U N I V E R SITÉ LIBRE DE BRUXELLES
Relativistic study of electron correlation effects
on polarizabilities, two-photon decay rates, and
electronic isotope-shift factors in atoms and ions
Ab initio and semi-empirical approaches
Th`ese pr´esent´ee en vue de l’obtention du titre de
Docteur en Sciences de l’Ing´enieur et Technologie
Livio Filippin
Promoteur
Prof. Michel Godefroid
Co-Promoteur
Dr. J¨orgen Ekman (Malm¨o H¨ogskola)
Service
Chimie quantique et Photophysique
Ann´ee acad´emique
Relativistic study of electron correlation effects
on polarizabilities, two-photon decay rates, and
electronic isotope-shift factors in atoms and ions
Ab initio and semi-empirical approaches
Th`ese pr´esent´ee en vue de l’obtention du titre de
Docteur en Sciences de l’Ing´enieur et Technologie
Livio Filippin
Promoteur
Prof. Michel Godefroid
Co-Promoteur
Dr. J¨orgen Ekman (Malm¨o H¨ogskola)
Service
Chimie quantique et Photophysique
Acknowledgments
Aucun travail ne peut être réalisé sans le bon-vouloir, l’énergie et la collaboration de
plusieurs personnes. A l’heure où je rédige ces lignes concluant un chapitre de quatre
années, je voudrais remercier toutes les personnes ayant contribué à ce projet.
J’adresserai tout d’abord mes remerciements à Michel Godefroid, mon promoteur de
thèse mais également de mémoire, pour son enseignement, son engouement et son
enthou-siasme tout au long de ces années de recherche. Dès mon arrivée en tant que mémorant,
il a su m’accueillir parfaitement au sein du service de Chimie quantique et Photophysique
dont il est co-directeur. Je voudrais le remercier pour son soutien et son aide, sa présence
et ses conseils, mais surtout pour la confiance qu’il m’a accordée dans mon travail au
quo-tidien. En effet, en tant que doctorant je me suis senti libre de mener à bien les projets qui
m’intéressaient, et ce de manière autonome mais toujours sous sa bienveillance. En plus
de féliciter l’excellent promoteur et professeur, je voudrais aussi remercier la personne aux
grandes qualités humaines. Toujours optimiste et faisant preuve d’attention, il est
con-sidéré à juste titre comme le “papa gâteau” du service CQP. En outre, je me souviendrai
longtemps de nos soirées de conférence, qu’elle se soient passées autour d’un prosecco
à Trento, d’un whisky à Lund, ou d’une caipirinha à São Paulo. De tout cœur merci
pour ces moments de complicité partagés hors du cadre de la recherche. Je trouve que
nous avons formé une superbe équipe, de loin la plus petite unité du service puisqu’elle
n’a compté que deux membres.
Je lui souhaite d’ailleurs quatre années de recherche
fructueuse avec Sacha Schiffmann, qui sera vraisemblablement le dernier doctorant de
son impressionnante carrière académique.
J’en viens à présent à la deuxième personne qui a largement contribué à ce travail. Je
remercie Daniel Baye, co-promoteur durant mon mémoire mais également collaborateur et
conseiller très précieux durant ma thèse, pour son expertise inestimable, sa disponibilité
ainsi que la grande attention qu’il a portée à mes différents projets.
Il m’a souvent
indiqué la voie à suivre, a corrigé mes défauts, et m’a appris à réfléchir en projetant mon
travail au-delà de mes ambitions premières. Ses nombreuses idées et inspirations lors de
discussions scientifiques avec Michel m’ont grandement éclairé dans la compréhension de
la méthode des réseaux de Lagrange, et ont augmenté mon envie d’aller plus loin. Ce fut
également un plaisir d’étendre à ses côtés cette méthode à la fois simple et élégante au
calcul relativiste de propriétés atomiques. Cette collaboration s’est révélée très fructueuse,
ayant abouti à la rédaction de quatre articles.
I would like to warmly thank Per Jönsson and Jörgen Ekman for welcoming me during
two months in their offices at Malmö Högskola. You both made my stay in Sweden very
enjoyable and memorable thanks to your kind attentions to me. I will cherish our baths
in the cold water of the Öresund sound, but also those during the CompAS meeting
in summer, much more convenient for a rookie like me. I will also remember my last
dinner at Monster restaurant, eating those massive triple-layer burgers, as well as our
relaxing times at the beautiful skybar of Malmhattan. Turning to science, your expertise
in computational atomic physics helped me a lot to get my hand in the grasp
2
k code,
and to understand subtleties in electron correlation models. Our collaboration was very
flourishing, leading to the writing of four papers.
Acknowledgments
Je termine mon doctorat dans le bureau où j’ai commencé mon mémoire. En cinq
ans, j’ai eu l’occasion de connaître trois personnes hautes en couleur, qui sont tour à
tour devenues des amis, à savoir mes collègues de bureau Ludo, Milaim et Guillaume.
Je voudrais les remercier pour les excellents moments que nous avons passés ensemble.
Ludo finissant sa thèse dans les mêmes temps que moi, je souhaite à Milaim ainsi qu’à
Guillaume une excellente dernière année de travail, et un bon courage pour la rédaction.
Je voudrais également remercier tous les membres du CQP pour l’ambiance unique et
excellente qui y règne. Celle-ci n’est certainement pas étrangère au bon déroulement de
ma thèse et au plaisir que j’ai eu à la réaliser. Je souhaite bon courage aux doctorants
qui finiront leur thèse d’ici un an.
Un grand merci à mes amis doctorants extérieurs au CQP, à savoir Michael, Robin
et Piotr, pour les agréables moments de détente sur le campus mais également hors de la
vie universitaire.
Je tiens à présent à remercier mes amis et ma famille, en particulier ma mère, pour
m’avoir toujours soutenu et entouré, pour m’avoir toujours poussé à réaliser mes projets.
Son amour indéfectible m’a sans conteste porté là où je suis aujourd’hui, et je tiens à
exprimer la chance immense de la compter dans ma vie. Elle m’apporte cette sérénité au
quotidien, cette insouciance qui m’a permis de me concentrer sur ce projet de thèse sans
devoir faire face aux tracas et aux imprévus de la vie. Mais c’est promis, je prendrai mon
envol très prochainement.
Abstract
List of publications
This thesis includes the results presented in the following peer-reviewed papers. The
latter are classified into two main groups, labeled A and B.
Group A: Relativistic calculations of polarizabilities and two-photon decay rates in
hydrogenic and alkali-like ions.
A
IAccurate solution of the Dirac equation on Lagrange meshes
D. Baye, L. Filippin, and M. Godefroid,
Phys. Rev. E
89, 043305 (2014).
A
IIRelativistic polarizabilities with the Lagrange-mesh method
L. Filippin, M. Godefroid, and D. Baye,
Phys. Rev. A
90, 052520 (2014).
A
IIIRelativistic two-photon decay rates with the Lagrange-mesh method
L. Filippin, M. Godefroid, and D. Baye,
Phys. Rev. A
93, 012517 (2016).
A
IVRelativistic semiempirical-core-potential calculations in Ca
+, Sr
+, and
Ba
+ions on Lagrange meshes
L. Filippin, S. Schiffmann, J. Dohet-Eraly, D. Baye, and M. Godefroid,
Submitted to Phys. Rev. A; arXiv:1709.07672v2 (2017).
Group B: Relativistic multiconfiguration calculations of electronic isotope-shift factors
in many-electron atoms.
B
ICore correlation effects in multiconfiguration calculations of isotope shifts
in Mg i
L. Filippin, M. Godefroid, J. Ekman, and P. Jönsson,
Phys. Rev. A
93, 062512 (2016).
B
IIMulticonfiguration calculations of electronic isotope shift factors in Al i
L. Filippin, R. Beerwerth, J. Ekman, S. Fritzsche, M. Godefroid, and P. Jönsson,
Phys. Rev. A
94, 062508 (2016).
B
IIIMulticonfiguration calculations of electronic isotope-shift factors in Zn i
L. Filippin, J. Bieroń, G. Gaigalas, M. Godefroid, and P. Jönsson,
Phys. Rev. A
96, 042502 (2017).
This thesis does not include the results presented in the following peer-reviewed papers.
The latter are classified in a third group, labeled C.
C
IEvolution of nuclear structure in neutron-rich odd-Zn isotopes and
iso-mers
C. Wraith et al.,
Phys. Lett. B
771, 385 (2017).
C
IIAb initio
calculations of hyperfine structures of zinc and evaluation of
the nuclear quadrupole moment Q(
67Zn)
List of acronyms
AS
active space . . . .
28
ASF
atomic state function . . . .
26
CAS
complete active space . . . .
28
CC
core-core . . . .
28
CI
configuration interaction . . . .
5
CompAS
the Computational Atomic Structure group . . . .
24
CP
core polarization . . . .
5
CSCO
common set of commuting observables . . . .
12
CSF
configuration state function . . . .
21
CV
core-valence . . . .
28
D
double (substitution) . . . .
28
DC
Dirac-Coulomb . . . .
8
DCB
Dirac-Coulomb-Breit . . . .
25
DHF
Dirac-Hartree-Fock . . . .
8
DHFCP
Dirac-Hartree-Fock plus core polarization . . . .
6
DKB
dual kinetic balance . . . .
37
E1
electric dipole . . . .
3
E2
electric quadrupole . . . .
6
FC
frozen core . . . .
29
FS
field shift . . . .
3
grasp2k
the General-purpose Relativistic Atomic Structure Package . . . .
7
HF
Hartree-Fock . . . .
19
IS
isotope shift . . . .
3
ITO
irreductible tensor operator . . . .
39
KB
kinetic balance . . . .
36
LMM
Lagrange-mesh method . . . .
5
LW
Long wavelength . . . .
51
M 1
magnetic dipole . . . .
6
MBPT
many-body perturbation theory . . . .
5
MCDHF
multiconfiguration Dirac-Hartree-Fock . . . .
8
MR
multireference . . . .
27
MRSD
multireference single and double (process) . . . .
28
MS
mass shift . . . .
3
NMS
normal mass shift . . . .
7
PCF
pair-correlation function . . . .
28
QED
quantum electrodynamics . . . .
16
ratip
the Relativistic calculations of Atomic Transition, Ionization,
and recombination Properties package . . . .
8
ris3
the Relativistic Isotope Shift program . . . .
7
S
single (substitution) . . . .
28
SCF
self-consistent field . . . .
24
SE
self energy . . . .
25
SMS
specific mass shift . . . .
7
TP
transverse photon . . . .
25
VP
vacuum polarization . . . .
25
Contents
I
Theory and methodology
1
1 Background and motivation
3
1.1
Polarizabilities . . . .
3
1.2
Two-photon decay rates
. . . .
6
1.3
Electronic isotope-shift factors . . . .
7
1.4
Thesis outline . . . .
8
2 Dirac equation
9
2.1
Relativistic context . . . .
9
2.2
Hydrogenic ions . . . .
12
2.2.1
Radial Dirac equations . . . .
15
3 Dirac-Hartree-Fock method
19
3.1
Relativistic context . . . .
19
3.2
DHF equations . . . .
20
3.3
grasp2k package . . . .
24
4 Beyond the DHF method
25
4.1
Corrections to the DC Hamiltonian . . . .
25
4.2
Electron correlation . . . .
26
4.2.1
MCDHF method . . . .
26
4.2.2
Semi-empirical approach . . . .
29
5 Lagrange-mesh method
31
5.1
Principle . . . .
31
5.2
Lagrange functions . . . .
31
5.2.1
Gauss-quadrature rule . . . .
31
5.2.2
Lagrange conditions
. . . .
32
5.3
Meshes based on orthogonal polynomials . . . .
33
5.4
Regularized Lagrange functions . . . .
34
5.4.1
Regularized Laguerre mesh . . . .
35
6 Static dipole polarizabilities
39
6.1
Definition . . . .
39
6.2
Hydrogenic ions . . . .
39
Table of contents
6.2.2
Tensor polarizability . . . .
44
6.3
Alkali-like ions . . . .
44
7 Radiative multipole transitions
45
7.1
Interaction Hamiltonian
. . . .
45
7.2
Multipole expansion of the radiation field . . . .
46
7.2.1
Gauge transformation
. . . .
47
7.2.2
One-electron transition amplitudes . . . .
47
7.3
One- and two-photon decay rates . . . .
48
7.3.1
Hydrogenic ions . . . .
48
7.3.2
Alkali-like ions
. . . .
51
8 Electronic isotope-shift factors
53
8.1
Isotope shift theory . . . .
53
8.2
Relativistic recoil Hamiltonian . . . .
55
8.2.1
Normal mass-shift factors
. . . .
55
8.2.2
Specific mass-shift factors
. . . .
55
8.3
Relativistic field-shift factors . . . .
56
8.4
Direct diagonalization of the Hamiltonian matrix
. . . .
58
9 Conclusions and perspectives
59
Bibliography
61
II
Papers
71
Paper A
I73
Paper A
II85
Paper A
III99
Paper A
IV115
Paper B
I133
Paper B
II143
Paper B
III155
List of Figures
2.2.1
Dirac energy spectrum for (a) a hydrogenic system, (b) a free electron, (c) an
anti-hydrogenic system.
. . . .
17
2.2.2
Radial wave functions
P
1s1/2(r) and Q
1s1/2(r) of Au
78+
(Z = 79).
. . . .
18
3.3.1
Schematic SCF algorithm solving the DHF equations.
. . . .
24
5.4.1
Basis of 4 Lagrange-Laguerre functions ˆ
f
i(1)(x). The 4 mesh points x
iverify
List of Tables
Part I
1
B
ACKGROUND AND MOTIVATION
This thesis is set within the framework of relativistic quantum mechanics, born from the
union of Einstein’s theory of special relativity and the formalism of quantum mechanics.
The first aim of this thesis is to perform relativistic calculation of polarizabilities and
two-photon decay rates in hydrogenic and alkali-like systems. Polarizabilities are defined
as the response of a system to an external electric field [1]. When immersed in the latter,
an atom or ion undercomes a distorsion of its charge distribution, and behaves as an
electric multipole [2]. Polarizabilities appear in a large number of fields and applications,
namely in cold atoms physics, metrology and chemical physics [3].
Two-photon transitions are part of a priori highly unlikely processes and are therefore
called forbidden radiative processes. The latter occur when angular and parity selections
rules forbid the spontaneous emission of one photon via an allowed electric dipole (E1)
transition from an excited state to a lower one in the energy spectrum. Long lifetimes
from metastable excited states in highly charged Be-like ions were measured in recent
storage-ring experiments, but their interpretation is problematic [4]. Hence, the study of
forbidden radiative processes is all obviously relevant [5, 6].
Another purpose of this thesis is to study the effect of the nuclear mass and charge
distribution on the electronic structure of an atom. Energies of atomic transitions are
slightly shifted from one isotope to another, due to the different masses and charge
distributions of the atomic nuclei [7]. These differences respectively lead to the so-called
mass shift (MS) and field shift (FS). Summing both contributions, the total shift is
referred to as the isotope shift (IS) of a given atomic transition [8]. The second aim of
this thesis is to perform relativistic ab initio calculations of electronic IS factors for a
set of transitions between low-lying levels of neutral systems. These electronic quantities
together with observed ISs between different pairs of isotopes provide the changes in
mean-square charge radii of the atomic nuclei [9].
1.1
Polarizabilities
There has been a renewal of interest in atomic and ionic polarizabilities in the last
decade [2]. Their relevance to various applications, e.g. in cold atoms physics and in
metrology, is discussed in the review of Mitroy et al. [3]. Among the large choice, let us
cite the following fields of research:
1.1 Polarizabilities
• measurement of refractive indices for the determination of the Boltzmann constant
k
Bwith a higher accuracy, which is useful for thermometry and macroscopic
stan-dards: the present definition of temperature is based on the triple point of water at
273.16 K. At present, the best estimate of
k
Bis obtained by measuring the speed
of sound in helium (He) gas [10, 11]. Using an accurate theoretical value of the
4He
polarizability [12] and an accurate measurement of the refractive index for helium
gas [13] yields a value for the universal gas constant,
R, which is proportional to
k
Bthrough the identity
R = N
Ak
B, where
N
Astands for the Avogadro constant,
• estimation of parity nonconservation (PNC) amplitudes in heavy atoms: search for
new physics beyond the standard model of the electroweak interaction. In the
exper-iment [14] on the cesium (Cs)
7s
→ 6s transition, the PNC amplitude was measured
relative to the Stark-induced scalar and tensor transition polarizabilities [15],
• calculation of atomic clock frequency shifts for the research of frequency standards:
redefinition of the SI second, currently based on the microwave transition between
the two hyperfine levels of the ground state of
133Cs [16]. The transition towards
optical atomic clocks for achieving better relative frequency uncertainties and
sta-bility requires the accurate determination of the dominant systematic shifts, such as
Zeeman, Stark, quadrupole moment and blackbody radiation (BBR) shifts [3, 17],
• determination of magic wavelengths for optical lattices of cold neutral atoms: the
laser field used to create the lattice can be tuned to a magic wavelength [18, 19],
where lattice potentials of equal depth are produced for the two electronic states of
the clock transition. Hence, the quadratic Stark shift induced by the laser, which
is proportional to the frequency-dependent polarizability, is equal for both states of
the clock transition, which cancels the differential shift between the two of them,
• research of decoherence minimization for quantum logical operations: quantum
computation schemes may involve qubits that are realized as internal states of
neutral atoms trapped in optical lattices or microtraps, such as the Rydberg gate
scheme [20] in which the qubit is based on two ground hyperfine states. A
two-qubit phase gate may be realized by exciting two atoms to Rydberg states. The
vibrational state of the atom may change in the latter state, leading to decoherence
due to motional heating. The aim is to minimize the heating by matching the
frequency-dependent polarizabilities of the atomic ground and Rydberg states [21],
• evaluation of radiative line strengths for the determination of abundancy of elements
in astrophysics: line strengths can be extracted from precisely known
polarizabili-ties that are dominated by a single strong transition, as it is the case for Cs where
the dipole polarizability of the ground state
6s is dominated by the 6s
→ 6p
jtransi-tions [22, 23]. Hence, one can extract the line strengths related to these transitransi-tions.
In the particular case of optical atomic clocks, which are based on transitions involving
long-lived metastable states, experiments in this field have reached such a high accuracy
that relativistic effects are visible and must be precisely accounted for in the
calcula-tions [3, 17]. Today’s most advanced atomic clocks report relative systematic frequency
uncertainties below
10
−17[24, 25]. Reaching higher accuracy is limited by small energy
shifts resulting from BBR and quadratic Stark effect [26–28], highly dependent on the
accuracy of static and dynamic polarizabilities [3].
1.1 Polarizabilities
• Direct method: the application of perturbation theory to the study of an atom or
ion immersed in a uniform electric field involves a response of the system under the
form of a first-order correction to the wave function, solution of an inhomogeneous
equation, weighted by the amplitude of the electric field. The zero- and first-order
wave functions enable a direct computation of the polarizability,
• Sum-over-states method: the second-order correction to the energy is expressed
as a infinite sum of contributions coupling, via an electric multipole operator, the
state whose polarizability is targeted and the stationary states of the zero-order
Hamiltonian. The polarizability, proportional to this correction, is computed by
explicitely summing these contributions, continuous spectrum (spectra) included.
The direct method consists either in solving the appropriate wave equation for different
values of the electric field strength and in extracting the polarizability from the perturbed
energies (finite-field approach) [29], or in adopting the variation-perturbation method by
using the method developed by Dalgarno and Lewis [30,31] for the research of the solution
to the inhomogeneous equation [32].
The sum-over-states method is essentially used for one- and two-valence-electron
sys-tems, for the reason that the polarizability is then dominated by some transitions towards
reasonably excited states. It is thus possible to construct an effectively complete set of
intermediate states, but this method is less suited for atoms with more valence electrons.
Within these two computational means, approaches are distinguished with respect to
the methodology adopted to describe electron correlation and relativistic effects: methods
in bases of explicitely correlated functions [33, 34], multiconfiguration methods combined
to configuration interaction (CI) [35], semiempirical-core-potential approaches [36],
many-body perturbation theory (MBPT) [37], coupled-cluster (CC) method [38], and hybrid
methods combining CI and MBPT [39] or linearized CC [40].
The studied systems are mostly one- and two-valence-electron systems, and the
com-putation of the polarizability is mostly realized with the sum-over-states method. The use
of the finite-field approach is limited to CC computations in Gaussian bases using
molecu-lar codes [29], while the Dalgarno and Lewis method is used in approaches that are better
adapted to the description of systems with more than two valence electrons [32, 41, 42].
With this respect, the calculation of this property with multiconfiguration methods is
currently limited to the nonrelativistic framework [35], using the civ
3
/civpol code [43].
Within this methodological and computational context, this thesis presents results
from relativistic calculations of static multipole polarizabilities in two types of atomic
and ionic systems. In a first step, calculations are performed in hydrogenic ions with the
sum-over-states method, by means of the Lagrange-mesh method (LMM) [44–46]. The
LMM is an approximate variational method taking the form of mesh equations thanks to
the use of a Gauss-quadrature approximation. The Dirac equation for the particular case
of hydrogenic ions is analytically solvable [47, 48], and exact expressions of the
ground-state static scalar dipole polarizabilities were derived in Refs. [49–52]. Besides, accurate
relativistic calculations were reported in Ref. [53] using
B-spline bases [54]. Hence, the
comparison with results from these benchmark references allows to assess the accuracy
of the present numerical method (see Papers A
Iand A
II).
In a second step, this thesis presents results from relativistic
semiempirical-core-potential calculations of static multipole polarizabilities in the alkali-like singly ionized
calcium (Ca
+), strontium (Sr
+), and barium (Ba
+). The latter have been proposed as
candidates for optical frequency standards due to the long lifetimes of their metastable
2D
1.2 Two-photon decay rates
to simulate the core-valence correlations for an atom with few valence electrons. The
latter approach is referred to as the Dirac-Hartree-Fock plus core-polarization (DHFCP)
approach, offering simplified expressions and reduced computational times in comparison
with ab initio methods [3, 55, 56]. The underlying methodology is originally discussed in
Refs. [57–61] within the nonrelativistic framework. Later on, nonrelativistic results of
po-larizabilities were published for alkali atoms [62, 63] and alkali-like ions [55, 56]. It is only
very recently that relativistic semiempirical-core-potential calculations were performed in
such systems [64–69]. It should be mentioned that the DHFCP approach is an
approxi-mate method, where the comparison with relativistic ab initio methods [26–28, 38, 70–75]
and experiments [76–78] should provide an assessment of the errors on polarizabilities
due to physical effects that are not included in the model (see Paper A
IV).
1.2
Two-photon decay rates
Two-photon transitions are part of the highly unlikely processes, such that they are
clas-sified in a large family containing all the forbidden radiative processes. In this category,
let us mention e.g. the one-photon magnetic dipole (M 1) and electric quadrupole (E2)
transitions, as well as the two-photon electric dipole (2E1) transition. The radiative
de-cay of the metastable
2s
1/2state to the
1s
1/2ground state of hydrogenic ions is one of
the most widely studied forbidden atomic transitions. The
2s
1/2state can decay via two
competing processes: either an
M 1 or a 2E1 transition. The earliest theoretical works on
two-photon processes in hydrogen were performed by Göppert-Mayer [79], and Breit and
Teller [80]. Both (nonrelativistic) studies concluded that the dominant two-photon
2E1
transition is the principal cause of the radiative decay of interstellar
2s hydrogen atoms.
Long lifetimes are attractive in the context of ultraprecise optical frequency standards,
and thus share with atomic and ionic polarizabilities a common field of applications.
Life-times of the
2s2p
3P
o0
state of highly charged Be-like ions (
33S
12+and
47Ti
18+) were
mea-sured in recent storage-ring experiments, but problems occur in their interpretation [4].
The study of the competition between forbidden radiative processes (one-photon beyond
E1, or multi-photon) and unexpected (induced by hyperfine coupling or by external fields)
is all obviously relevant [5, 6]. Very recently, two-photon transitions have been
theoreti-cally studied in other highly charged Be-like ions, in order to compute the
E1M 1 decay
rate of the
2s2p
3P
o0
→ 2s
2 1S
0transition [81, 82].
The ab initio calculation of two-photon radiative decay rates is very rare, although
their mathematical expressions strongly recall the expressions of the polarizability by
the presence of the infinite sum over the intermediate states from the sum-over-states
method [83]. For hydrogenic systems, the decay rates were computed by using
B-spline
bases [84] and more recently
B-polynomial bases [85]. Besides, an interesting physics
appears in the prediction of resonances and extinctions [86].
For many-electron systems, calculations are quasi inexistant [4, 5]. Apart from the
above-cited works on Be-like ions, some studies were carried out on angular distributions
correlation and polarizations of the two photons emitted in heavy He-like systems [87].
Besides, two-photon decay rates have been calculated in Ca
+, Sr
+, and Ba
+ions using the
ab initio all-order CC method [88,89]. Finally, Ref. [90] studied two-photon transitions in
neutral Ba induced by hyperfine coupling and by external magnetic fields. Hence, it is seen
that the evaluation of two-photon transition probabilities remains embryonic, although
there is a real appeal, mostly for highly ionized systems, for estimating the probability
of these decay modes relative to hyperfine- or magnetic-field-induced channels [4].
1.3 Electronic isotope-shift factors
the same theoretical models and numerical methods. The results of the
2s
1/2→ 1s
1/2and
2p
1/2→ 1s
1/2two-photon decay rates in hydrogenic ions obtained with the LMM
are compared with several benchmark theoretical values [84–86, 91, 92] (see Paper A
III),
while those of the
nd
3/2,5/2→ (n + 1)s
1/22E1 decay rates in Ca
+(n = 3), Sr
+(n = 4),
and Ba
+(n = 5) ions obtained with the DHFCP-LMM approach are compared with ab
initio values from Refs. [88, 89] (see Paper A
IV).
1.3
Electronic isotope-shift factors
Differences in mean-square charge radii of nuclear charge distributions along isotope
chains are important for testing nuclear models and interactions [8, 93]. The nuclear
size can be extracted from laser-spectroscopy IS measurements, assuming that the MS
and the electronic densities inside the nucleus can be computed [7,8]. The recent progress
in atomic structure theory enables to perform large-scale calculations for open-shell atoms
and ions, providing the necessary atomic input to analyze laser-spectroscopy data. These
studies have significantly contributed to the understanding of the evolution of nuclear
sizes as a function of neutron and proton numbers along the nuclide chart [94]. However,
the demand from the nuclear physics community is increasing, and further theoretical
atomic investigations are needed for more complex atoms.
For light elements, where the electron density can be considered as constant inside
the nucleus, the transition FS is proportional to differences in mean-square charge radii
between isotopes [95]. In heavier systems, the electron density is no longer constant
inside the nucleus, and the transition FS becomes sensitive to higher-order nuclear radial
moments, or Seltzer moments [96]. The contribution of the latter can be subtracted
from the FS using a pre-determined model of the nuclear charge distribution, in order
to reach better estimates of the mean-square charge radii differences [97]. The nuclear
deformation effects on the transition energies are usually neglected, although theoretical
investigations have recently been realized in heavy ions [98–101].
To determine nuclear properties from IS measurements, the MS must be extracted
from the total IS. The MS of a given atomic transition depends on the nuclear masses and
on the difference of electronic MS factors of the upper and lower levels of this transition.
These MS factors are expressed as expectation values of the so-called nuclear recoil
oper-ator, and are calculated using approximate solutions of the many-electron wave equation.
Since the derivation of relativistic corrections to the recoil operator by Shabaev [102, 103]
and Palmer [104], several papers [99, 105–107] have shown the importance of relativistic
effects on ISs. The tensorial expression of the relativistic recoil operator [108] has been
implemented in the ris
3
(Relativistic Isotope Shift) module [109, 110] of the grasp
2
k
package [111, 112] (General-purpose Relativistic Atomic Structure Package).
1.4 Thesis outline
complex systems. All-order CC methods for relativistic atomic structure calculations have
been applied to predict properties of heavy atoms, from one- to three-valence-electron
systems, but the SMS factor turns out to be the most difficult to accurately evaluate [117].
The combined CI-MBPT approach is attractive for few-valence-electron systems, and the
SMS factor is evaluated with the use of the finite-field technique [118, 119], although it is
currently not part of the CI-MBPT package [120]. With this respect, variational methods
based on relativistic multiconfiguration/CI expansions are more flexible, and have been
applied to few-electron systems to yield highly accurate results [121,122]. MS and FS have
also been estimated in complex neutral systems such as barium [107], copper [123, 124],
manganese [125], polonium [126], and osmium [127].
Within this context, this thesis presents results from relativistic ab initio calculations
of electronic IS factors by using the multiconfiguration DHF (MCDHF) method
imple-mented in the ris
3
/grasp
2
k and ratip [128] (Relativistic calculations of Atomic
Tran-sition, Ionization, and recombination Properties) program packages. Using the MCDHF
method, two different approaches are adopted for the computation of the electronic IS
factors for a set of transitions between low-lying levels of three neutral systems: Mg i (see
Paper B
I), Al i (see Paper B
II), and Zn i (see Paper B
III). The first approach is based
on the estimate of the expectation values of the relativistic nuclear recoil Hamiltonian
for a given isotope, combined with the calculation of the total electron densities at the
origin. In the second approach, the IS factors are extracted from the calculated transition
shifts for given triads of isotopes. Combining the IS factors with observed ISs between
different pairs of isotopes provides the changes in mean-square charge radii of the atomic
nuclei. Within this computational approach for the estimation of the MS and FS factors,
different models for electron correlation are explored in a systematic way to determine a
reliable computational strategy, and to estimate theoretical error bars of the IS factors.
In Paper B
I, the computed IS factors of some well-known transitions in Mg i are
com-bined with available nuclear data. It allows to determine transition ISs for the
26Mg
−
24Mg
pair of isotopes, where experimental [129–133] and theoretical values [118, 119, 134–137]
are available for comparison. By contrast, the lack of accurate theoretical calculations
of IS factors in Al i and Zn i must be pointed out. Hence, Papers B
IIand B
IIIprovide
benchmark values for several transitions involving low-lying states of these two systems.
1.4
Thesis outline
2
D
IRAC EQUATION
2.1
Relativistic context
Relativistic corrections become important when the nuclear charge of an atom increases,
and therefore must be taken into account [138]. Indeed, it is shown that as the number
Z of protons inside the nucleus increases, the electrons of the innermost orbitals reach
speeds near the speed of light [139]. Hence, the framework of relativistic quantum
me-chanics must be adopted. The simplistic relativistic Bohr model, which can only describe
one-particle systems, suffices to illustrate it quantitatively. One of the statements of
special relativity is that the mass of a particle increases towards infinity as its velocity
v approaches the speed of light c in vacuum. The classical theory of special relativity
mathematically formulates the relativistic mass, denoted as
m
relin the following, as
m
rel= γ(v)m =
p
m
1
− (v/c)
2,
(2.1.1)
where
γ(v) is the Lorentz factor and m the mass of the particle at rest. Within the
relativistic Bohr model of a hydrogenic system, let us consider one electron of mass
m
eat rest on a circular orbit around a nucleus of positive point charge
Ze and infinite mass.
Its average speed on the Bohr orbit of principal quantum number
n reads
hvi
n=
Ze
24πε
0n
}
=
Zαc
n
,
(2.1.2)
where
−e is the electron charge, ε
0the vacuum permittivity,
} the reduced Planck
con-stant (=h/2π), and α the fine-structure concon-stant. The second equality of Eq. (2.1.2) is
obtained by using the expression of the fine-structure constant [140]
α =
e
2
4πε
0}c
= 1/137.035 999 139(31).
(2.1.3)
Similarly, the average orbital radius
hri
nand total energy
hEi
nof the atomic electron
are respectively given by
hri
n=
4πε
0n
2}
2m
rel eZe
2=
n
}
m
rel ehvi
n(2.1.4)
and
hEi
n=
−
Ze
24πε
0 2m
rele2n
2}
2.
(2.1.5)
2.1 Relativistic context
Let us also consider that its only electron is on the first Bohr orbit (n = 1), which is
closest to the nucleus and subject to strong electrostatic interaction from all 79 protons.
According to Eq. (2.1.2), its speed is
58% of c (v/c
≈ 79/137). Using Eq. (2.1.1) gives a
substantial mass increase
(m
rele− m
e)/m
eof
22% which in turn, according to Eq. (2.1.4),
implies a decrease of its Bohr radius of
18%, thus corresponding to a radial contraction.
The linear dependence on the electron mass in the total energy (2.1.5) implies a decrease
of
22% in the total energy of the n = 1 electron, i.e., an increase of its binding energy.
Note that if speeds around
v/c
∼ 0.1 are already considered as relativistic, the n = 1
electron of an hydrogenic ion becomes relativistic for nuclei with
Z
≥ 14. It should also
be mentioned that the nonrelativistic Bohr model [141,142], i.e., in which
m
releis replaced
by
m
e, is validated by the Schrödinger equation for one-electron systems [143]. Hence,
these considerations remain valid in the framework of Schrödinger’s theory, considering
the speed of the electron on the
1s orbital.
Furthermore, in order to describe the fine structure of atomic states from fundamental
principles, it is also necessary to treat the bound electrons relativistically [144]. The Dirac
equation describes the behavior of massive elementary particles of half-integer spin, i.e.,
fermions like electrons. This equation is simultaneously compatible with the principles
of quantum mechanics and those of special relativity [47, 48]. This thesis does not focus
in details on the history of the Dirac equation, the latter being fully described in e.g.
Refs. [138, 139].
Atomic units
At this stage, it is necessary to introduce the units used in this thesis, namely, the atomic
units (a.u.), defined by
a
0= m
e= e =
} = 4πε
0= 1,
(2.1.6)
where
a
0=
4πε0}2
mee2
is the Bohr radius and
m
ethe electron mass at rest. The atomic unit
of speed
v
B= αc is defined from expression (2.1.3) of the fine-structure constant α. This
implies that the speed of light
c equals 1/α
≈ 137 a.u. The atomic unit of energy is the
Hartree, defined as
1E
h=
}
2/(m
ea
20) = e
2/(4πε
0a
0)
≈ 27.2 eV, i.e., twice the ionization
energy of the hydrogen atom in its
1s ground state.
Free particle
The classical theory of special relativity reports that the total relativistic energy
E of a
free particle of mass
m at rest is given by the energy-momentum relation
E = γ(v)mc
2=
p
m
2c
4+ p
2c
2,
(2.1.7)
where
p
2is the square of the relativistic momentum
p = γ(v)mv of the particle, with v
denoting its 3-velocity. Applying the correspondence rules of quantum mechanics
E
→ i
∂
∂t
,
p
→ −i∇,
(2.1.8)
and considering the square of identity (2.1.7), the Klein-Gordon equation is obtained as
−
∂
2ψ
∂t
2= m
2
c
4ψ
− c
2∇
2ψ.
(2.1.9)
2.1 Relativistic context
derivative, which implies that the solutions of this equation do not obey the superposition
principle. Moreover, the wave function at time
t = 0 does not fully determine the system
for any further time, as it was the case for the Schrödinger equation through the evolution
operator
U (t, t
0= 0) [138, 145]. Finally, the square root of this equation is not invariant
under Lorentz transformations in the Minkowski space-time Mk
4, and the time and space
coordinates are not treated equally [138]. Indeed, in Mk
4the expression of the 4-vector
x
µin covariant form is written as
x
µ= (x
0≡ ct, x
1, x
2, x
3),
(2.1.10)
where each of the 4 components (µ = 0, 1, 2, 3) has the dimension of a length.
The incompatibilities with special relativity occuring in the Klein-Gordon equation
for massive particles led Dirac [47, 48] to propose another Hamiltonian for a free particle
of mass
m and half-integer spin s. Its expression reads
H
D= cα
· p + βmc
2,
(2.1.11)
where
α and β are the 4
× 4 Dirac matrices. Their expression is given by
α =
0 σ
σ 0
,
β =
I
0
0
−I
,
(2.1.12)
where
I is the 2
× 2 identity matrix, and σ is a vector operator containing the Pauli
matrices
σ
x=
0 1
1 0
,
σ
y=
0
−i
i
0
,
σ
z=
1
0
0
−1
.
(2.1.13)
The link between the operator
σ and the spin angular momentum S is S =
12σ. Matrices
(2.1.13) form with the identity matrix
I a basis of the 2
× 2 Hermitian matrices vector
space [142, 146]. It can be shown that
α and β are Hermitian operators:
α = α
†,
β = β
†,
(2.1.14)
which leads to the required Hermiticity of Hamiltonian
H
D[138]. Moreover, they satisfy
the anticommutation relations
α
2x= α
2y= α
2z= β
2= 1,
{α
x, α
y} = {α
y, α
z} = {α
z, α
x} = 0,
{α
x, β
} = {α
y, β
} = {α
z, β
} = 0,
(2.1.15)
arising from the fact that the solutions of the Dirac equation must also satisfy the
Klein-Gordon equation [138, 144]. The Dirac equation for a free particle reads
cα
· p + βmc
2ψ(r, t) = i
∂ψ(r, t)
∂t
.
(2.1.16)
Different equivalent formulations of the Dirac equation exist, but the present one is the
closest to the Schrödinger equation. As matrices
α and β are 4
× 4 matrices, the solution
ψ(r, t) must be a 4-column vector.
2.2 Hydrogenic ions
Charged particle in an external field
Let us consider a charged particle of charge
q, mass m and (half-integer) spin s, immersed
in an external electromagnetic field. Its Dirac Hamiltonian is given by [138]
H
D= cα
· (p − qA) + βmc
2+ qΦ,
(2.1.17)
where
A is the magnetic vector potential, and Φ the electric scalar potential. Therefore,
the Dirac equation of the particle under consideration reads
cα
· (p − qA) + βmc
2+ qΦ
ψ(r, t) = i
∂ψ(r, t)
∂t
.
(2.1.18)
2.2
Hydrogenic ions
The present section mainly follows the books by Grant [138] and by Johnson [139]. The
reader is invited to refer to these sources for further details.
In the particular case of a central field, potentials
A and Φ read A = 0 and qΦ = V (r),
where
r =
|r|. It is the case for hydrogenic ions, for which V (r) is the Coulomb potential
V
C(r) =
−Z/r. The Dirac equation for a hydrogenic ion describes the motion of an
electron of mass
m
e= 1 a.u., charge
−e = −1 a.u. and spin s = 1/2 in the field of an
atomic nucleus [47, 48]. Its expression reads
cα
· p + βc
2−
Z
r
ψ(r, t) = i
∂ψ(r, t)
∂t
.
(2.2.1)
Only stationary states will be discussed here. Hence, by writing
ψ(r, t) with the form
ψ(r, t) = φ(r) e
−iEt,
(2.2.2)
one obtains the following eigenvalue problem:
cα
· p + βc
2−
Z
r
φ(r) = Eφ(r),
(2.2.3)
where
E denotes the binding energy of the electron in the Coulomb potential of the
atomic nucleus.
One notices that
H
Ddoes not commute neither with
L
2, the square of the orbital
angular momentum
L, nor with S
2, the square of the spin angular momentum
S. Indeed,
[L, H
D] = iα
× p →
H
D, L
26= 0,
[S, H
D] =
−iα × p →
H
D, S
26= 0.
(2.2.4)
Hence, the common set of commuting observables (CSCO)
{L
2, L
z, S
2, S
z} is not
com-patible with
H
D. However,
H
Dcommutes with the total angular momentum
J = L + S.
Indeed,
H
Dis invariant under rotations, therefore it commutes with the generator of the
SO(3) group, namely
J [147]. Hence, because [H
D, J
x] = [H
D, J
y] = [H
D, J
z] = 0, H
Dalso commutes with
J
2:
H
D, J
2= [H
D, J
z] = 0.
(2.2.5)
According to Eq. (2.2.5), there exist common eigenfunctions to
H
D,
J
2and
J
z[145]. The
following change of CSCO is thus required:
2.2 Hydrogenic ions
The basis change towards a coupled representation is built by means of Clebsch-Gordan
coefficients:
|j m
ji =
lX
ml=−l 1/2X
ms=−1/2|l m
l1/2 m
sihl m
l1/2 m
s|j m
ji
=
1/2X
ms=−1/2|l m
j− m
s1/2 m
sihl m
j− m
s1/2 m
s|j m
ji,
(2.2.7)
where the second equality arises from the fact that the total projection
m
jmust be
the sum of
m
land
m
sso that the Clebsch-Gordan coefficient
hl m
l1/2 m
s|j m
ji is not
null [146, 148]. Hence, one has
m
l= m
j− m
s. According to the triangular relations,
the quantum number
j must verify
|l − 1/2| ≤ j ≤ l + 1/2, thus j is equal to |l ± 1/2|.
Moreover, one has
m
j=
−j, −j + 1, ..., j − 1, j.
Kets
|j m
ji are eigenstates of J
2and
J
zwith respective eigenvalues
j(j + 1) and
m
j. They correspond to the spherical spinors
χ
jlmj, constructed from combinations of
spherical harmonics
Y
lml
, eigenstates of
L
2
and
L
z
with respective eigenvalues
l(l + 1)
and
m
l, and of Pauli spinors
χ
ms, eigenstates of
S
2
and
S
z
with respective eigenvalues
s(s + 1) = 3/4 and m
s=
±1/2 [146]. Let us rewrite (2.2.7):
χ
jlmj(ˆ
r) =
1/2
X
ms=−1/2hl m
j− m
s1/2 m
s|j m
jiY
mlj−ms(ˆ
r)χ
ms,
(2.2.8)
or, with a spinor form,
χ
jlmj(ˆ
r) =
hl m
j− 1/2; 1/2 + 1/2|j m
jiY
mlj−1/2(ˆ
r)
hl m
j+ 1/2; 1/2
− 1/2|j m
jiY
mlj+1/2(ˆ
r)
!
,
(2.2.9)
where
r
ˆ
≡ (θ, ϕ). The explicit expression of the spherical spinors reads [138, 139]
χ
l+1/2,lmj(ˆ
r) =
q
l+mj+1/2 2l+1Y
l mj−1/2(ˆ
r)
q
l−mj+1/2 2l+1Y
l mj+1/2(ˆ
r)
,
(2.2.10)
χ
l−1/2,lmj(ˆ
r) =
−
q
l−mj+1/2 2l+1Y
l mj−1/2(ˆ
r)
q
l+mj+1/2 2l+1Y
l mj+1/2(ˆ
r)
.
(2.2.11)
Each of these eigenvectors satisfies the following eigenvalue equations:
L
2χ
jlmj= l(l + 1) χ
jlmj,
S
2χ
jlmj=
3
4
χ
jlmj,
J
2χ
jlmj= j(j + 1) χ
jlmj,
J
zχ
jlmj= m
jχ
jlmj.
(2.2.12)
The CSCO
{L
2, S
2, J
2, J
z
} is no longer compatible with H
Dbecause
[H
D, L
2]
6= 0 and
[H
D, S
2]
6= 0, even if [H
D, J
2] = [H
D, J
z] = 0. This is the reason why it is interesting to
introduce the operator
K defined as
K =
−(J
2− L
2− S
2+ 1)
=
−(2 S · L + 1)
2.2 Hydrogenic ions
where the second equality arises from
J
2= L
2+ S
2+ 2 S
· L. Spherical spinors χ
jlmj(ˆ
r)
are eigenfunctions of
σ
· L, and therefore eigenfunctions of K, following
K χ
jlmj= κ χ
jlmj,
(2.2.14)
whose (integer) eigenvalues
κ are
κ =
− (j +
1
2
) =
−l − 1, if j = l +
1
2
+ (j +
1
2
) = l,
if
j = l
−
1
2
.
(2.2.15)
The value of
κ determines both j and l quantum numbers. This is why spherical spinors
are written with the form
χ
κmin the following [138]. Table 2.1 displays some values of
l,
j and κ, as well as the corresponding spectroscopic notation of some relativistic subshells.
It is worthwhile to notice that
|κ| = j + 1/2.
Table 2.1:
Spectroscopic notation of some relativistic subshells.
s
1/2p
1/2p
3/2d
3/2d
5/2f
5/2f
7/2l
0
1
1
2
2
3
3
j
1/2
1/2
3/2
3/2
5/2
5/2
7/2
κ
−1
+1
−2
+2
−3
+3
−4
It can be shown that the operator K
= βK commutes with H
D[138, 139]. The
compatible CSCO with
H
Dis hence
{J
2, J
z, K
}. Hamiltonian H
Dis also invariant under
spatial inversion, therefore it commutes with P
= βΠ, where the parity operator Π is the
generator of the
Z
2group [147]. In spherical coordinates, the operator
Π transforms the
azimutal angle
ϕ into (ϕ + π) and the colatitude θ into (π
−θ). When applied to spherical
harmonics
Y
lml
(ˆ
r), this operator provides the eigenvalue π = (
−1)
l
, according to
Π Y
mll(ˆ
r) = Y
l ml(π
− θ, ϕ + π) = (−1)
lY
l ml(ˆ
r).
(2.2.16)
One concludes that spherical spinors are eigenfunctions of
Π with the associated
eigen-value
(
−1)
l. The two spherical spinors
χ
κm
and
χ
−κm, corresponding to the same
j value,
present
l values that differ by one unit. Hence, they are of opposite parity.
Finally, the eigenstates of the Dirac equation (2.2.3) have the form
φ
nκm(r) =
1
r
P
nκ(r) χ
κm(ˆ
r)
iQ
nκ(r) χ
−κm(ˆ
r)
,
(2.2.17)
where
P
nκ(r) and Q
nκ(r) respectively correspond to the large and the small radial
com-ponent [138, 139]. Each of these eigenvectors satisfies the following eigenvalue equations:
J
2φ
nκm= j(j + 1) φ
nκm,
J
zφ
nκm= m φ
nκm,
K
φ
nκm= κ φ
nκm,
P
φ
nκm= π φ
nκm.
(2.2.18)
2.2 Hydrogenic ions
As spherical spinors satisfy the orthogonality relations
Z
π 0sin θ dθ
Z
2π 0dϕ χ
†κ0m0(θ, ϕ) χ
κm(θ, ϕ) = δ
κ0κδ
m0m,
(2.2.19)
the normalization condition for the eigenstates
φ
nκm(r),
Z
R3
φ
†nκm(r) φ
nκm(r) d
3r = 1,
(2.2.20)
reduces to a normalization condition on the radial wave functions
P
nκ(r) and Q
nκ(r):
Z
∞0
P
nκ2(r) + Q
2nκ(r)
dr = 1.
(2.2.21)
2.2.1
Radial Dirac equations
The operator
cσ
· p can be written as [138]
cσ
· p = −icσ · e
r∂
∂r
−
σ
· L
r
,
(2.2.22)
where
e
ris the unit radial vector. The action of the operator
σ
· e
ron spherical spinors
χ
κm(ˆ
r) proceeds as
σ
· e
rχ
κm(ˆ
r) =
−χ
−κm(ˆ
r).
(2.2.23)
Hence, as
K + 1 =
−σ · L according to Eq. (2.2.13), one has
cσ
· p f(r) χ
κm(ˆ
r) = i
d
dr
f (r) +
κ + 1
r
f (r)
χ
−κm(ˆ
r).
(2.2.24)
This expression, together with expression (2.2.17) of the wave function
φ
nκm(r) allow to
derive from Eq. (2.2.3) the system of coupled radial Dirac equations [138]:
V
C(r) + c
2− E
nκ−c
drd−
κrc
drd+
κrV
C(r)
− c
2− E
nκP
nκ(r)
Q
nκ(r)
= 0.
(2.2.25)
This eigenvalue problem is analytically solvable. The detailed calculation is available in
Refs. [138, 139]. Here only its solutions are presented, i.e., the energies
E
nκof the bound
states, and the corresponding radial wave functions
P
nκ(r) and Q
nκ(r).
2.2.1.1 Dirac energies
Solving these eigenvalue equations, the energies of the bound states are given in a.u.
by [138]
E
nκ= c
2λ
γ + n
− |κ|
Z
=
c
2s
1 +
(αZ)
2(γ + n
− |κ|)
2= c
2p
γ + n
r(γ + n
r)
2+ (αZ)
2|
{z
}
N,
(2.2.26)
where
λ =
p
c
2− E
2 nκ/c
2and
γ =
p
κ
2− (αZ)
2. The link between the principal quantum
number
n and the quantum number n
ris given by
2.2 Hydrogenic ions
The number of nodes in the large component
P
nκ(r) is n
− l − 1, as in the nonrelativistic
case, while the number of nodes in
Q
nκ(r) is n
− l − 1 for κ < 0 and n − l for κ > 0 [139].
The number N involved in Eq. (2.2.26) is called the effective principal quantum
num-ber. Its expression reads
N
=
p
(γ + n
r)
2+ (αZ)
2=
p
n
2− 2n
r
(
|κ| − γ).
(2.2.28)
Its role is analogue to that of the quantum number
n within the nonrelativistic
frame-work [144]. One notices that for
n
r= 0, N takes the same value as n.
The energy levels only depend on
|κ| = j + 1/2. Hence, for a given n, the solutions
for
±κ are degenerate (e.g. 2s
1/2and
2p
1/2), which preserves the degeneracy in
l as in
the nonrelativistic case. Besides, the Dirac equation does not account for the Lamb shift,
which is a correction arising from quantum electrodynamics (QED) [149,150]. Indeed, the
Lamb shift lifts the degeneracy of these levels, that can be observed experimentally [145].
By contrast, two levels with the same values of
n and l but with different values of j,
e.g.
2p
1/2and
2p
3/2, do not have the same energy, and are separated by a fine-structure
splitting [138, 139].
Expanding
E
nκ(in a.u.) in powers of
(αZ)
2leads to
E
nκ= c
21
−
(αZ)
22n
2−
(αZ)
42n
4n
|κ|
−
3
4
+
· · ·
= c
2−
Z
22n
2−
α
2Z
42n
31
|κ|
−
3
4n
+
· · ·
(2.2.29)
The first term corresponds to the electron rest energy
c
2, the second one to the
nonrel-ativistic binding energy, and the next ones to relnonrel-ativistic corrections. Hence, if the rest
energy
c
2is subtracted from
E
nκ
, the energy
ε
nκis defined as the sum of the
nonrelativis-tic binding energy,
−Z
2/2n
2, and of relativistic corrections. The nonrelativistic binding
energy is therefore recovered at the
c
→ ∞ limit.
For
ε > 0 (or E > +c
2), the electron is no longer bound to the nucleus, and essentially
behaves as a free particle. Thus, there exists a continuous energy spectrum for these
values. For
−c
2< ε < 0 (or 0 < E < +c
2), there exists a discrete spectrum of bound
states of energy
E
nκ. Besides, within Dirac’s theory it is mathematically possible to
obtain a negative square root in the expression of
E
nκ[47,48]. Thus, there exists a second
discrete spectrum of bound states, corresponding to
−2c
2< ε <
−c
2(or
−c
2< E < 0).
Finally, a second continuous spectrum exists for
ε <
−2c
2(or
E <
−c
2). The existence
of this negative energy continuum implies that the Dirac Hamiltonian
H
Ddoes not have
a global lower bound. Hence, one can wonder whether the variational method, which lies
on the existence of such a bound (according to the Ritz theorem) [146], is still applicable
within the framework of the Dirac equation. However, it is possible to demonstrate the
convergence of the variational method for
H
D[138]. The discrete spectrum
E
1< E
2<
· · ·
over
(
−c
2, +c
2) has an accumulation point at +c
2. The existence of the continua for
E > +c
2and
E <
−c
2implies that there exists a lower bound such that
−c
2< E < E
1
for the discrete spectrum, as well as an upper bound
E
≤ −c
2for the negative energy
continuum.
2.2 Hydrogenic ions
E (a.u.)
+c
2−c
20
Z > 0
Z = 0
Z < 0
p.e.c.
n.e.c.
(a)
(b)
(c)
Figure 2.2.1:
Dirac energy spectrum for (a) a hydrogenic system, (b) a free electron, (c) an
anti-hydrogenic system.
2.2.1.2 Radial wave functions
The calculation of the radial wave functions
P
nκ(r) and Q
nκ(r) is detailed in Refs. [138,
139]. Their expressions read
P
nκ(r) =
p
1 + E
nκ/c
2N
nκe
−x/2x
γ[(N
− κ)
1F
1(
−n + |κ|, 2γ + 1, x)
−(n − |κ|)
1F
1(
−n + |κ| + 1, 2γ + 1, x)] ,
(2.2.30)
Q
nκ(r) =
p
1
− E
nκ/c
2N
nκe
−x/2x
γ[
−(N − κ)
1F
1(
−n + |κ|, 2γ + 1, x)
−(n − |κ|)
1F
1(
−n + |κ| + 1, 2γ + 1, x)] ,
(2.2.31)
where
x = 2λr = 2Zr/N using Eq. (2.2.26). The normalization factor N
nκis given by
N
nκ=
1
N
Γ(2γ + 1)
s
Z Γ(2γ + 1 + n
− |κ|)
2(n
− |κ|)!(N − κ)
,
(2.2.32)
where
Γ is the Eulerian function of the first kind [151]. The functions
1F
1involved in
Eqs. (2.2.30) and (2.2.31) are the confluent hypergeometric functions [151]
1