALµ (r, M ) = jL (kr) YLLµ (r̂) ,

1
ALµ (r, E) = ∇ × (jL (kr) YLLµ (r̂)) (3.46)
ωs s
L L+1
= i jL+1 (kr)YL L+1 µ −i jL−1 (kr)YL L−1 µ
2L + 1 2L + 1
are the vector potentials for the magnetic and electric multipoles, respectively, jL (kr) is the spherical
Bessel function. The vector spherical harmonic YJlM (r̂) is related to the usual spherical harmonics
Ylm (r̂) via
JM
X
YJlM (r̂) = C1µ,lm Ylm (r̂) ξ µ
mµ

with ξ 0 = ez and with CjJM

1 m1 ,j2 m2
denoting the Clebsch-Gordan coefficient for the coupling of angular
momenta j1 and j2 to J. The quantity r̂ is the unit vector in the direction of r.
Choosing the axis z along the momentum k1 of the initial photon in Compton scattering, we can
use the above expressions (with k = k1 ) to write the multipole transition operators as
X 
(µ)
QλL = ξ ∗µ · ǫ1 QλL , (3.47)
µ

where Z
(µ)
QλL = (2π)1/2 (2L + 1)1/2 iL dx j(r) · ALµ (r, λ), λ = E, M. (3.48)

Strictly speaking, such a multipole decomposition makes a sense in the center-of-mass of the nucleus.
In a more general situation the operators (3.48) carry both convectional (center-of-mass) and intrinsic
electromagnetic currents. Since the nucleus enters to the initial, final and intermediate states in the
direct and crossed term of the Compton scattering amplitude with different velocities, the convectional
current cannot be completely excluded. However, for low energies and heavy nuclei which we only
consider here such complications are irrelevant, and we can safely assume in most situations that the
transition operators always transform the ground state of the nucleus, |0i, into an excited state |νi
with ν 6= 0.
In Eq. (3.48) each operator contains all orders in ω starting from ω L for the magnetic multipoles and
from ω L−1 for the electric ones. When in Eqs. (3.46), (3.48) the substitution jL (ωr) → (ωr)L /(2L+1)!!
is made, one obtains the corresponding transition operators in the long-wavelength limit which we call
◦ (µ) ◦
unretarded and denote by QλL and QλL . In particular,
◦
Z
QE1 = dx j(r) · ǫ1 = i[H, ǫ1 · D]. (3.49)

22
The amplitude resulting from the unretarded operators will be called “unretarded”, while the exact
operators are understood to lead to the “retarded” amplitude. If the wavelength of the incoming
photon is of the same order as the nuclear radius, it is not effective to expand the photon plane wave
with respect to ωr since many powers of ωr and many multipoles contribute. At these photon energies
the effect of retardation is essential. The problem of retardation is one of the main topics of our review
and will be discussed in sections 4 and 10.
For the final photon which is not directed along the axis z, similar expressions can be used with a
rotation R : k1 → k2 applied to the polarizations ξ µ and multipoles ALµ (r, λ):
X 
(µ) R
Q̃R
λL = ξR ∗
µ · ǫ2 QλL , (3.50)
µ

where Z
(µ) R 1/2 1/2 L
QλL = (2π) (2L + 1) i dx j(r) · AR
Lµ (r, λ). (3.51)

The rotated and unrotated operators are related by ordinary D-functions of Wigner,
(µ) R X (µ′ )
L
QλL = QλL Dµµ′ (R). (3.52)
µ′

Following the general idea that it is possible to represent the total scattering amplitude as a sum,
where each contribution corresponds to a different physical mechanism of excitation, we now consider
the scattering by giant resonances only, together with the corresponding counterparts coming from
the seagull amplitude. Therefore, TGR = RGR + B + SGR with SGR = S(κ → κGR ) and S from Eq.
(3.20). One then arrives at the following decomposition of the giant resonance scattering amplitude:

Ze2 N Z e2 GR
TGR (ω, θ) = −ǫ1 · ǫ2 F1 (q) − ǫ1 · ǫ2 κ F2 (q)
M ( A M )
|hν|Q |0i2| |hν|Q |0i2|
λL λL
gλL (θ)
XX
− + ,(3.53)
λL ν
E0 − EνλL + ω + iΓλL ν /2 E0 − EνλL − ω − iΓλLν /2

where unrotated transition operators are used (with any polarization which is irrelevant) and all
the angular and polarization dependence of the resonance amplitude is translated into the angular
functions gλL (θ) which incorporate the D-function in (3.52) and the factors of µ in (3.45) for each
photon. If both photons are circularly polarized and have the helicities µ1 = ±1 and µ2 = ±1, then
h i∗ h i∗
gEL (θ) = DµL1 µ2 (R) , gM L (θ) = µ1 µ2 DµL1 µ2 (R) . (3.54)

Since we consider Compton scattering, where the final and initial internal nuclear states are equal,
and restrict ourselves to spin-0 nuclei, it follows from angular momentum and parity conservation that
only products of multipole components with the same λL contribute to Eq. (3.53). For simplicity,
we have skipped recoil corrections in the energy denominator, which are negligibly small in the giant
resonance region. We also neglected the additional ω-dependent terms in the kinetic seagull amplitude
(relativistic corrections, which are related to negative-energy intermediate states) and in the mesonic
seagull amplitude (modifications of the electromagnetic polarizabilities). Such corrections are small in
the giant resonance region. The separation of the scattering amplitude into multipoles and its appli-
cation to giant resonances has been discussed by Arenhövel, Danos and Greiner [118]. If the transition
operators are used in the long-wavelength approximation and, at the same time, the dependence of
the resonance denominators on the photon energy is neglected, one observes a strict cancellation for
each electric multipole EL 6= E1 between the contributions to the resonance amplitude and those
coming from the multipole expansion of the seagull amplitude. This general property of scattering

23
amplitudes is a consequence of low-energy theorems [69] and may be traced back to low-energy ex-
pansions of spherical Bessel functions. The terms from the electric dipole give rise to the Thomson
limit, in accordance with the low-energy theorem.
The interpretation of Eq. (3.53) in terms of individual nuclear levels is inappropriate in the nuclear
continuum because of the formation of giant resonances [191], where the strength of a large number
of individual nuclear states n (including the continuum) is concentrated into one GR nuclear state ν.
That is why we use in Eq. (3.53) the sum over νλL rather than a sum over n standing in the first-
principle Eq. (3.2). In contrast to the continuum states n, the collective states ν have finite widths
ΓλL
ν which give an imaginary part to the resonance amplitude RGR as is also prescribed by Eq. (3.2).
Note that the presence of the widths in the crossed term in (3.53) is not supported by Eq. (3.2), in
which the crossed term has no imaginary part at positive energies ω at all. However, the Lorentzian
form of (3.53) is a rather popular approximation often used to describe dispersion effects in dissipative
systems like an oscillator with a friction. The form given in Eq. (3.53) could arise if one uses the
widths ΓλL 3
ν depending on the photon energy, like, for example, the radiative widths Γγ ∼ (h̄ω) [192],
and naively extrapolates them to negative ω (where the widths, strictly speaking, must be zero). A
mathematical advantage of the Lorentzian approximation to partial resonance amplitudes is that the
Lorentzian function
1 1 1 1
 
RL (ω) = 2 2
=− − , (3.55)
ωr − ω − iωΓ0 2ω0 ω − ω0 + iΓ0 /2 ω + ω0 + iΓ0 /2
q
with ω0 = ωr2 − Γ20 /4, has no singularities in the upper semiplane Im ω > 0 of the complex energy
and thus obeys the causality condition. It exactly satisfies the dispersion relation

2 ∞ ω ′ dω ′
Z
ReRL (ω) = P ImRL (ω ′ ) ,
π 0 ω′ 2 − ω2
what makes its use especially convenient in dispersion calculations. A disadvantage is that the crossing
symmetry RL (ω) = RL (−ω) is not fulfilled for the imaginary part of RL . However, this does not create
problems as long as the Lorentzian approximation for absorption cross sections is used for positive
ω only. On the experimental side it is well known that heavy nuclei show GDR photoabsorption
cross-sections in the form of Lorentzian curves for spherical nuclei. These Lorentzian photoabsorption
cross sections can be generated in a formal way from the sum in Eq. (3.53) by keeping only one level.
The coupling of giant resonance modes and collective nuclear surface modes has been discussed by
Arenhövel et al. [118, 119]. In this case the terms in (3.53) are the components into which the giant
resonances are fragmented due to the coupling of collective surface and volume modes. In this case a
3/2
relation between widths Γν and energies Eν of the components of the form Γν ∼ Eν is assumed.
In a collective model, giant resonances which we are dealing with here are due to collective motions
of nucleons. In case of the giant-dipole resonance (GDR) this is the isovector electric-dipole motion
of protons against neutrons for which the dynamic model of Steinwedel and Jenssen [149] has been
developed. In a shell model this same GDR phenomenon corresponds to 1p−1h excitations of nucleons
between neighboring oscillator shells (1h̄ω transitions) (cf. e.g. [126]). The most elegant way to
incorporate two-body currents into this one-body picture is provided by the Fermi liquid theory. In
this model the strong interaction between nucleons leads to an effective mass M ∗ of quasiparticles
and a modification of the isovector current. Because of the special importance of this aspect for
the understanding of subnuclear degrees of freedom in isovector giant-resonances we devote a special
section to this problem (section 7).
Other important GR modes are the isoscalar and isovector giant-quadrupole resonances. Enhance-
ment effects on these GR modes have not been discussed except for a first approach in one of our

24
previous works [95]. Following the essential ideas of this approach we will discuss the enhancement of
higher multipoles in the present work on more general grounds.
As pointed out above, giant resonances may be considered as nuclear states which themselves are
coupled to a very large number of strongly overlapping individual nuclear states. These GR states
may be viewed as coherent superpositions of transitions of a given multipolarity between occupied and
unoccupied shell model states located below and above the Fermi level, respectively. In the following
we slightly rearrange the resonance part of the GR amplitude, Eq. (3.53). We introduce the resonance
energy (EνλL )r of the GR states instead of EνλL via (EνλL )2r = (EνλL − E0 )2 + (ΓλL 2
ν ) /4. Furthermore, we
separate the ω-independent contribution coming from the E1 transition. Using Eqs. (3.35) and (3.49),
we may transform the resonance part in such a way that the low-energy limit appears explicitly:

e2 N Z
RGR (ω, θ) = (1 + κGR ) gE1 (θ)
M A
(EνλL )2r − ω 2 + iωΓλL
ν
2 (EνλL − E0 ) ΘλL λL
XX
+ 2 ν (ω) g (θ), (3.56)
λL ν ((EνλL )2r − ω2) + (ωΓλL
ν )
2

where

ΘλL
ν = |hν| QλL |0i|2 , λL 6= E1 (3.57)
2
◦ ω2
ΘE1
ν = |hν| QE1 |0i|2 − hν| QE1 |0i + |hν|D|0i|2
3
are energy-dependent coefficients and the operator D is the dipole moment.
In the following the subscript r of (EνλL )r will be omitted for convenience. Due to the transfor-
mation of the electric dipole contribution, the lowest multipoles in the sum, E1, M 1 and E2, are all
proportional to ω 2 in the low-energy limit. Close to resonance energies a good approximation for the
angular dependence of the amplitude RGR can be obtained by multiplying each term in the sum with
respect to λL in Eq. (3.56) by the multipole angular functions gλL (θ). Using (3.54) and (8.7), one can
obtain that

gEL (θ) = AL (z) ǫ1 · ǫ2 + BL (z) s1 · s2 , gM L (θ) = AL (z) s1 · s2 + BL (z) ǫ1 · ǫ2 , (3.58)

where si = k̂i × ǫi , k̂i = ki /ω, z = k̂1 · k̂2 and

2 2
AL (z) = (P ′ (z) + zPL′′ (z)), BL (z) = − P ′′ (z). (3.59)
L(L + 1) L L(L + 1) L

Here PL (z) are Legendre polynomials. For the lowest multipoles the functions gλL (θ) are given by:

gE1 (θ) = ǫ1 · ǫ2
gM 1 (θ) = s1 · s2
gE2 (θ) = 2(ǫ1 · ǫ2 ) (k̂1 · k̂2 ) − s1 · s2
gM 2 (θ) = 2(s1 · s2 ) (k̂1 · k̂2 ) − ǫ1 · ǫ2
1 5
  2  
gE3 (θ) = 15 k̂1 · k̂2 − 1 ǫ1 · ǫ2 − k̂1 · k̂2 (s1 · s2 ) . (3.60)
4 2
In order to obtain the angular distribution at energies outside the resonance region, it is necessary
to use further properties of the scattering amplitude, which are closely related to the problem of
retardation. In other words, it is necessary to discuss the energy dependence of the quantity ΘλL ν
entering into (3.56). This energy dependence is quite different at small energies, in the center of the
GR region and in the high-energy limit.

25
3.5 Multipole and isospin decomposition of the seagull amplitude

For the further consideration it is useful to understand the isospin structure of the seagull amplitude,
as well as its multipole decomposition. The underlying idea has been adopted from the hydrodynamic
model of giant resonances, introduced by Steinwedel and Jensen [149] which describes the isovector
motion of protons against neutrons in a giant-dipole resonance excitation. In this model the relevant
dynamic quantity is the square of the difference of proton and neutron densities (ρp − ρn )2 leading to
an increase in potential energy in case of a local deviation from the average. In a similar way also the
in-phase, or isoscalar, variation of the local density leads to giant resonance modes with the isoscalar
giant-dipole mode being equivalent to Thomson scattering of the whole nucleus.
In accordance with Eq. (3.18) the total seagull amplitude Stot in the static limit is equal to
in o
Stot = hf | [ǫ1 · D , ǫ2 · (J[1] + J[2] )] + [ǫ2 · D , ǫ1 · (J[1] + J[2] )] |ii (3.61)
2
with
(i)
[1] e X 1 + τ3
J = pi
M i 2

and J[2] from Eq. (3.11). In order to single out the isoscalar and isovector contributions to Eq. (3.61)
we represent the dipole moment operator in the following form:
X 1 + τ (i) (i) (i)
!
3 N X 1 + τ3 Z X 1 − τ3 1 X
D=e ri = e ri − e ri + eZ ri . (3.62)
i
2 A i 2 A i 2 A i

The term in round brackets in Eq. (3.62) is the intrinsic dipole moment operator. It depends on
relative coordinates of protons and neutrons and hence has an isovector structure. The remaining
term in (3.62) represents the dipole moment of the nucleus as a whole. It does not distinguish
coordinates of protons and neutron and, therefore, is an isoscalar operator.2 It is evident from Eq.
(3.11) that the commutator of J[2] with the isoscalar part of the dipole moment vanishes. Therefore,
exchange currents do not contribute to the isoscalar part of the seagull amplitude. The commutator
of J[1] with the isoscalar part of the dipole operator is nonzero and leads to the isoscalar part of the
seagull amplitude:
(is) Z 2 e2
Stot = − ǫ1 · ǫ2 . (3.63)
A M
The isovector part of the dipole moment has a nonzero commutator with both J[1] and J[2] and, as a
result, the isovector seagull amplitude in the static limit is given by

(iv) ZN e2
Stot = − (1 + κ) ǫ1 · ǫ2 . (3.64)
A M
In Eq. (3.64) we use the quantity κ instead of κGR for the formal reason that the current operators
introduced in Eq. (3.61) do not discriminate between GR and QD excitations. Nevertheless, the
application of the results obtained here may be restricted giant resonances, as will be done in the next
section when discussing GR sum rules.
As soon as retardation corrections are included via the photon plane wave the corresponding form
factors appear in the seagull amplitude with F1 being related to the current J[1] and F2 to J[2] (cf.
Eqs. (3.15) and (3.19)):

(is) Z 2 e2 (iv) ZN e2
Stot = − F1 (q) ǫ1 · ǫ2 , Stot = − (F1 (q) + κ F2 (q)) ǫ1 · ǫ2 . (3.65)
A M A M
2
Using these terms in such a context we ignore that the electric charge of the nucleus itself, Z, has an isovector
component.

26
In order to obtain a multipole decomposition of the seagull amplitude Stot , we introduce one- and
two-body densities ρ1,2 , which are related to the form factors F1,2 via

dq
Z
ρi (r) = Fi (q) exp (iq · r) . (3.66)
(2π)3

Then ∞
X (i)
ǫ1 · ǫ2 Fi (q) = fl (ω)Gl (θ) (3.67)
l=0

with Z ∞
(i)
fl (ω) = 4π drr 2 ρi (r)jl2 (ωr), (3.68)
0

G0 (θ) = gE1 (θ) and, for l > 0,

1 1 1
Gl (θ) = (l + 2)gE(l+1) (θ) + (2l + 1)gM l (θ) + (l − 1)gE(l−1) (θ). (3.69)
2 2 2
Here gλL (θ) is the angular distribution function for the multipole λL, as before. A derivation of the
expansion given in Eqs. (3.67) and (3.69) can be found in appendix C. Substitution of these relations
(is) (iv)
into Eq. (3.65) leads to the multipole decomposition of the seagull amplitude Stot = Stot + Stot :

Ze2 (L + 1)fL−1 (ω) + LfL+1 (ω)

 
λL
(ω) gλL (θ) , EL
X
Stot = Stot Stot (ω) = − , (3.70)
λL
M 2
Ze2 2L + 1 N (2)
 
ML (1)
Stot (ω) =− fL (ω) , fl (ω) = fl (ω) + κf (ω).
M 2 A l
When one uses only the lowest-order terms with respect to ω for each multipole, one obtains expressions
for the unretarded multipole amplitudes entering into Stot :

Ze2 (L + 1) ω 2L−2 D 2L−2 E N D 2L−2 E

 
◦ EL
S tot (ω) =− r + κ r , (3.71)
M 2 [(2L − 1)!!]2 1 A 2
Z∞
Ze2 (2L + 1) ω 2L D 2L E N D 2L E
 
◦ ML D
2l
E
S tot (ω) =− r + κ r , r = 4π dr r 2l+2 ρ1,2 .
M 2 [(2L + 1)!!]2 1 A 2 1,2
0

As was already pointed out all contributions from electric multipoles in Eq. (3.71) except for E1 cancel
the corresponding low-energy expansion of the resonance amplitude. In the case of magnetic multipoles
such cancellations do not appear. This asymmetry between electric and magnetic multipoles becomes
important when we discuss sum rules in section 4. There it will become evident that sum rules in a
strict sense are only possible for electric multipoles.

27
4 Compton scattering by giant resonances

In the present section we discuss the status of Compton scattering by giant resonances. The usual way
of constructing the resonance (positive energy) scattering amplitude of Compton scattering by giant
resonances is given by the Lorentzian representation of photoabsorption cross sections and the appli-
cation of the optical theorem and the once-subtracted dispersion relation. The amplitudes obtained
in this way are valid close to the peaks of the resonances but quite obviously fail in the low-energy
and high-energy limits. These deficiencies can be cured by introducing nonretarded amplitudes which
are valid in the low-energy limit and by introducing retardation formfactors to take into account re-
tardation effects [95, 176]. This procedure is in accordance with the nonrelativistic Hamiltonian but
violates the requirements dispersion relations which may be understood as a violation of causality [165].
When correcting for this effect a satisfactory representation of the resonance (positive energy) ampli-
tude including the effects of retardation is obtained as will be shown in section 10 through a model
calculation. On the basis of this theoretical approach we discuss recent theoretical and experimental
results obtained for the giant resonances.
Before proceeding further it is worth to briefly explain why the second-order perturbation theory,
which results in Eq. (3.56) for the resonance amplitude RGR , is not fully suitable for data analyses
and why dispersion relations are invoked for an alternative construction for RGR . The main practical
problem with a use of Eq. (3.56) is that an evaluation of this equation involves both the spectrum of
excited states and the energy dependences of the transition matrix elements hν|QλL |0i. These matrix
elements are not known with the necessary precision when ωR ≃ 1 (here R is the nuclear radius), i.e.
when the energy ω lies well above the region of the giant dipole and giant quadrupole resonances and
when retardation effects come fully into play. At such energies matrix elements between the ground
state |0i and excited states |νi with L ∼ ωR are no longer suppressed by threshold factors like (ωR)L
or (ωR)L−1 . Therefore, these excited states, being fully negligible in Eq. (3.56) in the low-energy
region ωR ≪ 1, do give an essential contribution to the real part of RGR at high energies. We refer
to section 10 for a model example.
On the other hand, the dispersion relation (3.21) suggests that the high-energy behavior of
RGR (ω, 0) is controlled by photoabsorption in the low energy region at ωR ≪ 1, where the giant
dipole and quadrupole resonances are located, so that the knowledge of details of the high-energy
behavior of the transition matrix elements should be unnecessary. From this reason it is attractive to
try to formulate the resonance amplitude RGR by using dispersion relations and to extend Eq. (3.21)
to nonzero angles. One possible extension based on fixed-t dispersion relations is considered in section
8. As we will see, it can be developed for low t (or small angles) but also runs into trouble when the
> 1, is considered.
region of high t, i.e. −tR2 ∼
In this situation a phenomenological approach was developed [95] which combines the advantages
of the two approaches and is apparently applicable at all energies of practical interest — up to the
pion photoproduction threshold — as confirmed by a model study in section 10. To help with the
understanding of the different steps and to provide a dictionary of the notation used, we briefly
summarize the procedure of finding the partial resonance amplitudes RGR λL . To avoid confusion note

that all the partial amplitudes used in this section do not include the constant RGR (0, 0), Eq. (3.35),
so that all of them vanish at ω = 0.
At the first stage naive once-subtracted dispersion relations for partial amplitudes are postulated
and the corresponding “dispersion” amplitudes R̃GRλL are obtained as a sum of few Lorentzians with

parameters given by the structure of the total photoabsorption cross section. These “dispersion”
amplitudes are not assumed to be identical with the proper amplitudes RGR λL because they have either

a wrong threshold (at ω → 0) or a wrong asymptotic (at ω → ∞) behavior, or both. They are
◦
only used to construct unretarded, RλL λL
GR , and retarded, R̂GR , amplitudes and to fix their scale. It is

28
 ◦
assumed that the (total) retarded amplitude R̂GR is nothing but the full unretarded amplitude RGR
2
multiplied by retardation form factor squared, F(ret) (ω). The matching of the “dispersion” amplitude
R̃GR and the retarded amplitude R̂GR is done in such a way that these two amplitudes yield the same
absorption cross sections at the giant-resonance peaks. The retarded amplitudes have theoretically
sound behavior both at low and high energies as long as one stays on the real axis of complex ω-
plane. Being also correct at the resonance peaks, they are assumed to be in general very close to the
proper amplitudes RGR λL . However, a comparison with the requirement of the forward-angle dispersion

relation reveals a mismatch ∆R(ω) between the total “dispersion” amplitude R̃GR (ω, 0) and the total
retarded amplitude R̂GR (ω, 0). In order to remove this mismatch, the electric dipole amplitude RGRE1

is appropriately corrected with ∆R, and this gives the final resonance amplitude RGR . The whole
construction is chosen to ensure vanishing of the resulting resonance amplitude RGR in the high-energy
limit at any angle.
The described procedure cannot be strictly derived from first principles. Moreover, some features
of thus obtained resonance amplitude RGR are in contrast to those found in relativistic models; among
them is the absence of a q-dependent form factor F (q) in the asymptotics of RGR (ω, θ) (see section 10).
Nevertheless, that is the best what was proposed to describe Compton scattering through the giant
resonance region, and this Ansatz finds a numerical confirmation both experimentally and through
model studies.
Now we proceed with giving details of the described approach.

4.1 The giant resonance amplitude in the Lorentzian representation

The general form of the resonance amplitude for the forward direction given in (3.56) consists of a
Lorentzian part and a coefficient ΘλL
ν , which is energy dependent and is expected to be different at low
energies, in the center of the giant-resonance region and in the high-energy limit, respectively. In the
following we, therefore, will discuss three different regimes of the resonance amplitude corresponding
to the three energy regions.
The great advantage of Compton scattering experiments is the fact that in the forward direction
the amplitude is fixed through the optical theorem, the dispersion relation and the low-energy theorem.
Then the specific information of Compton scattering is contained in the angular distribution of the
differential cross section. For applications this means that the resonance amplitudes as formulated in
Eq. (3.56) has to be replaced by a more practical version. In the vicinity of the giant resonances this
is obtained through the following procedure. We consider the total photoabsorption cross section as
a superposition of Lorentzian lines

(ωΓλL
ν )
2
σ λL (ω) = σνλL
X XX
σ(ω) = 2 (4.1)
λL λL ν [(EνλL )2 − ω 2 ] + ω 2 (ΓλL
ν )
2

where the sum with respect to ν is taken over all Lorentzians contributing to the multipoles λL. The
quantities EνλL , ΓλL
ν and σν
λL denote the resonance energies, widths and peak cross sections of the

Lorentzians, respectively. A resonance amplitude R̃GR , which is applicable in the resonance region
and corresponds to the photoabsorption cross section of (4.1), then has the form

N Z e2
(1 + κGR )gE1 (θ) + R̃λL (ω, θ)
X
R̃GR (ω, θ) = (4.2)
A M λL
N Z e2 ω 2 X λL λL (EνλL )2 − ω 2 + iωΓλL ν
= (1 + κGR )gE1 (θ) + σ Γ gλL (θ).
A M 4π λL,ν ν ν [(EνλL )2 − ω 2 ]2 + ω 2 (ΓλL
ν )2

29
Here and below R̃λL (ω, θ) means the product R̃λL (ω) gλL (θ). In the low and high energy limit a
special discussion of the amplitude’s properties is required.
The energy-dependent multipole amplitudes R̃λL in (4.2) are built in such a way that they satisfy
the optical theorem,
ω λL
ImR̃λL (ω, 0) = σ (ω), (4.3)
4π
and obey the naive subtracted dispersion relations

ω2 ∞ σ λL (ω ′ )
Z
ReR̃λL (ω, 0) = P dω ′ . (4.4)
2π 2 0 (ω ′ )2 − ω 2

Here the subtraction is replaced by zero, what is allowed because the subtractions for all multipoles are
contained in the first term of the r.h.s. of (4.2) and, furthermore, are of electric-dipole multipolarity.
In the high-energy limit ω → ∞, the dispersion resonance amplitudes (4.4) are given by

1
Z ∞ X σνλL ΓλL
ν
ReR̃λL (∞, 0) = − σ λL (ω ′ ) dω ′ = − , (4.5)
2π 2 0 ν 4π

where the explicit form of the partial absorption cross sections, Eq. (4.1), was used. Since the giant
resonance amplitude has to vanish in the high-energy limit, we obtain from (4.2) and (4.5) the GGT
sum rule:
XZ ∞ X π 2
2 NZ e
σ λL (ω ′ )dω ′ = σνλL ΓλL
ν = 2π (1 + κGR ). (4.6)
λL 0 λL,ν
2 A M

A further useful property of the scattering amplitude in the Lorentzian form concerns the second
negative momentum of the integrated cross section. We discuss this property here as a reference for
further applications. One has

ReR̃λL (ω, 0) 1 ∞ σ λL (ω ′ ) ′ 1 λL 1 X σνλL ΓλL

Z
ν
lim = 2 dω ≡ σ −2 = , (4.7)
ω→0 ω2 2π 0 (ω ′ )2 2π 2 4π ν (EνλL )2

where in the last step the explicit (Lorentzian) form of the amplitudes R̃λL has been used again.
It is possible to relate the resonance amplitude, Eq. (4.2), to the forward-direction form, Eq. (3.56),
which yields a representation of the matrix elements ΘλL ν applicable in the resonance region:

ω 2 σνλL ΓλL
ν
ΘλL
ν (ω) = . (4.8)
8π EνλL

Note that in this form of the matrix elements ΘλL

ν all effects of retardation are included. Such effects
are important, when the product of the photon energy ω with the nuclear radius R is of the order of
unity, which is the case in an actual experiment.

4.2 The unretarded resonance amplitude

The dispersion amplitude (4.2) is built in such a way that it vanishes at high energies in the forward
direction. Such vanishing appears as a result of the exact balance between the first term in (4.2) and
the partial amplitudes R̃λL (∞, 0), as is stated by Eq. (4.6). However, this balance is destroyed at angles
θ 6= 0, when the angular coefficients gλL (θ) of different multipoles become different. Accordingly, Eq.
(4.2) does not vanish in the high-energy limit. Another deficiency of Eq. (4.2) which has to be cured is

30
that the multipole amplitudes R̃λL (ω, 0) do not have the proper ω dependence at low energies except
for the lowest multipoles λL = E1, E2, M 1.
◦
As the first step to remove these deficiencies we introduce now an unretarded form RGR of the
giant resonance amplitude RGR , where each multipole transition operator is taken into account only in
its lowest order in ω and manifestly results in the correct low-energy behavior. Our construction is as
follows. In the low-energy limit each multipole (electric and magnetic) in the total nuclear scattering
amplitude TA (ω, θ) is proportional to ω 2L (except for the Thomson E1 amplitude). This property
can be obtained from general considerations based on analytical properties of the amplitude, as will
be seen in section 8. The amplitude RGR (ω, θ) in the form of Eq. (3.56), together with (3.48) teaches
us that in the low-energy limit each electric multipole component is proportional to ω 2L−2 , while the
magnetic multipole starts from ω 2L . As was pointed out above the EL multipole contributions of
the order ω 2L−2 in the seagull amplitude cancel the corresponding terms in the resonance amplitude,
which leads to the correct low-energy behavior for the sum. Taking into account these properties we
may write
◦ N Z e2 ◦ X ◦ λL
RGR (ω, θ) = (1 + κGDR )gE1 (θ) + R (ω, θ) = (4.9)
A M λL
N Z e2 ◦ ω 2 X ◦ λL λL (EνλL )2 − ω 2 + iωΓλL ν
(1 + κGDR )gE1 (θ) + σ Γ ΩλL (ω)gλL (θ),
A M 4π λL,ν ν ν [(EνλL )2 − ω 2 ]2 + ω 2 (ΓλL
ν )2 ν

where

ΩE1
ν (ω) = 1
ΩEL EL 2L−4
ν (ω) = (ω/Eν ) for L ≥ 2
ΩM L
ν (ω) = (ω/EνM L )2L−2 for L ≥ 1. (4.10)
◦
In accordance with Eq. (3.37), the unretarded enhancement parameter κGDR stands in the unretarded
amplitude (4.9). Assuming that the E1 component of Eq. (4.9) vanishes at high energies, we find the
(modified) TRK sum rule
∞ ◦ Xπ◦ N Z e2
Z
◦
σ E1 (ω ′ )dω ′ = E1 σ ν ΓE1
ν = 2π
2
(1 + κGDR ). (4.11)
0 ν 2 A M

Here the corresponding unretarded absorption cross section is

◦ ◦ ◦ (ωΓλL
ν )
2
ΩλL
X X
σ(ω) = σ λL (ω) = σ λL
ν 2 ν (ω). (4.12)
λL λL,ν [(EνλL )2 − ω 2 ] + ω 2 (ΓλL
ν )
2

◦
The difference between this expression for σ(ω) and the corresponding quantity σ(ω) from Eq. (4.1)
◦ λL
lies in the appearance of the ω-dependent factor ΩλLν and in the unretarded peak cross sections σ ν .
◦
A comparison of σ λL (ω) and σ λL (ω) will be given in section 10 in the framework of a specific model.
It should be noted that the energy-dependent factor ΩλL ν becomes effective for electric octupoles,
magnetic quadrupoles and higher multipoles. At energies much larger than the resonance energies
this form of the unretarded multipole cross sections cannot be applied. Nevertheless, in the case
of electric multipoles energy-weighted sum rules can be derived, which serve as a guideline in the
interpretation of experimental data.
◦
The cancellation between the leading terms in the unretarded resonance amplitudes RλL and in
the expansion of the seagull amplitude leads to sum rules for electric multipoles. Since different
mechanisms of excitation are involved, it is possible to formulate such sum rules independently for the

31
isoscalar (is) and the isovector (iv) contributions. Due to the fact that we are restricting ourselves to
giant resonance phenomena, it is necessary to understand which part of the total seagull amplitude
Stot corresponds to this degree of freedom. Clearly, the kinetic seagull amplitude B has to be included.
Therefore, as mesonic effects appear in the isovector part only (cf. Eq. (3.64)), the isoscalar sum rules
can straightforwardly be written down:
◦ EL,(is)
◦ EL,(is) 2 R (ω, 0) Z 2 e2 L+1 D E
σ −(2L−2) = 2π lim = π2 r 2L−2 , L > 1. (4.13)
ω→0 ω 2L−2 AM [(2L − 1)!!]2 1

The r.h.s. of Eq. (4.13) follows from the decomposition of the seagull amplitude, Eq. (3.71), together
with Eq. (3.63). Note that no such sum rule exists for E1, as the observed electric dipole resonance
is an isovector excitation. In the case of isovector sum rules the situation is more complicated. In
section 3.2 we discussed the full mesonic seagull amplitude obtained in the static limit. Clearly, only
a part of this amplitude, previously denoted by SGR , enters into the giant resonance amplitude TGR ,
which we discuss now. However, it is impossible to extract SGR from S without further assumptions,
as long as we basically argue in the framework of the Fermi gas model which cannot discriminate
between GR and QD modes of excitation. Strictly speaking, it is not sufficient to only pass from κ to
◦
κGDR in the amplitude S, but also the dependence on momentum transfer is modified. This means
that the form factor F2GR entering into SGR is different from the two-body form factor F2 appearing
in the amplitude S. Such a modification can only be discussed in a model-dependent way, e.g. by
following the frequently applied (see e.g. [173]) line of thought that the main contribution to SGR
comes from the central part of the exchange potential. We discuss the quantitative consequences of
this assumption in appendix B. Here we denote the normalized density, which corresponds to the form
factor F2GR by ρGR
2 and write the isovector sum rule as
◦ EL,(is)
 
◦
R
(ω, 0) N Z e2 ◦
σ EL,(iv)
−(2L−2) = 2π 2 lim + (1 + κGDR )δL,1  (4.14)
ω→0 ω 2L−2 A M

2 NZ e2 L+1
D EGR 
◦ GDR
E D
2L−2 2L−2
= π r +κ r .
A M [(2L − 1)!!]2 1 2

The label GR given here to the two-body formfactor F2GR , the two-body density ρGR
2 and the two-
D EGR
body averaged radius r 2L−2 indicates that these quantities now are restricted to the GR d.o.f.
2
and, therefore, may be slightly different from the corresponding quantities introduced in section 3.5
where no distiction between the GR and QD d.o.f. has been made. The upper indices (is) and (iv)
for the absorption cross section correspond to restricting the sum with respect to ν in Eq. (4.12)
to isoscalar and isovector giant resonance excitations, respectively. It should be stressed that these
sum rules are valid for the unretarded photoabsorption cross sections and not immediately for the
experimental (retarded) photoabsorption cross sections. Sum rules for the latter case will be discussed
below in section 4.5 and in section 8. A useful relation for the purpose of data analysis as well as for
the following discussion is the sum of Eqs. (4.13) and (4.14) for the case of the electric quadrupole.
One has
π 2 Ze2 D 2 E N ◦ GDR D 2 EGR
 
◦ ◦ E2,(is) ◦ E2,(iv)
σ E2
−2 = σ −2 + σ −2 = r + κ r . (4.15)
3 M 1 A 2

In the particular case L = 1, Eq. (4.14) gives the sum rule (4.11).

4.3 The retardation problem

The principal goal of a theoretical investigation of nuclear Compton scattering is to obtain a consistent
description of the energy and angular dependence of the scattering amplitude in a wide energy region.

32
In the previous section different quantities have been introduced. The “dispersion” amplitudes R̃λL
provide a formulation of Compton scattering in the resonance region only. Since their parameters
are taken from experiment, these amplitudes are retarded quantities. The corresponding unretarded
◦
amplitudes RλL were designed to correctly describe the low-energy Compton scattering, but they
were expressed in terms of unretarded quantities. Therefore, in order to obtain in a wide energy
region (also including high energies) a representation of the Compton scattering amplitude expressed
only via observable quantities, it is necessary to establish a relation between unretarded and retarded
quantities. A formulation of this retardation problem in terms of the Bessel functions entering into the
multipole transition operators has been given in section 3.4 in the following way: In an expansion [194]
of the plane wave for the photon into terms with definite parity and total angular momentum L,
Eq. (3.45), each term contains a spherical Bessel function jL (ωr). An amplitude obtained using
the exact multipole operators is called “retarded”, while in the case where the substitution jL (ωr) →
(ωr)L /(2L+1)!! is made, the resulting amplitude is called “unretarded”. The importance of retardation
effects is different in the different energy regions, i.e. at low energies, in the resonance region and in
the high-energy limit, which for our present purpose can be identified with the pion threshold. Only at
very low energies the parameter of expansion ωR, with the nuclear radius R, is small. In the resonance
region this parameter is still smaller than unity, but no longer negligible, while at high energies one
has ωR ∼ 1.
Let us consider again the representation of the (retarded) scattering amplitude TGR coming from
second-order perturbation theory. In Eq. (3.53) the seagull part of the scattering amplitude includes
the effects of retardation through the form factors. A further important property of (3.53) is that in the
high-energy limit the last term, i.e. the resonance amplitude, vanishes and, therefore, the high-energy
limit of the scattering amplitude is represented by the first two (seagull) terms. Our solution of (i.e.
the Ansatz for) the retardation problem with respect to the resonance amplitude consists of two parts.
◦
We have to fix (i) how the low-energy form RGR (ω, θ) of Eq. (4.9) and the retarded Lorentzian form
R̃GR (ω, θ) of Eq. (4.2) are related to each other, and (ii) through which mechanism the high-energy
form of the resonance amplitude tends to zero. We follow the ideas of our previous work [95] and then
— in later sections — discuss the predictions of fixed-t dispersion relations and calculations within a
relativistic harmonic-oscillator model.
◦
The interpolation between RGR (ω, θ) and R̃GR (ω, θ) starts with the introduction of a “retarded“
resonance amplitude
◦ 2
R̂GR (ω, θ) = RGR (ω, θ) F(ret) (ω). (4.16)

This retarded resonance amplitude R̂GR (ω, θ) reflects our “classical” expectation of the retardation
effect. The quantity F(ret) (ω) is a form factor due to the spatial distribution of quasiparticles. It takes
into account the different phases of the electromagnetic wave with respect to the location of these
quasiparticles. In a classical picture one form factor in Eq. (4.16) appears as a result of calculating the
total force acting on an extended charged object. The second form factor results from the calculation
of the photon radiation off this extended object due to acceleration. This picture is confirmed by
the consideration of the relativistic oscillator model (cf. section 10). As a consequence (4.16) is the
logical form of a retarded amplitude in a classical picture. This type of retardation has already been
discussed by Barashenkov and Kaiser [196] in the case of the nucleon and by Petrun’kin [197] and
Ericson and Hüfner [175] in a quantum-mechanical treatment of Thomson scattering. More recently
it has been derived by Baron et al. [176] in a nonrelativistic model calculation of retardation effects
on the resonance part of the giant dipole resonance.
Strictly speaking, due to the close structural relation between the resonance and seagull parts of
the amplitude, it is not obvious, what type of form factor should be used for F(ret) (ω) in Eq. (4.16).
One may expect that different types of excitations lead to different form factors. For example, while

33
for isoscalar excitations the one-body form factor F1 seems appropriate, isovector excitations should
◦ GDR
2 = (F 2 + κ ◦
be related to an isovector form factor F(iv) 1 (F2GR )2 )/(1 + κGDR ).
◦
In order to fix the coefficients σ λL
ν contained in the low-energy form, Eq. (4.9), of the resonance
◦
amplitude RGR (ω, θ), we use the condition that the retarded amplitude yields through the optical
theorem the correct magnitude of the absorption cross section at the resonance peaks:
λL
Im R̂GR (EνλL , θ) = Im R̃GR
λL
(EνλL , θ). (4.17)

This gives relations

◦ 2 λL λL
σ λL
ν F(ret) (Eν ) = σν (4.18)
◦
between the unretarded σ λL λL
ν and the experimental peak cross sections σν .

Thus defined, the retarded resonance amplitude R̂GR (ω, θ) fulfills one of the most essential con-
ditions we are looking for since it apparently vanishes in the high-energy limit owing to the factor
F(ret) (ω). In spite of this and in spite of its classical plausibility, the resonance amplitude R̂GR (ω, θ) in
the form of (4.16) is not fully acceptable because it violates causality. This can easily be seen if for ex-
ample the proton density is approximated by a Gaussian distribution, i.e. F(ret) (ω) = exp(−ω 2 /(4λ))
where the parameter λ is fixed by demanding that it reproduces the usual mean square nuclear charge
radius [176]. In this particular form one can easily derive that R̂(ω, θ) of Eq. (4.16) tends to infinity
in part of the upper complex-ω half-plane for |ω| → ∞. Consequently, this form of the resonance
amplitude violates causality [165]. Similar difficulties (e.g. additional poles in the upper half-plane
of ω) may appear for other distributions of quasiparticles. However, as it will be seen in section
10 within the relativistic oscillator model, it is possible that such energy-dependent form factors are
compensated by the sum over all intermediate states, if all matrix elements are taken into account
exactly. In this way the correct analytical properties of the total amplitude may in principle be re-
stored. Since it is necessary to approximate these matrix elements in all applications of the general
formalism, a different approach for overcoming the difficulties of an incorrect analytical behavior has
been suggested in [95].
The incorrect analytical behavior of the “retarded” amplitude (4.16) manifests itself in that it does
not obey the GGT dispersion relation at the forward angle. That means that there is a mismatch

∆R(ω) = R̃GR (ω, 0) − R̂GR (ω, 0) (4.19)

between the exact (dispersion) amplitude R̃GR (ω, 0) and the “retarded” amplitude (4.16). It was
proposed in [95] to consider this difference as an electric dipole correction to the “retarded” amplitude
and to finally write the resonance amplitude RGR (ω, θ) including the effects of retardation in the
following form:
RGR (ω, θ) = R̂GR (ω, θ) + ǫ1 · ǫ2 ∆R(ω). (4.20)
By construction, the amplitude (4.20) is in agreement with forward-direction dispersion theory, just
because of the above specific choice of ∆R(ω). Simultaneously, the amplitude (4.20) preserves all
the favorable properties of the retarded amplitude R̂(ω, θ). The high-energy limit of the resonance
amplitude (4.20) is determined by R̃GR (ω, 0) and has the form

RGR (ω, θ) ≈ R̃GR (ω, 0) gE1 (θ)

 
N Z e2 ω 2 X λL λL (EνλL )2 − ω 2 + iωΓλL ν
=  (1 + κGR ) + σ Γ  g E1 (θ)
A M 4π λL,ν ν ν ((EνλL )2 − ω 2 )2 + ω 2 (ΓλL
ν )
2

1 X λL λL  λL 2 
→ 2
σν Γν (Γν ) − (EνλL )2 + iωΓλL
ν gE1 (θ). (4.21)
4πω λL,ν

34
For the last line in Eq. (4.21) the sum rule (4.6) has been used. The Ansatz (4.20) implies that we
attributed an electric dipole characteristic to this high-energy limit.3 In section 8 the properties of
Eq. (4.20) are discussed within the framework of fixed-t dispersion relations.
The adopted procedure is equivalent to using the dispersion relation for the dipole E1 amplitude
(however not with only the E1 cross section but rather with a modified one) and to introducing
threshold (ΩλL
ν ) and retardation (F(ret) ) factors to other “dispersion” multipoles. This is explicitly
shown by another form of Eq. (4.20) which summarizes all different peaces of the amplitude:

N Z e2
RGR (ω, θ) = (1 + κGR )gE1 (θ)
A M
ω 2 X λL λL (EνλL )2 − ω 2 + iωΓλL ν
h i
+ σν Γ ν 2 ΩλL
ν (ω)gλL (θ) , (4.22)
4π λL,ν [(EνλL )2 − ω 2 ] + ω 2 (ΓλL
ν )
2 (ret)

where
h i
" 2
F(ret) (ω)
#
 
ΩλL λL
ν (ω)g (θ) (ret) =g λL
(θ) + ΩλL (ω) −1 gλL (θ) − gE1 (θ) . (4.23)
2
F(ret) (EνλL ) ν
◦
We may note that the unretarded enhancement parameter κGDR standing in the unretarded amplitude
◦
RGR (ω, θ) and in the retarded amplitude R̂GR (ω, θ) does not appear in the final amplitude (4.22).
Instead, only the retarded quantity κGR appears as it should, owing to the constraint of the GGT sum
◦
rule. So, the parameter κGDR cannot be directly determined from Compton scattering data. Given,
◦
however, the absorption cross section and the form factor, one can calculate κGDR through Eq. (4.11)
and compare it with κGR . This will be done in the following sections.
◦
The above-mentioned replacement of the unretarded enhancement parameter κGDR by the retarded
one, κGR , in Eq. (4.22) takes place only, because the correction ∆R(ω) in Eq. (4.20) was attributed
to the E1 multipolarity. Therefore, one can conclude that the electric dipole factor appearing in Eq.
(4.20) and persisting up to the asymptotics (4.21) is obligatory at low energies. At high energies it
might be modified, and this is exactly what is observed in the model considered in section 10, where
an additional form factor appears in the asymptotics of RGR . However, we have to repeat again (see
footnote 3 ) that the same model suggests that such modifications are not important numerically. (See
section 10.5 for more detail.)
The resonance amplitude RGR , Eqs. (4.20) or (4.22), together with the total giant-resonance seagull
contribution
e2  N Z GR GR  E1
B(ω, θ) + SGR (ω, θ) = − Z F1 (q) + κ F2 (q) g (θ), (4.24)
M A
composes the part of the total amplitude TA , Eq. (3.17), related to the giant-resonance degrees of
freedom. The remaining contributions in Eq. (3.17) are discussed in section 5.
The effects of retardation on the giant-multipole amplitudes are illustrated in Fig.4.1. The imag-
inary parts are not noticeably affected by the retardation factor F(ret)2 (ω) whereas the effect on the
real parts is very large, as can be seen from the differences between the solid and the dotted curves.
However, it is important to realize that the damping of the scattering amplitude through F(ret) 2 (ω) is a
physical effect only for multipoles other than electric dipole. For the electric dipole this damping (solid
curve in the upper figure) is an artifact of a nonrelativistic theory which violates the requirements
of dispersion theory. The corresponding correction given by ∆R(ω) of (4.20) not only restores the
3
Within the relativistic model considered in section 10, the asymptotics of the amplitude RGR (ω, θ) is more compli-
cated and contains also a retardation form factor dependent on the momentum transfer q. Numerically, however, such
a form factor is not very important, because it becomes only effective when the asymptotic amplitude is already small
(see section 10.5).

35
nondamped dotted curve of the upper figure but — even more — enhances this amplitude by adding
constructively those parts of the real amplitudes of the higher multipoles which have been damped
2
away by F(ret) (ω). As a consequence, retardation does not lead to a decrease of the real amplitude
as a whole but to a change in its multipolarity. The real amplitudes of the higher multipoles vanish
in the high-energy limit and the electric-dipole amplitude becomes larger by the same amount. The
total resonance amplitude RGR (ω, θ) vanishes in the high-energy linit due to a cancellation of the two
terms on the r.h.s. of the first line of (4.21).

Figure 4.1: Scattering amplitudes for the giant-dipole resonances (upper figure) and isovector giant-
quadrupole resonance (lower figure) of 208 Pb in the forward direction: Im RλL (ω, θ = 0) (dashed),
Re R̃λL (ω, θ = 0) (dotted) and Re R̂λL (ω, θ = 0) (solid).

4.4 Gerasimov’s argument

The developed model for nuclear Compton scattering in the giant-resonance region allows one to
◦
determine the parameter κGDR which is of great importance for understanding the effective mass of
the nucleon in a nucleus. As was emphasized above, the integrated strength of the unretarded E1
absorption cross section is determined through experimental data on photon absorption indirectly,
viz. by virtue of an extrapolation based on the specific form of retardation corrections, Eq. (4.18),
implemented in the model. In the next subsections we will see that such an extrapolation gives a
result very close to that predicted by Gerasimov’s argument [185] which is therefore considered here.
Gerasimov’s argument [185] says that the photoabsorption cross section integrated over all multi-
pole components is equal to the integrated unretarded dipole strength and that, therefore, retardation
corrections pertinent to the E1 absorption of real photons are compensated in the integral by other
multipoles. This argument has been criticized in a number of papers [186–190] (see also section 10) for
not being valid in general. In those papers it was explicitly shown that the alleged compensation does
not happen for a particle bound in a potential (without meson-exchange forces) and that relativistic
corrections destroy Gerasimov’s argument.
It is instructive to see where exactly a flaw is in the Gerasimov’s original derivation. It is in an
illegal combination of the dispersion relation (3.21) and the time-ordered perturbation theory (3.2)
which give seemingly similar but nevertheless different resonance amplitudes when applied in the usual
context of low energies ω ≪ M . The same total scattering amplitude T is represented differently in
the dispersion and perturbation theory:

T (ω, 0) = S disp (ω, 0) + Rdisp (ω, 0) = S p.t. (ω, 0) + Rp.t. (ω, 0) (4.25)

(in this discussion we omit subscripts like GR or GDR and assume that S means the total seagull
amplitude including the kinetic part B; for the sake of clarity, we denote the “dispersion” amplitude
R̃(ω, 0) by Rdisp (ω, 0) here). Considering as an example Compton scattering on a Dirac particle of mass
M bound in a potential and restricting oneself to energies ω ≪ M , one includes all negative-energy
intermediate states in Eq. (3.2) to the (effective) seagull S p.t. , so that Rp.t. is given by positive-energy
components of the total amplitude T :

Rp.t. (ω, 0) = T+ (ω, 0). (4.26)

When the retardation is turned off and only electric dipole transitions are included (this limit corre-
sponds to a formal use of k = 0 but ω 6= 0 in Eq. (3.2)), the corresponding unretarded E1 resonance
◦ ◦
scattering amplitude Rp.t.
E1 (ω, 0) = T +,E1 (ω, 0) vanishes at “high” energies ω ∼ ωm , where ωm is an

36
arbitrary energy which is much smaller than M but much larger than the binding energy of the particle
involved. Therefore the unretarded resonance amplitude satisfies an unsubtracted dispersion relation

◦ 1 ωm ◦ ω ′ 2 dω ′
Z
Re Rp.t.
E1 (ω, 0) = 2P σ E1 (ω ′ ) , (4.27)
2π 0 ω′2 − ω2
which gives
◦ 1
Z ωm ◦
Rp.t. (0, 0) ≡ Rp.t.
E1 (0, 0) = σ E1 (ω) dω. (4.28)
2π 2 0

As a simple analysis shows (see section 10.3 and Eq. (4.35) below), the positive-energy part alone
of the total amplitude T does not satisfy the unsubtracted dispersion relation, because T+ does not
vanish at “high” ω ∼ ωm . Therefore, in order to have the unsubtracted dispersion relation valid for
the “dispersion” resonance amplitude Rdisp ,

1 ωm ω ′ 2 dω ′
Z
Re Rdisp (ω, 0) = P σ(ω ′ ) (4.29)
2π 2 0 ω′2 − ω2
(note that the “dispersion” amplitude was defined through this dispersion integral) and accordingly
to have the relation Z ωm
disp 1
R (0, 0) = 2 σ(ω) dω, (4.30)
2π 0
the amplitude Rdisp has to include, at least partly, negative-energy contributions of T . Since the total
amplitude satisfies the GGT dispersion relation, we conclude that

Rdisp (ω, 0) = T (ω, 0) − T (ωm , 0), (4.31)

where the constant T (ωm , 0) ensures vanishing of the “dispersion” resonance amplitude at “high” en-
ergies. In this equation the amplitude T (ω, 0) includes both positive and negative energy contributions
and therefore Rdisp (ω, 0) does so.4 Comparing Eqs. (4.26) and (4.31), we conclude that their difference
at zero energy,

∆RGerasimov (0, 0) = Rdisp (0, 0) − Rp.t. (0, 0)

= T− (0, 0) − T (ωm , 0) = T− (0, 0) − T− (ωm , 0) − T+ (ωm , 0), (4.32)

is determined by the energy dependence of the negative-energy contribution (if any) and by the
asymptotics of the positive-energy one. Just this difference prevents the integrals (4.28) and (4.30) to
be identical and generally leads to a violation of Gerasimov’s argument.
As an illustration of what happens, let us consider positive and negative energy parts of the
amplitude T (ω, 0) in the case of a free Dirac particle having a momentum p [95]. Such a case is
easy for a treatment, and it is relevant because it shows the behavior of the scattering amplitude at
energies much higher than the binding energy. It is clear from scale arguments that the “high”-energy
(ω ∼ ωm ) asymptotics of the scattering amplitude is determined by a coherent sum of local scattering
contributions found for different points in the potential well with a constant local potential. Therefore
it is sufficient to consider scattering off free particles (with a local mass M (r) = M + Vs (r)) and then
to take a proper average. We can disregard a possible Lorentz-vector potential Vv , because it makes
only local shifts of energies E → E + Vv (r) which do not change any local observables including the
scattering amplitude (cf. [187, 195]).
4
Since the Lorentzian parameterization of the “dispersion” amplitude Rdisp , Eq. (4.2) at θ = 0, is constructed exactly
in a way to ensure its vanishing at high energies, this parameterization is an approximation to Eq. (4.31). Therefore it
intrinsically includes a part of the negative-energy contribution T− .

37
 Considering the case of forward scattering on the free Dirac particle, we have

(α · ǫ2 ) Λ+ (α · ǫ1 ) Λ+
" #
2 + p+k (α · ǫ1 ) p−k (α · ǫ2 )
T+ = e U + U,
Ep+k − Ep − ω Ep−k − Ep + ω
(α · ǫ2 ) Λ− (α · ǫ1 ) Λ−
" #
2 + p+k (α · ǫ1 ) p−k (α · ǫ2 )
T− = e U + U, (4.33)
−Ep+k − Ep − ω −Ep−k − Ep + ω

where U is the Dirac spinor (U + U = 1) and Λ±p are the projectors onto the positive and negative
energy states: !
± 1 α · p + βM q
Λp = 1± , Ep = M 2 + p2 . (4.34)
2 Ep
Keeping only the spin-independent part of the amplitude and the leading relativistic correction with
respect to 1/M , we find in the semirelativistic case of p ≪ M , ω ≪ M :
!
e2 ω2 e2
T+ = (ǫ1 · ǫ2 ) − − (ǫ1 · p) (ǫ2 · p),
M 4M 2 M3
!
e2 p2 Vs ω2 e2
T− = (ǫ1 · ǫ2 ) −1 + + + + (ǫ1 · p) (ǫ2 · p), (4.35)
M 2M 2 M 4M 2 M3

where we have explicitly restored the Lorentz-scalar potential Vs hidden in the particle mass M . Terms
with ω 2 in Eq. (4.35) are quantum corrections ∼ h̄2 which are specific for the spin 1/2 particle having
a magnetic moment eh̄/2M ; they are absent in the spinless case (see section 10.3). Moreover, the ω 2
terms cancel in the sum T = T+ + T− and therefore do not appear in the resonance amplitude (4.31).
The term with the energy Vs +p2 /(2M ) in T− describes a physical effect, which is a modification of the
high-energy behavior of the total amplitude T , in exact accordance with the findings of Goldberger
and Low [187, 195]:
" # !
e2 e2 p2 Vs
T (ω, 0) = −(ǫ1 · ǫ2 ) ≃ (ǫ1 · ǫ2 ) −1 + 2
+ . (4.36)
Ep M →M +Vs
M 2M M

Using Eq. (4.36) for the evaluation of T (ωm , 0) and Eq. (4.35) for the evaluation of T− (0, 0), we find
that the difference (4.32) of the resonance amplitudes and therefore the difference of the GGT and
TRK integrals is determined solely by the term with (ǫ1 · p) (ǫ2 · p) in T− , Eq, (4.35). Specifically,

e2
∆RGerasimov (0, θ) = (ǫ1 · p) (ǫ2 · p). (4.37)
M3
This perfectly agrees with a direct evaluation of the GGT integral done by Friar and Fallieros [187]. See
section 10 and, in particular, Eq. (10.47) for an alternative derivation of the (ǫ1 · p) (ǫ2 · p) component
of T− .
Applying these results to nucleons in the nucleus, taking a sum over Z protons, averaging over
nucleon momenta p and subtracting the center-of-mass contribution (which is assumably unaffected
by relativistic corrections), we conclude that the GGT sum rule including relativistic effects but not
including meson exchange currents is given by
* +!
1 ωm Ze2 1 p2 Z 2 e2
Z
σ(ω) dω = 1− + Vs − , (4.38)
2π 2 0 M M 2M MA

where MA ≃ AM is the mass of the nucleus. The relativistic correction arising in the r.h.s. of (4.38)
cannot be found without further assumptions about a splitting of the mean-field potential into the

38
Lorentz-scalar and Lorentz-vector parts Vs and Vv , respectively. We suggest that the net result might
be very similar to that contained in the last term in (4.38). That is, it might be roughly equivalent to
the replacement of the free nucleon mass M = 938.9 MeV by the mass M = MA /A ≃ 931.2 MeV (as
found for 12 C). If so, this would have only a tiny effect of about 0.8% on the sum rule and would give
◦
a small negative contribution −0.008 to the extracted value of κGR and κGDR .
Similarly, the difference of the GGT and TRK sum rules including relativistic effects but not
including meson exchange currents is given by
* +
1 ωm Ze2 p2
Z
◦
 
σ(ω) − σ E1 (ω) dω = . (4.39)
2π 2 0 M 3M 2
◦
This equation may suggest that the value of the unretarded enhancement parameter κ has to be
smaller than the value of the retarded one, κ, by

◦ A p2
κ−κ≃− h i ∼ −0.05. (4.40)
N 3M 2
Here the Fermi-gas estimate of the nucleon average momentum squared was used at the standard
◦
nuclear density ρ0 = 0.17 fm−3 and A/N was set to be 2. Such a large difference between κ and κ
would, of course, be important. However, within the simple potential model used it is not clear whether
such an expected difference is concentrated in the giant resonance region or in the quasi-deuteron one.
As we will see, the retardation model developed in the previous sections, leads to a close agreement
◦
between the enhancement parameters κGDR and κGR relevant to the giant resonances.

4.5 The effects of enhancement and retardation in electric multipole sum rules

The general phenomenology of giant resonances gives the possibility to predict the integrated strengths
of electric giant multipoles including the effects of retardation and enhancement. As each multipole
component in the photoabsorption cross section may in principle consist of several Lorentzians, it is
convenient for the further discussion to introduce a quantity (F 2 )(λL) which is the squared retardation
form factor at resonance energy averaged over the different multipole components:
◦ ! ◦ !−1
σ λL λL σ λL λL
ν Γν ν Γν
(F 2 )(λL) F 2 (E λL )
X X
= . (4.41)
ν (EνλL )2L−2 (ret) ν ν (EνλL )2L−2

In case of only one Lorentzian line contributing to the multipole λL Eq. (4.41) reduces to (F 2 )(λL) =
2
F(ret) (EνλL ) which will be used in the following applications. In principle it could be necessary to
distinguish between isoscalar and isovector averages of the retardation form factor. As in the case of
the retardation form factor itself we will neglect such differences.
From Eqs. (4.6), (4.11) and (4.18), together with Eq. (4.41) it follows that the relevant sum rule
for the giant-dipole resonance (GDR) is given by
Z∞
π X E1 E1 π X ◦ E1 E1 2
σ GDR (ω) dω = σ Γ = σ Γ F (E E1 )
2 ν ν ν 2 ν ν ν (ret) ν
0
N Z e2 ◦
= 2π 2 (1 + κGDR ) (F 2 )(E1) . (4.42)
A M
In (4.42) σ GDR (ω) is the giant-dipole resonance component of the experimental (retarded) photoab-
sorption cross section and σνE1 , ΓE1 E1
ν and Eν are the Lorentz parameters of the νth component of the

39
giant-dipole resonance, denoting the peak cross section, the width and the peak position, respectively.
The quantity (F 2 )(E1) on the r.h.s. of Eq. (4.42) characterizes the modification of the Thomas-Reiche-
Kuhn (TRK) sum rule due to the effect of retardation.
In the same way one obtains sum rules for the electric giant-quadrupole resonances, following from
Eqs. (4.7), (4.13) and (4.14):
Z∞ (is)
σE2 (ω) πX (is) σνE2 ΓE2 π 2 Z 2 e2 D 2 E
dω = ν
= r (F 2 )(E2) (4.43)
ω2 2 ν (EνλL )2 3 A M 1
0
Z∞ (iv)
σE2 (ω) πX (iv) σνE2 ΓE2 π 2 N Z e2
D E D EGR 
◦ GDR
2
dω = ν
= r 2
+κ r2 (F 2 )(E2) , (4.44)
ω 2 ν (EνλL )2 3 A M 1 2
0

which differ from the similar relations (4.13) and (4.14) by the fact that here retardation has been
taken into account explicitly by the averaged form factor appearing on the r.h.s.
From Eqs. (4.42)–(4.44) it is possible to make predictions for the properties of the three most
important giant resonances, the giant-dipole resonance (GDR), the isovector (IVGQR) and isoscalar
(ISGQR) giant-quadrupole resonances, if we also take into account experimental results, as far as
they are known. In order to clarify the quantitative influence of mesonic currents, retardation and the
relative strength of the different multipoles, it appears useful to introduce the following quantity:
∞  !−1
e2
2 NZ
Z
λL
γ =  σλL (ω) dω  2π , (4.45)
A M
0

which is the total multipole absorption strength given in units of the (unmodified) TRK sum rule.
The cross section σλL (ω) in Eq. (4.45) includes the effects of enhancement and retardation and may
in principle be the result of an experiment. In the case, where only one Lorentzian contributes, Eq.
(4.45) can be written for the three lowest electric multipoles, i.e. GDR, IVGQR and ISGQR, in the
form
◦
 
γ E1 = 2
1 + κGDR F(ret) (EE1 )
1 E2 2 D 2 E
 D EGR 
E2 ◦
γ(iv) = (E(iv) ) r + κGDR r 2 2
F(ret) E2
(E(iv) )
6 1 2

E2 Z D E
γ(is) = (E E2 )2 r 2 F(ret)
2 E2
(E(is) ). (4.46)
6N (is) 1

◦
Replacing in Eq. (4.46) the form factor by 1 and κGDR by 0 one obtains expressions for the integrated
multipole cross sections γ0λL without any effects of mesonic currents (enhancement) and retardation.
These quantities will be called the unmodified integrated cross sections which are theoretical quantities
to be obtained from sum rules not containing enhancement and retardation effects. The inclusion of
◦
the form factor corresponds to taking into account retardation , while κGDR is responsible for mesonic
contributions (enhancement). For the case of 208 Pb these and other quantities are listed in Table
4.1. The widths and peak positions of the giant resonances are taken from the following references:
GDR [198], IVGQR [25] and ISGQR [199]. The unmodified and the modified integrated cross sections
are related to each other through the relation

γ λL = γ0λL × MEF × F(ret)

2
(EλL ), (4.47)
◦
where MEF depends on λL and means a meson enhancement factor (any MEF = 1 when κGDR = 0).
Using this equation and the definitions of γ λL and γ0λL , the quantities MEF may easily be obtained

40
from Eqs. (4.46). For the calculation of F 2 (EλL ) the same prescriptions have been used as in our
1/2
previous paper [95]. For the one-body radius r 2 1 the charge radius as determined by electron
1/2
scattering has been used, i.e. for 208 Pb r 2 1 = 5.51 fm. For a big nucleus like 208 Pb this quantity
should be a good measure for the square-averaged radius defined through the distribution of (pointlike)
1/2
protons, viz. r 2 p and, therefore, also for the isoscalar radius, but differences may be expected [95]
◦ GR ◦
for the isovector radius ( r 2 1 + κGDR r 2 2 )/(1 + κGDR ). The two-body radius entering into this
latter quantity was determined [95] through experimental and theoretical arguments and a numerical
GR 1/2 1/2 ◦
 
value of r2 2 ≡ r2 pn = (4.3±0.8) fm has been obtained. The value of κGDR has been chosen
in such a way that the experimental integrated GDR cross section of 208 Pb, γ E1 =1.38, is reproduced.
This number has been obtained through photo-neutron experiments [198] on 208 Pb. Exactly the same
result is obtained as an average over a larger number of nuclei in the 208 Pb mass range, where the
peak cross sections have partly been redetermined through precise Compton scattering experiments.
This will be described in section 4.6. The importance of the content of Table 4.1 lies in the fact, that a
prediction for the “experimental” enhancement, i.e. the combination of retardation and enhancement
is obtained for the IVGQR which is an unknown quantity otherwise. The mesonic enhancement factor
◦ GR
of MEF = 1.28 calculated from κGDR and the radii r 2 1 and r 2 2 combines with the retardation
2
factor F(ret) E2 ) = 0.88 to give an “experimental” enhancement of MEF × F 2
(E(iv) E2
(ret) (E(iv) ) = 1.13. This
result of 13% is the only information we have on the “experimental” enhancement because experiments
are by far not precise enough to measure this quantity. Details will be given in section 4.7.

Table 4.1: Giant multipoles in 208 Pb: Peak energy EλL , width Γ, unmodified integrated cross section
γ0λL in units of the (unmodified) TRK sum rule, retardation factor F 2 (EλL ), mesonic enhancement
factor MEF and experimental or modified integrated cross section γ λL .

Parameter GDR IVGQR ISGQR

EλL 13.42 MeV 22.5 MeV 10.6 MeV
Γ 4.1 MeV 9.0 MeV 3.0 MeV
γ0λL 1 0.066 0.007
2
F(ret) (EλL ) 0.945 0.88 0.98
MEF 1.46 1.28 1
γ λL 1.38 0.074 0.007

It is interesting to note that the numbers contained in Table 4.1 are in line with the predictions
of Gerasimov’s argument. The sum of the integrated strengths of the three multipoles is 1 + κGR =
◦
1.38 + 0.074 + 0.007 = 1.46 whereas the corresponding non-retarded dipole strength is 1 + κGDR = 1.46
in units of the unmodified TRK sum rule. This certainly is an important confirmation of Gerasimov’s
argument since the two numbers compared here are (almost) completely independent of each other.

4.6 Scaling of giant resonance parameters via Compton scattering

The quantity we wish to determine is the unretarded enhancement constant for the giant-dipole res-
◦
onance, viz. κGDR , which is the universal parameter of giant resonance enhancement phenomena. In
principle this quantity can be determined from photoabsorption experiments. In a first place these are
photoneutron experiments [150, 151, 200] for heavy nuclei and total photoabsorption experiments [84]
for light nuclei. The reason for this difference in methods is that for heavy nuclei the photoneutron
channel largely dominates the photoproton channel and total photoabsorption measurements are in-
accurate because of the e+ e− pair production cross section which is much larger than the nuclear

41
photoabsorption cross section. On the other hand, at low mass numbers the photoproton and the
photoneutron channels are about equal in strength. Since low-energy protons are difficult to detect,
especially for solid targets the total photoabsorption cross section measurements are preferable.
In case of photoneutron experiments there may be one or more neutrons in the final state per
photoabsorption process. Furthermore, the neutrons may be emitted after a fission process. This
means that photoneutron experiments require a precise counting technique for neutrons and a method
to determine the neutron multiplicity. These experiments, therefore, gave an excellent overview of the
properties of giant resonances [150, 151, 200], but suffer from systematic uncertainties because of the
fact that the intensity of the photon beam and the detection efficiency of the neutron detector have to
be known on an absolute scale. As a consequence, the integrated cross sections measured in different
laboratories showed large deviations from each other [150, 151]. On the other hand, the relative cross
sections measured in one experiment for different photon energies, i.e. the shape of the cross-section
curve, turned out to be much more reliable. Therefore, general scaling factors of the cross sections
were required in order to improve on the integrated cross sections.
The amplitude for forward-angle Compton scattering is related to the total photoabsorption cross
section via optical theorem and dispersion relation. Therefore, the total photoabsorption cross sec-
tion may also be determined via Compton scattering provided the angular distribution of Compton
scattered photons is well enough known so that an extrapolation to zero angle becomes possible and
provided there is enough information to carry out the dispersion integral. Given these premises,
Compton scattering provides a rather precise information because there are no principal difficulties
in arriving at absolute numbers. The reason for this is that the projectiles and the reaction products
are photons of (almost) the same energy. This makes it possible to measure the rates of incident pho-
tons and scattered photons with the same detector and thus to avoid the determination of detection
efficiencies on an absolute scale. A further advantage of Compton scattering is that it is possible to
very largely reduce the effects of e+ e− pair production. This latter process enters into the Comp-
ton scattering amplitude via Delbrück [5] scattering. Since Delbrück scattering takes place in the
Coulomb field surrounding the nucleus this process is much more peaked to the forward direction
than nuclear Compton scattering and, therefore, can be avoided except for small scattering angles or
photon energies below 10 MeV where sizable contributions of Delbrück scattering are also observed at
large angles [5]. As a result the determination of the Compton differential cross section only depends
on the relative rates of incoming and scattered photons and makes this method a favorable tool to
scale the photon-neutron data. One experiment of this type has been carried for 209 Bi [63], being an
immediate neighbor of 208 Pb. The advantage of investigating 209 Bi instead of 208 Pb is, that because
of the much higher level density the cross section is a smooth function of energy whereas that of 208 Pb
is rather fragmented [64]. Furthermore, information on magnetic moments to compare with when
studying the Fujita-Hirata relation (see section 7) is most clear-cut for 209 Bi which has one valence
nucleon out side a closed core. The enhancement constant obtained in this way was rather large,
amounting to γ E1 = 1.46 ± 0.05. Later on it was realized that in spite of the smoothness of the
◦
cross section the exact value for κGDR depends on very fine details of the shape of the measured cross
sections [65, 66, 201] so that from one single experiment alone the desired systematic precision can
hardly be achieved. This point has been investigated by Fuhrberg et al. [66], showing that in spite of
precise scaling through a Compton scattering experiment minor differences in available photoneutron
cross sections lead to sizable differences in γ E1 . Details of this investigation are shown in Figs. 4.2–4.4.
The results of the two different Compton scattering experiments [63, 201] are shown in Fig. 4.2. The
data points at 9.0, 11.4 and 17.74 MeV depicted by closed circles have been measured using photons
from nuclear reactions [63], the other data points by tagged photons [201]. It is interesting to note
that the data from the two experiments coincide on a percent level of pecision at the high-energy side
of the spectrum at 17.4 Mev and that there are no data from tagged photons below 12.5 MeV. There
are two different photoabsorption experiments shown in Figs. 4.3 and 4.4. The photoabsorption data

42
of Fig. 4.3 are in agreement with the solid curve after applying a scaling factor of 1.35 whereas the
data of Fig. 4.4 are partly in favour of the solid curve and partly in favour of the dashed curve after
multiplying with a scaling factor of 1.09. As a whole the solid curve appears to be more justified than
the dashed curve but on the other hand the difference — at least at a first sight — appears to be
not dramatic. But nevertheless, when drawing conclusions with repsect to the retarded enhancement
constant κGDR = γ E1 − 1 the difference between curves become essential. For the solid curve we get
κGDR = 0.46 ± 0.05 whereas for the dashed curve κGDR = 0.29 ± 0.05. Summarizing we can say that
two different Compton scattering experiments lead to the same large scaling factors for the available
photoabsrption data. In spite of this, a slight difference in the shapes of the photoabsorption cross
sections on the low-energy sides leads to a sizable (3σ) difference in κGDR . In view of the rather
precise Compton cross sections shown in Fig. 4.2 at 9.0. and 11.4 MeV the larger of the two κGDR
results appears more likely to be true.

Figure 4.2: Differential cross sections for Compton scattering by 209 Bi. Scattering angle θ = 135◦ .
Data points at 9.0, 11.4 and 17.74 MeV are measured by [63]. Other data points are measured by [201].
Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the GDR
parameters of [201].

Figure 4.3: Photoabsorption cross sections for 209 Bi measured by [202] multiplied by a scaling factor
of 1.35. Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the
GDR parameters of [201].

Figure 4.4: Photoabsorption cross sections for 209 Bi measured by [203] multiplied by a scaling factor
of 1.09. Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the
GDR parameters of [201].

The conclusion from the foregoing is, that it has advantages to rely on averages over a larger
number of nuclei rather than on the result obtained for one single nucleus. On the theoretical side it
was realized [95] that instead of the experimental quantity (γ E1 − 1) = κGDR the unretarded quantity
◦
κGDR is of interest for a comparison with models, as outlined in the previous section.
An extensive investigation of Compton scattering in the giant resonance region in addition to total
photoabsortion measurements has been summarized in our previous paper [95]. The result of this
investigation is contained in Table 4.2. For nuclei in the spherical region around 208 Pb, i.e. for mass
numbers A = 197–209, the average integrated cross section amounted to γ E1 = 1.38 ± 0.05 leading to
◦
κGDR = 0.46 ± 0.05 (4.48)

as a best value for the unretarded GDR enhancement constant in the 208 Pb range [95]. By change this
number coincides with the one given in our original experiment [63] as documented in Figs. 4.2–4.4.
For the neighbouring mass regions of deformed nuclei the quantities γ E1 proved to be smaller than
for the spherical region as also shown in Table 4.2. This observation is rather interesting and will be
interpreted in connection with the effective mass M ∗ of the Fermi liquid theory (cf. section 7) quoted
in column 4 of Table 4.2.
In case of the total photoabsorption measurements which are the favorable method at low and
intermediate mass numbers there are also two major difficulties and, thus, require a rescaling through
Compton scattering. The first is that the measurement of the total photoabsorption cross section can
only be done as a measurement of the target-in target-out difference. The second is that even for
light nuclei e+ e− pair production is the dominating process so that this process has to be calculated

43
Table 4.2: Average experimental (retarded) integrated cross section γ E1 in units of the (unmodified)
◦
TRK value, average unretarded enhancement constant κGDR and effective mass M ∗ with M ∗ /M =
◦ GDR
1/(1 + 3κ /4) following from Fermi liquid theory (cf. section 7, Eq. (7.30)) for three different
nuclear mass ranges.

◦
A γ E1 κGDR M ∗ /M
181 1.27 0.34 0.80
197–209 1.38 0.46 0.75
232–238 1.27 0.34 0.80

precisely and subtracted from the data. The precise calculation of the e+ e− pair production cross
section is by far not without problems because of quantum-electrodynamic corrections entering into
the calculation. Therefore, also in this case a scaling via Compton scattering leads to improvements
on the photoabsorption cross section in general and to tests on the structure of the photoabsorption
cross section [42, 48].

4.7 The isovector giant-quadrupole resonance and higher multipoles

Giant multipole resonances other than the giant-dipole resonance are an important subject of pho-
tonuclear physics. In the previous sections we have mainly covered the theoretical aspects of these
investigations. An overview over the status of research both experimental and theoretical up to the
late 1970s is given in a conference proceedings edited by Bertrand [204].
The isoscalar and isovector magnetic-dipole resonances have been identified below particle emission
threshold as isolated levels or sequences of isolated levels [205, 206]. The favorable tool for these
investigations is nuclear resonance fluorescence.
The isoscalar electric giant-quadrupole resonance has been discovered [207] by electron scattering
experiments at energies below the peak of the giant-dipole resonance and investigated through different
nuclear reactions [208]. Because of this location at the low energy side of the giant dipole resonance
and because of the small strength located in this resonance is not possible to observe the isoscalar
quadrupole excitation through Compton scattering. Most of the information regarding the location
of the corresponding isovector excitation has also been obtained from inelastic electron scattering
experiments [209], which show a concentration of quadrupole strength at an excitation energy of
∼ 130×A−(1/3) MeV. However, the interpretation of the results in terms of excitation strength remains
uncertain. Other methods of determining location and strength mainly rely on the interference of the
quadrupole resonance with the predominant electric dipole resonance, leading to for-aft asymmetries.
Among these are photon-nucleon reactions, the radiative capture of nucleons and Compton scattering
experiments.
The IVGQR in 208 Pb has been investigated via Compton scattering of unpolarized [25] and po-
larized [33] photons where in the latter case linear polarization has been produced through off-axis
tagging. These experiments confirm that at least one isovector quadrupole sum rule is exhausted by
these resonances though the experimental errors forbid precise conclusions on the overfulfilment of
the sum rule. One example of Compton scattering experiments identifying the IVGQR is discussed in
Fig. 4.5. The quadrupole strength shows up as an interference of E2 and E1 resonance amplitudes.
A further property of the IVGQR of 208 Pb which only can be seen through Compton scattering at
high energy-resolution by using photons from nuclear reactions and Ge(Li) detectors [64] is its partial
fragmentation. This experiment [64] showed that there is E2 strength located at about 17.6 MeV.

44
Figure 4.5: Cross-section ration σ(θ = 150◦ )/σ(θ = 60◦ ) for Compton scattering of unpolarized photons
by 208 Pb [25]. Solid line: Calculated including the IVGQR. Dashed line: Calculated not including the
IVGQR. The IVGQR shows up as interference of the E2 amplitude with the predominant E1 amplitude
from the GDR.

Studies of (n, γ) reactions have provided valuable data on the IVGQR in nuclei ranging from 40 Ca
to208 P b [210–213]. However, neutron capture experiments are limited by the difficulty of producing
a sufficiently intense flux of high-energy monochromatic neutrons.
Data on the isovector quadrupole resonance of 40 Ca have been obtained via the photo-neutron
reaction (γ, n) and the (n, γ) neutron capture reaction [214]. Data analysis in terms of models estimate
the isovector quadrupole resonance to be at an energy of 31.0 MeV with a width of 16.0 MeV, and
exhausting most of the energy weighted IVGQR sum rule.
A study of Compton scattering by 12 C and 40 Ca carried out by Wright et al. [18, 19] did not lead
to the identification of localized IVGQR strength.

4.8 The electromagnetic polarizabilities of the nucleus

After first estimates of the electric polarizabilities of nuclei due to Migdal [215] (see also [216]), the
electromagnetic polarizabilities of the nucleus were addressed by Ericson and Hüfner [175] for the case
of no enhancement of the giant multipole strength and by Friar [69] on a more general ground. In this
section we want to use our formalism to include the effects of enhancement and to give a transparent
formulation of the retardation effects. For this purpose we start from the amplitude for Compton
scattering by giant resonances, viz.

TGR (ω, θ) = B(ω, θ) + SGR (ω, θ) + RGR (ω, θ) (4.49)

where B(ω, θ) is the kinetic seagull amplitude, SGR (ω, θ) is the GR part of the mesonic seagull ampli-
tude, which does not contain any energy dependence except of the form factor, as introduced before,
and RGR (ω, θ) is the resonance amplitude. The electromagnetic polarizabilities are obtained by ex-
panding the scattering amplitude up to the quadratic order of ω, i.e. at small ω the amplitude TGR
can be written as
Z 2 e2
ǫ1 · ǫ2 + ω 2 ᾱGR ǫ1 · ǫ2 + β̄GR s1 · s2 .


TGR (ω, θ) = − (4.50)
AM
This expansion is first given for the forward direction, making use of the optical theorem and the
once-subtracted dispersion relation (cf. Eq. (3.21)). One obtains a form valid for ω much below giant
resonance energies:

Z 2 e2 ω2 ∞ σ GR (ω ′ ) ′
Z
ReTGR (ω, 0) = − + 2 dω + O(ω 4 ). (4.51)
AM 2π 0 ω ′2
From (4.51) we read off the Baldin–Lapidus sum rule [217] for the sum of electric and magnetic
polarizabilities due to giant resonance excitations in the form

1 ∞ σ GR (ω) 1 X σνλL ΓλL

Z
ν
ᾱGR + β̄GR = 2 dω = . (4.52)
2π 0 ω2 4π λL, ν (EνλL )2

In order to arrive at the electromagnetic polarizabilities of the nucleus we have to take into account
that in addition to giant resonances the quasideuteron and nucleon-internal degrees of freedom lead

45
to contributions. Applying the Baldin–Lapidus sum rule to all three contributions we arrive at the
relation
ᾱtot + β̄tot = (ᾱGR + β̄ GR ) + (ᾱQD + β̄ QD ) + A(α̃N + β̃N ), (4.53)
where ᾱQD and β̄ QD are the QD electric and magnetic polarizabilities, while again α̃N and β̃N are
the in-medium electromagnetic polarizabilities of the nucleon. A rough estimate shows that the main
contribution to electromagnetic polarizabilities of the nucleus stems from the giant resonances with
the QD degree of freedom contributing additional 3 per cent and the nucleon internal excitations
giving 0.3 per cent. Therefore, our present discussion can be restricted to the GR degree of freedom.
The Baldin–Lapidus sum rule cannot simply be separated into an electric part and a magnetic
part. In order to nevertheless obtain explicit expressions for ᾱGR and β̄GR one can either make use
of dispersion sum rules written separately for the electric and magnetic polarizabilities (see [99, 226]
and section 8) or make use of second-order perturbation theory. The latter procedure is as follows:
We expand the representation (3.56) of the giant resonance amplitude RGR with respect to ω, as well
as the corresponding seagull amplitude B + SGR and take into account the fact that the resonant E2
contribution proportional to ω 2 cancels explicitly with the part coming from the seagull amplitude.
As a result we get
ᾱGR = α0 + ∆α , β̄GR = βpara + βdia , (4.54)
where
2 X | h0|D|ni |2 2 X | h0|M|ni |2
α0 = , βpara = , (4.55)
3 n6=0 En − E0 3 n6=0 En − E0

are the so-called “proper” electric and paramagnetic polarizabilities, respectively. The sums in Eq.
(4.55) are taken over all intermediate states corresponding to an internal excitation and M is the
magnetic dipole operator. The quantity ∆α is the retardation correction of the electric polarizability
and βdia the diamagnetic polarizability. For giant resonances these quantities are given by

1 Z 2 e2 D 2 E 1 Ze2 D 2 E ◦ GDR N D2
 D EGR 
∆α = r , βdia =− r +κ r2 − . (4.56)
3 AM 1 6 M 1 A 2 2AM

The contribution α0 is obtained from the part of ΘE1 ν in Eq. (3.57) containing the electric dipole
moment operator. The retardation correction ∆α has two contributions, one from the expansion of
the form factor in the seagull amplitude and the other from the expansion of the difference
◦ 2
|hν| QE1 |0i|2 − hν| QE1 |0i

in Eq. (3.57). Similarly, βpara corresponds to the contribution of ΘMν

1 in the resonance amplitude, Eq.

(3.56). The first term in Eq. (4.56) for βdia as in the case of ∆α is due to the expansion of the form
factors in the seagull amplitude, but in contrast to the expression for ∆α the mesonic contribution
◦
(proportional to κGDR ) appears explicitly. The second part of βdia is a recoil correction, which is
small as it contains an additional factor 1/A in comparison to the first term. In our review we
do not consider any recoil corrections, but write this term only for the sake of completeness. The
two quantities ∆α and βdia are common ingredients to all quantum mechanical treatments of the
electromagnetic polarizabilities, either of atoms [218] or of nuclei [69, 175] and hadrons [99, 103, 197].
An extensive investigation of the two quantities in the framework of a nonrelativistic theory is given
by L’vov and Schumacher [103]. According to all these investigations the quantity ∆α has its origin in
the form factor squared of the charged composite object at the momentum of the incoming or outgoing
photon. The formula (4.56) for ∆α, first established by Petrun’kin (see in [99, 197]), is independent
of the model used to describe the hadron or nucleus and also valid in the relativistic case, what was
recently re-emphasized in [219].

46
 A discussion of the electromagnetic polarizabilities of the nucleus including the effects of enhance-
ment has been given in [95] in the framework of the phenomenology outlined in section 3. These
polarizabilities have two contributions, one from the expansion of the form factors in the seagull am-
plitude B + SGR , the other from the resonance amplitude RGR . Using Eqs. (3.70), (3.71) and (4.16)
we arrive at the multipole decompositions
" #
Ze2 N ◦ GDR ω2 D 2E N ◦ GDR D 2 EGR
  
(E1)
(B + SGR ) =− 1+ κ − r + κ r gE1 (θ)
M A 3 1 A 2

Ze2 ω 2 D 2 E N◦
 D EGR 
(M 1)
(B + SGR ) =− r + κGDR r 2 gM 1 (θ)
M 6 1 A 2
Ze2 ω 2 D 2 E N◦
 D EGR 
(B + SGR )(E2) = − r + κGDR r 2 gE2 (θ)
M 6 1 A 2
" #
N Z e2  ◦ GDR
 ω 2 D E
2 ◦ GDR
D EGR 
2
R̂GR = 1+κ − r +κ r gE1 (θ)
A M 3 1 2

ω2 σνλL ΓλL
ν
gλL (θ),
X
+ 2 (4.57)
4π E1,M 1,E2, ν
F(ret) (EνλL ) (EνλL )2

where the sum for the resonance amplitude is written with taking into account only the first three
multipoles (E1, M 1 and E2), as in the order ω 2 higher multipoles do not contribute. Here we used
the retarded form R̂GR of the resonance amplitude, as it is most appropriate at energies below the
resonance region. In addition, we took into account the fact that different retardation form factors
should appear for the one-nucleon and the two-nucleon contribution. Inserting Eq. (4.57) into the
definitions, Eq. (4.55), we find
◦ E1
1 ∞ σ (ω) 1 X σνE1 ΓE1
Z
ν
α0 = dω = 2
2π 2 0 ω2 4π ν F(ret) (EνE1 ) (EνE1 )2
◦M1
1 ∞ σ (ω) 1 X σνM 1 ΓM 1
Z
ν
βpara = dω = 2 . (4.58)
2π 2 0 ω2 4π ν F(ret) (EνM 1 ) (EνM 1 )2

In the same way we may also obtain expressions for ∆α and βpara from Eq. (4.57), together with (4.56).
The quantity ∆α has two terms, ∆αs and ∆αr stemming from the seagull amplitude (B + SGR )(E1)
and from the E1 part of the resonance amplitude R̂GR , respectively, whereas βpara has only one term
coming from (B + SGR )(M 1) . These terms are

1 Ze2 D 2 E N◦
 D EGR 
∆αs = r + κGDR r 2
3 M 1 A 2
2
1 e NZ
D E E 
◦ GDR 2 GR
D
2
∆αr = − r +κ r
3M A 1 2
1Z e2 2 D E
∆α = ∆αs + ∆αr = r2
3 AM 1
1 Ze2 D 2 E N ◦ GDR D 2 EGR
 
βdia = − r + κ r . (4.59)
6 M 1 A 2

It is seen that a sizable dependence of the diamagnetic polarizability βdia on the enhancement constant
◦
κGDR exists.
The main assumption in putting together these expressions is the following: In Eq. (4.57) the
retarded resonance amplitude R̂GR has been used. Clearly, the influence of the function ∆R introduced
in section 4.4 should be small in the low-energy expansion. Nevertheless, it may be useful to understand

47
what contributions are coming from this correction. The angular dependence in Eq. (4.20) shows that
the function ∆R only influences the electric polarizability, but not the magnetic one. Substituting the
definitions of R̃GR and RGR in Eq. (4.20) and extracting the term proportional to ω 2 we obtain an
expression for the ∆R-contribution α(∆R) to the electric polarizability:
◦
1 X σνλL ΓλL 1 σ λL λL
ν Γν 1 N Z e2
D EGR 
◦ GDR
E D
ν 2 2
X
α(∆R) = λL 2
− + r +κ r . (4.60)
4π λL, ν (Eν ) 4π E1,M 1,E2, ν
(EνλL )2 3 A M 1 2

Making use of the relation (4.18) and noting again that

◦ GR
2 1 r2 + κGDR r 2
F(ret) (ω) ≈ 1 − 1
◦ GDR
2
ω2 ,
3 1+κ
we find a simple expression for α(∆R):
◦ λL
1 X σ λL
ν Γν
α(∆R) = . (4.61)
4π M 2,E3,...
(EνλL )2

The most important property of Eq. (4.61) is the fact that only multipoles higher than the electric
quadrupole contribute to the correction α(∆R). This statement can independently be confirmed via
the Baldin–Lapidus sum rule, Eq. (4.52). Subtracting all polarizability contributions, Eqs. (4.58)
and (4.59), from this relation, one easily comes to the same conclusion. Evidently, α(∆R) is highly
suppressed. Similar effects will be discussed on the basis of fixed-t dispersion relations in section 8.

Table 4.3: Electromagnetic polarizabilities of the nucleus 208 Pb and nucleon in units of 10−4 f m3 .
The nuclear polarizabilities are devided by the number A of nucleons in the nucleus. The nucleon
polarizabilities are proton-neutron averages ᾱN = 21 (ᾱp + ᾱn ) and β̄N = 21 (β̄p + β̄n ) of the free nucleons.

α0 /A 1202
∆αs /A 68
∆αr /A −54
ᾱGR /A 1216
βdia /A −36
ᾱN 11.3 ± 1.5
β̄N 3.7 ∓ 1.5

For the nucleus 208 Pb numerical values for the electromagnetic polarizabilities are obtained using
the parameters given in section 4.5. The results are given in Table 4.3 together with the proton-neutron
averages of the free-nucleon, the latter taken from section 5.3. For the magnetic polarizability of the
nucleus we only quote the diamagnetic component because the paramagnetic component is expected
to be considerably smaller. A fraction of 85% of the nuclear diamagnetism is due to kinetic currents
and the remaining 15% due to exchange currents.

48
5 Compton scattering in the quasideuteron range

In framework of our concept to describe nuclear Compton scattering in terms of three different degrees
of freedom, viz. giant-resonance (GR), quasi-deuteron (QD) and nucleon-internal (N), we now consider
the second of these contributions, i.e. the amplitude TQD (ω, θ). We may write the total nuclear
scattering amplitude in the from

Ttot (ω, θ) = TGR (ω, θ) + TQD (ω, θ) + TN (ω, θ) = TA (ω, θ) + TN (ω, θ) (5.1)

where TA (ω, θ) collects the two amplitudes, TGR (ω, θ) and TQD (ω, θ), which are related to the external
degrees of freedom of the nucleon. In the QD energy region we observe a superposition of contributions
from all three degrees of freedom with the QD part being the least important one. The reason for this
surprising property may be understood as follows. In the QD range the scattering amplitude Ttot (ω, θ)
consists of contributions from the imaginary part of TQD (ω, θ) and the superposition of the real parts
of all three amplitudes. The real part of the QD amplitude is suppressed in middle of the QD range
because of its sign change at the peak of the QD cross section. As a consequence the resulting scattering
amplitude is mainly due to the high-energy tail of the GR amplitude, TGR (ω, θ), and the low-energy
tail of the amplitude TN (ω, θ) which is related to the in-medium electromagnetic polarizabilities α̃N
and β̃N of the nucleons. The most interesting topic of Compton scattering experiments in the QD
range is the experimental determination of the in-medium polarizabilities and their comparison with
the values for the free nucleon.
In spite of the small size of the QD contribution to the Compton differential cross sections we have
to discuss this amplitude with care in order to take it into account with the highest possible precision.
This will be done in the following.

5.1 The quasideuteron amplitudes

The quasideuteron amplitude has two parts, viz.

TQD (ω, θ) = SQD (ω, θ) + RQD (ω, θ). (5.2)

The seagull amplitude SQD (ω, θ) has a static part S̃QD (ω, θ) which is energy-independent in the
forward direction and a dynamic, energy-dependent part which can be parameterized in terms of
mesonic corrections δα and δβ to the in-medium electromagnetic polarizabilities α̃N and β̃N of the
nucleon. By definition the quantities α̃N and β̃N are one-body effects which are related to the nucleon
internal coordinates only, whereas the quantities δα and δβ are related to two-body effects. The
important field of our current research concerning two-body effects will be discussed in section 9. The
static part of the mesonic seagull amplitude is of the same structure as the corresponding GR part.
This is
N Z e2 QD E1
S̃QD (ω, θ) = − κ g (θ)F2 (q) (5.3)
A M
where again κQD is the quasideuteron part of the enhancement constant κ = κGR + κQD and F2 (q) the
(QD)
two-body form factor. In principle the exact form factor F2 (q) for the quasideuteron contribution
to the static mesonic seagull-amplitude differs from the (total) two-body form factor F2 (q) introduced
(GR) (QD)
in section 3.2 and the following relation holds: κGR F2 (q) + κQD F2 (q) = κF2 (q), where the
(GR)
function F2 (q) appears in the GR part of the seagull amplitude as discussed in sections 4 and
(QD)
9 and in appendix B. However, the difference between F2 (q) and F2 (q) is much smaller than
(GR)
between F2 (q) and F2 (q) and may be neglected, as will be shown in section 9. In the following

49
 (QD)
discussion we will not distinguish between F2 (q) and F2 (q). The meson exchange corrections of
the electromagnetic polarizabilities enter into the scattering amplitude in the form

SQD (ω, θ) − S̃QD (ω, θ) = ω 2 A (δα ǫ1 · ǫ2 + δβ s1 · s2 ) F2 (q) + O(ω 4 ). (5.4)

Recently it was found [193] that the form factor F2 (q) in Eq. (5.4) is not identical with the two-body
form factor of Eq. (5.3). This difference, which is not negligible in the case of relatively light nuclei,
is due to the complicated structure of the two-nucleon correlation function and will be discussed in
section 9 within a modified Fermi gas model, together with the polarizability modifications δα and
δβ. For convenience all such energy dependences of the (total) mesonic seagull amplitude S(ω, θ) is
put into SQD (ω, θ), so that SGR (ω, θ) only consists of the term proportional to κGR . This is possible,
because any model investigation within the Fermi gas model may only address the total mesonic
seagull amplitude S(ω, θ) with no strict possibility to separate it into the GR and QD parts. The
reason for this is that the Fermi gas model does not take into account the specific differences between
GR and QD excitations. Therefore, this separation into GR and QD parts can only be understood in
the discussion of properties of the two resonance amplitudes RGR (ω, θ) and RQD (ω, θ) if we go beyond
the Fermi gas model, by explicitly taking into account the different nuclear excitation mechanisms
corresponding to the two processes. A possibility to extend the distinction between the GR and QD
parts to the mesonic seagull amplitude is provided by the Fermi liquid theory as discussed in section
7.
If we suggest that the resonance part of the quasi-deuteron contribution does not involve the nucleus
as a whole, but rather is connected with deuteron-like subsystems, its dependence on momentum
transfer should be given by the form factor F2 (q), which is in contrast to the giant resonance case,
where no such form factor appears. The energy dependence may be represented in a Lorentzian form,
which leads to
2 − ω 2 + iωΓQD
N Z e2 QD ω 2 EQD
 
RQD (ω, θ) = ǫ1 · ǫ2 F2 (q) κ + σQD ΓQD 2 . (5.5)
A M 4π (EQD − ω 2 )2 + ω 2 Γ2QD

Equation (5.5) expresses our supposition that the scattering amplitude of each of the contributing
quasideuterons follows an electric-dipole characteristic which should be the case. For the energy inde-
pendent term the factor F2 (q) is necessary to ensure vanishing of the resonance amplitude RQD (ω, θ)
in the high-energy limit for all angles. The assumption that the amplitude RQD (ω, θ) vanishes in the
high-energy limit leads again to the sum rule (3.42) for the integral over the absorption cross section
σQD . By explicitly using the Lorentzian form for σQD one finds

1 N Z e2 QD
σQD ΓQD = κ
4π A M
which is a useful relation between Lorentzian parameters and enhancement constant κQD . It is
interesting to note that as a result the seagull amplitude S̃ QD (ω, θ) and the energy-independent term
of the resonance amplitude RQD (ω, θ) cancel exactly.
In photon-nucleon investigations another phenomenological description other than a purely Lorentz-
ian form exists for σQD (ω). It is known as the Levinger representation ( [220], see also [221]) and has
the following form:
NZ
σQD (ω) = L σD (ω) exp(−D/ω) (5.6)
A
where L and D are some parameters and σD is the experimental photoabsorption cross section for
the deuteron. The exponential function takes into account that due to Pauli blocking not the total
deuteron absorption strength should appear in σQD . However, for the analysis of Compton scattering
data it is convenient and sufficient to use Lorentzian fits to the expression given in (5.6).

50
5.2 The single-nucleon contribution to Compton scattering up to quadratic order
in the photon energy

We now want to introduce the amplitude TN (ω, θ), which is connected with the third degree of freedom,
i.e. with the contribution from individual nucleons to the nuclear Compton scattering amplitude. Our
consideration starts from the spin-independent part of the scattering amplitudes Tp and Tn for the
free proton and neutron, respectively, [99] in the laboratory system,

e2 2κp + κp 2
Tp = − ǫ1 · ǫ2 + ω1 ω2 ǫ1 · ǫ2
M 4M 2
!
(1 + κp )2 1
− ω1 ω2 s1 · s2 k̂1 · k̂2 + ω1 ω2 s1 · s2
4M 2 4M 2
+ ω1 ω2 ᾱp ǫ1 · ǫ2 + ω1 ω2 β̄p s1 · s2 + O(ω 4 ), (5.7)

!
e2 κn 2 κn 2
Tn = ω1 ω2 ǫ 1 · ǫ 2 − ω 1 ω 2 s1 · s2 k̂1 · k̂2
M 4M 2 4M 2
+ ω1 ω2 ᾱn ǫ1 · ǫ2 + ω1 ω2 β̄n s1 · s2 + O(ω 4 ), (5.8)

The quantities κp and κn are the anomalous magnetic moments of the proton and neutron, respec-
tively, and ᾱp , β̄p , ᾱn and β̄n are the electromagnetic polarizabilities of the free proton and neutron,
respectively. Both amplitudes, Eqs. (5.7) and (5.8), can be divided into a non-Born part (containing
the electromagnetic polarizabilities) and a Born part (containing all other terms). When the nucleons
are embedded in a nucleus, recoil effects may be disregarded so that ω2 = ω1 to a good approximation.
Furthermore, form factors have to be introduced in order to take into account the spatial distributions
of nucleons in a nucleus. Clearly, it is also necessary to account for the Fermi motion of the nucleons.
In the energy region considered here this contribution is of the same order as relativistic corrections
and may thus be skipped for our present purpose.
The only problem in considering these single-nucleon contributions is to avoid any double-counting
with respect to the other parts of the nuclear scattering amplitude. It can directly be seen that (i) the
non-Born contributions have not yet been included in any other part of the amplitude Ttot and (ii)
the first term in Eq. (5.7) is evidently related to the kinetic seagull amplitude B(ω, θ). The remaining
terms in Eqs. (5.7) and (5.8) are corrections proportional to ω 2 and depend on the anomalous magnetic
moments of the nucleons. If one compares the values of ᾱN and β̄N given in section 5.3 with the
coefficients at ω 2 in the nucleon Born terms one can see that the latter are essentially smaller (of the
order of 0.4 × 10−4 fm3 ) and, therefore, may be neglected. Moreover, the contribution of the magnetic
moment is partly included into the resonance amplitudes RGR and RQD . Thus, we have to keep only
the nucleon non-Born contributions to the nuclear Compton amplitude which read
h  
T non−Born (ω, θ) = ω 2 Z ᾱp ǫ1 · ǫ2 + β̄p s1 · s2
  i
+ N ᾱn ǫ1 · ǫ2 + β̄n s1 · s2 + O(ω 4 ) F1 (q). (5.9)

In this representation, Eq. (5.9), use has been made of the assumption that protons and neutrons
follow the same one-body form factor F1 (q). From T non−Born (ω, θ) of Eq. (5.9) we may construct
the amplitude TN (ω, θ) of (5.1) by introducing the in-medium electromagnetic polarizabilities of the
nucleon α̃N and β̃N leading to
h  i
TN (ω, θ) = ω 2 A α̃N ǫ1 · ǫ2 + β̃N s1 · s2 + O(ω 4 ) F1 (q). (5.10)

51
For the special case of no modification of the electromagnetic polarizabilities in the nuclear medium
we have
Z N Z N
α̃N ≡ ᾱN = ᾱp + ᾱn β̃N ≡ β̄N = β̄p + β̄n . (5.11)
A A A A

5.3 Status of the free polarizabilities of the nucleon

Though electromagnetic polarizabilities of hadrons have been discussed already for a long time, some
clarifying remarks concerning these quantities are advisable at the beginning. The electromagnetic
polarizabilities ᾱp and β̄p of the free proton appearing in the Baldin-Lapidus sum rule or in the
differential cross section for proton Compton scattering of real photons (RCS) [99] are most natural
characteristics of the second-order proton response to external electromagnetic fields. Details of this
view are discussed below. The often used expressions like ᾱ = α0 + ∆α and β̄ = βpara + βdia introduce
quantities (they are normally termed the static electric polarizabilty α0 , the retardation correction of
the electric polarizability ∆α, the paramagnetic polarizability βpara and the diamagnetic polarizability
βdia ) which have only a theoretical significance but cannot be measured in real (or even gedanken)
experiments.
There is a persisting prejudice that ᾱ has not its own fundamental sense and rather appears in
Compton scattering as an artificial combination of the genuine polarizability α0 and the retardation
correction ∆α. The latter is thought to be caused by non-static effects pertinent to real photons.
They think that the static polarizability α0 determines a coupling of the particle to a static electric
field and therefore α0 alone is measured under static conditions which allow to expell the retardation
effects. Such a view point is wrong, what is easily demonstrated by a general consideration based on
effective Lagrangians [100].
A rigorous theorem, which, for the sake of simplicity, we formulate here for a spinless composed
hadron or nucleus of a known mass m and electric charge e, is the following. Whenever the particle
stays or slowly moves in the region outside external electromagnetic charges or currents jµext (r, t)
which create the probing electric E (or magnetic H) field, the internal structure of the particle and
its low-energy long-wavelength response to the field is characterized by the only additional parameter
ᾱ through an effective potential − 21 ᾱE2 (or by β̄ through − 21 β̄H2 , respectively) [100]. This implies
that the so-called static polarizability is irrelevant whenever the particle has a charge and an internal
size hr 2 i =
6 0, thus leading to ᾱ = α0 + e2 hr 2 i/(3m) 6= α0 . This theorem signifies that all standard
tools considered in textbooks as methods for measuring the electric polarizability (like placing the
particle into the field of an electric capacitor and looking at its energy shift or at an induced dipole
moment) give ᾱ rather than α0 [222]. There is no way to measure α0 instead of ᾱ but putting the
probing external charges j0ext (r, t) inside the particle, e.g. in an electron scattering experiment with
observation of a secondary real or virtual photon. In an experiment like this one could simultaneously
measure e2 hr 2 i and ᾱ and, therefore, determine α0 . This means that the polarizabilities ᾱ and β̄
introduced in Compton scattering studies are structure parameters which have a more general sense
than sometimes assumed. In essense, nothing changes when spin of the particle is included.
Since we are dealing here with Compton scattering itself, the relevance of the free-nucleon polar-
izabilities ᾱN and β̄N needs no further explanations. However we have to explain how the in-medium
polarizabilities of the nucleon, α̃N and β̃N , enter to nuclear Compton scattering. For a clear-cut defi-
nition of the in-medium polarizabilities [80,166] we have to distinguish between (i) effects of rebuiding
the internal (e.g., quark) structure of the nucleon entering into α̃N and β̃N as one-body quantities
and (ii) two-body effects of meson exchange currents between pn pairs which may be denoted by δα
and δβ. From a phenomenological point of view developed in sections 3 and 4 these latter quantities
are not part of the in-medium electromagnetic polarizabilities but are rather used to parametrize that
part of the “mesonic seagull amplitude” (Thomson scattering by correlated pn pairs) which is energy

52
dependent in the forward direction (cf. Eq. (5.4); for more details see section 9). In a first step
of the analysis of experimental data the convolution of the true in-medium electromagnetic polariz-
abilities and the meson exchange corrections are determined. This leads to the effective in-medium
electromagnetic polarizabilities
α̃ef f
N = α̃N + δα,
ef f
β̃N = β̃N + δβ. (5.12)
The quantities δα and δβ have to be calculated and, therefore, introduce some model dependence into
the determination of α̃N and β̃N . However, as will be shown later, the respective model dependent
uncertainties are not large.
We have to repeat our words said in the beginning of section 3 that it would be also possible
to accept the effective quantities (5.12) as an alternative (and different) definition of the in-medium
polarizabilities. Since the free-nucleon polarizabilities (at least, the electric one) are essentially deter-
mined by the pion cloud of the nucleon [223,224], the in-medium modifications of the pion contribution
to ᾱN and β̄N like a Pauli blocking [82, 104] have also a right to be attributed to the in-medium po-
larizabilities of the nucleon. However, in no way such effects should be added to the mesonic seagull
contribution because this leads to the double counting. Considering in the following the in-medium
polarizabilities we always mean the quantities α̃N and β̃N which are free from the mesonic seagull
correction.
In nuclei with Z = N which are investigated in the following the arithmetic averages of the proton
and neutron electromagnetic polarizabilities are observed. From the Baldin-Lapidus sum rule applied
to experimental photoabsorption cross sections the sum (ᾱ + β̄)N = 14.5 ± 0.5 (in units of 10−4 fm3 )
is very well known [225–228], where later estimates [227, 228] give values by a few per cent lower that
the older ones [225, 226]. Among the reasons for that is that experiments on Compton scattering
by the proton in the ∆-resonance range [229] and photoproduction experiments (like [146]) carried
out at MAMI (Mainz) lead to a more precise determination of the resonant M1+ amplitude of the ∆
photoexcitation. Now [230, 231] this amplitude is by about 3% lower than it was assumed a few years
ago in the SAID (SM95) parametrization of photo-meson amplitudes [232].
Though, apparently there is some room for discussion, the overall precision of (ᾱ + β̄)N is by far
good enough for the purpose of our present data analysis. Furthermore, adjustments in the predicted
differential cross sections are possible at small angles within the limits given by the errors of the nuclear
photoabsorption cross sections (cf. Fig. 5.2), making small differences in (ᾱ + β̄)N unobservable. Note
that the free parameter of our data analysis is the difference of in-medium polarizabilities (α̃ − β̃)N
which has no influence on the differential cross section at zero angle.
Separately, polarizabilities of the free proton have been measured in proton Compton scattering
experiments performed at photon energies below photo-meson threshold [233–237]. Their results are
partly summarized in the review paper [99] and in the Review of Particle Physics 1998 [238] (in
fact, the latter reference relies on a “global average” over experiments of 90’s derived in [237] on
the base of the Baldin–Lapidus sum rule and experimental data on the differential cross section of
proton Compton scattering). It was found that ᾱp = 12.1 ± 0.8 (stat + syst) ± 0.5 (theor) and
β̄p = 2.1 ± 0.8 (stat + syst) ± 0.5 (theor).
The knowledge of polarizabilities of the neutron is less certain. Using model-dependent dispersion
relations at fixed t, the differences ᾱp −ᾱn ≃ −1.4 and β̄p −β̄n ≃ 0.1 are theoretically expected [226,228].
On the experimental side the first meaningful number for the electric polarizability of the neutron
was measured by Rose et al. [239] through quasifree Compton scattering on neutrons bound in the
deuteron. The precision of this experiment was surpassed by an experiment on scattering of neutrons
in the Coulomb field of Pb nuclei, enriched in 208 Pb [240]. The reported number5 ᾱn = 12.6 ±
5
We give that number and a similar number below with a small correction of 0.62 added which takes care of relativistic
effects lost in the fully nonrelativistic treatment of the Coulomb scattering experiments [100, 222, 241].

53
1.5 (stat) ± 2.0 (syst) perfectly agrees with the above theoretical expectation. However, more recently
a new experiment on Coulomb scattering of the neutron [242] gave a rather different polarizability
ᾱn = 0.6 ± 5. It was also claimed that the high accuracy of the former Coulomb scattering result [240]
may possibly be grossly overestimated [243]. A detailed discussion of these points and a possible
experimental way out of this problem has recently been addressed by Wissmann et al. [244].
Given the experimental results for proton polarizabilities and the theoretical evaluations of the
proton–neutron differences, the works on nuclear Compton scattering discussed below use the following
values for averaged free-nucleon polarizabilities: ᾱN = 11.3 ± 1.5 and β̄N = 3.7 ∓ 1.5. These numbers
should be compared with what is inferred from nuclear Compton scattering experiments themselves.
The free-nucleon value of the sum is slightly shifted down in the nuclear medium to about (α̃ + β̃)N =
14.0 [34]. This observation of an approximately constant sum of electromagnetic polarizabilities does
not exclude that the relative sizes of the in-medium electric α̃N and magnetic β̃N polarizabilities may
be considerably different from the corresponding free-nucleon values due to meson exchange-currents
and/or modifications of the internal structure of the nucleon. In this connection it is of interest that
ef f
in-medium shifts as large as ∆αef f
exp = −8 and ∆β exp = +8 have been reported on the basis of analyses
of Compton scattering experiments on 16 O [43].

5.4 Meson exchange currents

In complex nuclei meson exchange currents are a consequence of the interaction between proton-
neutron pairs (quasideuterons). Following the notation of our previvous work [39] and the discussion
of section 5.1, the corresponding modification of the nuclear scattering amplitude up to the order ω 4
may be written in the form
 
S(ω, θ) − S̃(ω, θ) = δα gE1 (θ) + δβ gM 1 (θ) ω 2 F2 (q). (5.13)

The r.h.s. of (5.13) takes care of the fact that the meson exchange corrections δα and δβ are related to
a two-body effect and, therefore, go along with the formfactor F2 (q). The approximative definition in
the following Eq. (5.14) is more convenient, because in this case δα and δβ can be directly compared
with the electromagnetic polarizabilities of the free nucleon.
 
S(ω, θ) − S̃(ω, θ) ≈ δα(1) gE1 (θ) + δβ (1) gM 1 (θ) ω 2 F1 (q). (5.14)

A calculation of δα and δβ has been carried out by Hütt and Milstein [168] for nuclear matter. The
result is δα(1) (A = ∞) = −3.4 and δβ (1) (A = ∞) = +2.4 (cf. Table 5.1). For finite nuclei the same
authors [193] find the results also listed in Table 5.1. We see that the meson exchange corrections
of the electromagnetic polarizabilities become smaller with decreasing mass number. The reason for
this is that the possibility for two-body or — even more — three-body effects to take place becomes
smaller for decreasing mass number. Three-body effects are essential because they enter into the
process through correlations. In 4 He the possibility for two-body effects is largely reduced and that
for three-body effects almost absent.

5.5 Experiments on the bound-nucleon electromagnetic polarizabilities

Figure 5.1 shows a typical experimental arrangement as used at the MAX laboratory of the university
of Lund (Sweden) [51, 52]. The photon beam having an energy of about 100 MeV hits a thin metal
foil serving as a bremsstrahlung radiator. Quasi monochromatic photons are obtained through a
coincidence condition between an event in one of the NaI(Tl) detectors (A, B, C) and an event in
the tagger. The tagger consists of a magnetic spectrometer which directs the unused electrons into

54
Table 5.1: Meson exchange corrections of the in-medium electromagnetic polarizabilities in units of
10−4 f m3

A δα(1) δβ (1)
∞ −3.4 +2.4
40 −2.4 +2.1
16 −1.3 +1.0
4 ≈0 ≈0

the beam dump, and electrons which have lost energy through bremsstrahlung production on an
array of plastic scintillators. These plastic scintillators provide energy channels for the coincident
bremsstrahlung photons.

Figure 5.1: Typical experimental arrangement as used at the MAX laboratory of the university of Lund
(Sweden) [51,52]. The photon beam having an energy of about 100 MeV hits a thin metal foil serving as
a bremsstrahlung radiator. Quasi monochromatic photons are produced through a coincidence condition
between an event in one of the NaI(Tl) detectors (A, B, C) and an event in the tagger.

For the data interpretation the resonance amplitude R(ω, θ) and the static part of the mesonic
seagull amplitude S̃(ω, θ) have to be calculated in addition to the kinetic seagull amplitude B(ω, θ).
This is possible with very good precision for nuclei where the total photoabsorption cross section
is known. To illustrate this, the total photoabsorption cross sections of 16 O and 40 Ca measured by
Ziegler et al. [84] are shown in Fig. 5.2. The experimental photoabsorption cross section as fitted by
the solid line is partitioned into Lorentzian lines. Except for the sum of all these Lorentzian lines,
the only constraint is that the QD cross section as given by the dotted line resembles the Levinger
representation of Eq. (5.6) as close as possible. All the other Lorentzian lines may be interpreted in
terms of GR components. The experiments carried out to measure the in-medium electromagnetic
polarizabilities use photons of 50 to 80 MeV. At these energies only the tails of the GR components
are of relevance since the real part of the QD amplitude has a zero crossing in this energy range. This
makes the determination of the amplitude R(ω, θ) quite unambiguous. For the determination of the
static part of the mesonic seagull amplitude S̃(ω, θ) only the integrated GR and QD cross sections
have to be known in order to determine κGR and κQD . The formfactor F2 (q) is taken from model
calculations (cf. section 9).
On the basis of the prerequisites discussed in the preceding paragraph, we now discuss the present
status of our knowledge about in-medium electromagnetic polarizabilities based on recent results [48]
obtained for 40 Ca and 16 O. Throughout this section we will calculate the predictions for different
“effective” in-medium electromagnetic polarizabilities α̃ef f ef f
N and β̃N . These quantities are introduced
such that they may be inserted into Eq. (5.10) in order to take care of different tentative choices for
the in-medium electromagnetic polarizabilities including the effects of meson-exchange currents and
of modifications of the internal structure of the nucleon, while simultaneously the mesonic seagull-
amplitude S(ω, θ) is represented through its static approximation S̃(ω, θ). These choices are
(i) the free-nucleon electromagnetic polarizabilities,
(ii) the free-nucleon electromagnetic polarizabilities supplemented by meson exchange corrections pre-
dicted for the finite nucleus,
(iii) the free-nucleon electromagnetic polarizabilities supplemented by meson exchange corrections pre-
dicted for nuclear matter.

40 Ca: Figure 5.3 shows the energy distributions of the differential cross sections obtained in recent

55
Figure 5.2: Total photoabsorption cross sections [84] for 40 Ca (upper figure) and 16 O (lower figure)
partitioned into components having Lorentzian shapes. Dashed curves: GR components. Dotted curves:
QD components.

experiments [42, 47, 48] together with predictions. The nuclear part of the amplitudes, i.e. B(ω, θ) +
S̃(E, θ)+R(E, θ), is fixed through the experimental total photoabsorption cross section so that the only
parameters are the effective electromagnetic polarizabilities α̃ef f ef f
N and β̃N . All three curves are in good
agreement with each other and with the data points at θ = 45◦ . This shows that the nuclear part of the
scattering amplitudes has been calculated with sufficient precision, so that the angular dependence of
the differential cross section may be interpreted in terms of properties of the effective electromagnetic
polarizabilities. The only remaining precaution is concerned with the isovector giant-quadrupole
resonance (IVGQR) which also has an effect on the angular distribution of the differential cross
section. After taking the IVGQR into account using experimental photoabsorption data [212, 214],
the uncertainty of the strength and location of the IVGQR leads to modifications of the angular
distribution, being essentially smaller than the differences between the lower (dashed) curves and
the (solid) curves in the middle and, thus, is irrelevant for our conclusion that there is no noticeable
modification of the in-medium electromagnetic polarizabilities compared to the free ones. Indeed, there
is apparent preference of the experimental data for the solid curves i.e. the predictions including the
free-nucleon electromagnetic polarizabilities supplemented by the meson-exchange corrections δα(1)
and δβ (1) predicted for the finite nucleus. However, for 40 Ca it apparently makes no major difference
whether we make use of the nuclear matter (A = ∞, dotted curves) predictions or the predictions
for the finite nucleus (A = 40, solid curves). In the next paragraph we will see that such differences
become visible for the smaller nucleus 16 O.

Figure 5.3: Experimental elastic differential cross sections [42, 47, 48] for 40 Ca versus scattering
angle compared with predictions. Eγ = 58 MeV (upper Figure), Eγ = 75 MeV (lower Figure). The
curves are calculated for (i) the free-nucleon electromagnetic polarizabilities (dashed), (ii) the free-
nucleon electromagnetic polarizabilities supplemented by meson exchange corrections predicted for the
finite nucleus (solid), and (iii) the free-nucleon electromagnetic polarizabilities supplemented by meson
exchange corrections predicted for nuclear matter (dotted).

16 O: Figure 5.4 shows data obtained for 16 O versus scattering angle. For this smaller nucleus the
meson exchange corrections δα(1) and δβ (1) of the electromagnetic polarizabilities predicted for the
finite nucleus (A = ∞, solid curve) shows a sizable difference from the corresponding prediction for
nuclear matter (A = 16, dotted curve). The lower curves (dashed) have been calculated with the
meson-exchange corrections set to zero. It is apparent that the solid curves provide the optimum
fit to the majority of the experimental data — if we tentatively exclude the two data points for Eγ
= 75 MeV and about θ = 60◦ from the present consideration. This finding leads to the conclusion
that the predicted meson-exchange corrections are in line with our experiments and that there are no
indications of an additional in-medium modification of the electromagnetic polarizabilities.

Figure 5.4: Differential cross sections [39, 47, 48] for Compton scattering by 16 O versus scattering
angle compared with predictions. Eγ = 58 MeV (upper figure); Eγ = 75 MeV (lower figure). The
curves are calculated for (i) the free-nucleon electromagnetic polarizabilities (dashed), (ii) the free-
nucleon electromagnetic polarizabilities supplemented by meson exchange corrections predicted for the
finite nucleus (solid), and (iii) the free-nucleon electromagnetic polarizabilities supplemented by meson
exchange corrections predicted for nuclear matter (dotted).

56
5.6 Summary and Conclusions

The data presented above have led to a satisfactory consistency in our conclusion concerning the
in-medium electromagnetic polarizabilities. This conclusion is that the proton-neutron averages of
in-medium electromagnetic polarizabilities α̃N and β̃N are the same as the corresponding quantities
ᾱN and β̄N for the free nucleon within a precision of the order of ±2.5 × 10−4 fm3 , if we do not
put too much weight on some points discussed in the next paragraph. Furthermore, there are strong
indications that the data obtained for 40 Ca and 16 O are in favour of the predicted [193] meson-exchange
corrections δα and δβ of the electromagnetic polarizabilities. It was known before that the sum of
electromagnetic polarizabilities ᾱN + β̄N is not modified through binding. The essential additional
message of our result is that the difference of electromagnetic polarizabilities ᾱN − β̄N also remains
the same. In the dispersion theory of nucleon Compton scattering [101] the difference ᾱN − β̄N is
an independent input parameter which may be related to the exchange of the scalar σ meson in the
t-channel. On the other hand the scalar σ meson is believed to be responsible for the largest part of
the binding potential between nucleons in a complex nucleus. This observation suggests that there
might be a relation between the quantity α̃N − β̃N measured in nuclear Compton scattering and the
nucleon binding potential. The tentative conclusion from our experimental result, therefore, may be
that the role the σ meson plays for the structure of the nucleon is not modified when the nucleons are
imbedded in nuclear matter.
The conclusions drawn in the foregoing section are weakened to some extent by some unsolved
problems which should be cleared up in further studies: (i) The free-nucleon polarizabilities of the neu-
tron are not known with the desirable reliability. (ii) The angular distribution of Compton-scattering
differential cross-sections for 12 C and 16 O at an energy of 75 MeV deviates from the predictions at θ
= 60◦ by 20–30 % [39]. These deviations have not been removed by the recent theoretical studies of
Hütt and Milstein [168, 193].

57
6 Compton scattering above π meson threshold

Nuclear Compton scattering above π meson threshold has three regions of current interest where
experimental work has been carried out:
(i) Compton scattering in resonance range where the experiments were restricted to the energy region
of the ∆ resonance.
(ii) Compton scattering in the asymptotic range at small transverse momenta pT , where shadowing
and vector-meson dominance phenomena may be studied (see e.g. [128]).
(iii) Compton scattering in the asymptotic range at large transverse momenta pT , where the direct
coupling of photons to valence quarks may be studied. These experiments are of interest because the
wave function of quarks in the nucleon may be investigated. Data are available only for the proton
(see e.g. Kroll et al. [138, 139]).
In principle there is also a fourth type of Compton scattering experiments at high pT , viz. the
deep inelastic scattering of photons by quarks [245]. These deep inelastic Compton processes are not
coherent-elastic with respect to the nucleon or nucleus but resemble Compton scattering by single
quarks with subsequent hadronization. As a consequence a jet of hadrons is leaving the nucleon after
Compton scattering has taken place.
Our discussion here is rather brief and restricted to the ∆-resonance region of nuclei, i.e. to case
(i). The latter choice is made because this is the only region where recent experimental data are
available and because the theoretical ideas which are tested through the experiments have a large
overlap with those considered in the rest of the present paper.

6.1 Pioneering work on Compton scattering in the resonance range

The study of nuclear Compton scattering above π meson threshold started on the experimental side
with the work of Hayward and Ziegler [17]. The experiment has been carried out with a low-duty
factor linear accelerator then available in Mainz and with a comparatively small (25 cm⊘ × 36 cm)
NaI(Tl) detector positioned at one scattering angle of θ = 115◦ and in an energy range from 180 to 375
MeV. One difficulty in these experiments was the separation of elastic from inelastic photon scattering
processes. Therefore, the main interest in subsequent theoretical work [246, 247] was directed to this
problem. These investigations were based on ∆-hole model approaches [105, 107, 248, 249] including
inelastic processes and led to some agreement with the experimental data. These first investigations
showed that the proper experimental separation of elastic and inelastic components is essential for
this type of experiments. This was one of the major motivations for purchasing the large Mainz 48
cm⊘ × 64 cm NaI(Tl) detector [37].

12
6.2 Compton scattering by C in the ∆ range

In a first application of the large Mainz 48 cm⊘ × 64 cm NaI(Tl) detector elastic and inelastic photon
scattering from 12 C was investigated [37] at θLAB = 40◦ using the tagged-photon beam at the 855 MeV
microtron MAMI (Mainz). With a width of tagger intervals of ∆E= 2 MeV and an energy resolution
of the NaI(Tl) detector of 1.5 % the separation of the elastic and the inelastic lines was sufficient to
obtain elastic (Compton) and inelastic differential cross sections in an energy interval from 200 MeV
to 500 MeV. The differential cross sections for Compton scattering were compared with calculations
carried out in the ∆-hole model [108] leading to a remarkably good agreement between theory and
experiment at photon energies above 250 MeV and a drastic disagreement at photon energies below
250 MeV. The occurance of this discrepancy was very surprising because θLAB = 40◦ is close enough to
the forward direction so that any theory which reproduces the photoabsorption cross section and has a

58
correct low-energy limit should also reproduce the Compton scattering differential cross section. It is
interesting to note that the same observation was made in first interpretations of Compton differential
cross sections measured for 4 He at the same angle θLAB = 40◦ [44, 109] This observation seemed to
suggest that some omission should be in the ∆-hole approach as formulated in Refs. [107, 108]. In the
meantime this omission was identified to be the neglect of the seagull amplitudes B(ω, θ) + S̃(ω, θ)
which were discussed in section 3. Indeed after including the seagull amplitudes B(ω, θ) + S̃(ω, θ)
the agreement between theory and the θLAB = 40◦ data for 12 C and 4 He became much better, as is
further discussed below.

6.3 Compton scattering by 4 He in the ∆ range

Among the nuclei to be studied by Compton scattering 4 He is of special interest, because this nucleus
combines the well-known structure of a few-body system with the binding energy of a complex nucleus.
Furthermore, the first excited state is at 20 MeV, thus making the separation of elastic and inelastic
processes easy. Previous experiments on Compton scattering by 4 He in the ∆ range [20, 23, 29, 35, 40]
were carried out using the bremsstrahlung method and, therefore, may have suffered from the well-
known systematic errors contained in that method. In spite of that, the use of large NaI(Tl) detectors
made it possible to get good data on the elastic reaction at energies close to the end points of the
bremsstrahlung beams.
The first fully high-resolution Compton scattering experiments on 4 He were carried out only re-
cently with a remarkable progress. At LEGS (BNL) [45], the first data were obtained with a polarized
and tagged photon beam (∆E ≈ 5 MeV) produced by backscattering of laser light on 2.5 GeV elec-
trons. Photons were detected with a 48 cm⊘ × 48 cm NaI(Tl) detector. Measurements were performed
at six angles between 31◦ and 130◦ which produced angular dependences of the differential cross sec-
tions and beam asymmetries at 5 energies between 206 and 310 MeV. At MAMI (Mainz), for the first
time Selke et al. [44] performed measurements at a scattering angle of θLAB = 37◦ through the whole
∆ region from 120 to 510 MeV. The tagged-photon beam with ∆E ≈ 2 MeV and the large Mainz
48 cm⊘ × 64 cm NaI(Tl) detector were used in that experiment. The latter data were supplemented
by Kraus et al. [46] leading to additional data at θLAB = 37◦ taken with 48 cm⊘ × 64 cm NaI(Tl)
detector and to additional data at energies between 150 and 300 MeV for the two angles of θLAB = 93◦
and θLAB = 137◦ covered by smaller 25.4 cm⊘ × 35.6 cm NaI(Tl) detectors. In this work also the
beam asymmetry was determined between 230 and 270 MeV using linear polarization produced by
the coherent bremsstrahlung in a diamond crystal.
The results of the recent Mainz experiments [44, 46] for the differential cross section are shown in
Fig. 6.1 together with theoretical predictions to be discussed below. Not shown in that Figure are the
LEGS data [45] which, for example, lie in between the data points of Mainz [46] and Saskatoon [29] at
90◦ and agree with both, the Mainz and Saskatoon data, at 130◦ . A sample of the LEGS data can be
seen in Fig. 6.3 below. As was already found in the very first measurements at Bates (MIT) [20, 23],
any version of the ∆-hole model — the original version of Ref. [107], a simplified version [109] based
on a local-density approximation, and a version [108] corrected [46] for the seagull contributions to
the elementary γN scattering amplitude — fails to explain the large size of the differential cross
section observed at angles θLAB ∼ > 90◦ . This is in contrast to the fact that a nice agreement exists
between the experimental data and the modern predictions [46] at a small scattering angle of θLAB =
37◦ . Therefore, something fundamental must be wrong with our present understanding of Compton
scattering in the ∆ range.

59
Figure 6.1: Differential cross sections (dσ/dΩ)LAB for Compton scattering by 4 He [46] compared with
predictions. The full circles are calculated within the ∆-hole model including the seagull amplitudes
[46]. The results of the schematic model (dashed) and the extended schematic model (solid) are also
shown [46].

6.4 Calculations of Compton scattering in the ∆ range

The discrepancies found between the experimental 4 He data and the theory at large angles demonstrate
the necessity of comparing the main components of the existing different approaches with each other
and to search for possible lines of developments which may solve the problems, apparently existing in
all them. We will not discuss here whether the solution of these problems may be possible in a genuine
isobar-meson model or whether it might be necessary to introduce different degrees of freedom like
those of constituent quarks. Some ideas going into the latter direction have been published in recent
literature [250].
The schematic model:
A zero-order approach to a theoretical interpretation of Compton scattering in the ∆ range has been
attempted through the “schematic model”. It is based on the suppositions (i) that the absorption
and emission of photons takes place locally on one nucleon, (ii) that different nucleons contribute
equally and coherently, thus introducing into the nuclear Compton scattering amplitude Ttot (ω, θ) the
one-body nuclear form factor, and (iii) that the local spin-averaged γN scattering amplitude has a
pure magnetic dipole, M 1, characteristic (or optionally E1 but not both). In the schematic model
the energy dependence is not taken from the elementary γN scattering amplitude but rather from
the full nuclear amplitude at zero angle, Ttot (ω, 0). The latter quantity can be derived from the total
photoabsorption cross section σtot (ω) through the optical theorem and the GGT dispersion relation,
thus giving the dispersion-theoretical differential cross section dσGGT /dΩ. One then arrives at the
simple formula [17, 46] usually used in the CM frame:

dσ dσGGT 1 + cos2 θ
(ω, θ) = (ω, 0) |F1 (q)|2 . (6.1)
dΩ dΩ 2
Here q is the momentum transfer and F1 (q) the one-body formfactor. It is not expected that the
schematic model (6.1) makes a very realistic prediction except for very forward directions, because
e.g. of the neglect of any multipolarity except for M 1, which is an oversimplification for the ∆
energy-range.
The extended schematic model:
The nuclear scattering amplitude in the “extended schematic model” [46] is given by a combina-
tion of several pieces which are results of different reaction mechanisms. In close analogy with the
considerations given in sections 3 to 5, we write

Ttot (ω, θ) = TA (ω, θ) + TN (ω, θ),

TN (ω, θ) = T E1 (ω, 0) F1 (q) gE1 (θ) + T∆
nm
(ω, 0) F2 (q) gM 1 (θ)
m
+ T∆ (ω, 0) F1 (q) gM 1 (θ). (6.2)

In (6.2) the amplitude TA (ω, θ) contains all the relevant nuclear contributions as there are the ki-
netic and mesonic seagull amplitudes and some high-energy tail from the positive-energy part of the
quasideuteron amplitude. These pieces were discussed in sections 3 and 5. The amplitude TN (ω, θ)
is related to nucleon excitations including the πN continuum. It contains one term of electric-dipole
(E1) multipolarity and two terms of magnetic-dipole (M 1) multipolarity. The term T E1 (ω, 0) is cal-
culated via optical theorem and dispersion relation from the total photoabsorption cross section of

60
the nucleon excluding the ∆-resonance contribution. This is justified because the nonresonant one-
pion channel and the D13 (1520) resonance are of E1 multipolarity and also the nonresonant two-pion
channel is predominantly of E1 multipolarity. At low energies, a large part of T E1 comes from the
so-called Kroll-Ruderman mechanism of pion photoproduction. The ∆-resonance itself appears in
two terms, T∆ nm (ω, 0) and T m (ω, 0), corresponding to the nonmesonic (nm) and the mesonic (m)
∆
photoabsorption cross section of the nuclear ∆ resonance (see Fig. 2.6). Of the three terms on the
r.h.s. of (6.2) the nonmesonic ∆ term goes along with the formfactor F2 (q) because this term corre-
sponds to a two-body process. At forward angle the total amplitude (6.2) necessarily coincides with
the dispersion-theoretical amplitude TGGT (ω, 0) provided the sum of partial cross-sections used to
calculate the different pieces in (6.2) give σtot (ω), i.e. the total photoabsorption cross section of the
nucleus. Therefore, it is not surprising that the nuclear Compton scattering data at small angles find
an explanation in the framework of such a model. Really important is that by making use of both
multipolarities, M 1 and E1, and by including the seagull contributions improvements at large angles
are obtained.
The ∆-hole model:
Recent calculations of Compton scattering in the ∆-hole approach [46, 108, 109] originate from the
treatment of the problem given by Koch, Moniz and Ohtsuka [107]. Due to simplifications introduced
by the local density approximation, which was used in the later works, many parallels with the two
models described above become transparent. The scattering amplitude schematically reads
 
γ
Ttot (ω, θ) = AT∆ (ω)gM 1 (θ)F∆ (q) + ZTTγ (ω) + ATKR
γ
(ω) gE1 (θ)F1 (q). (6.3)

Here the first term gives the ∆-resonance part of the scattering amplitude. It appears from the model
as local and has a pure M 1 multipolarity. However, the form factor F∆ (q) it is accompanied with is not
identical with the one-body form factor F1 (q), because F∆ (q) counts both, the nucleon distribution
in the nucleus and a nuclear-density dependence of elementary Compton scattering on elements of
the nuclear volume through the ∆-hole excitation, which appears due to the density-dependent width
and self-energy of the dressed ∆ in the nuclear medium. The second term is of E1 multipolarity. It
represents the Thomson part of the γp scattering amplitude (which was omitted in [107, 108]) and
the Kroll-Ruderman part which appears due to an intermediate isovector excitation of the nucleon
to the s-wave πN state. It is assumed that the E1 part is not density dependent and, therefore, is
accompanied with the one-body form factor F1 (q).
Most recently such an approach was used in Ref. [109], in which the ∆ resonance amplitude takes
the form6
4 fγ2 2
Z h i
γ iq·r
AT∆ (ω)F∆ (q) = ω c dr ρ(r) e G∆h (ρ(r), s + ) + G∆h (ρ(r), s − ) . (6.4)
9 m2π
Here fγ is the γN ∆ coupling, ρ(r) the nuclear density, ωc the photon energy in the rest frame of the
produced ∆ , and G∆h the local density approximation to the ∆-hole Green function:
h√ i i−1
G∆h (ρ, s) = s − M∆ + Γ̃(ρ, s) − Σ∆ (ρ, s) , (6.5)
2

where Γ̃ and Σ∆ are the Pauli-blocked width and self-energy of the ∆ in the nuclear matter [107, 251,
252]. The effective energies s± in the s- and u-channels of the ∆-hole excitation include small shifts
caused by Fermi motion of the nucleons and read in the Fermi gas approximation:
 3 kF2 
s± = M 2 ± 2ω M + , (6.6)
5 2M
6
Strictly speaking, the form factor F∆ (q) in Eq. (6.4) depends on both, the momentum transfer q and the photon
energy ω.

61
where kF is the (local) Fermi momentum, kF3 = 23 π 2 ρ.
In the microscopic calculations of the ∆-hole model the constraint of the GGT dispersion relation
is not used. Therefore, an agreement of the calculated differential cross section with the experimental
data at small angles supports the validity of the model for nuclear Compton scattering. The agreement
at energies below 350 MeV is largly improved when the kinetic (Thomson) seagull contribution is
included [46, 109]. At energies above the ∆ peak the results of different implementations [46, 109]
of the ∆-hole model are different. A nice agreement with the experimental small-angle data [44]
was found in Ref. [46], in which the ∆ contribution was calculated with the parameters borrowed
from [108, 253].
Missing effects at large angles – nonlocality:
However, all the considered models fail to correctly describe the experimental results at large angles,
where the 4 He data [20, 23, 29, 45, 46] largely exceed the predictions. Comparing the different calcu-
lations, one may conclude that the density-dependent self-energy of the ∆ which results in a shift
and broadening of the ∆ peak is not the only effect which has to be taken into account to explain
nuclear Compton scattering in the ∆ range. At high momentum transfers the self-energy cannot be
considered as a local operator and therefore nonlocal effects should be taken into account. Nonlocal
two- or many-body E1 contributions might be important as well. This view was supported by a
calculation done by L’vov and Petrun’kin [129], in which large two-body E1 contibutions were ob-
tained as a result of effective contact interactions arising in a gauge-invariant description of Compton
scattering on a bound nucleon — see [133] for a more detailed motivation and derivation of such an
approach and its relation with the standard language. In spite of a phenomenological success of the
proposed scheme [45, 129], it cannot be fully applicable in the ∆ range, because e.g. the contact E1
contributions arise there as purely real, what cannot be true at energies exceeding the pion threshold.
The following simple model provides strong evidence that nonlocal effects missing in the relations
(6.3) and (6.4) may be responsible for the long-standing discrepancy between the ∆-hole model and
the 4 He Compton scattering data at large angles. The model we consider includes only one- and
two-body mechanisms of photon scattering by 4 He mediated by one-pion exchange. They are shown
in Fig. 6.2. We disregard the ρ-meson exchange. Pion absorption or rescattering by intermediate nu-
cleons is disregarded too what makes this model inapplicable close to the ∆-resonance peak where the
absorption is very strong. In spite of all these oversimplifications, the model is not fully unreasonable
at energies like 200 MeV, i.e. well below the ∆ peak.
Diagram a in Fig. 6.2 gives the impulse approximation. The corresponding amplitude is equal to

TIA (ω, θ) = A F1 (q) h TγN (ωef f , θ) i (6.7)

and is given by the spin-isospin average of the elementary γN scattering amplitude on the free nucleon.
The effective energy in Eq. (6.7),
q2  1
ωef f = ω + 1− , (6.8)
8M A
is obtained by assuming that the active nucleon has the momentum
q 1
pef f = − 1− (6.9)
2 A
in the Lab frame and that the rest of the nucleus is on shell (see e.g. [254]). The elementary γN
scattering amplitude could easily be approximated by the ∆-pole diagram plus a Thomson and Kroll-
Ruderman-like background as this was done e.g. in Refs. [107,109]. However, in the present context we
use instead numerical results obtained in the framework of the dispersion theory [101]. The differential
cross section for Compton scattering by 4 He obtained within the impulse approximation (6.7) at the
energy ω = 206 MeV is compared with predictions of the ∆-hole model [109] and with available

62
data [45, 46] in Fig. 6.3. Though the ∆-hole model includes the density-dependent modification of
the ∆-resonance contribution, this modification is numerically not a big effect at 206 MeV and the
predictions of the IA and the ∆-hole model are quite close to each other. The only exception is the
region of large angles where the difference is big. However, this difference is not related with the IA or
the ∆-hole model itself and is mostly caused by our choice for the effective energy (6.8) which includes
a q-dependent correction absent in the Ansatz (6.6).

Figure 6.2: One- and two-body contributions to nuclear Compton scattering. Diagram a is the impulse
approximation. Diagram b describes real pion production followed by pion absorption. Diagrams c are
required by the gauge invariance; in particular, the last diagram contains the contact γπN ∆ vertex.

Figure 6.3: Differential cross sections (dσ/dΩ)CM for Compton scattering by 4 He at Eγ = 206 MeV.
Data are from [45] (solid circles) and from [46] (open circles). Shown are predictions of the ∆-
hole model in the local-density approximation [109], impulse approximation, Eq. (6.7), and the total
contribution including the two-body MEC, Eq. (6.13).

The two-body contributions are given by diagrams b and c in Fig. 6.2. Diagram b is determined by
amplitudes of pion photoproduction which we take in the tree approximation with fixed nucleons, i.e.
including the Kroll-Ruderman γπN N vertex, the pion-in-flight current, and ∆ excitations in the direct
and crossed channels. In the diagram b the intermediate pion has the energy ω and, therefore, can be
real (not only virtual). Accordingly, even without ∆ contributions the diagram b contributes to the
imaginary part of the nuclear Compton scattering amplitude. In contrast to this, the diagrams c do not
contain real-pion propagators. Actually, the diagrams c do not represent a self-dependent contribution
but they rather ensure the gauge invariance of the total two-body amplitude. In particular, the last of
the diagrams c contains the contact γπN ∆ vertex which is labeled as the ∆–Kroll–Ruderman vertex
in Fig. 6.2. Explicit formulas for the kernel amplitude h Tb+c (k1 , k2 , Q) i which corresponds to the
diagrams b and c with omitted wave functions of external photons and nucleons and spin-isospin
averaged over nucleons in the nucleus are very similar to those given in Ref. [168] (see also section
9, Eqs. (9.3) and (9.21)). The kernel amplitude depends on the photon momenta k1 , k2 and on
the internal momentum Q which defines the sharing of the total momentum transfer q between the
momenta 12 q − Q, 12 q + Q transferred to each of the two interacting nucleons. Assuming oscillator
wave function for 4 He,
Cp  2
n 1 o
Ψ(p1 , . . . pA ) = const × exp p1 + . . . + p2A − P2 , P = p1 + . . . + pA ,
− (6.10)
2 A
we fix the oscillator parameter Cp through the one-body formfactor:
 Cp A − 1 2   q2 
F1 (q) = exp − q = exp − 2 , (6.11)
4 A 4α
where α = 165 MeV is determined from the electric radius of 4 He and that of the proton.Then the
two-body form factor is equal to7
 Cp A − 2 2 
F2 (q) = exp − q . (6.12)
8 A
For A = 4 we have F2 (q) = [F1 (q)]1/3 . Then the total MEC amplitude corresponding to diagrams b
and c reads
dQ  C
Z 
p 2
TM EC (ω, θ) = A(A − 1) F2 (q) exp − Q h Tb+c (k1 , k2 , Q) i. (6.13)
(2π)3 2
7
Note that we use the oscillator wave function only in order to evaluate the two-body form factor. Finding the
amplitude (6.7), we use a more accurate form factor found through experimental data on electron scattering by 4 He [256].

63
 In the limit ω = 0 the mesonic amplitude (6.13) determines the (unretarded) enhancement pa-
◦ ◦
rameter κ (see section 9 for more detail). Actually the value of κ obtained through the two-body
◦
diagrams of the lowest order in the πN N coupling is rather small [168], so that κ is dominated by
two-pion exchanges and three-body interactions. In the chiral limit of mπ → 0 the amplitude (6.13)
vanishes at ω = 0 due to the exact compensation between the diagrams b and c. However, when the
energy ω becomes higher than the pion threshold, such compensation is destroyed. In particular, the
diagram b gets an imaginary part at ω > mπ , whereas the diagram c without ∆ contributions remains
real. At energies above pion threshold the amplitude (6.13) is not suppressed any more by the small
pion mass and is numerically large. Adding the meson-exchange amplitude TM EC to the amplitude
of the impulse approximation TIA , we find a strong enhancement in the differential cross section at
backward angles shown in Fig. 6.3 at the energy 206 MeV which brings the predictions of the model
into a qualitative agreement with the data. At the energy 253 MeV the agreement with available
data [45, 46] is almost perfect but this may be accidental because the model is not expected to work
well close to the ∆ peak. On the other hand, the IA and the ∆-hole model grossly underestimate the
differential cross section at backward angles at these energies.
The increase obtained at backward angles is caused by mainly three reasons. First, the two-body
formfactor does not suppress the MEC contribution as much as the one-body formfactor does the IA
contribution; for instance, F2 (q)/F1 (q) ≃ 2.5 at 206 MeV and backward angles. Second, the medium
correction to photon scattering through E1 photoproduction which is described by diagrams b and c
with only electric couplings (these are diagrams without ∆) is quite large and predominately imaginary.
It explains about 50% of the obtained enhancement (in the amplitude) at 206 MeV and backward
angles. Such a correction was not included into the ∆-hole calculations of nuclear Compton scattering.
Third, equally large and also predominately imaginary is the nonlocal correction due to propagation
of the ∆-excitation from one nucleon to an other one through intermediate pion propagation.
The model considered here clearly is too simple to provide a perfect description of the 4 He
Compton-scattering data, especially close to the ∆ peak where multiple processes including pion
absorption and rescattering are very important. In addition to pion exchange, ρ-meson exchange has
also to be included. Moreover, nucleons should not to be considered fixed and also couplings to N N
continuum-states should be incorporated; in the ∆-hole model these couplings are a large part of the
so-called spreading potential of the ∆-hole states. It might also be important to calculate two-body
contributions with a more accurate wave function of 4 He including the D-wave. Despite all these
omissions, we believe that we can conclude that realistic calculations of nuclear Compton scattering
at high momentum transfers should properly take into account the nonlocal effects.

64
7 Fermi liquid theory and nuclear Compton scattering

An appropriate method to describe enhancement phenomena of the giant dipole resonance is within
Landau’s Fermi liquid theory [255,257] of finite systems as first applied by Migdal [86,258]. The Fermi
liquid theory is concerned with strongly interacting particles, which are replaced by weakly interacting
quasiparticles analogous to particles in an ideal Fermi gas with a residual interaction.
In a finite system, the problem has to be treated in close analogy to the nuclear shell model, with
the main difference that particle-hole excitations are replaced by quasiparticle-quasihole excitations.
Giant-dipole excitations involve transitions between neighboring oscillator shells from below the Fermi
energy-level to above it. In case of particle-hole excitations these transitions exhaust exactly one TRK
sum rule so that the transition to the quasiparticle picture should contain the overfulfilment of the
◦
TRK sum rule as represented by κGDR . This will be shown in this section.
The spectrum of collective excitations (e.g. sound and spin waves of giant resonances) and the
response of the system to an external (electromagnetic) field are determined by the quasi-particle
scattering amplitude near the Fermi surface for zero scattering angle. This amplitude is a function of
the angle between the initial quasi-particle momenta and is satisfactorily described by the first two
Legendre polynomials. The coefficients of these polynomials are the constants introduced into the
theory and are the main parameters of Landau’s theory of Fermi liquids [255].
In the Fermi liquid theory the system of particles with the spin 1/2 is replaced by a system of
quasiparticles with the same spin. This theory is aimed at the description of low-lying excitations.
Each quasiparticle is characterized by the momentum p and the quasiparticle number n(p). Strictly
speaking, this quasiparticle number is a matrix n̂(p) ≡ nα β (p) in spin and isospin spaces (for the sake
of simplicity we assume that each index simultaneously corresponds to spin and isospin variables).
The total number of particles N is equal to the total number of quasiparticles:
N dp
Z
Tr n̂(p) dτ = , dτ = , (7.1)
V (2π)3
(0)
where V is the volume of system. In the ground state one has nα β (p) = δα β θ(pF − p) as in a Fermi
gas. Due to the interaction between quasiparticles, the total energy of the system is not the sum of
quasiparticle energies. The energy matrix ǫ̂ ≡ ǫα β (p) of the (individual) quasiparticle is defined as a
variational derivative of the total energy E :
δE
Z
= Tr ǫ̂(p)δn̂(p) dτ . (7.2)
V
In the ground state ǫα β (p) = ǫ(0) (p)δα β and

∂ǫ(0) (p) pF
ǫF = ǫ(0) (pF ) , = vF = , (7.3)
∂p p=pF
M∗

where M ∗ is the effective mass of the quasiparticle. Giant-dipole excitations involve transitions be-
tween neighboring oscillator shells from below the Fermi energy to above it. Therefore, the relevant
nucleons have momenta close to the Fermi momentum. Near the Fermi surface ǫ(0) (p) = ǫF +vF (p−pF ).
Including quasiparticle excitations the energy matrix ǫα β (p) depends on the quasiparticle numbers
nα β (p) : Z
ǫα β (p) = ǫ(0) (p)δα β + fαγ,βδ (p, p′ )δnδ γ (p′ )dτ ′ , (7.4)
(0)
where δnδ γ (p) is the difference between nδ γ (p) and nδ γ (p) and fαγ,βδ (p, p′ ) describes the quasiparticle-
quasiparticle interaction. In this function one can put p = p′ = pF . Therefore, it depends only on the

65
directions of the vectors p and p′ . In matrix notations one can write Eq. (7.4) as
Z
ǫ̂(p) = ǫ̂ (0)
(p) + Tr′
fˆ(p, p′ )δn̂(p′ )dτ ′ , (7.5)

where Tr′ denote the trace with respect to indices corresponding to momentum p′ . Using (7.2) and
(7.5) one can write an expression for the excitation energy:

δE 1
Z Z Z
= Tr ǫ̂ (0)
(p)δn̂(p) dτ + Tr Tr′ fˆ(p, p′ )δn̂(p′ )δn̂(p)dτ dτ ′ . (7.6)
V 2

Usually the following representation is used for the operator fˆ(p, p′ ) of the interaction between quasi-
particles:
N0 fˆ(p, p′ ) = F + τ · τ ′ F ′ + σ · σ ′ G + τ · τ ′ σ · σ′ G′ . (7.7)
Here
2pF M ∗
Z
N0 = 4 δ(ǫ(0) (p) − ǫF ) dτ = (7.8)
π2
is the density of states on the Fermi surface (the factor four corresponds to two possible states in
both, spin space and isospin space), τ and σ are Pauli matrices for isospin and spin, respectively. The
functions F, F ′ , G and G′ depend on the angle θ between p and p′ and can be expanded with respect
to Legendre polynomials: X
F = Fl Pl (cos θ) , (7.9)
and the same for the other functions. The coefficients in this series are some phenomenological
constants which should be extracted from experiment. Usually only the zeroth and first harmonics
are essential. It possible to describe a big variety of processes inside the nucleus with only a few
coefficients Fl , Fl′ , Gl and G′l (see [258]).
It follows from Galilei invariance that the momentum density is equal to the density of mass flow.
Since the number of particles is equal to the number of quasiparticles, one has
∂ǫ̂
Z Z
Tr p n̂(p) dτ = M Tr n̂(p) dτ , (7.10)
∂p

where M is the nucleon mass. Substituting n̂(p) = n̂(0) (p) + δn̂(p) and using (7.5), we get for the
isoscalar current:
Z " #
p ∂ǫ(0) ∂n(0) (p′ ) ˆ ′
Z Z
j0 = Tr δn̂(p) dτ = Tr − Tr′ f (p , p)dτ ′ δn̂(p)dτ . (7.11)
M ∂p ∂p′

Taking into account the relation

∂n(0) (p) ∂ǫ(0) (p)

= −δ(ǫ(0) (p) − ǫF ) (7.12)
∂p ∂p

and also the representation (7.7), we finally obtain

p ∂ǫ(0) F1
Z Z  
j0 = Tr δn̂(p) dτ = Tr 1+ δn̂(p)dτ , (7.13)
M ∂p 3

where F1 = (F1pp + F1pn )/2 with F1pp = F1nn 6= F1pn = F1np being the coefficients at l = 1 for the
interaction of pp, nn and np quasiparticles (see Eqs. (7.7) and (7.9)). It is seen that due to the inter-
action between quasiparticles any motion of particles is accompanied by a backflow of quasiparticles

66
(corresponding to the term proportional to F1 ). From Eqs. (7.3) and(7.13) it follows that the effective
mass M ∗ is equal to:
F1
 
M∗ = M 1 + . (7.14)
3
For the isovector current j1 one obtains in the same way:

∂ǫ̂ p 1 + F1′ /3
Z Z
j1 = Tr τ3 n̂(p) dτ = Tr τ3 δn̂(p)dτ . (7.15)
∂p M 1 + F1 /3

The electromagnetic current is given by jem = (e/2)(j0 + j1 ). Therefore,

e p 1 F1′ − F1
Z  
jem = Tr 1 + τ3 + τ3 δn̂(p)dτ . (7.16)
2 M 3 1 + F1 /3

Note that F1′ = (F1pp − F1pn )/2. The application of a relativistic approach [88, 259, 260]leads to some
modification of the mentioned results. A review of the numerous articles devoted to this subject can
be found in [88].
In order to obtain an information on the parameters of Fermi liquid theory for nuclei it is possible
to start from a pion-exchange and ρ-meson-exchange interaction in the Born approximation. Then
one can apply the Brueckner approach to take into account higher orders of perturbation theory with
respect to the interaction between particles (see [87, 261–263] and the review [264]). The operator for
the one-pion-exchange potential is of the form

f2
Vπ (p) = − (σ · p)(σ ′ · p)τ · τ ′ = (7.17)
m2π (m2π + p2 )
" #
f2 ST 1 ′ m2π
− + σ · σ − σ · σ′ τ · τ ′ ,
m2π m2π + p2 3 3 (m2π + p2 )

where ST = (σ · p)(σ ′ · p) − (σ · σ ′ ) p2 /3. The potential arising from ρ-exchange is

fρ2
Vρ (p) = − [σ · σ ′ p2 − (σ · p)(σ ′ · p)] τ · τ ′ = (7.18)
m2ρ (m2ρ + p2 )
fρ2 2m2ρ
" #
ST 2 ′
− 2 − 2 + σ · σ − σ · σ′ τ · τ ′ .
mρ mρ + p 2 3 3 (m2ρ + p2 )

Taking into account ω- and σ-exchange potentials one is lead to replacing fρ2 in (7.18) by f˜ρ2 = 0.4 fρ2
[87]. Equations (7.17) and (7.18) correspond to the contributions to the nuclear exchange potential
given in Eq. (3.12), except that here a separation into central and tensor parts is given.
Brown and Rho [87] arrived at a quantitative derivation of the effective mass M ∗ making use of
such a model, where the isospin dependent interactions between nucleons consist of π and ρ exchanges.
The effective mass M ∗ is obtained by evaluating the self energy stemming from the Fock term. The
results of this investigation are summarized in the following. Writing ǫ(p) as

p2
ǫ(p) = + Σ(p, ǫ(p)) , (7.19)
M
where Σ(p, ǫ(p)) is the quasiparticle self-energy. Then using the definition of the effective mass (7.3)
one has [87]:
1 1 1 ∂Σ(p, ǫ(p)
∗
= + (7.20)
M M pF ∂p p=pF

67
Substituting this relation into (7.14) and calculating the one-pion-exchange contribution to the self-
energy, one obtains an expression for the pion contribution to F1 [87]:
" #
9f 2 M ∗ m2π + 2p2F m2π + 4p2F M∗
F1 π =− 2 ln − 2 = −0.46 . (7.21)
8π pF 2p2F m2π M

The contribution of ρ-exchange to F1 reads [87]:

M∗
F1 ρ = −0.56 . (7.22)
M
Taking a sum of (7.21) and (7.22) and then using (7.14) one has:
M∗
= 0.75 , F1 = −0.76 . (7.23)
M
The numerical value for M ∗ /M is in a good agreement with the experimental value given in Table 4.2
for the mass region A = 197–209 of spherical nuclei. It is interesting to note that the A-dependence
of the effective mass M ∗ also finds a satisfactory explanation in the work of Brown and Rho [87].
The explanation which can be extracted from that work is as follows. First of all we have to realize
that the excellent agreement between theory and experiment found in (7.23) and the A = 197–209
mass range discussed in Table (4.2), respectively, partly comes as a surprise because in its theoretical
derivation it was assumed [87] that most of the difference M − M ∗ comes from the Fock terms for
π− and ρ− exchange, whereas it was discussed in that paper that quasiparticle-phonon couplings also
have an effect on the difference M − M ∗ but with the opposite sign. The mathematical procedure of
treating the two effects is very much the same. The only difference is that the pion propagator has to
be replaced by the appropriate energy denominator for the intermediate state involving quasiparticle
plus vibration. The essential point is that the contributions from couplings to vibrations leads to
effects close to the nuclear surface only. High-l orbitals are confined by the centrifugal barrier more to
the interior of nuclei than those of low-l ones [87]. Since states with high angular momentum dominate
at the 208 Pb shell closures this property is exactly what we need to understand the A-dependence of
M ∗ /M with a minimum in the 208 Pb mass range.
The operators of the pion-exchange and ρ-exchange potentials are proportional to

τ · τ ′ = 2(τ+ τ−′ + τ− τ+′ ) + τ3 τ3′ ,

where τ+ and τ− are isospin raising and lowering operators, respectively. Therefore, the following
relation between F1pp and F1pn is obtained [87]:

F1pn = 2F1pp (7.24)

Starting from Eqs. (7.17) and (7.18) one can obtain the contributions of pion-exchange and ρ-
exchange to the quasiparticle-quasiparticle interaction (7.7). For this purpose one can calculate the
forward scattering amplitude of a quasiparticle with p > pF and a hole with momentum p < pF .
After the Fourier transform to configuration space the momentum independent terms in (7.17) and
(7.18) are proportional to δ(r) . Due to the strong short-range repulsion from the ω-meson, the wave
function vanishes at zero distance between the nucleons. Therefore, the momentum-independent terms
in (7.17) and (7.18) can be omitted. Neglecting also the tensor part of the interaction (proportional
to ST ), we have [264] :
f2
fˆπ (p, p′ ) = σ · σ′ τ · τ ′ + (7.25)
3 m2π
f2
(−9 + 3σ · σ′ + 3τ · τ ′ − σ · σ ′ τ · τ ′ ) .
12 (m2π + (p − p′ )2 )

68
Here the first term corresponds to particle-hole annihilation and the second one to particle-hole ex-
change. In the last case the following identities have been used:
3 1
σ αβ σ δγ = δαγ δδβ − σαγ σ δβ
2 2
and the same for isospin matrices. The contribution of ρ-exchange has the form:

2f˜ρ2
fˆρ (p, p′ ) = σ · σ′ τ · τ ′ + (7.26)
3 m2ρ
f˜ρ2
(−9 + 3σ · σ′ + 3τ · τ ′ − σ · σ ′ τ · τ ′ ) .
6 (m2ρ + (p − p′ )2 )

Multiplying both sides of (7.25) by P1 (x) = x = p · p′ /p2 and taking the integral over x from minus
one to one, we obtain expressions for the Landau parameters at l = 1. As it should be, the result for
F1 π agrees with (7.21). It is clear from (7.25) and (7.26) that
1 1
F1′ = G1 = − F1 = 0.25 , G′1 = F1 = −0.08 . (7.27)
3 9
Even though F1′ is not known with great precision, a 10–20 per cent uncertainty in it [96] only has
a 1–2 per cent effect in the quantities of interest here. The discussion of the Landau parameters at
l = 0 as well as the contribution of tensor interaction can be found in the review [264].
◦
Let us now pass to the consideration of the enhancement constant κGDR and its connection with
the parameters of the Fermi liquid. It was shown in [86] that without taking into account tensor
correlations the following relation is valid for the modified TRK sum rule:
Z∞
◦ e2 N Z ◦
σ GDR (ω) dω = 2π 2 (1 + κGDR ) =
M A
0
 
1 ′
e2 NZ 1 e2N Z 1 + 3 F1
 
2π 2 1 + F1′ = 2π 2
, (7.28)
M∗
 
A 3 M A 1 + 1 F1
3

◦
where σ GDR (ω) is the unretarded electric giant-dipole cross section for nuclear photoabsorption and
◦ ◦
κGDR is the unretarded enhancement constant. It is seen from (7.28) that κGDR is related not only
◦
to the effective mass M ∗ but also to the constant F1′ . Besides, if F1pn = 0 then κGDR = 0 (in this case
◦
F1′ = F1 ). Equation (7.28) gives a relation for κGDR in the form:
1 ′
◦ 3 F1− 31 F1
κGDR = . (7.29)
1 + 13 F1
◦
Alternatively, one may also express the ratio M ∗ /M of the effective mass and the free mass via κGDR .
Using Eqs. (7.14), (7.24) and (7.29) one obtains
−1
M∗ 3◦

= 1 + κGDR . (7.30)
M 4
Using the relations given above it is easy to see that the theoretical results for M ∗ /M , F1 and F1′ are
◦
nicely consistent with the experimental quantity κGDR = 0.46 (see also Table 4.2).
We now return to tensor correlations. The calculation of the total enhancement constant κ with
the use of its representation as a double commutator [172] gives κ ∼ 1, if the tensor correlations in

69
the ground state are taken into account. In this paper [172] it was noted that the total enhancement
constant κ consists of two pieces, κ = κ′ + ∆κ, where only κ′ has a relation to the Sachs magnetic
◦
moment. Therefore, we have to conclude that the quantities κGDR and κ′ may be identified with
each other. One consequence of this identification is that the short-range tensor correlations giving
rise to ∆κ are not important for giant resonances as anticipated above. In fact short-range tensor
correlations give the main contribution to the quasi-deuteron part of the total enhancement constant
(see [172]).
◦
We present here a simple derivation of (7.28). The cross section σ GDR is of the form:

◦ 4π 2 X
σ GDR (ω) = ω | hn| d |0i |2 δ(E0 + ω − En ) , (7.31)
3 n

where d is the dipole moment operator with respect to the center of mass of the nucleus,
NX ZX
d=e ri p − e ri n . (7.32)
A i A i

The additional indices p and n indicate that in that term the sum is taken over only protons and only
neutrons, respectively. Using the relation jem = i[H, d] , we get
Z∞
◦ 2π 2
σ GDR (ω) dω = h0| [d, jem ] |0i . (7.33)
3i
0

It follows from (7.16) that in the c.m. frame the electromagnetic current is given by
    
e  ◦ X X pj p + pj n ◦ GDR X X pj p + pj n
jem = (2 + κGDR ) pi p − −κ pi n − 
2M i j
A i j
A
◦ " #
e(1 + κGDR ) N X ZX
= pi p − pi n . (7.34)
M A i A i

Substituting (7.32) and (7.34) into (7.33) and evaluating the commutator, we get (7.28).
Next we discuss the low-energy limit of the Compton scattering amplitude. Within the framework
of second-order perturbation theory the resonance amplitude in the long-wavelength approximation is
given by:
X  hf | jem · ǫ2 |νi hν| jem · ǫ1 |ii hf | jem · ǫ1 |νi hν| jem · ǫ2 |ii

R(ω, θ) = − + (7.35)
ν E0 − Eν + ω E0 − Eν − ω

Again expressing the current via the commutator of the electric dipole moment with the Hamiltonian
it is easy to extract the low-energy limit of the resonance amplitude:
1
R(0, θ) = {h0| [ǫ1 · d, ǫ2 · jem ] |0i + h0| [ǫ2 · d, ǫ1 · jem ] |0i} . (7.36)
2i
Then it follows from (7.32) and (7.34) that

e2 ZN ◦
R(0, θ) = ǫ1 · ǫ2 (1 + κGDR ) . (7.37)
M A
Using the low-energy theorem we may also obtain the low-energy limit of the relevant seagull amplitude
Z 2 e2 Ze2 N◦
 
B(0, θ) + SGR (0, θ) = −ǫ1 · ǫ2 − R(0, θ) = −ǫ1 · ǫ2 1 + κGDR . (7.38)
MA M A

70
It is possible to derive Eq. (7.38) directly, when the explicit expression, Eq. (3.61), for the total seagull
amplitude is translated into the language of Fermi liquid theory. One has
i
B(0, θ) + SGR (0, θ) = {hf | [ǫ1 · D , ǫ2 · J] + [ǫ2 · D , ǫ1 · J] |ii} (7.39)
2
with
eZ X eZ X
D=d+ (ri p +ri n ) , J = jem + (pi p +pi n ) . (7.40)
A i AM i
Calculating the commutator we obtain again Eq. (7.38).
When nucleons inside a nucleus have an angular momentum l they also produce an angular (Sachs)
magnetic moment. Meson exchange currents modify this angular magnetic moment and it is easy to
derive from Eq. (7.34) that for the proton this modification is given by

N ◦ GDR
δglmeson (p) = κ . (7.41)
A
For the neutron one has
Z◦
δglmeson (n) = − κGDR . (7.42)
A
As a result, we obtain that the isovector part of the angular momentum correction is given by

(iv) 1 1◦
δgl = (δglmeson (p) − δglmeson (n)) = κGDR . (7.43)
2 2
(iv) ◦
This relations between δgl and κGDR is known as the Fujita-Hirata relation [87, 172, 173, 265–267].
A detailed investigation of it has been performed in [172] with special attention on the role of tensor
correlations.
The Fujita-Hirata relation opens the possibility to compare different measurable quantities related
to meson exchange currents with each other as there are the strength of the giant-dipole resonance and
the magnetic moments of nucleons in a nucleus. Taking the result for nuclei with a 208 Pb double-magic
◦
core, (cf. (4.48)) κGDR = 0.46 ± 0.05, we arrive at

δglmeson (p) = 0.28 ± 0.02, δglmeson (n) = −0.18 ± 0.02 (7.44)

corresponding to
(iv)
δgl = 0.23 ± 0.03. (7.45)
Given the validity of the Fujita-Hirata relation we may consider these quantities in (7.44) and (7.45)
as experimental results which may be compared with predictions. We find good agreement with the
prediction of Brown and Rho [87], viz.

δglmeson (p) = 0.27, δglmeson (n) = −0.17 (7.46)

and with the prediction of Hyuga et al. [268], viz.

δglmeson (p) = 0.27, δglmeson (n) = −0.14. (7.47)

The quantities δglmeson are not directly observable in measurements of magnetic moments because
of the effects of higher-order configuration mixing. For a certain fraction of time protons and neutrons
are correlated to each other producing an additional angular momentum, and related to this a correc-
tion δglhigh to the orbital g-factor. An analysis of mesonic effects on nuclear magnetic moments has

71
been carried out by Yamazaki [269]. After correcting for first-order configuration mixing he arrives at
anomalous orbital g-factors of

δglobs (p) = 0.15 ± 0.02, δglobs (n) = −0.05 ± 0.02, (7.48)

consistent for all measured g-factors of nuclear states having a 208 Pb core. Comparing these quantities
with our results of (7.44) we find

δglobs − δglmeson = −(0.13 ± 0.03) τ3 , (7.49)

[τ3 (p) = +1, τ3 (n) = −1]. This result (7.49) may be compared with the predicted [268, 270] effects of
higher-order configuration mixing, viz.,

δglhigh (p) = −0.15, δglhigh (n) = +0.10. (7.50)

We notice that the “experimental” result (7.49) is purely isovector, whereas the predictions (7.50)
suggests some isoscalar component. However, in view of the error given (7.49) it is not possible to
draw any conclusion from this difference.
A summary of these results is given in Fig. 7.1. The first column shows the result of the measure-
ment of magnetic moments carried out by Yamazaki [269]. The second column shows the quantities
◦
δlmeson obtained from the experimental κGDR on the basis of the Fijita-Hirata relation together with
higher-order correction δglhigh obtained as a difference of δlmeson in column II and δglobs in column I.
Column III shows the predictions of Hyuga et al. [268] and column 4 the prediction of Brown and
Rho [87].

Figure 7.1: Summary of results on the Fujita-Hirata relation. Column I: Experimental result of
◦
Yamazaki [269]. Column II: δgLmeson calculated from the experimental κGDR and δgLhigh obtained as
a difference between δlmeson of column II and δgLobs of column I. Column III: Predictions of Hyuga et
al. [268]. Column IV: Predictions of Brown and Rho [87].

The quasiparticle aspect of nucleons in a nucleus may be discussed in different forms. Instead of
considering the dynamics of nucleon interactions explicitly, as done in Landau’s Fermi liquid theory
and its interpretation by Brown and Rho [87] there are attempts to view the role of the nuclear
environment as a modification of the vacuum [271]. The advantages of this approach are that it is
possible to make use of different models like the chiral mean field and the Walecka model of nuclear
matter [272] and to arrive at new predictions which are not that easy to obtain within the framework
of the original Fermi liquid theory. One disadvantage is that these new vacua are less familiar than
the building blocks of Fermi liquid theory, thus making it desirable to build bridges between these
different aspects of nuclear matter. The most recent works going into this direction are those of G.E.
Brown [96], Friman and Rho [97], Brown and Rho [98] and Rho [273].
In [96] essentially the same effective mass M ∗ , which is related to the Migdal parameter F1 as in
Eq. (7.14), is also related to a (negative) scalar mean field Φ of the Walecka type [272] via

M ∗ = M + gσN N Φ. (7.51)

In terms of (7.51) the primary process is the lowering of the mass from M to M ∗ . The physical
reason for this lowering of the mass is that the scalar mean field with the coupling constant gσN N
gives an attractive scalar potential. Both descriptions, viz. the velocity dependence of quasi-particle
interactions in the Fermi liquid and the lowering of the in-medium mass lead to the same increase of
the quasiparticle velocity v = p/M ∗ as compared to p/M and these two interpretations are equivalent.

72
This equivalence is seen in a relativistic formulation, where Lorentz invariance provides a connection
between the two interpretations. These matters are reviewed in Brown et al. [88].
A further aspect of the in-medium modification of the nucleon-nucleon interaction has been inves-
tigated Brown and Rho in [91]. They discuss the enhancement of the ρ meson tensor coupling to the
nucleon. The ρ-meson is assumed to acquire an effective mass inside the nuclear medium with
m∗ρ M∗ 3
≃ ≃ (7.52)
mρ M 4

as suggested by the general concept of Brown and Rho scaling [92] and by the Fermi liquid theory
◦
together with the experimental value of κGDR = 0.46. It is shown [91] that this implies that in the
nuclear medium the ρ-meson tensor force is enhanced by a factor 16/9.

73
8 Dispersion relations at fixed momentum transfer for nuclear Comp-
ton scattering

In the previous sections phenomenological approaches have been applied in order to obtain the angular
dependence of the Compton scattering amplitude. One of the main guidelines has been the fulfilment of
a dispersion relation in forward direction, together with the requirement that the resonance parts of the
amplitude should vanish in the high-energy limit at arbitrary scattering angle. Another ingredient in
this construction of the Compton amplitude has been the correct low-energy behaviour of each (electric
and magnetic) multipole. In this section we investigate the angular dependence of the nuclear Compton
amplitude within a different approach based on a dispersion theory for the invariant amplitudes at
fixed momentum transfer. This approach gives a correct form of dispersion relations for the resonance
and partial-wave amplitudes which sometimes are written too naively.

8.1 General structure of the Compton amplitude

In the case of spin-0 nuclei, which are considered in this review, the general form (see e.g. [99]) of the
Compton scattering amplitude T is
n    o
T (ω, t) = ω 2 A1 (ω, t) + A2 (ω, t) ǫ1 · ǫ2 + A1 (ω, t) − A2 (ω, t) s1 · s2 , (8.1)

where A1 (ω, t) and A2 (ω, t) are invariant amplitudes and

t = −q2 = −2ω 2 (1 − cos θ) (8.2)

is the square of momentum transfer. Here all recoil corrections are neglected. The form (8.1) for the
amplitude is most convenient, when a discussion of low-energy properties, such as sum rules and the
behaviour of multipole amplitude below resonance energies, is intended. The invariant amplitudes Ai ,
which appear in a Lorentz-invariant form of the Compton amplitude as soon as unphysical (kinemati-
cal) singularities and zeroes are eliminated, are related to the helicity non-flip amplitude T1,1 and the
helicity flip amplitude T1,−1 for a circularly polarized photon via

T1,1 (ω, t) T1,1 (ω, t) T1,−1 (ω, t) T1,−1 (ω, t)

A1 = 2
=2 , A2 = = −2 . (8.3)
ω (1 + cos θ) (4ω 2 + t) 2
ω (1 − cos θ) t
Due to the conservation of angular momentum only T1,1 differs from zero in forward direction. For
linearly polarized photons the two non-vanishing amplitudes are

T| | = T1,1 − T1,−1 , T⊥ = T1,1 + T1,−1 , (8.4)

where T| | (T⊥ ) corresponds to the polarization of both photons being parallel (perpendicular) to the
scattering plane.
Furthermore, we will apply the usual partial wave expansion of the helicity amplitudes (see e.g.
[274, 275]),
∞
j
(ω) dj1,±1 (θ),
X
T1,±1 (ω, θ) = T1,±1 (8.5)
j=1

where the d-functions are given by the formulae [275, 276]

j θ θ
d1,1 (θ) = cos2 F (−j + 1, j + 2, 1, sin2 ),
2 2
j j(j + 1) θ θ
d1,−1 (θ) = sin2 F (−j + 1, j + 2, 3, sin2 ). (8.6)
2 2 2

74
Here F (a, b, c, x) = 1 + (ab/c)(x/1!) + . . . is the hypergeometric function. Equations (8.6) can be
rewritten as
1±z  ′ 
dj1,±1 (θ) = Pj (z) + (z ∓ 1)Pj′′ (z)
j(j + 1)
1∓z
= ±Pj (z) + P ′ (z), z = cos θ, (8.7)
j(j + 1) j
dj1,−1 (θ) = (−1)j−1 dj1,+1 (π − θ).
j
The functions dλλ′ satisfy an orthogonality relation:

Z 1 ′ 2
j j
dλλ′ (θ)dλλ′ (θ) d(cos θ) = δjj ′ . (8.8)
−1 2j + 1
j
The partial waves T1,±1 (ω) contain electric (T Ej ) and magnetic (T M j ) multipoles:
j
T1,±1 (ω) = T Ej ± T M j , (8.9)

what can be easily inferred from (3.45). Due to the optical theorem the imaginary parts of the partial
amplitudes T Ej and T M j give rise to partial absorption cross sections σ Ej and σ M j via
ω λj
σ (ω) = Im T λj (ω), (8.10)
4π
where λ = E, M .
In order to obtain a definite multipole amplitude, let us start from a dispersion relation at fixed
momentum transfer t for the invariant amplitudes Ai . From Eq. (8.3) one can expect that at fixed
t the amplitude A1 tends to zero as ω tends to infinity and hence satisfies an ordinary unsubtracted
dispersion relation. At the same time the amplitude A2 does not generally tend to zero and therefore
a dispersion relation for A2 should include either a subtraction or a safe regularization. The regular-
ization can be implemented by cutting the dispersion integral at some high maximal energy ωmax and
by adding an asymptotic contribution a(t) which is energy independent (but is momentum-transfer
dependent) at low and medium energies ω 2 ≪ ωmax 2 — see a parallel discussion of the nucleon case
in [99, 101]. For the sake of simplicity we disregard here such a regularization which can be easily
restored. Since the asymptotic piece a(t) does not depend on the energy, it does not affect the reso-
nance part R(ω, t) of the Compton scattering amplitude and enters only into the seagull S(ω, t), to
its helicity-flip energy-dependent part.
So, we write the dispersion relation for either amplitude Ai in the same form with the dispersion
integral taken up to infinity and without the explicit addition a(t) to the amplitude A2 :
Z∞
Z 2 e2 2 ω ′ dω ′
Ai (ω, t) = − + ImAi (ω ′ , t) . (8.11)
2AM ω 2 π ω ′2 − ω 2 − i0
0

Here the pole contribution (viz. the first term on the r.h.s.) ensures the Thomson limit for the total
amplitude T (ω, t).
In the following we apply the partial-wave expansion (8.5) and write the imaginary parts of Ai
through the partial absorption cross sections σ Ej and σ M j . Such a procedure may be invalid at
high |t| exceeding a binding threshold for disintegrating the nucleus into separate parts. In this
case we consider the resulting formulas in the sense of an analytical continuation. Nevertheless, this
procedure allows us to investigate the deviation of the exact partial scattering amplitudes from the
naive assumption of a purely Lorentzian form for its resonance part.

75
 The use of Eq. (8.3) leads to dispersion relations for the helicity amplitudes. With the help of Eq.
(8.5) and Eq. (8.10) for the r.h.s. of Eq. (8.11), one obtains
∞
∞ Z
dj1,±1 (θ ′ ) σ Ej (ω ′ ) ± σ M j (ω ′ )
( )
Z 2 e2 ω2 X
T1,±1 (ω, θ) = (1 ± cos θ) − + 2 dω ′ (8.12)
2AM 2π j=1 1 ± cos θ ′ ω ′2 − ω 2 − i0
0

where θ ′ is a function of θ, ω and ω ′ , due to the condition that the square of the momentum transfer
q 2 = −t is fixed,
t
ω ′2 1 − cos θ ′ = ω 2 (1 − cos θ) = − .


(8.13)
2
Here the θ ′ -dependent factors in the braces of Eq. (8.12) are polynomials in cos θ ′ and hence polyno-
mials in x = ω 2 /ω ′2 .
Let us multiply both sides of Eq. (8.12) by dJλ,λ′ and take the integral with respect to θ using Eqs.
(8.5) and (8.8). As a result we obtain

∞ Z ∞
EJ MJ Z 2 e2 ω2 X (±) σ EL (ω ′ ) ± σ M L (ω ′ ) ′
T (ω) ± T (ω) = − δJ,1 + 2 ΦJL (ω, ω ′ ) dω . (8.14)
AM 2π L=1 ω ′2 − ω 2 − i0
0

(±)
The function ΦJL has the following integral representation:

Z1
(±) 2J + 1 1 ± cos θ J
ΦJL (ω, ω ′ ) = d (θ)dL ′
1,±1 (θ ) d(cos θ). (8.15)
2 1 ± cos θ ′ 1,±1
−1

(±)
In all cases the function ΦJL is a polynomial in x, whoes properties follows from the orthogonality
relation (8.8) and can be summarized in the following way:

 0, J > L,

(±)
ΦJL (ω, ω ′ ) = xJ−1 , J = L, (8.16)
(±)
xJ−1 (1 − x)Φ̃JL (x), J < L,


(±)
where Φ̃JL (x) is a polynomial of the (L−J −1)th power in x. By applying Eq. (8.16) to the dispersion
relation (8.14) via (8.15) we obtain dispersion representations for the multipole amplitudes of nuclear
Compton scattering:

ω ′ 2 σ EJ (ω ′ )
2J 
(Ze)2 1 ∞ ω
Z
EJ
T (ω) = − δJ1 + 2 +
AM 2π 0 ω′ ω ′ 2− ω 2 − i0
Xh i
σ EL (ω ′ )AJL (x) + σ M L (ω ′ )BJL (x) dω ′ , (8.17)
L>J

and

ω ′ 2 σ M J (ω ′ )
∞ 2J 
1 ω
Z
T M J (ω) = +
2π 2
0 ω ′ 2 − ω 2 − i0 
ω′
Xh i
σ M L (ω ′ )AJL (x) + σ EL (ω ′ )BJL (x) dω ′ . (8.18)
L>J

Here AJL (x) and BJL (x) are polynomials of (L − J − 1)th power in x, with x = (ω/ω ′ )2 :
1 (+) (−) 1 (+) (−)
AJL (x) = [Φ̃JL (x) + Φ̃JL (x)], BJL (x) = [Φ̃ (x) − Φ̃JL (x)]. (8.19)
2 2 JL

76
The first few of them are given by

A12 (x) = 2, B12 (x) = −1,

7 21
A13 (x) = 2 − 4 x, B13 (x) = − 52 + 15
4 x,
11 77 77 2
A14 (x) = 2 − 4 x+ 5 x , B14 (x) = − 92 + 63 63 2
4 x− 5 x ,
15
A23 (x) = 4 , B23 (x) = − 54 ,
39
A24 (x) = 4 − 13x, B24 (x) = − 21
4 + 7x,
28
A34 (x) = 5 , B34 (x) = − 75 . (8.20)

Since each partial absorption cross section is nonzero only above an energy threshold corresponding
to the first excitation level of the nucleus, no divergence at ω ′ = 0 appears in Eqs. (8.17) and (8.18).
Similar multipole dispersion relations are used in the analysis of pion photoproduction data [277]. The
partial-wave series in (8.17), (8.18) may become divergent at high ω and should be understand in the
sense of an analytical continuation in such a case.
One can see from the relations (8.17) and (8.18) that at ω → 0 the partial photoabsorption
cross sections σ EJ and σ M J give contributions to the partial amplitudes T EJ and T M J , which are
proportional to ω 2J . The asymptotic contribution a(t) which is disregarded in the above consideration
and which makes equal but opposite additions to T EJ and T M J does not change this threshold
′
behavior. The nonzero contributions of σ λ L to T λJ with L > J is another important property of the
partial scattering amplitudes, which follows from (8.17) and (8.18). These properties may be used as
a starting point to investigate the deviation of multipole amplitudes from a purely Lorentzian form.

8.2 Resonance and seagull amplitudes at fixed momentum transfer

In the previous section we obtained a representation for the total Compton scattering amplitude in
non-forward direction, Eq. (8.12). Starting from this expression we now want to construct the t-
dependence of the giant resonance, the quasi-deuteron and the seagull parts, which enter into the
Compton amplitude TA . The additional contribution TN , which contains the electromagnetic po-
larizabilities of the nucleon, is not considered here. This decomposition should be consistent with
the definitions (3.21), (3.22) and (3.29) given for the case of forward direction. Here our dispersion-
theoretical method (as opposed to the operator-based techniques found e.g. in [70, 121, 278]) can be
efficiently used to reconstruct these contributions to the Compton amplitude on the base of an exper-
imental information. The t-dependence can consistently be constructed due to the use of dispersion
relations at fixed momentum transfer.
Let us define the contributions of each of the three parts to the total (nuclear) helicity amplitudes
A
T1,±1 by
A GR QD tot
T1,±1 (ω, t) = R1,±1 (ω, t) + R1,±1 (ω, t) + S1,±1 (ω, t). (8.21)

The representations of the two resonant parts should fulfill certain requirements. In the dispersion
GR (ω, t) instead of the full partial wave cross section σ λj only
integral contributing to the amplitude R1,±1
λj
σGR should appear. In all other aspects it has the same form as the one in Eq. (8.12). Furthermore,
some function of t, which does not depend on ω and ensures vanishing RGR at high ω, has to be
QD
added. The amplitude R1,±1 (ω, t) should have similar properties. It also is represented as a sum of a
λj tot (ω, t), the
dispersion integral containing σQD and some function of t. As for the seagull amplitude S1,±1
integration in the dispersion contribution should start from pion mass mπ with the integrand depending
λj
on the difference σ λj − σQD . The seagull amplitude has no resonance structure and corresponds to

77
the scattering by an object of small size in comparison with the nuclear radius R. More specifically, it
can be argued (see e.g. [68, 72] and the discussion in sections 3.1 and 3.2) that two distinct processes
contribute to Stot , namely the Thomson scattering by individual nucleons (kinetic seagull) and the
scattering by correlated nucleon pairs (mesonic seagull). For the latter, the characteristic size is of
the order of 1/mπ . The comparison of the nuclear radius R with the relevant interaction range for the
processes contributing to the seagull amplitude leads to the conclusion that the t-dependence of the
total seagull amplitude is similar to the nuclear form factor. We will investigate this point in more
detail in section 9. Finally, using Eq. (3.43) and taking the above-mentioned properties into account,
we get
Ze2 N
  
S1,±1 (ω, θ) = (1 ± cos θ) − F1 (t) + κ F2 (t) + (8.22)
2M A
∞ Z∞ j
ω2 X d1,±1 (θ ′ ) ∆σ Ej (ω ′ ) ± ∆σ M j (ω ′ ) ′

dω ,
2π 2 j=1 1 ± cos θ ′ ω ′2 − ω 2
mπ

with ∆σ(ω) = σtot (ω) − σQD (ω) − AσN bound (ω), the function F (t) being the nuclear one-body form
1
factor and F2 (t) corresponding also to a nuclear form factor but taking into account the finite size of
a correlated nucleon pair (cf. sections 3.5 and 9).
Let us discuss the behaviour of Eq. (8.22) in the case ω ≪ mπ . Then (−t) ≪ m2π due to (8.13).
Since the integral with respect to ω ′ starts from mπ , the same relation shows that also θ ′ ≪ 1. We
can omit ω 2 in the denominator of the integrand in Eq. (8.22). However, it is not possible also to
neglect θ ′ , which would correspond to t = 0, because as described above the t-dependence of any
contribution to the seagull amplitude should approximately be given by a nuclear form factor. Since
quasi-deuteron production occures from small regions of size ∼ 1/mπ ≪ R scattered over the whole
volume of the nucleus, many multipoles contribute to the integrand in (8.22), thus making the partial-
wave expansion almost useless. It is physically clear, however, that the t-dependence of the integrand
reflects a distribution of participating pn pairs in the nucleus, so that the integral is proportional to the
two-body form factor F2 (t). The coefficient of the proportionality can be identified with modifications
of the electric and magnetic polarizabilities of the nucleon, δα and δβ, caused by the mesonic seagull
amplitude:
( )
(1 ± cos θ) Ze2 N
 
S1,±1 (ω, θ) = − F1 (t) + κ F2 (t) + Aω 2 (δα ± δβ) F2 (t) . (8.23)
2 M A
The important quantities δα and δβ will be discussed in section 9 within a specific model, where it will
also be shown that form factors accompaning κ, δα and δβ in (8.22) are actually not identical, as a
consequence of a different configuration size of the correlated pn pairs determining these values. In this
section, however, we disregard this nontrivial feature and use the same form factor F2 (t) everywhere.
It is also possible to discuss this particular form (8.23) of the seagull amplitude on the level of nuclear
matrix elements, as was done in section 3.
Now, taking into account Eqs. (3.35), (3.41) and (8.22), we obtain an explicit representation of
GR (ω, t) we have
the resonance amplitudes from Eq. (8.12). For R1,±1

(1 ± cos θ) Ze2 Z N
  
GR
R1,±1 (ω, θ) = − + F1 (t) + κGR F2 (t) + (8.24)
2 M A A
∞
∞ Z dj ′ Ej Mj
)
ω2 X ′ ′
1,±1 (θ ) σGR (ω ) ± σGR (ω )
dω ′ .
π 2 j=1 1 ± cos θ ′ ω ′2 − ω 2 − i0
0

This representation of the giant resonance amplitude, in which form factors appear explicitly, is
convenient, as in the dispersive part each multipole contains the correct ω-dependence at low energies.

78
 QD
Similarly, for R1,±1 (ω, t) one has

(1 ± cos θ) e2 ZN

QD
R1,±1 (ω, θ) = κQD F2 (t) + (8.25)
2 M A
∞ Ej Mj
∞ Z
dj1,±1 (θ ′ ) σQD (ω ′ ) ± σQD (ω ′ )
)
ω2 X
dω ′ .
π 2 j=1 1 ± cos θ ′ ω ′2 − ω 2 − i0
0

In contrast to (8.24), the integral here involves many multipoles and therefore is less useful. It’s
physically clear again that the t-dependence of the integral mainly follows the two-body form factor
F2 .
Now we should fulfill the requirement that the resonance amplitudes tend to zero at high ω and
GR and RQD the external factor (1 − cos θ)/2 = −t/(4ω 2 ) tends
fixed t. Since in both amplitudes R1,−1 1,−1
to zero in this limit, they vanish independently of the explicit shape of the form factors. Moreover,
the asymptotic piece a(t) can also affect the helicity-flip amplitudes. Therefore, model-independent
restrictions appear only from the asymptotic behaviour of R1,1GR and RQD . In the limit ω → ∞ the
1,1
requirement of vanishing R1,1GR leads to

∞
Ze2 Z GR N
∞ Z
1 X dj1,1 (θ ′ )
 
− + F1 (t) + κ F2 (t) = 2 [σ Ej (ω ′ ) + σGR
Mj
(ω ′ )] dω ′ , (8.26)
M A A π j=1 1 + cos θ ′ GR
0

QD
while the same condition for R1,1 gives
∞
e2 N Z ∞ Z
1 X dj1,1 (θ ′ )
κ QD
F2 (t) = 2 Mj
[σ Ej (ω ′ ) + σQD (ω ′ )] dω ′ . (8.27)
M A π j=1 1 + cos θ ′ QD
0

GR of Eq. (8.24) via the absorption

Using Eq. (8.26) we can express the giant resonance amplitude R1,±1
λj
cross sections σGR only, without the explicit appearance of form factors. Putting t equal to zero in
Eqs. (8.26) and (8.27) we obtain again the sum rules (3.36) and (3.42), since
∞   ∞  
Ej Mj Ej Mj
X X
σGR (ω) = σGR (ω) + σGR (ω) , σQD (ω) = σQD (ω) + σQD (ω) .
j=1 j=1

Expanding both sides of (8.26) with respect to t and comparing the coefficients, we obtain additional
sum rules
Ze2 π 2 (n!)2 22n+1 D 2n E N D 2n E
 
r + κGR r = (8.28)
M (2n + 1)! 1 A 2
∞ Z∞
X (j + n + 1)! dω  Ej Mj

σ (ω) + σ (ω) , ,n > 0
j=n+1
j(j + 1)(j − n − 1)! ω 2n GR GR
0

where the ordinary definition of the average values of r 2n is used. In the particular case of n = 1 Eq.
(8.28) gives
∞ Z∞
Ze2 4π 2 D 2 E GR N dω  Ej
 D E  X 
2 Mj
r +κ r = (j + 2)(j − 1) σ (ω) + σ (ω) . (8.29)
M 3 1 A 2
j=2
ω 2 GR GR
0

Restricting ourselves to the lowest electric multipole on the r.h.s. of Eq. (8.28) we reclaim the usual
energy-weighted sum rules [279, 280]. In reality, only the lowest multipoles E1,M 1,E2 are experimen-
tally observed and taken into account in the analysis of scattering data [19, 32, 39]. For this case Eq.

79
(8.24) is of the following form:

(1 ± cos θ) Ze2 Z N GR
  
GR
R1,±1 (ω, θ) = − + F1 (t) + κGR F (t) + (8.30)
2 M A A 2
Z∞ " !#
ω2 dω ′ E1 ′ M1 ′ E2 ′ 2ω 2
σ GR (ω ) ± σ GR (ω ) + σ GR (ω ) 2 − (1 − cos θ) ∓ 1 .
2π 2 ω ′2 − ω 2 − i0 ω ′2
0

At very low energies it agrees with the standard formalism, as is discussed below. However, this
approximation fails at high energies where higher multipoles are not negligible — see section 10.
QD
If as before we assume that the quasi-deuteron contribution R1,±1 to the resonance amplitude
corresponds to the scattering by correlated nucleon pairs, its dependence on t is described by the form
factor F2 (t). From Eq. (8.25), together with (3.42), one has approximately
Z∞
QD 1 ± cos θ 1 ω ′2 dω ′ σQD (ω ′ )
R1,±1 = F2 (t) 2 (8.31)
2 2π ω ′2 − ω 2 − i0
0

For practical purposes the absorption cross section σQD (ω) may then be expressed using a phenomeno-
logical description as discussed in section 5. It is also possible to represent σQD as a Lorentzian with
a very big width [25, 34].
A straightforward application of the fixed-t dispersion relations discussed in this section is the
dispersion representation of the nuclear polarizabilities ᾱA and β̄A . Expanding the amplitudes T E1
and T M 1 from Eqs. (8.17) and (8.18) with respect to ω up to the order ω 2 we obtain [99, 226]
(
1
Z ∞ dω 1 X EL

L(L + 1)
 
ᾱA = σ E1 (ω) + σ (ω) 1 +
2π 2 0 ω2 2 L>1 2
L(L + 1)
 
ML
+ σ (ω) 1 − ,
2
Z ∞ (
1 dω 1 X EL L(L + 1)
  
M1
β̄A = σ (ω) + σ (ω) 1 −
2π 2 0 ω 2 2 L>1 2
L(L + 1)
 
ML
+σ (ω) 1 + . (8.32)
2

One can see that not only σ E1 and σ M 1 give a contribution to the electromagnetic polarizabilities,
but also higher multipoles of the photoabsorption cross section. In general, the asymptotic piece a(t)
also contributes and gives equal but opposite additions ±a(0) to ᾱA and β̄A , respectively. There are
also recoil corrections ∼ 1/AM [99, 226] which are again equal but opposite for ᾱA and β̄A . When
in Eq. (8.32) the sum ᾱA + β̄A is taken, one recovers the usual Baldin-Lapidus sum rule. If only the
lowest three multipoles are taken into account, one has
1 ∞
Z
dω

ω E1

ᾱA = σ E1 (ω) + 2σ E2 (ω) + σ (ω) + a(0),
2π 2 0 ω 2 AM
Z ∞
1 dω ω E1
 
M1 E2
β̄A = σ (ω) − σ (ω) − σ (ω) − a(0), (8.33)
2π 2 0 ω2 AM
where the asymptotic contributions and recoil corrections are explicitly written for the sake of com-
pleteness though the latter are negligible for nuclei with A ≫ 1. Evaluating these formulas with a
nonrelativistic sum rules technique and using a closure, one can reproduce known answers for the
polarizabilities of nonrelativistic systems [100,281]. Clearly, when in Eqs. (8.32) and (8.33) one passes

80
from the total cross sections σ λJ to the giant resonance contributions σGR
λJ the quantities ᾱ
GR and β̄GR
are obtained. In particular, the diamagnetic correction (4.56) is derived using the sum rule (8.29).
Furthermore, it is instructive to have a closer look at the interplay of the seagull and resonance
parts in the total amplitude. To this end we again restrict ourselves to the lowest three multipoles.
We make use of the previously defined quantities R̃λj (see Eqs. (4.2) and (4.4)), i.e.
Z∞
λj ω2 σ λj (ω ′ ) dω ′ λj
R̃ (ω, θ) = 2 g (θ) (8.34)
2π ω ′2 − ω 2 − i0
0

to express the giant resonance amplitude for this special case. From Eqs. (8.30) and (8.26) we obtain

RGR (ω, θ) = R̃E1 (ω, θ) + R̃E2 (ω, θ) + R̃M 1 (ω, θ)

Z∞
1  
+ 2 dω ′ σGR
E1
(ω ′ ) + σGR
M1 ′ E2
(ω ) + σGR (ω ′ ) gE1 (θ). (8.35)
2π
0

Similarly, the seagull amplitude has the form

B(ω, θ) + SGR (ω, θ) =

Z∞
 
Z 2 e2 1  
− gE1 (θ)  + 2 dω ′ σGR
E1
(ω ′ ) + σGR
M1
(ω ′ ) + σGR
E2
(ω ′ ) 
AM 2π
0
Z∞
ω2 dω ′ E2
 
+ σGR (ω ′ ) 2gE1 (θ) − gM 1 (θ) − gE2 (θ) . (8.36)
2π 2 ω ′2
0

If one takes into account the modified TRK sum rule, Eq. (3.36), for the last term on the r.h.s. of Eq.
(8.35), this expression is in agreement with the formulation of the giant resonance amplitude given in
Eq. (4.2). As before, when considered at fixed momentum transfer, the amplitude RGR (ω, θ) vanishes
in the high-energy limit. However, the individual multipole contributions from E2 and M 1 do not
vanish in this limit. Note that the applicability of fixed-t dispersion relations is restricted to small
momentum transfer, since due to the relation (8.13) for some ω ′ the angle θ ′ becomes imaginary at high
−t and the cosine cos θ ′ runs away the range of convergence of partial-wave series (viz. the Lehmann
ellipse). Thus, for large ω and t, but fixed θ the sum with respect to j in Eq. (8.12) is not necessarily
convergent. Unfortunately, this is precisely the regime, where the effects of retardation as discussed in
section 4.4 become important. Therefore, at this point one has to get back to the phenomenological
discussion of retardation given in section 4.4.

81
9 The mesonic seagull amplitude

It was pointed out above that the seagull amplitude has no imaginary part below pion threshold.
It contains contributions from two fundamentally different physical sources, Thomson scattering on
individual nucleons inside the nucleus and the scattering by correlated nucleon pairs. Such correlations
occur as a result of the nucleon-nucleon interaction, which can be described in terms of mesonic
exchange between the nucleons. As discussed before, the amplitude for the first process contains the
form factor F1 (t), while the second contribution is proportional to F2 (t). Since the correlation radius
is of the order 1/mπ , i.e. much smaller than the nuclear radius R, one can expect that the form factor
F2 does not differ tremendously from F1 . Nevertheless, this difference produces an observable effect.
We will confirm this statement within a model calculation. The same model will be used to investigate
the different contributions to the enhancement constant κ, as well as the energy-dependence of the
mesonic seagull amplitude, which is expressed in terms of polarizability modifications δα and δβ and
the influence of the finite nuclear size on all these quantities.

9.1 Construction of the mesonic seagull amplitude for nuclear matter

In this section the contribution of correlated nucleon pairs to the seagull amplitude, Eq. (8.23), will
be discussed on the basis of the results obtained in [168]. There the contribution of mesonic exchange
currents to the nuclear Compton scattering amplitude was investigated within the framework of a
(modified) Fermi gas model of nuclear matter in the non-relativistic limit. Such an approach was
suggested in [70, 282] and for the case of forward direction used in [80, 83]. There it was also shown
that the nuclear degrees of freedom can be taken into account by representing the amplitude S as a
convolution of a two-body spin-isospin correlation function with matrix elements corresponding to the
amputated irreducible Feynman diagrams for meson exchange. Such a correlation function consists
of a central and a tensor part. In a pure Fermi gas model the tensor correlator is zero, but the
importance of tensor correlations was pointed out in [83, 87, 172]. In its effect on the enhancement
constant κ, for example, the central correlator makes a comparatively small contribution due to the
strong compensation between the different Feynman diagrams taken into account. In [83] the results
of numerical calculations for the correlators in nuclear matter were used and compared to a similar
calculation based on the variational principle [283].
In [168] corrections to the pure Fermi gas wave function were calculated up to first order in a non-
covariant perturbation theory. As a result of this model calculation the values of the enhancement
constant κ and the polarizability modifications δα and δβ were obtained and the central and tensor
part of the correlation function due to pion exchange between the nucleons were represented in an
analytical form. In section 9.3 we will use this representation as a starting point to investigate the
difference between the form factors F1 and F2 . In the case of the present section, for the nuclear
radius tending to infinity, the correlation function is proportional to F1 . Since in [168] a Fermi gas
model has been used, these results are only applicable for heavy nuclei, where it can be expected that
the quasi-classical approximation is valid. The effects of a finite nuclear size as well as of a realistic
nuclear density have been investigated in [193] and will also be discussed in section 9.3.
Let us start our discussion from the static Hamiltonian for the interaction of the nucleon with the
pion field φ (see e.g. [173]),
f
HπN N = − (σ · ∇) (τ · φ) , (9.1)
mπ
together with a minimal coupling to the photon field. For all coupling constants we use the same
notation and values as given in [173], in particular f 2 /(4π)=0.08. The mesonic seagull amplitude S

82
can be written in the following form [70, 80, 83, 168]:
dQ
Z
S= F ij (Q) Tij (Q). (9.2)
(2π)3
ij
The contribution T(π) of π-meson exchange to Tij , corresponding to the diagrams shown in Fig. 9.1,
is given by:

2e2 f 2  ǫi ǫj ǫi1 Qj2 ǫ2 · Q2 ǫi2 Qj2 ǫ1 · Q2
ij 1 2
T(π) = −2 −2
m2π  D1 D1 d2 D2 d2

Qj Qi ǫ1 · Q 1 ǫ2 · Q 2 Qj Qi ǫ1 · ǫ2 i↔j
  
+ 4 2 1 − 2 1 + , (9.3)
D1 d1 d2 d1 d2 Q ↔ −Q 

where the following abbreviations have been used:

q
D1,2 = (Q ± K)2 + m2π − ω 2 , Q1,2 = Q ± , di = Q2i + m2π , K = (k1 + k2 )/2.
2
The correlator F ij entering Eq. (9.2) has the following general form:
(+) i j
F ij = h0| τa(−) τb ei Q·(xb −xa ) e−i q·(xa +xb )/2 |0i .
X
σa σb (9.4)
a6=b

(±)
Here the summation with respect to a and b is performed over all nucleons, τa = (τa1 ± iτa2 )/2 are
the isospin raising and lowering operators, while σai /2 denotes the i-th component of the spin operator
for the a-th nucleon.

ij
Figure 9.1: Typical diagrams contributing to T(π) . The wavy lines denote photons and dashed lines
ij
denote pions. The amputation indicates that T contains only the nucleon vertices, but not its wave
functions.

For the simplest case of a pure Fermi gas model the correlator (9.4) only has a central part:
F ij = FC δij , where
dp1 dp2
Z Z
FC = −2 dx1 dx2 e−i(x1 +x2 )q/2 ei(x1 −x2 )((p1 −Q−p2 ) . (9.5)
(2π)6
The ranges of integration for the nucleon momenta p1 and p2 are the proton and neutron Fermi
spheres, with the radii
1/3 1/3
Z N
 
(p) (n)
pF = 2.27 mπ , pF = 2.27 mπ , A = N + Z.
A A
This form is easily obtained in a usual Fermi gas model, where the proton (i = p) and neutron (i = n)
densities are written as
(i) 3
 
Z
dp (i) (i)
pF
ρ(i) = 2 n (p) , n(i) (p) = θ(pF − |p|) , i.e. ρ(i) = .
(2π)3 3π 2
Here θ(x) is the step function. Together with the assumption of a homogeneous distribution of protons
and neutrons inside the nuclear volume one has
 3  3
(p) (n)
4 pF R 4 pF R
=Z, = N.
9π 9π

83
 (i)
Finally, by writing R = 1.2 A1/3 fm one finds the above relations for pF . In Eq. (9.5) the integrations
with respect to xi are taken over a sphere with the radius R.
The dependence of the correlation function (9.5) on the nuclear radius R, on N and Z and on the
momentum transfer q were investigated in [168]. In the following we will put N = Z = A/2. Let
us pass in Eq. (9.5) to the variables X = (x1 + x2 )/2 and r = x1 − x2 , which denote the c.m. of
the nucleon pair and the relative distance between the two nucleons, respectively. Taking the integral
with respect to X over the region |X| < R and with respect to r over an infinite range, one obtains as
a first approximation for the correlator
(0)
FC = FC (q) F1 (t), (9.6)
(p) (n)
where q = Q/pF , pF = pF = pF = 1.8 mπ and
2 
A q q
 
(0)
FC (q) =− 1− 1+ θ(2 − q), (9.7)
2 2 4
which is a well-known result (see e.g. [126]) of the Fermi model. In the model under consideration the
form factor F1 (q) is equal to F (Rq), where
3 sin x
 
F (x) = 2 − cos x . (9.8)
x x
Thus, the use of the approximate result (9.6) for the correlator leads to a proportionality of the mesonic
seagull amplitude (9.2) to a form factor F1 (q).
Although this conclusion was obtained for the pure Fermi gas model, where only the central part
of the correlator contributes, the same is valid for the general case, if one integrates in Eq. (9.4) with
respect to xb − xa over an infinite range. It is important to understand the difference between the
mesonic seagull amplitude arising from the approximate correlator and the exact amplitude, which is
obtained by integrating in (9.4) over a finite volume with respect to both, xa and xb . In the amplitude
this difference manifests itself as a difference between the form factors F1 (q) and F2 (q). At low
energies the dependence of the amplitude S on momentum transfer q is determined by the distribution
of nucleon pairs inside the nucleus. In the case of heavy nuclei the scale of nucleon correlations is
essentially smaller than the nuclear radius R. Thus, one can expect that the q-dependence of S is
similar to the nuclear charge form factor F1 (q). However, experimental data clearly indicate [25,32,95]
that this q-dependence cannot fully be identified with the form factor F1 In [71, 72] it was suggested
to use for the amplitude S another form factor F2 instead of F1 . It has been proposed to apply
F2 (q) = F12 (q/2), which corresponds to the distribution of uncorrelated nucleon pairs [19,174]. A first
attempt to quantitatively discuss the function F2 within a model calculation has been made in [75]. As
a first step, we will consider this effect calculating the q-dependence of the term in S proportional to
the enhancement constant κ (cf. Eq. (8.23)). Putting K and q equal to zero in Eq. (9.3), substituting
(9.5) into (9.2) and integrating first with respect to Q and then with respect to the variables pi and
xi we get the following representation of F2 :
J(Rq)
F2 (q) = , (9.9)
J(0)
where
Z2
J(ζ) = dx x3 exp(−mπ R x) F 2 (pF R x) × (9.10)
0
 q 
2
1− x4 !
1 x 3 1 x2
   Z
2


3 1 − F (ζ(1 − x/2)) + dy sin(ζy) 1 − y − 
2 ζx 4 
1− x2
 

84
and F (x) is given in Eq. (9.8). The integral with respect to y in (9.10) can easily be taken analytically,
but we represent the result in the form (9.10) for the sake of brevity. In Fig. 9.2 the two form factors
F1 (q) and F2 (q) as given in Eqs. (9.8) and (9.9), respectively, are compared for A=40. As expected,
the difference is not very big. In Fig. 9.2 we also show the widely used approximation F12 (q/2). As one
sees, this approximation is not valid. Note that in Fig. 9.2 the form factor F (x) is not the experimental
(charge) form factor. For the general discussion of exchange form factors here this difference is not
important. However, experimental form factors F1 will be used as a reference in section 9.3 and
appendix B, where exchange form factors will be discussed in a more quantitative form.

Figure 9.2: Comparison of the form factors from Eqs. (9.8) (dotted curve) and (9.9) (full curve) for
the case of 40 Ca. The approximation F12 (q/2) is shown as a dashed curve.

A statement better suited for practical purposes than the analytical result (9.10) is obtained by
considering the slope of the form factor. We represent the form factors Fi at small momentum transfer
in the form:
q2 D 2E
Fi (q) = 1 + r . (9.11)
6 i

As known, r 2 1 = (3/5) R2 . In order to get the value of r 2 2 , it is necessary to expand the r.h.s. of
Eq. (9.10) up to the order (Rq)2 . The resulting ratio χ(A) = r 2 2 / r 2 1 is shown in Fig. 9.3. We
checked numerically that a good approximation for the function F2 in a wide range of t is given by
F2 (q) ≈ F1 (χt). In a range for A between 4 and 200 the relation
 2
0.32 fm 
χ ≈ 1 − q (9.12)
hr 2 i1

accounts for the dependence of χ on A within a few percent. This corresponds to a finite shift of the
mean charge radius: q q
hr 2 i2 ≈ hr 2 i1 − 0.32 fm. (9.13)
In the following section it will become evident that tensor correlations influence mostly the short-range
behavior of the correlation function. We can therefore expect that the difference between F1 and F2 is
mainly governed by central correlations. The more refined discussion in the next section will confirm
this.

Figure 9.3: Ratio χ(A) = r 2 2/ r2 1 as a function of the nuclear mass number A

9.2 Corrections to the correlation function

In order to obtain the correct behavior of the mesonic seagull amplitude, it is necessary to take into
account not only the central part of the correlator, but also its tensor part. To this end the correction
to the nuclear wave function has been calculated in [168] using perturbation theory with respect to
π- and ρ-meson exchange. These corrections are important, because they determine an essential part
of the nuclear correlation functions. In [168] all diagrams up to the order f 4 have been taken into
account, where f is the meson-nucleon coupling constant. In the course of that calculation it has been
realized that three-body corrections to the correlator are as important as two-body corrections. A word
of caution is necessary concerning the gauge invariance of the amplitude obtained in such a model.
In [168] it was shown that the modifications of electromagnetic polarizabilities are, indeed, gauge
invariant quantities. However, in higher order with respect to ω 2 terms violating gauge invariance

85
appear, which by definition have to cancel with some corresponding terms in the resonance part of
the amplitude. Such cancellations either can be incorporated implicitly in the phenomenology of the
resonance amplitude or may give rise to additional photonuclear sum rules.
As pointed out in the last section all dependence of the correlator on the momentum transfer is
given by the form factor F2 (q). So, putting q = 0 in Eq. (9.4) we represent without change of notation
the correlator F ij in the following form:

3Qi Qj
F ij = FC δij + FT tij , tij = − δij . (9.14)
Q2
The diagrams, which correspond to the two-body part of the correlator correction, are shown in Fig.
9.4. Explicit evaluation of these diagrams leads to the following result:
( )
4M f 2 dp1 dp2 4Qi Qj 2r i r j − r 2 δij
Z
ij
F(1) = − V + ×
m2π (2π)6 Q2 + m2π r 2 + m2π
n(p1 )n(p2 ) [1 − n(p2 + Q)] [1 − n(p1 − Q)]
, (9.15)
Q·r
where r = p1 −p2 −Q and V is the nuclear volume. The function n(p) = θ(pF −|p |) is the occupation
number in the Fermi gas model.

Figure 9.4: Set of diagrams, from which the two-body correction to the correlator is extracted. Here
the same symbolic abbreviation is used as in Fig. 9.1. The nucleon spin projections are denoted by λi .

Figure 9.5: Three-body diagrams yielding an additional correction to the correlator, which is of the
same order in f /mπ as the two-body diagrams shown in Fig. 9.4. The notations are the same as in
Fig. 9.1 and Fig. 9.4.

It is interesting to note that the three-body diagrams displayed in Fig. 9.5 also give a significant
contribution. The corresponding general expression for the three-body contribution to the correlator
reads
( )
16M f 2 dp1 dp2 2Qi Qj 2r i r j − r 2 δij
Z
ij
F(2) = V + ×
m2π (2π)6 Q2 + m2π r 2 + m2π
n(p1 )n(p2 )n(p1 + Q)
. (9.16)
Q·r
The result of the explicit analytical integration for Eqs. (9.16) and (9.15) can be found in [168]. As
before, the correction (9.16) contributes to the central part of the correlator, as well as to its tensor
part. Thus, at this stage of our discussion the correlation function F ij has the following form:
(0) ij ij
F ij = FC δij + F(1) + F(2) . (9.17)

Note that central and tensor parts of the corrections can be extracted by using the definitions in
(9.14). In Fig. 9.6 the full central part of the correlator as a function of Q/pF is shown, together
(0)
with the zeroth order approximation FC and the one, where only two-body correlations have been
taken into account. It can be seen that almost over the whole range of Q the asymptotic form of
the central correlator is modified, although the overall shape remains the same. Figure 9.7 shows the
tensor correlator with and without three-body effects. Here a strong modification due to three-body
correlations occurs at comparatively low Q, where we found a strong damping that is in fact essential

86
for obtaining reasonable physical results. One can see from Fig. 9.7 that the contribution of the high-Q
region is not negligible. It is known that taking into account only the one-pion exchange leads to a
wrong behavior in the tensor part of the nucleon-nucleon potential at small distances (i.e. at large Q)
(see e.g. [173]). Thus, it is necessary to include ρ-meson exchange in the calculation of the tensor part
of the correlator. This can be done [87, 284] by replacing FT by:
" #
q 2 + m2π /p2F
F̃T = FT 1−2 2 , (9.18)
q + m2ρ /p2F
where mρ is the ρ-meson mass. This inclusion of ρ-meson exchange improves essentially the behaviour
of the correlator at Q ∼ pF and higher (cf. Fig. 9.8).

Figure 9.6: Central part of the correlator with three-body corrections (full curve), without three-body
corrections (dotted curve) and Fermi correlator without any corrections (dashed curve).

Figure 9.7: Tensor part of the correlator with three-body corrections (full curve) and without three-body
corrections (dashed curve).

Figure 9.8: Comparison of the tensor part of the correlator with (full curve) and without (dashed
curve) the ρ-meson contribution.

Even below pion threshold the contribution of a virtual ∆-isobar excitation to the mesonic part
of the seagull amplitude is not negligible. General properties of such an effect were discussed in [74].
In [168] this contribution was investigated quantitatively using the same diagrammatic approach as
ij
for the amplitude T(π) . In the static limit the corresponding Hamiltonians, which determine the
interaction, are of the form (see e.g. [173])
efγN ∆ +
HγN ∆ = − S · (∇ × A) T3+ (9.19)
mπ
and
f∆ +
HπN ∆ = − (S · ∇)(T+ · φ), (9.20)
mπ
with the hermitian conjugate to be added in both cases. Here S and T are the 1/2-to-3/2 transition
operators in spin space and isospin space, respectively. Evaluating the corresponding diagrams on this
basis, one obtains the following expression for the isobar contribution to the tensor T ij entering into
Eq. (9.2):
hi1 l2j + l1i hj2 Qi1
" #
ij j
T(∆) = η + [ǫ1 × (k2 × ǫ2 ) − ǫ2 × (k1 × ǫ1 )] , (9.21)
D1 d1
with
2(Qa · ǫa )Qa
ha = ǫa − , la = (Q + K) × (ka × ǫa )
da
and the coefficient
8Ωe2 f∆ fγN ∆ f
η=− .
9(ω 2 − Ω2 )m3π
In (9.21) the same abbreviations have been used as in Eq. (9.3) and, in addition, Ω = M∆ − M is
the mass difference between the ∆-isobar and the nucleon. Also, the symmetry with respect to the
ij
substitution Q → −Q, which is due to the integration in Eq. (9.2), has been used to bring T(∆) into
the form (9.21). Comparison of this mesonic seagull amplitude for nuclear matter with Eq. (8.23)
allows us to give the numerical results in terms of explicit values for κ, δα and δβ. These numerical
values are given in appendix A.

87
9.3 The mesonic seagull amplitude in finite nuclei

In order to discuss the properties of the mesonic seagull amplitude for the case of specific nuclei rather
than for nuclear matter, it is convenient to consider the Fourier transform of the correlation function
discussed in section 9.2. Then the function F ij can be written as
Z
F ij = dx1 dx2 e−iq·(x1 +x2 )/2 e−iQ·(x1 −x2 ) × (9.22)
h i
gC (x1 − x2 ) δij + gT (x1 − x2 ) tij ,

with tij from (9.14). In Eq. (9.22) the functions gC and gT describe the central and tensor correlations
of two nucleons, while the exponential function depending on q is responsible for the distribution of
such nucleon pairs inside the nucleus. The functions gC and gT are related to the momentum space
correlation functions FC and FT via
1 dQ
Z
gC,T (ρ) = FC,T (Q) ei ρ·Q . (9.23)
V (2π)3
Note that in the functions FC and FT the contribution from ρ-meson exchange has been taken into
account. In the case of a pure Fermi gas model we have
!2
(0) p3F (0)
gC (ρ) = −2 F 2 (pF ρ) , gT (ρ) = 0 (9.24)
6π 3
ij
with the function F from Eq. (9.8). Next, we expand T(π) in Eq. (9.2) with respect to k1 and k2 up to
2
O(ω ), pass to the variables ρ = x2 − x1 and ξ = (x1 + x2 )/2. Then, taking the integral with respect
ij
to Q and the angles of ρ and ξ we obtain the contribution of T(π) to the mesonic seagull amplitude:

Ae2 ω2

S(π) = Φ1 (q) ǫ1 · ǫ2 + 2 Φ2 (q) ǫ1 · ǫ2 +
4M mπ
1

Φ3 (q) (ǫ1 × k1 ) · (ǫ2 × k2 ) , (9.25)
m2π
where
Z1
2M f 2 h i
Φi = (2RpF )3 dx GC T
i (ρ1 ) g̃C (ρ2 )+ Gi (ρ1 ) g̃T (ρ2 ) ×
3mπ π 2
0
 √ 
1−x Z1−x2
sin ξRq sin ξRq 1 − −x2 ξ2
Z
x2 e−ρ1  dξ ξ + dξ . (9.26)
 
Rq Rq 2x
0 1−x

In Eq. (9.26) the following abbreviations have been used:

!2
1 6π 2
g̃C,T (ρ2 ) = gC,T (ρ2 /pF ) , ρ1 = 2Rmπ x , ρ2 = 2RpF x .
2 p3F

The functions GC,T

i are of the form
2ρ21 − 12
GC
1 (ρ1 ) = ρ1 , GT1 (ρ1 ) = ,
ρ1
150 + 30ρ1 − ρ31 24 + 12ρ1 − ρ31
GC
2 (ρ1 ) = , GT2 (ρ1 ) =
60 30
3 − 21ρ + 2ρ 2 2ρ2 − 3ρ − 15
1 1 1
GC
3 (ρ1 ) = , GT3 (ρ1 ) = 1
12 6

88
The integral with respect to ξ in (9.26) can easily be taken analytically, but we represent the result in
this form for the sake of brevity. By comparing Eq. (9.25) with the corresponding terms in Eq. (8.23)
one sees that the parameters appearing in the mesonic seagull amplitude are given by the functions
Φi (q) at q=0:
e2 e2
κ = −Φ1 (0) , δα = Φ 2 (0) , δβ = Φ3 (0). (9.27)
4M m2π 4M m2π
It is evident from Eq. (9.26) that three different form factors

(i) Φi (q)
F2 (q) = (9.28)
Φi (0)
appear instead of only F2 .
In order to account for the contribution of ρ-meson exchange to T ij one may follow the prescription
of [87] and make the substitution f 2 → 2f˜ρ2 for the central part of each quantity and f 2 → −f˜ρ2 for
the tensor part. The pion mass m is substituted in all cases by the ρ-meson mass mρ .
In the case of the magnetic polarizability δβ the contribution of the ∆-isobar excitation to the
mesonic seagull amplitude should also be taken into account [74,168]. In our notation this corresponds
to an additional contribution to the function Φ3 , which is of the following form:
Z1
8M f∆ fγN ∆ f (2RpF )3 h i
δΦ3 = dx GC T
∆ (ρ1 ) g̃C (ρ2 )+ G∆ (ρ1 ) g̃T (ρ2 ) ×
81(M∆ − M )π 2
0
 √ 
1−x Z1−x2
sin ξRq sin ξRq 1 − − x2 ξ2
Z
x2 e−ρ1  dξ ξ + dξ , (9.29)
 
Rq Rq 2x
0 1−x

where again M∆ is the ∆-isobar mass and

6 − 12ρ1 12 − 6ρ1
GC
∆ (ρ1 ) = , GT∆ (ρ1 ) = .
ρ1 ρ1
The coupling constants appearing in Eq. (9.29) are taken to be f∆ = 2f and fγN ∆ = 0.35.
Up to now we have considered a constant nucleon density n0 = p3F /3π 2 inside the nucleus, which
is normalized as n0 V = Z. Using a local-density approximation we will extend our consideration
to realistic nuclear densities n(r). With the help of the usual plane-wave expansion via Legendre
polynomials Pl (x) we obtain the realistic-density (rd) form
Z∞ Z∞
(rd) 64πM f 2 2 −2mπ x
Φi = dx x e dr r 2 n2 (r) × (9.30)
3mπ Z
0 x
h i
GCi (2xmπ ) g̃C (2xp(r)) + GTi (2xmπ ) g̃T (2xp(r)) ×
∞
X
jl (r∆) jl (x∆) (Pl−1 (x/r) − Pl+1 (x/r)),
l=0

where p(r) = (3π 2 n(r))1/3 is the local Fermi momentum. In numerical calculations a three-parameter
Fermi parameterization of the densities n(r) has been used with values for the different nuclei taken
from [256].
Equations (9.25)–(9.30) form the starting point of our numerical investigation of the mesonic seagull
amplitude. Note that due to the use of either the Fermi gas model or a local density approximation
the accuracy of our results decreases with decreasing Z.

89
 First we discuss the numerical results for the different contributions to the enhancement constant
κ. Note that, as it was argued in [172] and is also discussed in [173], the main contribution to κGR
comes from central correlations, while κQD is mainly determined by tensor correlations. Obviously, it
is impossible to give a precise separation of κ into a GR part and a QD part, but this approximation
seems reasonable, as the GR contribution should be associated with a small momentum transfer
between the two nucleons in comparison with the QD part. In the pure Fermi gas model, where
(0)
gC = gC and gT =0, the contribution to κ from pion exchange κπ is approximately equal to the ρ-
meson contribution κρ . For nuclear matter (infinite nuclear radius) one has κπ = κρ =0.2, which is in
agreement with a variety of model calculations, e.g. [87, 171, 172]. For finite nuclei, again in the pure
Fermi gas, the value for κπ decreases slightly with decreasing Z, whereas κρ remains the same. With
inclusion of the full correlation functions gC and gT the situation for κ changes drastically. Now the
main contribution to κ comes from tensor correlations related to pion exchange. The pionic central
contribution is still of the same order as before, while κρC becomes negligible. The only significant
contribution from ρ-meson exchange is now due to tensor correlations and has a negative value. All
these relations between the different ingredients to κ remain valid, when realistic nuclear densities are
considered. In Fig. 9.9 the different contributions to κ are shown as a function of Z for the modified
Fermi gas model. In addition, the realistic-density result κ(rd) , which is obtained from Eq. (9.30), is
shown in the same figure.

Figure 9.9: Dependence of enhancement constant κ on proton number Z. The dashed curve corresponds
to the pionic tensor contribution κπT , the dash-dotted curve includes also the central contribution κπC
and the dotted curve gives the total κ, including the contribution from ρ-meson exchange. The realistic-
density result κ(rd) for the full enhancement constant (cf. Eq. (9.30)) is shown as a full curve.

In the case of electric and magnetic polarizabilities δα and δβ the contributions from ρ-meson ex-
change are suppressed by a factor of m2π /m2ρ in comparison with the pion contributions and, therefore,
are negligible. The values of δα and δβ are determined mainly by pionic central correlations, as can be
seen in Figs. 9.10 and 9.11. In the case of δβ the inclusion of the ∆-isobar intermediate state produces
a noticeable effect (cf. Fig. 9.11). In this contribution the values due to central and tensor correlations
are of the same order. Note that κ, δα and δβ get close to their asymptotic (nuclear matter) values
calculated in [168] only at extremely high Z. The size of both, δα and δβ becomes noticeably smaller,
when a realistic density is taken into account. This effect gains importance with decreasing Z. The
ratio of central and tensor contributions to δα is approximately the same for a realistic density as in
a modified Fermi gas model. For δβ the influence of tensor correlations in the realistic-density case is
slightly stronger than for homogeneous nuclear density.
We consider now the dependence of the mesonic seagull amplitude S on momentum transfer q,
(i)
which is determined by the form factors F2 (q) (cf. Eq. (9.28)). Here these form factors are given for
40 Ca. A wider range in A is covered in Appendix B, where explicit results are given for 208 Pb, 16 O
(2) (3)
and 12 C. The results from [193] indicate that F2 for the term proportional to δα and F2 (for δβ)
(1)
are equal with high accuracy, but differ significantly from the form factor F2 for the term containing
(i)
κ. All three functions F2 differ noticeably from F1 . Figure 9.12 shows the corresponding curves
for 40 Ca, for the case of a realistic density. One can see that the frequently used phenomenological
approximation F2 (q) = F12 (q/2), which is also shown in Fig. 9.12, is not in agreement with the q-
dependence of the amplitude S obtained here. For very small q it is convenient to represent the form
(i)
factors as F2 = 1 − q 2 ri2 /6. Then for calcium we find r1 =3.0 fm and r2 =2.5 fm. For other nuclei the
corresponding numbers are given in Appendix B. In the case of 208 Pb, the result from [75] coincides
(2)
within good accuracy with the corresponding form factor F2 from [193].

90
Figure 9.10: Pion-exchange contribution to electric polarizability δα as a function of Z. The dashed
curve corresponds to the central contribution δαC and the dotted curve gives the total δα = δαC + δαT .
The use of a realistic density leads to the full curve.

Figure 9.11: Pion-exchange contribution to magnetic polarizability δβ as a function of Z. The dashed

curve corresponds to the central contribution δβC and the dash-dotted curve gives the sum δβC + δβT .
Adding the contribution of the ∆-isobar excitation as given in Eq. (9.29) leads to the total value of δβ
given as the dotted curve. The use of a realistic density leads to the full curve.

(i) (1) (2)

Figure 9.12: Form factors F2 (∆) for 40 Ca. The dashed curve is F2 and the full curve is F2 . For
comparison the (experimental) charge form factor F1 is also shown (dash-dotted curve), as well as the
function F12 (∆/2) (dotted curve).

91
10 Retardation effects in Compton scattering. A model study with
a relativistic oscillator

10.1 Motivation and aims

In many earlier works [16,19,127,166], a simple ansatz for the Compton scattering amplitude through
the GR region was widely used. It is formulated in terms of the lowest partial-wave amplitudes,
λL
(ω)gλL (θ),
X
RGR (ω, θ) = RGR (10.1)
λL=E1,M 1,E2

whose imaginary parts, ImRGR λL (ω) = (ω/4π)σ λL (ω), exhaust photoabsorption in the GR region. The
GR
real parts of these partial amplitudes are assumed to be given by the dispersion integrals

ω2 ∞ λL (ω ′ ) dω ′
σGR
Z
λL E1
ReRGR (ω) = RGR (0)δλL, E1 + 2P , (10.2)
2π 0 ω ′2 − ω 2
where ∞
1
Z
E1
RGR (0)= 2 σGR (ω) dω. (10.3)
2π 0
This ansatz was constructed in a way that ensured the fulfillment of the Gell-Mann–Goldberger–
Thirring dispersion relation in forward direction, when all gλL = 1:

1 ∞ ω ′2
Z
RGR (ω, 0) = σGR (ω ′ ) dω ′ . (10.4)
2π 2 0 ω ′2 − ω 2 − i0
In particular, when ω → ∞, the amplitude RGR (ω, 0) vanishes, and the total Compton scattering
amplitude TGR is determined by the seagull contribution S GR .
This nice feature is destroyed by the E2 and M 1 contributions when the scattering angle θ is not
zero. For example, when θ = π, the asymptotically nonvanishing part of the resonance amplitude
reads
1
Z  
E2 M1
RGR (∞, π) = 2 σGR (ω) + σGR (ω) dω 6= 0. (10.5)
π
However, the vanishing of RGR for ω = ∞ and for all angles is physically necessary. Such a vanishing
was used in section 8.2 to analyze the seagull contribution emerging from the fixed-t dispersion relations
and to derive a set of related sum rules.
Phenomenologically, the dispersion relations (10.2) for the partial amplitudes could be corrected
through introducing special ω-dependent form factors F (ω) [95,176] which erase the unphysical asymp-
totic pieces. Though well motivated by retardation effects calculated in the frame of nuclear mod-
els [176], such a procedure may however need a further theoretical justification because the form factors
F (ω) generally destroy the asymptotic behavior of the amplitude at high complex energies and violate
the validity of dispersion relations. Only taken consistently by taking into account the contributions
of higher multipoles and relativistic corrections, they lead to a correct result. Accordingly, the present
section is aimed to illustrate an interplay between the retardation effects, higher multipoles, relativis-
tic corrections, asymptotic behavior and the dispersion relations by using a model which contains the
Compton scattering amplitude with all correct analytical properties. Such a model has to be relativis-
tic, and therefore we consider a relativistic harmonic oscillator, which allows an analytical treatment
of this problem. In addition to its theoretical transparency, this model has the advantage of having a
very close relation to the physical problem we are considering here, i.e. to giant resonances. The 1h̄ω0
transition from the ground state to the n = 1 oscillator state corresponds to the electric giant-dipole
resonance, the 2h̄ω0 transition to the d-component of the n = 2 oscillator states corresponds to the

92
electric giant-quadrupole resonance, etc. This interpretation shows that each electric giant-multipole
resonance is accompanied by higher harmonics which are closely related to the retardation problem.
Magnetic giant resonances are not taken into consideration.
In the following subsections we explicitly formulate the model. We find an analytical representation
of the Compton scattering amplitude8 T = S + R, study the high-energy behavior of T and show that
the amplitude satisfies the fixed-t dispersion relation. We discuss why relativistic effects are important
for the validity of the dispersion relation. Finally, we study the retardation effects on the amplitude at
low and high energies and on the E1 and E2 sum rules and discuss how the retardation form factors
can appear in the Compton scattering amplitude and in the dispersion relations.

10.2 Basics of the model. The seagull and resonance amplitudes

We consider a scalar particle with the electric charge e described by the Klein-Gordan equation with
an oscillator Lorentz-scalar potential,
h ∂ 2 i
i − eA0 − (i∇ + eA)2 − (µ2 + γ 4 r 2 ) ψ(r, t) = 0. (10.6)
∂t
± of the above Klein-
When the electromagnetic potential Aµ is absent, the normalized solutions ψnα
Gordan equation with positive and negative energies read

± 1
ψnα (r, t) = √ exp(∓iEn t)φnα (r), (10.7)
2En
where n = 2nr + j is the quantum number which determines the energies ±En of degenerated levels
with different radial quantum numbers nr and angular momenta j, and the generic parameter α = (jm)
describing the angular momentum j and its projection m. The eigen functions φnα satisfy the equation
Z
− ∇2 φnα (r) + γ 4 r 2 φnα (r) = (En2 − µ2 )φnα (r), |φnα (r)|2 dr = 1, (10.8)

and the eigen energies En are

q q
En = µ2 + (2n + 3)γ 2 = E02 + 2nγ 2 , n = 0, 1, 2, . . . . (10.9)

The wave functions and the energy spectrum are characterized by two parameters, γ and E0 . The
parameter γ determines the scale for the wave function of the ground state,
1
 
φ0 (r) = N0 exp − γ 2 r 2 , |N0 |2 = γ 3 π −3/2 , (10.10)
2
q
2 i1/2 = 3 −1
the corresponding r.m.s. electric radius, hrE 2γ , and the form factor of the charge distribu-
tion in the ground state,
!
q2
Z
2
F (q) = exp(iq · r) |φ0 (r)| dr = exp − 2 . (10.11)
4γ

The parameter 2E0 determines the energy gap between positive and negative parts of the spectrum.
In the nonrelativistic limit, E0 becomes the mass of the particle, and the quantity

γ2
ω0 ≡ (10.12)
E0
8
We drop here and in the following the subscript GR

93
becomes the oscillator frequency. The scale of relativistic effects is determined by the dimensionless
ratio
γ2 ω0 ω2
η= 2 = = 20 , 0 < η < ∞. (10.13)
E0 E0 γ
In the nonrelativistic regime, η ≪ 1. When η ∼ > 1, the energies which are necessary to excite the
oscillator or create a particle-antiparticle pair become of the same order.

Figure 10.1: Diagrams of Compton scattering off the relativistic oscillator (crossed terms are not
shown)

The total amplitude of photon scattering on the particle confined in the potential wall consists of
two parts. The first one, TD , is the so-called Delbrück scattering amplitude which describes scattering
by the empty potential wall via virtual particle-antiparticle pairs (see the recent review [5] and refer-
ences therein). It is usually small except for a region of near-forward scattering angles. For the case
of the oscillator potential, the Delbrück amplitude was calculated in [285]. The second part T , which
solely will be discussed in the following, is the amplitude for Compton scattering by the particle itself.
It is given by diagrams of the noncovariant perturbation theory which are shown in Fig. 10.1. There,
the diagram a gives the seagull contribution S, and the diagrams b and c correspond to the resonance
part R of the amplitude T = S + R. The seagull contribution reads

e2 e2
S=− (ǫ1 · ǫ2 )hφ0 | exp(iq · r)|φ0 i = − (ǫ1 · ǫ2 )F (q) (10.14)
E0 E0
with q = k1 − k2 , whereas the resonance amplitude R has the form
X e2 hφ0 |(ǫ2 · p) exp(−ik2 · r)|φnα ihφnα |(ǫ1 · p) exp(ik1 · r)|φ0 i
R = +
nα E0 En En − ω − E0 − i0
X e2 hφ0 |(ǫ2 · p) exp(−ik2 · r)|φnα ihφnα |(ǫ1 · p) exp(ik1 · r)|φ0 i
+
nα E0 En En + ω + E0 − i0
(ω → −ω, k1 ↔ −k2 , ǫ1 ↔ ǫ2 ). (10.15)

The factors 1/E0 in (10.14) and 1/E0 En in (10.15) are related with the proper normalization of the
wave functions, cf. Eq. (10.7). The first sum in (10.15) corresponds to the contribution of positive-
energy intermediate states (the diagram b), whereas the second one accounts for negative-energy states,
or virtual pair production (the diagram c).9 Taken together, they give the resonance amplitude in the
Feynman approach with the Feynman propagator (En2 − (ω + E0 )2 − i0)−1 . The crossed term has a
similar structure and interpretation. The k-dependent exponents in (10.15) describe the retardation
effects. Keeping in the operators (ǫ · p) exp(±ik · r) the leading-in-k terms of a given total angular
momentum j (cf. section 3.4), we can obtain unretarded contributions to different multipole amplitudes
Rλj . The unretarded amplitudes can also be extracted from leading terms in an expansion of the total
amplitude (10.15) in powers of photon momenta, k1 and k2 , considered at arbitrary absolute values
of k1 6= ω and k2 6= ω. The momenta are still assumed to be orthogonal to the photon polarizations,
9
In terms of physical states having positive energies and consisting of particles and antiparticles (i.e. p and p̄), the
two sums in (10.15) have a different meaning. The first sum is related with transitions of the particle p from the ground
state to an excited state (p0 ↔ pnα ). The second sum is related with pair production from the vacuum. Most of the pair
production is included through the Delbrück scattering amplitude TD which takes into account all possible transitions
vac ↔ pi p̄j . However, due to Bose correlations, the presence of the particle p0 in the initial state doubles the probability
of pair production to the occupied state i = 0. Respectively, the second sum in Eq. (10.15) takes into account these
additional transitions vac ↔ p0 p̄j caused by the initial particle p0 . Note that in the case of a spin-1/2 particle the second
sum would change its sign. It would describe Fermi correlations with p0 , i.e. Pauli blocking of pair production to the
occupied state p0 .

94
ǫ1 · k1 = ǫ2 · k2 = 0. Since it is equally easy, in our simple model, to calculate the amplitude R for such
a general case of arbitrary k1 and k2 , we follow this way to make the retardation effects manifestly
evident. In other words, we calculate the amplitude of virtual Compton scattering with transversely
polarized photons.
A powerful tool to calculate the resonance amplitude (10.15) is provided by the Greens function
which is explicitly known for the oscillator potential. Introducing
X hr2 |φnα ihφnα |r1 i
G(r2 , r1 |E) = , (10.16)
nα E 2 − En2 + i0

we rewrite the resonance amplitude like

2e2 ∗
Z 
R=− exp(ik1 · r1 − ik2 · r2 ) ǫ2 · ∇φ0 (r2 ) G(r2 , r1 |ω + E0 ) ×
E0
 
ǫ1 · ∇φ0 (r1 ) dr1 dr2 + crossed term. (10.17)

Substituting the Greens function for the relativistic oscillator [285, 286],

iγ
Z ∞ ds
G(r2 , r1 |E) = − ×
2 0 (2πi sin s)3/2
( )
is 2 2 iγ 2 h 2 2
i
exp (E − µ + i0) + (r1 + r2 ) cos s − 2r1 · r2 (10.18)
2γ 2 2 sin s

and taking simple Gaussian integrals with respect to r1 and r2 , we finally get

e2 F (q) (ǫ1 · k2 )(ǫ2 · k1 )

 
R= g(a, x)(ǫ1 · ǫ2 ) + g(a − 1, x) + (ω → −ω). (10.19)
2E0 2γ 2
Here the function g, a close relative of the standard incomplete gamma-function, reads

i
Z 2π  
g(a, x) = exp −ias + x(eis − 1) ds, (10.20)
e−2πia − 1 0

and the quantities x and a are defined as

k1 · k2 ω 2 + 2ωE0
x= , a= − 1 + i0. (10.21)
2γ 2 2γ 2

In this form, Eqs. (10.14) and (10.19) for the seagull and the resonance amplitude, respectively, are
valid for any k1 and k2 . We have to set k1 = k2 = ω whenever we consider Compton scattering with
real photons, including the retardation effects.
For further references note that
∞ ∞
xn xn
g(a, x) = e−x
X X
=− (10.22)
n=0
n!(n − a) n=0
a(a − 1) . . . (a − n)

and
xg(a − 1, x) = ag(a, x) + 1. (10.23)
It is seen that g(a, x) is a regular function of x and a except for simple poles at a = 0, 1, 2, . . . .
Asymptotic expansions of g at low energies (when x ≪ a ≈ −1) or at “nonrelativistically high energies”
(i.e. when ω ∼ γ ≫ ω0 and ω ≪ E0 , or, in other words, when a ≫ x ∼ 1; such a regime is possible,
provided η ≪ 1) directly follow from Eq. (10.22). The important case of a ≈ x ∼ (a − x)2 → ∞,

95
which arises for real Compton scattering at ultra-relativistic energies, is described by the saddle-point
asymptotics

π  (a − x)2 
r
exp −
g(a, x) = − ctg (πa) +
2x 2x
Z ∞ a − x 
i 2 1
√ sin √ s e−s /2 ds + O( ). (10.24)
x 0 x x
√ √
Here the sign of x must be chosen in such a way that Re x > 0.
Given (10.22), the amplitude R is recast to the form10
∞
e2 X ω0 F (k1 ) F (k2 )
R= ×
2E0 n=1 nω0 − ω − ω 2 /2E0 − i0
( )
xn−1 xn−2 (ǫ1 · k2 )(ǫ2 · k1 )
(ǫ1 · ǫ2 ) + + (ω → −ω), (10.25)
(n − 1)! (n − 2)! 2γ 2

which is useful to single out the individual contributions of different intermediate states. In particular,
from (10.25) one can directly read off individual contributions of oscillator excitations and pair pro-
duction to the amplitude (10.15). These contributions correspond to the pole parts of the ω-dependent
propagator in (10.25),

ω0 /E0 γ2 1 1
 
= + . (10.26)
nω0 − ω − ω 2 /2E0 E0 En En − E0 − ω En + E0 + ω

The sum over n in (10.25) starts from n = 1 because the matrix elements in (10.15) vanish at n = 0.
Also, in (10.25) it should be taken into account that (−1)! = ∞. For a weakly relativistic oscillator
(i.e. η ≪ 1) and low energies ω ∼ ω0 , the resonance amplitude R is dominated by the dipole term
(n = 1), whereas the higher-level corrections are generally suppressed as η n−1 .
Equation (10.25) is convenient for obtaining the multipole structure of the resonance amplitudes.
All the angular dependence of R is contained in the x-dependent braces in (10.25), where x ∝ cos θ ≡ z.
Applying the same technique of helicity amplitudes as was used in section 8.1, we find the partial Ej-
components of the resonance amplitude as a sum over radial excitations,11
n−1
e2 ω0 F (k1 ) F (k2 ) k1 k2 Cjn

REj =
X
2
+ (ω → −ω). (10.27)
2E0 n=j,j+2,j+4,...
nω 0 − ω − ω /2E0 − i0 2γ 2 (n − 1)!

Here the coefficients Cjn determine the multipole expansion of the braces in (10.25),

z n−1 (ǫ1 · ǫ2 ) + (n − 1)z n−2 (ǫ1 · k̂2 )(ǫ2 · k̂1 ) =

X
Cjn gEj . (10.28)
j
P
They are normalized as j Cjn = 1 and are equal to

2j + 1
Z 1  
j
Cjn = nz n−1 − (n − 1)z n−2 (1 + z) d1,1 (θ) dz
4 −1
[(j−1)/2]
1  j(j + 1)(2j + 1) n − j + 2i
1 + (−1)n−j
Y
= , (10.29)
4 (n + j + 1)(n + j − 1) i=1
n + j − 2i − 1
10
This form can be easily derived from (10.15) by using, instead of the Greens function method, the well-known
technique of the creation and annihilation operators for the harmonic oscillator.
11
The partial amplitudes here are normalized as R(ω, θ) = j REj (ω) g Ej (θ).
P

96
provided n ≥ j (otherwise Cjn = 0). In particular,
3 15
C1n = , C2n = , etc. (for n − j = even). (10.30)
n(n + 2) (n + 1)(n + 3)
The (magnetic) M j-components of R are identically zero in our model due to parity conservation.
Keeping in (10.27) the leading-in-k term, i.e. omitting the form factors F (k1 ) and F (k2 ) and
retaining only the lowest n = j contribution, we obtain the unretarded multipoles,
j−1
e2 ω0 (jω0 − ω 2 /2E0 ) k1 k2 Cjj

◦ Ej
R = . (10.31)
E0 (jω0 − ω 2 /2E0 )2 − ω 2 − i0 2γ 2 (j − 1)!
For a weakly relativistic oscillator (η ≪ 1), the main difference between the unretarded and retarded
multipoles REj at low energies ω ∼ < jω0 is described by the k-dependent form factor squared of the
ground state. Also, the form factor damps the amplitude at the resonance peak, e.g. to the amount
of F 2 (ω0 ) = e−η/2 in the case of the E1-multipole at ω ≃ ω0 .
In order to understand what happens at somewhat higher energies ω (but still a few ω0 ), note
that all the coefficients Cjn are positive. They determine relative contributions of different radial
excitations to REj and are proportional to the matrix element squared of the transition operator
(ǫ1 · p) exp(ik1 · r) between the ground state |φ0 i and a state |φn(jm) i of the energy En and the
angular momentum j. So, this feature does not depend on a specific shape of the potential. Then, in
the region between the energies ω0 and 3ω0 , the positive contributions O(η 2 ) of the radial excitations
to RE1 tend to cancel the leading negative contribution from the first level n = 1 and thus slightly
intensify the damping effect O(η) caused by the k-dependent form factors alone. Similar small local
strengthening of the damping effects just above the first peak holds for higher multipoles as well.
At even higher energies, ω ∼ γ ≫ ω0 , which we would call intermediate energies, when the
k-dependent form factors produce essential damping of individual transitions, the contributions of
radial excitations are not negligible at all. They are enhanced by powers of the same parameter
k1 k2 /2γ 2 which determines the magnitude of the form factors. As a result, many multipoles REj
become as large as RE1 , which dominates at low energies. They act coherently at small angles and
tend to compensate the effect of the form factors. Below we explicitly demonstrate this important
feature.

10.3 Asymptotic behavior and fixed-t dispersion relations

The resonance amplitude (10.25) at k1 = k2 = ω is given by a convergent series in x = (2ω 2 − q 2 )/4γ 2

and has no singularities at finite energies ω and momentum transfer q 2 = −t, except for the poles at
ω = ωn± and ω = −ωn± , where the quantities
ωn± = En ∓ E0 (10.32)
describe the energies of the oscillator excitation and pair production, respectively. Moreover, the seag-
ull amplitude (10.14), considered as a function of the energy, is a (t-dependent) constant. Therefore,
the validity of the fixed-t dispersion relation for the Compton scattering amplitude T and for the
appropriate invariant amplitudes depends on whether or not they vanish at high complex ω.
Using (10.14) and (10.19), we find coefficients of the structures (ǫ1 · ǫ2 ) and (s1 · s2 ) in the total
amplitude T and then obtain the invariant amplitudes A1,2 (ω, t). They read
e2 n o
ω 2 (A1 + A2 ) = F (q) (1 + a)g(a, x) + (ω → −ω) , (10.33)
2E0
e2 n o
(A1 − A2 ) = − 2 F (q) g(a − 1, x) + (ω → −ω) . (10.34)
4γ E0

97
Note that, at low energies ω → 0 and fixed t, the parameter a → −1 and x → t/4γ 2 . In this limit,
the r.h.s. of Eq. (10.33) vanishes. Respectively, both the amplitudes A1,2 (ω, t) are finite at ω = 0 and
thus are free from the 1/ω 2 pole. This perfectly agrees with the low-energy theorem, since the system
of “the particle plus the potential wall” has an infinite mass and, therefore, its Thomson amplitude
has to vanish.
At high energies, ω → ∞, and fixed t, both a and x go to infinity like ω 2 , and (a − x)2 = O(ω 2 )
as well. Using Eq. (10.24), we find that both amplitudes A1,2 (ω, t) vanish like O(1/ω) in the complex
plane of ω and, therefore, satisfy unsubtracted fixed-t dispersion relations (8.17). The ultra-relativistic
asymptotics of the resonance amplitude reads
e2 π (ǫ1 · k2 )(ǫ2 · k1 )
r   
R≃− F (q) e−1/η ctg (πa) + (ω → −ω) (ǫ1 · ǫ2 ) + , (10.35)
2E0 2x 2γ 2
√
where Re x > 0, like in (10.24). Beyond the real axis of ω, the amplitude R is predominately
imaginary,
e2 √ (ǫ1 · k2 )(ǫ2 · k1 )
 
R ≃ i sign(Im ω) F (q) πη e−1/η (ǫ1 · ǫ2 ) + . (10.36)
ω 2γ 2
When the relativistic parameter η is small, the coefficient of the leading 1/ω term in this asymptotics is
exponentially suppressed. Then the asymptotic regime (10.36) holds only at extremely high energies,
whereas at intermediate energies of ω ∼ γ ≫ ω0 the resonance amplitude is predominately real and
follows the 1/ω 2 dependence,
( )
e2 ω02  q2   q 2  (ǫ1 · k2 )(ǫ2 · k1 )
R≃− F (q) 1 − (ǫ 1 · ǫ 2 ) + 2 − . (10.37)
E0 ω 2 4γ 2 4γ 2 2γ 2

The remarkable feature of these asymptotics is that the retardation k-dependent form factors F (k1 ) and
F (k2 ), which accompany all individual contributions of the transitions |0i → |nαi and exponentially
grow at high complex energies, are canceled when many transitions come into play. In the asymptotic
regime, the resulting amplitude depends on the form factor only through the momentum transfer q
rather than the energy ω and, therefore, is free from the exponential growth at ω → i∞. This is why
the fixed-t dispersion relation holds.
As an important particular case, note that the asymptotics (10.37) of the exact amplitude R at
θ = 0 and at intermediate energies behaves like the asymptotics of the naive unretarded electric dipole
amplitude
◦ E1,naive e2 ω02
R = . (10.38)
E0 ω02 − ω 2 − i0
This feature is connected with a validity of the dispersion relation and its consequence, the E1-sum
rule, is discussed in the next subsection.
It is instructive to compare the high-energy behavior of the Compton scattering amplitude off
the relativistic and nonrelativistic (nonrel) oscillator and to see where a difference comes from. The
nonrelativistic oscillator has the energy spectrum

Ennonrel = E0nonrel + nω0 (10.39)

(only with positive energies) and is described by the same eigen functions φnα as in (10.8), with the
substitution En2 − µ2 → (2n + 3)γ 2 . The nonrelativistic oscillator frequency ω0 = γ 2 /M is determined
by the size of the charge distribution in the ground state, 1/γ, and the mass of the particle, M . The
nonrelativistic resonance amplitude Rnonrel is given by expressions similar to (10.15) and (10.25),
X e2 hφ0 |(ǫ2 · p) exp(−ik2 · r)|φnα ihφnα |(ǫ1 · p) exp(ik1 · r)|φ0 i
Rnonrel =
nα M2 En − ω − E0 − i0

98
 + (ω → −ω, k1 ↔ −k2 , ǫ1 ↔ ǫ2 ) (10.40)

∞
( )
X e2 γ 2 F (k1 ) F (k2 ) xn−1 xn−2 (ǫ1 · k2 )(ǫ2 · k1 )
= (ǫ1 · ǫ2 ) +
n=1
2M 2 (nω0 − ω − i0) (n − 1)! (n − 2)! 2γ 2
+ (ω → −ω) (10.41)

e2 F (q) (ǫ1 · k2 )(ǫ2 · k1 )

 
= g(a0 , x)(ǫ1 · ǫ2 ) + g(a0 − 1, x)
2M 2γ 2
+ (ω → −ω), a0 = ω/ω0 − 1 + i0. (10.42)

The absence of ω 2 -pieces in the denominator in (10.41) and in a0 drastically changes the behavior
of the amplitude Rnonrel at very high complex energies. When ω → i∞, the function g(a0 , x) grows
exponentially like e−x with x given by Eq. (10.21).. Respectively, the resonance amplitude Rnonrel
grows like the retardation form factors F 2 (ω) and does not obey the fixed-t dispersion relation.12
A difference between the asymptotics of the relativistic and nonrelativistic amplitudes appears not
only at extremely high energies. The retardation corrections become essential already at intermediate
energies, ω ∼ γ, and the amplitude Rnonrel behaves like

e2 ω02 (ǫ1 · k2 )(ǫ2 · k1 )

 
Rnonrel ≃ − 2
F (q) (1 + x)(ǫ1 · ǫ2 ) + (2 + x) (10.43)
M ω 2γ 2

at such ω, with x = (−q 2 + 2ω 2 )/4γ 2 . Therefore, it shows no tendency to decrease, in contrast to

(10.37).
It is again instructive to obtain the nonrelativistic asymptotics (10.43) directly from Eq. (10.40)
and to trace a difference with the relativistic consideration. Since the sum in (10.40) is saturated
by intermediate states with the typical momentum ∼ k and, hence, with the excitation energies
En − E0 ∼ k2 /(2M ) ≪ ω, the denominator in (10.40) can be expanded in the inverse powers of ω,

1 1 En − E0
=− − − ···. (10.44)
En − E0 − ω ω ω2
Here the leading term is odd and is canceled by the crossed term. The asymptotics is determined
by the next term. Using closure, one can recast the result in terms of the double commutator of the
electromagnetic currents and the nonrelativistic Hamiltonian H = p2 /(2M ) + M ω02 r 2 /2,
1
Rnonrel ≃ − hφ0 | [j2 , [H, j1 ] |φ0 i. (10.45)
ω2
Here j1 = (e/M )(ǫ1 · p) exp(ik1 · r) and j2 = (e/M )(ǫ2 · p) exp(−ik2 · r). Calculating the above matrix
element, we arrive at Eq. (10.43). The derivatives of the k-dependent exponents in the operators j1
and j2 , coming from the commutator with the kinetic energy, lead to the ω 2 -terms in the braces of
(10.43). With similar steps for the relativistic oscillator, we would get a difference due to the negative-
energy intermediate states which results in an ω 2 -correction in the energy denominator. So, instead
of Eq. (10.44), we get the terms

1 1 En − E0 − ω 2 /(2M )
= − − − ··· (10.46)
En − E0 − ω − ω 2 /(2M ) ω ω2
12
The violation of the dispersion relation for the nonrelativistic Compton scattering amplitude was discussed in Ref.
[186–188] where it was connected with additional poles or cuts at ω ≃ 2M . In the nonrelativistic oscillator model such
singularities are absent, but the exponential growth at ω → i∞ does make standard dispersion relations inapplicable. It
is still possible to write a ”practical” dispersion relation with a Cauchy loop of a finite size ωmax with γ ≪ ωmax ≪ M
and use the asymptotics discussed in the text for allowing for high-ω contributions.

99
in this case. Here, the relativistic correction ω 2 /(2M ) is of the same order of magnitude as En − E0
itself . It gives an additional ω-independent contribution to the asymptotic resonance amplitude,
1
R− = hφ0 |j1 j2 + j2 j1 |φ0 i. (10.47)
2M
This R− , which is the contribution of negative energies, compensates the ω-independent term, which
appears in (10.43) through the terms containing x, and results in a relativistic 1/ω 2 -asymptotics,
Eq. (10.37), consistent with the fixed-t dispersion relation.

10.4 Retarded and unretarded sum rules

Since the amplitude R vanishes at ω → ∞ and fixed t, it obeys the unsubtracted dispersion relation.
That means that both the coefficients of (ǫ1 · ǫ2 ) and (ǫ1 · k2 )(ǫ2 · k1 ) obey the relation:
2 ∞ ω ′ dω ′
Z
R(ω, t) = Im R(ω ′ , t)
π
0 ω ′2 − ω 2 − i0
∞ 2
XX e γ ωn± F 2 (ωn± )
2
= ± 2 2
×
E E
± n=1 0 n (ωn ) − ω − i0
( )
(x±
n)
n−1 (x± )n−2 (ǫ1 · k2 )(ǫ2 · k1 )
(ǫ1 · ǫ2 ) + n . (10.48)
(n − 1)! (n − 2)! 2γ 2

Here the sign of ± refers to the contributions of positive and negative energies, and x± ± 2
n = (2(ωn ) +
2
t)/4γ . Matching the r.h.s. of Eq. (10.48) with the relation

e2
R(0, t) = −S(t) = (ǫ1 · ǫ2 )F (t), (10.49)
E0
which strictly follows from the low-energy theorem for the total amplitude, which reads T (ω = 0, t) = 0
in our case, we arrive at the series of sum rules as discussed in section 8.2. In the following we consider
two of them which are closely related with the E1 and E2 photoabsorption, respectively.
(i) The first one, the Gell-Mann–Goldberger–Thirring (GGT) sum rule, corresponds to the partic-
ular case of t = 0 in (10.48). It reads
∞ 2π 2 e2
Z
Σ≡ σ(ω) dω = 2π 2 R(0, 0) = , (10.50)
0 E0
where
4π
Im R(ω, 0)
σ(ω) = (10.51)
ω
is the total photoabsorption cross section. In accordance with Eqs. (10.25) and (10.48), this cross
section consists of the contribution of particle excitation, σ+ , and that of pair production, σ− ,
∞
2π 2 e2 γ 2 An−1 ω2
e−A
X
σ = σ+ + σ− , σ± (ω) = δ(ω − En ± E0 ), A= . (10.52)
n=1
ωE0 En (n − 1)! 2γ 2

Respectively, the integral in (10.50) consists of two parts as well,

∞
2π 2 e2 γ 2 (A±
n)
n−1 (ωn± )2
exp(−A± A±
X
Σ = Σ+ + Σ− , Σ± = ± n ), n = . (10.53)
n=1 ωn E0 En
(n − 1)! 2γ 2

Both of them are needed to strictly fulfil the GGT sum rule. However, in the weakly relativistic case,
the pair contribution is exponentially suppressed by the retardation form factors, Σ− ∝ exp(−1/η),

100
and the sum rule is saturated with the positive-energy contribution Σ+ . Even when η = 1, Σ−
contributes still only 5% to the total magnitude of Σ. Only in the extreme case of η ≫ 1 both parts
of Σ become compatible. Σ− = Σ+ = 21 Σ in this limit.
When the retardation in Eq. (10.50) is switched off, i.e. when k1 and k2 are set to zero, we find
the unretarded E1-amplitude,
◦ E1 e2 γ 2 1
R (ω) = 2 2
+ (ω → −ω). (10.54)
E0 2γ − ω − 2ωE0
It also satisfies the unsubtracted dispersion relation, and, being matched with the low-energy theorem
(10.49), gives the unretarded partner of the GGT sum rule, which is the Thomas–Reiche–Kuhn (TRK)
sum rule:
2π 2 e2
Z ∞
◦ ◦ ◦
Σ ≡ σ E1 (ω) dω = 2π 2 RE1 (0) = . (10.55)
ωthr E0
◦ ◦
Here the unretarded photoabsorption cross section, σ E1 = (4π/ω)ImRE1 , also consists of the contri-
butions of excitation and pair production,

◦ ◦ ◦ ◦ 2π 2 e2 γ 2
σ E1 = σ E1,+ + σ E1,− , σ E1,± = δ(ω − E1 ± E0 ). (10.56)
ωE0 E1
◦
Corresponding parts of Σ read
◦ 2π 2 e2 γ 2
Σ± = , (10.57)
ω1± E0 E1
◦ ◦ ◦
and their sum, Σ = Σ+ + Σ− , coincides with the retarded integral in Eq. (10.50),
◦
Σ = Σ. (10.58)

The last equation was discussed in literature many times after S. Gerasimov [95, 185–190]. It strictly
holds in our model. Moreover, as is seen from the above derivation, it is valid in a more general
case, with a general form of the current operators or the energy spectrum, provided the unsubtracted
dispersion relation is valid for the retarded amplitude R(ω, 0). Other ingredients of the derivation,
◦ ◦
which are the unsubtracted dispersion relation for RE1 (ω) and the identity R(0, 0) = RE1 (0), are
model independent.13
However, Gerasimov’s relation (10.58) is valid only when both positive- and negative-energy parts
are included. It is not valid for each of these parts separately. For a weakly relativistic oscillator, the
contribution of pair production is exponentially small, Σ− ∝ exp(−1/η), whereas it is of order O(η)
for the unretarded counterpart. Respectively,
◦ ◦ ◦ 1
Σ+ − Σ+ = −(Σ− − Σ− ) ≃ −Σ− ≃ − ηΣ (10.59)
2
in this case. In other words, the positive-energy TRK sum rule differs from the GGT sum rule.
Saturation of the full TRK sum rule needs the consideration of relativistically high energies, whereas
such energies are irrelevant for the GGT sum rule. When the GGT and TRK integrals are truncated
below the pair production threshold (that is what is usually done in most nuclear applications), they
become different in order O(η). Such a “practical” violation of the Gerasimov’s relation was studied
and exemplified in [186–188, 190]. As was said, another reason for the violation of Eq. (10.58) can be
13
These requirements are not fulfilled for a Dirac particle, for which the seagull contribution is absent and R(∞, 0) =
T (∞, 0) 6= 0. In that case a sensical comparison of the GGT and TRK sum rule needs a consideration of positive-energy
components [121, 287].

101
caused by a nonvanishing amplitude R at high real or complex energies, which makes the unsubtracted
dispersion relation inapplicable for R (see an example in [189]).
Note that the retardation form factor in Σ+ reduces the dominating E1 contribution by F 2 (ω1 ) ≃
1 − 12 η. Since the E2 contribution also increases Σ+ by η, the resulting GGT magnitude is bigger than
◦
the nonrelativistic TRK by 21 η. Only after adding the negative-energy part of TRK, Σ− ≃ 21 ηΣ, the
balance between GGT and TRK is achieved.14 Equation (10.59) perfectly agrees with the evaluation
of the difference between positive-energy GGT and TRK sum rules for a bound particle done by Friar
and Fallieros [187] who showed that this difference is equal to −2π 2 e2 hp2 /3M 3 i (note that for the
oscillator model hp2 i = 23 γ 2 ). The origin of the difference (10.59) can be traced to the asymptotic
behavior of the unretarded amplitude (10.54) at intermediate energies ω0 ≪ ω ≪ M . This asymptotics
◦
is determined by the energy-independent contribution RE1
− of negative-energy states (see Eq. (10.47),
in which one has to use j1,2 with k1,2 = 0) and reads
◦ E1 ◦ e2 γ 2 e2 hp2 i
R (ω) ≃ RE1
− ≃ = . (10.60)
2M 3 3M 3
Accordingly, the dispersion integral (10.55) truncated at an intermediate energy ωm becomes
◦ ωm ◦ ◦ ◦ γ2 
Z  
Σ (ωm ) ≡ σ E1 (ω) dω = 2π 2 RE1 (0) − RE1 (ωm ) ≃ 1 − Σ, (10.61)
ωthr 2M 2
from which Eq. (10.59) follows.
Usually in low-energy applications, all contributions of negative-energy states which do not depend
on the energy ω and thus survive at high ω are included into an effective seagull amplitude. Doing
so, we conclude from Eq. (10.60) that the effective (i.e. vanishing at high ω) unretarded resonance
◦ ◦
amplitude is exactly the positive-energy contribution RE1 E1
+ (ω). Due to R− 6= 0, it is different from
the retarded (i.e. dispersion) amplitude R(ω, 0) at zero energy, and, correspondingly, the effective
◦ ◦
unretarded seagull S (ω, θ) = S(ω, θ) + RE1 E1
− g (θ) is different from the retarded seagull S(ω, θ) as well.

Considering unretarded and retarded amplitudes which vanish at intermediate energies, we find a
◦
difference in that these vanishing amplitudes are exactly the positive-energy part RE1 + and the total
amplitude R = R+ + R− , respectively. The latter includes both positive- and negative-energy states,
because the positive-energy part R+ alone does not vanish, as was discussed in section 10.3. Therefore,
using dispersion relations with an intermediate-energy cut ωm , we obtain through the retarded cross
section σ(ω) the amplitude R which includes both positive- and negative-energy states. At the same
◦
time, the dispersion integral of the unretarded cross section σ E1 (ω) gives only the positive-energy
◦
part of RE1 . That is why the obtained amplitudes are not exactly the same at zero energy and
that is why Gerasimov’s argument, which is a combination of the (retarded) dispersion relation and
◦
of the equation R(0, 0) = RE1 (0), Eq. (3.39), has a flaw when applied to photonuclear reactions at
intermediate energies.
(ii) The second sum rule, which is briefly discussed below along the same line, appears as a
derivative d/dt of the relation (10.48) taken at the point t = 0. More specifically, (i) we consider the
case of equal photon helicities, λ1 = λ2 , so that (ǫ1 · k2 )(ǫ2 · k1 ) = (t/2)(ǫ1 · ǫ2 ), and (ii) we do not
apply the derivative to the product (ǫ1 · ǫ2 ).15 We therefore have
∞ ∂ 2π 2 e2 1 2 π 2 e2
Z
Σt ≡ σt (ω) dω = 2π 2 R1 (0, 0) = hrE i = 2 , (10.62)
0 ∂t E0 6 2γ E0
14
Since the relativistic corrections affect the energy E0 of the bound state, they change also the high energy limit of
the Compton scattering amplitude T and the magnitude of the GGT and TRK sum rule, cf. [187, 195]. However, this in
itself does not destroy Gerasimov’s theorem (10.58). What does is a projection to positive-energy states.
15
In other words, we consider the t-derivative of the invariant amplitude A1 . Cf. Eq. (8.29).

102
where R1 (ω, t) = R(ω, t; λ1 = λ2 , ǫ1 · ǫ2 = 1) and the integrand

4π ∂
σt (ω) = ImR1 (ω, 0) (10.63)
ω ∂t
is generally determined by the partial Ej and M j cross sections,
∞
(j + 2)(j − 1) 1 E2
(σ Ej + σ M j ) = (σ + σ M 2 ) + higher j.
X
σt = (10.64)
j=2
4ω 2 ω2

(In fact, the magnetic cross sections are absent in our model.) The derivative cross section σt consists
of contributions of the excitation and the pair production,
∞
X π 2 e2 −A An−2 ω2
σt = σt,+ + σt,− , σt,± (ω) = e δ(ω − En ± E0 ), A= , (10.65)
n=2
ωE0 En (n − 2)! 2γ 2

and so does the corresponding integral,

∞
π 2 e2 (A±
n)
n−2 (ωn± )2
exp(−A± A±
X
Σt = Σt,+ + Σt,− , Σt,± = ± n ), n = . (10.66)
ω E
n=2 n 0 n E (n − 2)! 2γ 2

With the relativistic parameter η ≪ 1, the pair contribution is exponentially suppressed, Σt,− ∝
exp(−1/η).
Again, we can consider the unretarded analogue of the sum rule (10.62) which involves the unre-
tarded t-derivative of the amplitude R(ω, t),

◦ 1 ◦ E2 e2 1
Rt (ω) = 2 R = 2 2
+ (ω → −ω), (10.67)
ω 2E0 4γ − ω − 2ωE0
◦ ◦
and the unretarded cross section σ t = (1/ω 2 )σ E2 . Its positive- and negative-energy components read

◦ ◦ ◦ ◦ π 2 e2 ω
σ E2 = σ E2 E2
+ + σ− , σ E2
± = δ(ω − E2 ± E0 ). (10.68)
E0 E2
The corresponding integral

◦ ∞ ◦ dω ◦ ◦ ◦ π 2 e2
Z
Σt = σ E2 (ω) 2 = Σt,+ + Σt,− , Σt,± = (10.69)
0 ω ω2± E0 E2

exactly satisfies the Gerasimov-like relation

◦
Σt = Σt . (10.70)

Again, this relation does not hold separately for the + and − components. In essence, the relation
(10.70) is based only on the assumption that the amplitude Rt (ω) satisfies an unsubtracted dispersion
◦
relation. Other ingredients of the derivation, viz. an unsubtracted dispersion relation for Rt (ω) and
◦
the identity Rt (0) = Rt (0) are model independent.
◦
Since Σt,− ≃ ηΣt , when η is small, the “practical” violation of the Gerasimov E2-theorem (10.70)
happens in first order in η whenever the integrals are truncated at nonrelativistic energies ω ≪
E0 . Details of the balance between the retarded and full unretarded sum rules are as follows. The
retardation form factor reduces the dominating E2 contribution to Σt by F 2 (ω2 ) ≃ 1 − 2η, but the E3

103
 ◦
contribution (additional +3ηΣt ) increases Σt with respect to Σt,+ by ηΣt . After taking into account
◦
Σt,− the balance is restored.
The important conclusion, which holds in a more general situation than the oscillator model sug-
gests, is that relativistic corrections to the TRK-like sum rules are of the same scale as the retardation
effects and effects of higher multipoles in the GGT-like sum rules. Taken together, they lead to an
exact balance between retarded and unretarded sum rules. Incomplete saturation of the unretarded
sum rules with nonrelativistically low energies results in an error of the same scale as the relativistic
and retardation corrections themselves.
Relativistic TRK sum rule matches the GGT sum rule, nonrelativistic TRK does not.

10.5 Comparison with the phenomenological approach

The amplitude R ≡ Rosc suggested by our oscillator model is too simple to include all details of nuclear
Compton scattering in the giant-resonance region. For example, it cannot describe the ensemble of
isoscalar giant resonances which begins with the quadrupole excitation. (The dipole isoscalar mode is
a translation.) The oscillator model better works for the isovector modes, for which the E2/E1 ratio
of excitation energies and widths is close to the oscillator value of 2. (We can include the widths into
the oscillator model by using a complex parameter ω0 , although this gives an imaginary part to a few
other parameters such as the oscillator radius or the ground-state energy E0 .) An other restriction
of the oscillator model, which is caused by a shape of the oscillator potential, is that it keeps a firm
relation between the nuclear form factor and the quadrupole photoabsorption strength which may not
work well for all nuclei. In terms of the oscillator frequency ω0 and of the small relativistic parameter
η, this relation reads
η
Z ∞ Z ∞
2 2 E2
F (ω0 ) ≃ 1 − , F (2ω0 ) ≃ 1 − 2η, σ (ω) dω ≃ η × σ E1 (ω) dω. (10.71)
2 0 0

So, we cannot directly use the oscillator-model amplitude for the comparison with or the interpretation
of experimental data on Compton scattering. But we observe that the phenomenological amplitude
RGR from Eq. (4.20), which is used for practical data analyses, indeed incorporates some gross features
of the amplitude Rosc at energies in the GR region. These include (i) the fulfillment of the GGT sum
rule, (ii) retardation form factors accompanying the lowest resonance peaks, and (iii) the asymptotic
tail O(ω −2 ). One difference which does exist between these two amplitudes is that the asymptotics of
Rosc contains a q-dependent form factor which erases the amplitude at large angles. This feature is
absent in the phenomenological amplitude RGR where the vanishing in the high-energy limit and at
all angles is caused by an exact cancellation of the high-energy limit of the ω-dependent part of RGR
and the ω-independent subtraction.
We can check how the prescription (4.20), being applied to the oscillator-model cross sections,
matches the amplitude Rosc which is known exactly. For a numerical illustration we take the param-
eters ω0 and η as determined by the isovector giant dipole and quadrupole resonances in 208 Pb [95]
(see Table 4.1):
ω0 = (13.4 − 2.1i) MeV, η = 0.06 . (10.72)
q
With these parameters, the r.m.s. radius of the oscillator is hr 2 i1/2 = ω0−1 32 η ≃ 4.4 fm. The energy
region of our interest will be ω ≤ 100 MeV, that is well below the threshold of pair production
2E0 = 2ω0 /η ≃ 450 MeV.
Applying the prescription (4.20), we act in line with the phenomenological procedure and keep in
the dispersion integrals (10.2) only the lowest E1 and E2 peaks. Thus, we omit radial excitations,

104
which give corrections of order O(η 2 ), and the pair contribution. Then we get

RGR (ω, θ) = gE1 (θ)RGR (ω, θ = 0) + (gE2 (θ) − gE1 (θ))R̂GR

E2
(ω) , (10.73)

where
ω1+ F 2 (ω1+ ) ω2+ F 2 (ω2+ ) (ω2+ )2
( )
e2 γ 2
RGR (ω, θ = 0) = + (10.74)
E0 ((ω1+ )2 − ω 2 )E1 ((ω2+ )2 − ω 2 )E2 2γ 2
and
E2 e2 γ 2 ω2+ F 2 (ω) ω 2
R̂GR (ω) = . (10.75)
E0 E2 (ω2+ )2 − ω 2 2γ 2
By construction, Rosc and the amplitudes from (10.73) are very close at zero energy (the difference
there is O(η 2 ), see the previous discussion of the GGT sum rule), near the E1 and E2 peaks, and at
forward angle. They remain very close at other energies and angles, see Fig. 10.2.
This would not be the case if the retardation form factors were removed in Eqs. (4.20), (10.74),
(10.75). Then, in particular, the backward scattering amplitude RGR (ω, 180◦ ) at energies above the GR
region would behave as RGR (∞, 180◦ ) ≃ 2ηe2 /E0 and would show a wrong asymptotic tail 2η = 0.12
in Fig. 10.2.
Thus, the oscillator model suggests that the phenomenological prescription (4.20) provides a very
good approximation to the exact amplitude.

Figure 10.2: Phenomenological, Eq. (10.73), (dashed lines) and exact (solid lines) amplitudes RGR in
the relativistic oscillator model at ω0 = (13.4 − 2.1i) MeV and η = 0.06. Units are e2 /E0 .

105
A Parameters of the mesonic seagull amplitude for different nuclei

In section 9 the parameters κ, δα and δβ, which constitute the mesonic seagull amplitude, are shown
as a function of the proton number Z. In this appendix the corresponding numerical values are given,
including all values for the various contributions to each of these parameters. In the following tables
this has been done for both, a pure Fermi gas model and a realistic nuclear density based upon the
parameterizations in [256]. The results for the mesonic seagull amplitude in nuclear matter, which
has been discussed in sections 9.1 and 9.2 are also included in the tables (denoted “A → ∞”), as
well as the case of an asymptotic correlation function F0ij as in section 9.1 (superscript “0”). The
polarizability modifications δα and δβ are given in the usual units 10−4 fm3 .

Table A.1: Numerical values for the various contributions to the enhancement constant κ shown for
nuclear matter (A → ∞) and different finite nuclei. The quantities κC C
π and κρ denote the pion and
ρ-meson contributions due to central correlations, respectively, while κTπ and κTρ denote the contribu-
tions due to tensor correlations. When only the asymptotic correlation function has been used, the
contributions are termed κ0π and κ0ρ . The right-hand half of the table shows the values for realistic
nuclear densities based on the Fermi parameterizations from [256].
constant density realistic density
A→∞ 208 Pb 40 Ca 12 C 208 Pb 40 Ca 12 C 4 He

κ0π 0.19 0.16 0.13 0.11 0.15 0.11 0.09 0.05

κCπ 0.15 0.13 0.11 0.09 0.12 0.09 0.07 0.03
κπT 0.86 0.85 0.84 0.83 0.79 0.68 0.67 0.60
κ0ρ 0.19 0.17 0.16 0.15 0.16 0.11 0.10 0.08
κρC 0.01 0 0 0 -0.03 -0.01 -0.01 -0.18
κTρ -0.11 -0.11 -0.11 -0.10 -0.06 -0.07 -0.07 -0.07
κ 0.91 0.87 0.84 0.84 0.82 0.69 0.66 0.52

Table A.2: Same as Table A.1, but for the electric polarizability modification δα. The ρ contribution
is negligible in comparison with the π contribution.
constant density realistic density
A → ∞ 208 Pb 40 Ca 12 C 208 Pb 40 Ca 12 C 4 He

δα0 -4.17 -3.58 -3.11 -2.59 -3.26 -2.44 -1.94 -1.22

C
δα -2.87 -2.42 -2.05 -1.62 -2.21 -1.51 -1.07 -0.41
δαT -0.15 -0.09 -0.04 -0.02 -0.06 0.02 0.08 0.13
δα -3.02 -2.51 -2.09 -1.64 -2.27 -1.49 -0.99 -0.28

106
Table A.3: Same as Table A.1, but for the magnetic polarizability modification δβ. The ρ contribution
is negligible in comparison with the π contribution. The subscript ∆ denotes contributions to δβ from
diagrams with a ∆-isobar as an intermediate state.
constant density realistic density
A→∞ 208 Pb 40 Ca 12 C 208 Pb 40 Ca 12 C 4 He

δβ 0 1.86 1.55 1.30 1.03 1.42 1.10 0.83 0.47

δβ C 1.54 1.31 1.13 0.91 1.25 0.93 0.69 0.30
δβ T 0.11 0 -0.10 -0.20 -0.06 -0.21 -0.30 -0.38
δβ∆0 0.29 0.24 0.19 0.19 0.22 0.17 0.12 0.05
δβ∆C 0.38 0.33 0.29 0.29 0.30 0.22 0.17 0.09
δβ∆T 0.30 0.29 0.28 0.28 0.27 0.22 0.21 0.18
δβ 2.33 1.93 1.60 1.28 1.76 1.16 0.77 0.19

B Exchange form factors

At low energies the dependence of the mesonic seagull amplitude on momentum transfer is given by
exchange form factors. In this appendix the numerical results for such form factors are presented on
the basis of the model calculation discussed in section 9. Notations for the different types of exchange
form factors are given in Table B.1.

Table B.1: Notations for the different form factors and their formulation within a specific model.

notation description model realization

F1 one-nucleon form factor taken from ref. [256]
(1) (1)
F2 exchange form factor for the static F2 (q) = Φ1 (q)/Φ1 (0)
part of the mesonic seagull amplitude
(2) (1) (2)
F2 same as F2 , but for F2 (q) = Φ2 (q)/Φ2 (0)
the energy-dependent part
(GR) (1) (GR)
F2 same as F2 , but for F2 (q) = (Φ1 (q)/Φ1 (0))|gT =0
the giant resonance contribution only
(QD) (1) (QD)
F2 same as F2 , but for F2 (q) = (Φ1 (q)/Φ1 (0))|gC =0
the quasideuteron contribution only

Figure B.1 (for 208 Pb) and Fig. B.2 (for 12 C) show the form factors in comparison with both,
the nuclear charge form factor and the form factor for uncorrelated nucleon pairs. Furthermore, our
model prediction for the giant resonance and quasideuteron exchange form factors are shown for for
40 Ca (Fig. B.3) and 12 C (Fig. B.4). Again in these two figures the nuclear charge form factor and the

approximation of uncorrelated nucleon pairs is shown for the sake of comparison.

Table B.2 gives the radii corresponding to the exchange form factors shown in Figs. B.1, B.2 and
(i) (i)
9.12, which are obtained by representing the functions F2 in the form F2 = 1 − q 2 ri2 /6.

107
Table B.2: Numerical values for the radii appearing in the expansion of the exchange form factors
from Figs. (B.1), (B.2) and (9.12).
A r1 [fm] r2 [fm]
208 5.0 4.7
40 3.0 2.5
12 1.9 1.4

(i) (1)
Figure B.1: Exchange form factors F2 (q) for the case of 208 Pb. The dashed curve represents F2 ,
(2)
while the full curve corresponds to F2 . For comparison the (experimental) charge form factor F1
(dash-dotted curve) and the function F12 (q/2) (dotted curve) are also shown. The parameters for the
form factor F1 were taken from ref. [256].

Figure B.2: Same as Fig. B.1, but for 12 C.

(GR) (QD) (QD)

Figure B.3: Form factors F2 (∆) and F2 (∆)for 40 Ca. The dashed curve is F2 and the full
(GR)
curve is F2 . In addition, the (experimental) charge form factor F1 (dash-dotted curve) and the
function F12 (∆/2) (dotted curve) are shown for the sake of comparison.

Figure B.4: Same as Fig. B.3, but for 12 C.

108
C Multipole angular distribution functions

Here the partial-wave expansion of the seagull amplitude (3.67) given in Section 3 in terms of the
functions gλL is derived. When both photons have definite helicities λ1 , λ2 = ±1 and the scattering
plane is xz, these functions read

gEl (θ) = dlλ1 ,λ2 (θ), gM l (θ) = λ1 λ2 dlλ1 ,λ2 (θ).

To expand the product ǫ1 · ǫ∗2 Pl (cos θ), we use ǫ1 · ǫ∗2 = (1 + λ1 λ2 cos θ)/2 = d1λ1 ,λ2 (θ) and Pl (cos θ) =
dl0,0 (θ) and apply the general rule of adding angular momenta for the d-functions:

djm11 ,n1 (θ) djm22 ,n2 (θ) = Cjjm C jn dj (θ).

X
1 m1 ,j2 m2 j1 n1 ,j2 n2 m,n
jmn

jλ
Using specific values for the Clebsch-Gordan coefficients C1λ,l0 with λ = ±1, we get

l+2 λ1 λ2 l l−1
d1λ1 ,λ2 (θ) dl0,0 (θ) = dl+1
λ1 ,λ2 (θ) + dλ1 ,λ2 (θ) + dl−1 (θ).
2(2l + 1) 2 2(2l + 1) λ1 ,λ2

Coming back to the functions gλL , we find

l + 2 E(l+1) 2l + 1 M l l − 1 E(l−1)
Gl (θ) ≡ (2l + 1)ǫ1 · ǫ∗2 Pl (cos θ) = g (θ) + g (θ) + g (θ)
2 2 2
which is Eq. (3.69).

Acknowledgements

The authors are indebted to Prof. G.E. Brown for his continuing interest to this work. They thank
Deutsche Forschungsgemeinschaft for support of several visits of A.I.M. and A.I.L. in Göttingen and
Deutscher Akademischer Austauschdienst for support of visits of M.T.H. in Novosibirsk. One of us
(M.S.) thanks B. Schröder (MAX–LAB, Lund) and B. Ziegler, J. Ahrens and Th. Walcher (MAMI,
Mainz) for fruitful cooperations during the experiments and Deutsche Forschungsgemeinschaft for the
support of the experimental work.

109
 [1] A.H. Compton, Phys. Rev. 21 (1923) 484

[2] P.P. Kane, L. Kissel, R.H. Pratt, S.C. Roy, Phys. Rep. 140 (1986) 75

[3] P.P. Kane, Phys. Rep. 218 (1992) 67

[4] P. Papatzacos and K. Mork, Phys. Rep. 21 (1975) 81

[5] A.I. Milstein, M. Schumacher, Phys. Rep. 243 (1994) 183

[6] E.R. Gaerttner, G.L. Yeater, Phys. Rev. 76 (1949) 363

[7] R. Dressel, M. Goldhaber, A.O. Hanson, Phys. Rev. 77 (1950) 754

[8] M. Goldhaber and E. Teller, Phys. Rev. 74 (1948) 1046

[9] M.B. Stearns, Phys. Rev. 87 (1952) 706

[10] E.G. Fuller, E. Hayward, Phys. Rev. 94 (1954) 732; 95 (1954) 1106

[11] E.G. Fuller, E. Hayward, Phys. Rev. 101 (1956) 692

[12] E. Hayward, W.C. Barber, J. Sazama, Phys. Rev. C 8 (1973) 1065

[13] W.R. Dodge, E. Hayward, R.G. Leicht, B.H. Patrick, R. Starr, Phys. Rev. Lett. 44 (1980) 1040

[14] E. Hayward, R.G. Leicht, Annals of the New York Academy of Sciences 40 (1980) 99

[15] R. Leicht, M. Hammen, K.P. Schelhaas, B. Ziegler, Nucl. Phys. A 326 (1981) 111

[16] W.R. Dodge, E. Hayward, R.G. Leicht, M. McCord, R. Starr, Phys. Rev. C 28 (1983) 8

[17] E. Hayward, B. Ziegler, Nucl. Phys. A 414 (1984) 333

[18] D.H. Wright, A.M. Nathan, L.J. Morford, P.T. Debevec, Phys. Rev. Lett. 52 (1984) 244

[19] D.H. Wright, P.T. Debevec, L.J. Morford, A.M. Nathan, Phys. Rev. C 32 (1985) 1174

[20] E.J. Austin, E.C. Booth, E.K. McIntyre, J.P. Miller, B.L. Roberts, D.A. Whitehouse, G. Dodson,
Phys. Rev. Lett. 57 (1986) 972

[21] A.M. Nathan, P.L. Cole, P.T. Debevec, S.D. Hoblit, S.F. LeBrun, D.H. Wright, Phys. Rev. C
34 (1986) 480

[22] S.F. LeBrun, A.M. Nathan, S.D. Hoblit, Phys. Rev. C 35 (1987) 2005

[23] E.J. Austin, E.C. Booth, D. Delli Carpini, K.P. Gall, E.K. McIntyre, J.P. Miller, D. Warner,
D.A. Whitehouse, G. Dodson, Phys. Rev. Lett. 61 ( 1988) 1922

[24] A. Baumann, P. Rullhusen, K.W. Rose, M. Schumacher, J.M. Henneberg, N. Wieloch-Laufen-

berg, B. Ziegler, Phys. Rev. C 38 (1988) 1940

[25] K.P. Schelhaas, J. Ahrens, J.M. Henneberg, M. Sanzone-Arenhövel, N. Wieloch-Laufenberg, U.

Zurmühl, B. Ziegler, M. Schumacher, F. Wolf, Nucl. Phys. 489 (1988) 189

[26] K.P. Schelhaas, J.M. Henneberg, N. Wieloch-Laufenberg, U. Zurmühl, B. Ziegler, M. Schu-

macher, F. Wolf, Nucl. Phys. A 506 (1990) 307

110
[27] B. Ziegler, Phys. Lett. B 237 (1990) 334

[28] D.P. Wells, Doctoral Thesis, University of Illinois at Urbana-Champaign, 1990 (unpublished)

[29] D. Delli Carpini, E.C. Booth, J.P. Miller, R. Igarashi, J. Bergstrom, H. Caplan, M. Doss, E.
Hallin, C. Rangacharyulu, D. Skopic, M.A. Lucas, A.M. Nathan, D.P. Wells, Phys. Rev. C 43
(1991) 1525

[30] K.P. Schelhaas, J.M. Henneberg, N. Wieloch-Laufenberg, U. Zurmühl, B. Ziegler, M. Schu-

macher, F. Wolf, Nucl. Phys. A 528 (1991) 189

[31] A. Baumann, P. Rullhusen, K.W. Rose, M. Ludwig, M. Schumacher, A.I. Milstein, J.M. Hen-
neberg, N. Wieloch-Laufenberg, B. Ziegler, Nucl. Phys. A 536 (1992) 87

[32] K. Fuhrberg, G. Martin, D. Häger, M. Ludwig, M. Schumacher, B.-E. Andersson, K.I. Blomqvist,
H. Ruijter, B. Sondell, E. Hayward, L. Nilsson, B. Schröder and R. Zorro, Nucl. Phys. A548
(1992) 579

[33] D.S. Dale, R.M. Laszewski, R. Alarcon, Phys. Rev. Lett. 68 (1992) 3507

[34] M. Ludwig, B.-E. Andersson, A. Baumann, K.I. Blomqvist, K. Fuhrberg, E. Hayward, G. Müller,
D. Nilsson, A. Sandell, B. Schröder, M. Schumacher, Phys. Lett. B 274 (1992) 275

[35] J.P. Miller, E.J. Austin, J. Bergstrom, E.C. Booth, H. Caplan, D. Delli Carpini, G. Dodson,
M. Doss, K.P. Gall, E. Hallin, R. Igarashi, M.A. Lucas, E.K. McIntyre, A.M. Nathan, C. Ran-
gacharyulu, D. Skopic, D. Warner, D.P. Wells, D.A. Whitehouse, Nucl. Phys. A 546 (1992)
199c

[36] D.P. Wells, D.S. Dale, R.A. Eisenstein, F.J. Federspiel, M.A. Lucas, K.E. Mellendorf, A.M.
Nathan, A.E. O’Neill, Phys. Rev. C 46 (1992) 449

[37] F. Wissmann, J. Peise, M. Schmitz, M. Schneider, J. Ahrens, I. Antony, R. Beck, B. Dolbilkin,

S.J. Hall, S. Herdade, J. Herrmann, A. Hünger, P. Jennewein, J.D. Kellie, R. Kondratjev, H.-
P. Krahn, K.-H. Krause, V. Kusnezov, V. Lisin, G.J. Miller, A. Polonski, M. Schumacher, J.
Sobolewski, Th. Walcher, A. Zabrodin, Phys. Lett. B 335 (1994) 119

[38] K. Fuhrberg, D. Häger, T. Glebe, B.-E. Andersson, K. Hansen, M. Hütt, B. Nilsson, L. Nilsson,
D. Ryckbosch, B. Schröder, M. Schumacher, R. Van de Vyver, Nucl. Phys. A 591 (1995) 1

[39] D. Häger, K. Fuhrberg, T. Glebe, M. Hütt, M. , M. Schumacher, B.E. Andersson, K. Hansen,

B. Nilsson, B. Schröder, L. Nilsson, H. Freiesleben, E. Kuhlmann, Nucl. Phys. A 595 (1995) 287

[40] R. Igarashi, J.C. Bergstrom, H.S. Caplan, K.G.E. Doss, E.L. Hallin, D.M. Skopik, D. Delli
Carpini, E.C. Booth, E.K. McIntyre, J.P. Miller, M.A. Lucas, B.R. MacGibbon, A.M. Nathan,
D. Wells, Phys. Rev. C 52 (1995) 755

[41] T. Glebe, Doctoral Thesis, University of Göttingen 1996 (unpublished)

[42] C. Pöch, Doctoral Thesis, University of Göttingen 1996 (unpublished)

[43] G. Feldman, M.E. Mellendorf, R.A. Eisenstein, F.J. Federspiel, G. Garino, R. Igarashi, N. R.
Kolb, M.A. Lucas, B.E. MacGibbon, W.K. Mize, A.M. Nathan, R. E. Pywell, D.P. Wells, Phys.
Rev. C 54 (1996) R2124

[44] O. Selke, A. Hünger, A. Kraus, M. Schumacher, R. Wichmann, J. Ahrens, H.J. Arends, R. Beck,
J. Peise, M. Schneider, F. Wissmann, Phys. Lett. B 369 (1996) 207

111
[45] D. Moricciani, V. Bellini, M. Capogni, A. Caracappa, L. Casano, R.M. Chasteler, A. D’Angelo,
F. Ghio, B. Girolami, S. Hoblit, L. Hu, M. Khandaker, O.C. Kistner, L.H. Kramer, C.M.
Laymon, A.I. L’vov, B. Marks, L. Micele, V.A. Petrun’kin, B.J. Rice, A.M. Sandorfi, C. Schaerf,
C.E. Thorn, D.E. Tilley, H.R. Weller, Few Body Systems Suppl. 9 (1995) 349

[46] A. Kraus, O. Selke, F. Wissmann, J. Ahrens, H.-J. Arends, R. Beck, G. Galler, M.-Th. Hütt, B.
Körfgen, J. Peise, M. Schumacher, F. Smend, R. Wichmann, Phys. Lett. B 432 (1998) 45

[47] S. Proff, Doctoral Thesis, University of Göttingen 1998 (unpublished)

[48] S. Proff, C. Pöch, T. Glebe, J.-O. Adler, K. Fissum, K. Hansen, M.-Th. Hütt, O. Kaltschmidt,
M. Lundin, B. Nilsson, B. Schröder, M. Schumacher, D. Sims, F. Smend, F. Wissmann, Nucl.
Phys. A (in press)

[49] T.J. Bowles, R.J. Holt, H.E. Jackson, R.M. Laszewski, A.M. Nathan, J.R. Specht, R. Starr,
Phys. Rev. Lett. 41 (1978) 1095

[50] C.E. Thorn, G. Giordano, O.C. Kistner, G. Matone, A.M. Sandorfi, C. Schaerf, C.S. Whisnant,
Nucl. Instr. Meth. A 285 (1989) 447

[51] J.-O. Adler, B.-E. Andersson, K.I. Blomqvist, B. Forkman, K. Hansen, L. Isaksson, K. Lindgren,
D. Nilsson, A. Sandell, B. Schröder, K. Ziakas, Nucl. Instr. Meth. A 294 (1990) 15

[52] J.-O. Adler, B.-E. Andersson, K.I. Blomqvist, K.G. Fissum, K. Hansen, L. Isaksson, B. Nilsson,
H. Ruijter, A. Sandell, B. Schröder, D.A. Sims, Nucl. Instr. Meth. A 388 (1997) 17

[53] I. Anthony, J.D. Kellie, S.J. Hall, G.J. Miller, J. Ahrens, Nucl. Instr. Meth. A 301 (1991) 230

[54] S.J. Hall, G.J. Miller, R. Beck, P. Jennewein, Nucl. Instr. Meth. A 368 (1996) 698

[55] J.M. Vogt, R.E. Pywell, D.M. Skopic, E.L. Hallin, J.C. Bergstrom, H.S. Caplan, K.I. Blomqvist,
W. Del Bianco, J.W. Juri, Nucl. Instr. Meth. A 324 (1993) 198

[56] D. Lohmann, J. Peise, J. Ahrens, I. Anthony, H.J. Arends, R. Beck, R. Crawford, A. Hünger,
K.H. Kaiser, J.D. Kellie, Ch. Klümper, H.-P. Krahn, A. Kraus, U. Ludwig, M. Schumacher, O.
Selke, M. Schmitz, M. Schneider, F. Wissmann, S. Wolf
Nucl. Instr. Meth A 343 (1994) 494

[57] A. Kraus, O. Selke, F. Rambo, G. Galler, M. Schumacher, F. Smend, R. Wichmann, F. Wiss-

mann, J. Ahrens, H.-J. Arends, R. Beck, J. Peise,
Phys. Rev. Lett. 17 (1997) 3834

[58] F. Rambo, G. Galler, A. Kraus, M. Schumacher, F. Smend, F. Wissmann, S. Wolf,J. Ahrens,

H.-J. Arends, R. Beck, H.-P. Krahn, J. Peise, Phys. Rev. C 58 (1998) 489

[59] K.-H. Krause, J. Sobolewski, D. Hauff, J.M. Henneberg, N. Wieloch-Laufenberg, J. Ahrens, H.

Gimm, A. Zieger, B. Ziegler, Nucl. Instr. Meth. A 310 (1991) 577

[60] S. Kahane, R. Moreh, Phys. Rev. C 9 (1974) 2384

[61] M. Schumacher, F. Smend, W. Mückenheim, P. Rullhusen, Z. Physik A 300 (1981) 193

[62] P. Rullhusen, U. Zurmühl, F. Smend, M. Schumacher, H.G. Börner, S.A. Kerr, Phys. Rec. C 27
(1983) 559

[63] R. Nolte, A. Baumann, K.W. Rose, M. Schumacher, Phys. Lett. B 173 (1986) 388

112
[64] K. Fuhrberg, G. Mondry, D. Novotny, H.-J. Brinkmann, M. Ludwig, G. Müller, W. Scharfe, M.
Schumacher, Europhys. Lett. 9 (1989) 427

[65] R. Nolte, F. Schröder, A. Baumann, K.W. Rose, K. Fuhrberg, M. Schumacher, P. Fettweis, R.

Carchon, Phys. Rev. C 40 (1989) 1175

[66] K. Fuhrberg, R. Nolte, A. Baumann, P. Rullhusen, M. Schumacher, Phys. Rev. C 40 (1989)

2394

[67] J.S. Levinger and H.A. Bethe, Phys. Rev. 78 (1950) 115

[68] P. Christillin and M. Rosa-Clot, Phys. Lett. B51 (1974) 125

[69] J.L. Friar, Ann. Phys. 95 (1975) 170

[70] J.L. Friar, Phys. Rev. Lett. 36 (1976) 510

[71] P. Christillin and M. Rosa-Clot, Nuovo Cimento 28A (1975) 29

[72] P. Christillin and M. Rosa-Clot, Nuovo Cimento 43A (1978) 172

[73] P. Christillin, in: Proceedings of the Workshop “Intermediate Energy Nuclear Physics with
Monochromatic and Polarized Photons” (Frascati 1980)

[74] H. Arenhövel, Z. Phys. A 297 (1980) 129

[75] W.A. Alberico and A. Molinari, Z. Phys. A 309 (1982) 143

[76] M. Ericson and M. Rosa-Clot, Nuovo Cim. 76A (1983) 180

[77] P. Christillin, J. Phys. G: Nucl. Phys. 10 (1984) L65

[78] M. Weyrauch, H. Arenhövel, Phys. Lett. 134B (1984) 21

[79] P. Christillin, J. Phys. G: Nucl. Phys. 11 (1985) 795

[80] M. Rosa-Clot, M. Ericson, Z. Phys. A 320 (1985) 675

[81] P. Christillin, J. Phys. G: Nucl. Phys. 12 (1986) 837

[82] M. Ericson and M. Rosa-Clot, Z. Phys. A 324 (1986) 373

[83] M. Ericson, M. Rosa-Clot, Phys. Lett. B 188 (1987) 11

[84] J. Ahrens, H. Borchert, K.H. Czok, H.B. Eppler, H. Gimm, H. Gundrum, M. Kröning, P. Riehn,
G. Sita Ram, A. Zieger, B. Ziegler, Nucl. Phys. A251 (1975) 479

[85] A.J.F. Siegert, Phys. Rev. 52 (1937) 787

[86] A.B.Migdal, A.A.Lushnikov and D.F.Zaretsky, Nucl. Phys. 66 (1965) 193.

[87] G.E. Brown and M. Rho, Nucl. Phys. A338 (1980) 269

[88] G.E. Brown, W. Weise, G. Baym, J. Speth, Comments Nucl. Part. Phys. 17 (1987) 39

[89] G.E. Brown, M. Rho, Phys. Lett. B 222 (1989) 324

[90] G.E. Brown, H. Müther, M. Prakash, Nucl. Phys. A 506 (1990) 565

[91] G.E. Brown, M. Rho, Phys. Lett. B237 (1990) 3

113
 [92] G.E. Brown, M. Rho, Phys. Rev. Lett. 66 (1991) 2720

[93] W. Weise, Nucl. Phys. A 553 (1993) 59c

[94] G.E. Brown, M. Rho, M. Soyeur, Nucl. Phys. A 553 (1993) 705c

[95] M. Schumacher,A.I. Milstein, H. Falkenberg, K. Fuhrberg, T. Glebe, D. Häger, M.-Th. Hütt,

Nucl. Phys. A 576 (1994) 603

[96] “Chiral Symmetry and Changes of Properties in Nuclei” G.E. Brown, in:“Symmetries and fun-
damental interactions in nuclei” E. Henley and W. Haxton Eds., World Scientific 1995

[97] B. Friman, M. Rho, Nucl.Phys. A 606 (1996) 303

[98] G.E. Brown, M. Rho, Nucl. Phys. A 596 (1996) 503

[99] V.A. Petrunkin, Sov. J. Part. Nucl. 12 (1981) 278

[100] A.I. L’vov, Int. J. Mod. Phys. A 8 (1993) 5267

[101] A.I. L’vov, V.A. Petrun’kin, M. Schumacher, Phys. Rev. C 55 (1997) 359

[102] D. Drechsel, A. Russo, Phys. Lett. 137 B (1984) 294

[103] A.I. Lvov, M. Schumacher, Nucl. Phys. A 548 (1992) 613

[104] G.G. Bunatian, Sov. J. Nucl. Phys. 55 (1992) 1781

[105] E. Oset, W. Weise, Nucl. Phys. A 319 (1979) 477

[106] E. Oset, W. Weise, Nucl. Phys. A 368 (1981) 375

[107] J.H. Koch, E.J. Moniz, N. Ohtsuka, Ann. Phys. 154 (1984) 99

[108] B. Körfgen, F. Osterfeld, T. Udagawa, Phys. Rev. C 50 (1994) 1637

[109] B. Pasquini, S. Boffi, Nucl. Phys. A 598 (1996) 485

[110] T. Bar-Noy, R. Moreh, Nucl. Phys. A 229 (1974) 417

[111] T. Bar-Noy, R. Moreh, Nucl. Phys. A 275 (1977) 151

[112] A.M. Nathan, R. Moreh, Phys. Lett. 91B (1980) 38

[113] T.J. Bowles, R.J. Holt, H.E. Jackson, R.M. Laszewski, R.D. McKeon, A.M. Nathan, J.R. Specht,
Phys. Rev. C 24 (1981) 1940

[114] R. Alarcon, A.M. Nathan, S.F. LeBrun, S.D. Hoblit, Phys. Rev. C 39 (1989) 324

[115] G. Mondry, F. Wissmann, G. Müller, F. Schröder, P. Rullhusen, F. Smend, M. Schumacher, P.

Fettweis, R. Carchon, Nucl. Phys. A 531 (1991) 237

[116] U. Fano, Natl. Bureau of Standards Technical Note 83 (1960) 1

[117] H. Arenhövel, H.J. Weber, Nucl. Phys. A 91 (1967) 145

[118] H. Arenhövel, M. Danos, W. Greiner, Phys. Rev. 157 (1967) 1109

[119] H. Arenhövel, W. Greiner, Progr. Nucl. Phys. 10 (1969) 167

114
[120] H. Arenhövel, in: International Conf. Photonuclear Reactions and Applications, Asilomar 1973,
B.L. Berman Ed., Oak Ridge, Tennessee (1973)

[121] H. Arenhövel, in: New Vistas in Electro-Nuclear Physics, E.L. Tomusiak, H.S. Caplan and E.T.
Dressler Eds., NATO ASI Series B 142, Plenum Press, New York and London (1985) 251
[122] H. Arenhövel, M. Weirauch, Nucl. Phys. A457 (1986) 573

[123] U. Kneissl, Prog. Part. Nucl. Phys. 24 (1990) 41, ibid. 28 (1992) 331

[124] E.G. Fuller,E. Hayward,in: Photonuclear Reactions II, P.M. Endt, P.B. Smith Eds., North.
Holland (1962) 113
[125] E. Hayward, in: Photonuclear Reactions I, S. Costa and C. Schaerf Eds., Lecture Notes in
Physics 61, Springer, Heidelberg, (1977) 342

[126] A. Molinari, Phys. Rep. 64 (1980) 283

[127] B. Ziegler, in: New Vistas in Electro-Nuclear Physics, E.L. Tomusiak, H.S. Caplan and E.T.
Dressler Eds., NATO ASI Series B 142, Plenum Press, New York and London (1985) 293
[128] P. Rullhusen, in: Perspectives on Photon Interactions with Hadrons and Nuclei, M. Schumacher
and G. Tamas Eds., Lecture Notes in Physics, Springer, Heidelberg, (1990) 23

[129] A.I. L’vov and V.A. Petrun’kin, in: Perspectives on Photon Interactions with Hadrons and
Nuclei, M. Schumacher and G. Tamas Eds., Lecture Notes in Physics, Springer, Heidelberg,
(1990) 123
[130] P. Christillin, Phys. Rep. C 190 (1990) 63

[131] M.A. Lucas, PhD Thesis, Univ. Illinois at Urbana-Champaign 1994.

[132] T. Wilbois, P. Wilhelm and H. Arenhövel, Few-Body Syst. Suppl. 9 (1995) 263.

[133] M.I. Levchuk and A.I. L’vov, Few-Body Syst. Suppl. 9 (1995) 439
[134] M.I. Levchuk and A.I. L’vov, Proc. 7th Conf. Mesons and Light Nuclei, Prague 1998 (World
Scientific, eds. J. Adam et al.); nucl-th/9809034
M.I. Levchuk and A.I. L’vov (submitted).

[135] J.-W. Chen, H. Grießhammer, M.J. Savage and R.P. Springer, Nucl. Phys. A 644 (1998) 245;
J.-W. Chen (submitted), nucl-th/9810021.
[136] J.J. Karakowski, PhD thesis, Univ. of Washington (1998), nucl-th/9901011
J.J. Karakowski and G.A. Miller (submitted), nucl-th/9901018.

[137] T.H. Bauer, R.D. Spital and D.R. Yennie, Rev. Mod. Phys. 50 (1978) 261

[138] P. Kroll, W. Schweiger, M. Schürmann, Int. Jour. of Mod. Physics, A6 (1991) 4107
[139] P. Kroll, M. Schürmann, P. Guichon, Nucl. Phys. A 598 (1996) 435

[140] X. Ji, Phys. Rev. D 55 (1997) 7114

[141] A.V. Radyushkin, Phys. Rev. D 56 (1997) 5524

[142] T.A. Armstrong, W.R. Hogg, G.M. Lewis, A.W. Robertson, G.R. Brookes, A.S. Clough, J.H.
Freeland, W. Galbraith, A.F. King, W.R. Rawlinson, N.R.S. Tait, J.C. Thompson, D.W.L.
Tolfree, Phys. Rev. D 5 (1972) 1640; Nucl. Phys. B 41 (1972) 445.

115
[143] Lecture Notes in Physics 365, M. Schumacher, G. Tamas Eds., Springer, Berlin (1990)

[144] M. Schumacher, J. Phys. G: 14 Suppl. (1988) S235

[145] U. Zurmühl, P. Rullhusen, F. Smend, M. Schumacher, H.G. Börner, S.A. Kerr, Z. Phys. A 314
(1983) 171

[146] M. MacCormic, J. Habermann, J. Ahrens, G, Audit, R. Beck, A. Braghieri, G. Galler, N. d’

Hose, V. Isbert, P. Pedroni, T. Pinelli, G. Tamas, S. Wartenberg, A. Zabrodin, Phys. Rev. C55
(1997) 1033.

[147] J. Ahrens, private communication

[148] J. Ahrens, Nucl. Phys. A 446 (1985) 229c

[149] H. Steinwedel und J.H.D. Jensen, Z. Naturforsch 5a (1950) 413

[150] B.L. Berman, ATOMIC DATA AND NUCLEAR DATA TABLES 15 (1975) 319

[151] S.S. Dietrich, B.L. Berman, ATOMIC DATA AND NUCLEAR DATA TABLES 38 (1988) 199

[152] A. Leprêtre, H. Beil, R. Bergère, P. Carlos, J. Fagot, A. Veyssière, J. Ahrens, P. Axel, U. Kneissl,
Phys. Lett. 79B (1978) 43

[153] M. Schumacher, U. Zurmühl, F. Smend, R. Nolte, Nucl. Phys. A 438 (1985) 493

[154] J.S. Levinger, Phys. Rev. 84 (1951) 43

[155] K. Gottfried, Nucl. Phys. 5 (1958) 557

[156] J. M. Laget, Lecture Notes in Physics 137 (1981) 148

[157] M.-Th. Hütt, A.I. Milstein, M. Schumacher, Nucl. Phys. A 618 (1997) 483

[158] R. Wichmann,J. Ahrens, H.-J. Arends, R. Beck, G. Galler, A. Hünger, A. Kraus, I.J.D. Mac-
Gregor, J. Peise, B. Seitz, M. Schmitz, M. Schneider, O. Selke, M. Schumacher, F. Wissmann,
Z. Physik A 355 (1996) 169

[159] M. Wakamatsu and K. Matsumoto, Nucl. Phys. A392 (1983) 323

[160] R.C. Carrasco and E. Oset, Nucl. Phys. A536 (1992) 445

[161] J. Ryckebusch, M. Vanderhaeghen, L. Machenil, M. Waroquier, Nucl. Phys. A568 (1994) 828

[162] M. Vanderhaeghen, K. Heyde, J. Ryckebusch, M. Waroquier, Nucl. Phys. A595 (1995) 219

[163] J. Ryckebusch, L. Machenil, M. Vanderhaeghen, V. Van der Sluys and M. Waroquier, Phys.
Rev. C49 (1994) 2704

[164] J. Ryckebusch, L. Machenil, M. Vanderhaeghen, V. Van der Sluys and M. Waroquier, Phys.
Lett. B291 (1992) 213

[165] M. Gell-Mann, M.L. Goldberger, W. Thirring, Phys. Rev. 95 (1954) 1612

[166] M. Schumacher, P. Rullhusen and A. Baumann, Nuovo Cim. 100A (1988) 339

[167] D. Kurath, Phys. Rev. 130 (1963) 1525

[168] M.-Th. Hütt and A.I. Milstein, Nucl. Phys. A 609 (1996) 391

116
[169] E. Hayward, Radiat. Phys. Chem. 41 (1993) 739

[170] H. Falkenberg, K. Fuhrberg, T. Glebe, D. Häger, A. Hünger, M. Hütt, A. Kraus, M. Schumacher

and O. Selke, in: “Mesons and Nuclei at Intermediate Energies”, Dubna, Russia,1994, Mikhail
Kh. Khankhasayev and Zh.B. Kurmanov Eds., World Scientific 1995

[171] W.T. Weng, T.T.S. Kuo and G.E. Brown, Phys. Lett. 46B (1973) 329

[172] A. Arima, G.E. Brown, J. Hyuga, M. Ichimura, Nucl. Phys. A 205 (1973) 27

[173] “Pions and Nuclei”, T. Ericson and W. Weise, Clarendon Press, Oxford 1988

[174] D.O. Riska, Phys. Rep. 181 (1989) 207

[175] T.E.O. Ericson and J. Hüfner, Nucl. Phys. B 57 (1973) 604

[176] N. Baron, G. Baur, M. Schumacher, Nucl. Phys. A 530 (1991) 267

[177] W. Franz, Z. Physik 98 (1935) 314

[178] G.E. Brown and D.F. Mayers, Proc. Roy. Soc. 242 (1957) 89 and references therein

[179] M. Schumacher, Phys. Rev. 182 (1969) 7

[180] W. Weise, Phys. Rev. Lett. 31 (1973) 773

[181] W. Weise, Phys. Reports 13 (1974) 53

[182] J. Ahrens, L.S. Ferreira, W. Weise, Nucl. Phys. A 485 (1988) 621

[183] R. Silbar, C. Werntz and H. Überall, Nucl. Phys. A107 (1968) 655

[184] W. Thirring, Phil. Mag. 41 (1950) 1193

[185] S.B. Gerasimov, Phys. Lett. 13 (1964) 240

[186] T. Matsuura and K. Yazaki, Phys. Lett. B46 (1973) 17

[187] J.L. Friar and S. Fallieros, Phys. Rev. C 11 (1975) 274; ibid. p. 277

[188] H.P. Schröder and H. Arenhövel, Z. Phys. A 280 (1977) 349

[189] H. Arenhövel and D. Drechsel, Phys. Rev. C 20 (1979) 1965

[190] K.-M. Schmitt and H. Arenhövel, Z. Phys. A 320 (1985) 311

[191] G.E. Brown, M. Bolsterli, Phys. Rev. Lett. 3 (1959) 472

[192] A. Bohr and B.R. Mottelson , Nuclear Structure, Vol. II Nuclear Deformations, W.A. Benjamin
Reading, Massachusetts (1975) 477

[193] M.-Th. Hütt and A.I. Milstein, Phys. Rev. C 57 (1998) 305

[194] T. deForest, J.D. Walecka, Adv. Phys. 15 (1966) 1

[195] M.L. Goldberger and F.E. Low, Phys. Rev. 176 (1968) 1778

[196] V.S. Barashenkov and H.J. Kaiser, Fortschr. Phys. 10 (1962) 33

[197] V.A. Petrun’kin, Nucl. Phys. 55 (1964) 197

117
[198] A. Veyssière, H. Beil, R. Bergère, P. Carlos, A. Leprêtre, Nucl. Phys. 159 (1970) 561

[199] F.E. Bertrand, Ann Rev. Nucl. Sci. 26 (1976) 457; A. van der Woude, in [204] p. 65

[200] B.L. Berman, R.E. Pywell, S.S. Dietrich, M.N. Thompson, K.G. McNeill, J.W. Jury, Phys. Rev
C 36 (1987) 1286

[201] D.S. Dale, A.M. Nathan, F.J. Federspiel, S.D. Hoblit, J. Hughes, and D. Wells, Phys. Lett. B
214 (1988) 329

[202] R.R. Harvey, J.T. Caldwell, R.L. Bramblett, S.C. Fultz, Phys. Rev. B 136 (1964) 126

[203] L.M. Young, Ph.D. thesis, University of Illinois at Urbana-Champaign, 1972 (unpublished)

[204] Giant Multipole Resonances, ed. F.E. Bertrand, Harwood academic publisher, Chur, London,
New York (1980)

[205] R.M. Laszewski, P. Rullhusen, S.D. Hoblit, S.F. LeBrun, Phys. Rev. Lett. 54 (1985) 530

[206] K. Wienhard, K. Ackermann, K. Bangert, U.E.P. Berg, C. Bläsing, W. Naatz, A. Ruckelshausen,

D. Rück, R.K.M. Schneider, R. Stock, Phys. Rev. Lett. 49 (1982) 18

[207] R. Pitthan, Th. Walcher, Phys. Lett. 36B (1971) 563

[208] A. Van Der Woude, in [204] p. 69

[209] R. Pitthan, in [204] p. 161

[210] D.M. Drake, S. Joly, L. Nilsson, S.A. Wender, K. Aniol, I. Halpern, D. Storm, Phys. Rev. Lett
47 (1981) 1581

[211] R. Zorro, I. Bergqvist, S. Crona, A. Hakansson, A. Likar, A. Lindholm, L. Nilsson, N. Olsson,

Nucl. Phys. A 472 (1987) 125

[212] I. Bergqvist, R. Zorro, A. Hakansson, A. Lindholm, L. Nilsson, N. Olsson, A. Likar, Nucl. Phys.
A 419 (1984) 509

[213] C.M. Laymon, R.O. Nelson, S.A. Wender, L.R. Nilsson, Phys. Rev. C 46 (1992) 1880

[214] D.A. Sims, G.J. O’Keefe, R.P. Rassool, A. Kuzin, M.N. Thompson, J.O. Adler, B.-E. Andersson,
K.G. Fissum, K. Hansen, L. Isaksson, B. Nilsson, H. Ruijter, A. Sandell, B. Schröder, J.R.M.
Annand, G.I. Crawford, P.D. Harty, J.C. McGeorge, G.J. Miller, Phys. Rev. C 55 (1997) 1288

[215] A. Migdal, Journ. Phys. USSR 8 (1944) 331

[216] J.S. Levinger, Nuclear Photo-Disintegration (Oxford University Press, 1960)

[217] A.M. Baldin, Nucl. Phys. 18 (1960) 318;

L.I. Lapidus, Sov. Phys. JETP 16 (1963) 964

[218] J.L. Friar, S. Fallieros, Am. J. Phys. 49 (1981) 847

[219] A.I. L’vov, Nucl. Phys. A 638 (1998) 756

[220] J.S. Levinger, Phys. Lett. B82 (1979) 181

[221] J.M. Laget, Nucl. Phys. A312 (1978) 265

[222] A.I. L’vov, preprint Lebedev Phys. Inst. 344 (1987) (unpublished)

118
[223] R. Weiner and W. Weise, Phys. Lett. B 159 (1985) 85

[224] V. Bernard, N. Kaiser and U.-G. Meissner, Int. J. Mod. Phys. E 4 (1995) 193

[225] M. Damashek and F.J. Gilman, Phys. Rev. D1 (1970) 1319

[226] A.I. L’vov, V.A. Petrun’kin and S.A. Startzev, Sov. J. Nucl. Phys. 29 (1979) 651

[227] D. Babusci, G. Giordano and G. Matone, Phys. Rev. C 57 (1998) 291

[228] D. Babusci, G. Giordano, A.I. L’vov, G. Matone and A.M. Nathan, Phys. Rev. C 58 (1998) 1013

[229] A. Hünger, J. Peise, A. Robbiano, J. Ahrens, I. Antony, H.-J. Arends, R. Beck, G.P, Capitani,
B. Dolbilkin, H. Falkenberg,G. Galler, S.J. Hall, J.D. Kellie, R. Kondratjev, R. Kordsmeier, V.
Lisin, A.I. L’vov, G.J. Miller, C. Molinari, P. Ottonello, A.R. Reolon, M. Sanzone, M. Schmitz,
M. Schneider, M. Schumacher, O. Selke, Th. Walcher, F. Wissmann, S. Wolf, A. Zucchiatti,
Nucl. Phys. A. 620 (1997) 385

[230] R.A. Arndt, I.I. Strakovsky and R.L. Workman, Phys. Rev. C 56 (1997) 577

[231] O. Hanstein, D. Drechsel and L. Tiator, Nucl. Phys. A 632 (1998) 561

[232] R.A. Arndt, I.I. Strakovsky, and R.L. Workman, Phys. Rev. C 53 (1996) 430. The SAID program
is accessible at URL http://clsaid.phys.vt.edu/ CAPS/

[233] V.I. Goldansky, O.A. Karpukhin, A.V. Kutsenko and V.V. Pavlovskaya, Nucl. Phys. 18 (1960)
473

[234] P. Baranov, G. Buinov, V. Godin, V. Kuznetsova, V. Petrun’kin, L. Tatarinskaya, V. Shirchenko,

L. Shtarkov, V. Yurchenko and Yu. Yanulis, Phys. Lett. B 52 (1974) 122

[235] F.J. Federspiel, R.A. Eisenstein, M.A. Lucas, B.E. MacGibbon, K. Mellendorf, A.M. Nathan,
A. O’Neill and D.P. Wells, Phys. Rev. Lett. 67 (1991) 1511

[236] A. Zieger, R. Van de Vyver, D. Christmann, A. De Graeve, C. Van den Abeele and B. Ziegler,
Phys. Lett. B 278 (1992) 34

[237] B.E. MacGibbon, G. Garino, M.A. Lucas, A.M. Nathan, G. Feldman and B. Dolbilkin, Phys.
Rev. C 52 (1995) 2097

[238] C. Caso et al., Review of Particle Physics, Eur. Phys. J. C 3 (1998) 1

[239] K.W. Rose, B. Zurmuhl, P. Rullhusen, M. Lugwig, A. Baumann, M. Schumacher, J. Ahrens, A.

Zieger, D. Christman and B. Ziegler, Nucl. Phys. A 514 (1990) 621

[240] J. Schmiedmayer, P. Riehs, J.A. Harvey and N.W. Hill, Phys. Rev. Lett. 66 (1991) 1015

[241] M. Bawin and S.A. Coon, Phys. Rev. C 55 (1997) 419

[242] L. Koester, W. Waschkowski, L.V. Mytsina, G.S. Samosvat, P. Prokofjevs and J. Tambergs,
Phys. Rev. C 51 (1995) 3363

[243] T.L. Enik, L.V. Mitsyna, V.G. Nikolenko, A.B. Popov and G.S. Samosvat, Phys. Atom. Nucl.
60 (1997) 567

[244] F. Wissmann, M.I. Levchuk and M. Schumacher, Eur. J. Phys. A 1 (1998) 193

119
[245] G. Wormser, in: Lecture Notes in Physics 365, M. Schumacher , G. Tamas Eds., p. 100 Springer
(1990)

[246] H. Arenhövel, M. Weyrauch, P.G. Reinhard, Phys. Lett. 155B (1985) 22

[247] J. Vesper, N. Ohtsuka, L. Tiator, D. Drechsel, Phys. Lett. 159B (1985) 233;
J. Vesper, D. Drechsel, N. Ohtsuka, Nucl. Phys. A 466 (1987) 652

[248] F. Lenz, E. J. Moniz, Phys. Rev. C12 (1975) 909

[249] K. Klingenbeck, M.G. Huber, J. Nucl. Phys. G6 (1980) 961

[250] M.G. Huber, B. Ch. Metsch, H.R. Petry, in: Lecture Notes in Physics 365, M. Schumacher , G.
Tamas, Eds. Springer (1990)145

[251] E. Oset and L.L. Salcedo, Nucl. Phys. A 468 (1987) 631

[252] G. Garcı́a-Recio, E. Oset, L.L. Salcedo, D. Strottman and M.J. López, Nucl. Phys. A 526 (1987)
685

[253] T. Udagawa, S.-W. Hong and F. Osterfeld, Phys. Lett. B 245 (1990) 1

[254] R.H. Landau and A.W. Thomas, Nucl. Phys. A 302 (1978 ) 461

[255] L.D. Landau, Zh. Exp. Teor. Fiz. 35 (1958) 97

[256] C.W. deJager, H. deVries and C. deVries, Nucl. Data Tables 14 (1974) 479

[257] D. Pines and P. Noziéres, Theory of Quantum Liquids, Vol. I (W.A. Benjamin, New York and
Amsterdam, 1966)

[258] A.B. Migdal, Theory of Finite Fermi Systems and Applications to Atomic Nuclei (Interscience,
New York, 1957)

[259] G. Baym and S.A. Chin, Nucl. Phys. A 262 (1976) 527

[260] W. Bentz, A. Arima, H. Hyuga, K. Shimizu, K. Yazaki, Nucl. Phys. A 436 (1985) 593

[261] M.R. Anastasio and G.E. Brown, Nucl. Phys. A285 (1977) 516

[262] W.H. Dickhoff, A. Fässler, J. Meyer-ter-Vehn and H. Müther, Phys. Rev. C23 (1981) 1154

[263] W.H. Dickhoff, Nucl. Phys. A399 (1983) 287

[264] S.-O. Bäckman and G.E. Brown, Phys. Reports 124 (1985) 1

[265] J.I. Fujita and M. Hirata, Phys. Lett. B 37 (1971) 237

[266] J.I. Fujita, S. Ishida, M. Hirata, Supp. Prog. Theor. Phys. 60 (1976) 73

[267] J.I. Fujita and M. Ichimura, in: Mesons in nuclei, eds. M. Rho and D.H. Wilkinson (North-
Holland, Amsterdam, 1979) Chap. 15

[268] H. Hyuga, A. Arima, and K. Shimizu, Nucl. Phys. A 336 (1980) 363

[269] T. Yamazaki, in Mesons in Nuclei, edited by M. Rho and D.H. Wilkinson (North-Holland,
Amsterdam, 1979) Chap. 16

120
[270] A. Arima and H. Hyuga, in Mesons in Nuclei, edited by M. Rho and D.H. Wilkinson, (North-
Holland, Amsterdam, 1979) Chap. 17

[271] Y. Nambu and G. Jona-Lasinio, Phys. Rev. 122 (1961) 345

[272] B.D. Serot, J.D. Walecka, Adv. Nucl. Phys. 16 (1986) 1

[273] M. Rho, “Chiral Symmetry in Nuclear Physics” nucl-th/9812012

[274] M. Jacob and G.C. Wick, Ann. Phys. (N.Y.) 7 (1959) 404

[275] Y. Hara, Prog. Theor. Phys. Suppl. 51 (1972) 96

[276] A. Erdélyi, “Higher Transcendental Functions”, McGraw-Hill 1953

[277] F.A. Berends, A. Donnachie and D.L. Weaver, Nucl. Phys. B4 (1967) 1

[278] P. Stichel and E. Werner, Nucl. Phys. A145 (1970) 257

[279] J.S. O’Connell, in: “Photonuclear Reactions and Applications”, B.L. Berman Ed., Pacific Grove
1973

[280] E. Hayward, in [204] p. 275

[281] J.L. Friar and S. Fallieros, Phys. Rev. C 15 (1977) 365

[282] P. Christillin, Phys. Lett. B146 (1984) 375

[283] A. Fabrocici and S. Fantoni, Nucl. Phys. A435 (1985) 448

[284] B. Castel, I.S. Towner, “Modern Theories of Nuclear Moments”, Oxford Univ. Press 1990

[285] A.I. L’vov and A.I. Milstein, Phys. Lett. A192 (1994) 185

[286] R.P. Feynman and A.R. Hibbs, Quantum mechanics and path integrals, McGraw-Hill, New York,
1965

[287] J.S. Levinger, M.L. Rustgi and K. Okamoto, Phys. Rev. 106 (1957) 1191

121
 total absorption cross section σγp [µb]

600

500

400

300

200

100

0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
photon energy ω [GeV]

Figure 2.1: Photoabsorption cross section of the proton schematically separated into partial cross
sections [142] containing nucleon resonances and a nonresonant component.

3
10
total cross section σγA /A [mb]

102
∆
GDR
Born

1 QD
10 GQR

2 5 10 20 50 100 200
photon energy ω [MeV]

Figure 2.2: Schematic view of the nuclear photoabsorption cross section of 208 Pb from the giant reso-
nance to the ∆ resonance range. GDR: giant-dipole resonance, GQR: isovector giant-quadrupole res-
onance, QD: quasideuteron mode, ∆: Delta resonance of nucleons in the nucleus, Born: nonresonant
photoexcitation of nucleons in the nucleus through the Born terms of the photo-pion amplitudes [144].

122
 600
total cross section σγA /A [µb]

500

400

300

200

100

0
300 400 500 600 700 800
photon energy ω [MeV]

Figure 2.3: Total photoabsorption cross sections per nucleon. Data are shown for 1 H (black circles), 2 H
(gray circles), 3 He (open circles) and 4 He (grey squares) [146]. The solid line represents the universal
curve for complex nuclei.

0.2
total cross section σγA /A [mb]

p
d
Be
C
Al
0.1 Cu
Pb

0
10 3 10 4 10 5 10 6
photon energy ω [MeV]

Figure 2.4: Fits to the total nuclear photoabsorption cross sections per nucleon as an example of
shadowing in the asymptotic region [148].

123
 total absorption cross section σγp [mb]
800
700 (E1)1
600
500
400 ×10
300
200 (M1)1 (E2)1
100 (M1)0 (E2)0

0
10 20 30 40
photon energy ω [MeV]

Figure 2.5: Schematic representation of giant resonance multipoles for 208 Pb. (E1)1 : isovector giant
electric-dipole resonance (GDR), (M1)0 : isoscalar giant magnetic-dipole resonance, (M1)1 : isovector
giant magnetic-dipole resonance,(E2)0 : isoscalar giant electric-quadrupole resonance, (E2)1 : isovector
giant electric-quadrupole resonance [95].
total cross section σγA /A [µb]

400

200

0
50 100 150 200 250 300 350 400
photon energy ω [MeV]

Figure 2.6: Total nuclear photoabsorption cross sections per nucleon. Data are shown for nuclei
between 4 He and 238 U [148]. In addition the result of a model calculation for the universal photoab-
sorption curve is shown [157]. The left curve extending underneath the ∆ resonance is the QD cross
section. The ∆-resonance cross section is partitioned into two curves, into a non-mesonic part (left
resonant curve) and a mesonic part (right resonant curve). The nonresonant curve in the ∆ range
corresponds to the Born term of meson photoproduction.

124
 0.3
scattering amplitude [fm]
0.2
0.1
0
-0.1
-0.2
-0.3
20 40 60 80 100
photon energy [MeV]

0.015
scattering amplitude [fm]

0.01

0.005

-0.005

-0.01
20 40 60 80 100
photon energy [MeV]

Figure 4.1: Scattering amplitudes for the giant-dipole resonances (upper figure) and isovector giant-
quadrupole resonance (lower figure) of 208 Pb in the forward direction: Im RλL (ω, θ = 0) (dashed),
Re R̃λL (ω, θ = 0) (dotted) and Re R̂λL (ω, θ = 0) (solid).

125
 10

1
dσ/dΩ [mb/sr]
10-1

10-2

10-3

10-4
8 10 12 14 16 18 20
photon energy ω [MeV]

Figure 4.2: Differential cross sections for Compton scattering by 209 Bi. Scattering angle θ = 135◦ .
Data points at 9.0, 11.4 and 17.74 MeV are measured by [63]. Other data points are measured by [201].
Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the GDR
parameters of [201].

1000
total cross section σγA [mb]

500

200
100
50

20

10 15 20 25
photon energy ω [MeV]

Figure 4.3: Photoabsorption cross sections for 209 Bi measured by [202] multiplied by a scaling factor
of 1.35. Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the
GDR parameters of [201].

126
 1000

total cross section σγA [mb]

100

20

8 10 12 14 16 18
photon energy ω [MeV]

Figure 4.4: Photoabsorption cross sections for 209 Bi measured by [203] multiplied by a scaling factor
of 1.09. Solid curve: calculated using the GDR parameters of [63]. Dashed curve: calculated using the
GDR parameters of [201].

1.6
1.4
σ(θ=150°) / σ(θ=60°)

1.2
1
0.8
0.6
0.4
0.2

0 10 20 30 40 50
photon energy ω [MeV]

Figure 4.5: Cross-section ration σ(θ = 150◦ )/σ(θ = 60◦ ) for Compton scattering of unpolarized photons
by 208 Pb [25]. Solid line: Calculated including the IVGQR. Dashed line: Calculated not including the
IVGQR. The IVGQR shows up as interference of the E2 amplitude with the predominant E1 amplitude
from the GDR.

127
 concrete
lead
NaI(Tl) tagger
plastic
electron beam

C
metal foil

target

photon beam

1m

Figure 5.1: Typical experimental arrangement as used at the MAX laboratory of the university of Lund
(Sweden) [51,52]. The photon beam having an energy of about 100 MeV hits a thin metal foil serving as
a bremsstrahlung radiator. Quasi monochromatic photons are produced through a coincidence condition
between an event in one of the NaI(Tl) detectors (A, B, C) and an event in the tagger.

128
 90
total cross section σ [mb] 80
70
60
50
40
30
20
10
0
10 20 30 40 50 60 70 80
photon energy ω [MeV]

35

30
total cross section σ [mb]

25

20

15

10

0
10 20 30 40 50 60 70 80
photon energy ω [MeV]

Figure 5.2: Total photoabsorption cross sections [84] for 40 Ca (upper figure) and 16 O (lower figure)
partitioned into components having Lorentzian shapes. Dashed curves: GR components. Dotted curves:
QD components.

129
 11000

10000

9000
dσ/dΩ [nb/sr]

8000

7000

6000

5000

4000

3000
50 100 150
scattering angle θ [deg]

9000

8000

7000
dσ/dΩ [nb/sr]

6000

5000

4000

3000

2000

1000
50 100 150
scattering angle θ [deg]

Figure 5.3: Experimental elastic differential cross sections [42, 47, 48] for 40 Ca versus scattering
angle compared with predictions. Eγ = 58 MeV (upper Figure), Eγ = 75 MeV (lower Figure). The
curves are calculated for (i) the free-nucleon electromagnetic polarizabilities (dashed), (ii) the free-
nucleon electromagnetic polarizabilities supplemented by meson exchange corrections predicted for the
finite nucleus (solid), and (iii) the free-nucleon electromagnetic polarizabilities supplemented by meson
exchange corrections predicted for nuclear matter (dotted).

130
 2200

2000

1800

1600
dσ/dΩ [nb/sr]

1400

1200

1000

800

600
50 100 150
Scattering angle Θ [deg]

1600

1400
dσ/dΩ [nb/sr]

1200

1000

800

600

400
50 100 150
scattering angle θ [deg]

Figure 5.4: Differential cross sections [39, 47, 48] for Compton scattering by 16 O versus scattering
angle compared with predictions. Eγ = 58 MeV (upper figure); Eγ = 75 MeV (lower figure). The
curves are calculated for (i) the free-nucleon electromagnetic polarizabilities (dashed), (ii) the free-
nucleon electromagnetic polarizabilities supplemented by meson exchange corrections predicted for the
finite nucleus (solid), and (iii) the free-nucleon electromagnetic polarizabilities supplemented by meson
exchange corrections predicted for nuclear matter (dotted).

131
 2
1.5
1
0.5
0
200 400
0.3
dσ/dΩ (µb/sr)

0.2

0.1

0
150 200 250 300
0.15

0.1

0.05

0
150 200 250 300
Eγ (MeV)

Figure 6.1: Differential cross sections (dσ/dΩ)LAB for Compton scattering by 4 He [46] compared with
predictions. The full circles are calculated within the ∆-hole model including the seagull amplitudes
[46]. The results of the schematic model (dashed) and the extended schematic model (solid) are also
shown [46].

132
  
 



 KR,
 y  







 


y



y
t t
t


{KR  
y t










a) b) c)

Figure 6.2: One- and two-body contributions to nuclear Compton scattering. Diagram a is the impulse
approximation. Diagram b describes real pion production followed by pion absorption. Diagrams c are
required by the gauge invariance; in particular, the last diagram contains the contact γπN ∆ vertex.

1000
∆-hole
IA
IA+MEC
dσ/dΩ (nb/sr)

100

10
0 30 60 90 120 150 180
θCM (deg)

Figure 6.3: Differential cross sections (dσ/dΩ)CM for Compton scattering by 4 He at Eγ = 206 MeV.
Data are from [45] (solid circles) and from [46] (open circles). Shown are predictions of the ∆-
hole model in the local-density approximation [109], impulse approximation, Eq. (6.7), and the total
contribution including the two-body MEC, Eq. (6.13).

133
 0.4
protons
π 1h 9/2
δglmeson
0.3

0.2 δglobs

δglhigh
0.1

—0.1

—0.2
I II III IV

—0.3 ν 1i11/2

neutrons

Figure 7.1: Summary of results on the Fujita-Hirata relation. Column I: Experimental result of
◦
Yamazaki [269]. Column II: δgLmeson calculated from the experimental κGDR and δgLhigh obtained as
a difference between δlmeson of column II and δgLobs of column I. Column III: Predictions of Hyuga et
al. [268]. Column IV: Predictions of Brown and Rho [87].

= +

+ + + ...

ij
Figure 9.1: Typical diagrams contributing to T(π) . The wavy lines denote photons and dashed lines
ij
denote pions. The amputation indicates that T contains only the nucleon vertices, but not its wave
functions.

134
 1
40Ca
0.8
form factor F2

0.6
0.4
0.2
0
0 50 100 150 200 250
momentum transfer ∆q [MeV]

Figure 9.2: Comparison of the form factors from Eqs. (9.8) (dotted curve) and (9.9) (full curve) for
the case of 40 Ca. The approximation F12 (q/2) is shown as a dashed curve.

1
1

0.8
ratio χ ( A) = r 2 2 / r 2

0.6

0.4

0.2

0
10 20 30 40 50 60
mass number A

Figure 9.3: Ratio χ(A) = r 2 2/ r2 1 as a function of the nuclear mass number A

135
 N ( pr , λ ) P( pr , λ )
2 2 1 1 N ( pr , λ ) P( pr , λ )
2 2 1 1
r r
Q Q

−Q
r
( p1 − Q − p2 )
r r r

P( pr , λ ) N ( pr , λ )
1 1 2 2 N ( pr , λ ) P( pr , λ )
2 2 1 1

N ( pr , λ ) P( pr , λ )
2 2 1 1 N ( pr , λ ) P( pr , λ )
2 2 1 1

−Q
r
( p1 − Q − p2 )
r r r

r r
Q Q

P( pr , λ ) N ( pr , λ )
1 1 2 2 N ( pr , λ ) P( pr , λ )
2 2 1 1

Figure 9.4: Set of diagrams, from which the two-body correction to the correlator is extracted. Here
the same symbolic abbreviation is used as in Fig. 9.1. The nucleon spin projections are denoted by λi .

P( pr , λ )
1 1 N ( pr , λ )
2 2 N ( pr , λ )
3 3 P( pr , λ )
1 1 N ( pr , λ )
2 2 N ( pr , λ )
3 3
r
Q
r
Q

( p1 + Q − p2 )
r r r + r
Q
+

N ( pr , λ )
3 3 P( pr , λ ) 1 1
N ( pr , λ )
2 2 N ( pr , λ )
3 3 N ( pr , λ )
2 2
P( pr , λ )
1 1

N ( pr , λ ) 3 3
P( pr , λ ) 1 1 N ( pr , λ )
2 2

( p3 − Q − p1 )
r r r

r
Q
+ cr

N ( pr , λ ) 2 2 N ( pr , λ ) 3 3
P( pr , λ ) 1 1

Figure 9.5: Three-body diagrams yielding an additional correction to the correlator, which is of the
same order in f /mπ as the two-body diagrams shown in Fig. 9.4. The notations are the same as in
Fig. 9.1 and Fig. 9.4.

136
 Q/pF
0 1 2 3 4
C

( A / 2)
0

-0.2

-0.4

-0.6

-0.8

-1

Figure 9.6: Central part of the correlator with three-body corrections (full curve), without three-body
corrections (dotted curve) and Fermi correlator without any corrections (dashed curve).

T
( A / 2)
0.3

0.2

0.1

-0.1

1 2 3 4
Q/pF

Figure 9.7: Tensor part of the correlator with three-body corrections (full curve) and without three-body
corrections (dashed curve).

137
 T

( A / 2)
0.3

0.2

0.1

-0.1

1 2 3 4
Q/pF

Figure 9.8: Comparison of the tensor part of the correlator with (full curve) and without (dashed
curve) the ρ-meson contribution.

1.1
enhancement constant κ

0.9

0.8

0.7

20 40 60 80 100
Z

Figure 9.9: Dependence of enhancement constant κ on proton number Z. The dashed curve corresponds
to the pionic tensor contribution κπT , the dash-dotted curve includes also the central contribution κπC
and the dotted curve gives the total κ, including the contribution from ρ-meson exchange. The realistic-
density result κ(rd) for the full enhancement constant (cf. Eq. (9.30)) is shown as a full curve.

138
 -1
δα [10—4 fm3]

-1.5

-2

-2.5

20 40 60 80 100
Z

Figure 9.10: Pion-exchange contribution to electric polarizability δα as a function of Z. The dashed

curve corresponds to the central contribution δαC and the dotted curve gives the total δα = δαC + δαT .
The use of a realistic density leads to the full curve.

2
1.75
δβ [10—4 fm3]

1.5
1.25
1
0.75
0.5
20 40 60 80 100
Z
Figure 9.11: Pion-exchange contribution to magnetic polarizability δβ as a function of Z. The dashed
curve corresponds to the central contribution δβC and the dash-dotted curve gives the sum δβC + δβT .
Adding the contribution of the ∆-isobar excitation as given in Eq. (9.29) leads to the total value of δβ
given as the dotted curve. The use of a realistic density leads to the full curve.

139
 1 40Ca

0.8
form factor F2

0.6
0.4
0.2
0
50 100 150 200 250
momentum transfer ∆q [MeV]

(i) (1) (2)

Figure 9.12: Form factors F2 (∆) for 40 Ca. The dashed curve is F2 and the full curve is F2 . For
comparison the (experimental) charge form factor F1 is also shown (dash-dotted curve), as well as the
function F12 (∆/2) (dotted curve).

k2 k1 k2 k1
k1

φ0
ψ n+α
ψ n−α
φ0
φ0 φ0 φ0 φ0
k2

Figure 10.1: Diagrams of Compton scattering off the relativistic oscillator (crossed terms are not
shown)

140
 Re R1,1 at θ = 0o Re R1,1 at θ = 45o Re R1,1 at θ = 90o
0.5 0.5 0.5

0 0 0

-0.5 -0.5 -0.5

0 50 100 0 50 100 0 50 100
ω (MeV) ω (MeV) ω (MeV)

Re R1,−1 at θ = 180o Re R1,−1 at θ = 135o Re R1,−1 at θ = 90o

0.5 0.5 0.5

0 0 0

-0.5 -0.5 -0.5

0 50 100 0 50 100 0 50 100
ω (MeV) ω (MeV) ω (MeV)

Figure 10.2: Phenomenological, Eq. (10.73), (dashed lines) and exact (solid lines) amplitudes RGR in
the relativistic oscillator model at ω0 = (13.4 − 2.1i) MeV and η = 0.06. Units are e2 /E0 .

141
 1 208
Pb
0.8
form factor F2

0.6
0.4
0.2
0
-0.2
50 100 150 200 250
momentum transfer ∆q [MeV]

(i) (1)
Figure B.1: Exchange form factors F2 (q) for the case of 208 Pb. The dashed curve represents F2 ,
(2)
while the full curve corresponds to F2 . For comparison the (experimental) charge form factor F1
(dash-dotted curve) and the function F12 (q/2) (dotted curve) are also shown. The parameters for the
form factor F1 were taken from ref. [256].

1 12
C
form factor F2

0.8

0.6

0.4

0.2
50 100 150 200 250
momentum transfer ∆q [MeV]

Figure B.2: Same as Fig. B.1, but for 12 C.

142
 1 40
Ca
0.8
form factor F2

0.6 GR
0.4
0.2 QD
0
50 100 150 200 250
momentum transfer ∆
q [MeV]

(GR) (QD) (QD)

Figure B.3: Form factors F2 (∆) and F2 (∆)for 40 Ca. The dashed curve is F2 and the full
(GR)
curve is F2 . In addition, the (experimental) charge form factor F1 (dash-dotted curve) and the
2
function F1 (∆/2) (dotted curve) are shown for the sake of comparison.

1 12C
form factor F2

0.8 GR

0.6
QD
0.4

0.2
50 100 150 200 250
momentum transfer ∆q [MeV]

Figure B.4: Same as Fig. B.3, but for 12 C.

143