0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)

70 Ansichten215 SeitenFeb 12, 2012

© Attribution Non-Commercial (BY-NC)

PDF, TXT oder online auf Scribd lesen

Attribution Non-Commercial (BY-NC)

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)

70 Ansichten215 SeitenAttribution Non-Commercial (BY-NC)

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

Sie sind auf Seite 1von 215

Dissertation zur Erlangung des akademischen Grades doctor rerum naturalium (Dr. rer. nat.) der Mathematisch-Naturwissenschaftlichen Fakultt a der Universitt Rostock a

Liste der Gutachter Prof. Dr. Gerd Rpke o Institut fr Physik u Universitt Rostock a

Prof. Dr. Wojciech Rozmus Theoretical Physics Institute University of Alberta Edmonton, Kanada Datum der Verteidigung: 28. Oktober 2008

Wenn aber gleich alle unsere Erkenntnis mit der Erfahrung anhebt, so entspringt sie darum doch nicht eben alle aus der Erfahrung. Immanuel Kant Kritik der reinen Vernunft

Kurzfassung

Bremsstrahlung ist ein fundamentaler Prozess der Licht-Materie Wechselwirkung. Sie bezeichnet die Strahlungsemission whrend der Streuung von geladenen Teilchen. Bremsstraha lungsspektroskopie an Plasmen ermglicht die genaue Bestimmung wichtiger Plasmapao rameter. In dichten Plasmen kommt es infolge von Vielteilcheneekten zu einer Modizierung des Emissionsspektrums. Beispiele hierfr sind die Abschirmung des Streupotenu tials und wiederholte Teilchenste whrend des Emissionsvorgangs. Diese fhren zu einer o a u Absenkung des Bremsstrahlungsspektrums bei kleinen Photonenenergien. Die in dieser Arbeit vorgestellte Vielteilchentheoretische Beschreibung von Bremsstrahlung erlaubt die systematische Darstellung des Einusses von statischen und dynamischen Korrelationen auf das Emissions und Absorptionsspektrum. In diesem Formalismus sind die Korrelationen in der Einteilchen-Selbstenergie enthalten, die die endliche Lebensdauer von Einteilchenzustnden im Medium beschreibt. Die Berechnung der Selbstenergie geschieht mithilfe der a GW (0) -Methode, wodurch uber die Mean-Field Nherung hinausgegangen wird. Vorher a (0) unbekannte Resultate fr die GW -Selbstenergie in einem nichtentarteten Plasma weru den prsentiert. Ergebnisse zur Modikation der Bremsstrahlung im Medium werden in a Form des Gauntfaktors angegeben.

Contents

Abstract/Kurzfassung Table of content i iv

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v

1 1 1 4 5 7 8 13 13 33 38 41 42 45 45 46 48 49 49 51 53 53 57 59

1 Introduction 1.1 Plasmas . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Characterization of plasmas and plasma parameters 1.3 Interaction of Radiation with Matter . . . . . . . . 1.4 Many-particle Quantum Electrodynamics . . . . . . 1.5 Bremsstrahlung in dense media . . . . . . . . . . . 1.6 Aim and structure of this work . . . . . . . . . . .

2 Many-particle theory and application to plasmas 2.1 Quantum Electrodynamics . . . . . . . . . . . . . . . . 2.2 Many-particle eld theory and the GW -approximation 2.3 Matsubara Green functions . . . . . . . . . . . . . . . . 2.4 Application: The self-energy in GW (0) approximation . 2.5 Optical properties of dense plasmas . . . . . . . . . . . 3 Plasmas and radiation 3.1 Basic processes and plasma diagnostics 3.2 Line emission . . . . . . . . . . . . . . 3.3 Recombination radiation . . . . . . . . 3.4 Bremsstrahlung radiation . . . . . . . . 3.5 Scattering . . . . . . . . . . . . . . . . 3.6 Radiation transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Selected problems 4.1 Bremsstrahlung vs. Thomson Scattering . . . . . . . . . . . . . . . . . . . 4.2 Bremsstrahlung and Line Spectroscopy of Warm Dense Al Plasma . . . . . 4.3 Landau-Pomeranchuk-Migdal Eect in Dense Plasmas . . . . . . . . . . . .

iv 4.4 4.5

CONTENTS Optical Properties and Spectral Function . . . . . . . . . . . . . . . . . . . Self-Consistent Spectral Function and Analytic Scaling Behaviour . . . . . 63 65 76 78 96

II

Published articles

97

99 117 133 147 155 185 200 201 205

5 Bremsstrahlung and the LPM eect 6 Bremsstrahlung vs. Thomson Scattering 7 Optical Properties and Spectral Function 8 Spectroscopy of Warm Dense Al Plasma 9 Self-Consistent Spectral Function and Scaling Behaviour 10 Single Particle Spectral Function for the Classical Plasma Curriculum vitae List of publications/Liste der Verentlichungen o Declaration of authorship/Selbststndigkeitserklrung a a

Chapter 1 Introduction

1.1 Plasmas

As to our latest knowledge, only 5% of the matter present in our universe is visible matter, i.e. sensitive to electromagnetic radiation [Kra01]. The remaining portion consists of dark matter (23%) and dark energy (72%). Out of the 5% visible matter, more than 99% [GB05] are in the so-called plasma state. The term plasma was originally introduced by Langmuir [Lan28] to describe an ionized gas in an arc discharge. Today, plasmas are understood as matter in a state characterized by a signicant (e.g. > 10%) degree of ionization. Typical every-day examples are tube discharges, ames or plasma lamps. Most plasmas occurring in nature are found in astrophysical objects, such as the interior of stars [Ich93], giant planets [Bag92] or in the interstellar medium [Alf60]. In this work, the term plasma is used mainly in the context of Coulomb systems, i.e. matter consisting of electrons and atomic nuclei, that interact via the Coulomb force. In this sense, also systems like excited semi-conductors and metals, where electrons in the conduction band behave similar to free electrons in a plasma, make part of the considerations. More fundamental particles (in the sense of elementary particle physics), e.g. hadrons and quarks, will only be mentioned in marginal remarks. It should be noted at this point, that also the state of hadrons dissolved into quarks and gluons is referred to as a plasma, the so-called quarkgluon plasma [HM96].

1.2

In general, plasmas are in a non-equilibrium state that needs a detailed characterization. The concept of the time dependent distribution function is used to describe the plasma kinetics [Ich92]. However, quite often, conditions are met where the plasma state can be assumed to be in a global or local thermodynamic equilibrium, characterized by a few parameters such as temperature and density, possibly depending on the position. Near

Introduction

equilibrium states can be described by the degree of ionization as a function of time and position and assuming dierent temperatures for the various subspecies, such as atoms, unbound electrons, and ions [Chi00]. Fluctuations near equilibrium are considered in linear response theory, see e.g. [ZMR96a]. A very useful illustration of plasmas and Coulomb systems in general is their representation in the parameter plane spanned by the density of unbound electrons ne and the temperature T , as shown in Fig. 1.1. On the left plane we see the typical parameter

Figure 1.1: Density-temperature plane with important examples for plasmas and isolines for the plasma coupling parameter and the electron degeneracy parameter . Plasmas occurring in nature (astrophysical plasmas) are given on the left, the right plane shows regions where laboratory plasmas are found or are expected to be found. This graph courtesy of Dr. A. Hll. o ranges for plasmas occurring in astrophysical objects. The right plane shows the parameters of plasmas that are produced in a laboratory, i.e. fusion plasmas, such as magnetically conned plasmas (Tokamak) and plasmas generated during inertial connement fusion (ICF). Furthermore, the parameter ranges of typical condensed matter objects (metals and semiconductors, the electron density is the concentration of carriers in the conduction band) are shown as well as the parameters for ion traps. Besides the examples, Fig. 1.1 also shows isolines of the electron coupling parameter e (dark blue lines) and the electron degeneracy parameter e (light blue lines). The coupling parameter c =

2 Zc e2 4 0 d kB T

(1.1)

is the ratio of the Coulomb interaction among two particles of species c at the mean interparticle separation d = (3/4nc )1/3 and the thermal energy kB T . Plasmas characterized by c < 1 are called weakly coupled. Their behaviour is largely governed by the temperature of the system. The individual particles can be considered as free particles, each moving independently of the others. Plasma observables are easily calculated by averaging the corresponding single-particle quantity over the Maxwell-Boltzmann distribution function. In the opposite case, c > 1, the term strongly coupled plasma is used. The mutual Coulomb interaction is larger than the mean kinetic energy of each particle. Hence, the calculation of observables is much more complicated, since the interaction of the particles has to be accounted for. Examples for strong coupling eects are the formation of bound states (at low temperatures, e.g. in traps), the dissolution of bound states due to the Mott eect (at high densities), the structuring of the plasma and formation of a crystal lattice in the limit of low temperatures and high densities (liquids, semiconductors, metals). Note, that the coupling parameter looses its meaning as the ratio of potential energy to kinetic energy, when the temperature becomes smaller than the Fermi energy. In this case, the mean kinetic energy of the system scales with the Fermi energy rather than the temperature. As the relevant parameter characterizing dense and cold systems, one introduces the degeneracy parameter c as the ratio of the thermal energy to the Fermi energy, 1/3 2 2 kF,c 6 2 nc kB T , EF,c = , kF,c = , (1.2) c = EF,c 2mc 2sc + 1 with kF,c being the Fermi momentum and mc being the mass of particles of species c, sc is the species spin quantum number. The limit c 1 characterizes plasmas that can be treated as ideal Fermi gases, here, the quantum nature of the particles becomes important and determines the plasmas comportment. This happens usually at high densities, where the mean particle separation becomes so small, that the wavefunctions overlap and quantum interference sets in. Observable eects are the Pauli blocking for Fermions or quantum condensation like Bose-Einstein condensation (BEC) or Bardeen-Cooper-Schrieer-pairing (BSC). In degenerate systems, the Fermi-Dirac distribution function nF,c (p) = [1 + exp(

2 2

p /2mc c ]/kB T )

(1.3)

approaches a step function with a discontinuity at the Fermi energy EF,c = c (T = 0); c is the chemical potential of species c. In the limit 1 we speak of a classical plasma, the particles can be treated as classical objects, i.e. their equilibrium momentum distribution function can be approximated by a Maxwell-Boltzmann distribution, nM,c (p) = nc 3 c exp( 2 p2 /2mc ) , 2sc + 1 (1.4)

Introduction

Most real-world plasmas are neither ideal quantum gases nor classical, weakly coupled plasmas. The theoretical description of these plasmas is dicult. The Coulomb interaction among the particles and their quantum nature leads to an enormous degree of complexity. Whereas in ideal systems, the observables can be derived from the single-particle behaviour, spatial and temporal correlations among the constituents govern the physics of real plasmas and lead to the emergence of new phenomena. Collective modes appear, leading to a modication of the response of the plasma to external elds. To deal with these complexities, it is helpful to introduce the notion of quasi-particles interacting with each other via eective forces [KKER86]. Also bound states can be included in this picture as a distinct particle species. E.g. the many-electron problem of an electron scattering on bound electrons in an atom is reduced to the one-particle problem of an electron scattering in the eective potential of the atom [RDO05]. Similarly, plasmon modes are described as quasi-particles giving rise to the dynamical screening of the Coulomb potential. The weakening of the interaction due to screening can lead to the disintegration of bound states above a certain critical density, the so-called Mott-eect [KSK05, FIK06]. In this way, temperature and density directly inuence the thermodynamic properties of the plasma, i.e. its equation of state. Another important category of plasma properties is the response of the plasma to external perturbations. Examples are the plasmas DC conductivity (response to electric elds), and magnetic susceptibility (response to magnetic elds), optical properties like opacity, reectivity and emission (response to electromagnetic elds) and the response to particle beams, like the stopping power.

1.3

The interaction of plasmas with electromagnetic radiation is a long standing issue in both theoretical and experimental physics. Since light propagation in a medium is limited to wavelengths below the critical wavelength c = 2c/pl , with the electron plasma frequency pl = ne e2 / 0 me , investigations of condensed matter and plasmas near solid density or above require light sources in the far UV or x-ray regime. Light sources of the so-called fourth generation, such as the free electron laser at DESY-Hamburg (FLASH) [A+ 06], provide unprecedented parameters in brilliance, temporal and spatial coherence and at the same time short wavelengths [AAA+ 07]. Notably in the domain of plasma diagnostics, FELs oer intriguing possibilities, e.g. via collective Thomson scattering [HBC+ 07]. Further challenging projects aim at time resolved investigations, e.g. of chemical and biological processes [Tsc04, TT05, HRCK+ 07]. On the other hand, the development in short pulse and high power laser technology opened up a whole catalogue of interesting new phenomena to investigate [GF96]. Intense lasers provide an excellent tool to generate high energy density plasmas, e.g. in inertial connement fusion research [Lin95]. Eects related to high intensities are multi-photon ionization, generation of higher laser harmonics [TPS+ 03], attosecond pulse generation [CMK97, WCL+ 08], and the laser self-focussing eect [SS99]. At relativistic laser in-

tensities (> 1018 W/cm2 ), when the quiver energy of electrons becomes comparable or larger than the electrons rest mass [Gib06], bubble-acceleration [Puk03] sets in, opening new perspectives for small-scale particle accelerators [GTT+ 04, FGP+ 04]. From a theoretical point of view, ultrarelativistic laser-plasma eects expected at highest intensities (> 1021 W/cm2 ), e.g. the birefringence of the vacuum [HLA+ 06], pair creation by vacuum tunneling [BPR+ 06] and the Schwinger mechanism [Sch51a, DW97, SW92], i.e. the break down of the quantum vacuum, are of special interest. These eects relate to fundamental questions of quantum eld theory. A quantum eld theoretical formulation of laser-plasma interaction at such conditions is indispensable. The most general approach to the interaction between charged particle systems and electromagnetic elds is quantum electrodynamics (QED). Its basic concepts will be outlined briey in section 2.1, with special emphasis on the response of plasmas to electromagnetic elds. For more detailed presentations, I refer to textbooks e.g. by Itzykson and Zuber [IZ80], Gross [Gro93], or Bjorken and Drell [BD65]. In the end, expressions for the absorption, emission, and the scattering of electromagnetic waves or photons will be presented, that are obtained from QED via a perturbation expansion of the QED Hamiltonian. The question of non-perturbative treatments will be covered in section 2.2.

1.4

A theoretical approach to the interaction of electromagnetic elds with correlated matter, that can be extended to such extreme conditions as outlined above, needs to attend three main issues: First, the correlations in the system have to be considered as well as (second) the interaction between the light and the matter constituents. Both these questions should be addressed within a common framework. Only in this way, modications of the basic processes of absorption, emission, and scattering due to correlations and the creation or destruction of correlations in the plasma through radiation elds, notably coherent radiation elds, can be accounted for systematically and without double counting of one or the other. Thirdly, taking on the example of a plasma created by interaction of intense laser light with condensed matter, it is clear that also non-equilibrium aspects need to be incorporated, such as density and temperature gradients, instabilities and relaxation processes. A eld-theoretical generalization of linear response theory is straightforward [MRH02a, MRH02b]. In plasma physics, many-body theoretical methods [Mah81, FW71], have a long tradition of application to thermodynamic properties [KKER86], transport cross sections [RRRW00], and collective excitations. Correlations due to the Coulomb interaction among the constituent particles and their Bose- or Fermi statistics are calculated from the systems Hamiltonian after a perturbative expansion in terms of the interaction. Often, summation schemes are applied in order to include also inherently non-perturbative phenomena, such as the occurrence of bound states or the dielectric screening of the interaction potential. Standardly, the Hamiltonian contains the free movement of the particles, i.e. the kinetic energy, and the electrostatic interaction (Coulomb interaction).

Introduction

On the other hand, Quantum Electrodynamics is successfully applied to the calculation of those processes, were electrically charged particles, that can be regarded as free particles in good approximation, interact with electromagnetic elds, i.e. photons. Here, the Hamiltonian, besides the free movement, contains the coupling between Fermions and gauge Bosons (photons), which is treated as a perturbation. Quantum electrodynamics is relativistically covariant and both massive particles and gauge bosons are considered as quantum objects. However, the Coulomb interaction and the resulting correlations and collective phenomena, which are accounted for in many-body theory, is either not dealt with at all or it is regarded as a part of the unperturbed Hamiltonian. The latter allows for the treatment of interaction of atoms, molecules, or ions with the radiation eld. Besides established analytic techniques, based on perturbation theory, also numerical approaches have matured over the last years. By use of Dirac equation solvers [BSG99, MK04], various eects of interaction between ultra-intense (> 1020 W/cm2 ) laser pulses and atoms have been investigated, such as bremsstrahlung emission by laser accelerated electrons [SLJK07], high harmonic generation [FLK06] or induction of nuclear reactions [BEK06] have been accessed. However, complex plasma dynamics, involving dynamical screening, ionization and recombination, or occurrence of collective modes are often not accounted for in QED. Manyparticle quantum electrodynamics [Rei05], i.e. the synthesis of both theoretical concepts, provides a systematic approach to the description of strongly correlated systems exposed to external electromagnetic elds. The interplay between statical and dynamical correlations on the one side and the coupling of matter to the radiation eld can be studied within a common, consistent theoretical framework. The starting point is the Lagrangian of QED. A central task is to nd consistent non-perturbative approximations. A systematic approach to this problem is the -derivable technique [Bay62] and Dyson-Schwinger equations [RW94]. Further details are provided in chapter 2. This work is contained in the frame of an ambiguous and progressive research center situated at the Institute of Physics of the University of Rostock, the Collaborative Research Center 652: Strong Correlations and Collective Phenomena in Radiation Fields: Coulomb systems, Clusters, and Particles. In this center, correlated systems submitted to electromagnetic radiation are studied under various aspects, both experimentally and theoretically. The dynamical response of metal clusters in strong, short-pulse laser elds, plasmas interacting with free electron laser radiation, quantum dynamics of ions and atoms in traps, optical properties of highly excited semiconductors, and quantum condensation in strongly correlated electron-hole systems are the main topics, that are investigated in 13 projects. The subject of this thesis work, emanates from the project A4 (Many-body Quantum Electrodynamics and Dielectric Response). Besides bremsstrahlung in dense plasmas, another PhD thesis in this project deals with optical properties of metal clusters, investigated by means of molecular dynamic simulations [RRR07].

1.5

Bremsstrahlung denotes the emission of a photon by an accelerated charge, e.g. during the scattering of an electron in the Coulomb potential of an ion. A relativistic and quantum mechanical description of bremsstrahlung is provided by QED within second order perturbation theory. On the other hand, bremsstrahlung is the dominant process of emission of radiation from a hot and highly ionized plasma. The inverse process, absorption of photons during particle collisions, so-called inverse bremsstrahlung, is used e.g. for plasma heating [SRB06, CR85, BRS+ 03]. The precise prediction of the bremsstrahlung spectrum is of primary importance for plasma diagnostic purposes [Hut87]. Bremsstrahlung spectroscopy allows for the determination of many important plasma parameters, such as the density of the plasma, its composition, and its temperature. In principle, the electron velocity distribution function can be reconstructed from the bremsstrahlung spectrum [BEH+ 06], and thus also non-equilibrium features, such as non-thermal components, can be characterized. But also from a more theoretical point of view, bremsstrahlung is an interesting scientic issue: The emission (or absorption) of radiation by free or quasi-free particles (e.g. electrons) requires the presence of a third particle or collision partner (the ion) to carry away the recoil momentum, otherwise, conservation of momentum would be violated. This means that in order to describe bremsstrahlung in a many-particle environment, one has to adopt not only the formalism of Quantum Electrodynamics to describe the coupling of emitted or absorbed radiation and the particles, but one also has to account for the correlation and interaction among charged particles. Thus, bremsstrahlung is particularly well suited to study the interplay between correlations in the plasma and the coupling to the radiation eld. An example for the modication of the bremsstrahlung emission spectrum due to correlations in the plasma is the so-called Landau-Pomeranchuk-Migdal (LPM) eect [LP53, Mig56, Kle99]. At small photon energies, when the wavelength of the emitted photon is of the same order as the mean distance between nuclei, the scattering amplitudes of successive scatterings interfere. In a disordered system, this interference is mainly destructive, which leads to a suppression of the bremsstrahlung emission at small photon energies. The LPM eect was observed in high energy electron scattering experiments by many dierent groups, see e.g. [Kas85, ABSB+ 97, HUB+ 03]. Besides successive scattering, also the screening of the scattering potential is known to eect the bremsstrahlung emission. Ter-Mikaelyan [TM53] showed, that screening of the interaction potential leads to a further decrease of the bremsstrahlung emission and at very low frequencies (in the vicinity of the mediums plasma frequency) dominates the LPM-suppression. Being approved unambiguously in the case of high energy electron scattering, the question remains, if the same eects are also present in the bremsstrahlung emission from a hot and dense plasma. In a consistent, many-body eld theoretical approach to bremsstrahlung they should naturally be included. The approach is straightforward: In general, optical properties of charged particle

Introduction

systems, e.g. absorption and emission, are described by the dielectric tensor of the medium [Jac75]. The latter is a functional of the single-particle Green function and the electronphoton vertex function [KKER86]. These functions are well dened objects of quantum statistics, describing the propagation of particles in the medium and the coupling to the radiation eld, respectively. Thus, we have a relation between the macroscopic properties, such as emission and absorption spectra, to the microscopic determinants of the system. The account for many-particle eects in the calculation of the Green function and the vertex opens the possibility to describe medium eects on optical properties within a self-contained formalism. Here, the focus lies on the calculation of the inuence of manyparticle eects on the bremsstrahlung emission and inverse bremsstrahlung absorption from a plasma in thermal equilibrium. However, further macroscopic quantities, such as transport coecients (electrical and thermal conductivity, stopping power) or the equation of state of the plasma can be accessed within the same approach [KKER86].

1.6

This work aims at the description of bremsstrahlung from dense, correlated media, in the framework of many-particle QED. This general approach enables us to account for the various eects of correlations on the bremsstrahlung emission in a systematic way. Traditionally, corrections of the bremsstrahlung spectrum due to medium and quantum eects, are known as the Gaunt factor [Gau30, BH62]. It gives the deviation of the bremsstrahlung spectrum with respect to an expression that goes back to Kramers [Kra23], who neglected correlations and quantum eects altogether. In this sense, the work aims at a many-body eld theory of the Gaunt factor. Being a prototypical process of QED, the study of bremsstrahlung in a plasma directly leads to a number of questions that derive from the more general problems discussed in the previous sections. The electron-ion scattering becomes screened, which also aects the bremsstrahlung emission. Furthermore, the picture of the emitting electron as a free particle before and after the scattering breaks down. The account for the nite life time of single-particle states, i.e. momentum states, in the calculation of bremsstrahlung is a mayor topic of this work. Other important issues are e.g. bremsstrahlung in non-equilibrium systems [May07], coherence eects, such as bremsstrahlung of laser-accelerated electrons [SLJK07], coherent superposition of bremsstrahlung photons from electron scattering in ion lattices [BK05] or the question of bremsstrahlung in magnetized plasmas [NV83]. Bremsstrahlung is mainly emitted during electron-ion collisions. Electron-electron bremsstrahlung is negligible at least at non-relativistic energies. Therefore, a rst central point in this work is the analysis of the electron-ion scattering process beyond Born approximation, i.e. beyond the perturbative result provided by QED. Born approximation is appropriate for small transfer momenta. At large transfer momenta, it drastically overestimates the scattering cross section[RRRW00]. Here, the concept of the t-matrix has to be used, i.e. the summation of single photon exchange diagrams that make up the Born series. This leads to the Sommerfeld approximation for the Gaunt factor [WMR+ 01].

The Sommerfeld Gaunt factor is applied to answer a question that is related to the comparison of bremsstrahlung to another plasma diagnostic technique, i.e. Thomson scattering. Thomson scattering provides a well suited tool to infer the plasma temperature and density directly from the scattering spectrum [LGE+ 01, GGL+ 03, GLN+ 07]. In the case of high energy density plasmas, the problem comes up, if the scattered intensity overcomes the bremsstrahlung background level. A detailed analysis of the ratio of Thomson scattering to bremsstrahlung emission over a broad range of plasma densities and temperatures is performed. Results are presented in Ref. [FRR+ 06]. In particular, threshold conditions for the Thomson probe intensity and the spectral resolution of the detector are determined assuming the scenario of a Thomson scattering experiment at the free electron laser at DESY-Hamburg (FLASH) as proposed in Ref. [HBC+ 07]. As a result, bremsstrahlung does not present a major obstacle for such an experiment if the minimum requirements regarding the FEL intensity (I > 1012 W/cm2 ) and detector resolution (/ < 102 ) are met. Another application of the Sommerfeld Gaunt factor is presented in section 4.2. Here, XUV emission spectra from aluminum plasmas are analysed. The plasma was generated by interaction of a solid Al target with intense free electron laser radiation at the FLASH facility. By comparison of the experimental data to the theoretical prediction, the plasma temperature (kB T = 40 eV) and the free electron density (ne = 4 1022 cm3 ) is inferred. These values are in excellent agreement with data from the analysis of the spectral lines which were measured in the same experiment and also with radiation hydrodynamics simulations. These results prove the excellent applicability of bremsstrahlung spectroscopy for diagnostics of high energy density plasmas. In a dense plasma, the successive scattering of the emitting electron on dierent ions becomes important as known from the LPM eect, which was discussed above, see section 1.5. Formally, the account for successive scattering is achieved by resummation of certain subclasses of self-energy diagrams, describing the single scattering process. In Ref. [FRRW05] calculations for the absorption coecient using the full self-energy are compared to the Born-approximation. In order to preserve gauge invariance and conservation of charge, also vertex corrections are performed on the level of the polarization function as dictated by the Ward-Takahashi identities [War50, Tak57]. It is found, that the account for successive collisions in fact leads to a decrease of the absorption spectrum at low photon energies, while at high photon energies, the results converge with the Born approximation. For thermal bremsstrahlung from a dense plasma, this has not been shown before. As a further novelty, an increase of the absorption coecient at intermediate photon energies, is discovered. Both, decrease at low energies and increase at intermediate energies, are in the range of a few percent to a few ten percent compared to the Born approximation, depending on the plasma parameters. Details are discussed in section 4.3 and in the publications [FRRW05] and [FRW07]. Also, partial compensation between self-energy and vertex corrections is observed. The role of vertex corrections is investigated also on the level of the self-energy itself. This is done by comparing the two self-energy diagrams that appear in the second order of the interaction. Also here, partial compensation between the pure self-energy contributions

10

Introduction

and the vertex corrections is obtained. The self-energy decreases by 20% due to the vertex term. A second central point is the consideration of electron-electron correlations. Electronelectron correlations are of a more dynamical nature than electron-ion correlations, due to the large mass of the atomic nucleus compared to the electron mass. Well known eects of electron-electron correlations are for example the dynamical screening of the Coulomb potential and the appearance of collective plasma modes, so-called plasmons. As before, also the electron-electron correlations are included via the electron selfenergy. This is achieved in the framework of the so-called GW approximation, which was introduced for the homogeneous electron gas at zero temperature by Hedin [Hed65], but which can also be applied at nite temperature and also for multicomponent systems [FSK+ 06]. The GW approximation allows the non-perturbative treatment of dynamic electronelectron correlations. The self-energy is calculated beyond the mean-eld level (HartreeFock approximation). Its foundations and various applications will be discussed in detail in section 2.2. The GW self-energy (i.e. the GW (0) -variant, see section 2.4) was calculated for a broad range of plasma densities and temperatures to address a number of key questions: How does the electron spectral function, as the determinant for the optical response and other properties, behave under various plasma conditions? How is the transition from a weakly coupled, dilute system to a degenerate quantum liquid, reected in the spectral function? Under which conditions do collective excitations, visible as peaks in the spectral function, appear? Which are the dominant processes of damping of collective states and of single-particle states? A computer code was developped, that gives the GW (0) self-energy for arbitrary density and temperature. The numerical results for the self-energy and the spectral function will be discussed in sections 4.4 and 4.5. Complementary to the pure numerical treatment, much eort was made to get an analytic solution for the GW (0) self-energy. In this way, deeper insight into the relevant physics behind the mathematical formalism can be gained than from the numerical data. Of course, analytic solutions can only be obtained after performing certain approximations, which are valid only in limited cases of the plasma parameters. Here, I focus on the case of weakly coupled, low density and classical plasmas. This case oers the opportunity to precisely elaborate on the eects of increasing density. Thereby, the inuences of decreasing degeneracy parameter = kB T /EF (see equation (1.2)) and increasing plasma coupling parameter (equation (1.1)) become accessible. At the center of these considerations is the imaginary part of the single-particle self-energy. The imaginary part of the self-energy describes the damping of single-particle states due to the interaction with other plasma particles. Evaluated at the quasi-particle dispersion (see section 4.5 and Refs [For08, For09] for details), it is equal to the inverse of the life time of single-particle states. For free particles, the self-energy tends to zero, and the life time is innite. Naively, one could expect that in weakly coupled plasmas, a series expansion of the selfenergy in terms of the coupling parameter gives a sucient approximation to the exact solution. This is not the case. The perturbative quasi-particle self-energy is far away from

11

the precise numerical solution. Furthermore, it is independent of density [FW74, KKER86], see also Ref. [Le 96]. This means, that the single-particle life time stays nite even in the vacuum, which is an unphysical result. This paradox remained unresolved for a long time. In Refs. [For08] and [For09], it is shown, that a physically sound description of the single-particle life time in a many-particle environment can only be obtained from a nonperturbative ansatz. In the rst of these papers, the quasi-particle self-energy at small momenta is investigated. It is shown, that the Born approximation for the self-consistent self-energy can be solved analytically and the result complies very well with the full numerical solution. The derived expression for the quasi-particle damping or inverse life time is valid in the regime of plasma densities and temperatures where the Debye screening parameter = ne e2 / 0 kB T is smaller than the inverse Bohr radius. In this regime, it behaves proportional to n1/4 . Thus, the particle life time becomes innite in the limit of vanishing density, as one would expect from simple physical arguments. At larger densities, and also at larger momenta, the inuence of collective excitations becomes dominant, which can no longer be accounted for using the Born approximation. The case of large momenta is studied in a second paper [For09]. Here, the relevant damping mechanism is the coupling to longitudinal electron plasma oscillations, the plasmons. Using an appropriate approximation to the screened interaction, i.e. the plasmonpole approximation [KKER86], an analytic solution for the imaginary part of the selfconsistent quasi-particle self-energy could be derived also in this case. At large momentum p, the quasi-particle damping behaves as T /p ln(p2 ), i.e. it decreases very slowly. The low momentum limit and the large momentum limit are nally combined via a twopoint Pad approximation [Mag82], to form an interpolation formula for the quasi-particle e damping width that covers the complete range of momenta from p = 0 to p . The spectral function, computed via the interpolation formula complies with the full numerical solution within an error bar of less than 10%. The validity is limited to densities and temperatures where the Debye screening parameter < 1 a1 . B Having the single-particle spectral function disposable in an analytic form, greatly reduces the numerical cost for calculations of higher order correlation functions that need the spectral function as an input quantity. An examples is the already mentioned dielectric function which is given by a convolution of two spectral functions and the vertex function. The discussion of bremsstrahlung in dense plasmas and associated problems, as outlined above, will be limited to non-relativistic plasmas, i.e. the plasma temperature is small compared to the electron rest mass, kB T me c2 . A generalized linear response approach to relativistic plasmas is given in Refs [MRH02a, MRH02b]. Secondly, the intensity of the external radiation eld is always assumed to be so small that no non-linear eects appear and that no relativistic electrons are generated during the absorption of the elds. To give an order of magnitude, non-linear eects (e.g. birefringence of the vacuum, generation of higher order harmonics, self-focussing of laser beams, etc.) appear at intensities of roughly 1018 W/cm2 [Jac75]. Inverse bremsstrahlung in strong laser elds was recently treated by Brantov et al. [BRS+ 03], see also Bornath et al. [BSHK01]. Furthermore, the plasma is assumed to be of innite extension in each direction; no boundary eects, such as reection, refraction, transition radiation, etc. will be considered. For calculations of

12

Introduction

inverse bremsstrahlung in clusters, see the work by Mulser et al. [MKH05]. Out of the various problems discussed so far, two will be analyzed in more detail in this work: The question is raised, how the nite life time of the scattering particles can be calculated in a consistent way. This leads to the GW (0) approximation for the selfenergy. Using the self-energy, it is subsequently investigated how the account for the nite life time aect the bremsstrahlung emission, i.e. the emission spectrum. Furthermore, the applicability of bremsstrahlung spectroscopy for plasma diagnostic purposes will be demonstrated. The work is organized as follows: After this introductory chapter, the next chapter will give a brief outline of the theoretical framework. After a short discussion of quantum electrodynamics as the basic theory for the interaction of charged particles with electromagnetic radiation, the fundamental aspects of many-particle eld theory will be outlined. Central objects within this approach are the Dyson-Schwinger equations for the particle and eld propagators and their self-energies. As a special approximation and truncation scheme of Dyson-Schwinger equations, the rainbow-ladder approximation will be discussed and the relation to the GW -approximation, known from solid-state theory will be illuminated. Also, a short discussion of the GW approximation in the context of the so-called -derivable theory by Kadano and Baym [BK61, Bay62] will be presented. After a short summary of the Matsubara Green function technique, the GW (0) approximation will be introduced, which is the working horse in this work. The chapter closes with a discussion of the theory of optical properties of charged particle systems in the framework of many-particle eld theory. In chapter 3 the basic processes of interaction between photons and charged particles will be discussed, i.e. emission, absorption, and scattering via various interaction channels. In particular, the role of the various processes for the purpose of plasma diagnostics will be delineated. Also, the problem of radiation transport is presented shortly. In chapter 4, selected problems connected with the many-body eld theoretical description of bremsstrahlung in dense plasmas will be discussed in more detail. The main results of the present thesis are shortly summarized. The detailed presentation of the results can be found in Part II, which contains a compilation of three published, one accepted, and two submitted original articles.

2.1 Quantum Electrodynamics

Quantum Electrodynamics (QED) is todays most precise theory for the behaviour of electrically charged particles (electrons, nucleons, i.e. fermions), electromagnetic elds, i.e. photons, and their mutual interaction. The quantum nature of both massive particles and elds is contained, and the theory is relativistically invariant. Precision measurements e.g. of the 1s 2s transition in atomic hydrogen comply with theoretical predictions obtained from QED up to relative errors of 1017 [H 06]. QED is the most general starting point for a theoretical description of plasmas and their interaction with external electromagnetic elds. The rle of the elds is twofold: On the o one hand, the interaction between the charged particles is mediated by the electromagnetic eld through the exchange of virtual photons. On the other hand, one wishes to study the behaviour of a plasma submitted to an external eld, e.g static electric or magnetic elds or time-dependant elds, such as laser pulses. The emission, absorption, and scattering of radiation by plasmas is of fundamental interest, since these processes and the respective spectra heavily depend on the macroscopic and microscopic state of the plasma, e.g. its density- and temperature distribution, its composition and so forth. Precise measurements of emission- absorption- and scattering spectra therefore are an important tool for plasma diagnostics. The objective of this chapter is to outline the fundamental aspects of QED and show how expressions for emission, absorption and scattering of radiation in a plasma can be obtained.

2.1.1

Lagrangian formalism

A convenient starting point is the Lagrangian density of Quantum Electrodynamics [IZ80, PS95]. It can be written as the sum L(x) = LD (x) + LM (x) + Lint (x) , (2.1)

14

over the Dirac part LD (x) describing the free movement of massive fermions, the Maxwell part LM describing free electromagnetic waves, while Lint is the interaction part, i.e. the coupling between fermions and the electromagnetic elds. Here and in the following, x denotes a four vector, i.e. x = (ct, r), with time t, coordinate vector r = (x, y, z) and the velocity of light in vacuum c = 299792458 ms1 . The Einstein convention for the scalar product of 4-vectors is used throughout this chapter, i.e. x y := 3 ,=0 x g x = x0 y0 x1 y1 x2 y2 x3 y3 , with the standard metric tensor g = g [IZ80]. The dierent parts of the Lagrangian density read LD (x) =

c

c (x) i c mc c2 c (x) ,

0c

c

c (x) and c (x) are the Dirac spinor and its hermitean adjoint for fermions of species c, F is the eld tensor of the electromagnetic eld. It is derived from the four potential A(x) = ((x)/c, A(x)) ((x) is the scalar potential, A(x) is the vector potential) via F (x) = A (x) A (x). The four-current jc (x) = (cc (x), jc (x)) is given by the particle elds (Dirac spinors), jc (x) = cc (x) c (x), the s are the Dirac gamma matrices. Zc is the charge-number of species c, e = 1.602176487 1019 As is the elementary charge and 12 As/Vm is the vacuum electrical permittivity. 0 = 8.854187817 10 From the Lagrangian density, the equations of motion of the particle and radiation elds follow from the principle of extremal action, i.e. the variation of the action S = d4 xL(x) with respect to a dynamical eld vanishes, S = 0. Evaluation of the (functional) variation yields the EulerLagrange equations L L =0. x (/x ) For example, variation with respect to leads to the Dirac equation, (i c mc2 + Zc e A (x))c (x) = 0 , (2.6) (2.5)

and variation with respect to (x) leads to the hermitean adjoint Dirac equation. The variation with respect to A (x) gives ( )A (x) = j (x) + ( A (x)) . (2.7)

The last equation, written for every component A (x) leads to the more familiar Maxwell equations B(r, t) = 0 j(r, t) + 0 E(r, t)/t and E(r, t) = B(r, t)/t(r, t). The remaining Maxwell equations E(r, t) = (r, t)/ 0 and B(r, t) = 0 are equivalent to the denition of the potentials (r, t), and A(r, t): B(r, t) = A(r, t) and E(r, t) =

15

(r, t) A(r, t)/t. These equations are not explicitly time-dependent and therefore they are not obtained through the variational principle. The Lagrange density (2.1) is gauge invariant. Local phase transformations of the Dirac elds (x) (x) exp(ief (x)/ c) and (x) (x) exp(+ief (x)/ c) do not alter the Lagrangian density, if the four potential A (x) is simultaneously transformed as A (x) + f (x). The proof is straightforward by inserting the transformed elds into the Lagrangian density (2.1). The term e f (x) from the dierentiation of (x) exp(ief (x)/ c) is compensated by the term e f (x) from the gauge-transformed four-vector potential. This is a characteristic feature of any gauge eld theory: The synchronous transformation of the particle elds and the interaction elds, or gauge elds, leave the Lagrangian density and thereby the equations of motion unchanged. This argument can also be turned around in the following sense: Suppose the scattering of a Dirac fermion on some scattering center. The scattering induces a phase shift in the fermions wave function. Due to the gauge invariance, this phase shift entails a perturbation of the electromagnetic eld, i.e. the emission or absorption of a photon whilst the scattering. This is precisely bremsstrahlung, or inverse bremsstrahlung in the case of the photon being absorbed. In a similar way, the request for gauge invariance leads to the W and Z bosons in the case of weak interaction and to the gluons as the gauge bosons in quantum chromodynamics (QCD). The gauge function f (x) has to satisfy certain general mathematical requirements which are not to be discussed here. For a detailed presentation of this issue, see Ref. [Wei96]. In a given physical problem, one has to x the gauge in order to obtain meaningful quantitative results. The two most frequently used gauges are the Lorentz gauge A (x) = 0 and the Coulomb gauge A(x) = 0. While the former has the advantage of being relativistically invariant, the latter is more suitable when it comes to formulate a quantization scheme for the Maxwell eld. Furthermore, the Coulomb gauge is convenient, since it separates the electromagnetic eld into a transverse part, the photon eld and a longitudinal part, which gives the Coulomb interaction among charged particles. The Lagrange density (2.1) is Lorentz invariant. It is valid in any special frame of reference related via x = x with the Lorentz boost tensor [LL97a]. In order to perform canonical quantization, a special frame of reference will be introduced. However, the results can be transformed into any frame of reference by means of Lorentz transform.

2.1.2

Hamiltonian

Besides using the principle of extremal action, the equations of motion for the dynamical elds can be derived via the Hamiltonian. The Hamiltonian formalism is more suited for quantization of the theory via introduction of commutation relations among the generalized coordinates and their conjugate momenta, whose equations of motion are given by the Hamiltonian of the system. The Hamiltonian is obtained from the Lagrangian L = d3 rL(x) via the Legendre transformation H = i Pi Qi L, with the generalized coordinates Q and the conjugate momenta Pi = L/ Qi . Note, that in order to dene spatial integration and dierentiation with respect to time, a special frame of reference has to be chosen. Since the only time

16

derivatives in the Lagrange density (2.1) are those of the Dirac elds (x) and the vector part A(x) of the four potential1 , only (x) and A(x) are generalized dynamical coordinates. The corresponding generalized momenta are P (x) = L/ (x) = i 0 = i (x) , and PA (x) = L/ A(x) = 0 (x) + A(x) = 0 E(x) . t (2.8) (2.9)

Applying the Legendre transform with respect to these two dynamical coordinates, the Hamiltonian H(, A; P , PA ) = = d3 r A(x) A L t 2 1 2 0c + mc2 ec A + ( A)2 + P 2 20 A d3 r i +

0

(x) +

P 0 i c i P 0 = d3 r i c i

(2.10)

+ mc2 ec A +

c2 B + E 2

is obtained. The rst term is recognized as the Hamiltonian for a Dirac fermion in an electromagnetic eld, whereas the second term is the energy density of the electromagnetic eld 0 (c2 B2 + E2 )/2. Two remarks should be made at this point: In the derivation of the above equation, Gauss law E = / 0 , which is one of Maxwells equations, has been used. However, this does not aect our scheme of deriving the dynamical equations of motion from the Hamiltonian, since Gauss law does not represent such a dynamical equation, i.e. it is time independent and simply reects the conservation of charge. Secondly, it was used, that the term (PA ) vanishes when the volume integral d3 r is performed, which is again a consequence of charge conservation. Equation (2.10) is the Hamiltonian of relativistic quantum mechanics. The elds , A and their conjugate momenta are not quantized. The equations of motion for the generalized coordinates Q lead to the Dirac equation and the Maxwell equation. The solution of the Dirac equation gives the wave function for a Dirac Fermion in a classical electromagnetic eld, which is itself governed by Maxwells equations. The next step consists in the quantization of the Dirac elds and the Maxwell eld. In this way, eects like particle production, the spontaneous decay of excited atoms, or the vacuum polarization become accessible[Gro93]. In the following section the quantization of both the Dirac and the Maxwell eld will be delineated. As already mentioned, this task requires to x a system of reference for the denition of commutation relations among the eld operators. Secondly, the Maxwell eld has to be gauged. Here, the Coulomb gauge will be applied. This choice is advantageous

In the Lagrange density (2.1) only the term /t appears, not /t. The time derivative of the scalar potential does not appear as can be seen by expanding the Maxwell part of the Lagrangian using the denition of the Maxwell eld tensor.

1

17

in a twofold sense: First, canonical quantization is more simple to perform, since already a special frame of reference had to be chosen in order to x the Coulomb gauge itself, since the latter involves spatial derivatives. Second, in the theory of electron-ion plasmas, the ion subsystem denes a frame of reference, since ions barely move due to their comparatively large mass. Thus, the laboratory frame can be chosen as the one in which the Coulomb gauge is xed and where the quantization is dened. The Maxwell eld A(r, t) is transverse in the Coulomb gauge, i.e. the longitudinal component does not represent a dynamical degree of freedom. Consequently, the longitudinal part of the generalized momentum PA (r, t) has to be eliminated as a dynamical degree of freedom. This is done by decomposing PA (r, t) = PA (r, t) + PA (r, t) and identifying E (r, t) = PA / 0 with the Coulomb eld, satisfying Gauss law E (r, t) = / 0 . The Hamiltonian (2.10) in Coulomb gauge is rewritten as H= d3 r P 0 i c i + mc2 ec A + e c2 B + E2 + 2 2

0

(2.11)

with the scalar potential (x). The scalar potential does not represent a dynamical degree of freedom, since it is related in an instantaneous manner to the charge density e(x). It can be given completely in terms of the Dirac elds, i.e. the charge density e(x) = e (x)(x), (x) = e 2 d3 x (x ) . 4 0 |r r | E = 0. (2.12)

2.1.3

In order to quantize the Hamiltonian (2.11), commutation relations have to be imposed on the eld operators (x), P (x), A(x), and PA (x), which replace the generalized coordinates and conjugate momenta. The eld operators are usually represented as a Fourier series over creation and annihilation operators of eld-modes. The creation and annihilation operators are particularly suited for the formulation of perturbation theory and for the calculation of cross-sections, e.g. the reaction of particles among themselves and their coupling to external electromagnetic elds.

2.1.4

In Coulomb gauge, A = 0, the transversality of the Maxwell eld has to be reected also in the commutation relation. Therefore, the commutation relation between A and PA reads

j j j Ai (x), PA (x ) = Ai (x)PA (x ) PA (x )Ai (x) = i ij (r r )(t t ) ,

2

(2.13)

For other quantization schemes, e.g. using Lorentz gauge, see Ref. [Reb05].

18

ij (r r ) =

d3 k ik(rr ) ki kj ij 2 e 3 (2) k

(2.14)

The transverse delta function is solenoidal, i.e. the vector potential reads A(x) = 0

0

d3 k 2k (2)3

=1

(2.15)

where the annihilation and creation operators b (k) and b (k) for a photon of wavevector k = (k /c, k), k k = 0 and polarization were introduced. The transversality of A k imposes the conditions k = 0 , k = , k k on the polarization vectors . Inversion of Eq. (2.15) leads to k b (k) = b (k) = i 3 2 (2) k 0 0 i (2)3 2k0 0 0

(2.16)

d3 x eikx 0 A(x) , k d xe

3 ikx

0 k

A(x) ,

(2.17)

where we used the short-hand notation f (x) g(x) = f (x)g(x)(f (x))g(x). By using the denition (2.15), the commutator between creation and annihilation operators is evaluated to b (k), b (k ) = (k k ) , [b (k), b (k )] = b (k), b (k ) = 0 . (2.18)

In order to prepare the next section, were the perturbation expansion for QED will be outlined, I also give the propagator for the transverse photon eld. The transverse photon propagator is dened as

Dij (x, x ) =

1 0| TB Ai (x)Aj (x ) |0 , i c

(2.19)

where the eld modes satisfy the Coulomb gauge condition. The propagator is the expectation value in the ground state |0 of the time-ordered product of two eld operators at dierent coordinates. The time-ordered product for Bose particles is dened as TB {A(x)B(x )} = (t t )A(x)B(x) + (t t)B(x )A(x). The propagator Dij (x, x )

19

can most easily be evaluated in momentum space, using the plane-wave decomposition Eq. (2.15). One obtains

Dij (x x ) = 2 Dij (k)

=1

(2.20) (2.21)

k2

1 , 0. + i

Note, that the propagator depends only on the dierence x x as a consequence of spacetime homogeneity.

2.1.5

In order to quantize the Dirac elds (x) and P (x) = i (x), anticommutators instead of commutators have to be used to reect the Fermi-Dirac statistics of spin 1/2 particles. In this way, the total energy of the Dirac eld is positive denite, using commutators would lead to unphysical negative energies. The anticommutator relations imposed on the Dirac eld operators are (x), P (x ) = (x)P (x ) + P (x ) (x) = (r r ) , (x), (x ) = P (x), P (x ) = 0 . (2.22)

The index denotes the spinor component. The plane-wave decomposition for the Dirac elds reads 4 mc2 d3 p (x) = a (p)u (p)e px , 3 Ep (2) =1 (2.23) 4 mc2 d3 p 0 u P (x) = i a (p) (p)e px , 3 Ep (2) =1 with = 1 for = 1, 2 and = 1 for = 3, 4. a (p) is an annihilation operator for fermions ( = 1, 2) in momentum state p and a creation operator for antifermions ( = 3, 4), while a (p) acts as a creation operator for fermions and annihilation operator for antifermions. Antifermions are solutions of Diracs equation with opposite charge as fermions and travelling backwards in time. The restmass m > 0 is the same for fermions and for antifermions. u (p) is the amplitude of the Fourier component for momentum p, i.e. the solution of the free Dirac equation (i c p mc2 )u (p) = 0. For the creation and annihilation operators the anticommutation relations a (p), a (p ) = (p p ) , {a (p), a (p )} = a (p), a (p ) = 0 , (2.24)

20

The time-ordered product for fermions TF {A(x)B(x)} = (tt )(x)(x )(t t)(x )(x) correctly reects the Fermi-Dirac statistics i.e. the anticommutation relations instead of the commutation relations as in the case of the Bose time-ordered product. Using Eq. (2.23) the free fermion propagator is obtained as S(x x ) = d4 p S(p) exp(ip(x x ) (2)4 + mc S(p) = 2 , 0. p m2 c2 + i (2.26) (2.27)

Note, that S(x x ) is a 4x4 matrix acting in the Dirac spinor space.

2.1.6

Using the quantized expressions for the Maxwell eld Eq. (2.15) and the Dirac elds Eq. (2.23), the Hamiltonian of QED is obtained as

4

H = 0

=1

4 2

emc

0

, =1 =1

5/2

d3 p (2)3 Ep

+

4

1 + 0 2 , =1

0 is the normalization volume, Ep is the fermion energy Ep = + m2 c4 + p2 c2 . This Hamiltonian features an innite contribution due to the zero-point energy d3 k k /2(2)3 of the photon eld. However, this contribution is treated as a constant (albeit innite), additive term to the total energy operator, which is not observable, since only energy differences can be measured. However, it should be noted that under special experimental arrangements, the zero-point energy of the electromagnetic eld has an observable eect. One such example is the Casimir eect [Cas48, PMG86], which describes the attractive

21

force between two uncharged metal plates due to small dierence in their respective vacuum energies. On the other hand, the well-known and every-day observation of the spontaneous radiative decay of excited atoms or molecules can only be understood as a result of the interaction of the atom with the vacuum uctuations of the electromagnetic eld [Mes99]. Also the free Fermion part (rst term) contains an innite contribution. This can be seen, by expanding the sum over Dirac components as

4

0

=1 2

4

0

=1

d3 p Ep . (2.29) (2)3

The rst two terms are the kinetic energies of the fermion and the antifermion subsystem, respectively (note that a3/4 is an antifermion creation operator), the third term stems from the antifermion commutation relation. As in the case of the zero-point energy of the photon eld, this constant but innite contribution is subtracted, since not contributing to observable eects. To avoid the appearance of innite energy contributions, it is common use in quantum eld theory, to write the Hamiltonian in normal order. Operators with the phase factor exp(it) alway appear left from those whose time dependence is given by exp(+it). In this way, the innite terms in the Hamiltonian are omitted right from the beginning, which means that only the physical masses appear. The normal ordering of two operators A and B is denoted by : AB :. The following identity holds AB =: AB : 0|T AB |0 , (2.30)

where the negative sign appears only if the operators are fermionic creation- or annihilation operators. The third and fourth term of the Hamiltonian describe the coupling of the Dirac fermions to the transverse photon eld (third term) and their mutual Coulomb interaction (fourth term). The third term exhibits the typical structure of QED processes, where the interaction of fermions and photons is described as the creation and annihilation of fermions in dierent energy- and momentum states and the creation or annihilation of photons with matching energy and momentum, such that the total four-momentum is conserved. The coupling term is the starting point for the formulation of perturbation theory, which leads to the cross-sections for the various interaction channels in QED.

22

2.1.7

Non-relativistic limit

Since typical energies considered in this work are non-relativistic, I should give the nonrelativistic limit of the QED Hamiltonian, as well, H = 0

s

2

0

=1

j(k) k

0 2 0 k

1/2

b (k) + b (k)

1 + 3 2 0

ss

Here, the non-relativistic single-particle energy of particles of species s, Es (p) = 2 p2 /2ms was introduced. Note, that the index s labels dierent species and further quantum numbers such as the particles spin. Only particles, no antiparticles appear in the non-relativistic Hamiltonian, since the typical energy scales (given e.g. by the plasma temperature) are small compared to twice the rest mass of the particles, kB T 2mc c2 . Only at this energy scale, pair production sets in. The current density operator j(k) is dened as j(k) =

s

(2.32)

2.1.8

Perturbation theory

As already mentioned, the objective of QED is the calculation of cross sections for various processes involving the emission, absorption, and scattering of photons by electrically charged particles. To this purpose, a perturbation series known as the S-matrix expansion is developed in terms of the interaction part of the Hamiltonian (2.28), i.e. the coupling of electrons to transverse photons and the Coulomb interaction. The S-matrix expansion is most conveniently treated in the interaction picture or Dirac picture. In the Dirac pic ture, the time-evolution of an observable O due to the interaction part of the Hamiltonian Hint = H H0 is formally eliminated via the unitary transform OD (t) = eiH0 t/ OS eiH0 t/ . (2.33) OS is the observable in the Schrdinger picture. One can easily show that the equation of o motion for the observable in the Dirac picture involves only the non-interacting Hamiltonian, i.e. i D O (t) = H0 , OD (t) . (2.34) t On the other hand, the state vectors in the Dirac picture D (t) = eiH0 t/ |S (t) obey a wave equation which involves only the interaction part of the Hamiltonian, i D D (t) = Hint (t) D (t) . t (2.35)

2.1 Quantum Electrodynamics This equation can formally be integrated as D (t) = D (t0 ) + 1 i

t t0 D dt1 Hint (t1 ) D (t1 ) .

23

(2.36)

D (t0 ) is the quantum state at a given initial time t0 , which is characterized by a vanishing interaction part of the Hamiltonian. To eliminate any ambiguity in the choice of t0 , the limit t0 is considered and the state D (t ) is denoted the initial state |i . Since the interaction part of the Hamiltonian is assumed to vanish at t , the initial state in the Dirac picture is equal to the initial state in the Schrdinger picture. By o successive iteration of Eq. (2.36) on arrives at the Dyson series for the wave function, 1 (t) = |i + i

D t D dt1 Hint (t1 ) |i

(2.37) The S-matrix is dened as the transition amplitude (t )|i . Its matrix elements Sf i = f | S| i measure the probability amplitude that in the limit t the quantum system is in the nal state |f . As for the initial state, the interaction is assumed to vanish for t and the Dirac- and Schrdinger picture are equivalent. By projecting the Dyson o series (2.37) on the nal state f |, the S-matrix elements can be calculated iteratively. In compact form, using the time-ordered product T , which sorts its arguments such that late times stand to the left of early times, one obtains

D

1 + (i )2

t1

Sf i = f | (t ) = f |

n=0

1 i T

1 n!

dt1

dt2 . . .

dtn1

dtn

The factor 1/n! is the correct normalization to the number of possible permutations of the operators in the time-ordered product. Note, that this equation is only valid, if the number of fermion creation and annihilation operators is even. In QED, this is indeed the case as becomes clear from the Hamiltonian (2.28). The evaluation of the perturbation expansion of the S-matrix directly gives the transition amplitudes for various QED processes, where a transition from the initial state |i to the nal state |f appears. The evaluation of the rst order is simple. Since only one time argument appears, the time-ordering T has no eect. The interaction term with transverse photons gives eight terms, which correspond to photon emission and absorption by fermions and by antifermions (4 terms) and fermion-antifermion creation (annihilation) together with photon creation (annihilation) (4 terms). However, non of these processes is a real physical process, since energy- and momentum cannot be conserved simultaneously. In higher orders, the emerging terms contain a large number of products of creation- and annihilation operators for photons and fermions. Out of these many products only those give a nite contribution to the amplitude, which contain the right number of annihilation operators to destroy all particles in the initial state and the right number of creation operators to create all particles in the nal state.

24

Wick [Wic50] showed that the time-ordered product of N normal ordered eld operators, as is the case in the interaction Hamiltonian, can be written as a sum over products of M propagators and N 2M external operators and the sum extends from M = 0 to M = N/2. The evaluation of the second order of the perturbation series will be given as an example. In second order of the S-matrix expansion, terms appear which involve the interaction of charged particles with free, transverse photons. Due to Wick, the time-ordered product in the second-order contribution to the S-matrix S (2) = e2 2(i )2 d 4 x1 d 4 x2 T (x1 ) (x1 )A (x1 )(x2 ) (x2 )A (x2 ) (2.39)

can be decomposed into the sum S (2) = e2 d4 x1 d4 x2 2(i )2 : (x1 ) (x1 )A (x1 )(x2 ) (x2 )A (x2 ) : + 2 : (x1 ) iS(x2 x1 ) (x2 )A (x1 )A (x2 ) : + : (x ) (x )(x ) (x )iD (x x ) :

1 1 2 2 2 1 + 2 : (x1 ) iS(x2 x1 ) (x2 )iD (x2 x1 ) : + : iS(x2 x1 ) iS(x1 x2 ) A (x1 )A (x2 ) : + : iS(x2 x1 ) iS(x1 x2 ) iD (x2 x1 ) :

(2.40)

These terms can most conveniently be depicted by Feynman diagrams [Vel94], which allow for a very intuitive interpretation of the dierent terms and to sort out those terms which describe a special process. The rst term is again vanishing since it describes two parallel processes of rst order. In the second term, one nds the Compton scattering of photons on electrons (positrons) and the process of pair production by two-photon decay and pair annihilation by two-photon production. These processes are depicted by the Feynman graphs in Fig. 2.1 Whereas the second term contains the interaction with an external transverse eld, the third term contains the fermion-fermion scattering (Mller scattering) and fermionantifermion scattering (Bhabha scattering). It is depicted by the Feynman graphs in Fig. 2.2 In the fourth term in (2.40) appears what is called the one-loop fermion self-energy, i.e. the product of one fermion propagator and one photon propagator. The one-loop self-energy gives the rst order correction to the bare fermion propagator. The fth term contains the photon self-energy or polarization function, i.e. the product of two fermion propagators. The self-energy corrections to the bare propagators are depicted by the Feynman diagrams in Fig. 2.3 The evaluation of the one-loop fermion self-energy leads to a divergent integral. Similar to the vacuum uctuations of the Maxwell eld, the electron self-energy is treated as a

25

(d) Fermion Compton scattering (e) Antifermion Compton scat(exchange term) tering (exchange term)

Figure 2.1: Feynman diagrams of fermion(antifermion)-photon interaction constant, innite contribution to the electrons rest mass. To obtain the correct physical mass of the electron, the so-called bare mass, i.e. the mass without self-energy corrections, is itself xed a innity, such that the sum of the bare mass and the self-energy gives the physical mass. The mathematical rendition of this concept is known as renormalization [Kak93]. Figure 2.3(c) gives the one-loop photon self-energy or polarization function [Gro93]. Finally, the last term is the so-called vacuum contribution ground state energy, see Fig. 2.4 It can be shown, that this graph belongs to a series of terms that constitutes a phase factor of the total S-matrix and therefore has no observable eect. On the other hand, these completely closed vacuum graphs are the starting point for the path-integral formulation of quantum eld theory [GR96, FH65]. In a many-particle system, not only the transverse photons contribute to the interaction part of the Hamiltonian, but also the Coulomb interaction has to be considered. In principle, all the graphs in Figs. 2.1, 2.2, 2.3, and 2.4 also exist with the transverse photons replaced by the longitudinal Coulomb potential. For example, the diagram in

26

Figure 2.3: Fermion and photon one-loop self-energy corrections gure 2.2(a), with the transverse photon replaced by the longitudinal Coulomb potential, describes the elastic scattering among two fermions. Lets consider the case when one of the scattering partners is an ion and the other one is an electron. Due to its heavy mass, the ion can be treated as xed in the laboratory frame of reference. Evaluation of the transition amplitude, averaging over all initial spin polarizations and summation over all nal spin polarizations leads to the elastic scattering cross section d d =

elastic

(2.41)

Here, the nestructure constant = e2 /4 0 c 1/137 was introduced. In the nonrelativistic limit, this expression reduces to the well-known Rutherford cross-section. One gets the same result for elastic positron scattering [GR84]. Further terms in second order of the Coulomb interaction are the self-energies due to the Coulomb self-interaction, and the screening of the Coulomb interaction as well as the vacuum contribution of the Coulomb eld. The mixed terms, where one Coulomb interaction and one transverse photon appear in the Feynman diagram are shown in Fig. 2.5. They describe the emission of a photon whilst the scattering of a fermion in the Coulomb potential of an ion, so-called bremsstrahlung emission.

27

Out of the many processes that are obtained in second order of the S-matrix expansion, four are of special importance also for plasma physics. For the interaction with external elds these are Compton scattering and bremsstrahlung and for the internal elds the electron and photon self-energy. Compton scattering describes the scattering of electromagnetic waves on electrons. As will be discussed later, the scattering spectrum is sensitive to the macroscopic properties of the plasma, i.e. temperature, density, and the equation of state. Therefore, scattering of electromagnetic waves is a suitable tool for plasma diagnostics. On the other hand, bremsstrahlung is the most dominant mechanism of absorption and emission in a hot, highly ionized plasma. Also in this case, details in the emission or absorption spectrum can give valuable information about the plasmas state. In a many-particle environment, the electron and the photon self-energy give the modications of single-particle properties due to correlations. An example is the change of the energy-momentum relation for massive particles due to the eective mass. The effective mass is given by the fermion self-energy. Similarly, the propagation of photons in the medium, commonly described via the dielectric function, is determined by the photon self-energy, i.e. the polarization function.

28

2.1.9

Bremsstrahlung is understood as the emission of a photon by a charged particle which is submitted to an external eld. In the case of the external eld being a magnetic eld, the emitted radiation is usually called synchrotron radiation, while the term bremsstrahlung is mostly associated with emission of radiation by electrons in the Coulomb eld of other charged particles. In the most simple case, an electron scatters on an ion, which is assumed not to move during the process and whose nite extension is neglected, such that its eld can be taken as a pure Coulomb eld. Evaluation of the Feynman diagrams in Fig. 2.5 gives the well known Bethe-Heitler formula for bremsstrahlung [BH34]. This approximation will be referred to as Born approximation in the following. Note, that the presence of the scattering center (the ion) breaks the translational invariance, and therefore conservation of three-momentum is no longer required. The Bethe-Heitler cross section reads [IZ80] |pf | Z 2 3 d3 = 2 dk0 d de (2) |pi | |k|4 k0

|pi |2 sin2 i |pf |2 sin2 f 2 0 2 (4(pi ) |k| ) + 0 (4(p0 )2 |k|2 ) f 0 2 2 (pf |pf | cos f ) (pi |pi | cos i ) |pi | |pf | sin i sin f cos if |pi |2 sin2 i + |pf |2 sin2 f 2 0 0 0 (pf |pf | cos f )(pi |pi | sin i ) (pf |pf | cos f )(p0 |pi | cos i ) i (4p0 p0 |k|2 + 2 2 ) . (2.42) i f

2 2k0

f and i are the angles enclosed by the photons momentum k and the electrons initial and nal momentum pi and pf , respectively. if is the angle enclosed by the planes spanned by k and pi , and pf , respectively. Two important limiting cases should be discussed at this point. In the limit of vanishing photon energy k0 = 0 one obtains d3 = lim k0 0 dk0 de d d de

d3 k |pf |2 sin2 f |pi |2 sin2 i e2 + 0 0 2k0 (2)3 (k 0 p0 |k| |pf | cos f )2 (k pi |k| |pi | cos i )2 f |pf | |pi | sin f sin i cos if 2 0 0 . (2.43) (k pi |k| |pi | cos i )(k 0 p0 |k| |pf | cos f ) f The prefactor is the elastic scattering cross section Eq. (2.41). Eq. (2.43) shows that the Bethe-Heitler cross section diverges as 1/k0 = 1/ when the photon energy goes to 0. This is known as the infrared divergence of bremsstrahlung. However, the cross section is not a directly measurable quantity. In a scattering experiment, one measures the radiation intensity, i.e. the radiated power per unit area of the detection device. It

elastic

29

is proportional to the product of the cross section and the photons energy and therefore tends towards a nite value at small photon energy. Another way of discussing the infrared divergence of the bremsstrahlung cross section takes into account terms in the third order of the S-matrix expansion: Assuming that in an experiment we want to measure the

Figure 2.6: Radiative corrections of the elastic scattering amplitude bremsstrahlung spectrum with a spectrometer of spectral bandwidth . For photon frequencies < , we cannot distinguish between elastically scattered electrons and those which have emitted a bremsstrahlung photon. Therefore, in the theory, both elastic and inelastic scattering amplitudes have to be added coherently up to the same order in QED . Since the bremsstrahlung cross section is one order of QED higher than the elastic scattering cross section, also higher order contributions to the elastic scattering amplitude have to be considered, so-called radiative corrections. The corresponding diagrams are shown in Fig. 2.6. It can be shown, that these diagrams are also infrared divergent and that the divergent terms exactly cancel with those terms appearing in the bremsstrahlung amplitude. me c. Then, The second limiting case discussed here is the non-relativistic case, p0 i the bremsstrahlung cross section turns into Ei + Ei 16 Z 2 e2 e2 c2 d , (2.44) = 2 ln 2 d nr 3 4 0 c 4 0 me c2 vi vi is the initial velocity. Again, this expression is infrared divergent.

2.1.10

Quantum eld theory led us to an expression for the bremsstrahlung cross-section, based on second order perturbation theory, the Bethe-Heitler formula (2.42). Historically, expressions for the emission of electromagnetic radiation from a point charge, accelerated in the Coulomb eld of an ion, where rst given by Kramers [Kra23]. He used Maxwells theory of electromagnetism and assumed classical trajectories of the emitting particles: The radiated energy per frequency interval dW ()/d of emission from accelerated point charges is given by Larmors law, dW () 2 e2 = |a()|2 , d 3c3 4 0 (2.45)

30

with the Fourier components of the acceleration a() = (t) exp(it)dt. Thus, the r knowledge of the particles trajectory r(t) allows the calculation of the power spectrum. Unbound electrons in a Coulomb eld follow hyperbolic trajectories, i.e. the distance r(t) between the ion and the electron as a function of time is given by r(t) = b2 b (1 + 1 + (b/b )2

1/2

cos (t))

(2.46)

with the impact parameter b, (t) being the angle between the initial velocity vector vi = v(t ) and the velocity at time t, v(t); and the impact parameter for 90 scattering 2 b = Ze2 /4 0 me v0 . In general, one is interested in the power spectrum, i.e. in the radiated energy per unit time. This is obtained by integrating the radiated energy spectrum over all scattering angles and impact parameters and multiplying by the density of ions ni and the velocity of the incoming electrons, i.e. dP (; v0 ) = ni v0 2 d

0

db b

dW (; b) . d

(2.47)

The integrations can be performed analytically [LL97b], which leads to the expression 16 Z 2 ni dP (; v0 ) = d 3 3 m2 c3 v0 e e2 4 0

3

G() .

(2.48)

Here, the frequency dependent Gaunt factor G() was introduced, 3 iZe2 (1) (1) G() = uHu (u) Hu (u) , u = , 3 4 4 0 me v0

(1) (1)

(2.49)

with the Hankel function Hn (z) of rst kind and of order n. The derivative Hn (z) of the Hankel function is given by

(1) Hn

(1)

(2.50)

The Gaunt factor has the special property, that for large dimensionless frequencies u, u 1, it converges towards unity, G() = G(u()) 1 [Miy80]. Thus, Kramers formula for the bremsstrahlung power spectrum predicts a frequency independent behaviour, if either the frequency becomes large, or if the initial velocity becomes small. Immediately, this behaviour leads to an ultraviolet catastrophe, since the total radiated energy, W = ddP (/d) diverges at the upper integration limit. This is due to the fact that the 0 quantum nature of electromagnetic waves has been neglected. A natural upper cuto for the frequency integration is introduced, if the photon energy is accounted for in the 2 conservation of energy during the emission, i.e. me v0 /2 = me v 2 (t )/2 + . Since the

31

energy of the emitted photons cannot overcome the initial kinetic energy of the scattering 2 electrons, the spectrum terminates at = mv0 /2 . When the theory of bremsstrahlung, as sketched here, was developed, the calculation of Hankel functions was considerably complicated. Therefore, various approximations for the Gaunt factor have been developed that lead to more simpler expressions and which are suitable in dierent regions of the parameter space spanned by the initial velocity, the frequency and the charge of the central ion. A summary can be found in the review article by Karzas and Latter [KL61]. Nowadays, the term Gaunt factor is used in a more general context. It is common use to normalize calculations of the bremsstrahlung power spectrum to the frequency independent part of Kramers formula (2.48) and refer to the remaining factor as the Gaunt factor, that describes corrections beyond the high frequency limit of Kramers classical formula [BH62, KM59]. Such corrections take into account e.g. the quantum nature of the emitting particles. In principle, the classical trajectories are replaced by Coulomb wavefunctions [Som49]. The resulting expression, named Sommerfeld Gaunt factor, reads

d d |F (if , i)i, 1, )|2

G() = 3

(1 exp(2f ))(exp(2i ) 1)

, i/f = Z

2Ryme /pi/f

(2.51)

with the complete hypergeometric function F (, , ; z). The derivative is given by dF (, , ; z)/dz = F ( + 1, + 1, + 1, z). Initial and nal momentum are related by conservation of momentum, |pi pf | = /c. In the situation when both incoming and outgoing electron energies are large compared to the potential energy at the point of closest approach, i.e. f , i 1, the derivative of the complete hypergeometric function can be approximated by F (1, 1, 1, ) f i F (1, 1, 2, ) = f i ln(1 )/ and the complete hypergeometric function itself tends to F (0, 0, 1, ) = 1. Then, we nd again the Born approximation for the non-relativistic case, see Eq. (2.44) . The Bethe Heitler formula (2.42) represents the Gaunt factor for relativistic electronion bremsstrahlung. The Born-Elwert approximation [Elw39] represents an improved nonrelativistic Born approximation, that reproduces the full Sommerfeld result up to photon energies only slightly below the endpoint of the spectrum. Also in the case of bremsstrahlung emission from a plasma, the term Gaunt factor is used. Here, it describes corrections due to many-particle eects, such as screening of the Coulomb potential [WMR04], multiple scattering [WMR+ 01], correlations among the scattering centers [FRRW05], and correlations among the scattering electrons [FRW07], which go beyond the classical Kramers result, averaged over the distribution function of the emitting particles in the medium. Expressions for the bremsstrahlung emission from a plasma will be given in section 2.5. In principle, these are obtained by averaging the bremsstrahlung cross section for the individual scattering event with the electron momentum distribution function.

32

2.1.11

The cross section for scattering of polarized photons on a single electron (Compton scattering) is obtained by evaluation of the Feynman diagrams 2.1(a) and 2.1(d). One obtains the Klein-Nishina formula [KN29], i (1 cos ) . m e c2 KN (2.52) The second equation reects conservation of 4-momentum, i/f is the frequency of the photon before and after the scattering, respectively, i/f are the initial and nal polarization vectors. is the solid angle the photon scatters into. Furthermore, re denotes the so-called classical electron radius, re = e2 /4 0 me c2 . Summing over all initial and nal polarizations yields the unpolarized cross-section, = f = i 1 + d d =

KN 2 re 2

d d

2 re 2

f i

f i + + 4(f i )2 2 , i f

f i

f i + sin2 i f

(2.53)

2 re d = (1 cos2 sin2 ) . d 2

is found. As already noted, Compton or Thomson scattering play an important rle in o plasma diagnostics. First, scattering on uncorrelated electrons results in a Doppler broadened scattering spectrum due to the thermal movement of the electrons. The Doppler width is related to the temperature of the system or the Fermi energy in the case of Fermi degenerate systems. Secondly, in correlated electron systems, such as dense plasmas, the scattering takes place on collective excitations (plasmons, phonons), which show up as peaks in the scattering spectrum. Their position and amplitude are also directly related to the systems temperature and density [Thi07]. The role of Compton scattering or Thomson scattering as a tool for plasma diagnostic purposes will be discussed in more detail in section 3.5. As a nal remark, it should be noted, that both bremsstrahlung and Compton scattering can be understood as the scattering of photons on electrons. In the case of bremsstrahlung, one of the transverse photons, present in the Compton scattering diagrams, is replaced by a virtual photon, the longitudinal Coulomb interaction. Both processes are contained in the 4x4 Klein-Nishina matrix, which gives the scattering cross section of polarized photons on electrons. The 0-components of this matrix give the bremsstrahlung cross-section, while the other components give the true Compton scattering cross-sections, accounting for the polarization of the incoming and the outgoing photon.

33

2.2

2.2.1

Dyson-Schwinger equations

Perturbative QED, as outlined section 2.1, provides a systematic and powerful tool to calculate cross-sections for the various interaction processes between the radiation eld (gauge bosons) and electrically charged particles (fermions). Examples where given in sections 2.1.9 and 2.1.11 for bremsstrahlung and Compton scattering, respectively. However, there also exist phenomena that cannot be described by perturbation theory. Examples are the occurrence of bound states, e.g. atoms, or the appearance of collective modes in a correlated many-particle system, such as plasmons (collective plasma oscillations), phonons (collective modes of crystal lattices) or giant-resonances in atomic nuclei. To account for these kind of phenomena, non-perturbative methods have to be applied. These methods can be regarded as an extension of perturbation theory in the sense that certain sub-classes of diagrams are summed up to innite order, and thereby give a solution to the underlying equations of motion, e.g. of the particle propagator, that also contains the non-perturbative eects. Examples will be discussed below. Formally, the many-particle problem can be cast into a hierarchy of non-linear coupled integro-dierential equations, known as the Dyson-Schwinger equations [Dys49, Sch51b, Sch51c]. This approach was originally formulated for the vacuum, i.e. at zero temperature and zero density. Its generalization to nite temperature and nite density was accomplished more recently, see e.g. the review by Roberts and Schmidt [RS00] and references therein. An excellent introduction to Dyson-Schwinger equations can also be found in [HRW06]. We will discuss the lowest order Dyson-Schwinger equations. These are the equations for the two-point functions, i.e. the single-particle Green function G and the photon propagator D and their respective self-energies, the single-particle self-energy and the polarization function . A detailed derivation can be found in Ref. [HRW06]. These equations are represented by the following Feynman diagrams

G

G(0)

D

G(0)

(2.55)

= (0) =

,

D

(2.56) , (2.57)

D (0)

G

D (0)

(0) G

(2.58)

34

Note, that the wiggly lines denote full photon propagators. In section 2.1, the wiggly lines where used to denote free photons, which are now represented by the dashed lines. Furthermore, the full vertex is introduced as the solid dot. It will be discussed further below. Equation (2.55) is the Dyson-Schwinger equation for the single-particle Green function, which involves the self-energy as the driving term or inhomogeneity. The self-energy describes the inuence of many-particle correlations on the behaviour of the fermions. Observable eects are the nite life time of single-particle states in the medium (contained in the imaginary part of the self-energy) and modications of the dispersion relation, expressed via the real part of . Equation (2.56) is the Dyson-Schwinger equation for the fermion self-energy itself. It describes that correlations enter the self-energy via the full photon propagator D, see equation (2.56). The fermion, being electrically charged, polarizes the surrounding medium (or vacuum) via emission of a virtual photon and the polarized medium acts back on the fermion. The virtual photon, as well as real photons, obey the Dyson-Schwinger equation (2.57) (rigorously, equation (2.57) contains a summation over the polarization states of the photon, which is omitted here to keep the notation short. See also section 2.1.3 for a complete notation and remarks concerning the gauge of the photon eld). Here, the polarization function plays the same role as the fermion self-energy for the fermion Green function G. Hence, each equation is directly or indirectly connected to each of the other three equations, yet the system is not closed. The vertex function , which describes the eective fermion-photon coupling constant, obeys another integral equation, which involves the two-particle eective interaction K, i.e. a 4-point function. The vertex equation can be depicted as

(0) +

(2.59)

The label G at the full fermion Green functions has been omitted for convenience. The Dyson-Schwinger equations represent an exact, non-perturbative quantum-eld theoretical approach to many-particle systems. In fact, the quantum nature of both the massive matter particles (fermions) as well as of the gauge bosons and the external elds are accounted for. On the other hand, many-particle correlations are consistently integrated. Albeit having been developed originally for QED, also other quantum eld theories, such as the electroweak theory and quantum chromodynamics are formulated within the same framework [RS00]. However, in order to perform practical calculations e.g. of experimentally accessible observables, Dyson-Schwinger equations themselves are only of limited use. One has close the Dyson-Schwinger hierarchy. This means that in order to determine Dyson-Schwinger functions up to an order n by solving the corresponding Dyson-Schwinger equations, the n + 1-point function either has to be parametrized by using an appropriate ansatz or

35

neglected altogether. This procedure is similar to the virial expansion in statistical thermodynamics [AP92] and the introduction of closure relations in the BBGKY hierarchy of kinetic equations [ZMR96b].

2.2.2

The rst non-trivial closure of the Dyson-Schwinger hierarchy consists in neglecting higher order term in the vertex function , i.e. setting (12, 3) = (1, 2)(2, 3). In this way, a closed system of equations for the two-point functions G and D and their respective self-energies is obtained. In quantum eld theory, notably in QCD, this approximation is known as the rainbow-ladder approximation [RW94]. The rainbow-ladder approximation belongs to the class of so-called -derivable approximations introduced by Baym and Kadano [BK61, Bay62]. The theory of Baym and Kadano states, that an approximation to the self-energy is energy, momentum, and charge conserving, if and only if there exists a functional , such that the self-energy can be written as the functional derivative = [G]/G. The -functional is obtained as the sum of all closed two-particle irreducible diagrams, i.e. those diagrams, that cannot be separated into two disconnected graphs by removing two fermion lines. For the rainbow-ladder approximation, the functional is given by the following sum RL = rainbow + ladder

+ ...

+ ... (2.60)

Dierentiation with respect to a Green function G corresponds to the removal of any one

36

+ ...

(2.61)

+ ...

T

= + . In the last line, we made use of the Dyson-Schwinger equation (2.57) for the photon to resum all fermion loop diagrams. Furthermore, the T -matrix was introduced [KKER86], i.e. the resummation of all ladder -diagrams. Since the Born approximation (third term in the rst line) is contained in both the rainbow and the ladder term, it has to be subtracted once, to avoid double counting of this contribution. In the context of Coulomb systems, equation (2.61) is also known to as the Gould-De Witt approximation [GD67] for the self-energy. The rst terms describes the inuence of the dynamical screening, while the second term describes strong collisions, i.e. collisions that involve large transfer momenta. Weak scatterings are contained in the Born approximation, i.e. the third term. The rst term, known as the GW approximation, can be seen as a modication of the Fock term (second term in equation (2.61)), with the Coulomb interaction replaced by the dynamically screened interaction. Being a -derivable approximation, the rainbow-ladder approximation, as well as the GW approximation lead to expressions for the Green function, which allows energy, mo-

37

mentum, and charge conserving calculations of higher order correlation functions, e.g. two-particle correlation functions or collision integrals. The drawback of -derivable approximations is, that the Ward-Takahashi identities are violated. Ward-Takahashi identities provide an exact relation between the vertex function , i.e. the eective electron-photon coupling in the medium, and the self-energy, which in relativistic notation read q (p + p, q) = q q + (p) (p ) . (2.62)

Ward-Takahashi identities follow directly from the exact Dyson-Schwinger equations. They reect the gauge invariance of the theory. In the GW or rainbow-ladder approximation, they are violated simply because corrections to the vertex beyond zero order are neglected altogether. As an example consider the self-energy in second order of the screened interaction, given by (2) = + . (2.63)

While the rst term is contained in the GW approximation as the second iteration, the second term is missing. It represents a vertex correction, similar to the radiative correction discussed in the context of bremsstrahlung in section 2.1.9. The problem of unfullled Ward-Takahashi identities in -derivable approximations touches on a fundamental problem in many-body theory and eld theory, namely the question how to preserve gauge invariance in an eective, i.e. approximative theory, without violating basic conservation laws. A detailed analysis of this question with application to nuclear physics is presented in a series of papers by van Hees and Knoll [HK01, HK02b, HK02a]. In particular, they show how to implement vertex corrections on top of the self-consistent solution of -derivable approximations, in a way that the desirable features of the -derivable approximations are retained and gauge invariance is restored. Approximations for the self-energy, that also contain the vertex are often referred to as GW approximations. An application to solid state physics can be found in Ref. [Tak01], where the spectral function of electrons in aluminum is calculated using a parametrized vertex function. An interesting result obtained in that work is that vertex corrections and self-energy corrections entering the polarization function, largely cancel. This is as a consequence of Ward-Takahashi identities. Thus, and in order to reduce the numerical cost, it is a sensible choice to neglect vertex corrections altogether, and to keep the polarization function on the lowest level, i.e. the random phase approximation (RPA) which is the convolution product of two non-interacting Green functions in frequency-momentum space. The corresponding self-energy is named the GW (0) self-energy and has been introduced by Holm and von Barth [BH96], who were also the rst to study the fully self-consistent GW approximation [HB98]. For a detailed presentation of various implementations of the GW approximations and

38

applications in the eld of solid state physics, I refer to the review articles [AG98, ORR02, Mah94, Hed99]. Before coming to the evaluation of the GW (0) approximation for the self-energy, the method of thermodynamic Green functions will be introduced in the next chapter. With its help, many-particle Feynman diagrams can be evaluated at nite temperatures.

2.3

In this section the method of thermodynamic Green functions will be outlined briey. It is also referred to as the Matsubara Green function technique. For a detailed presentation of this subject I refer to the textbooks by Mahan [Mah81] and Baym and Kadano [KB62]. The Matsubara Green function for fermions of species c in the momentum state p is dened as (2.64) Gc (p, , ) = T ac,p ( )a ( ) . c,p and are imaginary times, being dened on the interval i i. The dependence of the creation and annihilation operators a ( ) and ap ( ) on the imaginary time is given p in the modied Heisenberg picture, ac,p ( ) = eiH ac,p eiH , (2.65)

with the eective Hamiltonian H = H c,p c a ac,p , c is the chemical potential of c,p particles of species c, and H is the Hamiltonian of the many-particle system (see equation (2.11)), i.e containing the mutual interaction among the particles and external elds, if present. The bracket . . . in Eq. (2.64) denotes the average in the grand canonical ensemble with the equilibrium statistical operator O = Tr {eq O} = e Tr eH O , (2.66)

1 = ln Tr eH is the normalization of the grand-canonical statistical operator. Tr denotes the trace of an operator. Finally, the time-ordering operator T sorts the operators to its right in descending order, i.e. the smallest time is standing at the rightmost position. The Matsubara Green functions are very convenient, because they are directly connected to physical properties. For example, the single-particle Green function Gc (p, , ) or its Fourier transform, gives the spectrum of single particle excitations, i.e. the probability to nd a particle with a certain energy and momentum. Such probability can be measured in angular-resolved photoemission spectroscopy (ARPES) [Din98]. (PES). I shall now discuss two important properties of thermodynamic Green functions. The rst property is, that the Green function depends only on the dierence of its timearguments, Gc (p, , ) = Gc (p, ). This can easily be seen by setting = 0, = in Eq. (2.64) and using the cyclic properties of the trace. The second property is the socalled Kubo-Martin-Schwinger condition (KMS),

(2.67)

39

i.e. the Green function is periodic on the imaginary times axis with period 2i. From this information one immediately knows the Fourier spectrum of the Green function, i Gc (p, ) =

i

eiz Gc (p, z ) ,

=

Gc (p, z ) = i

d eiz Gc (p, ) ,

(2.68)

The Green function for free, i.e. non-interacting fermions can easily be calculated, using eiH0 ac,p eiH0 = eic (p) ac,p , with the unperturbed eective Hamiltonian H0 = at the expression G(0) (p, z ) = c

c,p ( 2 2

(2.70)

1 , z c (p)

where c (p) = 2 p2 /2mc c is the single-particle energy of species c, relative to the chemical potential c . Summation over all Matsubara frequencies leads to the momentum distribution function, 1 1 2 1 1 = = nF,c (p). z c (p) exp(c (p)) + 1 (2.72)

The Green function for interacting fermions is the solution of Dyson equation Gc (p, z ) = G(0) (p, z ) + G(0) (p, z ) c (p, z ) Gc (p, z ) = c c 1 , (2.73) z c (p) c (p, z )

where the self-energy c (p, z ) has been introduced. It contains the interaction among the fermions and the external elds. The Green function can be expressed by a Cauchy integral over its imaginary part,

Gc (p, z) =

d Im Gc (p, ) . z

(2.74)

The imaginary part of the Green function on the real frequency axis is also called the spectral function, i.e. 2Im c (p, + i) . [ c (p) Re c (p, )]2 + [Im c (p, + i)]2 (2.75) In a similar way, the photon Green function (propagator) in the Matsubara framework can be derived, starting from the time-ordered product Ac (p, ) = 2Im Gc (p, + i) = D (q, , = T A (q, )A (q, ) . (2.76)

40

Due to the permutation symmetries of the bosonic creation- and annihilation operators which appear in the eld modes A(q, ), the KMS condition for the bosonic propagator reads D (q, ) = D (q, + ) . (2.77) The Fourier coecients for D(q, ) are

D (q, ) =

d exp(i ) D (q, ) ,

2 , = 0, 1, 2 .

(2.78)

(0) D (q, ) =

4( q q /q 2 ) . 2 2 + q

(2.79)

The full photon Green function obeys the photon Dyson equation

(0) (0) D (q, ) = D (q, ) + D (q, ) (q, )D (q, ) .

(2.80)

Remind, that throughout this chapter, the Coulomb gauge A(q, ) = 0 is applied, which separates the electromagnetic eld into photons, and the longitudinal Coulomb interaction. The Coulomb potential V (r) = Z1 Z2 e2 /4 0 r2 is the propagator of the Poisson equation for an isolated point charge, i.e. for the non-interacting case. V (r)/Z2 e = Z1 e (r) . (2.81)

This equation is easily solved in momentum space, one obtains the well-known relation V (q) = Z1 Z2 e2 / 0 q 2 . For the case of an interacting many-particle system, the Green function or potential W (q, ) obeys the Dyson equation, W (q, ) = V (q) + V (q) (q, ) W (q, ) = V (q) , 1 V (q) (q, ) (2.82)

c.f. equation (2.57). Comparing equation (2.82) to the denition of the internal interaction potential as known from electrodynamics, W (q, ) = V (q)/ (q, ), the following relation between the longitudinal polarization function and the longitudinal component of the dielectric tensor can be established: (q, ) = 1 V (q) (q, ) . (2.83) Finally, the spectral representation for the eective interaction potential reads

W (q, z) = V (q) 1 +

d Im

(q, + i)

(2.84)

41

2.4

W G

The GW -approximation for the self-energy is given by the diagram (c.f. section 2.2)

(p, z ) =

(0)

(0)

= T

q

G(p q, z ) W (q, ) .

(2.85)

The index c will be ignored in the following and electronic quantities will be assumed. Equation (2.85) is evaluated as follows: First, the Green function G and the screened interaction W are replaced by their spectral representations (2.74) and (2.84). One arrives at

(p, z ) = T

q,

V (q)

d A(p q, ) 2 z

1+

d Im 1 (q, ) RPA

(2.86)

Note that the dielectric function is taken in the random phase approximation (RPA) [AB84]; this denes the GW (0) approximation. In GW , also the screened interaction has to be calculated as a functional of the Green functions, c.f. equation (2.57). After summation of the bosonic Matsubara frequencies,

(p, z ) =

q

V (q)

d A(p q, ) 2

1 nF ( ) +

d Im

1 RPA (q,

) [nB ( ) + 1 nF ( )] z

is obtained. This expression contains the Hartree-Fock self-energy of the interacting system, d HF (p) = A(p q, )nF ()V (q) , (2.87) int 2 q and the correlated self-energy corr (p, z ) =

q

V (q)

d A(p q, ) 2 (2.88)

d Im

1 RPA (q,

) [nB ( ) + 1 nF ( )] . z

For convenience, the upper index corr is skipped in the following and it is only distinguished between the frequency dependent self-energy (p, + i) and the Hartree-Fock term HF (p), in the following. int

42

After analytic continuation z z = +i, 0, the imaginary part of the correlated self-energy is evaluated using Diracs identity lim0 1/(x i) = P1/x i (x), Im (p, + i0+ ) = 1 nF ()

q

Im

1 RPA (q,

where the exact relation nB ( ) + 1 nF ( ) = nB ( ) nF ( )/nF () was used. This equation represents a non-linear integral equation for the imaginary part of the selfenergy Im (p, + i0+ ), since the latter also enters the spectral function via the Dyson equation 2.75. The real part of the self-energy can be obtained from the Kramers-Kronig relation Re (p, ) = HF (p) + int

d Im (p, + i) .

(2.90)

2.5

The knowledge of the polarization function or, equivalently, the dielectric function allows for the determination of various optical properties of the system under consideration. The transverse polarization function is related to the index of refraction n(q, ) and the absorption coecient (q, ) via

(q, )

= n(q, ) + i

c 2

(2.91)

The absorption coecient is dened as the relative attenuation per unit length of the intensity of an electromagnetic wave. The above relations can be solved for the index of refraction and the absorption coecient, themselves, 1 n(q, ) = [Re (q, ) + | (q, )|]1/2 2 (q, ) = Im (q, ) . c n(q, ) (2.92) (2.93)

Since in the limit q 0, the transverse dielectric function coincides with the longitudinal one, we may neglect the distinction between both cases and write (q, ) = (q, ) (q, ), instead, as long as the wavelengths involved are large compared to atomic dimensions, i.e. aB = 0.53 1010 m. As will be explained further below in section (3.6), the knowledge of the absorption coecient also allows for the determination of the emission spectrum from the plasma. Furthermore, the reectivity 1 n(q, ) r(q, ) = , (2.94) 1 + n(q, )

43

is known via the dielectric function. Thus, in order to determine the optical properties of a plasma, a theory of the dielectric function is needed. A quantum statistical approach to the dielectric function, which allows for the systematic inclusion of many-particle correlations is described in detail in [Rei05]. Central to this formulation is the so-called generalized Drude ansatz, which relates the dielectric function in the limit of long wavelengths to the so-called complex collision frequency (),

2 pl . (q 0, ) = 1 ( + i())

(2.95)

The collision frequency, on the other hand, can be expressed via a force-force correlation function, 0 z z (2.96) lim Jk , Jk +i , () = 2 k0 0 pl

z with Jk the longitudinal component of the force operator

Jk =

1 0

c,p

ec p a c,pk/2 ac,p+k/2 , mc

(2.97)

0 is the equilibrium statistical operator. The evaluation of the force-force correlation function is performed by making use of its representation in terms of a force-force Green function GJJ [ZMR96a]

z z J0 , J0 +i

d 1 1 Im GJJ ( + i) . + i

(2.98)

0

lim

1 1 = P x i x

i(x) ,

(2.99)

z z Re J0 , J0 +i

1 Im GJJ ( + i) .

(2.100)

z J0,e =

z H, J0,e =

ie me 0

pkq

(2.101)

using the non-relativistic many-particle Hamiltonian (2.31). With Eq.(2.101), we identify the Green function as a four particle Green function. Its diagrammatic representation is shown in gure 2.7. Further evaluation requires the replacement of the four-particle Green function G4 , that in principle contains all correlations, by a suitable approximation. Evaluation of the forceforce correlations using various approximations is demonstrated e.g. in Refs [RRRW00,

44

G4

Figure 2.7: Diagrammatic representation of the current-current Green function GJJ () as a four-particle Green function. WMR+ 01]. In [FRRW05], it is investigated, how the use of full Green functions instead of their non-interacting counterparts modies the result for the force-force correlation function and hence for the absorption coecient, compared the Born approximation, where all Green functions are non-interacting Green functions, and no interaction inside the correlation function is taken into account. From the force-force correlation function, the absorption coecient for inverse bremsstrahlung can be calculated via equations (2.96), (2.95), and (2.93). The results can be cast into the form 16 2 Z 2 ne ni () = 3me c 3 e2 4 0

3

2 3me kB T

1/2

[1 exp( /kB T )] g () ,

(2.102)

with the averaged free-free Gaunt factor g (). The averaged Gaunt factor contains the medium and quantum corrections. A central point in this thesis is the calculation of the averaged Gaunt factor. To this end, the in-medium electron propagator and the in-medium vertex is calculated and plugged in the force-force correlation function to determine the absorption coecient and by comparison to equation (2.102) the Gaunt factor. The Born approximation, discussed above, results in a simple expression for the Gaunt factor, namely 3 B exp ( /2kB T ) K0 ( /2kB T ) . (2.103) g () = K0 (x) is the modied Bessel function of order 0 [AS70]. In most cases, instead of comparing to the Kramers approximation g = 1, the results for the absorption coecient will be B compared to the Born approximation B () = ()g (). In this way, the inuences beyond the Born approximation, i.e. beyond the perturbative treatment of the Gaunt factor are studied.

3.1 Basic processes and plasma diagnostics

Plasma spectroscopy is a versatile too to infer macroscopic properties from the plasma, like the plasma temperature and the density of its various components (electrons, ions, neutrals). Also non-equilibrium properties, like gradients in temperature and density, instabilities, and non-linear eects can be detected and characterized by spectroscopic analysis. In particular, absorption and emission spectroscopy should be mentioned as important methods of plasma diagnostics [Gri64], besides other techniques e.g. refractometry and magnetic diagnostics [Hut87]. From the details of the spectrum, detailed information about the plasmas state can be extracted. The intensity of spectral lines, resulting from radiative transitions in ions, atoms, or molecules gives access to the plasma temperature and the degree of ionization. The form (width and position) of the lines also allows for the determination of the plasma density [Hut87]. The continuum part of the spectrum, due to free-bound and free-free transitions, is very sensible to the plasma temperature, density, and composition, as well. Fig. 3.1 (solid line) shows an exemplary emission spectrum from a laser produced plasma at an electron temperature of kB Te = 170 eV compared to the blackbody continuum at the same temperature (dashed line). The gure is taken from Ref. [Sig91]. Clearly, the three typical features of an emission spectrum can be identied, namely the spectral lines, the bound-free continuum, limited by sharp edges at the series limits, and the free-free continuum, which is the only contribution at long wavelength, i.e. small photon energies. The blackbody and the bremsstrahlung continuum merge at long wavelengths ( > 300 nm). This behaviour is due to the re-absorption of the bremsstrahlung (inverse bremsstrahlung) and subsequent thermalisation of the radiation, which therefore becomes blackbody radiation. This point will be discussed in more detail in section 3.6. The characteristics of the emission or absorption spectrum are sensitive to the density, temperature, and composition of the plasma. In order to interpret the experimental data and to determine these plasma parameters, an accurate theory of the optical properties of a realistic plasma is required. The complex dynamics of the interaction of electromagnetic

46

Figure 3.1: Exemplary emission spectrum from a laser produced plasma compared to the black-body spectrum at the same temperature. Taken from [Sig91] radiation with a highly correlated system, such as a partially or fully ionized plasma, has to be investigated. In this spirit, the present work is devoted to the theory of optical properties of dense plasmas. The focus will lie especially on the question how the free-free continuum (bremsstrahlung) behaves in a dense plasma. Other authors, using an approach similar to the one presented in this work (c.f. section 2.3), have concentrated on the theory of line spectra, such as Gnter [Gn95], Sorge u u + et al.[SWR 00], Koennies et al. [KG94] for hydrogen spectra, Omar et al. for helium like spectra [OGWR06], and Lorentzen [Lor08] for hydrogen like lithium spectra.

3.2

Line emission

The transition of a bound electron from an excited state |n to a nal state |m by emission of a photon of energy = En Em results in a sharp spectral line in the detection device. For the purpose of plasma diagnostics, the determination of the spectral lines width and the shift from its position as measured in a dilute system (gas) or known from rst principle

47

calculations, is of interest. To this end, the resolution of the spectrometer needs to be ne enough (typically below 1 nm). To understand the underlying principle of plasma diagnostics using spectral lines, one has to look at the several mechanisms which are responsible for the nite spectral line width. First of all, the line has a natural line width, since the life time of an excited atomic state even in vacuum is nite. This is due to the coupling of the atom to the vacuum uctuations of the electromagnetic eld as discussed in section 2.1.1. The natural line width nat is connected to via via the Heisenberg uncertainty relation, i.e. nat = 1/ . Typical values for the natural line width for dipole transitions range from 107 Hz to 109 Hz, which corresponds to life times in the order of 1 ns to 100 ns. E.g. for the hydrogen Balmer- line (3p 2s), one nds for the natural line width = 6.5 107 Hz [WSG66]. Dipole forbidden transitions, e.g. the 2s 1s transition in atomic hydrogen, have a much smaller natural line width of about 1.3 Hz.1 In a hydrogen gas or a plasma, the thermal movement of atoms or molecules leads to the so-called Doppler broadening. If the radiator moves away from the detector in the moment of the radiative decay, the wavelength of the emitted photon is shifted to larger wavelengths (redshift), while in the opposite case, when the radiator moves towards the detector, the Doppler eect leads to a blueshift. The distribution of wavelength shifts around the unperturbed position reects the velocity distribution function of the radiators in the plasma. Since line spectroscopy is usually carried out on dilute, hot plasmas, the momentum distribution function can well be approximated by the Maxwell-Boltzmann distribution. Its width (FWHM) is given by the temperature of the system, = 0 2 ln(2) kB T , mAt c2 (3.1)

with the atoms mass mAt and the unperturbed line position 0 . Therefore, Doppler spectroscopy allows for the determination of the plasma temperature. The Doppler width of the hydrogen Balmer- line at a plasma temperature of 1 eV is 2 1010 Hz, e.g. about two orders of magnitude larger than the natural line width of the same transition. This means, that the natural line width can be neglected in most cases where temperature measurements using line spectroscopy of hot plasmas are performed. Finally, in a dense plasma, the spectral line width and position is strongly inuenced by the interaction of the radiator with surrounding particles. The width of the spectral line due to the interaction with third particles is called pressure broadening. The electromagnetic eld exerted on the radiator by the surrounding particles leads to a shift (Stark shift) in the case of an electric eld and possibly Zeeman splitting (in the case of strong currents and external magnetic elds) of the levels involved in the particular transition. Since the magnitude of the shift of the upper and lower level involved in the particular transition depend on the respective quantum numbers, these processes result in a shift in wavelength of the emitted photon. The eld strength is distributed according to the spatial distribution

This feature makes the 2s 1s transition suitable for high precision measurements. An example is the investigation of the variation of fundamental physical constants with time [H 06].

1

48

of the perturbing particles, which is reected in a broadening and a shift of the spectral line. As opposed to the Doppler broadening, pressure broadening heavily depends on the density of the medium. Therefore, the measurement of the spectral line width and shift also allows for the determination of the plasma density. As an example, the hydrogen Balmer- line width due to pressure broadening is reported in [Hut87] as 3 1010 Hz at a plasma temperature of kB T = 1 eV and ion density of ni = 1021 m3 . This value is of the same order of magnitude as the Doppler broadening of the same line at the same temperature. Thus, to accurately determine the temperature and the density in dense plasmas from the spectral line broadening, both eects have to be taken into account. The pressure shift (Stark shift) is of the same order of magnitude as the Stark broadening. For further details on spectral line broadening, I refer to the extensive literature on this subject, e.g. the monograph by Griem [Gri64], the collection [LH95], and the original articles [OGRW06, OGWR06], plus further references therein. Another method to determine the plasma temperature from the line spectrum uses a statistical argument. Assuming that the population of the levels involved in a particular radiative transition follows a Maxwell-Boltzmann distribution, the temperature of the plasma can be obtained by comparison of the integrated line intensities of two distinct spectral lines [LH95]. This method is particularly useful, if the spectrometer does not allow for a detailed analysis of the spectral line shape, i.e. if the spectrometers spectral resolution is not sucient to determine the Doppler broadening and/or the pressure broadening. The statistical distribution of the levels involved in the radiative transition is also used in the temperature measurement via spectroscopy of continuum radiation, as shown e.g. in Ref. [ZFF+ 08], see also section 4.2.

3.3

Recombination radiation

When the plasma is partially ionized, two additional processes contribute to to the emission spectrum of the plasma. Besides transitions between continuum states (bremsstrahlung), free-bound transitions occur. The case of a transition from a continuum state |k at energy Ek to a bound state |n at a discrete energy En and the accompanying emission of a photon of energy = Ek En is called recombination radiation, whereas the opposite process is referred to as photoionization. The corresponding emission (absorption) spectrum has a sharp edge at the line series limit as the most remarkable feature. Besides these edges, which allow for the identication of ion species and ionization stages, the intensity of bound-free transitions in a plasma in thermal equilibrium behaves roughly proportional to exp( /kB T ), as follows from averaging the individual transition probability with the Maxwell-Boltzmann distribution function of the initial states. This allows for the determination of the temperature from the slope of the frequency (energy) spectrum. Furthermore, by measuring the height of the absorption edge, the penetration depth of the radiation into the medium can be determined, provided, the absorption coecient at both sides of the edge is known.

49

3.4

Bremsstrahlung radiation

Bremsstrahlung is the radiative transition between two continuum states. In section 2.1.9, various expressions for the bremsstrahlung cross-section have been reviewed, i.e. the BetheHeitler formula (2.42) (full relativistic Born approximation), the Sommerfeld expression (2.51) (quantum-mechanical, non-relativistic), and the Kramers formula (2.48) (classical, non-relativistic). By averaging the cross-section, which depends on the initial velocity (momentum) of the emitting particle, with the momentum distribution function, an expression for the bremsstrahlung emission from a plasma can be derived. Thus, the accurate measurement of the bremsstrahlung spectrum emitted from a plasma allows for the determination of many plasma parameters [Hut87]. In general, bremsstrahlung spectra can be used to reconstruct the momentum distribution of electrons in the plasma [BEH+ 06] by deconvolution of the bremsstrahlung scattering cross-section from the measured spectrum. In the case of a classical, completely ionized plasma in thermal equilibrium, where the electron distribution function is a Maxwell-Boltzmann distribution, i.e. completely characterized by the electron temperature, the bremsstrahlung photon energy spectrum has an exponential decay behaviour at large photon energies, Ibs () exp( /kB T ), c.f. section 2.5. In this case, it is sucient to determine the slope of the bremsstrahlung continuum to obtain the electron temperature [Ric95]. For partially ionized plasmas, the situation is more complicated, since especially in the high frequency part of the spectrum, bound-free transitions become important. However, the latter have a similar frequency dependence and can be used for temperature measurements in the same way. An example of application can be found in Ref. [SWR+ 00]. In non-equilibrium plasmas, where we have additional electron modes (instabilities), the bremsstrahlung spectrum is no longer exponential but rather displays a power behaviour [HSSE03, BEH+ 06]. The bremsstrahlung emission law 2 carries a prefactor Zne / Te . Thus, if the emission spectrum is absolutely calibrated, also the electron density can be determined. Bremsstrahlung spectroscopy is widely used in astronomy and astrophysics to determine important parameters of the surface of stars, e.g. its temperature, density, and electron distributions. An example can be found in [KEM+ 07].

3.5

Scattering

Besides emission and absorption, the scattering of photons is the third class of processes belonging to the response of the medium to external elds. In general, two kinds of scattering have to be distinguished, namely elastic and inelastic scattering. Elastic scattering, also called Rayleigh scattering, denotes those processes, where the energy of the photon is conserved, only the momentum changes, i.e. the photon is scattered into some angle with respect to its initial trajectory. In contrast, inelastic scattering involves also a change in the photons energy, i.e. energy is transferred from the scattering particles to the scattered particles or vice versa. Several phrases are used in connection with scattering, denoting various channels of energy and momentum transfer. For a detailed discussion, see e.g. the

50

monograph by Sheeld [She75]. Thomson scattering refers to the elastic scattering of photons on free or weakly bound electrons. For long wavelengths, i.e. in the optical and the infrared regime, the classical picture of an oscillatory movement of the electron in the external eld applies, which leads to dipole radiation from the electron [LGE+ 01]. Due to the thermal movement of the electrons, there appear red- and blue shifted photons in a detector that is set up at a particular position. Thus, the scattering spectrum is symmetrically broadened and the width gives the plasma temperature, assuming, the velocity distribution is given by the Maxwell-Boltzmann distribution. In cold systems, e.g. metals, when the temperature is comparable or below the Fermi energy, the width of the scattering spectrum is given by the Fermi energy, which allows for the determination of the electron density [LNC+ ew]. Note that scattering of electromagnetic waves in cold targets requires the use of x-rays, since optical probes cannot penetrate the target when the probe frequency is below the plasma frequency of the target. Scattering of photons involving atomic or molecular transitions is called Raman scattering. During the interaction, a bound electron is transferred from its initial state to some intermediate, unoccupied state and from there to a nal state. Depending on weather the nal state is energetically higher or lower than the initial state, the wavelength of the scattered photon is shifted towards longer wavelengths (Stokes line) or shorter wavelengths (Anti-Stokes line), respectively. The term Raman scattering is also used for the scattering of infrared photons involving transitions between vibrational or rotational states of molecules. Finally, Brillouin scattering denotes the inelastic scattering of photons on collective excitations of the medium. Typically, the term stands for those processes that involve the absorption or creation of phonons by light scattering. But also scattering on other excitations, such as plasmons or magnons (spinwaves), can be referred to as Brillouin scattering. However, the scattering on plasmons as the longitudinal excitations in a plasma is mostly called collective Thomson scattering, since it involves scattering on free electrons which perform collective motion. Optical Thomson scattering has been applied since many years in diagnostics of dilute gases and plasmas, such as Tokamak plasmas [Hug75] or arc discharges [SLR93]. Also electron-hole plasmas in semi-conductors have been investigated using Thomson scattering [UW77]. Nowadays, researchers are looking for reliable and robust techniques to diagnose also dense systems. To a large extent this is due to the increasing research activities in inertial connement fusion research [Lin95] and laboratory astrophysics [TLS+ 06]. In these elds, the structure and the dynamics of plasmas which are at the same time hot (with temperature between several eV and several hundred eV) and dense, i.e. whose density ranges in the vicinity of solids, near 1023 cm3 , are investigated. At these parameters, the plasma is no longer penetrable by optical probes, since the plasma frequency is larger than the probe frequency, i.e. the radiation is reected and penetrates the plasma only in the skin-layer of several nm in depth. However, latest achievements in the development of narrow bandwidth, high intensity UV and x-ray sources, such as backlighter sources [GGL+ 03, SGK+ 07] and free electron lasers [AAA+ 07], made plasma diagnostics using

51

Thomson scattering also applicable in this so-called warm dense matter regime [GLN+ 07, GND+ 08].

3.6

Radiation transport

Dealing with radiation from and inside a plasma of nite spatial extensions, emission, absorption as well as scattering have to be treated on equal footing. External elds get absorbed and scattered in the medium, at the same time radiation is produced by radiative transitions inside the matter and possibly reabsorbed. The question of radiation transport becomes important. Radiation transport accounting for both absorption and emission and scattering is a complicated issue, that cannot be covered in these introductory remarks. Therefore, I focus on radiation transport that includes only absorption and emission. Since all processes of emission and absorption are assumed to be isotropic (we neglect e.g. the case of magnetized plasmas), the problem of radiation transport can be formulated in one dimension. Note, that when including scattering, this simplication is no longer valid, since the momentum transfer during the scattering involves a change of the photons direction while propagating through the medium. The central quantity is the absorption coecient , dened as the relative attenuation per unit length of the intensity of electromagnetic waves propagating through a medium. In general, the absorption coecient depends on both time and place. Also, it depends heavily on the frequency of the electromagnetic wave, i.e. its Fourier spectrum. The basic equation that governs the propagation of radiation through a medium is the macroscopic radiative transfer equation, dI(; s) = (; s)I(; s) + j(; s) , ds (3.2)

which is a balance equation for the loss and the gain of the radiation intensity I(; s), i.e. the radiation power emitted into the solid angle d and frequency interval d, measured at a detector of cross-section dA at the distance s from the incident surface of the medium, I(; s) = dPrad (; s)/dA d d. Here, the discussion is restricted to the onedimensional transfer equation, for the much more complicated three-dimensional case, see the monograph by Chandrasekhar [Cha60]. The loss of intensity is described by the rst term in equation (3.2), the second term describes the gain, i.e. the emitted power per unit frequency interval, unit solid angle and unit volume at s. It is convenient to dene the dimensionless optical depth as the integral of the absorption coecient along the path taken by the radiation through the medium of length l as

l

(; l) =

ds (; s) .

(3.3)

j(; s) is the emittance, i.e. the radiated power per unit volume, l is the length of the

(3.4)

with the source function S(), dened as the ratio of the emissivity and the absorption coecient, S() j(; s)/(; s). Eq. (3.4) has the formal solution I(; ()) = I(, 0)e

()

() 0

e( ()

())

S(, ()) d () .

(3.5)

I(, 0) is the intensity of the wave at the incident surface. In the case of a homogeneous medium, the optical depth is given by the product () = l () and the solution of the radiation transport equation becomes I(; l) = I(, 0)e()l + S()(1 e()l ) . (3.6)

In general, the source function can have arbitrary complexity. In the case of thermal radiation, i.e. radiation in thermodynamic equilibrium, the source function is the Planck blackbody spectrum, given by ST () = or, as a function of the wavelength, ST () = 4 c2 /5 d ST () = 2 c/k T . B e 1 (3.8) 3 /4 3 c2 , e /kB T 1 (3.7)

Two limiting cases of Eq. (3.6) should be discussed more closely. In the limit () 1, we have a transparent medium, only a small amount of radiation is absorbed. In this case, Eq. (3.6) turns into Kirchhos law [RL75], i.e. the emitted intensity is I(; l) I(; 0) = ()ST () l jT () l. When we have a strongly absorbing medium, external elds and self-emission are repeatedly absorbed and re-emitted, such that thermodynamic equilibrium is established between the matter and the radiation. The resulting self-emitted intensity I(; l) I(; 0) = ST () is just Plancks law of a black-body radiator. We have now dened the quantities that describe the propagation of electromagnetic waves through a given medium, i.e. the absorption coecient and the source function, and we have outlined the relations between them. Next, we need to establish a link to the microscopic level, i.e. to the physical behaviour of the elementary constituents of the medium. Since we will consider the electromagnetic spectrum only for photon energies smaller than -ray energies ( < 1 MeV), we are allowed to treat electrons and ions as elementary particles and we do not have to consider hadronic or partonic degrees of freedom.

4.1

4.1.1

Introduction

Thomson scattering is a well established tool of plasma diagnostics. Applications range from dilute, classical plasmas [SLR93, Mur04] to high energy density plasmas at densities above solid density [GND+ 08]. Also to semiconductors and insulators the technique of Thomson scattering diagnostics is applicable [UW77]. The plasma temperature and the plasma composition can be determined directly from the scattering spectrum. Thereby, it is important to dierentiate between two regimes of plasma scattering: The main parameter, determining the scattering spectrum qualitatively, is the scattering parameter = /k. Here, k is the modulus of the momentum which is transfered from the scattering photon to the scattered electron. It is determined by the wavelength 0 of the scattering photons and the scattering angle as k = 4/0 sin /2. Furthermore, is the screening parameter. In classical plasmas, is given by the inverse Debye screening length, = ne e2 / 0 kB T , which is the typical length scale of density uctuations in a plasma. At small transfer momenta, i.e. for > 1, the scattering takes place on collective density oscillations, i.e. plasmons. In the scattering spectrum, two plasmon peaks appear, which are up and down shifted from the frequency of the scattering photon by roughly the plasma frequency pl = ne e2 / 0 me . Thus, from the position of the plasmon peaks, the electron density can be inferred. The precise location of the plasmon resonances is a function of the transfer momentum, the plasma density and the temperature. A quantum statistical approach to the plasmon resonance position was recently presented by Thiele et al. [TBF+ 08]. The plasma temperature can be obtained from the scattering spectrum by comparing the scattered intensity I() of the up and down shifted plasmon resonance. This is due to the detailed balance relation, I() = exp( /kB T )I() [HBC+ 07]. At larger transfer momentum, i.e. < 1, the photons scatter mainly on the uncorrelated individual charges. Under these conditions, the scattering spectrum is a Doppler broadened resonance, whose width gives the plasma temperature.

54

Selected problems

In order to resolve the plasmon resonances and their separation from elastic Rayleigh scattering, small bandwidth sources are required. However, the applicability of optical lasers is limited to rather low density plasmas, due to the critical density of about 1020 cm3 for optical wavelengths. For solid density plasmas, narrow bandwidth x-ray sources are the method of choice. Pioneering work in this direction was performed by Glenzer et al. showing the possibility of Thomson scattering in both the non-collective scattering regime ( < 1) [GGL+ 03] as well as in the collective scattering regime [GLN+ 07]. Recently, the same technique was also demonstrated for shock compressed matter, see Refs. [GND+ 08, LNC+ ew]. The regime of near solid density (ne = 1021 cm3 1023 cm3 ) at temperatures of several eV is of particular interest for Thomson scattering. In this so-called warm dense matter (WDM) regime, a complex interplay between strong coupling eects (collisions) and quantum eects can be observed, since both the plasma coupling parameter as well as the degeneracy parameter are close to 1. The theory of Thomson scattering has to account for these eects to allow for precise prediction of the scattering spectrum. A many-body theoretical approach is presented in Ref. [TRRR06]. A collective Thomson scattering experiment at the free electron laser facility FLASH at DESY-Hamburg, operating at XUV wavelengths (currently 6-100 nm), was proposed in Ref. [HBC+ 07]. An important issue related to the feasibility of this experiment is the question of the signal to noise ratio. At the sought plasma conditions, bremsstrahlung represents a considerable source of background photons, which has to be overcome by the scattering signal. A detailed comparison of the bremsstrahlung level and the intensity of scattered FEL photons over a broad range of plasma parameters was performed in Ref. [FRR+ 06]. Different from previous studies of the same subject, various Gaunt factors were compared. Besides the standard expression for bremsstrahlung, i.e. Kramers formula, the Born approximation and the Sommerfeld expression were applied. In particular the latter approximation allowed to study the inuence of strong collisions on the bremsstrahlung emission. The Thomson scattering spectrum was calculated in random phase approximation [HRR+ 04]. Two key questions were of primary interest: What is the minimum intensity of the FEL required to overcome the bremsstrahlung level at various densities and temperatures? What is the required detector resolution to achieve this? The detector resolution is taken into account by performing convolution of the simulated scattering spectra [HBC+ 07] with a Gaussian of the corresponding width.

4.1.2

Results

We consider Thomson scattering of photons with a wavelength of 0 = 32 nm on Z = 2 fold ionized aluminum under an angle of = 120 . Figure 4.1 (black dashed curve) shows the raw scattering spectrum, the solid curve is obtained by convolution of the raw spectrum with a Gaussian function of relative width / = 102 to model the detector resolution. The laser intensity is IL = 5 1010 W/cm2 . The electron temperature is

410 5 Bremsstr. (Kramers) Bremsstr. (Born) Bremsstr. (Sommerfeld) Thomson (IL=5x10 10 W/cm 2) Thomson, not convoluted

55

310 5

210 5

110 5

0 30

31

32 [nm]

33

34

Figure 4.1: (a) Thomson scattering spectrum for laser wavelength 0 = 32 nm at scattering angle = 120 for aluminum at ionization degree Z = 2. The laser intensity is IL = 5 1010 W/cm2 . The thin dashed curve gives the raw scattering spectrum, the solid curve accounts for the detector resolution (/ = 102 ) via Gaussian convolution. The grey curves give the bremsstrahlung emission level assuming dierent expressions for the Gaunt factor, i.e. Kramers approximation (dashed), Born approximation (dotted), and Sommerfeld approximation (solid). The plasma parameters are ne = 1020 cm3 and kB Te = 10 eV. The scattering parameter is = 1.25. xed at Te = 10 eV/kB , the electron density is ne = 1020 cm3 . At these conditions, the scattering parameter is = 1.25, i.e. we are in the collective scattering regime. Correspondingly, plasmon peaks appear in the spectrum, separated from the central Rayleigh peak by roughly 0.5 nm. The grey curves mark calculations of the bremsstrahlung emission for the same conditions, applying dierent expressions for the averaged Gaunt factor, c.f. equation (2.102), i.e. Kramers approximation (g = 1, dashed curve), Born approximation (equation (2.103), dotted curve), and the temperature averaged Sommerfeld expression (equation (2.51), solid curve). All bremsstrahlung curves are below the peaks of the Thomson scattering signal, i.e. at the present condition, the Thomson scattering experiment is feasible. Comparing the dierent bremsstrahlung curves, the Sommerfeld approximation gives the most conservative estimation, i.e. the highest bremsstrahlung intensity. Born approximation, gives a much lower bremsstrahlung level. The Kramers result, i.e. neglecting quantum and medium eects is close to the Sommerfeld result at the present conditions. In the same manner, the maximum Thomson scattering intensity is compared to the

56

10 24

10 14 W/cm 2

Selected problems

10 23

10 13 W/cm 2

ne [cm -3]

10 22

10 12 W/cm 2

10 11 W/cm 2

10 21

10 9 W/cm 2

10 20 10

10 10 W/cm 2

1000

Figure 4.2: Threshold intensity as a function of electron temperature Te and free electron density ne . For each curve, (ne Te )-points above and to the right are not accessible by Thomson scattering since the bremsstrahlung level is above the Thomson signal. Below and left of each curve, the Thomson signal is stronger than the bremsstrahlung background. Various expressions for the Gaunt factor have been used, Kramers (dashed), Born (dotted) and Sommerfeld (solid). The detector resolution is xed at / = 104 , the FEL wavelength is 32 nm and the scattering angle is 120 . The target material is Al, Z=2. bremsstrahlung intensity over a broad range of electron density and temperature. In this way, for each point in the (ne Te ) plane, a threshold intensity can be determined, from which on the Thomson scattering intensity overcomes the bremsstrahlung background. The result is shown in gure 4.2 for the same scattering parameters as above. This time, the detector resolution is set to 104 . The curves are isolines of the threshold intensity as a function of electron density and temperature. Having a certain laser intensity available, all points in the (ne Te ) diagram below and to the left of the corresponding isoline allow for Thomson scattering, whereas bremsstrahlung is dominant for points above and to the right of the isoline. Here, higher laser intensities are needed. As before, dierent Gaunt factors are compared. While the Born approximation (dotted curves) shows a rather large deviation from the Kramers result (dashed), especially at increased densities, the Sommerfeld result (solid) is closer again to Kramers. The Sommerfeld expression for the Gaunt factor is the most accurate expression which contains the exact scattering wavefunctions. Hence, it should be used to estimate the bremsstrahlung intensity. It gives slightly higher thresholds as compared to the Kramers result, which was used in Ref. [BDFR02].

57

Summarizing, we determined threshold intensities for a Thomson scattering experiment using free electron laser radiation at warm dense matter conditions. It was shown that moderate intensities of 1012 W/cm2 are sucient to perform Thomson scattering in the WDM regime. Furthermore, a comparatively low detector resolution of 102 is sucient, 104 would be largely enough.

4.2

4.2.1

Introduction

The production of a hot plasma (i.e. at temperatures of several eV to several hundreds of eV) at solid density is a challenging task to researchers. The main interest in such a plasma stems from its relevance for inertial connement fusion research [Lin95] and laboratory astrophysics [TLS+ 06]. Among the various strategies, shock wave experiments [DSCC+ 97], heavy ion beams [HBN+ 05, HFL+ 02], and high power lasers are the most successful. A promising further option to produce high energy density plasmas are x-ray light sources of the fourth generation, free electron lasers [A+ 06]. Due to the shorter wavelength as compared to optical lasers, the radiation can penetrate deep into the target and can produce free electrons via direct photoionization [FZG04]. A homogeneous, volumetric plasma can be produced. In contrast, optical laser radiation is usually absorbed in a thin skin-layer of several nm producing a plasma with high density and temperature gradients.

4.2.2

A pioneering proof-of-principle experiment [ZFF+ 08] has recently been conducted at the free electron laser at DESY-Hamburg (FLASH) [A+ 06, AAA+ 07]. FEL radiation of 13.5 nm wavelength was focused on a sample of aluminum under 45 degree incident angle. The FEL was run in multibunch mode at 5 Hz repetition rate and with 20 pulses in each bunch. The pulse length of each individual pulse was 15 fs and the pulse energy was 33 J. The focal spot size was 30 m in diameter which gives an intensity of 2 1014 W/cm2 . The plasma emission was measured with a high throughput transmission grating XUV spectrometer [JTT+ 94] under 90 angle with respect to the FEL. The setup is shown in gure 2 in Ref. [ZFF+ 08]. The resulting spectra are shown in gure 4.3. The summed photon yield of ve separate measurements (total integration time was 13.5 minutes) is shown as a function of the photon wavelength. Various features are discernible: The main peak at 13.5 nm is the elastically scattered FEL light. In its right wing, a spectral line at = 12.7 nm can be identied originating from Al4+ (Al V). Further spectral lines mainly from Al IV ions are located at 11.6 nm and at 16.2 nm. These lines have been identied with the help of NIST data tables [KM91]. At 17.0 nm, the usual L-shell absorption edge shows up, accompanied by

58

Rayleigh scattering

Selected problems

-1

emitted photons

10

10

-1

10

10

17.0 nm (L-edge)

10

10 12 14 wavelength [nm]

16

18

Figure 4.3: Plasma emission spectrum featuring spectral lines from Al IV, Al V, the L-shell absorption edge and Rayleigh scattering. The solid curves show bremsstrahlung calculations at three dierent temperatures. Best t is obtained for kB Te = 40.5 eV.

an LII/III transition line at 17.2 nm. These features sit on a background that is mainly caused by bremsstrahlung. At short wavelengths ( < 10 nm), bound-free transitions show up as step-like structures.

4.2.3

Data analysis

From the spectroscopic data, valuable information about the plasma parameters can be inferred. Notably the plasma temperature is of interest, since it gives information about the eectiveness of plasma heating using FEL radiation. The plasma temperature can be determined in two complementary ways: First, the background radiation was compared to calculations for the bremsstrahlung radiation at various temperatures. These calculations are given in gure 4.3 as coloured lines (red: kB T = 46 eV, green: kB Te = 40.5 eV, blue: kB Te = 35 eV). The best t to the experimental data is given by the 40.5 eV bremsstrahlung curve. In this analysis, the Gaunt factor in Sommerfeld approximation was used, see section 2.1.10. Alternatively, the plasma temperature is given by the ratio of integrated line intensities [LH95]. We compared the Al IV lines at 11.6 nm and 16.2 nm, and obtained kB Te = (34.6 6) eV for the plasma temperature. Within the error bars, this value agrees well with the temperature that was obtained from the bremsstrahlung background.

59

Since the spectrum gives absolute photon numbers, also the electron density can be determined via the bremsstrahlung background. The value of ne = 4.21022 cm3 indicates that in fact a hot plasma at solid density conditions was produced. Furthermore, these values comply well with radiation hydrodynamics simulations using the code HELIOS [MGW06]. In conclusion, the reported experiment has shown that FEL lasers provide a well suited tool to produce warm dense matter under dened laboratory conditions by homogeneous plasma heating. The emitted radiation in the XUV spectral range, notably the bremsstrahlung continuum, allows for an accurate characterization of the main plasma parameters, i.e. density and temperature. In this case, temperatures around 40 eV and densities of 4.2 1022 cm3 have been found.

4.3

4.3.1

Introduction

The modication of the bremsstrahlung emission spectrum in a dense medium is a long standing issue. Since the pioneering work by Bethe and Heitler [BH34, Hei54], it was clear, that perturbative Quantum Electrodynamics provides a rather incomplete picture of the bremsstrahlung process. Notably the infrared-divergence d/d 1/ of the bremsstrahlung cross-section was discussed. Two main lines of arguments can be identied: On the one hand, the more academic problem of the isolated bremsstrahlung event, i.e. the scattering of a single electron on a single ion is considered going beyond the Born approximation. Compensation between divergent higher order terms in the elastic scattering cross section and the infrared divergence of the Bethe-Heitler formula is found, as was already discussed in section 2.1.9. A dierent approach takes its arguments from the fact that bremsstrahlung never occurs completely isolated but rather in a more or less correlated environment. The basic idea is the following: If the oscillation period of the emitted photon is of the same order as the average time between collisions with ions, i.e. the life time of the single-particle state, the scattering amplitudes of successive scatterings have to be added coherently in order to nd the total scattering probability. In disordered media, the interference of successive scattering amplitudes is destructive, the bremsstrahlung cross-section is reduced as compared to the isolated bremsstrahlung process. For the case of two scattering centers, this was rst shown by Landau and Pomeranchuk [LP53] using semiclassical arguments. Later, a quantum mechanical theory was developed by Migdal [Mig56]. He described the propagation of the electron through the scattering medium as a diusion process. The ratio of the resulting bremsstrahlung cross-section in the medium to the Bethe-Heitler result is . Hence, LPM suppression does not remove the infrared divergence proportional to but the order of divergence is lowered from ( )1 to ( )1/2 . The inuence of screening of the ion potential and dispersion of the emitted photons

60

Selected problems

in the medium was rst considered by Ter-Mikaelyan [TM53] using the dielectric function. The resulting bremsstrahlung cross-section behaves proportional to at low photon energies, i.e. the infrared divergence is removed. This result is often referred to as the dielectric suppression of bremsstrahlung. Both LPM suppression and dielectric suppression have been investigated by various experimental groups. The rst experiments made use of highly energetic cosmic ray particles. Particular features in the energy distribution of secondary particles could be traced back to the LPM eect [Kas85]. First accelerator based studies of the LPM eect were performed by Varfolomeev et al. [V+ 75]. However, the results suered from poor statistics. More clear evidence of LPM suppression was obtained by Anthony et al. [ABSB+ 97], although other eects, such as dielectric suppression and also transition radiation [Kle99] could not always be separated. The experiments by Hansen et al. [HUB+ 03, HUB+ 04] are today regarded as the rst unambiguous conrmation of LPM suppression. Due to the large electron energies of several hundred GeV, all energy scales could be separated, i.e. the various eects were identiable. A review of the theory of the LPM eect and experimental studies can be found in Ref. [Kle99]. In that work, the author also mentions the possibility of LPM suppression at non-relativistic energies in the thermal bremsstrahlung emission from hot and dense media, e.g. high energy density plasmas.

4.3.2

In Ref. [FRRW05], the eect of successive scattering on the thermal bremsstrahlung emission and absorption in a dense plasma is analyzed. A non-relativistic, fully-ionized hydrogen plasma is assumed. Within the framework of linear response theory [ZMR96b], the absorption coecient is related to equilibrium force-force correlation functions, which are calculated with the help of thermodynamic Green functions. Successive scatterings are accounted for by resummation of the self-energy that describes the single scattering event, i.e. solving the corresponding Dyson equation. A similar approach was also suggested by Knoll and Voskresensky [KV96]. They showed, that the account for the nite life time of single-particle states, i.e. a nite width of the spectral function, leads to a suppression of the bremsstrahlung spectrum at small photon energies. The emissivity is obtained from the polarization tensor which, in one-loop approximation, is given as the convolution of two spectral functions. However, the imaginary part of the self-energy, was not explicitly calculated but rather set as a pure parameter. Thus, only qualitative results could be obtained. A microscopic approach to the self-energy and thus to the LPM eect in plasmas was still missing. In this work, a self-consistent equation for the self-energy is derived, that describes the

61

i

(p, ) =

VD i VD G

(4.1)

The approach is similar to the GW approximation [Hed65], with the exception that a statically screened Debye potential is used instead of a dynamically screened one. In this way, the Born approximation is automatically included as the lowest order contribution to the self-energy. Eects due to dynamical screening, such as the dielectric suppression, are excluded. Hence, the LPM eect can be studied independently. Having the full electron propagator at our disposal, we proceed to calculate the bremsstrahlung spectrum via the force-force correlation function. In order to fulll WardTakahashi identities [War50, Tak57], also vertex corrections to the force-force correlation function are performed. For a broader discussion of this issue, see section 2.2.1. Only the lowest order vertex correction that stems from electron-ion scattering is considered here. The self-consistent calculation of the vertex function in the same manner as for the selfenergy, i.e. the solution of the corresponding Bethe-Salpeter equation was not performed.

4.3.3

Results

The central result of this study is shown in gure 4.4. The absorption coecient () for inverse bremsstrahlung, calculated from the force-force correlation, is given as a function of the photon energy . The absorption coecient is normalized to the result obtained in Born approximation, see equation (2.103). The plasma parameters are ne = 106 a3 = 6.7 1018 cm3 for the electron density and kB T = 2 Ry = 27.2 eV for the plasma temperature, i.e. a classical, weakly coupled plasma is studied. Two results are compared: The solid line represents the complete calculation including the self-energy corrections as well as the lowest order vertex correction. The dashed line belongs to the absorption coecient with only the self-energy corrections taken into account while neglecting vertex corrections. In both cases, a signicant deviation of the absorption coecient from the Born approximation is obtained. Notably at small photon energies ( < 0.08 Ry), the absorption is largely suppressed. At high photon energies ( > 0.5 Ry) the eect vanishes, the curves converge with the Born approximation, i.e. ()/B () 1. The suppression was found to increase as a function of the imaginary part of the self-energy. This coincides with the results by Knoll and Voskresensky [KV96], using an energy and momentum independent self-energy. At intermediate photon energies, an increase of the absorption coecient with respect to the Born approximation is found with a maximum of ()/B () reached at = 0.2 Ry. This resonant behaviour is a novel feature, that has not been observed before and must be attributed to the dynamical character of the self-energy. Note, that the plasma frequency

62

Selected problems

Figure 4.4: Inverse bremsstrahlung absorption coecient as a function of the photon energy for a classical plasma. (ne = 106 a3 6.7 1018 cm3 and kB T = 2 Ry = 27.2 eV). B The results are normalized to the Born approximation result B (), see equation (2.103). Solid line: Sum of self-energy and vertex contribution. Dashed line: Only self-energy contribution. at the present density is pl = 7.1 103 Ry, thus the increase in the absorption can certainly not be attributed to plasmon resonance. This enhancement feature certainly deserves further attention but goes beyond the scope of this work. By comparison between the solid curve (vertex corrections + self-energy) and the dashed curve (only self-energy), the eect of the vertex corrections can be studied. The increase of absorption at intermediate frequencies is lowered from ()/B () = 1.06 to a value of 1.04. This decrease also seen in the calculation of spectral line shapes [Gn95]. Also there, u vertex corrections largely cancel with self-energy eects. At low photon energies, the suppression of inverse bremsstrahlung absorption is amplied, here the vertex corrections and the self-energy add up to a large eect. It remains to be scrutinized if this large suppression is an artefact of the perturbative treatment of the vertex correction, i.e. if this behaviour changes if also the vertex function is determined self-consistently by solving the corresponding Bethe-Salpeter equation (2.59). Summarizing, it was shown, that the successive scattering of electrons on randomly distributed ions leads to a decrease in the inverse bremsstrahlung absorption eciency. This is comparable to the LPM eect that is well known from high energy scattering experiments. As a further interesting feature it was found, that the account for the self-

63

energy also leads to an increase of the absorption at intermediate photon energies which, however is reduced, if vertex corrections are considered. The latter is required to fulll Ward-Takahashi identities, i.e. to preserve gauge invariance of the theory.

4.4

4.4.1

Optical Properties and the One-Particle Spectral Function in Hot Dense Plasmas

Introduction

In the previous section, the Green function method was applied to determine the eect of successive scattering on the inverse bremsstrahlung absorption coecient in a classical, weakly coupled plasma, see also Ref. [FRRW05]. In this section, some further reaching questions shall be addressed: First, the eect of successive scattering is investigated at solar core conditions. The solar core plasma is a prototypical example for high energy density matter. Values for the mass density of = 1.5 102 g/cm3 and a temperature of T = 1.6 107 K 1360 eV are extracted from observations [BPW95]. By comparing to the results obtained in section 4.3, we can study the eect of increased density. Second, we analyze the importance of vertex corrections in the self-energy itself, whereas before, the vertex was only considered inside the polarization loop. To this end, GW calculations (see section 2.2.2) are performed.

4.4.2

Results

First, we consider the impact of successive electron-ion scattering on the absorption coecient following the same lines as in the previous section. As our model system, a completely ionized hydrogen plasma at solar core conditions is chosen. The density of ions and electrons is ni = ne = 7 1025 cm3 and the temperature is kB T = 100 Ry = 1360 eV. The results are shown in gure 4.5. Using only self-energy corrections (solid curve) , a similar behaviour as in the former case was found, i.e. suppression of bremsstrahlung at low photon energies, an enhancement at photon energies above the plasmon energy pl and convergence to the Born approximation in the high energy limit. Again, vertex corrections are applied to the force-force correlation function. At solar core parameters, the vertex corrections largely dominate the self-energy eects (dashed curve). In particular, the enhancement at intermediate photon energies is compensated. Similar results have also been obtained by [BTADS97]. The question of vertex corrections in the self-energy itself is investigated by evaluating the self-energy in second order of the electron-ion scattering, (2) (p, ) = + . (4.2)

64

1.1

Selected problems

1.05

1 / 0.95

V +

0.9

0.85 1

10

[Ry]

100

1000

Figure 4.5: Inverse bremsstrahlung absorption coecient as a function of the photon energy for the solar core plasma. Results including self-energy (solid curve) and self-energy plus vertex corrections (dashed curve) are shown, normalized to the Born approximation B ().

Whereas the rst diagram is included in the second iteration of equation 4.1, the second diagram represents a vertex correction, which is not included. Note, that the upper propagator lines in equation (4.2) represent the ions, while the base lines represent the electrons. It was found, that the vertex term gives a correction of at most 20%, see gure 4.6. Here, the rst and second iteration of the self-consistent equation for the self-energy (4.1) (dashed and dotted curves, respectively) are shown, as well as the converged result (solid curve). The inset shows the comparison between the iterations of the self-consistent selfenergy and the vertex correction in second order, see equation (4.2). The vertex term gives a correction of at most 20%. For the plasma parameters ne = 1021 cm3 and kB T = 1 Ry are chosen. A detailed discussion can be found in Ref. [FRW07]. In conclusion, it was shown, that vertex corrections are important for the calculation of optical properties because they partially compensate pure self-energy eects. This was seen in the case of the force-force correlation function, where the vertex correction compensates the enhancement feature at intermediate energies that is observed when taking only self-energy eects into account. On the other hand, also the self-energy itself is reduced signicantly (20% at the considered parameters) by the vertex counterterm.

65

0

GW , 1. iteration 0 GW , 2. iteration 0 GW , stable

0

Im (0.1,) [Ry]

-0.05

0.01 0.005

-0.1

Vertex correction

0 -0.005

-0.15 0

-0.01 0.9

0.5

[Ry]

1.5

Figure 4.6: Imaginary part of the self-energy at p = 0.1 a1 for electron-ion scattering, B see equation (4.1). The rst and second iteration (dashed and dotted curves, respectively) are shown as well as the converged result (solid curve). The inset shows the comparison between the iterations of the self-consistent self-energy and the vertex correction in second order, see equation (4.2). The vertex term gives a correction of at most 20%. Plasma parameters: ne = 1021 cm3 , kB T = 1 Ry.

4.5

4.5.1

Introduction

The single-particle spectral function A(p, ) gives the probability density to nd a particle at energy for a given momentum p. Starting from the spectral function, a number of interesting plasma observables can be determined, such as its equation of state [VSK04], transport properties like thermal and electric conductivity [RRRW00], optical properties [FRW07], see also the preceding sections, or the stopping power of fast particles traversing a medium [GSK96, MR96]. This is achieved by calculating higher order correlation functions, i.e. convolutions of single-particle spectral functions and vertex functions. The analytic structure of the spectral function, i.e. the location of resonances and their width, is determined by the real part and the imaginary part of the self-energy (p, ), respectively. In the case that the spectral function contains only a single peak, one speaks

66

Selected problems

of a broadened quasi-particle spectral function. The position of the quasi-particle resonance is given by the solution of the quasi-particle dispersion E(p) = p + Re (p, ) |=E(p) . 2m

2 2

(4.3)

The imaginary part of the self-energy at the quasi-particle dispersion Im (p, (p)) is a measure for the width of the quasi-particle resonance. Often, this quantity is referred to as the quasi-particle damping width or inverse of the quasi-particle life time. The quasiparticle life time is the average time in which the particle is in a dened momentum state before its momentum is changed due to collisions with other particles in the system. The GW (0) approximation for the self-energy is a well established method, which was already extensively discussed in section 2.4. It allows the self-consistent calculation of the self-energy beyond the Hartree-Fock level, i.e beyond the mean-eld approximation, where particle correlations are neglected. The GW (0) approximation yields a set of coupled integral equations for the electron self-energy and the spectral function A(p, ). The screened interaction potential is described by the random phase approximation for the polarization function. Furthermore, the chemical potential has to be adjusted to the spectral function to ensure conservation of the number of particles. Rigorously, this set of equations is solved only numerically, using an iteration algorithm that starts from a suitable initialization of the spectral function or the self-energy. Usually, the algorithm converges after about 10 iterations, i.e. the functions do not change any more from iteration to iteration.

4.5.2

Numerical Results

As a representative example, the spectral function for a one-component electron plasma is studied at conditions similar to the ones found at the solar core, i.e. the temperature is kB T = 1000 eV, the density is 7 1025 cm3 . A three-dimensional representation is given in gure 4.7. The spectral function is shown as a function of the particle energy (frequency) shifted by the chemical potential, + e for ve dierent momenta. At low momenta, plasmaron resonances show up as shoulders in the spectral function, shifted from the central quasi-particle peak by roughly the plasma frequency, which at the present density is pl = 23 Ry. At increased momenta a single, broadened quasi-particle peak is obtained. The spectral function at various densities is studied in gure 4.8. The temperature is the same for all calculations, kB T = 1360 eV, the momentum is xed at p = 0, where the collective eects become most visible. Again, at the solar core density n = 7 1025 cm3 , the characteristic central peak (quasi-particle peak) is accompanied by plasmaron satellites (indicated by arrows), shifted by the plasma frequency pl = 23 Ry from the quasi-particle maximum. The quasi-particle peak itself is shifted from the single-particle dispersion E(0) = by roughly 7 Ry and considerably broadened. Both, shift and width, decrease as the density is lowered. This can be understood a consequence of the decreased coupling constant n1/3 . In the limit of vanishing density, the spectral function converges towards

67

0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 01 2 mo 3 4 me ntu 5 6 7 mp [a -1 8 910 B ]

A(p,) [1/Ry]

Figure 4.7: Electron spectral function at solar core conditions (kB T = 1000 eV, ne = 71025 cm3 ). The spectral function is shown as a function of the particle energy shifted by the chemical potential e for ve dierent momenta. At low momenta, plasmaron satellites show up in the spectral function, whereas at increased momenta a single, broadened quasiparticle peak is obtained. The solid black line at the bottom denotes the free single-particle dispersion = 2 p2 /2m . The chemical potential is = 1880 eV (138 Ry). an on-shell delta distribution A(p, ) 2((p) )) with the free particle dispersion (p) = 2 p2 /2m .

4.5.3

It is worth to look for an analytic expression for the quasi-particle damping width p = Im (p, E(p)) for the following reasons: On the one hand, analytic results can be used as benchmarks for the numerical algorithm and give accurate results in limiting cases that are hard to access numerically. In particular, this applies to the case of large momenta, and low densities or low temperatures. In each case, the spectral function features very narrow resonances that are dicult to resolve on the numerical grid. Second, analytical expressions permit a deeper insight into the relevant physical processes that lead e.g. to the nite spectral width and shift of the spectral function. Naturally, analytical results can only be obtained through approximations to the exact expressions and equations that otherwise can only be solved numerically. These approxi-

68

Selected problems

1.2 spectral function A(p=0,) [Ry] 1 0.8 0.6 0.4 0.2 0 -40 -20 0

n=7.0x1025 cm-3 n=7.0x1024 cm-3 n=7.0x1023 cm-3 n=7.0x1022 cm-3 n=7.0x1021 cm-3

20

40

energy h+ [Ry]

Figure 4.8: Spectral function at vanishing momentum p = 0 as a function of energy + . The dierent curves correspond to dierent densities between 7 1021 cm3 and 7 1025 cm3 . The temperature is kept constant at kB T = 1360 eV. At low density, the spectral function becomes a narrow, unstructured quasi-particle resonance, while at increased density, the plasmaron satellites appear in the wings of the central peak (arrows). mations are only valid for certain plasma parameters. Here, I focus on the case of weakly coupled, classical plasmas, i.e. plasmas at low density and high temperature. As a further approximation, only temperatures that are large against the binding energy of hydrogen are considered, kB T > 13.6 eV = 1 Ry. Thus, bound states do not need to be taken into account in the calculation of the self-energy. At lower temperatures, bound states become important and must be included in the self-energy. This can be achieved by use of the t-matrix [KKER86]. The limit of weak coupling is of special interest, since here a vanishing self-energy is expected, such that the spectral function converges to a delta-distribution. This behaviour was also seen in the numerical results, see gure 4.8. At small coupling parameters, one might expect that the self-energy is already given by the rst iteration of the corresponding integral equation (2.89) starting from the quasi-particle limit, i.e. replacing the spectral function on the r.h.s. by an on-shell delta-distribution. This so-called quasi-particle damping was investigated e.g. by Fennel and Wilfer in Ref. [FW74]. However their result is not satisfying. The quasi-particle damping is far from the self-consistent numerical solution. Secondly, an expression is obtained, which is independent of density. This means, that a nite quasi-particle life time is obtained regardless of the density and even in the vacuum

69

limit n = 0. This is unphysical, since free particles are always in a dened momentum state. This paradox can only be resolved if the self-consistency implied in the equation for the self-energy is consequently elaborated also in the analytical treatment. In Refs [For08] and [For09], the analytic solution for the self-consistent quasi-particle damping is derived. Expressions are found, that give the quasi-particle damping for a weakly coupled onecomponent plasma in the limit of vanishing momentum p = 0 and in the asymptotic case p . By means of a two-point Pad approximation, an interpolation formula is e constructed that covers the complete p range. The diculty in solving the integrals in equation (2.89) is mainly caused by the term Im 1 (q, ). Here, suitable approximations have to be used. It was found, that at small momenta p, the one-loop approximation (Born approximation) for the screened potential leads to the same form of the self-energy than the full RPA expression. Replacing the inverse dielectric function by the one-loop approximation and applying certain approximations to the remaining integrands leads to the analytic form of the self-energy valid at small momenta, (p, ) =

2 2

p /2m + 2

2 2

sign(

p /2m + )

+ i

2 2

p /2m +

e2 kB T 4 0

1/2

. (4.4)

For the derivation, see Ref. [For08]. Equation (4.4) reproduces the numerical data for the self-energy and the spectral function at vanishing momentum with an accuracy of better than 7%. This is true as long as no plasmaron resonances appear in the spectral function. At the single-particle dispersion = E(p), equation (4.4) simplies to 0 := Im (p, E(p) = e2 kB T /4

0

(4.5)

the so-called non-collective damping width. It can also be expressed by the plasma coupling parameter as 0 = (33 )1/4 kB T . (4.6) The range of plasma parameters, where equation (4.4) is valid, can be estimated from the ratio of the plasma frequency (i.e. the energy scale where plasmaron peaks appear) and the non-collective damping width. The plasmaron peaks are hidden in the broadened quasi-particle peak, as long as the parameter = pl /0 < 1. The parameter is identical to the square-root of the Debye screening parameter, measured in inverse Bohr radii, (4.7) = aB . The non-collective damping width 0 behaves proportional to n1/4 . It vanishes in the limit n 0 as expected from rst principle arguments. The high accuracy by which this simple expression reproduces the numerical results is illustrated in gure 4.9.

70

3/4

Selected problems

10

-1

=1 =0.5

=0.5

=1

=1

10

-2

3/4

=0.5

=1

=1

=1

10

-3

10

-2

10

-1

Figure 4.9: Numerical results for the quasi-particle damping width at vanishing momentum, Im (0, E(0))/kB T , normalized to the temperature as a function of the plasma coupling parameter . Results are shown for three dierent temperatures, kB T = 10 Ry (solid), kB T = 100 Ry (dashed), and kB T = 1000 Ry (dash-dotted). At small coupling parameters and for = pl /0 = aB < 1, the numerical data comply with the derived scaling law for the non-collective damping 0 3/4 , see equation 4.6. When becomes larger than 1, the plasmaron satellites appear in the spectral function and the width of the quasi-particle peak decreases accordingly. The numerical data for the quasi-particle damping at vanishing momentum, 0 , normalized to the plasma temperature is shown as a function of the plasma coupling parameter for three dierent temperatures. At small values of (see equation (4.7)), the numerical data nicely follow the analytic expression (4.6), i.e 0 /kB T 3/4 . Only when becomes larger than 1, the numerical data deviate from the scaling law. This is due to the appearance of collective modes in the spectral function. Correspondingly, spectral weight is transferred from the central peak into the resonances and the quasi-particle resonance narrows. The regime of validity for the quasi-particle damping width p as given by (4.4) is illustrated in gure 4.10. The grey colored area is limited by the green line, here the parameter = aB is equal to 1, i.e. pl = 0 . At higher densities, the plasmaron resonances separate from the central peak and appear as distinct features in the spectral function. Furthermore, the grey region is limited by the horizontal line at kB T = 1 Ry = 13.6 eV, i.e. the binding energy of the ground state hydrogen atom. Below this line, bound

71

states in the plasma become frequent and have to be considered in the self-energy as well, see. e.g. the work by Schepe [SSTH98] and Schmielau [Sch01]. In gure 4.10, the lines where the plasma degeneracy parameter (red line) and the plasma coupling parameter are equal to 1, are shown as well. The present investigation mainly focusses on non-degenerate and weakly coupled plasmas. For the low density limit, an analytic formula for the self-energy was found, which accurately describes the damping of quasi-particle states up to densities where = 1 a1 and at temperatures above the B ionization energy of hydrogen. An important issue is the question how the self-energy and the spectral function behave at increased coupling ( > 1) and for degenerate systems ( < 1). The latter case, i.e. the spectral function for completely degenerate systems at T = 0 has been investigated intensively e.g. by Barth and von Holm [BH96, HB98] applying also the GW (0) approximation as well as the GW approximation. In general, the spectral function at such conditions is characterized by a very narrow quasi-particle peak (at the Fermi surface p = kF the quasi-particle peak is in fact undamped), and two plasmaron satellites, which vanish at increasing momentum. The onset of this behaviour can already be seen in the calculations for the solar core parameters shown in gure 4.8 (solid curve). At increased density, i.e. > 1 a1 , spectral B weight is transferred from the quasi-particle peak into the plasmaron satellites. Furthermore, the quantum degeneracy of electrons starts to play a role. In particular, collisions at small transfer momenta become less probable due to Pauli blocking. Therefore, the quasi-particle peak as well as the plasmaron resonances become more and more narrow as the density further increases or the temperature decreases. Summarizing, a new analytic expression for the single-particle self-energy, valid at small momenta and for < 1 a1 has been found that predicts a power-law behaviour of the B quasi-particle damping p n1/4 . This scaling law reproduces the numerical data, obtained from accurate GW (0) self-energy calculations, within error-bars of a few percent. A long standing issue in many-body physics, namely the unphysical behaviour of the self-energy in quasi-particle approximation has been resolved. The problem could be traced back to the inconsistent treatment of the self-energy in the corresponding integral equation (2.89).

4.5.4

Equation (4.4) reproduces the GW (0) self-energy only at small momenta. In particular, it does not describe the decrease of the quasi-particle damping at increased momentum. Following the same strategy that led to the quasi-particle damping at low p, explained in the previous subsection, the case of large p is treated in Ref. [For09]. The integral equation for Im (p, E(p)) is solved after replacing the inverse dielectric function by the plasmon pole approximation [KKER86]. In this way, the appearance of plasmaron resonances in the spectral function is accounted for. In addition, an expression for quasi-particle damping at arbitrary momenta is obtained as a two-point Pad approximation [Mag82] that interpolates between the small p limit e that was discussed in the previous section, see equation (4.5), and the large p limit. The

72

Selected problems

10

10 kBT [eV] 10

= 1 a-1 B

0 = (e kBT/40)

kBT = 1 Ry =1 =1

21

1/2

=1 =1

10

10 20 10

= 1 a-1 B

22

10

10

10 10 -3 ne [cm ]

23

24

10

25

10

26

10

27

Figure 4.10: Density-temperature plane and isolines of various plasma parameters. The grey colored region marks the validity region of the analytic scaling law (4.5) for the quasiparticle damping width. following form is obtained,

Pad p e =

a0 + a1 p (p) , 1 + b1 p + b2 p 2 3/2 a0 = T , a1 = , b1 = , b2 = , 2 2 2T 2T ln (p) ln (p) ln (p) 3 ln2 (p) ln2 (p) ln3 (p) (p) = (p) ln (p) + + + + , 3 (p) 2 (p) (p) 2 3 (p) 2 2 (p) 3 3 (p) (p) = ln(e + 2 2 p exp(A/T )/T ) , pl A = 1.3357 [nB (pl ) exp(pl /T ) nB (pl ) exp(pl /T )] . 2

(4.8)

Here, Rydberg units have been used, i.e. e2 /4 0 = 2, = 1, me = 1/2, and kB = 1. The quasi-particle damping width is plotted in gure 4.11 as a function of p (dashed curve). Density and temperature of the plasma are xed at n = 7 1020 cm3 and kB T = 100 eV, respectively. The solid curve is obtained by tting the numerical data for the

73

-0.2

GW -fit Interpolation formula

(0)

20

40 60 -1 momentum p [aB ]

80

100

Figure 4.11: Eective quasi-particle damping p as a function of momentum p for plasma density n = 7 1020 cm3 and temperature T = 100 eV. The t-parameters for the Gaussian t to the full GW (0) -calculations are given as solid line, the dashed line denotes the analytic interpolation formula (4.8). spectral function using a Gaussian of the form AG (p, ) = ( (p))2 2 exp 2 2 p 2p . (4.9)

The Pad formula agrees with the t data within an error bar of less than 10%. At p = 0, e the quasi-particle damping starts at the nite value 0 = e2 kB T /8 0 /21 and slowly decreases towards higher p, such that in the limit p , a narrow, on-shell quasi-particle spectral function is obtained. This is also illustrated in gure 4.12. Here, the full numerical solution for the spectral function in GW (0) approximation (solid curve) is compared to the Gaussian ansatz (4.9) (dashed curve) using the quasi-particle damping width p from equation (4.8). Results are shown for three dierent momenta, p = 0 (a), p = 50 a1 (b), and p = 100 a1 (c). The plasma density is ne = 7 1020 cm3 , B B the temperature is kB T = 100 eV. The Debye screening parameter is = 1.9 102 a1 , B

This value is obtained from equation 4.5 by multiplication with the factor /2 2. In this way, the maximum of the Gaussian function matches the maximum of the Lorentzian form of the spectral function (2.75).

1

74

Selected problems

i.e. we are in the range of validity of the non-collective damping width, as discussed in the previous subsection. At small momenta, the Gaussian does not completely reproduce the numerical solution, although the height of the maximum and the overall width coincide at least approximately. The steep rise in the wings of the numerical solution and the plateau around + e = 0 are not reproduced by the Gaussian ansatz. These structures result from the plasmaron resonances that merge with the central peak at the considered plasma parameters. At higher momenta an almost perfect agreement between the numerical data and the Gaussian spectral function is observed. Here, the plasmaron resonances are damped out and only the broadened quasi-particle peak remains. In conclusion, an analytic expression for the quasi-particle damping width was presented, that accurately reproduces the numerical data for the spectral function at arbitrary momentum p by use of a Gaussian ansatz. In the regime of validity (see gure 4.10), this result greatly facilitates the calculation of higher order correlation functions via the spectral function. An example is the polarization function which yields the optical properties of the system. Further interesting quantities, e.g. transport coecients [RRRW00], the stopping power[GSK96] or the equation of state [VSK04] can be accessed as well. The spectral function does not need to be recalculated by solving the GW (0) equations for each density and temperature but can be approximated by the Gaussian ansatz (4.9) in combination with the analytic expression for the quasi-particle damping p , equation (4.8). Furthermore, in those cases, that the spectral function needs to be known with high accuracy (i.e. better than 10%), the Gaussian ansatz should be used as initialization for the iterative algorithm. Convergence is then reached after only two or three iterations.

10 GW(0) Gauss fit

75

p=0 aB-1

0 -3 -2 -1 0 1 frequency +e [Ry] 2 3

(a) p = 0

10 GW(0) Gauss fit

p=50 aB-1

0 2497

2498

2502

2503

(b) p = 50 a1 B

10 GW Gauss fit spectral function A(p,) [1/Ry] 8

(0)

p=100 aB-1

0 9997

9998

10002

10003

(c) p = 100 a1 B

Figure 4.12: Spectral function in GW (0) -approximation (solid lines) and Gaussian ansatz (dashed lines) with quasi-particle damping width p taken from equation (4.8) for three dierent momenta. Plasma parameters: n = 7 1020 cm3 , T = 100 eV. The plasma coupling parameter is = 2.1 102 , the degeneracy parameter is = 3.5 102 , the Debye screening parameter is = 1.9 102 a1 . B

76

Summary

Summary

The present work is devoted to the process of bremsstrahlung occurring in a correlated medium, such as a dense plasma. A many-body quantum eld theoretical approach is presented. In this frame, the bremsstrahlung emissivity and the absorption due to inverse bremsstrahlung are expressed via the force-force correlation function or, equivalently, the polarization function. This allows for a microscopical description based on the Green function technique. In this way, many-particle eects are included in a systematic and intuitive manner. The key quantity within the many-particle theoretical description is the single-particle spectral function. With its help, single-particle properties can be calculated as moments of the spectral function. Higher order correlation functions, such as the polarization function or the two-particle Green function are given by convolution products of the single-particle spectral function, and the vertex function, which is itself related to the spectral function via the Bethe-Salpeter equation, see section 2.2. The inuence of correlations is described by the complex single-particle self-energy. The exact determination of the self-energy requires the rigorous solution of the Dyson-Schwinger equations which amounts to the exact diagonalization of the many-particle Hamiltonian. Appropriately chosen approximations have to be performed, depending on the special problem that is regarded. Here, special attention was payed to the question, how the bremsstrahlung emission and inverse bremsstrahlung absorption are modied due to correlations among plasma constituents. Both electron-ion and electron-electron correlations were analysed. The results are given as a Gaunt factor, i.e. a correction factor that Kramers formula for bremsstrahlung has to be multiplied with. As a well known example for the inuence of strong electron-ion correlations on the bremsstrahlung emission, the Landau-Pomeranchuk-Migdal eect was considered, i.e. the suppression of bremsstrahlung at small photon energies due to multiple scattering on different ions whilst the emission of the bremsstrahlung photon (c.f. section 4.3). Being theoretically predicted and experimentally conrmed in the case of ultrarelativistic electron scattering in dense targets, a quantitative analysis of this question for the case of thermal bremsstrahlung from a plasma was still missing. In our calculations, where we took into account the successive scattering via resummation of the self-energy in Born approximation, we obtained a similar eect than the LPM theory. At small photon energies, the emission spectrum is suppressed compared to the Bethe-Heitler formula, i.e. the Born approximation. The suppression can become signicantly large, in the case of a plasma near solar core conditions, 15% suppression have been observed. The suppression increases systematically as a function of the width of the spectral function, i.e. the imaginary part of the self-energy. As the self-energy, the suppression reaches a maximum in the regime of moderately coupled ( 1) and moderately degenerate ( 1) plasmas, where also the imaginary part of the self-energy peaks as a function of density and temperature. At large photon energies, the results converge towards the Bethe-Heitler result for bremsstrahlung. At an intermediate energy, slightly

Summary

77

above the plasma frequency, an enhancement of bremsstrahlung emission was obtained. However, it remains unclear, if this enhancement is an artefact which could result from an inconsistent treatment of the vertex function, i.e. counterparts to the pure self-energy corrections. This assumption is enforced by the observation that the enhancement is reduced if vertex corrections are taken into account on a perturbative level. The question of vertex corrections was furthermore investigated on the level of the selfenergy itself in section 4.4. To this end, the self-energy arising from electron-ion scattering was evaluated in second order. As predicted by rst principle relations (Ward-Takahashi identities), partial compensation of the direct self-energy term through the exchange term was observed. In the considered case, a reduction of 20% was found. In order to study electron-electron correlations, special care has to be taken for their dynamical character. Dynamical screening of the potential and the emergence of collective excitations in the spectral function and the self-energy are well known eects due to electron-electron correlations. A straightforward method to compute the electron self-energy beyond the mean-eld level is the GW approximation. Here, we used the GW (0) variant, i.e. the screened interaction is xed on the level of the random phase approximation. A systematic behaviour of the single-particle spectral function was obtained. In the low density limit, the spectral function consists of a single peak, the quasi-particle peak. Its width decreases slowly with increasing momentum. At large momenta, the spectral function converges towards a narrow on-shell delta function. With increasing density, the spectral function becomes more complex. In this case, coupled electron-plasmon modes (plasmarons) appear as shoulders in the wings of the quasi-particle peak at small momenta. At increased momenta, the plasmarons decay and only the quasi-particle remains, becoming again a narrow on-shell resonance. The simple, single-peak behaviour at low densities and high temperatures was analyzed in more detail. Using a Gaussian ansatz for the spectral function, the quasi-particle damping width was derived by analytic solution of the GW (0) equation. The obtained formulae reproduce the numerical data with less than 10% relative deviation. The quasi-particle damping width scales with density as n1/4 . Thus, in the vacuum limit (n = 0), free particles are obtained consistently. This is in contrast to the quasi-particle approximation to the self-energy which yields a density independent damping width. Using the Sommerfeld expression for the Gaunt factor, bremsstrahlung as a competing process of emission in a Thomson scattering experiment was analyzed in section 4.1. By comparing the bremsstrahlung photon yield and the Thomson scattering cross-section as a function of the plasma temperature and density, threshold conditions for the laser intensity and the detector resolution were determined. In particular, the case of a Thomson scattering experiment at the free electron laser facility Hamburg (FLASH), operating at XUV wavelengths, was considered. It was found, that the typically available FEL intensities of about 1012 W/cm2 and moderate detector resolutions of / 102 are sucient for the Thomson scattering signal to exceed the thermal bremsstrahlung background. The paramount importance for bremsstrahlung as a plasma diagnostic tool was demonstrated in the context of an experiment aiming at the production of a solid density, homoge-

78

Outlook

neously and volumetrically heated plasma. Ultrashort FEL pulses were focussed on a solid Al target and the resulting plasma radiation was measured in an XUV spectrometer. Both bremsstrahlung and line components of the spectrum concordantly yield a plasma temperature of about 40 eV at an electron density of 4.2 1022 cm3 . Thus, warm dense matter conditions were accessed. This experiment established a new method for the production of warm dense matter under dened laboratory conditions using FEL radiation.

Outlook

Further investigations of the single-particle self-energy for dense plasmas are currently performed. The focus is on the microscopical description of plasmas at arbitrary degeneracy. This is of large interest e.g. to trace the excitation of a hot dense plasma by interaction of intense laser radiation with cold solid state material, such as the one outlined above. Having the spectral function disposable over the complete range of density and temperature, a unied picture of the absorption in the cold target, the dynamic evolution of the plasma and also recombination processes could be given. Other open questions in this eld are related to the consistent treatment of the vertex. Here, it is worth to consider strategies that are successfully applied in quantum eld theory, notably in hadron physics. The basic idea is to postulate a vertex function rst, constrained by sum-rules and identities such as the Ward-Takahashi identities [BC80a, BC80b, EAC+ 08]. Subsequently, the propagators and self-energies are obtained by functional dierentiation of the vertex function. In this way, the Ward-Takahashi identities are automatically fullled. Furthermore, this ansatz being -derivable, conservation of energy, momentum, and particle number is assured on the level of the expectation values. An application to the problem of particle beam-plasma interaction in a hot QED plasma, following these lines, was recently discussed by Morozov et al. [MR06]. As has been demonstrated, bremsstrahlung plays an important role in the eld of plasma diagnostics of warm dense matter. Thus, it is obligatory to advance also the theory of bremsstrahlung. As an important result of this work, it should be kept in mind that many-particle eects can signicantly modify the bremsstrahlung emission spectrum or the inverse bremsstrahlung absorption spectrum. The detection of features such as the suppression at low photon energies or the enhancement above the plasma frequency oer new possibilities also for plasma diagnostic techniques.

Bibliography

[A+ 06] V. Ayvazyan et al., First operation of a free-electron laser generating GW power radiation at 32 nm wavelength, Eur. Phys. J. D 37, 297 (2006), doi:10.1140/epjd/e2005-00308-1. http://dx.doi.org/10.1140/ epjd/e2005-00308-1. W. Ackermann, G. Asova, V. Ayvazyan, A. Azima, N. Baboi, J. Bhr, a V. Balandin, B. Beutner, A. Brandt, A. Bolzmann, R. Brinkmann, O. I. Brovko, M. Castellano, P. Castro, L. Catani, E. Chiadroni, S. Choroba, A. Cianchi, J. T. Costello, D. Cubaynes, J. Dardis, W. Decking, H. DelsimHashemi, A. Delserieys, G. D. Pirro, M. Dohlus, S. Dsterer, A. Eckhardt, u H. T. Edwards, B. Faatz, J. Feldhaus, K. Flttmann, J. Frisch, L. Frhlich, o o T. Garvey, U. Gensch, C. Gerth, M. Grler, N. Golubeva, H.-J. Grabosch, o M. Grecki, O. Grimm, K. H. U. Hahn, J. H. Han, K. Honkavaara, T. Hott, M. Hning, Y. Ivanisenko, E. Jaeschke, W. Jalmuzna, T. Jezynski, R. Kamu mering, V. Katalev, K. Kavanagh, E. T. Kennedy, S. Khodyachykh, K. Klose, V. Kocharyan, M. Krfer, M. Kollewe, W. Koprek, S. Korepanov, D. Kostin, o M. Krassilnikov, G. Kube, M. Kuhlmann, C. L. S. Lewis, L. Lilje, T. Limberg, D. Lipka, F. Lhl, H. Luna, M. Luong, M. Martins, M. Meyer, P. Michelato, o V. Miltchev, W. D. Mller, L. Monaco, W. F. O. Mller, O. Napieralski, o u O. Napoly, P. Nicolosi, D. Nlle, T. Nuez, A. Oppelt, C. Pagani, R. Paparella, o n N. Pchalek, J. Pedregosa-Gutierrez, B. Petersen, B. Petrosyan, G. Petrosyan, L. Petrosyan, J. Pger, E. Plnjes, L. Poletto, K. Pozniak, E. Prat, D. Proch, u o P. Pucyk, P. Radclie, H. Redlin, K. Rehlich, M. Richter, M. Roehrs, J. Roensch, R. Romaniuk, M. Ross, J. Rossbach, V. Rybnikov, M. Sachwitz, E. L. Saldin, W. Sandner, H. Schlarb, B. Schmidt, M. Schmitz, P. Schmser, J. R. u Schneider, E. A. Schneidmiller, S. Schnepp, S. Schreiber, M. Seidel, D. Sertore, A. V. Shabunov, C. Simon, S. Simrock, E. Sombrowski, A. A. Sorokin, P. Spanknebel, R. Spesyvtsev, L. Staykov, B. Steen, F. Stephan, F. Stulle, H. Thom, K. Tiedtke, M. Tischer, S. Toleikis, R. Treusch, D. Trines, I. Tsakov, E. Vogel, T. Weiland, H. Weise, M. Wellhfer, M. Wendt, I. Will, A. Wino ter, K. Wittenburg, W. Wurth, P. Yeates, M. V. Yurkov, I. Zagorodnov, and K. Zapfe, Operation of a free-electron laser from the extreme ultraviolet to the water window, Nature Photonics 1, 336 (2007), doi:10.1038/nphoton.2007.76. http://dx/doi.org/10.1038/nphoton.2007.76.

[AAA+ 07]

80 [AB84]

Bibliography N. R. Arista and W. Brandt, Dielectric response of quantum plasmas in thermal equilibrium, Phys. Rev. A 29, 1471 (1984), doi:10.1103/PhysRevA.29. 1471. http://dx.doi.org/10.1103/PhysRevA.29.1471.

[ABSB+ 97] P. L. Anthony, R. Becker-Szendy, P. E. Bosted, M. Cavalli-Sforza, L. P. Keller, L. A. Kelley, S. R. Klein, G. Niemi, M. L. Perl, L. S. Rochester, and J. L. White, Bremsstrahlung suppression due to the Landau-PomeranchukMigdal and dielectric eects in a variety of materials, Phys. Rev. D 56, 1373 (1997), doi:10.1103/PhysRevD.56.1373. http://dx.doi.org/10.1103/ PhysRevD.56.1373. [AG98] F. Aryasetiawan and O. Gunnarsson, The GW method, Rep. Prog. Phys. 61, 237 (1998), doi:10.1088/0034-4885/61/3/002. http://dx.doi.org10.1088/ 0034-4885/61/3/002. H. Alfvn, Cosmical Electrodynamics, American Journal of Physics 28, 613 e (1960), doi:10.1119/1.1935919. http://dx.doi.org/10.1119/1.1935919. A. Alastuey and A. Perez, Virial Expansion of the Equation of State of a Quantum Plasma, Europhys. Lett. 20, 19 (1992), doi:10.1209/0295-5075/20/ 1/004. http://dx.doi.org/10.1209/0295-5075/20/1/004. M. Abramowitz and A. Stegun (eds.), Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables (Dover Publications, New York, 1970), 9th edn. F. Bagenal, Giant Planet Magnetospheres, Ann. Rev. Earth Planet. Sci. 20, 289 (1992), doi:10.1146/annurev.ea.20.050192.001445. http://dx.doi.org/ 10.1146/annurev.ea.20.050192.001445. G. Baym, Self-Consistent Approximations in Many-Body Systems, Phys. Rev. 127, 1391 (1962), doi:10.1103/PhysRev.127.1391. http://dx.doi.org/10. 1103/PhysRev.127.1391. J. S. Ball and T.-W. Chiu, Analytic properties of the vertex function in gauge theories. I, Phys. Rev. D 22, 2542 (1980), doi:10.1103/PhysRevD.22.2542. http://dx.doi.org/10.1103/PhysRevD.22.2542. J. S. Ball and T.-W. Chiu, Analytic properties of the vertex function in gauge theories. II, Phys. Rev. D 22, 2550 (1980), doi:10.1103/PhysRevD.22.2550. http://dx.doi.org/10.1103/PhysRevD.22.2550. J. D. Bjorken and D. Drell, Relativistic Quantum Fields (McGraw-Hill, New York, 1965).

[Alf60] [AP92]

[AS70]

[Bag92]

[Bay62]

[BC80a]

[BC80b]

[BD65]

Bibliography [BDFR02]

81

H. A. Baldis, J. Dunn, M. E. Foord, and W. Rozmus, Thomson scattering diagnostic of solid density plasmas using x-ray lasers, Rev. Sci. Instrum. 73, 4223 (2002), doi:10.1063/1.1515386. http://dx.doi.org/10.1063/1.1515386. J. C. Brown, A. G. Emslie, G. D. Holman, C. M. Johns-Krull, E. P. Kontar, R. P. Lin, A. M. Massone, and M. Piana, Evaluation of Algorithms for Reconstructing Electron Spectra from Their Bremsstrahlung Hard X-Ray Spectra, ApJ 643, 523 (2006), doi:10.1086/501497. http://dx.doi.org/10. 1086/501497. T. J. Brvenich, J. Evers, and C. H. Keitel, Nuclear Quantum Optics u with X-Ray Laser Pulses, Phys. Rev. Lett. 96, 142501 (2006), doi:10. 1103/PhysRevLett.96.142501. http://dx.doi.org/10.1103/PhysRevLett. 96.142501. H. Bethe and W. Heitler, On the stopping of fast particles and the creation of positive electrons, Proc. Roy. Soc. A 146, 83 (1934), doi:10.1098/rspa.1934. 0140. http://dx.doi.org/10.1098/rspa.1934.0140. P. J. Brussaard and H. C. van de Hulst, Approximation Formulas for Nonrelativistic Bremsstrahlung and Average Gaunt Factors for a Maxwellian Electron Gas, Rev. Mod. Phys. 34, 507 (1962), doi:10.1103/RevModPhys.34.507. http://dx.doi.org/10.1103/RevModPhys.34.507. U. von Barth and B. Holm, Self-consistent GW0 results for the electron gas: Fixed screened potential W0 within the random-phase approximation, Phys. Rev. B 54, 8411 (1996), doi:10.1103/PhysRevB.54.8411. http://dx.doi. org/10.1103/PhysRevB.54.8411. G. Baym and L. P. Kadano, Conservation Laws and Correlation Functions, Phys. Rev. 124, 287 (1961), doi:10.1103/PhysRev.124.287. http://dx.doi. org/10.1103/PhysRev.124.287. V. N. Baier and V. M. Katkov, Radiation from polarized electrons in oriented crystals at high energy, Nucl. Instrum. Methods Phys. Res., Sect. B 234, 106 (2005), doi:10.1016/j.nimb.2005.01.006. http://dx.doi.org/10.1016/ j.nimb.2005.01.006. D. B. Blaschke, A. V. Prozorkevich, C. D. Roberts, S. M. Schmidt, and S. A. Smolyansky, Pair Production and Optical Lasers, Phys. Rev. Lett. 96, 140402 (2006), doi:10.1103/PhysRevLett.96.140402. http://dx.doi.org/10.1103/ PhysRevLett.96.140402. J. N. Bahcall, M. H. Pinsonneault, and G. J. Wasserburg, Solar models with helium and heavy-element diusion, Rev. Mod. Phys. 67,

[BEH+ 06]

[BEK06]

[BH34]

[BH62]

[BH96]

[BK61]

[BK05]

[BPR+ 06]

[BPW95]

82

[BRS+ 03]

A. Brantov, W. Rozmus, R. Sydora, C. E. Capjack, V. Y. Bychenkov, and V. T. Tikhonchuk, Enhanced inverse bremsstrahlung heating rates in a strong laser eld, Phys. Plasmas 10, 3385 (2003), doi:10.1063/1.1586917. http:// dx.doi.org/10.1063/1.1586917. J. W. Braun, Q. Su, and R. Grobe, Numerical approach to solve the time-dependent Dirac equation, Phys. Rev. A 59, 604 (1999), doi:10.1103/ PhysRevA.59.604. T. Bornath, M. Schlanges, P. Hilse, and D. Kremp, Nonlinear collisional absorption in dense laser plasmas, Phys. Rev. E 64, 026414 (2001), doi:10.1103/ PhysRevE.64.026414. http://dx.doi.org/10.1103/PhysRevE.64.026414.

[BSG99]

[BSHK01]

[BTADS97] F. Bechstedt, K. Tenelsen, B. Adolph, and R. Del Sole, Compensation of Dynamical Quasiparticle and Vertex Corrections in Optical Spectra, Phys. Rev. Lett. 78, 1528 (1997), doi:10.1103/PhysRevLett.78.1528. http://dx. doi.org/10.1103/PhysRevLett.78.1528. [Cas48] H. Casimir, On the attraction between two perfectly conducting plates, Proc. K. Ned. Akad. Wetensch. 51, 793 (1948). http://relativity.livingreviews. org/refdb/record/4231. S. Chandrasekhar, Radiation Transfer (Dover Publications, New York, 1960). J. Chihara, Interaction of photons with plasmas and liquid metals - photoabsorption and scattering, J. Phys.: Condens. Matter 12, 231 (2000), doi:10. 1088/0953-8984/12/3/303. http://dx.doi.org/10.1088/0953-8984/12/3/ 303. I. P. Christov, M. M. Murnane, and H. C. Kapteyn, High-Harmonic Generation of Attosecond Pulses in the Single-Cycle Regime, Phys. Rev. Lett. 78, 1251 (1997), doi:10.1103/PhysRevLett.78.1251. http://dx.doi.org/10. 1103/PhysRevLett.78.1251. R. Cauble and W. Rozmus, The inverse bremsstrahlung absorption coecient in collisional plasmas, Phys. Fluids 28, 3387 (1985), doi:10.1063/1.865338. http://dx.doi.org/10.1063/1.865338. Z.-J. Ding, Self-energy in surface electron spectroscopy: I. Plasmons on a free-electron-material surface, J. Phys.: Condens. Matter 10, 1733 (1998). http://stacks.iop.org/0953-8984/10/1733.

[Cha60] [Chi00]

[CMK97]

[CR85]

[Din98]

Bibliography

83

[DSCC+ 97] L. B. Da Silva, P. Celliers, G. W. Collins, K. S. Budil, N. C. Holmes, T. W. Barbee Jr., B. A. Hammel, J. D. Kilkenny, R. J. Wallace, M. Ross, R. Cauble, A. Ng, and G. Chiu, Absolute Equation of State Measurements on Shocked Liquid Deuterium up to 200 GPa (2 Mbar), Phys. Rev. Lett. 78, 483 (1997), doi:10.1103/PhysRevLett.78.483. http://dx.doi.org/10.1103/ PhysRevLett.78.483. [DW97] S. Drr and A. Wipf, Finite Temperature Schwinger Model with Chirality u Breaking Boundary Conditions, Ann. Phys. NY 255, 333 (1997), doi:10.1006/ aphy.1996.5664. http://dx.doi.org/10.1006/aphy.1996.5664. F. J. Dyson, The S Matrix in Quantum Electrodynamics, Phys. Rev. 75, 1736 (1949), doi:10.1103/PhysRev.75.1736. http://dx.doi.org/10.1103/ PhysRev.75.1736. G. Eichmann, R. Alkofer, I. C. Clot, A. Krassnigg, and C. D. Roberts, Pere spective on rainbow-ladder truncation, Phys. Rev. C 77, 042202 (2008), doi: 10.1103/PhysRevC.77.042202. http://dx.doi.org/10.1103/PhysRevC.77. 042202. G. Elwert, Verschrfte Berechnung von Intensitt und Polarisation im kona a tinuierlichen Rntgenspektrum1, Ann. Phys. 426, 178 (1939), doi:10.1002/ o andp.19394260206. http://dx.doi.org/10.1002/andp.19394260206. J. Faure, Y. Glinec, A. Pukhov, S. Kiselev, S. Gordienko, E. Lefebvre, J.-P. Rousseau, F. Burgy, and V. Malka, A laser-plasma accelerator producing monoenergetic electron beams, Nature 431, 541 (2004), doi:10.1038/nature02963. http://dx.doi.org/10.1038/nature02963. R. P. Feynman and A. R. Hibbs, Quantum Mechanics and Path Integration (MacGraw-Hill, New York, 1965). V. Fortov, I. Iakubov, and A. Khrapak, Physics of Strongly Coupled Plasmas (Oxford University Press, Oxford, 2006). R. Fischer, M. Lein, and C. H. Keitel, Enhanced Recollisions for Antisymmetric Molecular Orbitals in Intense Laser Fields, Phys. Rev. Lett. 97, 143901 (2006), doi:10.1103/PhysRevLett.97.143901. http://dx.doi.org/10.1103/ PhysRevLett.97.143901. C. Fortmann, Self-consistent spectral function for non-degenerate Coulomb Systems and Analytic Scaling Behaviour, J. Phys. A: Math. Theor. 41, 445501 (2008), doi:10.1088/1751-8113/41/44/445501. http://dx.doi.org/ 10.1088/1751-8113/41/44/445501.

[Dys49]

[EAC+ 08]

[Elw39]

[FGP+ 04]

[For08]

84 [For09]

Bibliography C. Fortmann, Single-particle spectral function for the classical one-component plasma, Phys. Rev. E 79, 016404 (2009), doi:10.1103/PhysRevE.79.016404. http://dxx.doi.org/10.1103/PhysRevE.79.016404. C. Fortmann, R. Redmer, H. Reinholz, G. Rpke, W. Rozmus, and A. Wiero ling, Bremsstrahlung vs. Thomson scattering in VUV-FEL plasma experiments, High Energy Density Phys. 2, 57 (2006), doi:10.1016/j.hedp.2006.04. 001. http://dx.doi.org/10.1016/j.hedp.2006.04.001. C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling, Bremsstrahlung from o dense plasmas and the Landau-Pomeranchuk-Migdal eect, in Condensed Matter Theory, vol. 20 (edited by J. Clark, R. Pano, and H. Li) (Nova Science, New York, 2005), vol. 20, 317. http://arxiv.org/abs/physics/0502051. C. Fortmann, G. Rpke, and A. Wierling, Optical Properties and Oneo Particle Spectral Function in Non-Ideal Plasmas, Contrib. Plasma Phys. 47, 297 (2007), doi:10.1002/ctpp.200710040. http://dx.doi.org/10.1002/ ctpp.200710040. S. V. Faleev, M. van Schilfgaarde, T. Kotani, F. Lonard, and M. P. Dese jarlais, Finite-temperature quasiparticle self-consistent GW approximation, Phys. Rev. B 74, 033101 (2006), doi:10.1103/PhysRevB.74.033101. http: //dx.doi.org/10.1103/PhysRevB.74.033101. A. L. Fetter and J. D. Walecka, Quantum Theory of Many-Particle Systems (McGraw-Hill, New York, 1971). W. Fennel and H. P. Wilfer, Quasiparticle Lifetimes of Fermi Systems, Ann. Phys. Lpz. 32, 265 (1974), doi:10.1002/andp.19754870406. http://dx.doi. org/10.1002/andp.19754870406. M. Fajardo, P. Zeitoun, and J.-C. Gauthier, Hydrodynamic simulation of XUV laser-produced plasmas, Eur. Phys. J. D 29, 69 (2004), doi:10.1140/epjd/ e2004-00004-8. http://dx.doi.org/10.1140/epjd/e2004-00004-8. J. A. Gaunt, Continuous Absorption, Proc. R. Soc. A 126, 654 (1930), doi: 10.1098/rspa.1930.0034. http://dx.doi.org/10.1098/rspa.1930.0034. D. A. Gurnett and A. Bhattacharjee, Introduction to Plasma Physics (Cambridge University Press, Cambridge, 2005). H. A. Gould and H. E. DeWitt, Convergent Kinetic Equation for a Classical Plasma, Phys. Rev. 155, 68 (1967), doi:10.1103/PhysRev.155.68. http://dx. doi.org/10.1103/PhysRev.155.68.

[FRR+ 06]

[FRRW05]

[FRW07]

[FSK+ 06]

[FW71] [FW74]

[FZG04]

Bibliography [GF96]

85

P. Gibbon and E. Frster, Short-pulse laser-plasma interactions, Plasma Phys. o Control. Fusion 38, 769 (1996), doi:10.1088/0741-3335/38/6/001. http:// dx.doi.org/10.1088/0741-3335/38/6/001. S. H. Glenzer, G. Gregori, R. W. Lee, F. J. Rogers, S. W. Pollaine, and O. L. Landen, Demonstration of Spectrally Resolved X-Ray Scattering in Dense Plasmas, Phys. Rev. Lett. 90, 175002 (2003), doi:10.1103/PhysRevLett.90. 175002. http://dx.doi.org/10.1103/PhysRevLett.90.175002. P. Gibbon, Short Pulse Laser Interaction with Matter (Imperial College Press, London, 2006). S. H. Glenzer, O. L. Landen, P. Neumayer, R. W. Lee, K. Widmann, S. W. Pollaine, R. J. Wallace, G. Gregori, A. Hll, T. Bornath, R. Thiele, V. Schwarz, o W.-D. Kraeft, and R. Redmer, Observations of Plasmons in Warm Dense Matter, Phys. Rev. Lett. 98, 065002 (2007), doi:10.1103/PhysRevLett.98. 065002. http://dx.doi.org/10.1103/PhysRevLett.98.065002. S. H. Glenzer, P. Neumayer, T. Dppner, O. L. Landen, R. W. Lee, R. J. o Wallace, S. Weber, H. J. Lee, A. L. Kritcher, R. Falcone, S. P. Regan, H. Sawada, D. D. Meyerhofer, G. Gregori, C. Fortmann, V. Schwarz, and R. Redmer, Compton Scattering Measurements from Dense Plasmas, J. Phys. Conf. Ser. 112, 032071 (2008), doi:10.1088/1742-6596/112/3/032071. http: //dx/doi.org/10.1088/1742-6596/112/3/032071. W. Greiner and J. Reinhardt, Theoretische Physik, vol. 7 (Harry Deutsch, Thun & Frankfurt am Main, 1984). W. Greiner and J. Reinhardt, Field Quantization (Springer, HeidelbergBerlin, 1996). H. R. Griem, Plasma Spectroscopy (McGraw-Hill, New York, 1964). F. Gross, Relativistic Quantum Mechanics and Field Theory (J. Wiley & Sons, New York, 1993). D. O. Gericke, M. Schlanges, and W.-D. Kraeft, Stopping power of a quantum plasma T-matrix approximation and dynamical screening, Phys. Lett. A 222, 5 (1996), doi:10.1016/0375-9601(96)00654-8. http://dx.doi.org/10. 1016/0375-9601(96)00654-8. C. G. R. Geddes, C. Toth, J. van Tilborg1, E. Esarey, C. B. Schroeder, D. Bruhwiler, C. Nieter, J. Cary, and W. P. Leemans, High-quality electron beams from a laser wakeeld accelerator using plasma-channel guiding, Nature 431, 538 (2004), doi:10.1038/nature02900. http://dx.doi.org/10. 1038/nature02900.

[GGL+ 03]

[GND+ 08]

[GTT+ 04]

86 [Gn95] u [H 06]

Bibliography S. Gnter, Optische Eigenschaften Dichter Plasmen, Habilitation thesis, Rou stock University, Rostock (1995). T. W. Hnsch, Einstein Lecture - Passion for Precision, Ann. Phys. Lpz. a 15, 627 (2006), doi:10.1002/andp.200610215. http://dx.doi.org/10.1002/ andp.200610215. B. Holm and U. von Barth, Fully self-consistent GW self-energy of the electron gas, Phys. Rev. B 57, 2108 (1998), doi:10.1103/PhysRevB.57.2108. http: //dx.doi.org/0.1103/PhysRevB.57.2108. A. Hll, T. Bornath, L. Cao, T. Dppner, S. Dsterer, E. Frster, C. Forto o u o mann, S. Glenzer, G. Gregori, and T. Laarmann, Thomson scattering from near-solid density plasma using soft x-ray free electron lasers, High Energy Density Phys. 3, 120 (2007), doi:10.1016/j.hedp.2007.02.033. http://dx.doi. org/10.1016/j.hedp.2007.02.033. D. H. H. Homann, A. Blazevic, P. Ni, O. Rosmej, M. Roth, N. A. Tahir, A. Tauschwitz, S. Udrea, D. Varentsov, K. Weyrich, and Y. Maron, Present and future perspectives for high energy density physics with intense heavy ion and laser beams, Laser Part. Beams 23, 47 (2005), doi:10.1017/ S0263034605050019. http://dx.doi.org/10.1017/S0263034605050019. L. Hedin, New Method for Calculating the One-Particle Greens Function with Application to the Electron-Gas Problem, Phys. Rev. 139, 796 (1965), doi:10. 1103/PhysRev.139.A796. http://dx.doi.org/10.1103/PhysRev.139.A796. L. Hedin, On correlation eects in electron spectroscopies and the GW approximation, J. Phys.: Condens. Matter 11, R489 (1999), doi:10.1088/0953-8984/ 11/42/201. http://dx.doi.org/10.1088/0953-8984/11/42/201. W. Heitler, The Quantum Theory of Radiation (Dover Publications, New York, 1954). D. H. H. Homann, V. E. Fortov, I. V. Lomonosov, V. Mintsev, N. A. Tahir, D. Varentsov, and J. Wieser, Unique capabilities of an intense heavy ion beam as a tool for equation-of-state studies, Phys. Plasmas 9, 3651 (2002), doi: 10.1063/1.1498260. http://dx.doi.org/10.1063/1.1498260. H. van Hees and J. Knoll, Renormalization in self-consistent approximation schemes at nite temperature: Theory, Phys. Rev. D 65, 025010 (2001), doi: 10.1103/PhysRevD.65.025010. http://dx.doi.org/10.1103/PhysRevD.65. 025010. H. van Hees and J. Knoll, Renormalization in self-consistent approximation schemes at nite temperature. III. Global symmetries, Phys. Rev. D

[HB98]

[HBC+ 07]

[HBN+ 05]

[Hed65]

[Hed99]

[HK01]

[HK02a]

Bibliography

87

66, 025028 (2002), doi:10.1103/PhysRevD.66.025028. http://dx.doi.org/ 10.1103/PhysRevD.66.025028. [HK02b] H. van Hees and J. Knoll, Renormalization of self-consistent approximation schemes at nite temperature. II. Applications to the sunset diagram, Phys. Rev. D 65, 105005 (2002), doi:10.1103/PhysRevD.65.105005. http: //dx.doi.org/10.1103/PhysRevD.65.105005. T. Heinzl, B. Liesfeld, K.-U. Amthor, H. Schwoerer, R. Sauerbrey, and A. Wipf, On the observation of vacuum birefringence, Opt. Commun. 267, 318 (2006), doi:10.1016/j.optcom.2006.06.053. http://dx.doi.org/10.1016/j. optcom.2006.06.053. J. W. Harris and B. Muller, The Search for the Quark-Gluon Plasma, Ann. Rev. Nucl. Part. Sci. 46, 71 (1996), doi:10.1146/annurev.nucl.46.1.71. http: //dx.doi.org/10.1146/annurev.nucl.46.1.71.

[HLA+ 06]

[HM96]

[HRCK+ 07] S. P. Hau-Riege, H. N. Chapman, J. Krzywinski, R. Sobierajski, S. Bajt, R. A. London, M. Bergh, C. Caleman, R. Nietubyc, L. Juha, J. Kuba, E. Spiller, S. Baker, R. Bionta, K. S. Tinten, N. Stojanovic, B. Kjornrattanawanich, E. Gullikson, E. Plnjes, S. Toleikis, and T. Tschentscher, o Subnanometer-Scale Measurements of the Interaction of Ultrafast Soft XRay Free-Electron-Laser Pulses with Matter, Phys. Rev. Lett. 98, 145502 (2007), doi:10.1103/PhysRevLett.98.145502. http://dx.doi.org/10.1103/ PhysRevLett.98.145502. [HRR+ 04] A. Hll, R. Redmer, G. Rpke, , and H. Reinholz, X-ray Thomson scattering o o in warm dense matter, Eur. Phys. J. D 29, 159 (2004), doi:10.1140/epjd/ e2004-00059-5. http://dx.doi.org/10.1140/epjd/e2004-00059-5. A. Hll, C. D. Roberts, and S. V. Wright, A Perspective on Hadron Physics, o in Particles and Fields: X Mexican Workshop, American Institute of Physics Conference Series, vol. 857 (edited by M. A. Prez, L. Urrutia, and L. Villasee qor) (2006), American Institute of Physics Conference Series, vol. 857, 4661, doi:10.1063/1.2359242. http://dx.doi.org/10.1063/1.2359242. G. D. Holman, L. Sui, R. A. Schwartz, and A. G. Emslie, Electron Bremsstrahlung Hard X-Ray Spectra, Electron Distributions, and Energetics in the 2002 July 23 Solar Flare, ApJ Letters 595, L97 (2003), doi: 10.1086/378488. http://dx.doi.org/10.1086/378488. H. D. Hansen, U. I. Uggerhj, C. Biino, S. Ballestrero, A. Mangiarotti, P. Sona, T. J. Ketel, and Z. Z. Vilakazi, Is the Electron Radiation Length Constant at High Energies?, Phys. Rev. Lett. 91, 014801 (2003), doi:10. 1103/PhysRevLett.91.014801. http://dx/doi.org/10.1103/PhysRevLett. 91.014801.

[HRW06]

[HSSE03]

[HUB+ 03]

88 [HUB+ 04]

Bibliography H. D. Hansen, U. I. Uggerhj, C. Biino, S. Ballestrero, A. Mangiarotti, P. Sona, T. J. Ketel, and Z. Z. Vilakazi, Landau-Pomeranchuk-Migdal eect for multihundred GeV electrons, Phys. Rev. D 69, 032001 (2004), doi:10.1103/ PhysRevD.69.032001. http://dx.doi.org/10.1103/PhysRevD.69.032001. T. P. Hughes, Plasmas and Laser Light (John Wiley & Sons, New York, 1975). I. H. Hutchinson, Principles of Plasma Diagnostics (Cambridge University Press, Cambridge, 1987). S. Ichimaru, Statistical Plasma Physics, vol. 1: Basic Principles (Addison Wesley, 1992). S. Ichimaru, Nuclear fusion in dense plasmas, Rev. Mod. Phys. 65, 255 (1993), doi:10.1103/RevModPhys.65.255. http://dx.doi.org/10.1103/ RevModPhys.65.255. C. Itzykson and J.-B. Zuber, Quantum eld theory (McGraw-Hill, New York, 1980). J. D. Jackson, Classical Electrodynamics (J. Wiley & Sons, New York, 1975), 2nd edn. J. Jasny, U. Teubner, W. Theobald, C. Wlker, J. Bergmann, , and F. P. Schfer, A Single-shot spectrograph for the soft x-ray region, Rev. Sci. Instrum. 65, 1631 (1994), doi:10.1063/1.1144851. http://dx.doi.org/10.1063/1.1144851. M. Kaku, Quantum Field Theory - A Modern Introduction (Oxford University Press, New York, 1993). K. Kasahara, Experimental examination of the Landau-Pomeranchuk-Migdal eect by high-energy electromagnetic cascade showers in lead, Phys. Rev. D 31, 2737 (1985), doi:10.1103/PhysRevD.31.2737. http://dx.doi.org/10.1103/ PhysRevD.31.2737. L. P. Kadano and G. Baym, Quantum Statistical Mechanics (W.A. Benjamin Inc., New York, 1962). E. P. Kontar, A. G. Emslie, A. M. Massone, M. Piana, J. C. Brown, and M. Prato, Electron-Electron Bremsstrahlung Emission and the Inference of Electron Flux Spectra in Solar Flares, ApJ 670, 857 (2007), doi: 10.1086/521977. http://dx.doi.org/10.1086/521977. A. Knnies and S. Gnter, Asymmetry and shifts of the L - and the L -line o u of hydrogen, J. Quant. Spectrosc. Radiat. Transfer 52, 825 (1994), doi:10. 1016/0022-4073(94)90170-8. http://dx.doi.org/10.1016/0022-4073(94) 90170-8.

[Kak93] [Kas85]

[KG94]

89

W. D. Kraeft, D. Kremp, W. Ebeling, and G. Rpke, Quantum Statistics of o Charged Particle Systems (Akademie-Verlag, Berlin, 1986). W. J. Karzas and R. Latter, Electron Radiative Transitions in a Coulomb Field., ApJ Suppl. Ser. 6, 167 (1961). http://adsabs.harvard.edu/abs/ 1961ApJS....6..167K. S. Klein, Suppression of bremsstrahlung and pair production due to environmental factors, Rev. Mod. Phys. 71, 1501 (1999), doi:10.1103/RevModPhys. 71.1501. H. W. Koch and J. W. Motz, Bremsstrahlung Cross-Section Formulas and Related Data, Rev. Mod. Phys. 31, 920 (1959), doi:10.1103/RevModPhys.31. 920. http://dx.doi.org/10.1103/RevModPhys.31.920. V. Kaufman and V. C. Martin, Wavelengths and Energy Level Classications for the Spectra of Aluminum (Al I through Al XIII), J. Phys. Chem. Ref. Data 20, 775 (1991). http://www.nist.gov/srd/PDFfiles/jpcrd421.pdf. O. Klein and Y. Nishina, On the Scattering of Radiation by Free Electrons According to Diracs New Relativistic Quantum Dynamics, Z. Phys. 52, 853 (1929). H. A. Kramers, On the theory of X-ray absorption and of the continuous X-ray spectrum, Philos. Mag. 46, 836 (1923). L. M. Krauss, Quintessence: The Mystery of the Missing Mass (Basic Books, New York, 2001). D. Kremp, M. Schlanges, and W.-D. Kraeft, Quantum Statistics of Nonideal Plasmas (Springer, Berlin, Heidelberg, New York, 2005). J. Knoll and D. Voskresensky, Classical and Quantum Many-Body Description of Bremsstrahlung in Dense Matter: LandauPomeranchukMigdal Eect, Ann. Phys. NY 249, 532 (1996), doi:10.1006/aphy.1996.0082. http://dx.doi.org/ 10.1006/aphy.1996.0082. I. Langmuir, Oscillations in Ionized Gases, Proc. Nat. Acad. Sci. 14, 627 (1928). M. Le Bellac, Thermal Field Theory (Cambridge University Press, Cambridge, 1996). O. L. Landen, S. H. Glenzer, M. J. Edwards, R. W. Lee, G. W. Collins, R. C. Cauble, W. W. Hsing, and B. A. Hammel, Dense matter characterization by X-ray Thomson scattering, J. Quant. Spectrosc. Rad. Transf. 71, 465 (2001), ISSN 0022-4073, doi:10.1016/S0022-4073(01)00090-5. http: //dx.doi.org/10.1016/S0022-4073(01)00090-5.

[Kle99]

[KM59]

[KM91]

[KN29]

90 [LH95]

Bibliography W. Lochte-Holtgreven, 3: Evaluation of Plasma Parameters, in Plasma Diagnostics (edited by W. Lochte-Holtgreven) (AIP Press, New York, 1995), 135. J. Lindl, Development of the indirect-drive approach to inertial connement fusion and the target physics basis for ignition and gain, Phys. Plasmas 2, 3933 (1995), doi:10.1063/1.871025. http://dx.doi.org/10.1063/1.871025. L. D. Landau and E. M. Lifschitz, Lehrbuch der Theoretischen Physik, vol. 1 Mechanik (Harry Deutsch, Thun & Frankfurt am Main, 1997). L. D. Landau and E. M. Lifschitz, Lehrbuch der Theoretischen Physik, vol. 2 Klassische Feldtheorie (Harry Deutsch, Thun & Frankfurt am Main, 1997). H. J. Lee, P. Neumayer, J. Castor, T. Dppner, R. W. Falcone, C. Fortmann, o B. A. Hammel, A. L. Kritcher, O. L. Landen, R. W. Lee, D. D. Meyerhofer, D. H. Munro, R. Redmer, S. P. Regan, S. Weber, and S. H. Glenzer, Xray Thomson scattering measurements of density and temperature in shockcompressed beryllium, Phys. Rev. Lett. (under review). S. Lorentzen, Temperatur- und Dichteabhngigkeit von Spektrallinienprolen a 2+ in Plasmen am Beispiel der Li -Ly- Linie, Diploma thesis, Rostock University, Rostock (2008). L. D. Landau and I. J. Pomeranchuk, The limits of applicability of the theory of bremsstrahlung by electrons and of the creation of pairs at large energies, Dokl. Akad. Nauk SSSR 92, 535 (1953). A. Magnus, On the structure of the two-point pad table, in Analytic Theory e of Continued Fractions (Springer, Berlin, Heidelberg, New York, 1982), doi: 10.1007/BFb0093313. http://dx/doi.org/10.1007/BFb0093313. G. D. Mahan, Many-Particle Physics (Plenum Press, New York and London, 1981), 2nd edn. G. D. Mahan, GW approximations, Comments Condens. Mat. Phys. 16, 333 (1994). M. Mayer, New electron-proton Bremsstrahlung rates for a hot plasma where the electron temperature is much smaller than the proton temperature, Astronomy & Astrophys. 461, 381 (2007), doi:10.1051/0004-6361:20066245. http://dx.doi.org/10.1051/0004-6361:20066245. A. Messiah, Quantum Mechanics, two volumes bound as one (Dover Publications, Mineola, New York, 1999).

[Lin95]

[Lor08]

[LP53]

[Mag82]

[Mes99]

Bibliography [MGW06]

91

J. J. MacFarlane, I. E. Golovkin, and P. R. Woodru, HELIOS-CR A 1-D radiation-magnetohydrodynamics code with inline atomic kinetics modeling, J. Quant. Spectrosc. Radiat. Transfer 99 (2006), doi:10.1016/j.jqsrt.2005.05.031. http://dx/doi.org/10.1016/j.jqsrt.2005.05.031. A. B. Migdal, Bremsstrahlung and Pair Production in Condensed Media at High Energies, Phys. Rev. 103, 1811 (1956), doi:10.1103/PhysRev.103.1811. http://dx.doi.org/10.1103/PhysRev.103.1811. K. Miyamoto, Plasma Physics for Nuclear Fusion (MIT Press, Cambridge, MA, 1980). G. R. Mocken and C. H. Keitel, Quantum dynamics of relativistic electrons, J. Comp. Phys. 199, 558 (2004), doi:10.1016/j.jcp.2004.02.020. http://dx. doi.org/10.1016/j.jcp.2004.02.020. P. Mulser, M. Kanapathipillai, and D. H. H. Homann, Two Very Efcient Nonlinear Laser Absorption Mechanisms in Clusters, Phys. Rev. Lett. 95, 103401 (2005), doi:10.1103/PhysRevLett.95.103401. http://dx. doi.org/10.1103/PhysRevLett.95.103401. K. Morawetz and G. Rpke, Stopping power in nonideal and strongly coupled o plasmas, Phys. Rev. E 54, 4134 (1996), doi:10.1103/PhysRevE.54.4134. http: //dx.doi.org/10.1103/PhysRevE.54.4134. V. G. Morozov and G. Rpke, Photon production by relativistic electrons in o plasmas, J. Phys.: Conf. Ser. 35, 110 (2006), doi:10.1088/1742-6596/35/1/ 009. http://dx.doi.org/10.1088/1742-6596/35/1/009. V. G. Morozov, G. Rpke, and A. Hll, Kinetic Theory of Quantum Electroo o dynamic Plasma in a Strong Electromagnetic Field: I. The Covariant Formalism, Theor. Math. Phys. 131, 812 (2002), doi:10.1023/A:1015979524409. http://dx.doi.org/10.1023/A:1015979524409. V. G. Morozov, G. Rpke, and A. Hll, Kinetic Theory of Quantum Electroo o dynamic Plasma in a Strong Electromagnetic Field: II. The Covariant MeanField Approximation, Theor. Math. Phys. 132, 1021 (2002), doi:10.1023/A: 1015979524409. http://dx.doi.org/10.1023/A:1015979524409. A. B. Murphy, Thomson scattering diagnostics of thermal plasmas: Laser heating of electrons and the existence of local thermodynamic equilibrium, Phys. Rev. E 69, 016408 (2004), doi:10.1103/PhysRevE.69.016408. http: //dx/doi.org/10.1103/PhysRevE.69.016408. W. Nagel and J. Ventura, Coulomb bremsstrahlung and cyclotron emissivity in hot magnetized plasmas, Astron. Astrophys. 118, 66 (1983). http://adsabs. harvard.edu/abs/1983A%26A...118...66N.

[Mig56]

[Miy80] [MK04]

[MKH05]

[MR96]

[MR06]

[MRH02a]

[MRH02b]

[Mur04]

[NV83]

92

Bibliography

[OGRW06] B. Omar, S. Gnther, G. Rpke, and A. Wierling, Hydrogen Balmer Specu o trum from a High-Pressure Arc Discharge: Revisited, Contrib. Plasma Phys. 47 (2006), doi:10.1002/ctpp.200710042. http://dx.doi.org/10.1002/ctpp. 200710042. [OGWR06] B. Omar, S. Gnter, A. Wierling, and G. Rpke, Neutral helium spectral lines u o in dense plasmas, Phys. Rev. E 73, 056405 (2006), doi:10.1103/PhysRevE.73. 056405. http://dx.doi.org/10.1103/PhysRevE.73.056405. [ORR02] G. Onida, L. Reining, and A. Rubio, Electronic excitations: densityfunctional versus many-body Greens-function approaches, Rev. Mod. Phys. 74, 601 (2002), doi:10.1103/RevModPhys.74.601. http://dx.doi.org/10. 1103/RevModPhys.74.601. G. Plunien, B. Mller, and W. Greiner, The Casimir Eect, Phys. Rep. u 134, 87 (1986), doi:10.1016/0370-1573(86)90020-7. http://dx.doi.org/10. 1016/0370-1573(86)90020-7. Peskin and Schrder, An Introduction to Quantum Field Theory (Addison o Wesley, 1995). A. Pukhov, Strong eld interaction of laser radiation, Reports on Progress in Physics 66, 47 (2003), doi:10.1088/0034-4885/66/1/202. http://dx.doi. org/10.1088/0034-4885/66/1/202. T. S. Ramazanov, K. N. Dzhumagulova, and Y. A. Omarbakiyeva, Eective polarization interaction potential charge-atom for partially ionized dense plasma, Phys. Plasmas 12, 092702 (2005), doi:10.1063/1.2008213. http:// dx.doi.org/10.1063/1.2008213. E. Rebhan, Theoretische Physik, vol. 2 (Elsevier-Spektrum, Mnchenu Heidelberg-Berlin, 2005). H. Reinholz, Dielectric and Optical Properties of Dense Plasmas, Ann. Phys. Fr. 30, 1 (2005), doi:10.1051/anphys:2006004. http://dx.doi.org/10.1051/ anphys:2006004. J. Richter, Radiation of hot gases, in Plasma Diagnostics (edited by W. Lochte-Holtgreven) (AIP Press, New York, 1995), chap. 1, 163. G. B. Rybicki and A. P. Lightman, Radiative Processes in Astrophysics (J. Wiley & Sons, New York, 1975). H. Reinholz, T. Raitza, and G. Rpke, Correlations, Collision Frequency o and Optical Properties in Laser Excited Clusters, Int. J. Mod. Phys. B 21, 2460 (2007), doi:10.1142/S0217979207043816. http://dx.doi.org/10. 1142/S0217979207043816.

[PMG86]

[PS95] [Puk03]

[RDO05]

[Reb05] [Rei05]

Bibliography

93

[RRRW00] H. Reinholz, R. Redmer, G. Rpke, and A. Wierling, Long-wavelength limit of o the dynamical local-eld factor and dynamical conductivity of a two-component plasma, Phys. Rev. E 62, 5648 (2000), doi:10.1103/PhysRevE.62.5648. http: //dx.doi.org/10.1103/PhysRevE.62.5648. [RS00] C. D. Roberts and S. M. Schmidt, Dyson-Schwinger equations: Density, temperature and continuum strong QCD, Prog. Part. Nucl. Phys. 45, S1 (2000), doi:10.1016/S0146-6410(00)90011-5. http://dx.doi.org/10.1016/ S0146-6410(00)90011-5. C. D. Roberts and A. G. Williams, Dyson-Schwinger equations and their application to hadronic physics, Prog. Part. Nucl. Phys. 33, 477 (1994), doi:10. 1016/0146-6410(94)90049-3. http://dx.doi.org/10.1016/0146-6410(94) 90049-3. J. Schwinger, On Gauge Invariance and Vacuum Polarization, Phys. Rev. 82, 664 (1951), doi:10.1103/PhysRev.82.664. http://dx.doi.org/10.1103/ PhysRev.82.664. J. Schwinger, On the Greens Functions of Quantized Fields I, Proc. Nat. Acad. Sci. 37, 451 (1951). http://www.pnas.org/content/37/7/452.full. pdf. J. Schwinger, On the Greens Functions of Quantized Fields II, Proc. Nat. Acad. Sci. 37, 455 (1951). http://www.pnas.org/content/37/7/455.full. pdf. T. Schmielau, Optische und Ein-Teilchen-Eigenschaften von Halbleitern, Ph.D. thesis, Rostock University, Rostock (2001). A. Sengebusch, S. H. Glenzer, A. L. Kritcher, H. Reinholz, and G. Rpke, Shift o of Cl K and K Lines in Laser Produced Dense Plasmas, Contrib. Plasma Phys. 47, 309 (2007), doi:10.1002/ctpp.200710041. http://dx.doi.org/10. 1002/ctpp.200710041. J. Sheeld, Plasma Scattering of Electromagnetic Waves (Academic Press, New York, 1975). R. Sigel, Physics of Laser-Plasma Interaction, in Handbook of Plasma Physics, vol. 3 (edited by A. Rubenchik and S. Witkowski) (Elsevier, Amsterdam, 1991). S. Schnez, E. Ltstedt, U. D. Jentschura, and C. H. Keitel, Lasero assisted bremsstrahlung for circular and linear polarization, Phys. Rev. A 75, 053412 (2007), doi:10.1103/PhysRevA.75.053412. http://dx.doi.org/ 10.1103/PhysRevA.75.053412.

[RW94]

[Sch51a]

[Sch51b]

[Sch51c]

[She75] [Sig91]

[SLJK07]

94 [SLR93]

Bibliography S. C. Snyder, G. D. Lassahn, and L. D. Reynolds, Direct evidence of departure from local thermodynamic equilibrium in a free-burning arc-discharge plasma, Phys. Rev. E 48, 4124 (1993), doi:10.1103/PhysRevE.48.4124. http://dx. doi.org/10.1103/PhysRevE.48.4124. A. Sommerfeld, Atombau und Spektrallinien, vol. 1 (Vieweg, Braunschweig, 1949). D. Semkat, R. Redmer, and T. Bornath, Collisional absorption in aluminum, Phys. Rev. E 73, 066406 (2006), doi:10.1103/PhysRevE.73.066406. http:// dx.doi.org/10.1103/PhysRevE.73.066406. H. Schillinger and R. Sauerbrey, Electrical conductivity of long plasma channels in air generated by self-guided femtosecond laser, Appl. Phys. B 68, 753 (1999), doi:10.1007/s003400050699. http:/dx.doi.org/10.1007/ s003400050699. R. Schepe, T. Schmielau, D. Tamme, and K. Henneberger, Damping and T-Matrix in Dense e-h Plasmas, Phys. Status Solidi B 206, 273 (1998), doi:10.1002/(SICI)1521-3951(199803)206:1 273::AID-PSSB273 3.0.CO;2-T. http://dx.doi.org/10.1002/(SICI)1521-3951(199803)206: 1<273::AID-PSSB273>3.0.CO;2-T. I. Sachs and A. Wipf, Finite temperature Schwinger model, ETH Zrich u preprint ETH-TH/91-15 (1992). http://www2.uni-jena.de/~p5anwi/ publications/papers/schwinger.ps.gz. S. Sorge, A. Wierling, G. Rpke, W. Theobald, R. Sauerbrey, and T. Wilhein, o Diagnostics of a laser-induced dense plasma by hydrogen-like carbon spectra, J. Phys. B: At. Mol. Opt. Phys. 33, 2983 (2000), doi:10.1088/0953-4075/33/ 16/304. http://dx.doi.org/10.1088/0953-4075/33/16/304. Y. Takahashi, On the generalised Ward identity, Nuovo Cimento 6, 370 (1957). Y. Takada, Inclusion of Vertex Corrections in the Self-Consistent Calculation of Quasiparticles in Metals, Phys. Rev. Lett. 87, 226402 (2001), doi:10. 1103/PhysRevLett.87.226402. http://dx.doi.org/10.1103/PhysRevLett. 87.226402. R. Thiele, T. Bornath, C. Fortmann, A. Hll, R. Redmer, H. Reinholz, o G. Rpke, A. Wierling, S. H. Glenzer, and G. Gregori, Plasmon resonance in o warm dense matter, Phys. Rev. E 78, 026411 (2008), doi:10.1103/PhysRevE. 78.026411. http://dx.doi.org/10.1103/PhysRevE.78.026411. R. Thiele, Thomsonstreuung in warmer und dichter Materie, Ph.D. thesis, Rostock University (2007).

[Som49] [SRB06]

[SS99]

[SSTH98]

[SW92]

[SWR+ 00]

[Tak57] [Tak01]

[TBF+ 08]

[Thi07]

95

N. A. Tahir, I. V. Lomonosov, A. Shutov, S. Udrea, C. Deutsch, V. E. Fortov, V. Gryaznov, D. H. H. Homann, J. Jacobi, V. Kain, M. Kuster, P. Ni, A. R. Piriz, R. Schmidt, P. Spiller, D. Varentsov, and K. Zioutas, Proposed studies of strongly coupled plasmas at the future FAIR and LHC facilities: the HEDgeHOB collaboration, J. Phys. A: Math. Gen. 39, 4755 (2006), doi:10.1088/0305-4470/39/17/S70. http://dx.doi.org/10. 1088/0305-4470/39/17/S70. M. Ter-Mikaelyan, The interference emission of high-energy electrons, Dokl. Akad. Nauk SSSR 94, 1033 (1953). U. Teubner, G. Pretzler, T. Schlegel, K. Eidmann, E. Frster, and K. Witte, o Anomalies in high-order harmonic generation at relativistic intensities, Phys. Rev. A 67, 013816 (2003), doi:10.1103/PhysRevA.67.013816. http://dx. doi.org/10.1103/PhysRevA.67.013816. R. Thiele, R. Redmer, H. Reinholz, and G. Rpke, Using the Gould-DeWitt o scheme to approximate the dynamic collision frequency in a dense electron gas, J. Phys. A 39, 4365 (2006), doi:10.1088/0305-4470/39/17/S09. http: //dx.doi.org/10.1088/0305-4470/39/17/S09. T. Tschentscher, Investigation of ultrafast processes using X-ray free-electron laser radiation, J. Chem. Phys. 299, 271 (2004), doi:10.1016/j.chemphys.2003. 12.013. http://doi.dx.org/10.1016/j.chemphys.2003.12.013. T. Tschentscher and S. Toleikis, Investigations of ultrafast phenomena in high-energy density physics using X-ray FEL radiation, Eur. Phys. J. D 36, 193 (2005), doi:10.1140/epjd/e2005-00268-4. http://dx.doi.org/10.1140/ epjd/e2005-00268-4. R. G. Ulbrich and C. Weisbuch, Resonant Brillouin Scattering of Excitonic Polaritons in Gallium Arsenide, Phys. Rev. Lett. 38, 865 (1977), doi:10.1103/ PhysRevLett.38.865. http://dx.doi.org/10.1103/PhysRevLett.38.865. A. A. Varfolomeev et al., Eect of the medium on the bremsstrahlung spectrum of 40-GeV electrons, Sov. Phys. JETP 42, 218 (1975). M. Veltman, Diagrammatica (Cambridge University Press, 1994). J. Vorberger, M. Schlanges, and W.-D. Kraeft, Equation of state for weakly coupled quantum plasmas, Phys. Rev. E 69, 046407 (2004), doi:10.1103/ PhysRevE.69.046407. http://dx.doi.org/10.1103/PhysRevE.69.046407. J. C. Ward, An Identity in Quantum Electrodynamics, Phys. Rev. 78, 182 (1950), doi:10.1103/PhysRev.78.182. http://dx.doi.org/10.1103/ PhysRev.78.182.

[TRRR06]

[Tsc04]

[TT05]

[UW77]

[War50]

96 [WCL+ 08]

Bibliography B. Wang, J. Chen, J. Liu, Z.-C. Yan, and P. Fu, Attosecond-pulse-controlled high-order harmonic generation in ultrashort laser elds, Phys. Rev. A 78, 023413 (2008), doi:10.1103/PhysRevA.78.023413. http://dx.doi.org/10. 1103/PhysRevA.78.023413. S. Weinberg, The Quantum Theory of Fields, vol. 1 (Cambridge University Press, 1996), 2nd edn. G. C. Wick, The Evaluation of the Collision Matrix, Phys. Rev. 80, 268 (1950), doi:10.1103/PhysRev.80.268. http://dx.doi.org/10.1103/ PhysRev.80.268.

[Wei96] [Wic50]

[WMR+ 01] A. Wierling, T. Millat, G. Rpke, R. Redmer, and H. Reinholz, Inverse o bremsstrahlung of hot, weakly coupled plasmas, Physics of Plasmas 8, 3810 (2001), doi:10.1063/1.1383025. http://link.aip.org/link/?PHP/8/3810/ 1. [WMR04] A. Wierling, T. Millat, and G. Rpke, Classical bremsstrahlung in o a non-ideal plasma with eective interaction., J. Plasma Phys. 70, 185 (2004), doi:10.1017/S0022377803002538. http://dx.doi.org/10.1017/ S0022377803002538. W. L. Wiese, M. W. Smith, and B. M. Glennon, Atomic Transition Probabilities (H through Ne - A Critical Data Compilation), in Natl. Stand. Ref. Data Ser., NSRDS-NBS, vol. 4 (Natl. Bur. Stand. (U.S.), 1966). http: //physics.nist.gov/PhysRefData/ASD/index.html. U. Zastrau, C. Fortmann, R. R. Fustlin, L. Cao, T. Dppner, S. H. Glenzer, a o G. Gregori, T. Laarmann, H. J. Lee, A. Przystawik, P. Radclie, H. Reinholz, G. Rpke, R. Thiele, J. Tiggesbumker, X. N. Truong, S. Toleikis, o a I. Uschmann, A. Wierling, T. Tschentscher, E. Frster, and R. Redmer, o Bremsstrahlung and Line Spectroscopy of Warm Dense Aluminum Plasma Generated by XUV Free Electron Lasers, Phys. Rev. E 78, 066406 (2008), doi:10.1103/PhysRevE.78.066406. http://dx.doi.org/10.1103/PhysRevE. 78.066406. D. Zubarev, V. Morozov, and G. Rpke, Statistical Mechanics of Nonequilibo rium Processes, vol. 2 (Akademie Verlag, Berlin, 1996). D. Zubarev, V. Morozov, and G. Rpke, Statistical Mechanics of Nonequilibo rium Processes, vol. 1 (Akademie Verlag, Berlin, 1996).

[WSG66]

[ZFF+ 08]

[ZMR96a] [ZMR96b]

Authors: Carsten Fortmann, Heidi Reinholz, August Wierling, and Gerd Rpke o Appeared in volume 20 of the series Condensed Matter Theories, Nova Science, New York, 2006, pages 317-332. Listing of contributions by authors: C.F.: Preparation of the manuscript (sections 1, 3A, 3B, 3C, 4), derivation of the algebraic equation for the self-energy, numerical calculations. H.R.: Preparation of manuscript (section 2) G.R.: Preparation of manuscript (section 1) A.W.: Preparation of manuscript (section 3D), numerical calculation of vertex contributions G.R.: Preparation of manuscript (section 1)

100

101

Bremsstrahlung from dense plasmas and the Landau-Pomeranchuk-Migdal eect C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling o Institute of Physics, Rostock University D-18051 Rostock, Germany

1. INTRODUCTION In a fully ionized plasma, bremsstrahlung and inverse bremsstrahlung are the only emission and absorption processes, respectively. For partially ionized plasmas these processes contribute to some extent to the continuous emission/absorption spectrum. We introduce the emission coecient j() as the rate of radiated energy per unit volume, frequency and solid angle, and the absorption coecient () as the relative attenuation of the intensity of electromagnetic waves propagating in the medium per unit length. For a thermally equilibrated plasma, these quantities are linked by Kirchhos law [1] j() = L()() , (1) with the Planck distribution L() = h 3 / 4 3 c2 (exp( /kB T ) 1) . Thus, it is h sucient to study one or the other, i.e. j() or () . Here, we choose the absorption coecient (). The absorption spectrum can be determined according to quantum electrodynamics (QED) [2] from the interaction part of the QED Lagrangian Lint (x) =

c Zc ej (x)A (x) ,

(2)

x being a four-dimensional space-time variable. This Lagrangian describes a minimal c coupling between the particle current j and the vector potential A . The index c denotes the species of the particles involved, carrying the charge Zc e. Introducing a Hamiltonian H, the transition rate between asymptotically free states of electrons e e | pin , | pout with the energies Ein , Eout respectively, follows from Fermis Golden rule as 2 2 e e (3) pout |Hint |pin (Ein + h Eout ) , win,out = h

102

with Hint = d3 x Lint (x). Assuming single, uncorrelated scattering from dierent ions only, the absorption coecient is given by () = ni 0 0 d3 pin 0 d3 pout e 1 f (Ein ) win,out . 3 3 (2) (2) c (4)

Here, ni 0 is the total number of ions in the volume 0 and win,out is the transition probability that a photon of momentum k = pout pin and polarization is absorbed by a single electron of momentum pin in the Coulomb potential of an ion, leaving the process with momentum pout . The momentum distribution function of the incoming 1 e electrons is denoted as f (Ein ) = exp ( 2 p2 /2me e )/kB T + 1) . The factor h in 1/c arises due to the current of incoming photons. Evaluation of the transition rate in Born approximation with pout |Hint |pin Zi e2 2 h = 0 me 0 (pout pin )z e2 2 0 0 |pout pin |2 h h (5)

gives the well known Bethe-Heitler formula [3] for the emission coecient Eq. (1). In the low frequency limit, the Bethe-Heitler cross section behaves roughly like 1/. Within QED, this infrared-divergence is discussed as a consequence of the neglect of vertex corrections [4]. The infrared-divergent terms can be shown to cancel with corresponding contributions to the form factor of the source particle which is a manifestation of the Ward-Takahashi identities [5]. In the non-relativistic limit and for soft photons, the absorption coecient for a hydrogen plasma (Zi = 1) is given by B () = C ne ni sinh h 3 h 2kB T K0 h 2kB T , (6)

3/2 1/2 where C 1 = 3 23 3/2 me c (kB T ) /e6 and K0 (x) = 0 cos(x sinh t)dt is the 0 modied Bessel function of zeroth order. Here, the electron density ne has been introduced, which is equal to the ion density ne = ni in charge neutral systems. This Born approximation can be improved by taking Coulomb wave functions for the initial and nal states. In this case, an analytic result for the absorption coecient was given by Sommerfeld [6]. Due to the occurrence of hyper-geometric functions, simpler approximations such as the Born-Elwert approximation [7] have been developed. In the classical limit, the Sommerfeld expression reduces to a result obtained earlier by Kramers [8]. In a dense plasma, the inuence of the collective behavior of the system and the modication of single-particle properties of the emitting and absorbing particles as well as the bremsstrahlung photons are important. Retaining the single-scattering picture of Eq. (4), medium eects can be taken into account e.g. by a modication of the potential. Instead of a Coulomb potential, a static or dynamically screened potential should be used [9]. A quantum-statistical approach based on a systematic perturbative treatment of the force-force correlation function has been developed and applied to inverse bremsstrahlung in Ref. [10].

103

How does multiple scattering of the emitting electron change the cross section? This question was rst treated by Landau and Pomeranchuk in a semi-classical way and soon afterwards by Migdal [11] using kinematic considerations. They showed that the account of successive collisions leads to a suppression of the bremsstrahlung cross section compared to the Bethe-Heitler result at photon energies low against the energy of the scattering electron. Both Bethe-Heitler and Landau-PomeranchukMigdal (LPM) theory give the same result in the limiting cases of high photon energies and/or low densities, i.e., in those cases where the Born approximation is applicable. Migdals result has been rederived more recently using dierent methods, e.g., path-integral calculations [12] and quantum kinetic equations [13]. A comprehensive overview of theoretical approaches is given in [14]. There, it is also pointed out that the LPM eect might play an important rle for the emission/absorption spectrum o of a plasma even in the non-relativistic regime due to the large number of free charge carriers. Due to the large energies required to observe notable eects, it took quite a long time until experiments could unambiguously approve LPM theory. Experimental investigations have been performed since the late 1950s using high energy electrons (some MeV to GeV) from cosmic rays and accelerators [14]. These early experiments suered from poor statistics and were unable to conrm LPM theory. More recent experiments at SLAC and CERN [15,16] have indeed shown the LPM eect. Migdals theory is not completely microscopic but relies on the denition of a macroscopic parameter, namely the coherence length l, initially introduced by TerMikaelyan [14] and rst applied to the theory of bremsstrahlung by Landau and Pomeranchuk [11]. The coherence length gives the scale of photon energy, below which the suppression of bremsstrahlung becomes important. It is basically determined by the density of the medium. Knoll and Voskresensky [17] were the rst to treat the question of the LPM eect in the context of a many particle system, where every constituent has to be regarded as an emitter of bremsstrahlung. They were able to show a suppression of the emission/absorption spectrum at low frequencies by using medium modied single particle propagators. The particles are assigned a nite lifetime , given by the width of their spectral function = h/ . However, this quantity is not calculated from a microscopic approach, but simply set as a parameter. It is related to the aforementioned coherence length by the simple relation l/c [14]. In this work, we present a completely microscopic calculation of the absorption spectrum, where the spectral function of the electron is obtained in a self consistent manner. Thus, we present for the rst time a fully microscopic treatment of the LPM eect for a many-particle system. The absorption coecient is dened in linear response theory and is expressed through thermodynamical correlation functions using a diagram technique equivalent to the well known Feynman diagrams [18]. Medium eects are accounted for systematically in terms of self-energy and vertex corrections, which are evaluated in certain approximations. We show that the Bethe-Heitler bremsstrahlung spectrum follows from this approach in lowest order perturbation theory using free particle propagators. The account of coherent successive scattering, as in LPM theory, can be achieved by a partial summation of self-energy diagrams. This procedure leads to a nite width of the single-particle spectral function which reects the modication of the energy-momentum dispersion

104

relation due to the medium. It is found that the absorption coecient calculated on the basis of these medium modied propagators is signicantly altered in the low frequency range in comparison to the Born approximation. In the high frequency limit as well as in the low density limit, the Born approximation is reproduced. Thus, the main features of LPM theory can be rederived within our microscopic approach. 2. LINEAR RESPONSE THEORY We consider the interaction of soft photons with a non-relativistic, homogeneous plasma. A key quantity to describe the propagation of electro-magnetic waves in a medium is the dielectric tensor ij (k, ) [19]. In the isotropic case, the tensor can be decomposed into a transverse t (k, ) and a longitudinal l (k, ) part with respect to the wave vector k. Here and in the following, we take k to point along the z-axis, k = kez . In the long-wavelength limit k 0, the longitudinal and the transverse part coincide () = limk0 l (k, ) = limk0 t (k, ). The absorption coecient can be obtained from the dielectric function according to () = Im () , c n() (7)

where the index of refraction n() is also linked to the dielectric function by 1 1/2 n() = (Re () + |()|) . 2 The relation between l (k, ) and the longitudinal response function l (k, ) e2 (k, ) l (k, ) = 1 0 k 2 l

1

(8)

(9)

allows for a microscopic approach to the dielectric function. Within linear response theory, the Kubo formula relates the response function to the current-current correlation function [20] k2 z z l (k, ) = i0 J ; J +i , (10) k k where the correlation functions for two observables A, B are dened according to 1 (A; B) =

d Tr[A(i )B 0 ] , h

0

A; B

+i

=

0

The time dependency of the operators is taken in the Heisenberg picture. The current density operator is given as Jk = 1 0 ec h c,pk/2 ac,p+k/2 . p a mc (12)

c,p

105

a and ac,p are creation and annihilation operators for momentum states, respecc,p tively. The species and further quantum numbers such as spin are labeled by c , 0 is a normalization volume, 0 is the equilibrium statistical operator, = 1/(kB T ) is the inverse temperature. Note, that in Eq.(10), a small but nite imaginary part has been added. In the nal results, the limit 0 is taken. The inverse response function can also be expressed as [10] 1 (k, ) = l i z z z z i(Jk ; Jk ) Jk ; Jk 2 (J z ; J z )2 0 k k k

+i

z z Jk ; Jk

z z +i Jk ; Jk +i z z Jk ; Jk +i

(13) This transformation of the current-current correlation function into a force-force cori z z relation function Jk ; Jk +i with J = h [H, J] , which has the meaning of a force as the time derivative of momentum, is more suited for a perturbative treatment [10]. Also, it is convenient to introduce a generalized collision frequency () in analogy to 2 2 the Drude relation [10] () = 1 pl / [ ( + i ())] where pl = c nc e2 /(0 mc ) c is the squared plasma frequency. By comparison with Eq. (9), we establish an expression for the collision frequency in terms of correlation functions () = 0 2 lim 0 pl k0 z z J k, J k

+i

z z Jk ; Jk

z z +i Jk ; Jk +i z z Jk ; Jk +i

(14)

z z 2 taking into account that (Jk ; Jk ) = 0 pl /0 is an exactly known quantity. Further details can be found in Ref. [10]. Making use of Eq. (7), the absorption coecient can be expressed as 2 pl Re () . () = 2 2 Im () + |()|2 )n() c (

(15)

In the high frequency limit pl , the index of refraction is unity and the collision frequency is small compared to the frequency . Then, we can consider the approximation () =

2 pl 0 z z Re () = Re J 0 , J 0 2 2 c c0 +i

(16)

where the collision frequency is given in the form of a force-force correlation function, cf. Ref. [10]. Thus, the absorption coecient is directly proportional to the real part of the force-force correlation function, which itself can be determined using perturbation theory. As well known, the deviation of the diraction index n() from unity at frequencies near the plasma frequency pl is responsible for the so-called dielectric suppression of the bremsstrahlung spectrum. We refer to the pioneering work of Ter-Mikaelyan [21]. In our approach, making use of Eqs. (8) and (14), the index of refraction can be determined from the force-force correlation function. However, due

106

Figure 1. Diagrammatic representation of GJJ ( ). (a) full account of all medium eects by a four particle Green function, (b) factorization into two polarization bubbles, (c) Born approximation, (d) account for successive scattering of electrons on ions via electronic spectral function (full electron propagator), (e) vertex correction. to the choice of the frequency range in consideration ( pl ), this eect will not be considered in the present work. We will focus only on the medium eects obtained directly from the evaluation of the force-force correlation function. 3. GREEN FUNCTION APPROACH AND DENSITY EFFECTS A convenient starting point for a perturbative treatment of the force-force correlation function is the representation in terms of a Green function GJJ in the limit k 0, see Ref. [22] z z J0 , J0

+i

1 1 d Im GJJ ( + i) . + i

1 xi

(17)

0 Im GJJ ( + i) . c0 3

(18)

The time derivative of the electron current density operator is calculated as i ie z z J0,e = H, J0,e = h me 0 with the Hamiltonian H=

c,k c Ek a ac,k + c,k ei vq qz a a ai,kq ae,p+q e,p i,k

(19)

pkq

1 2

c,d kpq

(20)

c and Ek = h2 k 2 /2mc . The spin is not given explicitly but is included into the single particle quantum number c. Due to conservation of total momentum of electrons, only electron-ion collisions contribute to Eq.(19). With Eq.(19), we identify the Green function as a four-particle Green function.

107

Its diagrammatic representation is shown in Fig. 1 (a). Here, G4 denotes a four-particle Green function that contains all interactions between the considered particles [10]. We perform a sequence of approximations by selecting certain diagrams contributing to G4 . Considering the electron-ion interaction determining the force z J0,e only in lowest order, i.e. in Born approximation, but the full correlations within the electron and ion subsystem, respectively, we are led to Fig. 1 (b) showing only those diagrams which can be factorized into two polarization bubbles. e denotes the electronic polarization function, i is the corresponding ionic quantity. In this way, we keep the description of the single scattering event on the level of the Born approximation. However, higher order interactions between the full electron and ion subsystem such as a ladder approximation (t-matrix) are ignored, cf. also Ref. [10]. The t-matrix corrections and in particular the reproduction of the Sommerfeld result [6] have been studied in Ref. [10]. Diagrams (d) and (e) account for successive scattering of electrons on ions via the electronic spectral function (d) and the vertex correction (e). They will be discussed in subsections 3.B, 3.C and 3.D, while in the following subsection 3.A we discuss the Born approximation. A. Born approximation As a simple example and prerequisite for further improvements, we consider the Born approximation. In this case, G4 is a product of four single particle propagators and all single particle propagators are replaced by free propagators. We obtain for the e Green function the diagram given in Fig. 1 (c). For a Maxwellian plasma (f (Ep ) = e ne 3 /2 exp(Ep )), we obtain e Im GBorn () = JJ

h ni ne 3 e6 (1 e ) e 24 3 2 4 2 0h e dEp eEp

e

e Ep

e h + Ep

( + h

h2 2 2 2me )

h2 2 me e h2 2 2Ep me

1 + ln 2

e Ep + h + e Ep + h

e Ep e Ep

2 2

+ +

h2 2 2me h2 2 2me

. (21)

Note, that the spin-degeneracy factor 1/2 for fermions is compensated by a factor 2 from the summation over spin variables in the calculation of the correlation functions. e = (2 2 /me kB T )1/2 is the thermal de-Broglie wavelength. The inverse screening h length occurs due to the use of a statically screened potential of Debye-Hckel type u Z i e2 ei 2 2 2 vq = 0 0 (q2 +2 ) , with = c nc Zc e /(0 kB T ), to ensure convergence at = 0. For = 0, we can consider the Coulomb limit = 0. Performing the integration we arrive at Eq. (6). Note, that this result is sometimes also written as

Born

ni ne 3 e6 e h (1 e ) () = 4 3 3 3 24 0 c h

e e dEp eEp

ln

e Ep + h + e Ep + h

e Ep e Ep

, (22)

see Ref. [23]. It is instructive to study the dierent terms in Eq.(22) in more detail: The integrand contains the distribution function (Maxwell distribution). Furthermore, a logarithm that depends on both electron and photon energy appears.

108

Figure 2. Lowest order correction to Born approximation. Self-energy (second and third diagram) and vertex correction from scattering of electrons on ions.

Taken with appropriate prefactors, this logarithm is equal to the dierential cross section for inverse bremsstrahlung in the non-relativistic limit (Bethe-Heitler formula) [19]. Thus, we have a reasonable relation between the absorption coecient as a macroscopic property and the underlying microscopic process, namely inverse bremsstrahlung. The absorption spectrum is obtained through integration of the cross section of the microscopic process weighted with the distribution function of the absorbing particles. We will now show how the Born approximation can be improved. As already mentioned above, improvements based on a more sophisticated description of the single scattering process via a t-matrix approach have been obtained recently [10]. Eects due to dynamical screening have also been considered. Both eects remove the infrared divergence mentioned before. In this work, we want to include medium eects such as the successive scattering of the absorbing particles on ions during the absorption of the photon, a process becoming more important with increasing density. This is also the basic idea of the LPM eect. The correction in lowest order to Born which takes account of medium eects in the propagator can be achieved by performing either one self-energy insertion or one vertex correction in the sense of Ward-Takahashi identities. The corresponding Feynman diagrams are given in Fig. 2. The rst loop on the r.h.s. of Fig. 2 yields the Born result as shown previously. In the following contributions we shown the self-energy and the vertex correction, respectively. We use i+ to indicate loops with ionic propagators. However, one does not obtain a nite result from this ansatz in the case of the self-energy correction. Instead, a partial summation of all self-energy terms leading to a spectral function is necessary. The results are presented in subsection 3.C. In constrast, the last diagram of Fig. 2 describing the vertex correction gives a nite contribution as shown in subsection 3.D. A full self-consistent treatment of the vertex, i.e. solving the corresponding Bethe-Salpeter equation [18], has not yet been performed. B. Spectral function We now discuss the inuence of multiple scattering of the source particles (electrons) on ions. This can be accounted for by using dressed propagators in the calculation of the force-force correlation function, i.e. by replacing the free electron Green

109

Ge (p, z ) =

h d Ae (p, ) h , 2 z h

(23)

where Ae (p, ) is the electronic spectral function. According to Dysons equah tion [18], the Green function can be represented by a complex electron self-energy e (p, z ). We perform the analytic continuation of the discrete Matsubara frequencies into the upper half of the complex energy plane via z = h + i and decom pose the self-energy into the real and imaginary part, e (p, + i) = e (p, ) h h ie (p, )/2. Then, the spectral function is related to the self-energy: h Ae (p, ) = h ( h

e Ep

(24)

For the self-energy, we describe the scattering of the electron on an ion by a statically screened ion potential of the Debye type as discussed before. The Hartree term vanishes due to charge neutrality. The Fock term of the electron self-energy is not relevant, since we assume a non-degenerate system. We consider here the selfconsistent rst loop correction to the Hartree-Fock self-energy due to the electron-ion interaction. This approximation can be improved by taking into account a partial summation of further loops, leading to the so-called GW approximation for the selfenergy [20]. After analytic continuation of the Matsubara Green function we have e (p, + i) = ni h d3 q v ei (2)3 q

2

h + i

e Ep+q

1 . e (p + q, + i) h

(25)

Eq. (25) can be solved numerically by iteration starting from a suitable initialization. From the form of the Debye potential, we note that the main contribution to the integral in Eq. (25) arises from terms with small momentum q. Therefore, we will discuss an approximation where the argument p + q in the self-energy is replaced by p, i.e. e (p + q, + i) e (p, + i) on the r.h.s. of Eq. (25). After dropping h h the shift of the momentum variable in the self-energy, the q integral can be performed analytically. The result e (p, + i) = h ni me e4 1 42 2 0h

(26) is solved numerically for the real and imaginary part. In Fig. 3 (a), we show the self-energy, the dispersion relation, and the resulting spectral function obtained from a self-consistent solution of Eq. (26). The spectral function in Fig. 3 (a) shows a broadened quasi-particle resonance at the energy h =

110

(a) 0.2

p = 1 aB 2 1 0 0.5 0 -0.5 0.5 0 -0.5 1000 500 0.8 0.9 1 1.1 _ h [units of Ry] 1.2

-1

(b)

0 0.7

0.8

1.2

Figure 3. Imaginary and real part of the electrons self-energy e (p, + i), disperh 1 sion relation, and spectral function Ae (p, ) for p = 1 aB (a) from self-consistent h calculation cf. Eq. (26) and (b) using free propagators, cf. Eq. (27). The parameters are: Electron density ne = 106 a3 , temperature kB T = 27.2 eV. B h 2 (p2 + 2 )/2me . As expected, its shape is primarily determined by the imaginary part of the self-energy. This can be seen by a comparison of Ae (p, ) with e (p, ). We mention, that analytic constraints on the self-energy function such as Kramers-Kronig h h relations as well as the rst sum-rule for the spectral function d Ae (p, ) = 1 , 2 are fullled within the numerically achievable precision. For the sake of comparison, we discuss a simplied calculation in which we neglect the self-consistent propagator and replace it by a free propagator. Thus, the self-energy on the r.h.s. of Eq. (25) disappears. Then we nd 0 (p, +i) h e me ni me4 1 2 2 me +p 2 2 ( +i)2i 2 2 ( + i) = h h 2 2 40 h h h

1

, (27)

which can be separated into real and imaginary part in the limit 0: 0 (p, ) = h e h 0 (p, ) = e ni me4 p2 /2 + 2 /2 me / h , 2 2 2 (p2 /2 + 2 /2 me / )2 + 22 me / h h 2(2) 0 h (28)

From the imaginary part 0 (p, ) we see that the contribution to the spectral h e e function near the free-particle energy h = Ep is damped out to a large extent. e These functions as well as the corresponding dispersion relation h Ep 0 (p, ) h e and spectral function are plotted on the r.h.s. of Fig. 4. The spectral function exhibits two separate peaks corresponding to the roots of the dispersion relation and

111

e no peak at the quasi-particle energy Ep . This contribution from the central root at e the free-particle energy is damped out due to the large value of 0 (p, = Ep ) at h e the free particle dispersion. Note also, the order of magnitude change in the damping e between the rst iteration shown on the r.h.s. of Fig. 4 and the self-consistent result shown on the l.h.s. Thus, these structures can clearly be identied as artifacts since they disappear in the self-consistent calculation, which is necessary to obtain relevant results.

C. Eects of the electron self-energy The Born approximation can be improved accounting for self-energy and vertex corrections, see Fig. 2. However, the contribution of the two diagrams containing the self-energy is diverging so that we performed partial summations of higher orders, leading to the Dyson equation and the spectral function discussed above. In this way, the free electron propagator is replaced by the full electron propagator when calculating the force-force correlation function. This approximation exactly reects the point made by Migdal and Landau/Pomeranchuk to account for a nite life-time (damping rate) of electron states propagating in a dense medium. The force-force Green function is obtained from the evaluation of the diagram shown in Fig. 1 (d): GJ ( ) J h 2 ni e6 = m2 2 e 0

d3 p (2)3

2 qz d3 q (2)3 (q 2 + 2 )2

d Ae (p q, ) 2 z + h

d Ae (p, ) h h . 2 z h

(29)

After summation over the fermionic Matsubara frequencies z and shifting variables, we obtain the imaginary part of GJ (). J Im GJ () = J =

ni e6 6m2 2 e 0

d3 p (2)3

ni e6 h = 2 2 (2)5 3me 0

dp dq

0 0

where the integrals over the angular parts have been performed.

112

Free propagator Self-consistent self-energy Lorentzian

() / ()

Figure 4. Correction factor ()/B () as a function of the photon energy h with free propagators in self-energy diagram, the self-consistent spectral function, and a Lorentzian ansatz. Parameter values: Electron density ne = 106 a3 , temperature B kB T = 27.2 eV . The further evaluation requires an expression for the spectral function. Note, that in the limit of free particles, where the spectral function is given by a -function, the Born approximation, Eq. (21), is recovered. We will use the result obtained above within our approximation for the self-energy. The result is shown in Fig. 4. The correction factor ()/B () (cf. Eq.(18)) is plotted as a function of the frequency for three dierent approximations with the parameters ne = 106 a3 and kB T = 27.2 eV. The full line presents the result B of the self-consistent treatment, Eq. (26), and is compared to a calculation with free propagators in the self-energy diagrams, see Eq. (27), and a calculation using a Lorentzian ansatz of the spectral function with a width taken at the on-shell energy e (p, = Ep ). This corresponds to the introduction of a nite life-time in the h approach of Knoll and Voskresensky [17]. For the lowest frequencies considered here, all approximations show a suppression of the absorption coecient in comparison to the Born result. At high frequencies, all curves tend towards unity, i.e., the Born result is recovered. For intermediate frequencies, an enhancement of up to 35 % is found for the calculation using free propagators. Making use of the self consistency, the enhancement is reduced to 6 % at most. For the Lorentzian ansatz, no enhancement at all appears. Thus, the width of the imaginary part of the self-energy as a function of frequency (Fig. 3) plays a crucial rle for the size of the enhancement. We expect that any increase in the width o of the self-energy will further decrease the enhancement or even lead to a suppression for all frequencies as in the case of the Lorentzian ansatz, which corresponds to an innite width of the imaginary part of the self-energy. A further broadening of the imaginary part of the self-energy could result from an extension to higher orders in the set of diagrams for the self-consistent calculation of the spectral function, e.g. by inclusion of vertex terms.

113

D. Vertex corrections As known from the Ward-Takahashi identities, self-energy and vertex corrections are intrinsically related. In particular, if medium corrections arise in a certain order of a small parameter like the density from the self-energy, also contributions from the vertex corrections are expected in the same order. This well-known fact is used to construct so-called conserved approximations [24]. Thus, it is necessary to study the vertex corrections corresponding to the self-energy considered before. However, the solution of the vertex-equation is a technically very challenging task and has been solved so far only in certain approximations [25]. In lowest order, the vertex correction is obtained through insertion of one ion-loop inside the electron-loop: We nd for the imaginary part of the force-force Green function Im GVJ ( J ni ne 3 e6 e + i) = 3 2 3 m 24 0 h e +

dp p e

0

2 p2 /2me h

(p) ln

e Ep + e Ep e Ep + e Ep

e Ep + h e Ep + h e Ep h e Ep h

h dp p e 2me / h

2 2

p /2me

V (p) ln

(32)

taken in lowest order in . This expression can be evaluated numerically. In the limit of high frequencies, where the second integral becomes negligible compared to the rst on, the absorption coecient is proportional to K1 ()/ 4 . Since the K1 function has the same asymptotic behavior as the K0 function, which appeares in the Born result for the absorption coecient Eq. (6), the ratio V /B (cf. Eq. (18)) behaves like 1/ in the high frequency limit. For all frequencies considered, a suppression with respect to the Born approximation is found. For the considered energy range, the corrections are small and decrease with increasing energy. Using the set of parameters ne = 106 cm3 and kB T = 27.2 eV, the expected high frequency behavior 1 arises for energies larger than 1 Ry. The corrections are small for low densities compared to higher densities. We consider the absorption coecient () including all of the improvements. Since the Born approximation is already included in the self-energy contribution () we have () = V () + (). (33) The relative change ()/B () is presented in Fig. 5. For the sake of comparison, the self-energy correction is shown as well. For small frequencies, the self-energy contribution and the vertex contribution add to a net suppression. For higher energies, the self-energy term shows an enhancement, which is partially compensated by the vertex. However, the net result is still an enhancement. In the high frequency limit, the Born result is reproduced.

114

() / ()

0.9 0.8 0.01

Figure 5. Total absorption coecient taken relative to the Born result V ()/B () as a function of the photon energy h. Parameter values: Electron density ne = 106 a3 , temperature kB T = 27.2 eV. B 4. CONCLUSIONS In this paper, we have studied the inuence of the surrounding medium on the bremsstrahlung spectrum in non-ideal plasmas. The interaction with the medium leads to a nite life-time of the electron states. Instead of free quasi-particles, the spectral function has to be used to describe the electron properties in the medium. Thus, the use of the single-particle spectral function is a quantum-statistically sound implementation of the original idea of successive scatterings by Landau/Pomeranchuk and Migdal. Our approach, namely the microscopic treatment of the dynamical self-energy, extends a recent work of Knoll and Voskresensky [17], where a Lorentzian ansatz for the spectral function with a frequency-independent quasi-particle lifetime was considered. The Lorentzian ansatz for the spectral function was discussed above in subsection 3.C., taking the imaginary part of the self-energy at the quasi-particle energy. Then, a suppression of the bremsstrahlung spectrum was observed. In general, the microscopical treatment leads a frequency-dependent imaginary part of the self-energy, and, according to the Kramers-Kronig relation, to a non-vanishing real part. In particular, the inclusion of the real part of the self-energy in the spectral function inuences the medium modication of the bremsstrahlung spectrum. In the present paper, the one-loop approximation was taken for the self-energy. It has been shown that a self-consistent treatment has to be used in order to avoid unphysical artifacts which arise, if instead of the full propagator the free propagator is taken to evaluate the self-energy. This is already known from the treatment of the spectral functions in plasma physics in the so-called GW approximation [20] where the interaction with the medium is implemented by a screened potential. It should

115

be mentioned that in this case a self-consistent treatment of the spectral function on the level of the GW approximation has been performed [26]. Any iterative solution starting from the free propagator leads to non-physical structures in the spectral function. Within our approach, we found a switch from a suppression at low frequencies to an enhancement at high frequencies for the bremsstrahlung spectrum. The approximation for the self-energy can be improved by considering further diagrams. In particular, the vertex correction would be of interest which modies the coupling to the interaction potential. The inclusion of vertex corrections is also necessary to obtain conserved approximations and has been shown in Subsection 3.D., where a further suppression of the bremsstrahlung spectrum was observed. The modication of the bremsstrahlung spectrum by the surrounding medium is sensitively dependent on the approximation used. We cannot elaborate further on the switch from suppression to enhancement seen in the self-energy correction. In order to verify the existence of such a switch, higher order calculations are necessary. A consistent procedure would consist in a) using the full propagator in the calculation of the vertex correction, b) solving the full vertex equation with full propagators and nally c) solve the Dyson equation for the single particle propagator with the solution of the vertex equation. The importance of vertex corrections in the self consistency relations has also been shown in the description of the spectral function of the homogeneous electron gas. There, notable dierences between a so-called GW approximation [25] including vertex terms and a GW approximation [27] arise. It should be mentioned that self-consistent Schwinger-Dyson equations for the self-energy have been considered in eld theory [28] to nd solutions for the QCD running coupling problem. A corresponding treatment would lead to a better description of the modication of the bremsstrahlung spectrum in a dense medium, but would exceed the frame of the present work. Also, at low frequencies, further eects such as the dielectric suppression are of importance. It has not been considered in this approach, but can easily be obtained from the force-force correlation function as well. ACKNOWLEDGEMENTS We would like to thank J. Knoll, D. Voskresensky and V. Morozov for stimulating discussions. C.F. would also like to thank the Gesellschaft fr Schwerionenforschung u (GSI) for its hospitality and the Studienstiftung des Deutschen Volkes for a scholarship. REFERENCES [1] H. R. Griem, Principles of Plasma Spectroscopy (Cambridge University Press, Cambridge, 1997). [2] C. Itzykson and J.-B. Zuber, Quantum eld theory (McGraw-Hill, New York, 1980).

116

[3] H. Bethe and W. Heitler, Proc. Roy. Soc. A 146, 83 (1934). [4] W. Heitler, The Quantum Theory of Radiation (Oxford University Press, Oxford, 1957). [5] B. R. Holstein, Topics in Advanced Quantum Mechanics (Addison-Wesley, Redwood, 1992). [6] A. Sommerfeld, Atombau und Spektrallinien (Vieweg-Verlag, Braunschweig, 1949). [7] G. Elwert, Ann. Phys. 34, 178 (1939). [8] H. A. Kramers, Phil. Mag. 46, 836 (1923). [9] G. Beke, Radiation Processes in Plasmas (Wiley, New York, 1966). [10] H. Reinholz, R. Redmer, G. Rpke, and A. Wierling, Phys. Rev. E 62, 5648 o (2000), A. Wierling, Th. Millat, G. Rpke, and R. Redmer, Phys. Plasma 8, o 3810 (2001). [11] L. D. Landau and I. J. Pomeranchuk, Dokl. Akad. Nauk SSSR 92, 535 (1953), A. B. Migdal, Phys. Rev. 103, 1811 (1956). [12] B. G. Zakharov, JETP Lett. 64, 781 (1996). [13] A. V. Koshelkin, J. Phys. A 35, 8763 (2002). [14] S. Klein, Rev. Mod. Phys.f 71, 1501 (1999). [15] J. F. Bak, Nucl. Phys. B 302, 525 (1988). [16] H. D. Hansen et al., Phys. Rev. D 69, 032001 (2004). [17] J. Knoll and D. Voskresensky, Ann. Phys. 249, 532 (1996). [18] W. D. Kraeft, D. Kremp, W. Ebeling, and G. Rpke, Quantum Statistics of o Charged Particle Systems (Akademie-Verlag, Berlin, 1986). [19] J. D. Jackson, Classical Electrodynamics (Wiley & Sons, New York, 1975). [20] G. D. Mahan, Many-Particle Physics (Plenum Press, New York and London, 1981). [21] M. L. Ter-Mikaelyan, Dokl. Akad. Nauk SSSR 94, 1033 (1953). [22] D. Zubarev, V. Morozov, and G. Rpke, Statistical Mechanics of Nonequilibrium o Processes (Akademie-Verlag, Berlin, 1996), Vol. 2. [23] I. H. Hutchinson, Principles of Plasma Diagnostics (Cambridge University Press, Cambridge, 1987). [24] G. Baym and L. Kadano, Phys. Rev. 124, 287 (1961). [25] Y. Takada, Phys. Rev. Lett. 87, 226402 (2001). [26] A. Wierling and G. Rpke, Contrib. Plasma Phys. 38, 513 (1998). o [27] B. Holm, Phys. Rev. Lett. 83, 788 (1999). [28] C. D. Roberts and S. M. Schmidt, Prog. Part. Nucl. Phys. 45, S1 (2000).

Authors:Carsten Fortmann, Ronald Redmer, Heidi Reinholz, Gerd Rpke, August Wiero ling, and Wojciech Rozmus Appeared as regular article in Journal of High Energy Density Physics, Volume 2, pages 57-69, May 2006. Listing of contributions by authors: C.F.: Preparation of manuscript, all numerical calculations R.R.: Preparation of manuscript H.R.: Preparation of manuscript G.R.: Preparation of manuscript A.W.: Preparation of manuscript W.R.: Code for Thomson scattering calculations

118

119

b

Received 21 February 2006; received in revised form 13 April 2006; accepted 14 April 2006 Available online 15 May 2006

Abstract We determine the spectral photon yield from a hot dense plasma irradiated by VUV-FEL light in a Thomson scattering experiment. The Thomson signal is compared to the emission background mainly caused by bremsstrahlung photons. We determine experimental conditions that allow for a signal-to-background ratio larger than unity. By derivation of the Thomson and the bremsstrahlung spectrum from linear response theory we present a consistent quantum statistical approach to both processes. This allows for a systematic treatment of medium and quantum effects such as dynamical screening and strong collisions. Results are presented for the threshold FEL-intensity as a function of density and temperature. We show that the account for quantum effects leads to larger thresholds as compared to previous work. 2006 Elsevier B.V. All rights reserved.

Keywords: Thomson scattering; Bremsstrahlung; Free electron laser; Plasma diagnostics; Threshold intensity; Dielectric function; Dynamic structure factor

pe th or 's

1. Introduction

Thomson scattering is a well-established technique for experimental investigation of plasma parameters. Examples can be found in Refs. [1e6]. Observables like particle density, temperature, composition, and degree of ionization can be spatially and temporally resolved by analysis of the scattering spectrum [7]. Until recently, coherent light sources have been available only for the visible and near UV part of the electromagnetic spectrum. Due to small critical density ncrit u2e0me/ e2 z 1020 cm3 of free charge carriers for optical probes, the applicability of Thomson scattering using coherent sources has been limited to targets of relatively low density. Glenzer et al. [8,9] have shown and explored the possibility of X-ray Thomson scattering in solid density targets using the Ti He-a line at 4.75 keV as probe light [10]. A new alternative emerged with the development of VUV-free electron lasers (VUV-FEL), providing pulses of coherent radiation in the far

* Corresponding author. E-mail address: carsten.fortmann@uni-rostock.de (C. Fortmann). 1574-1818/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.hedp.2006.04.001

Au

rs o

(vacuum-) ultraviolet. At the moment, the VUV-FEL at DESY Hamburg operates at 32 nm wavelength [11], corresponding to 38 eV photons. With this coherent light source, dense matter up to solid densities of 1023 cm3 can be penetrated, see Refs. [12,13]. Under these conditions, the Thomson spectrum permits the determination of electron temperature and density directly from the position and height of collective resonances, i.e. plasmons, showing up in the scattering signal [9]. First experiments will be performed in the near future at the VUVeFEL facility at DESYat l 32 nm FEL wavelength, while in later stages of the project, wavelengths from 13 nm (VUVeFEL) down to 0.1 nm (X-FEL) will be available. Due to the large number of free charge carriers at the temperatures and densities considered, thermal bremsstrahlung emission, resulting from inelastic freeefree scattering, contributes signicantly to the emission background. Therefore, experimental conditions such as scattering angles, spectral properties of the probe and the detector have to be chosen as to obtain a maximum signal-to-background ratio. Background is to be understood as bremsstrahlung radiation, whereas signal corresponds to the photons having undergone Thomson scattering.

na

co

a Institute for Physics, Rostock University, 18051 Rostock, Germany Theoretical Physics Institute, University of Alberta, Edmonton, Canada T6G 2J1

py

120

58

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

So far, classical formulas for the bremsstrahlung emission level going back to Kramers [14] have been used to determine threshold intensities of the external source to overcome the background due to bremsstrahlung [15,16]. The Kramers result, given below in Eq. (19), is derived from the assumption of Keplerian trajectories of the emitting electron in the Coulomb eld of an ion and integration over all initial velocities weighted with the Maxwellian velocity distribution function. Quantum features are only accounted for in a semiclassical way: the velocity integral extends over velocities y, fullling the condition my2/2 ! hu, i.e. the kinetic energy has to be larger than the photon energy. Further quantum properties, such as the nite photon momentum as well as the quantum mechanical nature of the scattering process are not accounted for. By comparing the Thomson signal strength at the laser wavelength l 14.7 nm to the bremsstrahlung photon yield calculated from Kramers formula, Baldis et al. found threshold intensities of 1013 W/cm2 for typical values of free electron density ne 1022 cm3 and temperature kBT 100 eV [16]. In this work, we evaluate threshold conditions (intensities) using improved expressions for the bremsstrahlung spectrum. As usual, corrections to Kramers formula for bremsstrahlung are described by the so-called Gaunt factor [17]. In the simplest approach it is obtained by taking into account collisions between electrons and xed ions in Born approximation. In dense plasmas, many-particle effects as dynamical screening become important. Moreover, strong collisions have to be accounted for. We show in this paper how the Gaunt factor can be derived from linear response theory [18] in a general way. Within this framework, modications of the emission spectrum beyond Born approximation can be included in a systematic manner [19] as will be discussed later. We then apply our formulas to determine the threshold intensities for a broad range of experimental parameters (wavelength, spectral properties of detectors, and different materials), relevant for future experiments at DESY. Furthermore we compare Thomson and bremsstrahlung photon yield over a nite spectral range. Thereby, and by taking improved expressions for the bremsstrahlung cross section, we show that even higher thresholds have to be reached in order to obtain a Thomson signal above the bremsstrahlung level at-least near the plasmon resonances. These peaks are much lower than the central peak, being essentially an ion feature. The present work is organized as follows: in the rst section we review the basic physics of Thomson scattering and bremsstrahlung and how they can be expressed in terms of the dynamic structure factor and the dielectric function, respectively. Since these two quantities are related to each other via the uctuationedissipation theorem [20], we are able to describe both processes on a common and consistent basis. We then compare the emission level due to bremsstrahlung to the Thomson signal whose strength is proportional to the ux of incoming photons, i.e. the power density of the external source. Thereby, we nd expressions for the threshold power density as a function of particle density and temperature. In the last section we discuss our results for various sets of experimental parameters relevant for future experiments at the VUV-FEL.

2. Thomson scattering and bremsstrahlung The central quantity of interest is the spectral power density dP/dV dl dU, i.e. the rate of energy radiated per unit scattering volume dV, wavelength dl, and solid angle dU. The total spectral power density is the sum of the corresponding quantity for every radiative process in the plasma. In this work we focus on Thomson scattering and bremsstrahlung, i.e. d3 Ptot d3 PTh d3 Pbr : 1 dV dl dU dV dl dU dV dl dU To unambiguously identify the Thomson signal, we require that the Thomson power density is at least equal to the bremsstrahlung level, d3 PTh d3 Pbr ! : 2 dV dl dU dV dl dU The Thomson spectrum is given by the intensity of the probe laser IL and the Thomson scattering cross section d2sTh/dudU. To account for the nite spectral bandwidth of the detector, one has to convolute each power spectrum with a detector function G(l). In practice, this is only relevant for the Thomson signal since the bremsstrahlung spectrum is slowly varying in the relevant frequency region. The Thomson power spectrum reads Z d2 sTh ul d3 PTh l IL dlGl l IL Rl: 3 dV dl dU dU du We have introduced the response function Rl, where the bar denotes the convolution with the detector function. Note that we assume an optically thin plasma, thus radiation transport is neglected. Also, due to the short pulselength of the VUV-FEL (20e120 fs), we neglect the heating of the plasma due to the probe beam. The bremsstrahlung spectrum does solely depend on the plasma parameters density and temperature, so that, with a suitable formula for the bremsstrahlung spectrum, which we will abbreviate by the notation d3Pbr/dV dl dU h j(l) in the following, Eq. (2) denes a threshold intensity Ithresh l jl : Rl 4

th

or

's

pe

Au

rs o

We will now briey describe how expressions for the Thomson scattering and the bremsstrahlung spectrum can be obtained from a common starting point, i.e. the dielectric function of the plasma.

2.1. Thomson scattering The cross section for Thomson scattering in a plasma can be given in terms of the dynamic structure factor (DSF), Sk; u: d2 sTh k; u dsU k1 Sk; u; dUdu dU Th k0 5

na

co

py

121

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69 59

Sk; u

th

or

This relation can be utilized by applying an appropriate approximation to the dielectric function. Due to the mass ratio me =mi ( 1, Thomson scattering on ions can be neglected. Therefore, we will approximate Sk; u by See k; u, the electronic DSF. 2.2. Bremsstrahlung giving the BetheeHeitler cross section, see Ref. [23]. The transition amplitude involves a longitudinal eld (Coulomb eld) i.e. a scattering partner, say an ion of effective charge Zeff. Thus, bremsstrahlung is naturally of second order in density. A free electron does not emit bremsstrahlung, unless collisions take place. The challenge is then to accurately describe the scattering process itself. Born approximation, as given by Eq. (10) does not describe the correct behaviour of the bremsstrahlung spectrum in the case of collisions involving high transfer momenta, so-called strong scatterings. These can be included by ladder-summation of all one-photon exchange processes, which leads to the t-matrix. Thereby, the electroneion interaction is treated accurately in all orders. In the Coulomb limit, one nds

Radiative freeefree transitions of electrons, known as bremsstrahlung, represent the main source of emission in a hot, fully ionized plasma. However, the bremsstrahlung process requires the presence of a scattering partner that carries the recoil momentum. Kinematically, emission by electrons which scatter on ions is dominant. For a thermally equilibrated plasma, emission, characterized by the spectral power density j(u), and the absorption coefcient a(u) as the relative attenuation of the intensity of electromagnetic waves propagating in the medium per unit length, are linked by Kirchhoffs law [25] j(u) L(u)a(u), with the Planck distribution Lu hu3 =

Au

's

pe

Here, re e2/4pe0mec2 is the classical electron radius, q is the scattering angle between k1 and k0, and 4 is the angle enclosed between the plane of polarization and the scattered wavevector k1. For unpolarized light, averaging over possible all 2 polarizations yields ds=dUTh re 1 cos2 q 2. For the VUV-FEL, the light is linearly polarized and Eq. (6) should be applied. However, details of the experimental setup with respect to the direction of the polarization plane are still open. Henceforth, we use the unpolarized Thomson cross section, thereby eliminating the dependence on the parameter 4. Furthermore, with regard to the comparison of Thomson scattering to bremsstrahlung, we keep the estimation of the Thomson signal as restrained as possible. Eq. (5) shows that plasma collective properties can be measured in a Thomson scattering experiment. However, this requires an accurate theory of the dynamic structure factor. The dynamic structure factor Sk; u is closely related to the longitudinal dielectric function by the uctuationedissipation theorem [24] 7

where the index of refraction n(u) is also linked to the transverse dielectric function by 1 1=2 nu pRe et u jet uj : 2 9

rs o

In the long wavelength limit, the longitudinal dielectric function, which appears in the uctuationedissipation theorem (7) and the transverse dielectric function coincide [27]. Thus, relations (7) and (8) enable us to treat both Thomson scattering and bremsstrahlung on a common basis, namely by using an appropriate theory for the longitudinal dielectric function. 2.3. Consistent approximations

The aim of this work is to compare both radiation processes, Thomson scattering and bremsstrahlung, calculated in a consistent approximation. The comparison has to be carried out between the contributions of either process in leading order of density. Bremsstrahlung occurs in second order of the coupling constant aQED 1/137 as can be seen from the transition amplitude wbr expressed by Feynman diagrams, fi

na

co

see Refs. [21,22] for details. The variables k; u are related to the transferred linear momentum and energy, respectively, while k0 u0/c and k1 jk0 kj are incoming and outgoing linear momenta of the laser eld, respectively. (ds(U)/dU)Th is the Thomson scattering cross section for the isolated scattering event. It is given by the KleineNishina formula, its derivation can be found in standard textbooks of quantum electrodynamics, e.g. Ref. [23]. In the nonrelativistic limit, the Thomson cross section for linearly polarized light is given by ds r2 e 1 cos2 4 sin2 q : 6 dU Th 2

au

u Im et u ; c nu

py

4p3 c2 exphu=kB T 1. The wavelengths which are of interest here are in the 10e100 nm-range and large against atomic length scales of a few A, so that the long wavelength limit k / 0 is applied throughout this paper. Therefore, the emissivity j(u) and absorption a(u) are only functions of frequency u in this limit. For effects of nonlocality on the bremsstrahlung spectrum, we refer to Ref. [26]. The absorption coefcient can be obtained from the imaginary part of the transverse dielectric function according to 8

122

60

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

rs o

where b 1/kBT is the inverse temperature, U0 is a normalization volume. The current operator for particles of species c is dened as Jk;c 1 X qc hpay pk=2;c apk=2;c : U0 p mc 15

Au

does not contain any longitudinal eld and therefore behaves regularly over the whole spectral range. Thus, the DSF, which describes the Thomson scattering in the medium, is taken in random phase approximation (RPA) that accounts for the dynamical screening. This is all the more important, as the Thomson scattering cross section is evaluated in the vicinity of the plasma frequency where dynamical screening is the dominant effect [20]. Collisions are neglected, since they occur in the next to leading order in density. Furthermore, we explore plasma parameters, where the electroneion collision frequency nei is small compared to the plasma frequency upl. To conclude, a consistent comparison between Thomson scattering and bremsstrahlung in media asks for the cross section of either process in leading order in density. This consideration leads us to the RPA for the dynamic structure factor and a t-matrix ladder-summation for the bremsstrahlung spectrum.

pe

's

one expresses the currentecurrent correlation function by _ _ a forceeforce correlation function hJ z ; J z 0 iuih , using the k;c k;c _ time derivative of the current J k;c iH; J k;c =h. The forcee force correlation function is more suited for a perturbative treatment, see Ref. [20] and Appendix A, where also denitions and useful properties of correlation functions are given. 4. Calculation of the bremsstrahlung spectrum Since the bremsstrahlung spectrum is evaluated at frequencies far above the plasma frequency, we can use the high frequency limit of Eq. (12) in order to derive the absorption coefcient. For details, we refer to Appendix A. In the high frequency limit, the absorption coefcient (8) is given by [20,19] au bU0 D _ z _ z E pU0 Re J 0;e ; J 0;e Im GJJ u ih: __ 2 uih ce0 u ce0 u3 16

th

or

Thomson scattering and bremsstrahlung can be determined by the longitudinal dielectric function el k; u. We will now briey outline how the longitudinal dielectric function can be obtained in the framework of linear response theory. A discussion of its fundamental aspects can be found in Ref. [18], for various applications to optical properties of plasmas, see Ref. [20]. In general, the longitudinal dielectric function

Here we introduced the forceeforce Green function GJ J u, __ dened in Eq. (A.6), which can conveniently be calculated using Feynman diagrams, see Fig. 1(a). G4 denotes a four-particle Green function that contains all interactions between electrons and ions [19]. In the long wavelength limit considered here, eee and iei collisions vanish due to momentum conservation. We perform a sequence of approximations by selecting certain diagrams contributing to G4.

na

Sommerfelds expression for bremsstrahlung, see Eq. (22) below and Ref. [28]. On the other hand, due to the long range behaviour of the Coulomb potential, Born approximation as well as the Sommerfeld result suffers from infrared divergencies. Here, the account for screening leads to convergent results. For details, we refer to Ref. [19]. In our calculations we will use Sommerfelds formula Eq. (22). Since we are interested in the bremsstrahlung spectrum at high frequencies, i.e. in the vicinity of the FEL frequency (hu0 38 eV), no screening effects are considered here. Note that the use of the t-matrix does not increase the order in density, instead, it gives the accurate Gaunt factor in leading order of density. Further many-particle effects like self-energy and vertex corrections lead to higher order contributions in density and, therefore, are not considered here. For a consistent treatment of the self-energy of the scattering electron, given by multiple scattering on ions and its impact on the bremsstrahlung spectrum, see Ref. [29]. In the case of Thomson scattering, we adopt the same strategy: we calculate the contribution from lowest order in density. Unlike bremsstrahlung, already the rst order in density gives a nite contribution. The amplitude for Thomson scattering in second order of the interaction

el k; u is given in terms of the dielectric susceptibility ccc0 k; u, where c and c0 label the different species in the plasma and further quantum numbers such as spin, el k; u P 1 : cc0 Vcc0 kccc0 k; u 12

Within linear response theory, the Kubo formula relates the susceptibility (response function) to the currentecurrent correlation function [27] ccc0 k; u ibU0 k2 D z z E ; J ;J 0 uqc qc0 k;c k;c uih 14

co

Vcc0 k

q c q c0 : e0 k 2

py

123

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69 61

K

19

multiplied with a correction factor gff u, called Gaunt factor [17], which takes into account medium effects as well as quantum corrections, i.e.

Fig. 1. Diagrammatic representation of GJ J um . (a) full account of all me__ dium effects by a four-particle Green function, (b) Born approximation, (c) t-matrix approximation.

co

4.1. Born approximation As the lowest order contribution with respect to the interaction potential, Fig. 1(b) shows the Born approximation. This approach reproduces the well-known BetheeHeitler formula in the nonrelativistic limit:

p 3 expbhu=2K0 bhu=2: p

py

ju jK u$gff u:

20

21

na

1 kB T

The Born approximation assumes free particles as the in and out states in the scattering amplitude. Taking the true scattering states, i.e. Coulomb wavefunctions, leads to the so-called Sommerfeld formula [28], gS u; T ff

l

N Z

rs o

22

with

pe

p 4 3 S gff u; Ei p

Ec is the free particle energy, for nonrelativistic particles p Ec h2 p2 =2mc holds. Details of the calculation are disp cussed in Ref. [29]. In Eq. (17) the effective ion charge Zeff is used to account for the screening of the charge Ze of the nucleus due to inner shell electrons. In Ref. [16], Zeff is calculated as a function of temperature and density in the framework of ThomaseFermi theory, see Ref. [30] for details. Alternatively, Zeff can be obtained in the socalled chemical picture by solving Sahas equation, see Ref. [31]. Especially for high temperatures, the chemical picture gives more reliable results, as the ionization equilibria between different ionization stages are satised. We will therefore use the chemical picture in this work and calculate the mean ionization level with COMPTRA04 [31]. The ThomaseFermi model will only be applied for the purpose of comparison of results obtained by Baldis et al. [16]. The logarithm in Eq. (17) is the nonrelativistic limit of the BetheeHeitler cross section [32,33]. The drawback of this result is the divergence of the bremsstrahlung spectrum in the limit u / 0. However, physically this is of no importance, since for low frequencies the index of refraction is different from unity and modies the spectrum signicantly. This is known as dielectric suppression [34]. It is in common use to write formulas for the emission and absorption due to (inverse) bremsstrahlung in terms of Kramers classical result [14]

h2 h2 2h2 h2 i f i f

23

and " p #l1 p pihi hf E E 1 4 Ei Ef i f Il p p2 epjhi hf j=2 p p Ei Ef 4 Ei Ef Gl 1 ihi G l 1 ihf G2l 2 2 F1 l 1 ihf ; l 1 ihi ; 2l 2; p ! 4 Ei Ef p p2 ; Ef Ei hu: Ei Ef 24

Au

th

or

's

2 h2 Zeff Ry=Ei=f is the Sommerfeld parameter with the Rydi=f berg energy Ry me e4 =24pe0 h2 x13:6 eV; 2 F1 a; b; c; d, is the hypergeometric function [35] and G(x) is the Gamma function. As shown in Ref. [19], Sommerfelds expression is also obtained by a t-matrix ladder-summation with a statically screened Debye potential (Fig. 1(c)) if the limit of vanishing inverse screening length (Coulomb limit) is taken. It gives the correct Gaunt factor in the low density limit, which is considered here.

124

62

3 Born Sommerfeld

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

uctuationedissipation theorem (7) in connection with Eq. (12) allows to express the electronic part of the dynamical structure factor via the electronic dielectric susceptibility cee k; u, See k; u Im cee k; u h : pne 1 exp hu=kB Te 26

1000 eV 1000 eV

_ gff()

h [eV]

Fig. 2. Gaunt factor gff u in Born approximation Eq. (21) and using Sommerfelds formula Eq. (22), respectively. Results are presented as a function of the photon energy -u for various temperatures.

V k; u is the screened interaction potential which acts between particles of species c and c0 . It satises the equation X sc sc Vcd kc0 k; uVdc0 k; u: 28 Vcc0 k; u Vcc0 k d

d

sc

Au

Fig. 2 shows the dependence of the Gaunt factor on the photon energy for electron temperatures kBT 10 eV, 100 eV and 1000 eV. Over the large energy interval shown, the Gaunt factor in either calculation for a xed temperature does not vary much and is of order unity, thus the widely used approximation to set gff uz1 is justied for estimates of the emission level as done in Refs. [15,16]. More specic, gff u 1 is a good approximation for photon energies comparable to the temperature -u=kB Tx1 and for temperatures in the vicinity of the ionization energy (kB TxZ 2 Ry for hydrogenic systems). If one of these conditions is not met, one should use the full Sommerfeld expression (22) or appropriate approximations, see the detailed discussion in Ref. [36]. For the VUV-FEL experiments, both requirements are satised only roughly: the laser provides photon energies of hux40 eV and the optical pump laser will excite the plasma to temperatures of kB Tx10.50 eV, which is of almost the same order of magnitude as the rst ionization energies of aluminum (6 eV, 19 eV, 28 eV), or hydrogen (13.6 eV) [37], materials presumably used as target in the experiments. Thus we take into account the Gaunt factor in t-matrix approximation (Eq. (22)) in our calculations. In Ref. [19], a consistent treatment of the impact of dynamical screening and strong collisions on bremsstrahlung based on the GouldeDeWitt scheme [38] is given. It is shown that the high frequency behaviour of the Gaunt factor is dominated by the t-matrix contribution, whereas dynamical screening can be neglected as long as the considered frequencies are large compared to the plasma frequency.

c0 k; u ibU0 c dcc0

rs o

na

The free susceptibility c0 k; u is obtained from Eq. (14), c taking the currentecurrent correlation function only in zeroth order with respect to the interaction, i.e.

p c Ec pk=2 Epk=2 hu ih

co

x Zc 0

py

10eV

We will consider the classical limit, where the denominator of Eq. (26) can be approximated by hu/kBTe. In RPA, the following equation for the susceptibility holds [40]: 27

29

fpc is the momentum distribution function of species c. Again, a small but nite imaginary frequency ih, h > 0 has to be introduced, in order to x the sign of the imaginary part and to obtain convergent results. In the case of a classical two-component plasma with different temperatures Tc, we take the Maxwell distribution. In the limit - / 0 the resulting susceptibility reads c0;cl k; u U0 nc Wxc =kB Tc ; c with the plasma dispersion function Wxc 1 2xc e

2 xc

pe

30

's

or

p 2 2 et dt i pxc exc ;

31

p and the dimensionless variable xc u2 mc =2k 2 kB Tc . Eqs. (27) and (28) can be solved algebraically for the RPA-susceptibilities cRPA k; u as shown in Appendix B, see also cc0 Ref. [40]. Evaluation of Eq. (26) with cRPA k; u, see Eqs. ee (29) and (B.5), yields the electronic DSF in RPA, i.e. 2 2 xe exp xe 1 Zeff a2 Te =Ti Wxi p See k; u 1 a2 Wxe a2 Zeff Te =Ti Wxi u p 2 a2 Wxe Zeff 1 a2 Wxe a2 Zeff Te =Ti Wxi xi exp xi2 p : 32 u p Here, the scattering parameter

The total dynamic structure factor is dened as [39] N Z 1 Scc k; u dt eiut hrc k; trc k; 0i; 25 2pN

N

h.i denotes the ensemble equilibrium average. Here, our discussion is focused on the electron structure factor See k; u. The

th

125

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69 63

a ke =k

p ne e2 =e0 kB Te =k

33

dP/dVdd [W/cm3 nm sr]

4105

th

has been introduced. It separates the regime of collective Thomson scattering a[1, where collective resonances (plasmons) appear in the spectrum, and the regime of noncollective Thomson scattering a ( 1, where the spectrum reects the single-particle distribution function of electrons without further structures. Expression (32) was also used by Baldis et al. [16]. Eq. (32) for the DSF contains two terms: the rst term is basically the DSF of a free electron gas. However, it also contains the ionic dispersion function, which accounts for the screening effect of ions. The second term gives the scattering signal from electrons which are close to ions and are therefore determined by the dynamics of these heavy particles. This term is dominant at small u where W(xe) can be approximated by its static limit, i.e. Wxe ( 1/1. For larger values of u, the rst term dominates the spectrum, because the ion part is damped out more rapidly than the electron part. In particular, plasmon resonances, which appear in the spectrum due to a vanishing real part of the denominator, only survive in the rst term, while they are damped out in the second. In socalled Salpeter approximation [41] the DSF (32) is separated into two terms depending only on electronic and ionic variables, respectively, electroneion correlations are neglected to a large extent. Therefore, the full expression (32) should be used, as done throughout this work. Chihara [22] gives the DSF in terms of the local eld correction factor, thereby including electroneion collisions. In appropriate limits, his expression coincides with Eq. (32), see Ref. [21]. On the other hand, due to the equivalence of the local eld correction and the collision frequency in the high frequency limit, Chiharas expression leads to the same formula for the absorption coefcient a(u) as given in Eq. (8) [42]. Thus, Chihara presents an alternative approach to the question of photoabsorption and emission and Thomson scattering in plasmas starting from the DSF, whereas in this work the dielectric function is the central quantity. Having the thermal emission spectrum (20) and the dynamic structure factor (32) at our disposal, we are now in the position to compare the signal from Thomson scattering to the bremsstrahlung background thereby treating both processes on a common basis.

3105

Bremsstr. (Kramers) Bremsstr. (Born) Bremsstr. (Sommerfeld) 10 2 Thomson (IL=5x10 W/cm ) Thomson, not convoluted

2105

1105

Fig. 3. Emission power spectra for Thomson scattering and bremsstrahlung. The black dashed curve represents the unconvoluted Thomson spectrum. Parameters: kBTe 10 eV, ne 1020 cm3, laser wavelength l 32 nm, scattering parameter a 1.25, Zeff 1.

or

's

pe

rs o

na

To model the detector, we assume a detector function G(l) of Gaussian shape, its width given by the relative spectral bandwidth Dl/l Du/u [16] 34

For the case of bremsstrahlung, we neglect the effect of the nite bandwidth of the detector, since the corresponding spectrum is only slowly varying with frequency. The effective ion charge Zeff, calculated with COMPTRA04 [31] is close to Zeff 1 in the case of hydrogen for the present values of electron density and temperature. In Fig. 3, the black dashed curve represents the pure Thomson signal, i.e. no convolution with the detector function G(l) has been performed. For the present parameters ne 1020 cm3 and kBTe 10 eV, the Thomson spectrum contains a very narrow ion peak, situated at the laser wavelength, and two satellites, which can be identied as electronic plasmon resonances. The central peak dominates these electronic resonances by a factor of 10, approximately. The solid black

8109

Bremsstr. (Kramers) Bremsstr. (Born) Bremsstr. (Sommerfeld) Thomson (IL=1013 W/cm2)

6109

6. Results

Au

4109

We now compare power spectra for Thomson scattering and bremsstrahlung emission. Figs. 3e6 show results for different combinations of electron density and temperature for a hydrogenic plasma. We consider densities ne 1020 cm3 and 1022 cm3 and temperatures of 10 eV and 50 eV as examples for possible experimental conditions with cryogenic targets. Furthermore, backscattering geometry (scattering angle 120 ) is chosen and a VUV laser wavelength of l 32 nm is assumed.

2109

0 30

co

31

0 30

py

32

31

33

34

[nm]

32

33

34

[nm]

Fig. 4. Emission power spectra for Thomson scattering and bremsstrahlung. Parameters: kBTe 10 eV, ne 1022 cm3, laser wavelength l 32 nm, scattering parameter a 12.5, Zeff 1.

126

64

5106

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

4106

3106

2106

1106

[nm]

Fig. 5. Emission power spectra for Thomson scattering and bremsstrahlung. Parameters: kBTe 50 eV, ne 1020 cm3, laser wavelength l 32 nm, scattering parameter a 0.56, Zeff 1.

th

81010

41010

21010

0 30

Au

61010

or

curve is obtained by convolution of the pure Thomson signal (dotted black curve) with the detector function. The detector resolution is set to Dl 0.32 nm, which corresponds to 1% of the central wavelength. Due to the broad detector function, the central peak is lowered and broadened such that the plasmon resonances do not show up as separate structures anymore. Instead, they only provide sidewings in the spectrum. In Fig. 4 (ne 1022 cm3 and kBTe 10 eV), the electronic peaks are totally absorbed in the central peak. In this case, the ion peak in the unconvoluted signal (not shown) is several orders of magnitude larger than the electronic resonances, which therefore do not contribute to the convolution integral (3). Fig. 5 shows spectra for ne 1020 cm3 and kBTe 50 eV. At these parameters, one obtains a 0.56 < 1 for the scattering parameter, we are in the noncollective regime. Already the unconvoluted Thomson signal is free of sharp electronic resonances, only two shoulders appear in the broad, noncollective signal. Finally, in Fig. 6 (ne 1022 cm3 and kBTe 50 eV), we have the same structure as in Fig. 4, no plasmon peaks are visible due to the dominance of the ion feature. In all four calculations of the Thomson spectrum, collisions have been neglected, since the electroneion collision frequency

31

32

's

33

[nm]

Fig. 6. Emission power spectra for Thomson scattering and bremsstrahlung. Parameters: kBTe 50 eV, ne 1022 cm3, laser wavelength l 32 nm, scattering parameter a 5.6, Zeff 1.

pe

34

rs o

na

co

30

31

32

33

34

nei [43] is small as compared to the plasma frequency. Even for the most unfavorable set of parameters (Fig. 4), the ratio nei/upl 0.16 is still allowing for collisionless approximation. In Figs. 3e6, the intensity of the laser is chosen such that the central peak is situated clearly above the bremsstrahlung level. Three approximations for bremsstrahlung are shown, namely Kramers formula (19), Born approximation (18) and Sommerfelds formula (22). Born approximation gives larger deviations from Kramers result than the Sommerfeld expression. This corresponds to the behaviour of the Gaunt factor shown in Fig. 2: the Sommerfeld result is always closer to unity than Born approximation. In the following, we evaluate the threshold intensity Ithresh(u) dened in Eq. (4) using Eqs. (3), (5), and (32) for the Thomson power spectrum and Eq. (20) with the Gaunt factor in either Born approximation (21) or t-matrix approximation (22) for the bremsstrahlung power spectrum. Figs. 7e10 show contour plots of the threshold intensity in part of the neeTe-plane. In Fig. 7 we compare Kramers result (dashed curves) to Born approximation (dotted curves) assuming aluminum as target material, scattering angle q 20 , Dl/l 104. In this case, the effective ion charge Zeff is calculated using ThomaseFermi theory [30] as was also done by Baldis et al. [16], who used the same set of parameters. Their results are reproduced by using Kramers formula for bremsstrahlung (dashed curve). For low temperatures, where the Gaunt factor in Born approximation is smaller than unity at the considered photon energy, cf. Fig. 2, higher densities are accessible as compared to Kramers result. For high temperatures, the opposite becomes true, gB u > 1 leads to lower accessible densities. ff In Figs. 8e10, three approximations have been calculated for the bremsstrahlung level. Besides Kramers formula (dashed curves) we show results for Born approximation (dotted curves) and Sommerfelds formula (solid curves). Comparing the Sommerfeld result to Born approximation, it can be noted that Born approximation gives larger corrections, while Sommerfelds theory leads to smaller deviations from Kramers result. This is one important result of this work: taking into account quantum effects in a rigorous way via Sommerfelds expression for the Gaunt factor leads only to small corrections of threshold intensities, while Born approximation tends to overestimate these effects. This underlines the importance to go beyond Born approximation. For moderate and high temperatures, the Sommerfeld result lies systematically below the Kramers result, due to the increasing Gaunt factor at high temperatures. Furthermore, we investigated the inuence of other experimental parameters, namely the laser wavelength, the material, and the spectral bandwidth of the spectrometer. The latter parameter turns out to be of great importance: by comparison of Fig. 8 with Fig. 9, one observes that notably higher densities are accessible in the case of small spectral bandwidth (Dl/ l 104 in Fig. 9) than for the relatively large bandwidth (Dl/l 102 in Fig. 8). Comparing Fig. 7 to Fig. 9, the inuence of the Z-number of the material becomes apparent: low Z-materials produce

py

127

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

1024

1015 W/cm2 1014 W/cm2

65

1023

Born Kramers

ne [cm-3]

1013 W/cm2

1011 W/cm2

108 W/cm2

1010 W/cm2

1020 10

100

co l

1015 W/cm2

2

1021

1012 W/cm2

kBTe [eV]

Fig. 7. Threshold curves in the densityetemperature plane for Al. Parameters: laser wavelength 14.7 nm, q 20 , Dl/l 104, Te/Ti 2. For bremsstrahlung, three approximations are applied: Kramers (dashed curves), Born (dotted curves), and Sommerfeld (solid curves).

less bremsstrahlung than high Z-elements due to the Z2proportionality of the bremsstrahlung cross section, cf. Eq. (19). On the other hand, the Z-dependence of the dynamical structure factor largely cancels out for the free electron part of the structure factor while it is roughly Z/(1 Z )2 for the ionic part, cf. Eq. (32). Finally, important differences are noted upon changing the laser wavelength from 32 nm (Fig. 9) to 13 nm (Fig. 10) especially at low temperatures. Both data sets have been calculated using the same spectral resolution (Dl/l 104) and material (H). Looking at Fig. 10 (l 13 nm) at low temperatures the threshold contours are almost parallel to the density axis. For higher temperatures, both wavelengths 32 nm and 13 nm give nearly equal threshold intensities, which do not depend

on temperature. This can be understood from the bremsstrahlung spectrum, which becomes nearly independent of frequency and temperature at low hu/kBT. Since the information about temperature and density is stored in the position and height of the plasmon resonances, we will now focus on experimental conditions to be met in order to separate the plasmon peak from bremsstrahlung background. Results for the threshold intensity are given in Table 1. The electron density is set to ne 1021 cm3, for ne 1020 cm3 and 1022 cm3 see Figs. 3e6. Two wavelengths are considered, l 32 nm, the momentary VUVeFEL wavelength at DESY, and l 13 nm, envisaged wavelength in the near future. Since the latter wavelength allows for very efcient X-ray optics to be applied, the spectral resolution has

1024

or

's

ne [cm-3]

th

1023

pe

Au

1022

10

21

12

1010 W/cm2

1020

10

rs o

1014 W/cm2 10 W/cm

13

100

na

kBTe [eV]

Fig. 8. Threshold curves in the densityetemperature plane for H. Parameters: laser wavelength 32 nm, q 120 , Dl/l 102, Te/Ti 2. For bremsstrahlung, three approximations are applied: Kramers (dashed curves), Born (dotted curves), and Sommerfeld (solid curves).

py

1000 1000

1022

128

66

1024

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

1014 W/cm2

1023

1013 W/cm2

ne [cm-3]

10 W/cm

1020 10

1010 W/cm2

100

kBTe [eV]

Fig. 9. Threshold curves in the densityetemperature plane for H. Parameters: laser wavelength 32 nm, q 120 , Dl/l 104, Te/Ti 2. For bremsstrahlung, three approximations are applied: Kramers (dashed curves), Born (dotted curves), and Sommerfeld (solid curves).

been reduced to Dl/l 104, while 32 nm wavelength allows only for Dl/l 102. Finally the scattering angle has been chosen such that the plasmon peak is on the one hand well pronounced but on the other hand shifted far enough from the central peak as to be resolved by the spectrometer. The wavelength shift of the plasmon resonance from the FEL wavelength is given in the fth column labelled by Dlres. For detailed discussion of the Thomson scattering spectrum at various plasma parameters, we refer to Refs. [12,13]. The rightmost column of Table 1 gives the Gaunt factor in Sommerfeld approximation (Eq. (22)). Note that in the case of the 32 nm wavelength, the threshold intensity is increased by 14e22% due to the Gaunt factor.

7. Conclusion

1024

or

's

pe

rs o

1012 W/cm2

In this work we have studied Thomson scattering and bremsstrahlung emission in warm dense matter in the context of plasma diagnostic experiments to be performed in near future. Out of these two competing, not only Thomson scattering is supposed to serve as a probe for plasma parameters, but also bremsstrahlung gives an important contribution to the photon yield from highly ionized plasmas. Thus, experimental conditions have to be determined that allow for a maximum signal-to-background ratio. Here, we focused on the laser intensity. To this purpose, expressions for the Thomson signal, which is given by the dynamical structure factor of the plasma as well as for the

na

1015 W/cm2 1014 W/cm2 1013 W/cm2 Born Kramers Sommerfeld

ne [cm-3]

Au

th

1023

10

22

1011 W/cm2

1021

108 W/cm2

1010 W/cm2

10

20

10

100

kBTe [eV]

Fig. 10. Threshold curves in the densityetemperature plane for H. Parameters: laser wavelength 13.0 nm, q 120 , Dl/l 104, Te/Ti 2. For bremsstrahlung, three approximations are applied: Kramers (dashed curves), Born (dotted curves), and Sommerfeld (solid curves).

co

1000 1000

1021

1011 W/cm2

py

1022

1012 W/cm2

129

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69 Table 1 Experimental parameters and resulting threshold intensities for different electron temperatures and laser wavelengths kBTe [eV] 10 20 50 10 20 50 l [nm] 32 32 32 13 13 13 q 120 120 120 120 45 45 Dl/l 102 102 102 104 104 104 Dlres [nm] 1.1 1.2 1.4 0.3 0.2 0.2 Ithresh [W/cm2] 1.37 1012 4.92 1012 9.29 1012 2.95 109 1.55 1011 3.49 1012 gS ff 1.14 1.17 1.22 1.03 1.04 1.06 67

Appendix A. Correlation functions and green functions Correlation functions for two observables A, B are dened according to 1 A; B b

b Z

N Z

dt eiuiht At; B;

py

:

A:1

Au

bremsstrahlung spectrum have been derived from a common starting point, namely linear response theory. This approach allows for a systematic treatment of medium and quantum effects, such as dynamical screening and strong collisions. We have applied our formulas to determine threshold intensities for the external photon source (FEL) as function of density and temperature as well as their dependence on further parameters, namely the laser wavelength, the spectral resolution of detectors, and the material. In the discussion we focused on the bremsstrahlung spectrum. It was shown that Born approximation overestimates the effect of collisions, i.e. it leads to relatively high Gaunt factors if compared to Sommerfelds expression. Sommerfelds formula (22) gives the correct Gaunt factor in a weakly coupled plasma. At temperatures kBTe ! 20 eV threshold intensities calculated with the Gaunt factor in Sommerfeld approximation are larger than those obtained by Baldis et al. [16], who used Kramers expression, i.e. gff 1. The Thomson signal was analysed within RPA which gives the contribution in lowest order of density. For densities investigated here, collisions are also relevant for the Thomson signal and lead to a pronounced change of the electronic part in Sk; l. This was investigated in Ref. [44]. However, the account for collisions does not alter the results for the threshold intensities performed in this work, since the Thomson scattering signal was evaluated at the laser wavelength, where the ionic feature of the DSF dominates. Finally, we have demonstrated that Thomson scattering can easily overcome the bremsstrahlung background, if laser intensities of 108e1013 W/cm2 are provided. From this point of view, Thomson scattering experiments for plasma diagnostics using VUVeFEL radiation seem feasible.

r0 is the equilibrium statistical operator. The time dependence of observables is taken in the Heisenberg picture, using the system Hamiltonian H X h2 p2

p;c

2mc

ay ap;c p;c

co

Dlres is the position of the plasmon resonance in the Thomson spectrum relative to the laser wavelength. Fixed parameters: Z 1, ne 1021 cm3.

A:2

pe

th

or

's

rs o

with

Performing integration by parts, the currentecurrent correlation function (16) can be expressed through a forceeforce correlation function as " z z z z i uq2 1 _ _ c1 k; u z z iu Jk ; Jk J k ; J k uih bU0 k 2 Jk ; Jk 2 # z z z _ _ J k ; Jk uih Jk ; J z uih k z z ; A:3 Jk ; Jk uih

h i. z _ J z i H; J0;e h: 0;e Also, it is convenient to introduce a generalized collision frequency n(u) in analogy to the Drude relation [20] P 2 eu 1 u2 =u2 iunuwhere u2 pl pl c nc e c = e0 mc is the squared plasma frequency. z z By comparison with Eq. (12) using Jk ; Jk e0 u2 =bU0 ; pl we establish an expression for the collision frequency in terms of correlation functions " # z z z _ _ J k ;Jk uih Jk ; J z uih z z bU0 k _ _ z z nu lim J k ; J k uih : A:4 e0 u2 k/0 Jk ;Jk uih pl Further details can be found in Ref. [20]. Making use of Eq. (8), the absorption coefcient can be expressed as au u2 Re nu pl 2 2u Im nu jnuj2 nu c u A:5

Acknowledgements

This work has been supported by the Virtual Institute VH-VI-104 Plasma Physics Research using FEL Radiation of the Helmholtz Society and the DFG-Sonderforschungsbereich 652 Starke Korrelationen und kollektive Phanomene im Strahlungsfeld: Coulombsysteme, Cluster und Partikel. We gratefully acknowledge A. Holl and J. Chihara for stimulating discussion. C.F. would like to thank DESY for hospitality and support during a two week visit.

In the high frequency limit u[upl , the index of refraction is unity, the imaginary part of the collision frequency tends to zero and the collision frequency is small compared to the frequency u. Then, we can considerthe approximation au _ _ u2 Re nu=u2 c bU0 Re J z ; J z uih =ce0 u2 , where the colpl 0 0 lision frequency is given in the form of a forceeforce correlation function, cf. Ref. [20]. Thus, the absorption coefcient is directly proportional to the real part of the forceeforce

na

130

68

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69

correlation function, which itself can be determined using perturbation theory. The relation to the thermodynamic Green function of two observables is given by i hA; Biuih pb

N Z

du 1 Im GAB u ih: u u ih u

A:6

component plasma [15]. It was used in Ref. [16] to compare the Thomson signal to the emission background caused by thermal bremsstrahlung photons. Generalizing for different electron and ion temperatures Te and Ti and inserting the explicit expression for the response function in RPA (30), one obtains Eq. (32).

Using Diracs identity one obtains RehA; Biuih Im GAB u ih=ub; which directly leads to Eq. (16). Appendix B. The RPA-susceptibility Eq. (27) is obtained from the exact relation for the response function ccc0 k; u, X sc ccc0 k; u c0 k; udcc0 c0 k; u Vcd k; ucdc0 k; u; c c

d

References

d

#

sc Vcd kc0 k; uVdc0 k; u d

c00 k; u c

c0 k; udcc0 c

X

d

cRPA ie

or

's

For a two-component plasma, i.e. an electroneion plasma c e, i, we obtain by matrix inversion c0 1 c0 Vii e i RPA cee ; B:3 1 c0 Vee c0 Vii c0 c0 Vii Vee Vei Vie i i e e B:4

cRPA and cRPA are obtained by interchanging indices i and e in ei ii Eq. (B.4) and Eq. (B.3), respectively. The imaginary part of the electronic susceptibility in RPA, cRPA k; u (Eq. (B.3)) for a hydrogen plasma (Vee Vii ee Vei Vie h V) now evaluates to 2 1 Vkc0 k; u 0 i Im cRPA k; u ee 1 Vkc0 k; u c0 k; u Im ce k; u i e 2 0 Vkce k; u 0 1 Vkc0 k; u c0 k; u Im ci k; u: i e

Au

th

This expression can also be derived from kinetic theory, i.e. the perturbative expansion of Vlasovs equation for a two-

pe

B:2 B:5

rs o

by truncation after the rst iteration, i.e. cdc0 c0 0 in the secdc ond term of Eq. (B.1). Insertion of the screened potential Eq. (28) yields the closed equation "

na

B:1

[1] S.C. Snyder, L.D. Reynolds, G.D. Lassahn, J.R. Fincke, C.B. Shaw Jr., R.J. Kearney, Phys. Rev. E 47 (1993) 1996. [2] S.C. Snyder, G.D. Lassahn, L.D. Reynolds, Phys. Rev. E 48 (1993) 4124. [3] S.C. Snyder, L.D. Reynolds, J.R. Fincke, G.D. Lassahn, J.D. Grandy, T.E. Repetti, Phys. Rev. E 50 (1994) 519. [4] R.E. Bentley, J. Phys. D 30 (1997) 2880. [5] C. Chenais-Popovics, V. Malka, J.-C. Gautier, S. Gary, O. Peyrusse, M. Rabec-Le Gloahec, I. Matsushima, C. Bauche-Arnoult, A. Bachelier, J. Bauche, Phys. Rev. E 65 (2002) 046418. [6] A.B. Murphy, Phys. Rev. E 69 (2004) 016408. [7] D.E. Evans, Plasma Phys. Controlled Fusion 12 (1970) 573. [8] S.H. Glenzer, G. Gregori, R.W. Lee, F.J. Rogers, S.W. Pollaine, O.L. Landen, Phys. Rev. Lett. 90 (2003) 175002. [9] G. Gregori, S.H. Glenzer, W. Rozmus, R.W. Lee, O.L. Landen, Phys. Rev. E 67 (2003) 026412. [10] G. Gregori, S.H. Glenzer, F.J. Rogers, S.M. Pollaine, O.L. Landen, C. Blancard, G. Faussurier, P. Renaudin, S. Kuhlbrodt, R. Redmer, Phys. Plasmas 11 (2004) 2754. [11] V. Ayvazyan, et al., Eur. Phys. J. D 37 (2006) 297. [12] A. Holl, R. Redmer, G. Ropke, H. Reinholz, Eur. Phys. J. D 29 (2004) 159. [13] R. Redmer, H. Reinholz, G. Ropke, R. Thiele, A. Holl, IEEE Trans. Plasma Sci. 33 (2005) 77. [14] H.A. Kramers, Philos. Mag. 46 (1923) 836. [15] D.E. Evans, J. Katzenstein, Rep. Prog. Phys. 32 (1969) 207. [16] H.A. Baldis, J. Dunn, M.E. Foord, W. Rozmus, Rev. Sci. Instrum. 73 (2002) 4223. [17] J.A. Gaunt, Proc. R. Soc. London, Ser. A 126 (1930) 654. [18] D. Zubarev, V. Morozov, G. Ropke, Statistical Mechanics of Nonequilib rium Processes, vol. 2, Akademie Verlag, Berlin, 1996. [19] A. Wierling, T. Millat, G. Ropke, R. Redmer, H. Reinholz, Phys. Plasmas 8 (2001) 3810. [20] H. Reinholz, R. Redmer, G. Ropke, A. Wierling, Phys. Rev. E 62 (2000) 5648. [21] J. Chihara, J. Phys. F 17 (1987) 295. [22] J. Chihara, J. Phys.: Condens. Matter 12 (2000) 231. [23] C. Itzykson, J.-B. Zuber, Quantum Field Theory, McGraw-Hill, New York, 1980. [24] S. Ichimaru, Statistical Plasma Physics, vol. 1, Addison-Wesley, New York, 1991. [25] H.R. Griem, Principles of Plasma Spectroscopy, Cambridge University Press, Cambridge, 1997. [26] R.D. Jones, K. Lee, Phys. Fluids 25 (1982) 2307. [27] G.D. Mahan, Many-Particle Physics, second ed. Plenum Press, New York and London, 1981. [28] A. Sommerfeld, Atombau und Spektrallinien, vol. 1, Vieweg, Braunschweig, 1949. [29] C. Fortmann, H. Reinholz, G. Ropke, A. Wierling, Condensed Matter Theory, vol. 28, submitted for publication. physics/0502051. [30] R.M. More, Atomic Physics in Inertial Connement Fusion Technical Report UCRL-84991-1, Lawrence Livermore National Laboratory, 1991. [31] S. Kuhlbrodt, B. Holst, R. Redmer, Contrib. Plasma Phys. 45 (2005) 73. [32] J.D. Jackson, Classical Electrodynamics, second ed. J. Wiley & Sons, New York, 1975.

co

py

131

C. Fortmann et al. / High Energy Density Physics 2 (2006) 57e69 [33] W. Heitler, The Quantum Theory of Radiation, Dover Publications, New York, 1984. [34] M. Ter-Mikaelyan, Dokl. Akad. Nauk SSSR 94 (1953) 1033. [35] M. Abramowitz, A. Stegun (Eds.), Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables, ninth ed. Dover Publications, New York, 1970. [36] W.J. Karzas, R. Latter, Astrophys. J. Suppl. Ser. 6 (1961) 167. [37] C.W. Allen, Astrophysical Quantities, The Athlone Press, London, 1973. [38] H.A. Gould, H.E. DeWitt, Phys. Rev. 155 (1967) 68. 69

[39] R. Kubo, J. Phys. Soc. Jpn 12 (1950) 570. [40] A. Selchow, G. Ropke, A. Wierling, H. Reinholz, T. Pschiwul, G. Zwicknagel, Phys. Rev. E 64 (2001) 056410. [41] E.E. Salpeter, Phys. Rev. 120 (1960) 1528. [42] J. Chihara, private communication. [43] J.D. Huba, NRL Plasma formulary, Naval Research Laboratory, Washington, DC., 1994. [44] R. Thiele, H. Reinholz, R. Redmer, G. Ropke, J. Phys. A 39, (2006) 4365.

Au

th

or

's

pe

rs o

na

co

py

132

Authors: Carsten Fortmann, August Wierling, and Gerd Rpke o Appeared as regular article in Contributions to Plasma Physics, Vol. 47, Issues 4-5, pages 297-308, June 2007. Listing of contributions by authors: C.F.: Preparation of manuscript (mainly sections 3-6), all numerical calculations A.W.: Preparation of manuscript (mainly sections 1-2,7) G.R.: Preparation of manuscript (section 1)

134

135

Contrib. Plasma Phys. 47, No. 4-5, 297 308 (2007) / DOI 10.1002/ctpp.200710040

C. Fortmann , G. R pke, and A. Wierling o

Institut f r Physik, Universit t Rostock, 18051 Rostock, Germany u a Received 19 December 2006, accepted 19 December 2006 Published online 6 June 2007 Key words Non-Ideal plasmas, bremsstrahlung, spectral function, vertex corrections. PACS 52.25.Mq,52.25.Os,52.27.Gr A basic concept to calculate physical features of non-ideal plasmas, such as optical properties, is the spectral function which is linked to the self-energy. We calculate the spectral function for a non-relativistic hydrogen plasma in GW -approximation. In order to go beyond GW approximation, we include self-energy and vertex correction to the polarization function in lowest order. Partial compensation is observed. The relation of our approach to GW and GW calculations in other elds, such as the band-structure calculations in semiconductor physics, is discussed. From the spectral function we derive the absorption coefcient due to inverse bremsstrahlung via the polarization function. As a result, a signicant reduction of the absorption as compared to the Bethe-Heitler formula for bremsstrahlung is obtained.

c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction

Spectroscopy can serve as a versatile tool to infer properties of dense and hot plasmas from the emitted radiation [1,2]. In non-ideal plasmas, where the coupling parameter can exceed unity, properties often deviate signicantly from their form in ideal plasmas due to the importance of interaction effects [3]. With the advent of femtosecond laser pulses [4], it has become possible to produce non-ideal plasmas with table-top systems. In particular, it is nowadays possible to create conditions similar to those in the center of astrophysical objects such as the sun or giant planets. Also, the FAIR-project at GSI Darmstadt combined with the high energy laser systems PHELIX and NHELIX available at GSI will provide ideal tools to produce and diagnose plasmas where the effects under discussion here are important [5]. Effects such as dynamical screening, dissolution of bound states, Pauli-blocking, and the importance of collisions have been observed [3]. Taking proper account of interaction effects is a challenge to any theoretical description of non-ideal plasmas. Many-body perturbation theory presents a toolbox to determine various properties of non-ideal plasmas [6 8]. The use of Greens function techniques allows for a systematic and intuitive consideration of many-body effects. In particular, one-particle properties can be obtained from the one-particle Greens function or its spectral representation, the one-particle spectral function. A number of important mechanisms in dense plasmas such as dynamical screening of the Coulomb interaction can be described by partial summation of certain diagrams. A particularly successful concept is the quasi-particle picture [68]. However, with increasing coupling the quasi-particle pictures breaks down. This shows up in a broadening of the spectral function. To go beyond the quasi-particle picture in a consistent way poses serious problems of self-consistency. Vertex and self-energy corrections have to be taken into account on the same footing to obey sum rules and other exact known properties. Using Ward identities [9] or the Kadanoff-Baym scheme of the conserving vertex [10] enables one to construct consistent sets of diagrams, alas the resulting integral-equations are complicated to solve. The GW approximation is a particularly scheme for the self-energy approximating it by a product of the Greens function G and an effective interaction W [11]. Being introduced by Hedin in 1965 [12], it has a long history of applications, which is reviewed in Ref. [13] and [14]. However, the full self-consistency implied in

Corresponding

author: e-mail: carsten.fortmann@uni-rostock.de, Phone: +49 381 498 6947, Fax: +49 381 498 6942

136

298

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o

Hedins original proposal was not carried out so far. Instead, the ideal Greens function or a quasi-particle picture was used. In this manner, the GW approximation was successfully applied to determine shifts in the ionization potential, approximations for the equation of state, and an effective interaction to describe bound states in a medium [8]. Also, the band structure of different types of materials, like semiconductors [15], alkali [16], and transition metals [17] was determined. Due to the efciency of modern computer technology, it has recently become feasible to address the self-consistency implied in Hedins scheme to some extent, see e.g. [1821]. Some of these calculations improve solely on self-energy corrections, while others take into account vertex contributions as well. It is customary to call the later GW approximations. In this paper, we perform calculations for the optical properties of a hydrogen plasma at solar core conditions, as an important example for astrophysical plasmas. Typical parameters of the solar core plasma are temperatures of about T = 100 Ry/kB 1360 eV/kB and particle densities reaching n 7 1024 cm3 . These parameters justify the model of a classical (non-degenerate) and weakly coupled plasma, being characterized by values of = kB T /EF 10 for the degeneracy parameter and = (Ze2 /4 0 )/(3/4n)1/3 kB T 0.03 for the coupling parameter. Central to the description of optical properties of plasmas is the dielectric function (k, ) [1]. Within the frame of the approach presented here, it can be obtained from the polarization function, which is a member of Hedins set of equations. In an earlier paper [22], the suppression of the bremsstrahlung cross section due to successive scattering has been treated by a taking into account self-energy and vertex corrections. However, these corrections were only considered within a one-loop approximation. In this way, dynamical screening effects were neglected. It is the objective of this communication to present results with an improved one-particle spectral function as obtained with the GW approximation. In Sec. 2, we will review the GW approximation making use of Hedins equations. Sec. 4 presents an illustrative example for the GW 0 approximation. Implications for the absorption coefcient are discussed in Sec. 6. A discussion and conclusions will be given in Sec. 7. If not otherwise indicated, we apply the Rydberg system of units where = kB = 1 , e2 = 2 , 0 = 1/4 , and me = 1/2.

2 The GW approximation

A convenient starting point of our approach are Hedins equations [23]. It is a closed set of equations relating the full Greens function G, the non-interacting Greens function G0 , the self-energy , the dynamically screened interaction Wab [11], the polarization function , and the vertex function . In detail, the full Greens function is given by Ga (12) = Ga,0 (12) + d(34) Ga,0 (13) a (34) Ga (42) . (1)

Here and in the following, the shorthand notation (1) (r1 , 1 , 1 . . .) for spatial variables r, imaginary times as well as quantum numbers such as spin is used. The self-energy a (12) is obtained from the dynamically screened interaction and the vertex function according to a (12) = i d(34) Ga (13) Waa (41) a (32, 4) . (2)

The dynamical screened interaction is given via Wab (12) = by the polarization function aa (12) = d(34)Ga (13)Ga (41)a (34, 2) . (4) Vab (12) +

c

(3)

www.cpp-journal.org

(12) (13) +

b

d(4567)

(5)

137

Contrib. Plasma Phys. 47, No. 4-5 (2007) 299

The set of equations represents a perturbation expansion in terms of the screened interaction Wab and is expected to show better convergence properties compared to an expansion in the inter-particle interaction Vab . By truncating the set of equations (1)-(5) on a certain level, various approximations for these quantities can be dened. A particular simple approximation is the G0 W 0 approximation [12, 23]. It is obtained by taking the bare vertex (12, 3) = (12)(13) and inserting non-interacting Greens functions into the expression for the polarization function. This leads to a self-energy given by GW (12) = iGa (12)Waa (21). Next, the Greens a functions in this expression as well as in the screened interaction W are taken as the free Greens function G0 . This expression turns out to be quite successful. However, in dense plasmas as well as in a number of materials in solid state physics, there is need to improve beyond this simple approximation. Such an improvement is the GW 0 approximation which was studied in Ref. [24] for the electron gas at T = 0 and in Ref. [25] for the solar core plasma. In the latter case, a considerably broadened quasiparticle and a featureless behaviour at the plasma frequency was found. No plasmon-like satellite structures survived the partial self-consistency treatment. Another straightforward extension is a self-consistent solution of Dysons equation and the screened interaction while keeping the vertex function in lowest order [19]. This will be termed GW approximation in this paper. In Ref. [19], an increase in the quasiparticle bandwidth and a featureless satellite structure is found in contradiction to experimental evidence. Also, such an approximation leads to a drastic violation of the f-sum for the inverse dielectric function. The results indicate the importance of vertex corrections [26]. Calculations for a one-dimensional semiconductor [27] lead to similar conclusions. Large cancellations between self-energy and vertex corrections have also been found in Ref. [2832]. Recently, a number of approximation schemes to take into account vertex corrections have been proposed [33]. In this work we will apply a sequence of approximations described in the following: Taking the self-energy in the GW -approximation, the functional derivative occurring in the vertex equation (5) yields in lowest order of Waa a (12) Gb (45)

0 = Waa (12)(14)(25)ab .

(6)

Note that the screened interaction is taken in the one-loop approximation for , the so-called random-phase approximation (RPA). This term leads to a ladder approximation for the vertex in terms of the dynamically screened interaction a (12, 3) = (12)(13) +

0 d(67)Waa (12)Ga (16)Ga (72)a (67, 3) .

(7)

Already this equation is challenging to solve [34], even in Shindo approximation [35]. As a result, the improved self-energy in this approximation is given in second order of the dynamically screened interaction by the term studied in Ref. [18], GW a

0

(12) =

Ga (12)W 0 (21) +

(8)

Also, the improved polarization function is given besides the loop diagram by an exchange diagram with respect to W 0 GG (12) = aa Ga (12)Ga (21) +

0 d(34)Ga (13)Ga (41)Waa (43)Ga (24)Ga (32) .

(9)

At this stage the rst iteration of Hedins equations (1)-(5) is completed. We will not go beyond this rst iteration, in particular the screened interaction potential Waa is kept on the level of RPA. To address the deviations from the quasi-particle picture, the introduction of the one-particle spectral function Aa (p, ) is convenient. With its help, the spectral representation of the full Greens function is given by Ga (p, z) = d Aa (p, ) , z 2

(10)

p and z denote momentum and energy (frequency) of the particle. The Greens function Ga (12) is obtained from the Greens function in momentum-frequency representation Ga (p, z) by means of Laplace transform. The

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

138

300

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o

quasi-particle approximation itself can be stated as a -like form of the spectral function Aa (p, ) = 2 ( Ea (p)) , (11)

where the energy Ea (p) is obtained as Ea (p) = p2 + Re a (p, Ea (p)) . 2ma (12)

Note, that the spectral function itself obeys a normalization relation d Aa (p, ) = 2

1 .

(13)

Furthermore, sum rules for the rst and second moment of the spectral function are known [24], d Aa (p, ) =

HF Ea ,

(14)

2

d 2 Aa (p, ) =

(15)

where the index HF refers to the energy in Hartree-Fock approximation and the index c indicates the use of the correlated self-energy. From the spectral function, a number of thermodynamic quantities can be obtained, e.g. the one-particle density [8, 36], na (a , ) = d3 p (2)3 d fa () Aa (p, ) , 2

(16)

with the distribution function fa () of particles of species a. Optical properties, which are under consideration in this work, can be obtained from the well-known relation between the dielectric function (q, ) and the polarization function Eq. (4), i.e. (q, ) = 1

a

(17)

with the interaction potential Vaa (q). In particular, the absorption coefcient, which gives the attenuation of electromagnetic radiation traversing the plasma is given by the imaginary part of the long wavelength limit of the dielectric function as (18) () = Im (q 0, ), c a relation that holds for wavelengths long against interatomic distances and frequencies high compared to the plasma frequency pl = 4(n)1/2 . For details, we refer to Refs. [1, 25].

The GW approximation has been very successfully used for a broad variety of problems in many-particle physics for a long time. Here we would like to mention only a few examples. One of the rst applications in solid state physics was the calculation of band structures combining density functional theory (DFT) and the GW -method. Northrup et al. [16] showed how this approach improves the bandgap problem of the local density approximation (LDA), which predicts too small band-gaps for most materials, whereas the use of GW helps to decrease the deviation of the theoretical from the experimental value below 0.1 eV in the case of silicon, compared to about 1 eV as obtained from pure LDA-DFT. Similar results were obtained for many different semi-conductors and insulators [14, 38]. Optical properties of highly excited semiconductors using a combined GW and T-Matrix approach have been studied extensively by Schmielau et al., see Refs [3941]. Also in the case of nite systems, the GW approximation has been applied successfully. For the case of the Na4 tetramer, good agreement between the theoretically predicted photoabsorption and experimental data was obtained [42].

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

139

Contrib. Plasma Phys. 47, No. 4-5 (2007) 301

Nuclear and quark matter have been investigated using techniques very similar to the approach presented here. The problem of dynamical chiral symmetry breaking, which cannot be described in a perturbative calculation of quark propagators, was demonstrated to emerge in a nonperturbative approach known in quantum eld theory as rainbow-ladder approximation [43, 44]. Fehr [45] and Wierling et al. [25] investigated spectral properties of electrons in nonideal plasmas making use of the GW approximation. Whereas Fehr applied the perturbative G0 W 0 approach (c.f. section 2) to both equation of state and optical properties, Wierling already used a self-consistent GW 0 approximation which is also used here in the following section.

In the GW 0 approximation, the set of Hedins equation reduces to a self-consistent solution of Dysons equation with the correlated part of the self-energy, Ae (p, + i) = e (p, + i) = 2Im e (p, + i)

2

(19) Ae (p q, ) . + i (20)

d d (2)2

d3 q 8 Im (2)3 q 2

1 RPA (q,

+ i0) [1 + nB ( )]

As indicated by W 0 , the dielectric function (q, ) is taken in random phase approximation (RPA) and the Bose 1 function is given as nB () = (exp (/T ) 1) . Here, as in the following, electrons labeled by a = e are considered. Note that the Hartree-Fock self-energy, which appears as an additional term in Eq. (20) is small for the considered parameters and is henceforth neglected. We give results for a hydrogen plasma using the conditions at the solar core center. We ignore other ions such as helium etc. in these exploratory calculations. In Fig. 1, the self-consistent spectral function calculated in GW 0 approximation, i.e. the solution of the set of Eqs. (19) and (20), is shown for a xed momentum of p = 0.21a1 . B The grey curve is the initial ansatz for the spectral function, i.e. the input for the r.h.s. of Eq. (20). It has been chosen of Lorentzian form with a width of = 10 Ry. The rst iteration is given by the dashed-dotted curve. The peak of the spectral function is shifted to smaller frequencies and the function is asymmetrically broadened. The second and third iteration give only minor modications to the rst iteration, the forth iteration (not shown) does not vary signicantly from the third iteration.

0.4 initial (Lorentzian, =10 Ry) QPA 1. iteration 2. iteration 3. iteration

0.2

0.1

0 -50

0 [Ry]

50

Fig. 1 Electron spectral function in GW 0 for solar core conditions (ne = 7 1024 cm3 , Te = 100 Ry 1360 eV) at momentum p = 0.21a1 . Dashed-dotted curve: rst iterB ation of GW 0 , (quasiparticle approximation [QPA]). Grey solid curve: Ansatz for the spectral function in the iterative solution of GW 0 . Dotted curve: 1st iteration, dashed: 2nd iteration, solid: 3rd iteration. The 4th iteration is not distinguishable from the 3rd iteration.

For comparison, we also show the G0 W 0 approximation, i.e. the rst iteration of Eq. (20) starting from a spectral function of vanishing width (delta-distribution). In this case, there is no quasiparticle peak at the quasiparticle resonance QP = p2 , but four satellites and minima appear. The latter are due to sharp peaks in the response function Im 1 (k, ) and have been observed earlier [46, 47]. RPA As a result of this calculation, we observe, that the self-consistent calculation leads to a spectral function that is physical easily understandable, i.e. it contains a broadened and shifted quasiparticle resonance. However, the

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

140

302

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o

signatures of collective effects, such as the dynamical screening, which are present in the G0 W 0 approximation (dash-dotted curve in Fig. 1) vanish completely in the self-consistent result. In contrast, the G0 W 0 result does not contain a broadened quasiparticle peak but is completely determined by the behaviour of the response function, which contains the collective excitations of the plasma (plasmons), as also shown by Fehr et al., see Ref. [48].

0 Im -2 -4 -6 4 2 0 -2 -4 0 0 -10 -20 -30 20 10 0 -10 -20 40 20 0 -20 -40 3 2 1 -40 -20 0 [Ry] 20 40 0 -40 -20 0 [Ry] 20 40 0

Re

-p -Re

Fig. 2 Self-energy (real and imaginary part), dispersion relation and spectral function calculated self-consistently by solving Eq. (21) (left) and quasiparticle approximation (right), i.e. the r.h.s of Eq. (21) evaluated with (p, ) = 0. Parameters: ne = 7 1024 cm3 , Te = 100 Ry 1360 eV. The momentum is xed at p = 0.21 a1 . B

Ae [1/Ry]

Alg. solution for Born approx. (Lorentz plasma) full RPA solution (TCP)

0.3

0.2

0.1

-40

-20

0 [Ry]

20

40

Fig. 3 Self-consistent spectral function computed in full RPA (two component plasma) (dashed curve) and in static Born approximation including only electron-ion collisions (solid curve). The neglectance of e e collisions leads to a smaller width of the spectral function. Solar core parameters: ne = 7 1024 cm3 , Te = 100 Ry.

Although convergence is already achieved after 3-4 iterations of Eqs. (19) and (20), the calculation is too time-consuming to be used in the computation of physical observables such as equation of state or the polarization function, the latter involving convolution integrals over two spectral functions in both momentum and frequency domain. Therefore, we make use of some further approximations. First, we replace the full RPA-like screened interaction by a one-loop approximation, which takes into account the scattering among particles in Born approximation. In this case we only consider electron-ion scatterings. Ions are treated in adiabatic approximation and the interaction is mediated by a statically screened potential of Debye-type, i.e. Vei (q) = 8/(q 2 + 2 ), where = (8n/T )1/2 is the inverse Debye screening length. This approximation is justied by the fact that the corresponding absorption cross-section leads to the nonrelativistic limit of the Bethe-Heitle formula for inverse bremsstrahlung in the limit of vanishing , as shown in [22].

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

141

Contrib. Plasma Phys. 47, No. 4-5 (2007) 303

For small , the main contributions to the integral in Eq. (20) come from small q. Therefore, we can neglect the shift in the momentum variable in the self-energy on the r.h.s. As shown in Ref. [22], one then obtains a particularly simple equation for the self-energy in this approximation, which we will denote by the sufx GW 0 s, GW s (p, +i) = 8

0 0 ni 2 + p2 i + GW s (p, + i) 2i + i GW0 s (p, + i)

(21) which can be solved by standard root-nding algorithms. Fig. 2 shows the solution of Eq. (21) for the same set of parameters as used in the RPA calculation (Fig. 1). To the left, the self-energy (imaginary and real part), as well as the dispersion relation, and the spectral function as obtained from the numerical solution of Eq. (21) are shown. The spectral function is centered around the quasiparticle energy. Its overall shape is given by the imaginary part of the self-energy, while the real part determines the dispersion. In the self-consistent solution, the dispersion relation contains only a single root near the quasiparticle energy, while there are three roots in the quasiparticle approximation (r.h.s of Fig. 2). The upper and lower root give rise to the two satellites in the corresponding spectral function, while the central root does not yield a resonance due to the large imaginary part 0 of G0 W s around = 1 Ry. The separation of the satellites from the quasiparticle energy is approximately given by the plasma frequency. This is a general feature, which has been observed in earlier calculations carried out at lower particle densities and temperatures [22].

[Ry]

120 A e (p, )[ Ry] 0.5 0.4 60 0.3 40 0.2 20 0.1 0

100

80

20 0 2 4 6 8 10

p [ a B -1 ]

Fig. 4 Contour plot of the self-consistent spectral function with simplied self-energy (Eq. (21) at solar core conditions as function of frequency and momentum p. The spectral function is asymetrically broadened and shifted from the quasi-particle energy QP = p2 . At high momentum the spectral function converges into a sharp quasi-particle resonance at the free-particle energy. Parameters: ne = 7 1024 cm3 , Te = 100 Ry 1360 eV.

In Fig. 3 we compare the spectral functions obtained from the two-component RPA and the simplied Born approximation containing only the scattering of electrons on xed ions (Lorentz-Plasma). Since electron-electron collisions are neglected in the latter case, the corresponding spectral function is narrower than in the RPA calculation. Furthermore, in the one-loop approximation there are no plasmon degrees of freedom which lead to further damping in the RPA calculation. On the other hand, the Lorentz plasma calculation already contains all characteristic features of the RPA spectral function. Namely the asymmetrical broadening and the shift of the peak towards lower energies show up. These features become less pronounced at higher momenta, where many-particle effects are expected to be less important. With increasing momentum, the width of the spectral function decreases, while at the same time the height of the quasiparticle peak increases, as can be seen in Fig. 4. This follows naturally from the normalization condition for the spectral function Eq. (13). In conclusion, we have shown that our simplied Lorentz plasma model for the self-energy leads to physically intuitive one-particle spectral function which includes the electron-ion collisions in a consistent way.

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

142

304

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o

In the last years, substantial effort has been made to go beyond the GW -approximation and to include the vertex to some extent. In a pioneering paper, Takada [21] demonstrated the feasibility of self-consistent GW approximation. He used an ansatz for the vertex-function which fulls certain sum-rules and conservation laws. As a result, he showed the subtle cancellation of contributions from self-energy and vertex-corrections to spectral properties of charged particles, exemplied for the damping of plasmons in Al. Very recently, Ziesche [49] has reviewed the calculation of direct and exchange contributions (vertex correction) to the on-shell self-energy of the homogenous electron-gas. For plasmas, Vorberger et al. [36] systematically studied contributions to the equation of state beyond MontrollWard, including all exchange terms (vertex corrections). In the same spirit, we will now investigate the lowest order vertex correction to the self-energy in second order of the interaction, given by the following diagrammatic expression.

(2) (p, ) =

(22)

The second diagram is the lowest order diagram which does not appear in the GW approximation, whereas the rst diagram is automatically included. The base lines are free electron propagators, while the upper loops are composed of ionic propagators taken in the adiabatic approximation. To ensure convergence, a statically screened potential Vei (q) is used. The numerical evaluation of Eq. (22) is shown in Fig. 5. The dashed and dotted curves give the rst and second iteration respectively of the GW 0 s approximation, the solid curve is the self-consistent result. For comparison, the rst order vertex correction as obtained from the exchange diagram in Eq. (22) is shown in the inset. The vertex term gives at most a 20% reduction of the self-energy.

0

GW , 1. iteration 0 GW , 2. iteration 0 GW , stable

0

Im (0.1,) [Ry]

-0.05

0.01 0.005

-0.1

Vertex correction

0 -0.005

-0.15 0

-0.01 0.9

0.5

[Ry]

1.5

Fig. 5 Imaginary part of the self-energy as calculated from Eq. (21). The rst (dashed) and second (dotted) iteration as well as the convergent result (solid) are shown. Inset: Comparison of GW approximation and lowest order vertex correction (dashed-dotted), cf. Eq. (22). The exchange diagram gives a correction of at most 20%. Parameters: ne = 1021 cm3 , Te = 1 Ry.

As an illustrative example, we calculate the impact of the broadened electron spectral function on the absorption coefcient due to inverse bremsstrahlung. The absorption coefcient for radiation in a plasma can be obtained from the dielectric function (q, ), Eq. (18). According to Eq. (4), nonideality effects enter via self-energy corrections of the full Greens functions and by vertex corrections beyond the bare vertex. Within the GW approximation, only self-energy corrections are considered, while the GW also accounts for vertex terms in the polarization function.

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

143

Contrib. Plasma Phys. 47, No. 4-5 (2007) 305

6.1 Self-energy corrections Using the self-consistent spectral function obtained in simplied GW 0 (Eq. (21)) to calculate the polarization function, leads us to the result plotted as solid curve in Fig. 6. We normalize our result to Kramers formula corrected by a Gaunt-factor in Born approximation [50, 51], which corresponds to the non-relativistic limit of the Bethe-Heitler cross section for inverse Bremsstrahlung [52]. In the infrared limit ( 0) the Born approximation shows a logarithmic divergence.

1.1

1.05

1 /

B

0.95

Gaussian SF, = 6.5 Ry Gaussian SF, =13.0 Ry Self-consistent SF

0.9

0.85 1

10 [Ry]

100

1000

Fig. 6 Free-free absorption coefcient calculated from broadened electron propagators. The absorption coefcient is normalized to the Born approximation. At low frequencies, suppression of inverse bremsstrahlung is obtained. For high photon energies, the improved result converges to the Born approximation, while around = 10 Ry enhancement sets in. Also shown: () calculated with Gaussian spectral functions (dotted and dashed line) of different widths. Here, the convergence into the Born result is much slower, since the Gaussian spectral function does not yield the correct quasiparticle limit at high frequencies. Parameters: ne = 7 1024 cm3 , Te = 100 Ry (solar core).

The result for solar core parameters is shown in Fig. 6. For small frequencies, a reduction of the free-free absorption of about 15% as compared to Born approximation is observed, while at high frequencies our approach converges to the Born approximation. At intermediate frequencies, we obtain a slight enhancement of our result relative to the Born approximation. These characteristics have already been observed in earlier calculations for lower densities and temperatures [22]. We compare our results to calculations which use parametrized spectral functions of Gaussian shape in the polarization loop (dashed and dotted curve). The width is given. Also in this case, an enhancement is observed which decreases with decreasing width as can be seen by comparing the dotted ( = 6.5 Ry) and the dashed curve ( = 13 Ry). The value = 6.5 Ry gives a Gaussian spectral function of similar shape as the self-consistent calculation at small momenta. As shown in Fig. 4, the self-consistently calculated spectral function converges to a quasi-particle resonance at large momenta, i.e. large frequencies. This behaviour leads to the faster convergence of the absorption coefcient to the Born result as compared to the calculation using Gaussians with frequency and momentum independent widths. Remember that the Born result for the absorption coefcient is obtained by inserting delta-like spectral functions in the polarization function Eq. (4). 6.2 Vertex corrections We have shown that the account of a broadened one-particle spectral function leads to a suppression of the infrared behaviour of the inverse bremsstrahlung spectrum. However, the calculations have been carried out on the level of GW 0 -approximation. Vertex corrections contribute to the polarization function in the same order with respect to the statically screened potential. Moreover, these vertex corrections tend to cancel the self-energy correction to some extent. In order to account for both types of corrections on the same footing, we focus on contributions from the vertex equation, Eq. (5), up to rst order in the screened interaction in the polarization, Eq. (4),

(q, ) =

(23)

In Fig. 7 we compare the suppression of the absorption coefcient obtained from the inclusion of self-energy only (solid curve) and with included vertex correction in the polarization loop (dashed curve), c.f. Eq. (23). At the present parameters, the vertex correction is the most dominant term. Over the whole range of frequencies considered here, we obtain a suppression of the absorption coefcient as compared to the Born approximation.

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

144

306

1.1

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o

1.05

1 /

B

0.95

V +

0.9

0.85 1

10 [Ry]

100

1000

Fig. 7 Absorption coefcient () for solar core conditions (ne = 7 1024 cm3 , T = 100 Ry) as function of the photon energy . () is calculated from Eq. (4) with broadened electron propagators using the self-energy obtained from Eq. (21) (solid curve). Additionally, the rst vertex correction is calculated, cf. Eq. (22). The sum of both terms is given by the dashed curve. The vertex correction dominates the behaviour of the absorption coefcient. In particular, the enhancement at 10 Ry present in the pure self-energy calculation vanishes completely.

Also, the enhancement of absorption at frequencies around = 10 Ry, which was observed in the calculation using only self-energy corrections (solid curve), vanishes if the vertex is taken into account (dashed curve). This shows the importance of vertex corrections. Here, the vertex correction is only taken in lowest order of the interaction and density, while the self-energy based result contains a summation to all orders of density. In order to compare self-energy and vertex contributions to the modication of the absorption spectrum in a fully consistent way, one would have to go beyond the perturbative calculation presented here and solve for the vertex equation Eq. (5) using at least free-particle propagators. This task goes beyond the scope of this paper, where only the lowest order corrections are to be studied. Finally, we remark, that in the mentioned earlier calculations presented in Ref. [22], the vertex correction modies the self-energy result only to a minor extent, i.e. the enhancement at intermediate frequencies is reduced by 40%. This is due to the lower density (1019 cm3 ) used in that work. In all calculations, the high frequency behaviour converges nicely to the Born result. For the infrared part of the spectrum, the behaviour of the absorption coefcient needs further considerations, going beyond the present work. As shown in Fig. 7, the self-energy correction in the polarization function reduces the absorption coefcient by a constant factor of about 15% in the low frequency limit, but does not regularize the infrared divergence of the Born approximation. The vertex term, on the other hand, induces a large suppression so that in the low frequency limit higher orders of the vertex correction have to be considered. Furthermore, it is well-known, that below the plasma frequency (7 Ry for the solar core parameters) dynamic screening plays an important r le. This needs a further summation of diagrams going beyond the present scheme. The impact of o dynamical screening on bremsstrahlung emission has been discussed by Ter-Mikaelyan in Ref. [53].

7 Conclusions

A systematic treatment of optical properties in non-ideal plasmas is possible in the framework of Greens function methods. Corrections beyond the quasi-particle pictures can be generated using Hedins equations. However, a consistent solution of Hedins equations is a formidable task. Here, we considered a simplied set of equations. Notably the GW0 -approximation, mainly used in solid state physics, has been shown to lead to sensible results also in the eld of plasma physics. Results for the self-energy and the spectral function are presented. Plasmon-like structures, present in perturbative calculations as performed in Ref. [48] vanish completely. An asymmetrically broadened and shifted spectral function is obtained. Furthermore, it is shown, that a simplied model, where the dynamically screened interaction is approximated by static screening leads to similar results for both self-energy and spectral function as compared to the full RPA results. Deviations can be understood as a consequence of neglecting collisions among particles of equal species. The simplied GW 0 approximation for plasmas is now available for a broad range of parameters, i.e. density and temperature. Also, the correct asymptotic behaviour, i.e. convergence to a delta-like quasiparticle resonance at large momenta, is obtained. In this paper we used parameters corresponding to the solar core, whereas Ref. [22] contains similar results for lower density and temperature.

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

145

Contrib. Plasma Phys. 47, No. 4-5 (2007) 307

Vertex corrections to the self-energy have been studied on a perturbative level. So far, only the high frequency behaviour of the exchange diagram second order in the screened interaction was evaluated. It leads to 20% reduction of the self-energy with respect to the second order of GW 0 . The availability of the one-particle spectral function for any momentum and frequency makes it possible to use it in calculations of further physical observables as the equation of state and the optical properties. Here we focused on the absorption of electromagnetic radiation due to free-free transitions, which is obtained from the polarization function. Insertion of the broadened particle propagators in the one-loop approximation for leads to a signicant modication of the absorption spectrum at low frequencies. Besides the suppression in the infrared, enhancement of absorption is observed at intermediate frequencies. Since the one-loop approximation using full propagators is an inconsistent approximation with regards to conservation laws, such as Ward identities, we investigated the vertex-correction inside the polarization loop in lowest order. It turned out, that the vertex correction is by far the most important correction to Born approximation, i.e. the self-energy effects contained in the full spectral functions, are dominated by the vertex correction. However, this is not a general feature as becomes clear if comparing to calculations for other parameters in Ref. [22]. On the other hand, a consistent comparison of both self-energy and vertex corrections on the same level of approximation necessitates the summation of vertex-corrections by solution of the Bethe-Salpeter equation (5). This task will be accomplished in the future.

Acknowledgements This article was supported by the DFG within the Sonderforschungsbereich 652 Starke Korrelationen und kollektive Ph nomene im Strahlungsfeld: Coulombsysteme, Cluster und Partikel. A.W. would like to thank the Center a of Atomic and Molecular Technologies of Osaka University for its hospitality. C.F. acknowledges stimulating discussion with J. Vorberger and W.D. Kraeft and thanks R. Zimmermann for many helpful advice.

References

[1] H.A. Griem, Principles of Plasma Spectroscopy (Cambridge University Press, Cambridge, 1997). [2] I.H. Hutchinson, Principles of Plasma Diagnostics (Cambridge University Press, Cambridge, 1987). [3] see the proceedings of earlier workshops on the Physics of nonideal Plasmas, e.g. Contrib. Plasma Phys. 39, 5-184 (1999); 41, 119-302 (2001); 43, 241-397 (2003). [4] M.D. Perry and G. Mourou, Science 264, 917 (1994); C.J. Joshi and P.B, Corkum, Physics Today 108, 36 (1995). [5] D.H.H. Hoffmann, A. Blazevic, P. Ni, O. Rosmej, M. Roth, N.A. Tahir, A. Tauschwitz, S. Udrea, D. Varentsov, K. Weyrich, and Y. Maron, Laser and Particle Beams, 23, 47 (2005). [6] A.L. Fetter and J.D. Walecka, Quantum Theory of Many Particle Systems (McGraw-Hill, New York, 1971). [7] G.D. Mahan, Many-Body Physics (Plenum, New York, 1990). [8] W.D. Kraeft, D. Kremp, W. Ebeling, and G. R pke, Quantum Statistics of Charged Particle Systems (Plenum, New o York, 1986). [9] J. Ward, Phys. Rev. 78, 182 (1950); Y. Takahashi, Nuovo Cimento 6, 370 (1957); T. Toyoda, Ann. Phys. (N.Y.) 173, 226 (1987). [10] G. Baym and L.P. Kadanoff, Phys. Rev. 124, 287 (1961). [11] In plasma physics, the use of V s instead of W is more common. [12] L. Hedin, Phys. Rev. 139, A796 (1965). [13] G.D. Mahan, Comments Condens. Mater. Phys. 16, 333 (1994). [14] F. Aryasetiawan and O. Gunnarsson, Rep. Prog. Phys. 61, 237 (1998); G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. 74, 601 (2002). [15] R.W. Godby, M. Schluter, and L.J.Sham, Phys. Rev. B 37, 10159 (1988). [16] J.E. Northrup, M.S. Hybertson, and S.G. Louie, Phys. Rev. Lett. 59, 819 (1987). [17] F. Aryasetiawan, Phys. Rev. B 46, 13051 (1992). [18] E.L. Shirley, Phys. Rev. B 54, 7758 (1996). [19] B. Holm and U. von Barth, Phys. Rev. B 57, 2108 (1998). [20] A. Schindlmayr and R.W. Godby, Phys. Rev. Lett. 80, 1702 (1998). [21] Y. Takada, Phys. Rev. Lett. 87, 224602 (2001). [22] C. Fortmann, H. Reinholz, A. Wierling, and G. R pke, Proceedings of the 28th International Workshop on Condensed o Matter Theories, in print, arXiv: physics/0502051. [23] L. Hedin and S. Lundqvist, in Solid State Physics, edited by H. Ehrenreich, F. Seitz, and D. Turnbull, Vol. 23, p.1 (Academic Press, New York, 1969). [24] U. von Barth and B. Holm, Phys. Rev. B 54, 8411 (1996). [25] A. Wierling and G. R pke, Contrib. Plasma Phys. 38, 513 (1998). o

www.cpp-journal.org c 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

146

308 [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53]

C. Fortmann, G. R pke, and A. Wierling: Optical properties and one-particle spectral function o M. Hindgren and C.-O. Almbladh, Phys. Rev. B 56, 12832 (1997). H.J. de Groot, P.A. Bobbert, and W. van Haeringen, Phys. Rev. B 52, 11000 (1995). T.M. Rice, Ann. Phys. (N.Y.) 31, 100 (1965). D.J.W. Geldart and R. Taylor, Can. J. Phys. 48, 155; 48, 167 (1970). G.D. Mahan and B.E. Sernelius, Phys. Rev. Lett. 62, 2718 (1989). F. Bechstedt, K. Tenelsen, B. Adolph, and R. Del Sole, Phys. Rev. Lett. 78, 1528 (1997). R.T.M. Ummels, P.A. Bobbert, and W. von Haeringen, Phys. Rev. B 57, 11962 (1998). B. Tanatar and E. Demirel, Phys. Rev. B 62, 1787 (2000). I.A. Nechaev and E.V. Chulkov, Phys. Rev. B 71, 115104 (2005). K. Shindo, J. Phys. Soc. Japan 29, 287 (1970). J. Vorberger, M. Schlanges, and W.D. Kraeft, Phys. Rev. E 69, 046407 (2004). A. Wierling, Th. Millat, G. R pke, R. Redmer, and H. Reinholz, Phys. Plasmas 8, 3810 (2001). o Faleev et al., Phys. Rev. B 74, 033101 (2006). T. Schmielau, G. Manzke, D. Tamme, and K. Henneberger, Phys. Stat. Sol. (b) 221, 215 (2000). R. Schepe, T. Schmielau, D. Tamme, and K. Henneberger, Phys. Stat. Sol. (b) 206, 273 (1998). G. Manzke, T. Schmielau, and K. Henneberger, Contrib. Plasma Phys. 41, 207 (2001). G. Onida, L. Reining, R.W. Godby, R. Del Sole, and Wanda Andreoni, Phys. Rev. Lett. 75, 818 (1995). D. Blaschke, C. Roberts, and S. Schmidt, Phys. Lett. B 425, 232 (1998). A. H ll, C.D. Roberts, and S.V. Wright, nucl-th/0604029. o R. Fehr and W.D. Kraeft, Contrib. Plasma Phys. 35, 463 (1995); W.D. Kraeft, R. Fehr, Contrib. Plasma Phys. 37, 173 (1997). B.I. Lundqvist, Phys. Konden. Mater. 6, 193; 6, 206 (1967); 7, 117 (1968). Ben Yu-Kuang Hu, Phys. Rev. B 47, 1687 (1993). R. Fehr, PhD thesis (Greifswald, 1997). P. Ziesche, Ann. Phys. (Leipzig) 16, 45 (2007). W.J. Karzas and R. Latter, Astrophys. J. Suppl. Ser. 6, 167 (1967). C. Fortmann, R. Redmer, H. Reinholz, G. R pke, W. Rozmus, and A. Wierling, J. High Energy Dens. Phys. 2, 57 (2006). o H. Bethe and W. Heitler, Proc. Roy. Soc. London A 146, 83 (1934). M.L. Ter-Mikaelyan, Dokl. Akad. Nauk SSSR 94, 1033 (1953).

www.cpp-journal.org

Chapter 8 Bremsstrahlung and Line Spectroscopy of Warm Dense Aluminum Plasma generated by XUV Free Electron Laser

Authors: Ulf Zastrau, Carsten Fortmann, Roland Rainer Fustlin, Lei Feng Cao, Tilo a Dppner, Stefan Dsterer, Siegfried H. Glenzer, Gianluca Gregori, Tim Laarmann, Haeja o u Lee, Andreas Prsystawik, Paul Radclie, Heidi Reinholz, Gerd Rpke, Josef Tiggesbumker, o a Robert Thiele, Xuan Truong Ngyuen, Ingo Uschmann, Sven Toleikis, August Wierling, Thomas Tschentscher, Eckhart Frster, and Ronald Redmer o Appeared as regular article in Physical Review E, Vol. 78, Issue 6, pages 066406 15, Dec. 2008. Listing of contributions by authors: U.Z.: Execution of experiment, data processing, writing of manuscript C.F.: Bremsstrahlung calculations, writing of manuscript R.R.F.: Execution of experiment, HELIOS simulations, writing of manuscript A.W.: Calculation of relative line intensities, writing of manuscript L.F.C.: Execution of experiment, writing of manuscript R.R.: Writing of manuscript G.R.: Writing of manuscript H.R.: Writing of manuscript S.H.G.: Execution of experiment, writing of manuscript all others: Execution of experiment

148

149

PHYSICAL REVIEW E 78, 066406 2008

Bremsstrahlung and line spectroscopy of warm dense aluminum plasma heated by xuv free-electron-laser radiation

U. Zastrau,1,* C. Fortmann,2 R. R. Fustlin,3 L. F. Cao,1 T. Dppner,4 S. Dsterer,3 S. H. Glenzer,4 G. Gregori,5 T. Laarmann,3 H. J. Lee,6 A. Przystawik,2 P. Radcliffe,3 H. Reinholz,2 G. Rpke,2 R. Thiele,2 J. Tiggesbumker,2 N. X. Truong,2 S. Toleikis,3 I. Uschmann,1 A. Wierling,2 T. Tschentscher,3 E. Frster,1 and R. Redmer2

Institut fr Optik und Quantenelektronik, Friedrich-Schiller-Universitt, Max-Wien Platz 1, 07743 Jena, Germany 2 Institut fr Physik, Universitt Rostock, 18051 Rostock, Germany 3 Deutsches Elektronen-Synchrotron DESY, Notkestrasse 85, D-22607 Hamburg, Germany 4 Lawrence Livermore National Laboratory, University of California, P.O. Box 808, Livermore, California 94551, USA 5 Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, United Kingdom 6 Department of Physics, University of California, Berkeley, California 94720, USA Received 25 April 2008; revised manuscript received 20 October 2008; published 30 December 2008 We report the creation of solid-density aluminum plasma using free-electron laser FEL radiation at 13.5 nm wavelength. Ultrashort pulses were focused on a bulk Al target, yielding an intensity of 2 1014 W / cm2. The radiation emitted from the plasma was measured using an xuv spectrometer. Bremsstrahlung and line intensity ratios yield consistent electron temperatures of about 38 eV, supported by radiation hydrodynamics simulations. This shows that xuv FELs heat up plasmas volumetrically and homogeneously at warm-dense-matter conditions, which are accurately characterized by xuv spectroscopy. DOI: 10.1103/PhysRevE.78.066406 I. INTRODUCTION PACS number s : 52.50.Jm, 52.25.Os, 52.27.Gr, 52.70.La

1

The physics of warm dense matter WDM 1 has gained increasing interest because of its location in the transition region from cold condensed materials to hot dense plasmas. WDM states are of paramount importance to model astrophysical objects such as giant planets 2 or brown dwarfs 3 . Furthermore, WDM occurs as a transient state in novel experiments to generate high-energy densities in materials, most notably the realization of inertial connement fusion 4,5 . The rst experimental investigations of WDM have been performed, e.g., with shock-wave experiments 6 and with laser-excited plasmas 711 . WDM describes materials at temperatures of several eV at solidlike densities. Its creation and investigation under controlled conditions in the laboratory is a difcult task. Using common optical short-pulse lasers, nonlinear absorption leads to rapid temporal variations, steep spatial gradients, and a broad spectrum of plasma physical processes. Pioneering techniques such as laser-driven shock heating, x-ray heating, ion-heating techniques 6,1215 , and x-ray Thomson scattering 16 have been developed in order to improve the plasma heating mechanism. In WDM the electron temperature is comparable to the Fermi energy, i.e., the degeneracy parameter = kBTe / EF is close to unity. Furthermore, the ion coupling parameter = Z2 / 4 0kBTi 4 ni / 3 1/3 is greater than or equal to unity, i.e., the interparticle Coulomb correlation energy is equal or exceeds the thermal energy; Z is the ion charge, and ni is the ion density. Thus, electrons as well as ions exhibit strong temporal and spatial correlations which depend strongly on the plasma parameters, temperature, and density. A proper description of WDM is also a tremendous challenge to many-particle physics. Both the theory for ideal

plasmas and for condensed matter fail in this regime. Classical plasma theory based on expansions of the correlation contributions in powers of the coupling parameter breaks down since 1, and strong-coupling effects among the various species have to be taken into account. On the other hand, the plasma is too hot to be considered as condensed matter, i.e., expansions in terms of the degeneracy parameter also fail. Thus, precise knowledge of physical properties as a function of the plasma parameters temperature and density is of primary importance. In this article, we demonstrate that xuv free-electron lasers FELs open a promising possibility to heat matter volumetrically and homogeneously. Furthermore, we show that xuv bremsstrahlung and line spectroscopy allow the determination of the plasma temperature and free-electron density.

II. EXPERIMENTAL SETUP A. Free-electron laser characteristics

*zastrau@ioq.uni-jena.de

1539-3755/2008/78 6 /066406 5

The fourth-generation light source FLASH free-electron laser in Hamburg uses the self-amplied spontaneous emission SASE principle to generate brilliant xuv pulses 17,18 . In the experiment, pulses of 91.8 eV photon energy wavelength = 13.5 nm , 15 5 fs duration were focused to a 30- m spot by a carbon-coated ellipsoidal mirror with grazing incidence angle of 3 and 2 m focal length, as shown in Fig. 1. Since the FEL process starts from spontaneous radiation, it shows an intrinsic shot-to-shot pulse energy spread. The relevant distribution of pulse energies for this experiment is shown in Fig. 2. The total number of pulses in our measurement was 81 000, including 3% of zero-energy events and a signicant fraction of high-energy pulses up to 130 J. The mean value is 48 J, measured at the end of the undulators. Since the beamline transmission is known to be T = 0.68, the average pulse energy at the experiment is 33 J.

2008 The American Physical Society

066406-1

150

ZASTRAU et al.

PHYSICAL REVIEW E 78, 066406 2008

Rayleigh scattering

-1

10 10

emitted photons

-1

10 10

17.0 nm (L-edge)

10

10 12 14 wavelength [nm]

16

18

FIG. 1. Color online Experimental setup. a The xuv pulses from the FEL undulators are focused by an elliptical beamline mirror on the bulk aluminum target, creating a 30- m focus. b The target is hit under 45 and plasma is created. A high-resolution xuv spectrometer observes the plasma emission under 45 to the surface normal. B. xuv plasma spectroscopy

FIG. 3. Color Experimental xuv photon spectrum per solid angle = 4 104 sr and wavelength interval = 0.025 nm symbols with error bars and bremsstrahlung calculations for different electron temperatures. The spectrum is corrected for the spectrometer throughput and detection efciency.

Each pulse irradiates a bulk aluminum target under 45 incident angle, resulting in an average intensity of 2 1014 W / cm2; see Fig. 1. The polarization is linear in the horizontal plane. At the chosen wavelength, the critical density for penetration into the bulk, ncrit = 2 c 2 0me / e2 2 = 6.1 1024 cm3, is about 40 times higher than the valence electron density in cold solid aluminum, ne = 1.6 1023 cm3; therefore, the initial absorption length is 40 nm 19 . The pulse energy is deposited in a target volume of 15 m 2 40 nm, generating WDM. The number of 1012 atoms in this volume is in the same order of magnitude as the incident photon number.

At xed target position, the emission spectrum of about 104 exposures was recorded before moving to a fresh site. Since the very rst FEL pulse ablates a few-nanometer-thin surface layer, a further surface cleaning technique was not necessary. The target was at ambient temperature, and vacuum was kept constant at 107 mbar. The FEL was run in multibunch operation mode at a 5-Hz repetition rate, with 20 bunches per train giving 100 FEL pulses per second. Five separate measurements of different durations adding up to a total interaction time of 13.5 min were performed. Since the individual spectra look identical, we assume that plasma formation and emission processes vary very slowly during the measurement. xuv emission spectra in the region of 6 18 nm were measured with a high-throughput spectrometer, featuring a toroidal Ni-coated focusing mirror and a free-standing transmission grating. The spectrometer is described in detail in Ref. 20 . The spectral resolution was limited to 0.2 nm due to a slightly uctuating plasma position. A back-thinned xuv charge-coupled device CCD camera with 13 13 m2 pixel size and a quantum efciency of = 0.45 served as the detector. From the measured spectra, absolute photon numbers per wavelength interval and solid angle are calculated using the tabulated efciency of all components.

III. DATA ANALYSIS

FIG. 2. Color online Histogram of the energy spread of 13.5-nm xuv pulses generated by self-amplied spontaneous emission SASE . Only pulses contributing to the measured spectrum of the performed experiment are shown. The top ordinate shows the corresponding irradiation intensities on target.

Figure 3 shows the sum of all spectra in logarithmic scale after correction for the spectrometer throughput and detection efciency. The error bars arise essentially from statistical signal-to-noise ratios beside uncertainties of the spectrometer components. The main peak at 13.5 nm stems from Rayleigh scattering of FEL photons by bound aluminum electrons. It is broadened symmetrically by 0.4 nm due to artifacts originating from the support grid of the transmission grating. Spectral lines from Al IV and Al V are identied using the NIST tables 21 as listed in Table I. The continuum emission is formed by free-free transition radiation bremsstrahlung and free-bound recombination radiation.

066406-2

151

BREMSSTRAHLUNG AND LINE SPECTROSCOPY OF WARM TABLE I. List of the spectral lines emitted from the plasma, identied with the NIST database 21,23 . Experimental Reference Relative Oscillator data data intensity strength f nm nm NIST NIST 11.6 12.7 16.2 17.2 0.2 0.2 0.2 0.2 11.646 12.607 16.169 17.14 250 800 700 0.247 0.332 PHYSICAL REVIEW E 78, 066406 2008 B. Transition line ratio

Transition Al IV: 2s22p6 2s22p5 2 p0 4d 1/2 Al V: 2s22p5 2s22p4 1D 3s Al IV: 2s22p6 2s22p5 2 p0 3s 1/2 LII/IIIM

Independently, the electron temperature can be obtained from the ratio of integrated line intensities I for the identied transition lines from the Boltzmann distribution 29 as follows: I1 = I2

3 1 f1 e 3 2 f2

1 2 /kBTe

Since the target is heated volumetrically, the continuum emission is partly reabsorbed and we observe a steplike feature at 17.0 nm that is consistent with the absorption LII/III edge 19 . This indicates deep deposition of energy into the target, as expected for xuv photon-matter interactions. The corresponding absorption L edge at a similar excitation ux was also analyzed in laser-excited silicon 22 . Finally, the LII/IIIM uorescence line 23 is observed at 17.2 nm.

A. Bremsstrahlung

The experimental spectra allow the determination of the plasma parameters using fundamental relations 24 . The electron temperature and density are inferred from the continuum background radiation due to bremsstrahlung. We compare the experimental data to Kramers law 25 e2 4

0 3

and oscillator with the corresponding photon frequencies strengths f , having in mind that the plasma is optically thin. Here, the Al IV lines doublets at 1 = 16.169 nm and 2 = 11.646 nm, with oscillator strengths given in Table I, are used. Integration was performed from 15.9 to 16.3 nm and from 10.8 to 11.9 nm after subtraction of the bremsstrahlung continuum, respectively. The resulting temperature is 34 6 eV. Within the error bounds, this is consistent with the temperature inferred by analysis of the bremsstrahlung continuum, so that we can state the plasma temperature with about 38 eV. A full compliance of line and continuum temperature cannot be expected, since the plasma dynamics affects both emission processes in different ways. At early times after the laser-target interaction, the system is still very dense and the excited levels of the transitions under consideration are possibly dissolved into the continuum. Only after expansion are the excited levels well dened and radiative transitions take place. Bremsstrahlung, on the other hand, is most relevant at early times due to the n2 dependence of its emissivity; see e Eq. 1 . Thus, in the bremsstrahlung emission we expect a higher temperature than in the line spectrum.

C. Relative ion abundance

jff

n2 e

16 Ze2 3mec2

2

c/ kBTe

6 k BT em e

gT

for the free-free emissivity jff . Here, me is the electron is the wavelength-dependent Gaunt factor mass and gT 26 , accounting for medium and quantum effects. It is calculated in a Sommerfeld approximation 27 . We assume an average ion charge of Z = 4, which is supported by calculations of the relative ion abundances using the code COMPTRA04 28 ; see below. In the wavelength range from 7.1 nm to 8.0 nm marked by the dotted box in Fig. 3 , we expect no essential contributions from bound-bound and bound-free transitions. Statistical analysis of the data in this range yields 40.5 eV for the temperature with a rms error of 5.5 eV. Bremsstrahlung spectra for 35 eV, 40.5 eV, and 46 eV are shown in Fig. 3. Reabsorption was considered using tabulated opacity data 19 ; in this way, the L edge at 17.0 nm is reproduced. The height of the L edge corresponds to a transmission through 40 nm of cold aluminum. Kramers law Eq. 1 depends on the square of ne. From the absolute photon number at = 17.0 nm, Nphoton 1 = 51 424 1 , we calculate the free-electron density using Eq. 1 as ne = 4.0 1022 cm3, taking the inferred plasma temperature of 40.5 eV. This value is consistent with radiation hydrodynamics simulations; see below.

The relative abundance of aluminum ions was calculated with the code COMPTRA04 28 . Results for electron temperatures from 10 to 50 eV are shown in Fig. 4, assuming a solid density sol = 2.7 g / cm3 and 0.5 soli.e., slightly expanded. The concentration of ion species for the relevant temperatures between 34 and 40 eV complies with the observed line emission spectrum. At an averaged temperature of 38 eV, the ion fractions of Al IV, Al V, and Al VI take the values of about 7%, 45%, and 46%, respectively. Amounts of Al I III as well as Al VII and higher are negligible. All expected spectral lines from Al IV in the observed spectral range are either observed at 16.2 nm or covered by the strong FEL signal at 12.9 nm . For transition details see Table I. Weak lines from Al V between 11.9 and 13.2 nm overlap with the Rayleigh peak and only the 12.607-nm line can be identied. Transition lines from Al VI, which are located between 8.6 and 10.9 nm, are not signicantly present compared to the detector noise. This indicates that the corresponding high-lying excited levels are dissolved and do not contribute. Emission lines from Al VII and higher are expected to play a signicant role only at temperatures exceeding 40 eV, as shown in Fig. 4. These lines have also not been found, but have previously been observed in optical laser-matter interaction experiments 30 . This contrast is well understood by

066406-3

152

ZASTRAU et al.

1 0.8 0.6 0.4 0.2 0 20 Al III, Z=2 Al IV, Z=3 Al V, Z=4 Al VI, Z=5 Al VII, Z=6

PHYSICAL REVIEW E 78, 066406 2008

0.3 0.2 0.1 laser 0 -0.1

0

[1/cm3]

23

2.5x10

0.025

[eV]

25.7

time [ps]

0.0

0.1

30 40 electron temperature kBTe [eV] 50

depth [m]

-0.1

FIG. 4. Color Calculation of the relative Al ion species abundance from COMPTRA04 as a function of the electron temperature. The solid line is for the Al density sol = 2.7 g / cm3, dashed line 0.5 sol = 1.35 g / cm3.

FIG. 5. Color HELIOS simulation results for the electron density left and electron temperature right as a function of time and radius.

scrutinizing the different mechanisms of absorption and ionization in the case of optical light as opposed to the case of xuv photons.

IV. DISCUSSION A. Different photon absorption mechanisms

equilibrate at a temperature of several eV up to hundred femtoseconds 34,35 , forming typical WDM. During this time, electrons can also recombine with the 2p holes by emitting uorescence radiation at 17.2 nm wavelength. This state of matter cools down on a picosecond time scale by energy transfer to the lattice via electron-phonon scattering 24 .

B. Hydrodynamics simulation

In optical laser-matter interactions, dominant absorption mechanisms are multiphoton ionization, nonlinear processes, inverse bremsstrahlung, and resonance absorption 31 . When the critical free-electron density of the optical laser is exceeded, most light is reected and absorption is limited to the skin layer, leaving behind a plasma with steep density and temperature gradients. For xuv photons, nonlinear absorption is negligible at the considered intensity and resonance absorption is not important since the plasma is undercritical. Thus, hot-electron production in the keVMeV range is unlikely. Photoexcitation and inverse bremsstrahlung are the only possible mechanisms. Since there are no prepulses in the FEL case, the deposition of photons starts in a cold target and the energy is distributed volumetrically and homogeneously throughout the interaction zone. A change in the polarization of the FEL e.g., using circularly polarized light could lead to less heating and certainly to a reduction in the scattered intensity Rayleigh peak . For atomic systems, a systematic decrease in the photoionization yield has been reported upon changing the lasers polarization 32 , while such studies for solid-state targets at xuv conditions remain to be done. The plasma production mechanism presented here takes advantage of photons exceeding a 2p-level binding energy of 72.8 eV. In particular, a 2p bound electron is photoionized with a cross section of PI = 7 Mbarn 19 . This electron is transferred into the conduction band, leaving behind a hole in the 2p shell. For partially ionized aluminum the photoabsorption cross section increases below the LII/III edge slightly with temperature 33 . Due to their high excess energy of about 20 eV, further electrons at lower energies are excited via impact ionization and Auger processes 1 . The electrons

To illustrate the hydrodynamic processes and to estimate the electron temperature, one-dimensional radiation hydrodynamics simulations using HELIOS 36 have been performed. HELIOS features a Lagrangian reference frame i.e., the grid moves with uid, separate ion and electron temperatures, and ux-limited Spitzer thermal conductivity. It allows the deposition of laser energy via inverse bremsstrahlung as well as bound-bound and bound-free transitions, using a SESAME-like equation of state. Per atom, 2.6 conduction band electrons were assumed to contribute to the laser absorption 37 . The results are shown in Fig. 5. On the time scale of the FEL pulse, both electron density and temperature rise up to values of ne 1023 cm3 and kBTe 26 eV, respectively, without any steep gradient. These values are in good qualitative agreement with the results for ne and Te obtained by the spectral analysis. This hydrodynamics simulation uses an average pulse energy of 33 J on the target as an input parameter. As discussed above, Fig. 2 illustrates that a signicant fraction of pulses have much higher pulse energies up to 130 J. Since the free-free emissivity Eq. 1 depends strongly nonlinearly on both electron temperature and density, we expect the observed radiation to be rather dominated by the high-energy fraction of the xuv pulses than by its average value. This nally explains the slight underestimation of the electron temperature in the hydrodynamics simulation compared to the experimental results. Additionally, the simulation shows that these WDM conditions exist for about 200 fs, at almost constant plasma density and temperature, and hydrodynamic motion is negligible 38 . Plasma expansion as well as electron diffusion to the cold matter of the bulk target and heat conduction becomes important during the rst several picoseconds 39 , while the density decreases about a factor of 2, inuencing the relative

066406-4

153

BREMSSTRAHLUNG AND LINE SPECTROSCOPY OF WARM PHYSICAL REVIEW E 78, 066406 2008

V. SUMMARY AND CONCLUDING REMARKS

portant for shock-wave physics, applied-material studies, planetary physics and inertial connement fusion, and other forms of high-energy density-matter generation.

ACKNOWLEDGMENTS

The capability of xuv FEL radiation to create WDM by interaction with a solid aluminum target was demonstrated. The analysis of the xuv line and continuum emission spectra yields an electron temperature of 34 6 eV and 40.5 5.5 eV, respectively. The observed line spectrum is compatible with predicted ion abundances. Together with radiation hydrodynamics modeling, we get a sound picture of complex xuv laser-plasma interaction dynamics. The simulations conrm the volumetric heating of the target without strong gradients. Our results provide complementary information to results that were reported for optical laser-matter interactions 30 . Further and detailed studies of WDM will include spatially and temporally resolved experiments to determine the electron temperature and density. For this regime, novel diagnostic techniques such as x-ray interferometry 40,41 and x-ray Thomson scattering 34,42,43 have been developed. In combination with these techniques, the xuv FEL will be a unique platform for WDM investigations. This will be im-

We thankfully acknowledge nancial support by the German Helmholtzgemeinschaft via the Virtual Institute VH-VI104, the German Federal Ministry for Education and Research via Project No. FSP 301-FLASH, and the Deutsche Forschungsgemeinschaft DFG via the Sonderforschungsbereich SFB 652. T.L. acknowledges nancial support from the DFG under Grant No. LA 1431/2-1. R.R.F. received DFG funds via Grant No. GRK 1355. The work of S.H.G. was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344. S.H.G. was also supported by LDRDs 08-ERI-002, 08-LW-004, and the Alexander von Humboldt foundation. The work of G.G. was partially supported by the Science and Technology Facilities Council of the United Kingdom. Finally, the authors are greatly indebted to the machine operators, run coordinators, and scientic and technical teams of the FLASH facility for enabling an outstanding performance.

1 2 3 4 5 6 7

8 9 10

11

12 13 14

15 16

17 18 19

R. W. Lee et al., Laser Part. Beams 20, 527 2002 . N. Nettelmann et al., Astrophys. J. 683, 1217 2008 . D. Saumon et al., High Press. Res. 16, 331 2000 . J. D. Lindl et al., Phys. Plasmas 11, 339 2004 . T. R. Dittrich et al., Phys. Plasmas 6, 2164 1999 . L. B. Da Silva et al., Phys. Rev. Lett. 78, 483 1997 . Y. Ping, D. Hanson, I. Koslow, T. Ogitsu, D. Prendergast, E. Schwegler, G. Collins, and A. Ng, Phys. Rev. Lett. 96, 255003 2006 . T. Ao, Y. Ping, K. Widmann, D. F. Price, E. Lee, H. Tam, P. T. Springer, and A. Ng, Phys. Rev. Lett. 96, 055001 2006 . K. Widmann, T. Ao, M. E. Foord, D. F. Price, A. D. Ellis, P. T. Springer, and A. Ng, Phys. Rev. Lett. 92, 125002 2004 . G. W. Collins, P. M. Celliers, L. B. Da Silva, R. Cauble, D. M. Gold, M. E. Foord, N. C. Holmes, B. A. Hammel, R. J. Wallace, and A. Ng, Phys. Rev. Lett. 87, 165504 2001 . A. Saemann, K. Eidmann, I. E. Golovkin, R. C. Mancini, E. Andersson, E. Forster, and K. Witte, Phys. Rev. Lett. 82, 4843 1999 . T. S. Perry et al., Phys. Rev. E 54, 5617 1996 . J. J. MacFarlane et al., Phys. Rev. E 66, 046416 2002 . P. K. Patel, A. J. Mackinnon, M. H. Key, T. E. Cowan, M. E. Foord, M. Allen, D. F. Price, H. Ruhl, P. T. Springer, and R. Stephens, Phys. Rev. Lett. 91, 125004 2003 . S. Sakabe, R. Sigel, G. D. Tsakiris, I. B. Foldes, and P. Herrmann, Phys. Rev. A 38, 5756 1988 . S. H. Glenzer, G. Gregori, R. W. Lee, F. J. Rogers, S. W. Pollaine, and O. L. Landen, Phys. Rev. Lett. 90, 175002 2003 . W. Ackermann et al., Nat. Photonics 1, 336 2007 . V. Ayvazyan et al., Phys. Rev. Lett. 88, 104802 2002 . B. L. Henke et al., At. Data Nucl. Data Tables 54, 181 1993 .

20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

41 42 43

J. Jasny et al., Rev. Sci. Instrum. 65, 1631 1994 . V. Kaufman et al., J. Phys. Chem. Ref. Data 20, 775 1991 . K. Oguri et al., Appl. Phys. Lett. 87, 011503 2005 . J. A. Bearden, Rev. Mod. Phys. 39, 78 1967 . H. R. Griem, Principles of Plasma Spectroscopy Cambridge University Press, Cambridge, England, 1997 . H. A. Kramers, Philos. Mag. 46, 836 1923 . J. A. Gaunt, Proc. R. Soc. London, Ser. A 126, 654 1930 . C. Fortmann et al., High Energy Density Phys. 2, 57 2006 . S. Kuhlbrodt et al., Contrib. Plasma Phys. 45, 73 2005 . W. Lochte-Holtgreven, in Plasma Diagnostics, edited by W. Lochte-Holtgreven AIP, New York, 1995 , p. 135. U. Teubner et al., Appl. Phys. Lett. 59, 2672 1991 . P. Gibbon et al., Plasma Phys. Controlled Fusion 38, 769 1996 . K. J. LaGattuta, Phys. Rev. A 43, 5157 1991 . M. Fajardo et al., Eur. Phys. J. D 29, 69 2004 . T. Tschentscher et al., Eur. Phys. J. D 36, 193 2005 . A. L. Dobryakov et al., Phys. Scr. 60, 572 1999 . J. MacFarlane et al., J. Quant. Spectrosc. Radiat. Transf. 99, 381 2006 . R. LaVilla et al., Phys. Rev. Lett. 9, 149 1962 . M. M. Murnane et al., Science 251, 531 1991 . A. Krenz et al., Eur. Phys. J. D 36, 199 2005 . L. B. Da Silva, T. W. Barbee, R. Cauble, P. Celliers, D. Ciarlo, S Libby, R. A. London, D. Matthews, S. Mrowka, J. C. Moreno, D. Ress, J. E. Trebes, A. S. Wan, and F. Weber, Phys. Rev. Lett. 74, 3991 1995 . J. Filevich et al., Appl. Opt. 43, 3938 2004 . S. H. Glenzer et al., Phys. Rev. Lett. 98, 065002 2007 . A. Hll et al., High Energy Density Phys. 3, 120 2007 .

066406-5

154

Chapter 9 Self-Consistent Spectral Function for Non-Degenerate Coulomb Systems and Analytic Scaling Behaviour

Author: Carsten Fortmann Appeared as regular article in Journal of Physics A: Mathematical and Theoretical, Vol. 41, Issue 44, pages 445501 127, Sept. 2008

156

157

IOP PUBLISHING J. Phys. A: Math. Theor. 41 (2008) 445501 (27pp) JOURNAL OF PHYSICS A: MATHEMATICAL AND THEORETICAL

doi:10.1088/1751-8113/41/44/445501

Self-consistent spectral function for non-degenerate Coulomb systems and analytic scaling behaviour

Carsten Fortmann

Institute of Physics, Rostock University, 18051 Rostock, Germany E-mail: carsten.fortmann@uni-rostock.de

Received 29 May 2008, in nal form 11 September 2008 Published 7 October 2008 Online at stacks.iop.org/JPhysA/41/445501 Abstract Novel results for the self-consistent single-particle spectral function and selfenergy are presented for non-degenerate one-component Coulomb systems at various densities and temperatures. The GW (0) -method for the dynamical selfenergy is used to include many-particle correlations beyond the quasi-particle approximation. The self-energy is analysed over a broad range of densities and temperatures (n = 1017 cm3 1027 cm3 , T = 102 eV/kB 104 eV/kB ). The spectral function shows a systematic behaviour, which is determined by collective plasma modes at small wavenumbers and converges towards a quasi-particle resonance at higher wavenumbers. In the low density limit, the numerical results comply with an analytic scaling law that is presented for the rst time. It predicts a power-law behaviour of the imaginary part of the selfenergy, Im n1/4 . This resolves a long time problem of the quasi-particle approximation which yields a nite self-energy at vanishing density. PACS numbers: 52.27.Aj, 52.65.Vv, 71.10.Ca, 71.15.m

1. Introduction Strongly correlated Coulomb plasmas, found e.g. in planetary interiors [1, 2], fusion plasmas [3], and plasmas excited by lasers or ion beams [4], are characterized by a high degree of spatial and temporal correlations, which lead to the emergence of phenomena like collective plasma modes, dynamical screening of the interparticle interaction potential, and dissolution of bound states. In particular, laser-excited plasmas cover a broad range of densities and plasma temperatures. Values range from typical condensed matter conditions to hot, weakly coupled plasmas. Theoretical approaches to the physical properties of such systems have to deal with a great complexity. A particular challenge is the formulation of a coherent theory, which is valid over a wide range of densities (n) and temperatures (T), thereby allowing to describe matter

1751-8113/08/445501+27$30.00 2008 IOP Publishing Ltd Printed in the UK 1

158

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

in various states, e.g. a solid-state target, being transferred into a plasma by interaction with high-power lasers and its subsequent relaxation [5]. Many-particle perturbation theory [6] presents a general approach to many-body systems like condensed matter [7], partially and fully ionized plasmas [8], and nuclear matter, to mention only a few. Also for non-abelian systems, such as the quarkgluon plasma [9], there exist similar approaches to that described here for Coulomb systems, e.g. the concept of SchwingerDyson equations, see the review article [10]. The thermodynamic properties as well as the response to external perturbations of these systems in various situations can be studied systematically [11]. The central quantity within the many-body theoretical approach is the single-particle spectral function A(p, ). It represents a physical observable which can be measured via angular resolved photoemission spectroscopy (ARPES) [1214]. Starting from the spectral function, a number of interesting questions related to the physics of many-particle systems can be addressed. The equation of state [15], transport cross-sections [16] (e.g. electrical conductivity, thermal conductivity and stopping power [17]) and optical properties [18] (emission and absorption of electromagnetic radiation) become accessible. In this work, the focus is on the spectral function of plasmas. As an example, a onecomponent electron plasma is considered which is charge compensated by a homogeneously distributed background of positively charged ions (jellium model). The plasma is characterized by the degeneracy parameters and the plasma coupling parameter which are dened as = 2mkB T , h2 (3 2 n)2/3 = e2 4 0 kB T 4 n 3

1/3

(1)

Here, the electron mass m was introduced, kB is the Boltzmann constant. In this work, we consider only non-degenerate systems, 1, i.e. the thermal energy kB T is large compared to the Fermi energy EF = h2 (3 2 n)2/3 /2m. The calculation of the spectral function becomes challenging in the regime of strong coupling, i.e. when the plasma coupling parameter becomes comparable or larger than unity. The coupling parameter measures the ratio of the Coulomb interaction energy of two particles 1, particle collisions become frequent, at a mean distance to their thermal energy kB T . At involving transfer of both momentum and energy. The interparticle potential is screened due to the presence of nearby third particles. These correlations signicantly modify the plasma observables and have to be accounted for in the calculation of the spectral function. This is accomplished via the single-particle self-energy (p, ), which is a complex function of both wavevector p and frequency , leading to a structured spectral function. Though, the main task of many-particle theory, applied to strongly coupled systems, is to calculate the self-energy in a suitable approximation. The simplest approximation, often found in the literature on Coulomb systems, is the mean-eld or HartreeFock approximation [8]. One obtains a frequency-independent selfenergy which induces a shift in the spectral functions pole, the so-called HartreeFock or quasi-particle shift. For dilute plasmas, this correctly describes the lowering of the chemical potential due to the averaged eld of the plasma particles. Also, the shift of the ionization energy for bound states is obtained [19, 20]. However, in dense systems, the mean-eld approximation breaks down since the dynamical screening and collective excitations cannot be accounted for. One has to go beyond the quasi-particle picture. A particularly successful approximation for the self-energy, including these dynamical effects, is the so-called GW -approximation [21, 22]. Correlations are accounted for via the dynamically screened interaction potential W (q, ), rather than via the bare Coulomb interaction. The GW approximation knows a long history of applications in the eld of condensed matter theory. Examples are the calculation of single-particle spectra in the

2

159

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

homogeneous electron gas [2325], bandgaps in semiconductors [26], effective masses of metal electrons [27], optical and electronic properties of insulators [28], electronic structure of superconductors [29], but also atomic and molecular systems [3032]. In particular, GW self-energy corrections systematically improve band-gap calculations performed by means of density functional theory [3335]. Recently, the GW approximation has been applied also to dense plasmas. Whereas Fehr et al [36] performed lowest order (one-loop) self-energy corrections to the equation of state, Wierling et al [37] carried out pioneering self-consistent calculations of the electron self-energy in the solar-core plasma. An asymmetrically broadened, otherwise featureless spectral function was obtained. In this work, the GW self-energy and the corresponding spectral function is investigated for non-degenerate, one-component electron plasmas. Only unbound electrons are considered, bound state contributions can be accounted for via T-matrix calculations, as done in [38]. The self-energy is evaluated for a broad range of densities and temperatures, going from ideal, weakly coupled plasmas ( 1) to the strong coupling regime 1. As a novel contribution to the eld, an analytic scaling law for the GW self-energy at low densities is derived which accurately describes the numerical data in this limit. This expression can be combined with corresponding formulae that are valid in the EF , to construct a t formula for the self-energy which then degenerate case, when kB T covers a large portion of the densitytemperature plane. Formerly, analytic expressions for the self-energy have been derived that base on the quasi-particle approximation [39]. In particular, the completely degenerate electron gas at T = 0 was considered, using the plasmon-pole approximation [40], and also weakly coupled kB T ), using the Born approximation for the self-energy ( 1), classical plasmas (EF [8]. The latter result exhibits several problems: the imaginary part of the quasi-particle selfenergy is independent of density and carries a prefactor 1/ . Thus, there is an unphysical h nite damping of single-particle states even in the vacuum and the classical limit h 0 is not dened. On the other hand, from physical arguments, one expects that the self-energy vanishes at zero density and that it is a purely classical expression ( = 0), when h 1. This problem has remained unresolved up to now. The real part of the quasi-particle self-energy is well behaved, i.e. it vanishes at zero density and is purely classical. The new analytic expression for the self-energy presented in this paper is derived without the quasi-particle approximation, i.e. it is a non-perturbative result. It is shown that only this non-perturbative treatment leads to an expression that is classical for both the real and the imaginary part and vanishes exactly in the vacuum limit n 0. The work is organized as follows: after a brief recapitulation of the single-particle spectral function and the GW -method in section 2, numerical results for the self-consistent spectral function and self-energy will be discussed in section 3. Section 4 contains the derivation of the non-perturbative scaling law and comparison to the numerical results. In section 5, it will be analysed why the quasi-particle picture is incapable to give a physically consistent result for the imaginary part of the self-energy. Conclusions will be drawn in section 6. The appendix contains detailed calculations that are only summarized in the main part of the paper. 2. Spectral function and self-energy The derivation of the GW -approximation involves some lengthy manipulations. In this section, only the most relevant formulae are given, while appendix A contains the detailed steps. Central to the description of electronic properties in a many-body system, which is in thermodynamic equilibrium, is the thermodynamic electron single particle Green h function G(p, z ), dened at the discrete Matsubara frequencies z = (2 + 1) ikB T / ,

3

160

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

= 0, 1, 2, . . .. It is related to the single-particle self-energy equation G(p, z ) = G(0) (p, z ) + G(0) (p, z ) (p, z )G(p, z ) = G(0) (p, z )

(0) 1 1

(p, z )

(2)

with the free Green function G (p, z ) = [ z p ] . Also, the single-particle energy h p = h2 p2 /2m is introduced, is the electron chemical potential. G(p, z ) contains the thermodynamic properties of a single particle coupled to a thermal bath at a given temperature T. For example, the momentum distribution function is easily obtained by summation of the Green function over all Matsubara frequencies, n(p) = kB T

z

G(p, z ).

(3)

Instead of the complex Matsubara Green function, it is more convenient to operate on the real valued spectral function A(p, ), dened on the real frequency axis. It carries the same information as the Green function and is dened via the spectral representation of the latter, G(p, z ) =

d A(p, ) . 2 z

(4)

Here, is a real valued frequency. This relation can be resolved for A(p, ), A(p, ) = lim+ 2 Im G(p, + i)

0

(5) (6)

= lim+

0

i.e. the spectral function is obtained after analytic continuation of the Green function from the Matsubara frequencies to arbitrary complex frequencies as the imaginary part of G(p, + i), when approaches zero from positive values. In this way, the sign of the imaginary part of the self-energy is xed, i.e. Im (p, ) < 0 for > 0. The real part of the self-energy behaves unambiguous for = 0. The spectral function usually exhibits several resonances, including a central peak, located at the quasi-particle energy Ep , i.e. the solution of the quasi-particle dispersion Ep = p + Re (p, Ep / ), h (7) accompanied by symmetrically distributed satellites which are attributed to collective modes in the many-particle system [40]. The width of the resonances in the frequency domain is commonly identied with the inverse life-time of these excitations. Let us rst look at the lowest order approximation to the self-energy, the HartreeFock term HF (p). It is given by the convolution of a non-interacting Green function with the unscreened Coulomb potential V (q) = e2 / 0 q 2 0 ( 0 is a normalization volume),

HF

(p, z ) = kB T

,q

G(0) (p q, z )V (q)

HF

(8) (9)

=

q

(p),

with the Fermi distribution function nF ( ) = [exp( /kB T ) + 1]1 . In the rst line, h h h summation takes place over the Bosonic Matsubara frequencies = 2 ikB T / , =

4

161

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

0, 1, 2, . . . . The rst term q V (q) (Hartree term) diverges, but it is exactly compensated by the same term from the positive charge background. The second term (Fock term or exchange term) gives a nite contribution. Closed expressions can be given in the case of nondegenerate plasmas [8, 39] and completely degenerate Fermi gases [7]. One nds HF (p) n EF ) and HF (p) n1/3 in the quantum degenerate in the high temperature limit (kB T EF , see [8] for details. Thus, the HartreeFock self-energy fulls the physical case kB T constraint to vanish at zero density. The HartreeFock term is a real function of momentum, only. The corresponding spectral function is shifted from the free particle dispersion, AHF (p, ) = 2 (p +

HF

(p) h).

(10)

No imaginary part of the self-energy appears in this approximation, i.e. the life-time of the HartreeFock quasi-particles is innite. This is consequence of the mean-eld approximation, where no uctuations of the electric eld, i.e. no dynamics of the surrounding plasma particles are taken into account. Recently, also the second-order exchange contribution to the selfenergy has been obtained in closed form [41, 42], see also [43]. However, this term and all higher order terms, involving only the bare Coulomb potential, do not lead to a nite particle life-time, only a shift of the dispersion relation is obtained. To describe the situation in a dense and strongly correlated system, where the singleparticle states are spectrally broadened, i.e. they acquire a nite life-time, one has to go beyond the quasi-particle approximation, and take into account the screening of the interaction. The GW -approximation, can be regarded as the generalization of the HartreeFock theory to dynamically screened interactions. It was introduced by Hedin [23] for the homogeneous electron gas, and is dened as (p, z ) = kB T

q,

G(p q, z )W (q, ).

(11)

W (q, z) is the dynamically screened interaction. Note that the GW approximation is a selfconsistent ansatz, since the self-energy appears on the lhs as well as in the Green function on the rhs of (11). Also, the screened interaction W (q, ) is a functional of the Green function via the dielectric function (q, ), i.e. the polarization function (q, ): W (q, ) = V (q) V (q) = . (q, ) 1 V (q) (q, ) (12)

In GW -approximation, (q, ) is given by the inner product of two Green functions, (q, ) = kB T p,z G(q + p, z + )G(p, z ). The double self-consistency implied in this ansatz makes the GW -approximation complicated and a numerically demanding problem. On the other hand, the full GW -approximation suffers from deciencies due to the neglect of vertex-corrections [44], such as violation of the f -sum rule [45]. This problem can be avoided by keeping the dynamically screened interaction on the level of the random phase approximation (RPA) [46], dened by the RPA polarization function,

RPA (q, )

= kB T

p,z

(13)

RPA (q,

+ i) =

k

(14)

The use of the RPA polarization function leads to the so-called GW (0) approximation for the self-energy. It has been shown to give more accurate quasi-particle energies [25]

5

162

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

than the full GW -approximation. Additionally, it is known that higher order corrections beyond GW (0) , such as vertex-corrections and corrections in the polarization function beyond RPA, partially compensate. Therefore, ignoring them altogether is expected to give better results than accounting for one or the other [22]. The f -sum rule is fullled. Further sum rules, e.g. for the moments of the spectral function can be derived [24] which are useful to control the numerical treatment of the integral equations to solve. 1 The inverse dielectric function RPA (q, ) describes the propagation of electromagnetic waves in the plasma. As a main feature, it contains the longitudinal plasma oscillations or plasmons. These resonances show up as peaks in the inverse dielectric function, located at the roots of the plasmon dispersion Re RPA (q, ) = 0. For non-degenerate systems, as considered here, the plasmon dispersion can be expanded in powers of the wavenumber q, 2 2 h and one nds the GrossBohm relation [47] res (q) = pl (1 + q 2 / 2 ) + ( q 2 /2m)2 for the plasmon resonance frequency res (q). Here, the plasma frequency pl and the inverse Debye screening length , ne2 ne2 pl = , = , (15) 0m 0 kB T have been introduced. A detailed discussion of the plasmon resonance in dense plasmas can be found in [48]. For the present discussion, it is important to keep in mind that the collective plasma excitations are accounted for via the inverse dielectric function in RPA. This is the main advantage of the GW (0) -approximation compared to the mean-eld or HartreeFock approximation. Depending on the choice of parameters like density and temperature, these plasmon resonances determine the shape of the self-energy as a function of the frequency and thereby also the spectral function, where satellites besides the quasi-particle peak indicate coupled electronplasmon modes, often referred to as plasmarons [40]. It should be noted at this point that contributions from bound states to the self-energy are not accounted for in this work. The description is limited to fully ionized plasmas. Bound state contributions can be included using the concept of the T-matrix, see e.g. the work by Schmielau et al [38]. Using the spectral representation (4) and the screened interaction (12), the following equation for the imaginary part of the self-energy in GW (0) -approximation is obtained after summation over the Bosonic Matsubara frequencies , d h V (q)A(p q, ) Im (p, + i) = nF ( ) q 2 h

1 Im RPA (q, )nB ( )nF ( h ), h h (16) 1 h h with the BoseEinstein distribution function nB ( ) = [exp( /kB T ) 1] . The real part of the self-energy is obtained by means of Hilbert transform as d Im (p, ) HF Re (p, ) = int (p) + P . (17) HF P denotes the Cauchy principal value integration, int (p) is the HartreeFock self-energy of the interacting system, d HF A(p q, )nF ( )V (q). (p) = h h (18) int 2 q 1/2 1/2

Finally, to close the set of equations, the chemical potential has to be xed by inversion of the density relation d h A(p, )nF ( ). h (19) n(, T ) = 2 0 p 2

6

163

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

The factor 2 in front of the rhs stems from the summation over the spin components. Together with Dysons equation (6), (16)(19) constitute a system of nonlinear integral equations for the self-energy. Besides the normalization of the spectral function

d A(p, ) = 1, 2

(20)

similar sum-rules can be derived also for higher moments of the spectral function [24]. These are independent of the concrete approximation used for the self-energy. In second order, one obtains an equation relating the rst moment of the spectral function to the interacting HartreeFock self-energy (9), h2

d A(p, ) = p + 2

HF int (p).

(21)

Similarly, the second moment is related to the HartreeFock energy and the frequency integrated imaginary part of the self-energy, which is itself a conserved quantity, at least within the GW (0) approximation, see (23) below, h3

d 2 A(p, ) = h 2

d Im (p, + i) + p +

2 HF int (p) .

(22)

For the GW (0) self-energy, Holm and von-Barth have found the following identity, relating the integrals over the imaginary part of the self-energy to the totally integrated response function,

d Im (p, + i) = h

d V (q)Im 2

1 RPA (q, ).

(23)

In the next section, results for the self-energy will be presented that are obtained via numerical solution of (16). The sum rules given above are used to check the accuracy of the numerical results. 3. Numerical results The GW (0) -approximation is evaluated numerically for various sets of plasma parameters in the following. A typical example of a weakly coupled ( = 0.07), moderately degenerate ( = 2.2) plasma is the plasma at the solar core, with temperatures of T 100 Ry/kB 1360 eV/kB and electron densities of n 71025 cm3 [49]. The solar core plasma has been investigated using the GW (0) -method in a number of previous publications, see [18, 37, 50]. Here, most attention is paid to a systematic analysis of the single-particle spectral function and the self-energy over a broad range of densities and temperatures, however, sticking to nondegenerate plasmas and neglecting bound states. We therefore start with a plasma temperature that equals the solar core temperature and a density that is 10% of the solar core electron density. Later, higher and lower temperatures will be considered as well, i.e. kB T = 10 Ry and kB T = 1000 Ry. Note that kB T is always chosen large against typical binding energies of atoms which are usually of the order of several Ry. Thus, bound states can be neglected. The numerical solution of equation (16) is performed by means of an iterative algorithm, starting from a suitable initialization of the spectral function. Typically, the algorithm converges after 510 iterations. The threefold integral (16) is evaluated on a two-dimensional grid with roughly 100 nodes in the frequency coordinate and 1020 nodes in the momentum coordinate. The angular integral is performed rst, followed by the frequency integration and

7

164

J. Phys. A: Math. Theor. 41 (2008) 445501

Im (p=0,) [Ry] 0 -2 -4 -6 -8 -10 10 5 0 -5 -10 50 25 0 -25 -50 0.3 0.2 0.1 0.0 -40 -20 0 hpl h - pl energy h+ [Ry] 20 40 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40

C Fortmann

Figure 1. Self-energy (in units of the Rydberg energy, 1 Ry = 13.6 eV), dispersion relation (in units of Ry), and spectral function (in units of 1/Ry) for plasma density n = 7 1024 cm3 (10% of the solar core density) and temperature T = 100 Ry/kB = 1360 eV/kB . The spectral function contains two weakly pronounced plasmaron satellites, appearing at slightly smaller energies than the plasma frequency (dashed vertical lines). The chemical potential is = 377 Ry = 5133 eV.

the integration over the modulus of the wavenumber q. The result is checked for consistency in each iteration using the sum-rules (2123). Further details concerning the numerical implementation are provided in [51]. Figure 1 shows the numerical result for the self-energy (imaginary and real part), dispersion relation h + h2 p2 /2m Re (p, ), and the spectral function for plasma parameters chosen as n = 7 1024 cm3 for the plasma density, i.e. 10% of the solar core density, and T = 100 Ry = 1360 eV/kB for the plasma temperature. The chemical potential is = 377 Ry = 5133 eV. The momentum was xed at hp = 0. The spectral function at the chosen parameters is a broadened resonance with two satellites appearing at about h + = 5 Ry which is slightly below the plasma frequency hpl = 7.2 Ry at the chosen conditions. The plasma frequency is indicated by the dashed vertical lines. As already mentioned, these satellites are often referred to as plasmarons, i.e. a coupled mode between the single-particle resonance and the collective plasma oscillation [40]. The imaginary part of the self-energy (top graph) is peaked at the free dispersion h + = 0 and this peak leads to the small dip in the spectral function between the satellites, see (6). The

8

A(p=0,)

h+-Re

(p=0,)

Re (p=0,) [Ry]

165

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

0.6 A(p,) [1/Ry] 0.5 0.4 0.3 0.2 0.1 0 0 mo 5 me ntu 10 m p [ 15 a -1 B ] 20 -50 0

450

Figure 2. Spectral function for plasma density n = 71024 cm3 and temperature kB T = 1360 eV (solar core temperature) as a function of momentum and density. The black line on the bottom represents the free dispersion relation h = p = h2 p 2 /2m . At the present parameters the chemical potential is = 377 Ry.

real part of the self-energy (second graph from top) is a rather smooth function, leading to only small variations in the dispersion (third graph from top). Next, the dependence of the spectral function on the wavenumber p is analysed. In gure 2, the spectral function A(p, ) is shown for ve different wavenumbers, i.e. 1 1 1 1 p = 0, 5aB , 10aB , 15aB and 20aB , aB = 4 0h2 /me2 is the Bohr radius. The density and temperature are the same as before, n = 7 1024 cm3 , kB T = 100 Ry. At increased 1 wavenumber p 5aB , enhanced complexity of the spectral function is observed. The plasmaron peaks, which at hp = 0 appear as small shoulders in the otherwise broad central resonance, are better dened. The central quasi-particle peak itself becomes narrower and the 1 plasmaron peaks separate. At the highest momenta considered hp > 15aB , the plasmarons themselves are damped out, and a single, narrow resonance forms, located near the single particle energy h = p = h2 p2 /2m , i.e. the quasi-particle picture is restored. Some of these features, especially the plasmaron satellites are already known from literature [40]. Now that the general characteristics of the spectral function have been discussed, the central concern of this paper can be worked out, i.e. the analysis of the dependence of the self-energy and the spectral function on the plasma parameters density and temperature. In gure 3, the spectral function at p = 0 is shown for ve different densities between n = 71025 cm3 (solar core conditions) and 0.01% of the solar core density. The temperature is kept constant at T = 100 Ry/kB = 1360 eV/kB . The spectral function drastically changes with varied density. In the case of the highest density considered, a narrow quasi-particle peak accompanied by two separate plasmaron satellites (indicated by arrows) is observed. The quasi-particle peak is notably shifted from the free dispersion 0 = , due to the real part of the self-energy. Going to lower densities, the plasmaron satellites merge into the central peak, as can be seen in the case of the spectral function for n = 7 1024 cm3 and also the quasiparticle shift is reduced. Finally, at the lowest densities considered, n = 7 1022 cm3 and 7 1021 cm3 , a single, narrow quasi-particle resonance is obtained which is centred around the free dispersion. The width decreases with the density which is the expected behaviour in the low density limit.

9

166

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

25 -3

1.2 spectral function A(p=0,) [Ry] 1 0.8 0.6 0.4 0.2 0 -40 -20

0 20 energy h+ [Ry]

40

Figure 3. Spectral functions at p = 0 for different plasma densities, ranging from the density at the solar core, n = 7 1025 cm3 (solid curve) to 0.01% of the solar core (dash-dot-dotted curve). The plasma temperature is T = 100 Ry/kB = 1360 eV/kB for all ve curves.

_ -Im (0,E0/h) / k BT

10

-1

~n

1/4

~n

= 10 =1

1/4

= 10

=1

~n

1/4

10

-2

= 10

=1

10

17

10

18

10

19

10

20

10 10 10 10 10 -3 density n [cm ]

21

22

23

24

25

10

26

10

27

10

28

h Figure 4. Effective quasi-particle damping width at p = 0, Im (0, E0 / ) normalized to the thermal energy as a function of the plasma density. The arrows indicate for each temperature the density at which the degeneracy parameter takes the values = 10 and = 1.

In order to study the dependence of the self-energy on density and temperature in more h detail, the effective quasi-particle self-energy (p, Ep / ) as a function of the density at various temperatures is considered. This quantity gives the shift and width of the central peak h in the spectral function A(p, ), i.e. when is close to the quasi-particle frequency Ep / , see (7). The results for the imaginary part of the effective quasi-particle self-energy at p = 0 as a function of the plasma density are shown in gure 4. Three different temperatures have been assumed, T = 10, 100 and 1000 Ry/kB . Towards low densities, a systematic decrease

10

167

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

of Im (0, E0 / ) with the density is observed which is also known from the literature h [37]. The asymptotes to the low density behaviour, shown as thin dotted lines, indicate that h Im (p, Ep / ) scales proportional to n1/4 . This behaviour will be analysed in more detail in section 4, where an analytic solution for the GW self-energy is derived that exhibits the same n1/4 proportionality. At higher densities, the power law behaviour terminates and the self-energy starts to decrease. This can be understood by looking again at gure 3. Here, it was shown that at increased density, the plasmaron satellites separate from the central quasi-particle peak, i.e. spectral weight is shifted to the satellites and the central peak narrows. The calculations have only been performed for non-degenerate systems, i.e. for densities, where the degeneracy parameter = kB T /EF is still large compared to unity. The extension to degenerate systems is straightforward and is covered in another paper [51], but will not be treated in this work. The real part of the self-energy (effective quasi-particle shift), at the densities and temperatures considered here, was found to follow exactly the HartreeFock behaviour, i.e. h h Re (0, E0 / ) = 2 2 /2m n [8]; is the inverse Debye screening length, see (15). 4. Analytic solution for the GW (0) self-energy in Born approximation: classical limit 4.1. Derivation of the analytic solution As discussed, the spectral function in the low density limit is lacking any plasmaron resonances, only a broadened quasi-particle peak appears, see gure 3. In order to understand this behaviour, the GW (0) -equation (16) is reconsidered applying a sequence of approximations as described in the following. In this way, an analytic solution is found that is valid at low coupling parameters. It will be shown that the observed scaling is obtained correctly, if the imaginary part of the self-energy is kept nite also on the rhs of the self-energy integral equation (16). It therefore represents a generically non-perturbative result. Details of the calculations can be found in appendix B. Since collective excitations do not show up in the self-energy and the spectral function at low densities, it is obvious to neglect these features already in the screened interaction. Formally, this is achieved by replacing the complete inverse dielectric function by the Born approximation, Im

1

(q, )

(24)

For the static dielectric function appearing in the denominator, we use the Debye expression 2 2 D (q, 0) = 1 + /q , with the inverse Debye screening length. In other words, instead of the interaction via a dynamically screened potential, electronelectron collisions via a statically screened potential are considered using the Born approximation. Then, (16) turns into Im (p, + i) =

2mkB T e2 2 1 d cos d 3 4 0 1 h dq q m 2 exp exp 2 (q 2 + 2 )2 2q kB T 2kB T 0 Im (p q, + i ) . [ h pq Re (p q, )]2 + [Im (p q, + i )]2 h (25)

11

168

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

0 -0.2 -0.4 Im (p=0,) [Ry] -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 -4 -2 q=0 q finite 0 2 energy h+ [Ry] 4

Figure 5. Imaginary part of the self-energy at momentum hp = 0 for plasma parameters n = 7 1021 cm3 and T = 100 Ry/kB . The self-consistent Born approximation (nite q, dashed curve) is compared to the calculation where the momentum shift hq is neglected in the self-energy on the rhs of the self-energy equation (solid curve).

Note that the dielectric function is taken in the classical limit, i.e. the FermiDirac distribution is replaced by the Maxwell distribution, leading to the exponentials in the rst line of (25). Due to the statically screened Coulomb potential, important contributions to the q-integral stem from values q . Therefore, we neglect the shift of momentum in the self-energy (p, ). To justify this on the rhs of equation (25), i.e. we write (p q, ) approximation, we show the numerical solution for the imaginary part of the self-energy (25) in gure 5 (dashed curve). The solid curve corresponds to the solution that is obtained by neglecting the momentum shift in the argument of the self-energy on the rhs of (25). As can be seen, this approximation does not modify the result signicantly. In fact, the small deviations, which are only observable around h + 0, are already in the order of the numerical accuracy. Subsequently, the remaining terms in (25) are expanded in powers of q/, as described in detail in the appendix. Finally, the threefold integral can be performed and the equation h [Im (p, + i)]2 + [ 2 p2 /2m h + Re (p, )]2 = kB T e2 4 0 (26)

is obtained. The lhs is just the denominator of the spectral function, cf (6). Together with the spectral representation of the Green function (4), we then nd the equation [ z h2 p2 /2m + h (p, z)]1 = 4 0 e2 kB T (p, z), (27)

which, in the limit z = + i, 0+ has the solution (p, + i) = h h2 p2 /2m + sign( h2 p2 /2m + ) h 2 h + i h2 p2 /2m + 2

2

12

e2 kB T 4 0

1/2

(28)

169

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

(29)

ensures the correct sign of the imaginary part of the self-energy, i.e. Im (p, + i) < 0 for > 0.

4.2. Comparison to the numerical solution The imaginary part of (28) is plotted in gure 6 for T = 100 Ry/kB and n = 7 1021 cm3 , i.e. for the smallest density considered in gure (3). The analytic formula is compared to the full numerical solution for two different wavenumbers, p = 0 (a) and p = 1/aB (b). The dotted vertical line indicates the position of the quasi-particle dispersion Ep . In the rst case, both numerical and analytic calculation agree reasonably well, albeit the analytic solution lies systematically above the numerical data. However, the overall deviation is smaller than 7%. In the second case (p = 1/aB ), the upshifted plasmon peak, present in the numerical result, is not reproduced by the analytic formula. Thus, the analytic formula is applicable only for small momenta, while at higher momenta, the dynamical features of the interaction become important. On the other hand, the analytic formula is very useful to initialize the numerical algorithm. This is analysed in gure 7. Here, the spectral function, that is obtained in the rst iteration of the algorithm, was computed in two different ways for the same parameters as above, n = 7 1021 cm3 and kB T = 100 Ry. The dashed curve gives the rst iteration starting from the analytic formula (28) for the self-energy, the dotted curve is the same calculation but starting from a narrow Gaussian spectral function with a width of 0.3 Ry (FWHM). In plot (a) the wavenumber is p = 0, while in (b), p = 1/aB was chosen. For p = 0, the analytic ansatz leads to a good resemblance with the converged result (solid curve). The converged result is taken here as the 20 iteration starting from the Gaussian ansatz. The Gaussian ansatz, iterated once, results in a two-peak structure which is far from the converged spectral function. Also at p = 1/aB , starting from the analytic ansatz gives a much better overall correspondence than the calculation starting from a Gaussian spectral function, although subtle details like the plasmaron peak at h + 4 Ry is not reproduced in the rst iteration. In order to perform a quantiable comparison between both initializations and their impact on the convergence of the algorithm, we determine the mean squared deviation of the spectral function in a given iteration from the converged result S = N 1 N (A() (0, i ) i=1 A(20) (0, i ))2 with N the number of points on the -grid of the spectral function, i the grid points. The result is shown in gure 8 for the Gaussian ansatz (marked +) and the initialization using the analytic self-energy (marked ). During the rst four iterations, the mean squared deviation of the second method is by two orders of magnitude smaller than if using the Gaussian spectral function. While the mean squared deviation using the analytic self-energy becomes smaller than 102 already after ve iterations, it takes eight iterations for the Gaussian ansatz to get to this point. Also, it was found that a Gaussian with the width xed at the imaginary part of the effective quasi-particle self-energy, does not improve the convergence, since the special form of the self-energy and the spectral function with a broad plateau and steep edges cannot be reproduced by such an ansatz and the analytic self-energy given in (28) should be used instead.

13

170

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

0 energy h+ [Ry]

(a) p = 0

10

Im (p=1/aB,) [Ry]

(0)

0 energy h+ [Ry]

(b) p = 1/ a B

10

Figure 6. Imaginary part of the self-energy for plasma density n = 71021 cm3 and temperature T = 100 Ry/kB . Results for p = 0 (a) and for p = 1/aB (b) are shown. The self-consistent GW (0) -calculation (solid curve) is compared to the analytic formula (28) given as dashed curve. The dotted vertical line indicates the quasi-particle dispersion h = Ep .

4.3. Analytic solution at the quasi-particle dispersion In the following, the analytic solution (28) with the frequency xed at the quasi-particle h h dispersion = Ep / shall be considered in more detail. The numerical results for (p, Ep / ) at p = 0 have already been discussed in section 3, see gure 4. Since the only dependence on frequency and wavenumber is given by the trivial term h h2 p2 /2m + = h p , the discussion may be restricted to the case p = 0 and h = 0 = . Note that due to (28) Re (p, p ) = 0, therefore Ep = p . Then, the imaginary part of (28) reads e2 kB T = Im (0, / ) = h 4 0

14

e2 4 0

1/4

4 nkB T

(30)

171

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

converged result 1. iteration (starting from Gaussian) 1. iteration (starting from analytic formula)

-6

-4

-2

0

(a) p = 0

energy h+ [Ry]

converged result 1. iteration (starting from Gaussian) 1. iteration (starting from analytic formula)

-6

-4

-2

0

(b) p = 1/ a B

energy h+ [Ry]

Figure 7. Spectral function for plasma density n = 7 1021 cm3 and temperature T = 100 Ry/kB . Plot (a) shows the spectral function at p = 0, in (b) p = 1/aB was chosen. The rst iteration starting from a sharp quasi-particle spectral function (dotted curve) is compared to the rst iteration starting from the analytic expression for the self-energy (dashed curve).

which can also be given in terms of the plasma coupling parameter , Im (0, / ) = (3 h

3 1/4

kB T .

(31)

This value is the damping width neglecting the inuence of electronplasmon interaction via the dynamically screened potential. In the following, it is referred to as the non-collective damping width. The non-collective damping width depends solely on the temperature and the classical coupling parameter . Therefore, it is a purely classical result. We compare this result to the numerical solution for the GW (0) self-energy, presented in section 3. In gure 9, the effective quasi-particle self-energy at vanishing momentum h Im (0, E0 / ), in units of the plasma thermal energy, is shown as a function of . The

15

172

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

0

-5

10-6

10

11

12

# of iteration

Figure 8. Mean square deviation of the spectral function at p = 0 from the converged result as a function of the iteration number . Red symbols correspond to results obtained when starting with a sharp quasi-particle spectral function, the green symbols result from starting with the analytic self-energy and leads to relatively small deviations already in the rst iterations. High accuracy (S < 105 ) is obtained in both schemes only after about ten iterations. Plasma parameters: n = 7 1021 cm3 , T = 100 Ry/kB .

3/4

10

-1

=1 = 0.5

= 0.5

=1

10

-2

~

-3

3/4

=1 = 0.5

-2 -1

10

10

10

Figure 9. Effective quasi-particle damping width at p = 0, Im (0, E0 / ) normalized to the h thermal energy as a function of the plasma coupling parameter . The solid black line marks the derived scaling law following (30). The vertical lines mark for each temperature the onset of plasmon excitation and corresponding decrease of quasi-particle damping, i.e. the parameter = aB > 1.

numerical results perfectly agree with the derived scaling law in the low density limit. As long as the density is small, such that the plasma frequency is below the non-collective damping width, i.e. hpl 1 (32) = 2 k T /4 e B 0

16

173

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

the damping is mainly non-collective, and a single, broadened resonance appears in the spectral function, cf gure 3. The parameter can also be expressed through the plasma coupling parameter and the degeneracy parameter , or through the inverse Debye screening length , = 27 5/2 2 3 = aB .

1/6 1/4 1/2

0.9698

1/4 1/2

(33) (34)

Thus, the derived scaling law is only valid for large , i.e. classical systems. This was already shown in gure 4. When approaches 1, quantum effects set in. For example, collisions become less probable due to Pauli blocking which leads also to a decrease of the self-energy. Obviously, the Bohr radius aB sets the relevant length-scale that is to be compared to the inverse screening length in order to estimate the importance of non-collective damping. Non-collective damping is the dominant mechanism, as long as the screening length is large compared to the Bohr radius. When the screening length becomes smaller than the Bohr radius, i.e. > 1, which, due to (32), is equivalent to having the energy of plasma oscillations larger than the non-collective damping width, the plasmaron satellites begin to separate from the broadened quasi-particle peak. Spectral weight is transferred from the wings of the central peak into the plasmaron satellites leading to a more dened quasi-particle resonance, i.e. a decreased damping of the central peak, see gure 3. Concluding, the analytic result is only a good approximation at low densities, when 1. At higher densities, dynamical screening becomes important, leading to satellites in the spectral function. 5. Deciencies of the quasi-particle approximation The non-collective damping width was introduced above in (30) as the value of the imaginary part of the self-energy at vanishing momentum and frequency, p = 0, + / = 0. Of h course, the same result is also obtained if this choice of variables was already made at the very beginning of the calculations leading to (28). However, in the latter case, the manipulations can be performed in a different manner. At an intermediate step of the calculation, one h can identify the reason why the quasi-particle damping Im (p, Ep / ) as given in [8, 39], behaves unphysical in the low density and classical limits. Detailed calculations are given in appendix C, while here only the most important steps are summarized. Setting p = 0 and h + = 0 in (16), neglecting the momentum shift in the argument of the self-energy on the rhs and replacing the dynamically screened potential by the statically screened Born approximation as before, we obtain Im (0, / ) = h with h h2 q 2 + i2m Im (0, / ) . (36) 2 q 2mkB T h Most contributions to the integral stem from small values of the wavevector, q . Therefore, we may neglect the real part of z and write z = i 2m Im (0, / )/2 q kB T . h h Using the expansion 1 1 (37) 3 + O(x 5 ) lim exp(x 2 ) erfc(x) = x+ 2 x x z=

17

e2 2 0h

mkB T 2

dq q Re[exp(z2 ) erfc(iz)], (q 2 + 2 )2

(35)

174

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

in lowest order only, the q-integral can be performed, resulting in Im (0, / ) = e2 kB T /4 0 , h (38)

which coincides with (28) at p = 0 and h = . From (35), one can also derive the quasi-particle approximation for the imaginary part of the self-energy: if the imaginary part of z, i.e. the self-energy, is neglected on the rhs (this is just the quasi-particle approximation), and furthermore the limit q 0 is considered, the expression Im

QP

(0, / ) = h =

e2 2 0h e2 4 0h

mkB T 2

(q 2

dq q + 2 )2 (39)

2mkB T ,

is obtained. This coincides with the formula for the imaginary part of the quasi-particle self-energy as given in [8, page 114, equation (4.164)]. There, the spectral function on the rhs of the integral equation for (p, ) is replaced by an on-shell delta distribution (free particle spectral function), i.e. the self-energy is set to 0 on the rhs. The resulting integral is evaluated at the free particle dispersion h = p . As a result, one obtains the expression Im (p, p / ) = h e2 4 0h 2mkB T 1 F1 (1, 3/2; p /2kB T ), (40)

with 1 F1 (, ; z) being the conuent hypergeometric function [52]. Note that in the given reference, instead of the imaginary part of the self-energy, the quasi-particle damping h h (p, p / ) = 2Im (p, p / ) is given. Also, the original formula differs from (40) by a factor of 1/4. However, the formula given here was approved through private communication by W-D Kraeft. Obviously, (40) is independent of density. The neglect of Im (0, / ) in the complex h variable z leads to a different analytical structure of the equation. Therefore, the quasi-particle approximation has no chance to ever obtain the correct behaviour at low densities. Low densities, and therefore small inverse screening lengths shift the supporter of the q-integral to small q, where contributions from Im z are important, whereas the real part of z vanishes at q = 0 and leads to a result which is independent of . In the same way, one can understand why the quasi-particle limit diverges when considering the classical limit h 0. The imaginary part of z has h in the denominator which after the integration cancels the h in the prefactor in equation (35). No cancellation takes place, if the imaginary part is neglected, i.e. in the quasi-particle approximation. This leads to the divergence of the nal result.

6. Conclusion In this work, the single-particle self-energy of the one-component electron plasma was investigated. The spectral function was calculated self-consistently using the GW (0) approximation which allows the systematic treatment of dynamical correlations in the plasma. The spectral function contains at small momenta a broadened quasi-particle

18

175

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

peak and two plasmaron satellites which, at low densities, merge into the central quasiparticle resonance. At increased momenta, for a given density and temperature, the spectral function converges to a single, sharp quasi-particle resonance. Special attention was paid to a systematic investigation of the self-energy and the spectral at different densities and temperatures. Here, only non-degenerate plasmas were considered, i.e. the temperature is large compared to the Fermi temperature. Also, bound states were neglected. It was found that at low densities, the imaginary part of the on-shell self-energy, i.e. h the inverse single particle lifetime, follows a universal scaling law Im (p, Ep / ) n1/4 . For the rst time, an analytic result for the on-shell single-particle self-energy was found that contains the correct low-density limit, i.e. a vanishing self-energy at n = 0. This is a major progress compared to the well-known quasi-particle approximation that yields a nite damping width even at zero density. The new on-shell single-particle damping h width is Im (p, Ep / ) = (3 3 )1/4 kB T . Since it is derived in Born approximation, i.e. no collective excitations contribute to the damping mechanisms, this quantity is called the non-collective damping width. By comparison of the numerical results to the new analytic formula, the parameter = aB was identied to separate the regime of non-collective damping ( 1) from the regime, where the coupling between single-particle states and collective excitations dominantly determine the single-particle damping ( 1) at small momenta. This analysis complements earlier work on the electron spectral function based on the plasmon-pole approximation. For 1, the analytic formula (28) is a good approximation for the self-energy. Furthermore, the use of the analytic formula for the self-energy as an initialization of the iterative algorithm leads to signicantly faster convergence as compared to other methods, where a Gaussian ansatz is used as the initial spectral function. The non-collective damping is a purely classical result, no powers of h appear. This is fundamentally different from the quasi-particle approximation to the imaginary part of the selfenergy which has no classical limit, i.e. the self-energy diverges in the limit h 0. It could be shown that this problem, as well as the paradox of being density independent, stem from the inherently inconsistent treatment of the self-energy in the quasi-particle approximation. The long-time open question of the classical limit of the single-particle self-energy can now be regarded as settled. The results reported in this work are of paramount importance for many-particle theory and applications to dense plasmas. In particular, simple analytic expressions for the single-particle spectral function and self-energy in the classical and in the degenerate limit are needed to construct Pad -like interpolation formulae that cover the complete densitytemperature plane. e Such expressions would greatly simplify the calculation of equation of state, transport and optical properties of dense, high energy plasmas, solid state devices but also nuclear, hadronic and partonic matter, and provide benchmarks for numerical approaches, i.e. simulation techniques. One part of this task, the analytic formula for non-degenerate dilute plasmas has been accomplished in this work. Acknowledgments This work was performed with nancial support from the German Research Society (DFG) under grant SFB 652 (Collaborative Research Center Strong Correlations and Collective Phenomena in Radiation Fields: Coulomb Systems, Clusters, and Particles). Special thanks go to G R pke and A Wierling for many suggestions and comments on the paper and to W-D o Kraeft for stimulating discussions.

19

176

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

Appendix A. Details on the GW (0) -approximation Throughout the appendix, the Rydberg system of units will be applied to keep the formulae short and readable. In these units h = kB = 1, e2 = 2, 0 = 1/4 and m = 1/2. We start from the representation of the self-energy in terms of the full Green function G(p, z ), the dynamically screened potential W (q, ) and the vertex function (p, p + q; z , z + ), given by the diagram

W

(A.1)

(p, z ) =

(0) G

.

(0)

In the GW -approximation, the vertex is replaced by the bare vertex of the considered particles, electrons in this case,

W

(A.2)

(p, z ) =

(0) G (0)

= T

q,

G(p q, z )W (q, ),

(A.3)

which is equation (11). The dynamically screened interaction is taken in the random phase approximation [46], W (0) (q, ) =

RPA (q,

RPA (q,

+ i) = 1 V (q) + i) =

k

RPA (q,

Using the spectral representations of both the Green function (4) and the screened interaction in RPA, W (0) (q, z) = V (q) 1 + leads to (p, z ) = T

q,

d Im

1 RPA (q,

+ i)

(A.7)

V (q)

d A(p q, ) 2 z

1+

1 d Im RPA (q, ) ,

(A.8)

q

V (q)

d A(p q, ) 2

1 nF ( ) +

20

d Im

1 RPA (q,

) nB ( ) + 1 nF ( ) z

(A.9)

177

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

is obtained. This expression contains the HartreeFock self-energy of the interacting system,

HF int (p)

=

q

(A.10)

corr

(p, z ) =

q

V (q)

d A(p q, ) 2

1 RPA (q,

d Im

)[nB ( ) + 1 nF ( )] . z

(A.11)

For convenience, we skip the upper index corr in the following and only distinguish between HF the frequency-dependent self-energy (p, + i) and the HartreeFock term int (p), in the following. After analytic continuation z z = + i, 0, the imaginary part of the correlated self-energy is evaluated using Diracs identity lim0 1/(x i) = P 1/x i (x), Im (p, + i0+ ) = Im 1 nF ()

q

1 RPA (q,

)nB ( )nF ( ),

where the exact relation nB ( ) + 1 nF ( ) = nB ( )nF ( )/nF () was used. This equation is given as (16) in the main text. Appendix B. Analytic self-energy for the classical one-component plasma In the high temperature limit kB T EF , we replace the FermiDirac distributions in the self-energy equation (16) as well as in the dielectric function (14) by the MaxwellBoltzmann 3 distribution, nF (k ) f (k) = n 2 exp(k /T ) with the thermal de-Broglie wavelength = (4/T )1/2 . In this approximation, the polarization function takes the form [39], Re

RPA (q, ) =

n 2qT +q q

q q

1 F1

1 F1 1, 3/2,

q 2q T 2 T

2

1, 3/2,

q + 2q T 2 T

2

(B.1)

2

Im

RPA (q, ) =

q Tn 3 exp + 8 q 2q T 2 T Tn exp 8 q 2T

3

exp

q 2q T 2 T

(B.2) = exp q + 2T 4q 4T

2 2

1 . nB ()

2 2

(B.3)

q 1 exp 4q 2 T + 4T d cos dq d 3 q | (q, )|2 1 0 Im (p q, + i ) 1

exp

2T . (B.4)

21

178

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

Furthermore, the Born approximation is applied, i.e. the dielectric function in the denominator is replaced by the Debye expression, D (q, 0) = 1 + 2 /q 2 . Diagrammatically, the self-energy in this approximation is written as

RPA

(p, z ) =

VD G

VD

(B.5)

VD denotes the Debye potential VD (q) = e2 / 0 (q 2 + 2 ) 0 . Since the main contribution to the q-integral stems from momenta q < , we neglect the transfer wavenumber q in the argument of the self-energy on the rhs and write Im (p, + i) = 2 2 T 3/2

1

d cos

1

2

dq

0

d 2T . (B.6)

1 exp 4q 2 T + 3 2 2 q 1 + 2 q

q2 4T

exp

Furthermore, we neglect the term q 2 /4T in the exponential which is small for high temperatures and for q < , Im (p, + i) = 2 2 T 3/2

1

d cos

1 0

dq

d 2T (B.7)

q 2 exp 2 [q 2 + 2 ]2 4q T

exp

This equation was given in section 4 as (25). Now, the integration over the angle can be performed as Im (p, + i) = 2 2 T 3/2

2 1

d cos

1 0

dq

[q 2

q + 2 ]2

exp +

4q 2 T

exp

2T

Im (p, + i ) 4p2 q 2

2

2 1

(B.8) 1 2 exp 2 [q 2 + 2 ]2 4q T 2T

=

22

2 T 3/2 p

dq

0

exp

179

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

arctan arctan

(p q)2 + Re (p, ) , (B.9) Im (p, + i ) where the integral dx/[(a + x)2 + b2 ] = b1 arctan((a + x)/b) was used. In the limit q 0 the identity 1 2 2 lim e /4q T = ( ), (B.10) q0 2q T allows us to perform the frequency integration, 2 2q T exp 4q 2 T 2 T dq 2 d exp Im (p, + i) = 3/2 p [q + 2 ]2 2T 2q T 0 (p + q)2 + Re (p, ) arctan Im (p, + i ) 2 (p q) + Re (p, ) arctan Im (p, + i ) 2 2 T q = 3/2 dq 2 d exp ( ) 2 ]2 p 0 [q + 2T (p + q)2 + Re (p, ) arctan Im (p, + i ) 2 (p q) + Re (p, ) arctan Im (p, + i ) 2 2 T q = dq 2 p 0 [q + 2 ]2 (p + q)2 + Re (p, ) arctan Im (p, + i) 2 (p q) + Re (p, ) . (B.11) arctan Im (p, + i) Using the following power expansion of the arctan-function x arctan(1 + x) = + + O(x 3 ), (B.12) 4 2 i.e. (p q)2 + Re (p, ) (p + q)2 + Re (p, ) arctan arctan Im (p, + i) Im (p, + i) = 4pq Im (p, + i) 1 + p2 + Re (p, ) Im (p, + i)

2 1

+ O(q 3 ), (B.13)

we obtain Im (p, + i) = 2 2 T p

0

dq

[q 2

q + 2 ]2 p2 + Re (p, ) Im (p, + i)

2 1

4pq Im (p, + i) 1 +

, (B.14)

23

180

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

which can be turned into [Im (p, + i)]2 + [p2 + Re (p, )]2 = = 2T . 8 2 T

dq

0

(q 2

q2 + 2 )2 (B.15)

To solve this single equation for the two unknown Re (p, ) and Im (p, + i), we make use of the spectral representation of the Green function d A(p, ) G(p, z) = [z p2 (p, z)]1 = 2 z =

1 d Im (p, + i) (Im (p, + i))2 + ( + p2 Re (p, ))2 z 2 (B.16)

(p, z) d Im (p, + i) 1 = . (B.17) 2 z 2T 2T In the last step we also used the spectral representation of the correlated self-energy. The last equation has the solution = (p, z) = z p2 + 2 z p2 + 2

2 1/2

2T

(B.18)

With z = + i and having in mind that Im (p, + i) < 0 for > 0, we nally nd (p, + i) = + p2 sign( + p2 ) 2 + + i p2 2

2 1/2

2T

, (B.19)

i.e. equation (28). Appendix C. Details for the quasi-particle self-energy We start from (25) for the imaginary part of the self-energy using the Born approximation for the screened interaction potential: 2 1 T q Im (p, + i) = 3/2 d cos dq d 2 (q + 2 )2 1 0 2 exp 2 . (C.1) A(p q, ) exp 2T 4q T By assuming a frequency and momentum independent self-energy (p, ) (0, E0 / ), this becomes h q 2 T 2 1 Im (0, E0 / ) = h d cos dq d 2 + 2 2 3/2 1 0 q 2 h Im (0, E0 / ) exp 2 . (C.2) exp 2T 4q T [ + q 2 ]2 + [Im (0, E0 / )]2 h Since the self-energy is assumed to be independent of the frequency, the real part of the correlated self-energy vanishes exactly. For the HartreeFock part of the self-energy is proportional to n 3 in the classical limit [8], we also neglect this term, since it gives contributions of higher order in n, whereas we are only interested in the lowest order.

24

181

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

After eliminating Im (0, E0 / ) on both sides, performing the trivial integration over the h angle , which yields a factor 2, the frequency integration is performed by the help of [52]

0

0

(C.3)

leading to 1=4 T 2 3

0

dq q 2 + 2 )2 (q T 2 3

(C.4)

(C.5)

T 2

dq q Re[exp(z2 ) erfc(iz)], (q 2 + 2 )2

(C.6)

h q 2 + i Im (0, E0 / ) . 2q T

(C.7)

Equation (C.6) is given as (35) in section 5. It should be noted at this point that the integral converges only for nite , i.e. the Coulomb limit 0 does not yield a nite result. Most contributions to the integral stem from small values of the wavevector, q . h Therefore, we may neglect the real part of z and write z = i Im (0, E0 / )/2q T . Using the expansion

x+

(C.8)

0

(q 2

dq q + 2 )2

(C.9)

0

(q 2

dq q + 2 )2

(C.10)

0

0

(C.11)

8T 2 Im (0, E0 / ) h

dq q 2 8T 2 . = 2 + 2 )2 (q Im (0, E0 / ) 4 h

(C.12)

25

182

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

Im (0, E0 / ) = 2T , h

(C.13)

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] 26 Chabrier G, Saumon D and Potekhin A Y 2006 J. Phys. A: Math. Gen. 39 44119 Remington B A, Drake R P and Ryutov D D 2006 Rev. Mod. Phys. 78 755 Lindl J 1995 Phys. Plasmas 2 39334024 Hoffmann D H H et al 2005 Laser Part. Beams 23 47 Zastrau U et al 2008 (at press) Fetter A L and Walecka J D 1971 Quantum Theory of Many-Particle Systems (New York: McGraw-Hill) Mahan G D 1981 Many-Particle Physics 2nd edn (New York: Plenum) Kraeft W D, Kremp D, Ebeling W and R pke G 1986 Statistics of Charged Particle Systems (Berlin: Akademieo Verlag) Domitrovich P P and M ther H 1994 J. Phys. G: Nucl. Part. Phys. 20 1885900 u H ll A, Roberts C D and Wright S V 2006 Particles and Fields: X Mexican Workshop (American Institute of o Physics Conference Series vol 857) ed M A P rez, L Urrutia and L Villaseqor pp 4661 e Zubarev D, Morozov V and R pke G 1996 Statistical Mechanics of Nonequilibrium Processes vol 1 (Berlin: o Akademie Verlag) Kaminski A and Fretwell H M 2005 New J. Phys. 7 98 Ding Z J 1998 J. Phys.: Condens. Matter 10 173351 Ding Z J 1998 J. Phys.: Condens. Matter 10 175365 Vorberger J, Schlanges M and Kraeft W D 2004 Phys. Rev. E 69 046407 Kremp D, Schlanges M and Kraeft W D 2005 Quantum Statistics of Nonideal Plasmas (Berlin: Springer) Zwicknagel G, T pfer C and Reinhard P G 1999 Phys. Rep. 309 117208 o Fortmann C, R pke G and Wierling A 2007 Contrib. Plasma Phys. 47 297 o Ebeling W, Kraeft W D and Kremp D 1972 Theory of Bound States and Ionization Equilibrium in Plasmas and Solids (Berlin: Akademie Verlag) Seidel J, Arndt S and Kraeft W D 1995 Phys. Rev. E 52 5387 Aryasetiawan F and Gunnarsson O 1998 Rep. Prog. Phys. 61 237312 Mahan G D 1994 Comments Condens. Matter Phys. 16 333 Hedin L 1965 Phys. Rev. 139 796823 von Barth U and Holm B 1996 Phys. Rev. B 54 841119 Holm B and von Barth U 1998 Phys. Rev. B 57 210817 Northrup J E, Hybertsen M S and Louie S G 1987 Phys. Rev. Lett. 59 81922 Godby R W, Schl ter M and Sham L J 1988 Phys. Rev. B 37 1015975 u Akkus H and Mamedov A M 2007 J. Phys.: Condens. Matter 19 116207 Leb` gue S, Arnaud B, Rabiller P, Alouani M and Pickett W E 2004 Eur. Phys. Lett. 68 84652 e Stan A, Dahlen N E and van Leeuwen R 2006 Eur. Phys. Lett. 76 298304 Kheifets A S 1995 J. Phys. B: At. Mol. Opt. Phys. 28 3791803 Yerokhin V A, Artemyev A N, Beier T, Shabaev V M and Soff G 1999 Phys. Scr. T 80B 4957 Onida G, Reining L and Rubio A 2002 Rev. Mod. Phys. 74 60159 Faleev S V, van Schilfgaarde M, Kotani T, L onard F and Desjarlais M P 2006 Phys. Rev. B 74 033101 e Rinke P, Qteish A, Neugebauer J, Freysoldt C and Schefer M 2005 New J. Phys. 7 126 Fehr R and Kraeft W D 1995 Contrib. Plasma Phys. 35 46379 Wierling A and R pke G 1998 Contrib. Plasma Phys. 38 513 o Schepe R, Schmielau T, Tamme D and Henneberger K 1998 Phys. Stat. Solidi (b) 206 2739 Fennel W and Wilfer H P 1974 Ann. Phys., Lpz 32 26576 Lundquist B I 1967 Phys. Kondens. Mater. 6 206 Glasser M L and Lamb G 2007 J. Phys. A: Math. Theor. 40 12158 Ziesche P 2007 Phys. Status Solidi b 244 202236 Ziesche P 2007 Ann. Phys., Lpz 16 45 Ward J 1950 Phys. Rev. 78 182 Tamme D, Schepe R and Henneberger K 1999 Phys. Rev. Lett. 83 241 Arista N R and Brandt W 1984 Phys. Rev. A 29 147180

183

J. Phys. A: Math. Theor. 41 (2008) 445501 C Fortmann

[47] Bohm D and Gross E P 1949 Phys. Rev. 75 185164 [48] Thiele R, Bornath T, Fortmann C, H ll A, Redmer R, Reinholz H, R pke G, Wierling A, Glenzer S H and o o Gregori G 2008 Phys. Rev. E 78 026411 [49] Bahcall J N, Pinsonneault M H and Wasserburg G J 1995 Rev. Mod. Phys. 67 781808 [50] Fortmann C, R pke G and Wierling A 2007 Pulsed Power Conference (PPPS-2007) Digest of Technical Papers o (IEEE) ed E Schamiloglu and F Peterkin p 194 [51] Fortmann C, R pke G and Wierling A (in preparation) o [52] Abramowitz M and Stegun A (ed) 1970 Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables 9th edn (New York: Dover)

27

184

Author: Carsten Fortmann Appeared as regular article in Physical Review E, Vol. 79, Issue 1, pages 016404 1-11, Jan. 2009.

186

187

PHYSICAL REVIEW E 79, 016404 2009

C. Fortmann*

Institut fr Physik, Universitt Rostock, 18051 Rostock, Germany Received 15 August 2008; published 8 January 2009 The spectral function for an electron one-component plasma is calculated self-consistently using the GW 0 approximation for the single-particle self-energy. In this way, correlation effects that go beyond the mean-eld description of the plasma are contained, i.e., the collisional damping of single-particle states, the dynamical screening of the interaction, and the appearance of collective plasma modes. Second, a nonperturbative analytic solution for the on-shell GW 0 self-energy as a function of momentum is presented. It reproduces the numerical data for the spectral function with a relative error of less than 10% in the regime where the Debye screening 1a1. In the limit of low density, the nonperturbative parameter is smaller than the inverse Bohr radius, B 1/4 self-energy behaves as n , whereas a perturbation expansion leads to the unphysical result of a densityindependent self-energy Fennel and Wilfer, Ann. Phys. Leipzig 32, 265 1974 . The derived expression will greatly facilitate the calculation of observables in correlated plasmas transport properties, equation of state that need the spectral function as an input quantity. This is demonstrated for the shift of the chemical potential, which is computed from the analytical formulas and compared to the GW 0 result. At a plasma temperature of 100 eV and densities below 1021 cm3, the two approaches deviate by less than 10% from each other. DOI: 10.1103/PhysRevE.79.016404 PACS number s : 52.27.Aj, 52.65.Vv, 71.15. m

I. INTRODUCTION

The many-particle Green function approach 1 allows for a systematic study of macroscopic properties of correlated systems. Green functions have a long history of applications in solid state theory 2 , nuclear 3 , and hadron physics 4 , and also in the theory of strongly coupled plasmas 5 . In the last case, optical and dielectric properties 6,7 have been studied using the Green function approach, as well as transport properties like conductivity 8 and stopping power 9,10 , and the equation of state 11 . Modications of these quantities due to the interaction among the constituents can be accessed, starting from a common starting point, namely, the Hamiltonian of the system. The key quantity for electronic properties in a correlated many-body environment is the electron spectral function A p , , i.e., the probability density to nd an electron at for a given momentum p. It is related energy frequency 0, via to the retarded electron self-energy p, +i , Dysons equation A p, = 2 Im Re p, p

2

p, + i + Im

p, + i

2.

p

= p2

has been introduced, where e is the electron chemical potential. Note that here and throughout the paper the Rydberg system of units is used, where = 1, me = 1 / 2, and e2 / 4 0 = 2. Furthermore, the Boltzmann constant kB is set equal to 1, i.e., temperatures are measured in units of energy.

1539-3755/2009/79 1 /016404 11

The self-energy describes the inuence of correlations on the behavior of the electrons. A nite, frequency-dependent self-energy leads to a nite lifetime of single-particle states and a modication of the single-particle dispersion relation. Hence, the calculation of the electron self-energy is the central task if one wants to determine electronic properties, e.g., those mentioned above. The Hartree-Fock approximation 12 represents the lowest order in a perturbative expansion of the self-energy in terms of the interaction potential 13 . Because it is a meaneld approximation, effects due to correlations in the system cannot be described. Examples are the appearance of collective modes, the energy transfer during particle collisions, and the quasiparticle damping. The next-order term is the Born approximation, where binary collisions are taken into account via a bare Coulomb potential. However, the Born approximation leads to a divergent integral, due to the longrange Coulomb interaction. Therefore, the perturbation expansion of the self-energy has to be replaced by a nonperturbative approach, accounting for the dynamical screening of the interparticle potential. A nonperturbative approach to the many-particle problem is given by the theory of Dyson 14 and Schwinger 15,16 generalized to nite temperature and nite density 17 . An excellent introduction to the Dyson-Schwinger equations can also be found in 4 . The Dyson-Schwinger equation for the self-energy contains the full Green function G, the screened interaction W, and the proper vertex . Since each of these functions obeys a different Dyson-Schwinger equation itself, involving higher-order correlation functions, the Dyson-Schwinger approach leads to a hierarchy of coupled integro-differential equations. In order to provide soluble equations, this hierarchy has to be closed at some level, i.e., correlation functions of a certain order have to be either parametrized or neglected. One such closure of the Dyson-Schwinger hierarchy consists in neglecting the vertex, i.e., the three-point function, and considering only the particle propagators and their re2009 The American Physical Society

016404-1

188

C. FORTMANN

PHYSICAL REVIEW E 79, 016404 2009

spective self-energies, i.e., two-point functions. One arrives at the so-called GW approximation, introduced in solid-state physics by Hedin 18 . Hedin was led by the idea of including correlations in the self-energy by replacing the Coulomb potential in the Hartree-Fock self-energy by the dynamically screened interaction W. In this way, one obtains a selfconsistent, closed set of equations for the self-energy, the polarization function , the Green function, and the screened interaction. It can be shown 19 that the GW approximation is one of the so-called -derivable approximations 20,21 . As such, it leads to energy-, momentum-, and particle-numberconserving expressions for higher-order correlation functions. It has been successfully applied in virtually all branches of solid state physics. An overview of theoretical foundations and applications of the GW approximation can be found in the review articles 2224 . The drawback of the GW approximation is that the WardTakahashi identities are violated. The Ward-Takahashi identities provide an exact relation between the vertex function , i.e., the effective electron-photon coupling in the medium, and the self-energy, and follow from the Dyson-Schwinger equations. They reect the gauge invariance of the theory. In GW theory, they are violated simply because corrections to the vertex beyond zero order are neglected altogether. This issue touches on a fundamental problem in many-body theory and eld theory, namely, the question of how to preserve gauge invariance in an effective, i.e., approximate, theory, without violating basic conservation laws. A detailed analysis of this question with application to nuclear physics is presented in a series of papers by van Hees and Knoll 2527 . Approximations for the self-energy that also contain the vertex are often referred to as GW approximations. An example can be found in Ref. 28 , where the spectral function of electrons in aluminum is calculated using a parametrized vertex function. An interesting result obtained in that work is that vertex corrections and self-energy corrections entering the polarization function largely cancel. This can be understood as a consequence of the Ward-Takahashi identities. Thus, and in order to reduce the numerical cost, it is a sensible choice to neglect vertex corrections altogether, and to keep the polarization function on the lowest level, i.e. the random phase approximation RPA , which is the convolution product of two noninteracting Green functions in frequency-momentum space. The corresponding self-energy is named the GW 0 self-energy and has been introduced by von Barth and Holm 29 , who were also the rst to study the fully self-consistent GW approximation 30 . Throughout this work, the GW 0 self-energy will be analyzed. Having been used in solid state physics traditionally, the GW 0 method was recently also applied to study correlations in hot and dense plasmas. The equation of state 31,32 , as well as optical properties of electron-hole plasmas in highly excited semiconductors 33 and dense hydrogen plasmas 7 were investigated. In general, the calculation of such macroscopic observables of many-particle system involves numerical operations that need the spectral function as an input. Since the selfconsistent calculation of the self-energy, even in the GW 0

approximation, is itself already a numerically demanding task, it is worth looking for an analytic solution of the GW 0 equations which reproduces the numerical solution at least in a certain range of plasma parameters. Such an analytic expression then also allows study of the self-energy in various limiting cases, such as the low-density limit or the limit of high momenta, which are difcult to access in the numerical treatment. Furthermore, an analytic expression that is already close to the numerical solution permits the calculation of the full GW 0 self-energy using only few iterations. Analytical expressions for the single-particle self-energy have already been given by other authors, e.g., Fennel and Wilfer 34 and Kraeft et al. 12 . They calculated the selfenergy in rst order of the perturbation expansion with respect to the dynamically screened potential. Besides being far from the converged GW 0 self-energy, their result is independent of density, i.e., the single-particle lifetime is nite even in vacuum. As shown in 35 , this unphysical behavior is a direct consequence of the perturbative treatment. By using a nonperturbative ansatz, an expression for the selfconsistent self-energy in a classical one-component plasma was presented that reproduces the full GW 0 self-energy at small momenta, i.e., for slow particles. The behavior of the quasiparticle damping at larger momenta remained open and will be investigated in the present work. Second, based on the information gathered about the low- and high-momentum behavior, an interpolation formula will be derived that gives the quasiparticle damping at arbitrary momenta. The work is organized in the following way. After a brief outline of the GW 0 approximation in the next section, numerical results will be given in Sec. III for the single-particle spectral function for various sets of parameters, electron density n, and electron temperature T. In Sec. IV the analytic expression for the quasiparticle damping width is presented and comparison to the numerical results is given. Section V deals with the application of the derived formulas to the calculation of the chemical potential as a function of density and temperature. An Appendix contains the detailed derivation of the analytic self-energy. As a model system, we focus on the electron one-component plasma; ions are treated as a homogeneously distributed background of positive charges jellium model .

We start our discussion with the integral equation for the imaginary part of the single-particle self-energy in the GW 0 approximation: 1 nF Im d V q A p q, 2 nB nF . 3

Im

p, + i

q 1 RPA

q,

V q = 8 / q2 0 is the Fourier transform of the Coulomb potential with the normalization volume 0. It is multiplied by the inverse dielectric function in the RPA,

016404-2

189

SINGLE-PARTICLE SPECTRAL FUNCTION FOR THE PHYSICAL REVIEW E 79, 016404 2009

100 1 10-1 10-2 10-3 10-4 0 0 1 2 3 4 5 6 7 momentum p [a -1]

B

RPA

q,

=1V q

p

nF

nF + pq/2

pq/2

p+q/2 p+q/2

4

80 frequency +e [Ry] 60 40 20

to account for dynamical screening of the interaction. Furthermore, the Fermi-Dirac and the Bose-Einstein distribution functions, nF/B = exp / kBT 1 1 were introduced. Note that the dielectric function is determined only once, at the beginning of the calculation. In particular, the singleparticle energies p = p2 e entering Eq. 4 are determined from the noninteracting chemical potential, whereas during the course of the self-consistent calculation of the selfenergy, the chemical potential is recalculated at each step via inversion of the density relation n

e,T

10

10-5

=2

p

d A p, 2

nF

5

100 80 frequency +e [Ry] 60 40 20 0 10-1 10-2 10-3 10-4 1 2 3 4 5 6 7 momentum p [a -1]

B

Using the self-consistent chemical potential also in the RPA polarization function leads to violation of the f-sum rule, i.e., conservation of the number of particles. The real part of the self-energy is obtained via the Kramers-Kronig relation 2 . All quantities spectral function, self-energy, and chemical potential have to be determined in a self-consistent way. This is usually achieved by solving Eqs. 1 3 iteratively. The numerical algorithm is discussed in detail in Ref. 35 .

III. NUMERICAL RESULTS

-20 -40 0 8 9 10

10-5

The spectral function was calculated for the case of a hot one-component electron plasma. Temperature and density were chosen such that the plasma degeneracy parameter = T EF 6

is always larger than 1, i.e., the plasma is nondegenerate. Furthermore, the temperature is xed above the ionization energy of hydrogen, T 1 Ry, such that bound states can be neglected in the calculations. At lower temperatures, bound states have to be included, e.g., via the t matrix. For an application in electron-hole plasmas, see Ref. 33 . The electron density is adjusted such that the plasma coupling parameter = 2 4 n T 3

1/3

FIG. 1. Color online Contour plots of the spectral function as a function of momentum and frequency. The color scale is logarithmic. Results are shown for two different densities n = 7 1021 upper graph and 7 1025 cm3 lower graph . For these parameters, the plasma coupling parameter is = 4.4 103 and 9.6 102, respectively. The degeneracy parameter is = 7.5 102 and 1.6, respectively. The black line indicates the free particle dispersion = p.

which gives the mean Coulomb interaction energy compared to the thermal energy, is smaller than 1 in all calculations, i.e., we are in the limit of weak coupling. In Fig. 1, we show contour plots of the spectral function in frequency and momentum space for two different densities n = 7 1021 cm3 upper graph and 7 1025 cm3 lower graph . The temperature is set to T = 1000 eV in both calculations. The free particle dispersion = p is shown as a solid black line. In the rst case, the plasma is classical = 7.5 102 and weakly coupled = 4.4 103 . The spectral function is symmetrically broadened and the maximum is found at the

free dispersion, i.e., there is no notable quasiparticle shift in the present conditions. With increasing momentum, the width of the spectral function decreases, and the maximum value increases; the norm is preserved. The situation changes when we go to higher densities see the lower graph in Fig. 1 . The chosen parameters are typical solar core parameters 36 . The degeneracy parameter is now = 1.6 and the coupling parameter is = 0.096. The increased degeneracy and coupling result in a signicant modication of the spectral function as compared to the low-density case: A shift of the spectral functions maximum to smaller frequencies is observed, the Hartree-Fock shift. The shift due to dynamic correlations is still small in the present conditions; it becomes important in strongly degenerate systems 29 . Furthermore, shoulders appearing in the wings of the main quasiparticle peak at small momenta indicate the excitation of new quasiparticles, so-called plasmarons 37 . They can be seen more clearly in Fig. 7 solid curve . The plasmaron satellites are separated from the main peak by roughly the plasma frequency pl = 4 n, which is about 23 Ry at the present density. In the former case of lower density no plas-

016404-3

A(p,) [1/Ry]

A(p,) [1/Ry]

190

C. FORTMANN

14 12 spectral function A(p,) [1/Ry] GW Gauss fit

(0)

PHYSICAL REVIEW E 79, 016404 2009

14 12 spectral function A(p,) [1/Ry] 10 8 6 4 2 0 2497 p=50 aB

-1

GW Gauss fit

(0)

GW Gauss fit

(0)

10 8 6 4 2 0 p=0 aB

-1

(a)

-3

-2

-1

0 1 frequency +e [Ry]

(b)

2498

2502

2503

(c)

9998

10002

10003

FIG. 2. Spectral function in GW 0 -approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1019 cm3, T = 100 eV. The B B plasma coupling parameter is = 1.0 102, the degeneracy parameter is = 1.6 103, and the Debye screening parameter is = 6.0 103a1. B

marons appear; only a featureless, single resonance is obtained. At higher momenta, the plasmarons merge into the central peak. As in the low-density case, the position of the maximum approaches the single-particle dispersion, due to the decreasing Hartree-Fock shift at high p. Again, the width of the spectral function decreases with increasing momentum. This is visible more clearly in Figs. 27. Here, the solid curves represent the GW 0 spectral function as a function of frequency. Results are shown for three different momenta, 1 1 1 p = 0aB a , 50aB b , and 100aB c . Two different temperatures are considered, T = 100 Figs. 24 and 1000 eV Figs. 57 , and for each temperature three different densities are studied. With increasing momentum p, the spectral function becomes more and more narrow, converging eventually into a narrow on-shell resonance, located at the unperturbed single-particle dispersion + e = p2. As a general feature, one can observe an increase of the spectral functions width with increasing density and with increasing temperature. The increase with density is due to the increased coupling, while the increase with temperature reects the thermal broadening of the momentum distributhat enters the self-energy and thereby tion function nF also the spectral function. From these results, we see that the spectral function has a quite simple form in the limit of low coupling, i.e., at low densities and high temperatures. The numerical results are compared to a Gaussian ansatz for the spectral function, shown as the dashed curve in Figs. 27. The explicit form of the Gaussian is given as Eq. 11 ,

10 GW Gauss fit

(0)

below. Its sole free parameter is the width, denoted by p. An analytic expression for p will be derived in Sec. IV. The coincidence is in general good at high momenta, whereas at low momenta the spectral function deviates from the Gaussian. In particular, the steep wings and the smoother plateau that form at low momenta are not reproduced by the Gaussian. Also, the plasmaron peaks appearing in the spectral function at high density see Fig. 7 cannot be described by the single Gaussian. Determination of p via least-squares tting of the Gaussian ansatz to the numerical data at each p leads to the solid curve in Fig. 8, obtained in the case of n = 7 1020 cm3 and T = 100 eV. Starting at some nite value 0 at p = 0, the width falls off slowly toward larger p. The dashed curve shows p as obtained from the analytic formula that will be derived in the following section.

IV. ANALYTICAL EXPRESSION FOR THE QUASI-PARTICLE SELF-ENERGY

The solution of the GW 0 equation 3 requires a considerable numerical effort. So far see, e.g. the work by Fennel and Wilfer in 34 , attempts to solve the integral 3 analytically were led by the idea of replacing the spectral function on the right-hand side RHS by its noninteracting counter, i.e., going back to the perturpart, A 0 p , = 2 p bation expansion of the self-energy and neglecting the implied self-consistency. At the same time, the inverse dielectric function is usually replaced by a simplied expresGW Gauss fit

(0)

10

10

GW Gauss fit

(0)

p=0 aB

-1

p=50 aB

-1

p=100 aB-1

(a)

-3

-2

-1

0 1 frequency +e [Ry]

(b)

0 2497

2498

2502

2503

(c)

0 9997

9998

10002

10003

FIG. 3. Spectral function in GW 0 approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1020 cm3, T = 100 eV. The B B plasma coupling parameter is = 2.1 102, the degeneracy parameter is = 3.5 102, and the Debye screening parameter is = 1.9 1 102aB . 016404-4

191

SINGLE-PARTICLE SPECTRAL FUNCTION FOR THE

10 GW Gauss fit

(0)

GW Gauss fit

(0)

10

10

GW Gauss fit

(0)

p=0 aB

-1

p=50

-1 aB

p=100 aB-1

(a)

-3

-2

-1

0 1 frequency +e [Ry]

(b)

0 2497

2498

2502

2503

(c)

0 9997

9998

10002

10003

FIG. 4. Spectral function in GW 0 approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1021 cm3, T = 100 eV. The B B plasma coupling parameter is = 4.4 102, the degeneracy parameter is = 7.5 101, and the Debye screening parameter is = 6.0 102a1. B

sion, e.g., the Born approximation or the plasmon-pole approximation 12 . Whereas the second simplication is indispensable due to the complicated structure of the inverse dielectric function, the rst one, i.e., the recursion to the quasiparticle picture, is not necessary, as was shown by the author in Ref. 35 . In fact, the result that one obtains in the quasiparticle approximation is far from the converged result, at least in the high-temperature case. Second, if the quasiparticle approximation is used, the imaginary part of the selfenergy is not density dependent, i.e., a nite lifetime of the particle states is obtained even in vacuum. This unphysical result can be overcome only if one sticks to the selfconsistency of the self-energy, i.e., if one leaves the imaginary part of the self-energy entering the RHS of Eq. 3 nite. Using the statically screened Born approximation, which describes the binary collisions among electrons via a statically screened potential, a scaling law Im (p , QP p ) 3/4 was found 35 . Hence, the spectral function width vanishes when the plasma coupling parameter see Eq. 7 tends to 0. An expression for the self-energy was found that reproduces the converged GW 0 calculations at small mo. At higher momenta, the derived expression menta, p ceases to be valid. In this work, a different approximation to the dielectric function is studied, namely, the plasmon-pole approximation 12 . This means that the inverse dielectric function is replaced by a sum of two functions that describe the location of the plasmon resonances,

2 GW(0) Gauss fit

spectral function A(p,) [1/Ry] 2

Im

1 RPA

q,

2 pl

Im

1 PPA

q, +

q

For classical plasmas, the plasmon dispersion q can be approximated by the Bohm-Gross dispersion relation 38

2 q

2 pl

1+

q2

2

+ q4 .

Many-particle and quantum effects on the plasmon dispersion have recently been studied in 39 . The plasmon-pole approximation PPA allows one to perform the frequency integration in Eq. 3 , resulting in the expression Im p, + i =

2 pl

Vq

q q

1

q

A p q,

q/T

nB nB

exp

q

A p q, +

q/T

exp

10

We will rst study the case of high momenta, i.e., momenta that are large against any other momentum scale or inverse length scale, such as the mean momentum with respect to the Boltzmann distribution, = 3T / 2, or the inverse screening p length = 8 n / T.

GW Gauss fit

spectral function A(p,) [1/Ry]

(0)

GW Gauss fit

(0)

1.5 p=0 aB 1

-1

1.5 p=50 aB 1

-1

1.5 p=100 aB 1

-1

0.5

0.5

0.5

(a)

0 -10

-5

0 frequency +e [Ry]

10

(b)

0 2490

2495

2505

2510

(c)

0 9990

9995

10005

10010

FIG. 5. Spectral function in GW 0 approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1021 cm3, T = 1000 eV. The B B plasma coupling parameter is = 4.4 103, the degeneracy parameter is = 7.5 102, and the Debye screening parameter is = 1.9 1 102aB . 016404-5

192

C. FORTMANN

PHYSICAL REVIEW E 79, 016404 2009

As discussed in the previous section, the numerical results for the spectral function at high momenta can well be reproduced by a Gaussian. Thus, we make the following ansatz for the spectral function: 2

p

= 1.3357 nB

pl

2 2p

pl

nB

pl/T

pl

exp

pl/T

exp

T ln 2 2p

2 2

p/

2 p

. 15

AGauss p,

exp

2

2 p

HF

. 11

Note that only the Hartree-Fock contribution to the real part of the self-energy appears. The frequency-dependent part is usually small near the quasiparticle dispersion,

QP

The solution of this equation can be expanded for large arguments of the logarithm, yielding

p

= ln p p

T 2p ln

2

p , p p

p =

+ Re

p,

= QP p

12

p = +

p ln ln p

3

p +

2

QP p = p which therefore can be approximated as HF + Re p . In the following, we make use of the knowledge about the width parameter p that we gathered already through simple least-squares tting of the Gaussian ansatz to the spectral function in order to solve the integrals in Eq. 10 . First, we replace the spectral function on the RHS by the Gaussian ansatz 11 and evaluate the emerging equation at the single-particle dispersion QP p . By claiming that the Gaussian and the spectral function have the same value at the quasiparticle energy, we identify p = / 2Im (p , QP p ). Figure 8 shows that the quasiparticle damping p is a smooth function of p that varies only little on the scale of the screening parameter . Since the latter denes the scale on which contributions to the q integral are most important, we can neglect the momentum shift in the self-energy on the RHS, i.e., we can replace the spectral function on the RHS of Eq. 10 by HF

p ,

3 ln p ln2 p ln3 p + + 2 3p 2 2p 3 3p

+O p

p = ln

p2 exp A/T /T ,

pl/T

16

A = 1.3357 nB

pl

2

pl

nB

pl

exp .

exp

pl/T

A p q,

+ 2

p

p

p

q q

exp

2 p

pq

13 between

Equation 16 is a solution of Eq. 15 provided the argument of the inner logarithm is larger than Eulers constant e, i.e., 2 / p2 exp A / T / T e, i.e., at large p. The case of small p, where the previous inequality does not hold, has to be treated separately; see Appendix B. Together with an expression for the quasiparticle damping at vanishing momentum taken from 35 and scaled such that the maximum of the spectral function at p = 0 is reproduced, T / 2, an interpolation formula Pad formula was 0= derived that covers the complete p range:

Pad = p

and can now perform the integral over the angle the momenta p and q, 2

p 1

a0 + a1 p p , 1 + b1 p + b2 p2

d cos

1

ln p ln p ln p p = p ln p + + p 2 p p

p 2 2 q p q + 2pq cos + e 2

exp = Erf

2 p+q p 2 2

2 2

2 p q

2pq Erf

p = ln e + . 14

p2 exp A/T /T ,

p q 3/2

p q 2 p2

p

a0 =

T,

a1 =

b1 =

2T

b2 =

2T

The remaining integration over the modulus of the transfer momentum q can be performed after some further approximations, explained in detail in Appendix A. For large p, one nally obtains the transcendental equation

17 The function p in the last equation differs from p in Eq. 16 in that Eulers constant e 2.7183 has been added to the

016404-6

193

SINGLE-PARTICLE SPECTRAL FUNCTION FOR THE

1 GW(0) Gauss fit

spectral function A(p,) [1/Ry] 1 GW Gauss fit

spectral function A(p,) [1/Ry]

(0)

1 GW Gauss fit

(0)

0.8

0.8

0.8

0.6

p=0 aB

-1

0.6

p=50

-1 aB

0.6

p=100 aB

-1

0.4

0.4

0.4

0.2

0.2

0.2

(a)

0 -20

-15

-10

-5 0 5 frequency +e [Ry]

10

15

20

(b)

0 2480

2485

2490

2510

2515

2520

(c)

0 9980

9985

9990

10010

10015

10020

FIG. 6. Spectral function in GW 0 approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1023 cm3, T = 1000 eV. The B B plasma coupling parameter is = 2.1 102, the degeneracy parameter is = 3.5 101, and the Debye screening parameter is = 1.9 1 1 10 aB .

argument of the logarithm. In this way, the function p is regularized at small p and tends to 1 at p = 0, i.e., the quasiparticle damping goes to the correct low-p limit. At large p, this modication is insignicant, since the original argument rises as p2. For the detailed derivation, see Appendix B. Expression 17 , used in the Gaussian ansatz 11 , leads to a spectral function that well reproduces the numerical data from full GW 0 calculations: Figure 8 dashed curve shows the effective quasiparticle damping width p as a function of momentum p for the case n = 7 1020 cm3 and T = 100 eV. The solid curve gives the best-t value for p obtained via least-squares tting of the full GW 0 calculations assuming the Gaussian form 11 see Sec. III . The two curves coincide to a large extent. The largest deviations are observed in 1 the range of p 20aB . At this point, the validity of expression 16 as the solution of Eq. 15 ceases, since the argument of the logarithm becomes smaller than e. As already mentioned, we circumvented this problem by regularizing the logarithm, adding e to its argument. The deviation at p 1 20aB of up to 15% is a residue of this procedure. At higher momenta, the deviation is generally smaller than 10% and the analytic formula evolves parallel to the t parameters. At smaller densities, the correspondence is even better as can be seen by comparing the spectral functions shown in Figs. 27. The dashed curves give the Gaussian ansatz for the spectral function with the quasiparticle width taken from the interpolation formula 17 . As a general result, the ana1 GW(0) Gauss fit spectral function A(p,) [1/Ry] 1

lytic expression for the quasiparticle damping p leads to a spectral function that nicely ts the numerical solution for the spectral function at least at nite p. At very small values of p, the overall correspondence is still fair, i.e., the position of the maximum and the overall width match, but the detailed behavior does not coincide. In particular, the steep wings and the central plateau that form in the GW 0 calculation are not reproduced by the one-parameter Gaussian. For this situation, the analytic formula for self-energy given in 35 should be used instead. By comparing the numerical data for the spectral function to the Gaussian ansatz at different densities, it is found that the Gaussian spectral function is a good approximation as long as the Debye screening parameter is smaller than the 1 1aB . This becomes obvious by inverse Bohr radius, comparing Figs. 6 and 7. In the rst case n = 7 1023 cm3, T = 1000 eV , we have = 0.19, while in the second case n = 7 1025 cm3, T = 1000 eV , = 1.9 is found. As already noted in the discussion of the numerical results in Sec. III, in the case of increased density, the plasmaron satellites appear as separate structures in the wings of the central quasiparticle peak, whereas they are hidden in the central peak at smaller densities. Therefore, a single Gaussian is not sufcient to t the spectral function at increased densities. Since the position of the plasmaron peak is given approximately by the plasma frequency pl, whereas the width of the central peak at small p is just the quasiparticle width 0, we can identify the ratio , as the parameter that of these two quantities, pl / 0

GW(0) Gauss fit spectral function A(p,) [1/Ry] 1 GW(0) Gauss fit

0.8

0.8

0.8

0.6

p=0 aB

-1

0.6

p=50 aB

-1

0.6

p=100 aB

-1

0.4

0.4

0.4

0.2

0.2

0.2

(a)

(b)

(c)

9960

9980

10040

FIG. 7. Spectral function in GW 0 approximation solid lines and Gaussian ansatz dashed lines with quasiparticle damping width p taken from Eq. 17 for three different momenta p = a 0, b 50a1, and c 100a1. Plasma parameters: n = 7 1025 cm3, T = 1000 eV. The B B plasma coupling parameter is = 9.6 102, the degeneracy parameter is = 1.6, and the Debye screening parameter is = 1.9a1. Here, the B Gaussian t is no longer sufcient due to the appearance of plasmaron resonances in the spectral function shoulders at 30 and 20 Ry . 016404-7

194

C. FORTMANN

PHYSICAL REVIEW E 79, 016404 2009

1e-01

GW -fit Interpolation formula

(0)

- [Ry]

(0)

1e-02

T = 100 eV

20

40 60 -1 momentum p [aB ]

80

100

1e+19

1e+22

FIG. 8. Effective quasiparticle damping p as a function of momentum p for plasma density n = 7 1020 cm3 and temperature T = 100 eV. The t parameters for the Gaussian t to the full GW 0 calculations are given as the solid line; the dashed line denotes the analytic interpolation formula 17 .

FIG. 9. Shift of the chemical potential as a function of the plasma density for a plasma temperature T = 100 eV. Results for using the parametrized spectral function solid line are compared to full numerical calculations, using the GW 0 approximation dashed line .

tells us if plasmaron peaks appear separately pl 0 or not pl 0 . Since the plasma frequency increases as a function of n1/2, whereas the quasiparticle width scales as n1/4 see Eq. 17 , the transition from the single-peak behavior to the more complex behavior including plasmaron resonances appears at increased density. Neglecting numerical constants of order 1 in the ratio of plasma frequency to damping width, we see that pl / 0 1 is equivalent to 1, which was our observation from the numerical results. Therefore, we can identify the range of validity of the presented expressions for the spectral function and the quasiparticle damping. It is valid for those plasmas where we have 1. densities and temperatures such that The physical origin of the requirement 1 can be understood in the following way 35 . At length scales smaller than the Bohr radius, one typically expects quantum effects, e.g., Pauli blocking. These effects are not accounted for in the derivation of the quasiparticle damping. Therefore, it appears to be a logical consequence that the validity of the results is limited by the length scale at which typical quantum phenomena become important. The regime of validity of the analytic formula can also be expressed via the plasma coupling parameter and the temperature as T2/3. Since we restrict ourselves to plasma temperatures where bound states can be excluded, i.e., T 1 Ry, this is equivalent to saying that 1. Although the correspondence between the accurate GW 0 calculations and the parametrized spectral function at small momenta is not as good as in the case of large momenta, the parametrized spectral function can be applied in the regime of validity to the calculation of plasma observables without introducing too large errors. As an example, this will be shown for the case of the chemical potential in the next section.

V. APPLICATION: SHIFT OF THE CHEMICAL POTENTIAL

calculate the shift of the electrons chemical potential = free, i.e., the deviation of the chemical potential of the interacting plasma from the value of the noninteracting system free. The chemical potential of the interacting system is obtained by inversion of the density as a function of T and , Eq. 5 . The free chemical potential free is obtained in a similar way by inversion of the free density, nfree T,

free

=2

p

nF

free

18

Figure 9 shows the shift of the chemical potential as a function of the plasma density n for a xed plasma temperature T = 100 eV. Results obtained by inversion of Eq. 5 using the parametrized spectral function 11 with the quasiparticle damping width taken from Eq. 17 solid curve are compared to those results taking the numerical GW 0 spectral function dashed curve . The GW 0 result gives slightly smaller shifts than the parametrized spectral function, i.e., the usage of the analytical damping width leads to an overestimation of the shift of the chemical potential. However, the deviation remains smaller than 20% over the range of densities considered here, i.e., for 1. At small densities, i.e., for n 1020 cm3, the parametrized spectral function yields the same result as the full GW 0 calculation. The deviation at increased density can be reduced by improving the parametrization of the spectral function at small momenta. To this end, the behavior of the quasiparticle damping width at high momenta, Eq. 16 should be combined with the frequency-dependent solution for p at vanishing momentum, as presented in Ref. 35 . However, this task goes beyond the scope of this paper, where we wish to present comparatively simple analytic expressions for the damping width that yield the correct low-density behavior of plasma properties.

VI. CONCLUSIONS AND OUTLOOK

To demonstrate the applicability of the presented formulas for quick and reliable calculations of plasma properties, we

In this paper, the GW 0 approximation for the singleparticle self-energy was evaluated for the case of a classical

016404-8

195

SINGLE-PARTICLE SPECTRAL FUNCTION FOR THE PHYSICAL REVIEW E 79, 016404 2009

one-component electron plasma, with ions treated as a homogeneous charge background. A systematic behavior of the spectral function was found, i.e., a symmetrically broadened structure at low momenta and convergence to a sharp quasiparticle resonance at high p. At increased densities, plasmaron satellites show up in the spectral function as satellites besides the main peak. In the second part, an analytic formula for the imaginary part of the self-energy at the quasiparticle dispersion QP p = p + HF p was derived as a two-point Pad formula that interpolates between the exactly known behavior at p = 0 and p . The former case was studied in 35 , while an expression for the asymptotic case p was derived here. The result is summarized in Eq. 17 . In contrast to previously known expressions for the quasiparticle damping, based on a perturbative approach to the self-energy 34 , the result presented here shows a physically intuitive behavior in the limit of low densities, i.e., it vanishes when the system becomes dilute. Use of the Gaussian ansatz 11 for the spectral function in combination with the quasiparticle width leads to a very good agreement with the numerical data for the spectral function in the range of plasma parameters 1 1aB ; the relative deviation is smaller than 10% where under this constraint. Thus, a simple expression for the damping width of electrons in a classical plasma has been found, which can be used to approximate the full spectral function to high accuracy. This achievement greatly facilitates the calculation of observables that take the spectral function or the self-energy as an input, such as optical properties inverse bremsstrahlung absorption , conductivity, or the stopping power. Furthermore, it was demonstrated that the derived expressions allow for quick and reliable calculations of plasma properties without having to resort to the full self-consistent solution of the GW 0 approximation. As an example, the shift of the chemical potential was calculated using the parametrized spectral function, and compared to GW 0 results. For densities of n 1021 cm3, the two approaches coincide with a relative deviation of less than 10%, going eventually up to 20% as the density approaches 1022 cm3. At low densities both approaches give identical results. This shows the extreme usefulness of the presented approach for the calculation of observables via the parametrized spectral function. As a further important application of the results presented in this paper, we would like to mention the calculation of radiative energy loss of particles traversing a dense medium, i.e., bremsstrahlung. A many-body theoretical approach to this scenario is given by Knoll and Voskresensky 40 , using nonequilibrium Green functions. They showed that a nite spectral width of the emitting particles leads to a decrease in the bremsstrahlung emission. This effect is known as the Landau-Pomeranchuk-Migdal effect 41,42 . It has been experimentally conrmed in relativistic electron scattering experiments using dense targets, e.g., lead 43,44 . In 45 , it is shown that thermal bremsstrahlung from a plasma is also reduced due to the nite spectral width of the electrons in the plasma. In the cited papers, the quasiparticle damping width was either set as a momentum- and energy-independent parameter in 40 , or calculated self-consistently using simplied approximations of the GW 0 theory in 45 , which

itself is a very time-consuming task and prohibited investigations over a broad range of plasmas parameters. Now, based on this works results, calculations on the level of the full GW 0 approximation become feasible, since analytic formulas have been found that reproduce the GW 0 self-energy. Effects of dynamical correlations on the bremsstrahlung spectrum can be studied starting from a consistent singleparticle description via the GW 0 self-energy.

ACKNOWLEDGMENTS

The author acknowledges much helpful advice from Gerd Rpke and fruitful discussion with W.-D. Kraeft as well as with C. D. Roberts. Financial support was obtained from the German Research Society DFG via the Collaborative Research Center Strong Correlations and Collective Effects in Radiation Fields: Coulomb Systems, Clusters, and Particles SFB 652 .

APPENDIX A: ANALYTIC SOLUTION FOR THE GW(0) SELF-ENERGY USING THE PLASMON-POLE APPROXIMATION

After the angular integration which was performed in Eq. 14 , the imaginary part of the self-energy at the quasiparticle dispersion = p reads Im p,p2 =

2 pl

4p

dq q q 2

Erf

q2 + 2pq + 2

q p

q2 2pq +

p

nB

q

exp

q/T

q2 + 2pq 2

2 p q

q 2pq 2

p

nB

exp

q/T

A1 This equation represents a self-consistent equation for Im p , = p2 = 2 / p. Our aim is to derive an analytic expression that approximates the numerical solution of Eq. A1 for arbitrary p. To , this end, we rst look at the case of large momenta, p and later combine that result with known expressions for the limit of vanishing momentum p 0, to produce an interpolation Pad formula that covers the complete p range. We perform a sequence of approximations to the integral in A1 . First, we observe, that at large p, the term 2pq dominates in the argument of the error function. We rewrite Eq. A1 as

p=

2 Erf Erf

Im

p,p2 = nB nB

2 pl

2 4p

q

0 q/T

dq q q

Erf

2pq 2 2

p

2pq 2 2

p

exp

Erf

q/T

2pq

p

2pq

p

exp

A2

016404-9

196

C. FORTMANN

2 pl

PHYSICAL REVIEW E 79, 016404 2009

2 4p nB

q

2pq dq 2 Erf q q 2 p

q/T

nB

exp

q/T

p = 1.3357 2 pl

exp

A3

2 2p

pl/T

nB

pl

exp

pl/T

nB p/

2 p

pl

The integrand in Eq. A3 contains a steeply rising part at q p / p and a smoothly decaying part for at large q, i.e., when q p / p. Therefore, we separate the integral in the equation into two parts, one going from q = 0 to q = = q p / p and the other from to innity. In the rst part of the q integral, the values for q are so small that we can replace the plasmon dispersion by the plasma frequency pl. In the second term, the argument of the error function is large and the error function can be replaced by its asymptotic value at innity, limx Erf x = 1. This leads to

p

exp

T ln 1 + 2 2p

2 2

A8

At large p, the term 2 p2 / p dominates the argument of the logarithm, i.e., we can write ln 1 + 2 p2 / 2 ln 2 p2 / 2 . p p Then, we arrive at Eq. A1 , given in Sec. IV.

2 pl

q 0

2 4p nB nB

pl

2pq dq 2 Erf q pl 2 p

pl/T

nB

pl

exp

pl/T

From the knowledge of the behavior of p in the limits p 0 and p , a two-point Pad interpolation formula can be constructed. For the value of the quasiparticle damping width at p = 0 we take the expression

0

exp exp

+2

q

dq nB q q

exp

q/T

T,

B1

q/T

A4

Finally, we expand the last term in powers of q / T, which is justied at low densities q pl , and keep only the rst order, nB

q

which is the exact solution of the self-consistent Born approximation 35 . The Pad interpolation formula is constructed in the following way. We make the ansatz

Pad = p

exp

q/T

nB

exp

q/T

2T

q

a0 + a1 p p , 1 + b1 p + b2 p2

B2

+O

A5 We obtain

p= 2 pl q 0

where the function p contains the logarithmic terms present in the behavior of p at large p cf. Eq. 16 : ln ln ln p = ln + + 2

2 4p nB

pl

2pq dq 2 Erf q pl 2 p

pl/T

nB

pl

exp

pl/T

exp

+ 4T

dq 2 . q q

A6

B3

q 0

= ln e +

B4

2pq dq Erf =2 q 2 p

2 pq

p 2 p

2F 2

1/2,1/2;3/2,3/2;

p0 Pad = p

2p2 2/ q =2 2

a 0 + a 1 a 0b 1 p + O p 2 ,

B5

2F 2

1/2,1/2;3/2,3/2; 2 lim

p Pad = p

= 1.3357, dq 1 + q 2/ = 1 ln 1 + 2

2

a 1 a 0b 2 a 1b 1 + + O p3 p , b2 p b2 p2 2

B6

2 pl

/q2 =

1 ln 1 + 2

2 2

p/

2 p

A7 where = p / p was used. Note that, in the second integral, q the q4 term in the plasmon dispersion 9 is omitted. 2F2 a1 , a2 ; b1 , b2 ; z is the generalized hypergeometric function 46 .

and comparison to the behavior of p in these limiting cases, e.g., Eq. 16 for large p and Eq. B1 for p 0. Setting the slope of p at p = 0 to zero, as well as the coefcient in front of the p2 term of the asymptotic expansion, we arrive at the following equations for the coefcients of the interpolation formula: a0 = 2 T, a1 a0bq = 0, B7

016404-10

197

SINGLE-PARTICLE SPECTRAL FUNCTION FOR THE PHYSICAL REVIEW E 79, 016404 2009

b 2,

a0b2 a1b1 = 0.

B8

b1 =

2T

b2 =

2T

B10

3/2

a0 =

T,

a1 =

B9

1 L. P. Kadanoff and G. Baym, Quantum Statistical Mechanics W. A. Benjamin, New York, 1962 . 2 G. D. Mahan, Many-Particle Physics, 2nd ed. Plenum Press, New York, 1981 . 3 A. L. Fetter and J. D. Walecka, Quantum Theory of ManyParticle Systems McGraw-Hill, New York, 1971 . 4 A. Hll, C. D. Roberts, and S. V. Wright, in Particles and Fields: X Mexican Workshop, edited by M. A. Prez, L. Urrutia, and L. Villaseqor, AIP Conf. Proc. No. 857 AIP, Melville, NY, 2006 , pp. 4661. 5 A. Wierling, in Advances in Plasma Physics Research, edited by F. Gerard Nova Science, New York, 2002 , p. 127. 6 H. Reinholz, Ann. Phys. Paris 30, 1 2005 . 7 C. Fortmann, G. Rpke, and A. Wierling, Contrib. Plasma Phys. 47, 297 2007 . 8 H. Reinholz, I. Morozov, G. Rpke, and T. Millat, Phys. Rev. E 69, 066412 2004 . 9 D. O. Gericke, M. Schlanges, and W.-D. Kraeft, Phys. Lett. A 222, 241 1996 . 10 W.-D. Kraeft and B. Strege, Physica A 149, 313 1988 . 11 J. Vorberger, M. Schlanges, and W.-D. Kraeft, Phys. Rev. E 69, 046407 2004 . 12 W. D. Kraeft, D. Kremp, W. Ebeling, and G. Rpke, Quantum Statistics of Charged Particle Systems Akademie-Verlag, Berlin, 1986 . 13 W. Ebeling, W.-D. Kraeft, and D. Kremp, Theory of Bound States and Ionization Equilibrium in Plasmas and Solids Akademie-Verlag, Berlin, 1972 . 14 F. J. Dyson, Phys. Rev. 75, 1736 1949 . 15 J. Schwinger, Proc. Natl. Acad. Sci. U.S.A. 37, 451 1951 . 16 J. Schwinger, Proc. Natl. Acad. Sci. U.S.A. 37, 455 1951 . 17 C. D. Roberts and S. M. Schmidt, Prog. Part. Nucl. Phys. 45, S1 2000 . 18 L. Hedin, Phys. Rev. 139, A796 1965 . 19 U. von Barth, N. E. Dahlen, R. van Leeuwen, and G. Stefanucci, Phys. Rev. B 72, 235109 2005 . 20 G. Baym and L. P. Kadanoff, Phys. Rev. 124, 287 1961 . 21 G. Baym, Phys. Rev. 127, 1391 1962 . 22 F. Aryasetiawan and O. Gunnarsson, Rep. Prog. Phys. 61, 237 1998 . 23 G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. 74, 601

2002 . 24 G. D. Mahan, Comments Condens. Matter Phys. 16, 333 1994 . 25 H. van Hees and J. Knoll, Phys. Rev. D 65, 025010 2001 . 26 H. van Hees and J. Knoll, Phys. Rev. D 65, 105005 2002 . 27 H. van Hees and J. Knoll, Phys. Rev. D 66, 025028 2002 . 28 Y. Takada, Phys. Rev. Lett. 87, 226402 2001 . 29 U. von Barth and B. Holm, Phys. Rev. B 54, 8411 1996 . 30 B. Holm and U. von Barth, Phys. Rev. B 57, 2108 1998 . 31 R. Fehr and W.-D. Kraeft, Contrib. Plasma Phys. 35, 463 1995 . 32 A. Wierling and G. Rpke, Contrib. Plasma Phys. 38, 513 1998 . 33 R. Schepe, T. Schmielau, D. Tamme, and K. Henneberger, Phys. Status Solidi B 206, 273 1998 . 34 W. Fennel and H. P. Wilfer, Ann. Phys. Leipzig 32, 265 1974 . 35 C. Fortmann, J. Phys. A: Math. Theor. 41, 445501 2008 . 36 J. N. Bahcall, M. H. Pinsonneault, and G. J. Wasserburg, Rev. Mod. Phys. 67, 781 1995 . 37 B. I. Lundquist, Phys. Kondens. Mater. 6, 206 1967 . 38 D. Bohm and E. P. Gross, Phys. Rev. 75, 1851 1949 . 39 R. Thiele, T. Bornath, C. Fortmann, A. Hll, R. Redmer, H. Reinholz, G. Rpke, A. Wierling, S. H. Glenzer, and G. Gregori, Phys. Rev. E 78, 026411 2008 . 40 J. Knoll and D. Voskresensky, Ann. Phys. N.Y. 249, 532 1996 . 41 A. B. Migdal, Phys. Rev. 103, 1811 1956 . 42 S. Klein, Rev. Mod. Phys. 71, 1501 1999 . 43 H. D. Hansen, U. I. Uggerhj, C. Biino, S. Ballestrero, A. Mangiarotti, P. Sona, T. J. Ketel, and Z. Z. Vilakazi, Phys. Rev. Lett. 91, 014801 2003 . 44 H. D. Hansen, U. I. Uggerhj, C. Biino, S. Ballestrero, A. Mangiarotti, P. Sona, T. J. Ketel, and Z. Z. Vilakazi, Phys. Rev. D 69, 032001 2004 . 45 C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling, in Condensed Matter Theory, edited by J. Clark, R. Panoff, and H. Li Nova Science, New York, 2005 , Vol. 20, p. 317. 46 Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables, 9th ed., edited by M. Abramowitz and A. Stegun Dover Publications, New York, 1970 .

016404-11

198

200

Curriculum vitae

Curriculum vitae

24.07.1978 07/1989 06/1998 20.06.1998 10/1998 10/1999 10/1999 09/2001 07/2001 10/2001 09/2002 10/2002 09/2003 07/2003 10/2003 08/2004 Geboren in Eutin Besuch des Gymnasiums, Johann-Heinrich Voss Schule, Eutin Abitur Zivildienst, Universittsklinikum Freiburg. i. Br. a Grundstudium im Studiengang Physik (Diplom), Universitt Rostock a Vordiplomprfungen u Studienaufenthalt an der Unversit de Nantes, Frankreich e Hauptstudium im Studiengang Physik (Diplom), Universitt Rostock a Diplomprfungen u Diplomarbeit zum Thema Bremsstrahlung in Dichter Materie und Landau-Pomeranchuk-Migdal Eekt Whrend des Studiums gefrdert durch die a o Studienstiftung des Deutschen Volkes 09/2004 12/2004 Stipendiat des Gradiuiertenkollegs 567 Stark korrelierte Vielteilchensysteme 01/2005 06/2005 Wissenschaftlicher Mitarbeiter im Virtuellen Institut der Helmholtzgemeinschaft Plasma Physics Research using FEL Radiation 07/2005 heute Wissenschaftlicher Mitarbeiter im Sonderforschungsbereich 652 der DFG Starke Korrelationen und Kollektive Phnomene im Strahlungsfeld: a Coulombsysteme, Cluster und Partikel

List of publications

201

Eingeladene Vortrge a

1. Diagnostics of Dense Plasmas using Thomson Scattering C. Fortmann, R. Redmer, H. Reinholz, G. Rpke, R. Thiele, S. Glenzer, H. Lee o GSI Plasmaphysik-Seminar Juni 2008 2. Bremsstrahlung and Thomson Scattering in Dense Plasmas, C. Fortmann, G. Rpke, and A. Wierling, o Lawrence Livermore National Laboratory, Livermore, California, USA, Juni 2007.

Konferenzvortrge a

1. Spectral Function for Dense Plasmas, C. Fortmann, G. Rpke, and A. Wierling, o Frhjahrstagung der Deutschen Physikalischen Gesellschaft, u Darmstadt, Mrz 2008. a 2. Thomson Scattering at FLASH, C. Fortmann, Physics of High Energy Density Matter Conference, Hirschegg, Osterreich, Januar 2008. 3. Theory of Thomson scattering in Warm Dense Matter, C. Fortmann, Abschluss-Meeting des Virtuellen Instituts Plasma Physics Research Using FEL Radiation, Hamburg, Dezember 2007. 4. Thomson Scattering in Warm Dense Matter, C. Fortmann, 2. Laserlab Europe Workshop, Darmstadt, Oktober 2007. 5. FLASH scattering and absorption in liquid H2, C. Fortmann, Peak-Brightness Collaboration Meeting, Hamburg, September 2007. 6. Bremsstrahlung from Dense Plasmas: A Many-Body Theoretical Approach, C. Fortmann, G. Rpke, and A. Wierling, o IEEE Pulsed Power and Plasma Sciences Conference, Albuquerque, New Mexico, USA, Juni 2007.

202

List of publications

7. Einteilchen-Spektralfunktion und optische Eigenschaften dichter Plasmen, C. Fortmann, G. Rpke, A. Wierling, o Frhjahrstagung der Deutschen Physikalischen Gesellschaft, u Dsseldorf, Mrz 2007. u a 8. Bremsstrahlung und optische Spektren aus dichten Plasmen, C. Fortmann, G. Rpke, J. Tiggesbumker, T. Dppner, o a o Workshop des SFB 652 Starke Korrelationen und kollektive Phnomene im Strahlungsa feld: Coulombsysteme, Cluster und Partikel, Warnemnde, Februar 2007. u 9. Optical Properties and the One-particle Spectral Function in Non-ideal Plasmas, C. Fortmann, G. Rpke, and A. Wierling, o 12th International Workshop on Physics of Non-ideal Plasmas (PNP12), Darmstadt, September 2006. 10. Spectral function for charged particles: GW and beyond, C. Fortmann, G. Rpke, o Workshop des SFB 652 Starke Korrelationen und kollektive Phnomene im Strahlungsa feld: Coulombsysteme, Cluster und Partikel, Dorf Mecklenburg, April 2006. 11. Bremsstrahlung vs. Thomson scattering in VUV-FEL plasma experiments, C. Fortmann, R. Redmer, H. Reinholz, G. Rpke, W. Rozmus, and A. Wierling, o Physics of High Energy Density Matter Conference, Hirschegg, Osterreich, Januar 2006. 12. Landau-Pomeranchuk-Migdal Eect in Dense Plasmas, C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling, o Frhjahrstagung der Deutschen Physikalischen Gesellschaft, u Berlin, Mrz 2005. a 13. Bremsstrahlung in Dense Matter and LPM Eect, C. Fortmann, G. Rpke, o Workshop of the Virtual Institute Dense Hadronic Matter and QCD Phase Transition, Bad Honnef, Juli 2004.

1. Carsten Fortmann, Self-consistent Spectral Function for Non-Degenerate Coulomb Systems and Analytic Scaling Behaviour, J. Phys. A: Math. Gen. (accepted).

203

2. R. Thiele, T. Bornath, C. Fortmann, A. Hll, R. Redmer, H. Reinholz, G. Rpke, A. o o Wierling, S. H. Glenzer, G. Gregori, Plasmon resonance in warm dense matter, Phys. Rev. E 78, 026411 (2008). 3. S. H. Glenzer, P. Neumayer, T. Dppner, O. L. Landen, R. W. Lee, R. J. Wallace, o S. Weber, H. J. Lee, A. L. Kritcher, R. Falcone, S. P. Regan, H. Sawada, D. D. Meyerhofer, G. Gregori, C. Fortmann, V. Schwarz, and R. Redmer, Compton scattering measurements from dense plasmas, J. Phys.: Conf. Series 112, 032071 (2008). 4. A. Hll, Th. Bornath, L. Cao, T. Dppner, S. Dsterer, E. Frster, C. Fortmann, S.H. o o u o Glenzer, G. Gregori, T. Laarmann, K.-H. Meiwes-Broer, A. Przystawik, P. Radclie, R. Redmer, H. Reinholz, G. Rpke, R. Thiele, J. Tiggesbumker, S. Toleikis, N.X. o a Truong, T. Tschentscher, I. Uschmann, and U. Zastrau, Thomson Scattering from Near-solid Density Plasmas Using Soft X-Ray Free Electron Lasers, Journal of High Energy Density Physics 3,120 (2007). 5. C Fortmann, G. Rpke, and A. Wierling, o Optical Properties and the One-particle Spectral Function in Non-ideal Plasmas, Contrib. Plasma Phys. 47, 297 (2007). 6. C. Fortmann, R. Redmer, H. Reinholz, G. Rpke, W. Rozmus, and A. Wierling, o Bremsstrahlung vs. Thomson Scattering in VUV-FEL Plasma Physics Experiments, Journal of High Energy Density Physics 2, 57 (2006).

1. C. Fortmann, Single Particle Spectral Function for the Classical One-Component Plasma, eingereicht bei Phys. Rev. E. preprint: http://arxiv.org/physics/0808.2151. 2. U. Zastrau, C. Fortmann, R. R. Fustlin et al. a Bremsstrahlung and Line Spectroscopy of Warm Dense Aluminum Plasma generated by XUV Free-Electron Laser Radiation, eingereicht bei Phys. Rev. Lett.

204

List of publications

1. Carsten Fortmann, R. Redmer, H. Reinholz, G. Rpke, and R. Thiele Thomson o Scattering on Solid Density and Compressed Be, High Energy Density Physics with Intense Ion and Laser Beams: GSI Annual Report 2007. 2. Bremsstrahlung from dense matter and the Landau-Pomeranchuk-Migdal eect, C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling, o Condensed Matter Theories, Vol. 20, p. 317 (Nova Science, New York, 2006). 3. Carsten Fortmann, R. Redmer, H. Reinholz, G. Rpke, W. Rozmus, and A. Wierling, o Bremsstrahlung vs. Thomson scattering in VUV-FEL plasma experiments, High Energy Density Physics with Intense Ion and Laser Beams: GSI Annual Report 2005. 4. C. Fortmann, H. Reinholz, G. Rpke, and A. Wierling, o Bremsstrahlung in dense matter and Landau-Pomeranchuk-Migdal eect, GSI Plasma Physics Annual Report 2004.

Declaration of authorship

205

Erklrung a

Hiermit erklre ich an Eides statt, dass ich die vorliegende Arbeit selbstndig angefertigt a a und ohne fremde Hilfe verfasst habe, keine auer den von mir angegebenen Hilfsmitteln und Quellen dazu verwendet habe und die den benutzten Werken inhaltlich und wrtlich o entnommenen Stellen als solche kenntlich gemacht habe.

Carsten Fortmann

## Viel mehr als nur Dokumente.

Entdecken, was Scribd alles zu bieten hat, inklusive Bücher und Hörbücher von großen Verlagen.

Jederzeit kündbar.