Exact Results for Thermodynamics of the Hydrogen Plasma

Plasma: Low-Temperature Expansions Beyond Saha ... Abstract We study hydrogen in the Saha regime, within the physical picture in terms of a quantum pr...

0 downloads 147 Views 1MB Size
J Stat Phys DOI 10.1007/s10955-007-9464-0

Exact Results for Thermodynamics of the Hydrogen Plasma: Low-Temperature Expansions Beyond Saha Theory A. Alastuey · V. Ballenegger · F. Cornu · Ph.A. Martin

Received: 20 April 2007 / Accepted: 6 November 2007 © Springer Science+Business Media, LLC 2007

Abstract We study hydrogen in the Saha regime, within the physical picture in terms of a quantum proton-electron plasma. Long ago, Saha showed that, at sufficiently low densities and low temperatures, the system behaves almost as an ideal mixture made with hydrogen atoms in their groundstate, ionized protons and ionized electrons. More recently, that result has been rigorously proved in some scaling limit where both temperature and density vanish. In that Saha regime, we derive exact low-temperature expansions for the pressure and internal energy, where density ρ is rescaled in units of a temperature-dependent density ρ ∗ which controls the cross-over between full ionization (ρ  ρ ∗ ) and full atomic recombination (ρ  ρ ∗ ). Each term reduces to a function of ρ/ρ ∗ times temperature-dependent functions which decay exponentially fast when temperature T vanishes. Scaled expansions are ordered with respect to the corresponding decay rates. Leading terms do reduce to ideal contributions obtained within Saha theory. We consistently compute all corrections which are exponentially smaller by a factor exp(βEH ) at most, where EH is the negative groundstate energy of a hydrogen atom and β = 1/(kB T ). They include all effects arising from both the Coulomb potential and the quantum nature of the particles: excitations of atoms H , formation of molecules H2 , ions H2+ and H − , thermal and pressure ionization, plasma polarization, screening, interactions between atoms and ionized charges, etc. Scaled low-temperature A. Alastuey () Laboratoire de Physique, Université de Lyon, ENS Lyon, CNRS, 46 allée d’Italie, 69364 Lyon Cedex 07, France e-mail: [email protected] V. Ballenegger Institut UTINAM, Université de Franche-Comté, CNRS, 16 route de Gray, 25030 Besançon Cedex, France F. Cornu Laboratoire de Physique Théorique, Université Paris-Sud, CNRS, Bâtiment 210, 91405 Orsay Cedex, France Ph.A. Martin Institut de Théorie des Phénomènes Physiques, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland

A. Alastuey et al.

expansions can be viewed as partial resummations of usual virial expansions up to arbitrary high orders in the density.

1 Introduction Hydrogen is an important element, both at a theoretical level and for practical purposes. Here, we consider a non-relativistic quantum hydrogen plasma, made of protons and electrons with respective masses mp and me , which interact via the familiar 1/r-Coulomb potential (see Sect. 2.1). As far as thermodynamic properties of that system are concerned, an exact calculation at finite temperature T and finite density ρ, remains far beyond present human abilities. Nonetheless, by exploiting the exact knowledge of the spectrum of hydrogen atom and using Morita’s method [51], Ebeling [23] first computed low-density expansions for pressure and free energy up to order ρ 2 at fixed non-zero temperature, in a closed analytical form (see also Ref. [39]). When ρ goes to zero, the system becomes fully ionized (see Ref. [43] for a rigorous proof). At order ρ 2 , the recombination of a small fraction of charges into hydrogen atoms is exactly taken into account. Such low-density expansions have been more recently completed up to order ρ 5/2 [4–7]. Those results have been checked afterwards in Ref. [37], and their high-temperature form in the one-component case does coincide with that derived in Ref. [20]. In the opposite limit where ρ goes to infinity at zero temperature, the system behaves as a mixture of free Fermi gases, and expansions in inverse powers of ρ have been calculated (see Refs. [29] and [50] for the first calculations in the one-component case, and also Ref. [39] for similar works or extensions). The previous exact asymptotic expansions are suitable for regimes where the system is almost fully ionized. The purpose of the present paper is to derive a similar expansion in the so-called Saha regime, where a non-vanishing fraction of charges is recombined into hydrogen atoms. That regime was introduced long ago [60] in the framework of the chemical picture. Assuming that the system is an ideal mixture of protons, electrons, and hydrogen atoms, its composition is then determined by applying the usual mass action law [26]. More recently, by starting from the physical description of the system in terms of a quantum plasma, it has been proved through successive works by Fefferman [27], Lieb et al. [18], Macris and Martin [45], that Saha approach is asymptotically exact in a scaling limit mixing the temperature and the chemical potential (see Sect. 2.2). As argued in Sect. 2.3, that limit defines quite diluted and low temperature conditions, namely the Saha regime, under which non-ideal contributions are small perturbations. In order to compute the corresponding contributions, we consider a formalism that combines the path integral representation of a quantum gas to familiar Mayer diagrammatics (see Sect. 2.4). Our key starting point in that framework is the so-called screened cluster expansion (SCE) of particle densities in terms of fugacities [8], which turns to be quite appropriate for studying recombined phases as illustrated in Refs. [10–12] (dielectric response of an atomic gas) or [9] (partial screening of van der Waals forces by free charges). The physical content of SCE is close to ideas first introduced by Rogers [54] for describing atomic or molecular recombination within the physical picture. In that approach, virial coefficients are numerically estimated within a priori modelizations, which incorporate quantum effects at short distances and classical Debye screening at large distances. The corresponding so-called ACTEX method has been developed through successive works [55, 56, 58, 59]. It has also been applied to hydrogen [57], with quite good results at low and moderate densities as described in Ref. [48]. Nevertheless, in the Saha regime, exact asymptotic expansions with analytical prescriptions for computing the successive terms, have not been derived within ACTEX method.

Exact Results for Thermodynamics of the Hydrogen Plasma

In Sect. 3, using the parametrization of chemical potential in terms of temperature introduced in Ref. [45], we show that every contribution in SCE of particle densities, decays exponentially fast when T goes to zero. Thanks to available inequalities for the spectrum of the considered Coulomb Hamiltonian, we extract the leading terms which do arise from free (ionized) protons and electrons, as well as from atoms H in their groundstate with energy EH = −me4 /(22 ) where m is the reduced mass m = mp me /(mp + me ). Next corrections are ordered with respect to their decay rate in the zero-temperature limit. They account for a large variety of physical effects: plasma polarization, formation of molecules H2 , ions H2+ and H − , interactions between ionized charges and Hydrogen atoms. Such corrections are defined without any ambiguity or a priori modelizations, so they do not depend on any adjustable phenomenological parameter. In particular, SCE provides well-behaved expressions for the partition functions of a molecule H2 or ions H2+ and H − in the vacuum, which are the generalizations of quantum virial functions for the hydrogen atom [23] to more complex entities. Notice also that ionic contributions to charge neutrality or screening are consistently incorporated, as it should. The equation of state (EOS) is derived by using thermodynamic identities in Sect. 4. This leads to our main result, namely scaled low-temperature (SLT) expansion of the pressure P around ideal Saha pressure βP /ρ ∗ = βPSaha /ρ ∗ +


βPk /ρ ∗ ,



considered as a function of the temperature and of the dimensionless density variable ρ/ρ ∗ where ρ is the electron number density (which is equal to the proton number density by neutrality). The temperature-dependent reference density ρ ∗ defined by ρ∗ =

exp(βEH ) 2(2πλ2pe )3/2

with λpe = (β2 /m)1/2 ,


determines the cross-over between full ionization for ρ  ρ ∗ , and full recombination for ρ  ρ ∗ (see Sect. 2.2). Since EH is negative, EH  −13.6 eV, ρ ∗ decays exponentially fast at low temperatures. In expansion (1.1), it is convenient to express the pressure in units of kB T ρ ∗ which turns out to be the natural reference pressure in the Saha regime. Then, each term in (1.1) is dimensionless. The first term is the usual Saha pressure expressed in terms of ρ/ρ ∗ βPSaha /ρ ∗ = ρ/ρ ∗ + (1 + 2ρ/ρ ∗ )1/2 − 1.


We see indeed that for ρ  ρ ∗ , the system becomes fully ionized (βPSaha ∼ 2ρ), whereas for ρ  ρ ∗ all ionized charges recombine into neutral hydrogen atoms (βPSaha ∼ ρ). Each term in expansion (1.1) beyond that leading ideal contribution has the form of a non-linear function of ratio ρ/ρ ∗ , times a temperature-dependent function hk (β) (or possibly a polynomial in the hl (β), l ≤ k). The hk (β) decay exponentially fast when T vanishes and are ordered with respect to their decay rates hk (β) ∼ e−βδk , 0 < δ1 < δ2 < · · ·. Hence the expansion (1.1) is organized as a series of exponential terms with increasingly faster exponential decay as T → 0. The hk -functions and their decay rates are governed by a balance between energy and entropy involving the ground-state energy EN(0)p ,Ne of Coulomb Hamiltonian HNp ,Ne for Np protons and Ne electrons in mutual interaction. We determine the pressure in the Saha regime by computing exactly all terms in expansion (1.1) smaller than leading ideal

A. Alastuey et al.

contribution βPSaha /ρ ∗ of order 1 by exponentially decaying factors of maximum order exp(βEH ). We find βPk /ρ ∗ = (function of ρ/ρ ∗ ) × hk (β),

hk (β) ∼ e−βδk , k = 1, 2, 3, 4,

βP5 /ρ ∗ = (function of ρ/ρ ∗ ) × [h1 (β)]2


(0) (0) (0) with (δk in electronvolt units, E2,1 = EH + , E1,2 = EH − , E2,2 = EH2 ) 2

δ1 = |EH |/2  6.8, δ2 = |3EH − EH2 |  9.1, δ3 = 3|EH |/4  10.2,


δ4 = |2EH − EH + |  11.0. 2

The explicit forms of the density-dependent functions and of the hk (β) can be found in Sects. 4.1 and 4.2 together with a discussion of their interpretation and relative importance in different density and temperature regimes. In short, the hk -functions incorporate various corrections to the ideal Saha term which can be described by • • • •

h1 (β): plasma polarization around ionized charges h2 (β): formation of H2 molecules and atom-atom interactions h3 (β): atomic excitations and interactions between ionized charges h4 (β): formation of H2+ and H − ions, atom-charge interactions, and screening of atomic groundstate

The construction of SLT expansion (1.1), defined by taking the zero-temperature limit at fixed ratio ρ/ρ ∗ , is itself an important new result. It provides a non-trivial structure for the various corrections to ideal Saha pressure, which are properly ordered in that scaling limit. It turns out that keeping only the first correction βP1 /ρ ∗ , is equivalent to a modification of the Saha ionization rate which has been derived previously by several authors (see e.g. [41] and references quoted in [39]). To our knowledge, next terms βPk /ρ ∗ (2 ≤ k ≤ 5) are entirely new, and do not have counterparts in the literature. We provide their exact expressions, which involve suitably truncated few-body partition functions. Two-body truncated partition functions are merely related to quantum virial functions first introduced by Ebeling (see e.g. [39] and references quoted therein). Three- and four-body truncated partition functions are introduced and defined here for the first time. Previous terms (2 ≤ k ≤ 5) also account, beyond standard calculations, for interactions between recombined entities as well as screening effects. For instance, contributions of atom-atom interactions in βP2 /ρ ∗ are evaluated without any a priori modelization, while screening of atomic groundstate embedded in βP4 /ρ ∗ incorporates further corrections to the familiar Debye shift. Corrections in SLT expansion (1.1) are ordered with respect to their decay rates when the temperature vanishes at fixed ratio ρ/ρ ∗ . The behavior of such corrections along a given low-temperature isotherm when the density is varied, displays some interesting physics. For very small densities ρ  ρ ∗ , all density-dependent functions in front of the h k s can be expanded in powers of ρ. Then, we retrieve the well-known virial expansion at fixed temperature up to order ρ 2 included (see Sect. 4.2). In particular, the leading correction of order ρ 3/2 is the familiar classical Debye term arising from the polarization contribution βP1 /ρ ∗ . When the density is increased, virial density-expansion can no longer be used, but SLT expansion still works and accounts for non-perturbative effects with respect to finite

Exact Results for Thermodynamics of the Hydrogen Plasma

values of ρ/ρ ∗ . Up to moderate densities ρ  ρ ∗ , βP1 /ρ ∗ remains the leading correction to ideal Saha terms. Interestingly, that polarization contribution is reduced at higher densities ρ  ρ ∗ because most protons and electrons are recombined into atoms H . Then, molecular contributions embedded in βP2 /ρ ∗ provide the first correction to Saha pressure, since they also overcome contributions of atom-atom interactions, at least for a sufficiently low temperature isotherm. Ultimately, they are responsible for the breakdown of expansion (1.1) at too large densities. Our results clearly provide a better analytical knowledge of the thermodynamics in an extended part of the phase diagram, as illustrated by the validity domain drawn in Fig. 12 of Sect. 4.3. The SLT expansions can be viewed as infinite resummations of low-density expansions. We emphasize again that the EOS (1.1) incorporates the screening effects in a coherent and consistent way for the whole range of densities ρ  ρ ∗ (strongly ionized gas) and ρ  ρ ∗ (recombined gas). When the interaction is Coulombic, one has to face the divergence of the sum of bound state contributions to the partition function of an isolated atom arising from the infinite number of Rydberg states. That important and well-known problem is usually dealt with the Planck-Larkin prescription to cut off states of energies En larger than kB T (see e.g. the discussion in [24]). In our implementation of the physical picture for the recombined phase, no divergence occurs since the partition function of the hydrogen atom appears naturally in a convergent truncated form, as a consequence of collective screening effects. Only that truncated partition function embedding both bound and ionized states is free from ambiguity. More comments about that point are offered in Sect. 3.2. Collective screening effects also give raise to well-behaved partition functions for more complex entities, like ions H − and H2+ , or molecules H2 . Such partition functions are naturally defined according to a truncation procedure similar to that introduced for the atomic partition function. They also involve contributions from both bound and dissociated states. The molecular partition function accounts thus not only for molecular bound states, but also for diffusion states made with two protons and two electrons. The finiteness of few-body truncated partition functions is of course crucial in the analysis of their low-temperature behaviors, which are shown to be controlled by Boltzmann factors exp(−βEH ), exp(−βEH − ), exp(−βEH + ), exp(−βEH2 ), associated with the corresponding recombined entities in their 2 groundstate (as would trivially be expected in a system with short range forces [25, 35]). Contributions from excited or diffusion states are well-defined in those truncated partition functions, and they may be neglected when the temperature is low enough. An exact treatment of screening in the many-body problem is also required to establish the correct classification of terms in the expansion (1.1) according to decaying exponentials. For instance, in addition to obvious contributions of atomic bound states, there are correction terms proportional to the inverse screening length κ (or powers of it), which is itself √ proportional to the square root of the density κ ∼ ρ. Since the latter is also exponentially small in the Saha regime (see Sect. 2.2), contributions of collective screening effects have to be compared to pure atomic terms, and may be predominant as exemplified by the first correction βP1 /ρ ∗ . Such systematic classification could not have been obtained without a unified theory which deals exactly with the interplay between screening effects and the other physical phenomena at stake (primarily the formation of atomic and molecular bound states). Though SLT expansions are asymptotic, i.e. a priori valid in the zero-temperature limit, they can be used for quantitative purposes within a rather wide range of temperatures and densities. We have performed numerical calculations, for both the EOS and the internal energy. For T of the order a few thousand kelvins, the hk (β)’s are not accurately reproduced by their simple low-temperature asymptotic forms: further contributions, which arise

A. Alastuey et al.

in particular from excited states of the recombined entities, must also be taken into account. Within a simple criterion on the order of magnitude of the various corrections to Saha pressure, we draw the validity domain of SLT expansion (1.1) in the plane (β, ρ) (see Fig. 12 in Sect. 4.3). That validity domain exemplifies the quantitative interest of our calculations, which can be applied to physical systems under observable conditions, like the Sun photosphere for instance. Furthermore, we have compared our findings to those of Militzer and Ceperley [48] obtained within Path Integral Monte Carlo (PIMC) simulations (PIMC methods have been implemented through successive works [15–17, 47, 49]). The agreement is satisfactory, as it should since PIMC results are computationally exact within statistical errors (see e.g. Ref. [40]). The detail of that comparison, as well as all our numerical results, will be presented in a forthcoming paper [1]. From a mathematical view point, all manifestations of screening stem from the screened potential introduced in Sect. 2.4 and studied in [13]. That potential can be viewed, in the quantum mechanical context, as the analogue of the classical Debye-Hückel potential. Because of its central role, we have devoted the long Appendix A: to a number of related properties which are used in our analysis. In Appendix B:, the low temperature behaviors of the truncated atomic, ionic and molecular partition functions mentioned above, are determined by methods using Green functions and path integral representations. In Appendix C:, we compute the contributions of interactions between atoms and ionized charges.

2 Saha Regime and Screened Cluster Expansion 2.1 Definition of the Model The hydrogen plasma is a two-component system made of protons and electrons. In the present non-relativistic limit, the proton and the electron are viewed as quantum point particles with respective charges, masses, and spins, ep = e and ee = −e, mp and me , σp = 1/2 and σe = 1/2. The kinetic energy operator for each particle of species α = p, e with position x takes the Schrödinger form −2 /(2mα )Δ where Δ is the Laplacian with respect to x. Two particles separated by a distance r interact via the instantaneous Coulomb potential v(r) = 1/r. The corresponding Coulomb Hamiltonian HNp ,Ne for Np protons and Ne electrons reads HNp ,Ne = −

N  1 2 Δi + eα eα v(|xi − xj |) 2mαi 2 i =j i j i=1


where N = Np + Ne is the total number of particles. In (2.1), the subscript i is attached to protons for i = 1, . . . , Np and to electrons for i = Np + 1, . . . , Np + Ne , so the species index αi reduces either to p or e while xi denotes either the position Ri of the i-th proton or the position rj of the j -th electron (j = i − Np ). The system is enclosed in a box with volume Λ, in contact with a thermostat at temperature T and a reservoir of particles that fixes the chemical potentials equal to μp and μe for protons and electrons respectively. Its grand-partition function Ξ is Ξ = Tr exp[−β(HNp ,Ne − μp Np − μe Ne )].


In (2.2), the trace is taken over all states symmetrized according to the Fermionic nature of each species; the boundary conditions for the wave functions at the surface of the box can be

Exact Results for Thermodynamics of the Hydrogen Plasma

chosen of the Dirichlet type. Lieb and Lebowitz [44] have proved that the thermodynamic limit (Λ → ∞ at fixed β and μα ) exists, thanks to Fermi statistics and screening. Indeed, the Fermionic statistics of at least one species implies the H -stability [21, 22] HNp ,Ne > −B(Np + Ne ),

B >0


that prevents the collapse of the system. On the other hand, screening ensures that it does not explode. In a fluid phase, the infinite system maintains local neutrality, i.e. the homogeneous local particle densities ρp and ρe for protons and electrons remain equal for any choice of the chemical potentials μα . In other words, the common particle density ρ = ρp = ρe , as well as all other bulk equilibrium quantities, depend on the sole combination μ = (μp + μe )/2,


and not on the difference ν = (μe − μp )/2. In particular, in terms of the fugacities zα = exp(βμα ), this means that both the density ρ and the pressure P are functions of only β and z = (zp ze )1/2 = exp(βμ). Therefore individual chemical potentials μe , μp are not uniquely determined: we can choose their difference ν at will without changing the bulk densities. Among the possible choices, it is particularly convenient to set 3 λe μp = μ − kB T ln , 2 λp

3 λe μe = μ + kB T ln 2 λp


where λα = (β2 /mα )1/2 is the thermal de Broglie wavelength of species α. This choice guarantees that Maxwell-Boltzmann densities of free (no interactions) proton and electron gases, respectively ρpid = 2zp /(2πλ2p )3/2 , ρeid = 2ze /(2πλ2e )3/2 ,


are identical, i.e. ρpid = ρeid = 2z/(2πλ2 )3/2


with λ = (λp λe )1/2 . The factors 2 in (2.6) accounts for spin degeneracy. The enforced neutrality of the ideal mixtures is equivalent to the linear relation 

eα zα /(2πλ2α )3/2 = 0



between the activities zp and ze , sometimes called the pseudo neutrality condition. That condition can be imposed without loss of generality when dealing with fugacity expansions in the grand canonical ensemble. As shown in Sect. 3, it considerably simplifies the analysis of diagrammatic series for the interacting system. If we consider other fugacities (zp , ze ) which do not satisfy the pseudo neutrality condition, an infinite number of graphs contributes to any term with a given order in low-density expansions. The calculations of those terms then become rather cumbersome. Nevertheless, beyond that technical complication, their final expression would be identical to that derived by starting with the above fugacities satisfying both condition (2.8) and zp ze = zp ze , in agreement with Lieb and Lebowitz proof [44].

A. Alastuey et al.

2.2 Rigorous Results at Low Density and Low Temperature We briefly recall the Saha theory in its simplest form. From the elementary view point of the thermodynamic of ideal substances, equilibrium ionization phases can be considered in the so-called chemical picture [26] as mixtures of noninteracting gases of electrons, protons, and hydrogen atoms, with chemical potential μat obeying the law of chemical equilibrium μat = μe + μp . According to (2.7) the corresponding densities of electrons and protons are 




(mp me )1/2 =2 2πβ2


3/2 exp(βμ),


exp(−β(EH − 2μ))


whereas the atomic density is id ρat =4

M 2πβ2

where M = mp + me is the atomic mass and the factor 4 is the number of spin states. Apart from the binding energy EH of the Hydrogen atom, all other effects of the Coulomb interaction are disregarded, so the Saha EOS is that of a mixture of perfect gases id . βPSaha = ρpid + ρeid + ρat


We see in (2.9) and (2.10) that, when μ = EH , all densities are of the same exponential order at low temperatures: this corresponds to the coexistence of ionized and atomic phases. It is appropriate to characterize the set of ionization equilibrium phases by a temperaturedependent chemical potential [45] μ = μ(β) = EH + kB T ln w


where w is a fixed parameter 0 < w < ∞. As shown from (2.9) and (2.10), that parameter determines the relative proportion of atoms to ionized charges through  3/4 id M γ ρat = 2w = , id ρp,e m 2


where we have introduced the equivalent parameter γ = 4(M/m)3/4 w. According to the above definitions, the fugacity z = exp(βμ) can be seen as parametrized by either w or γ at fixed temperature, i.e. z = w exp(βEH ) or  z=

m M

3/4 γ exp(βEH )/4.


For further purposes, it is convenient to consider the temperature dependent reference density ρ ∗ defined by (1.2) in the Introduction. Then we can rewrite ideal densities as ρpid = ρeid = ρ ∗ γ


and id = ρ∗ ρat

γ2 . 2


Exact Results for Thermodynamics of the Hydrogen Plasma id id In terms of γ , the proton (or electron) density ρ = ρpid + ρat = ρeid + ρat and the Saha pressure (2.11) respectively read

ρ=ρ and

γ2 γ+ 2

  γ2 . βPSaha = ρ ∗ 2γ + 2



Inversion of relation (2.17) provides γ , and hence the chemical potential μ, as a function of the reduced density ρ/ρ ∗ . Substitution of that function in (2.18) finally yields the Saha EOS (1.3) for the dimensionless pressure written as a function of ρ/ρ ∗ (note that our density variable is half of the total number density). As said in the Introduction, ρ ∗ is the cross-over density between full ionization and atomic recombination. The Saha picture has been rigorously justified from the statistical mechanics of the full interacting electron-proton gas in the following asymptotic sense. When the temperature goes to zero at fixed negative values of μ, the system obviously becomes highly dilute because all fugacities then vanish exponentially fast. If low temperatures favor recombination of electrons and protons into bound entities with negative ground state energies, on the contrary low densities favor dissociation. The chemical composition of the system will result of those two competing energy and entropy effects. That problem has been studied in a rigorous way by Fefferman [27], who proved the two following results using a refined version of the stability of matter (2.3) (see the discussion after (2.30) and Ref. [14] for a review). First, when β → ∞ with μ < EH (μ fixed), the pressure tends to that of an ideal mixture of protons and electrons with respective densities ρpid and ρeid (2.9) i.e. the system becomes fully ionized. Second, there exists some δ > 0 such that, when β → ∞ with EH < μ < EH + δ (μ fixed), the pressure tends to that of an ideal gas of hydrogen atoms in their groundstate with density (2.10). In that case, there is full atomic recombination. The previous discussion of the Saha EOS suggests that ionized protons, ionized electrons and Hydrogen atoms should coexist at μ = EH . This has been firmly settled by Lieb et al. [18] and also Macris and Martin [45] who proved that, when one introduces the temperature dependent chemical potential (2.12) and let β → ∞, the EOS tends to that of an ideal mixture of protons, electrons, and Hydrogen-atoms in their ground state, namely id βP = (ρpid + ρeid + ρat )[1 + O(exp(−β))] = βPSaha [1 + O(exp(−β))]


for β large enough and  > 0. The original work of Fefferman provides a power-law bound 1/β to the error term; that bound was improved to an exponential one in [18]. Thus we see that all ideal densities vanish exponentially fast, while corrections to ideal terms in (2.19) decay exponentially faster. The mathematical methods used in [27] and [18] are adequate to obtain a rigorous control of the dominant term (the Saha pressure), but apparently not adapted to explicitly calculate the corrections. In this work, it is our purpose to develop tools that enable to systematically compute those corrections, by expanding the pressure beyond the Saha term in an exact way (see (1.1)). In order to characterize the Saha regime in our study of the interacting system, we shall still use the parametrization (2.14) of the fugacity z associated with the zero-temperature limit, in terms of the parameter γ .

A. Alastuey et al.

2.3 Simple Physical Considerations about the Fugacity Expansion Saha equation of state (2.19) can be recovered, at a heuristic level, from simple considerations on low fugacities series for the pressure. Those considerations will serve as a guide to the analytic estimations of various non-ideal contributions to the full EOS at finite temperature and density performed in next sections. Low fugacity series are easily inferred, at a formal level, from the identity ln Ξ , (2.20) Λ where the thermodynamic limit Λ → ∞ is implicitly taken, as in the whole paper. They read  N βP = zp p zeNe BNp ,Ne (2.21) βP =

(Np ,Ne ) =(0,0)

where Mayer coefficients BNp ,Ne in (2.21) can be expressed as suitable traces, BNp ,Ne =

1 Tr[exp(−βHNp ,Ne )]Mayer . Λ


The first Mayer operators [exp(−βHNp ,Ne )]Mayer read [exp(−βH1,0 )]Mayer = exp(−βH1,0 ), [exp(−βH0,1 )]Mayer = exp(−βH0,1 ), [exp(−βH1,1 )]Mayer = exp(−βH1,1 ) − exp(−βH1,0 ) exp(−βH0,1 ), 1 [exp(−βH2,0 )]Mayer = exp(−βH2,0 ) − exp(−βH1,0 ) exp(−βH1,0 ), 2 ...,


while a similar expression holds for any [exp(−βHNp ,Ne )]Mayer [exp(−βHNp ,Ne )]Mayer = exp(−βHNp ,Ne ) − · · · .


In (2.24), terms left over reduce to a linear combination of products of Gibbs operators exp(−βHMp ,Me ) (Mp ≤ Np , Me ≤ Ne ) associated with all possible partitions of Np protons and Ne electrons. Traces (2.22) must be taken over Fermionic states which are products of anti-symmetrized states for each set of degrees of freedom associated with a Gibbs operator exp(−βHMp ,Me ). For instance, in space of positions and spins, B2,0 reads B2,0 =

1 Λ


dR2 [2 R1 R2 | exp(−βH2,0 )|R1 R2 

− 2 R1 | exp(−βH1,0 )|R1  R2 | exp(−βH1,0 )|R2  − R2 R1 | exp(−βH2,0 )|R1 R2 ].


For the term exp(−βH1,0 ) exp(−βH1,0 ) subtracted in [exp(−βH2,0 )]Mayer , each Gibbs operator exp(−βH1,0 ) refers to a single proton, so no anti-symmetrization occurs and only diagonal matrix elements of exp(−βH1,0 ) appear in (2.25). Truncated Mayer operators can also be expressed in terms of Ursell operators [32–34].

Exact Results for Thermodynamics of the Hydrogen Plasma

Despite Mayer coefficients BNp ,Ne diverge, leading contributions to the equation of state can be easily picked out in formal series (2.21), as follows. For (Np = 1, Ne = 0) and (Np = 0, Ne = 1), we obtain the simple exact expressions B1,0 =

2 (2πλ2p )3/2


B0,1 =

2 . (2πλ2e )3/2



After multiplication by fugacity factors exp(βμp ) and exp(βμe ) respectively, we obtain the related contributions to pressure (2.21) which reduce, of course, to the ideal MaxwellBoltzmann densities of ionized protons (ρpid ) and ionized electrons (ρeid ). For (Np = 1, Ne = 1), it is reasonable to expect that hydrogen atoms with internal ground state energy EH provide the leading low-temperature contribution which reads 4 exp(−βEH ), (2πλ2H )3/2


with λH = (β2 /M)1/2 while factor 4 is due to spin degeneracy. The corresponding contriid of Hydrogen atoms bution to (2.21) is nothing but the ideal Maxwell-Boltzmann density ρat in their ground state. In the Saha regime, ideal densities of ionized protons (ρpid ), ionized id ) are all found to be of the same order of magnitude electrons (ρeid ) and hydrogen atoms (ρat exp(βEH ) disregarding powers of β, because of (2.12). All other contributions to the EOS are expected to be small corrections to Saha pressure, as suggested by the following simple arguments and estimations. The Saha regime defines quite diluted conditions since ρ vanishes exponentially fast. Therefore, ionized charges and hydrogen atoms are expected to be weakly coupled and weakly degenerate. Let us introduce the various length and energy scales defined in Table 1, where we assume that each atom Table 1 Length and energy scales in a quantum hydrogen plasma Symbol


Physical signification


2 /(me2 )

Bohr radius


(β/mp,e,at )1/2

de Broglie lengths



Bjerrum length



Mean interparticle distance

κ −1

(4πβe2 [ρpid + ρeid ])−1/2 κ −1 | ln(κλ)|

Quantum screening distance

2 /a 3 e2 aB

Atom-atom interaction energy

H −c

e2 aB /a 2

Atom-charge interaction energy


e2 /a

Charge-charge interaction energy


kB T

Classical kinetic energy


|EH | = me4 /(22 )

Atom ground-state energy



Debye screening length

Energy H −H

A. Alastuey et al.

Fig. 1 Hierarchy of a length and b energy scales in the Saha regime

carries, roughly speaking, an instantaneous dipole of order eaB , while the physical signification of lQ is given in next Sect. 2.4. According to the hierarchies between those length and energy scales described in Fig. 1, both exchange and interaction contributions for ionized charges and hydrogen atoms should be exponentially smaller than above ideal terms. Similarly, we can estimate the contributions of complex entities which result from the quantum mechanical binding of Ne electrons and Np protons, i.e. the existence of a bound state in the spectrum of HNp ,Ne with negative ground state energy EN(0)p ,Ne . The ideal contribution of a given complex entity is easily extracted from BNp ,Ne , and it is of order exp(−βEN(0)p ,Ne ). After multiplication by fugacity factor exp[β(μp Np + μe Ne )], we find a contribution to pressure (2.21) which is of order exp[−β(EN(0)p ,Ne − (Np + Ne − 1)EH ] exp(βEH ),


where we have used parametrization (2.12) of the chemical potential. Ideal contribution (2.29) decays exponentially faster than exp(βEH ) for (Np , Ne ) = (1, 1), (1, 0), (0, 1), by virtue of inequality EN(0)p ,Ne − (Np + Ne − 1)EH > 0,

(Np , Ne ) = (1, 0), (0, 1), (1, 1)


which is a key ingredient in Fefferman’s proof [27]. Although not yet proved, that inequality is satisfied by known complex entities [28] as illustrated below. Of course, and as for ionized charges and hydrogen atoms, exchange and interactions contributions for complex entities should be smaller than ideal ones. Above heuristic arguments suggest that corrections to ideal Saha pressure (2.11) decay exponentially faster than leading terms when T vanishes, in agreement with the rigorous bound involved in (2.19). A precise evaluation of those corrections will be performed in Sect. 3 by using screened cluster expansion described in next Sect. 2.4. That method removes all long range Coulomb divergencies which plague Mayer coefficients BNp ,Ne . It provides well-defined recipes for computing contributions from both interactions and complex entities. The simplest entities which appear are the molecule H2 with groundstate en(0) (0) = EH2  −31.7 eV, ion H2+ with E2,1 = EH +  −16.2 eV, and ion H − with ergy E2,2 2

(0) E1,2 = EH −  −14.3 eV. Notice that such groundstate energies do satisfy inequality 2.30. For complex entities made with five or more particles, we will assume inequality

EN(0)p ,Ne > (Np + Ne − 2)EH ,

Np + Ne ≥ 5.


That inequality, more constraining than (2.30), is indeed satisfied by known stable complex (0) (0) = EH +  −36.5 eV and E2,3 = EH −  −28.5 eV are indeed entities. For instance, E3,2 3 2 larger than 3EH  −40.8 eV. Previous groundstate energies are computed within the method

Exact Results for Thermodynamics of the Hydrogen Plasma

Fig. 2 Geometrical representation of inequalities (2.30). Consider a line of slope μ < 0 which goes through (0) the point associated with the hydrogen atom. If all the points (Np + Ne , EN ,N ) associated with other p e entities lie above that line, the inequalities (2.30) hold for that value of μ, and the system tends to a dilute atomic gas in the limit β → ∞

described in [52]. The corresponding values are in excellent agreement with experiments and reported data in the literature. The resulting stability regimes of ionized, atomic and molecular phases are shown in Fig. 2. 2.4 Screened Cluster Expansion within Loop Formalism Screened cluster expansions are devised within an auxiliary classical system of charged loops. As first shown by Ginibre [30], Ξ is identical to the grand-partition function of a classical system made with loops. That transformation starts with the expression of Ξ in space of positions and spins, and use of Feynman-Kac formula [38, 53, 61, 62],

x 1 · · · x N | exp(−βHNp ,Ne )|x1 · · · xN  =

 N N  exp[−(x i − xi )2 /(2λ2αi )]  (2πλ2αi )3/2


 × 0



dsv(|(1 − s)(xi − xj ) + s(x i

 D(ξ i ) exp −

β eα eα 2 i =j i j

− x j ) + λαi ξ i (s) − λαj ξ j (s)|)

. (2.32)

A. Alastuey et al. Fig. 3 A loop L = (α, q, X, η(s)) made up of 5 particles

In the r.h.s. of (2.32), functional integrations are performed over Brownian bridges ξ i (s) (ξ i (0) = ξ i (1) = 0) with the normalized Gaussian measure D(ξ ) defined by its covariance (see (2.34) with q = 1). Each Brownian bridge ξ i (s) defines a path (1 − s)xi + sx i + λαi ξ i (s) associated with a given particle. A loop L is then defined as the collection of open paths associated with particles exchanged in a given permutation cycle. This leads to the identity [14, 19, 46] Ξ = Ξloop =

  ∞ N   1 D(Li )z(Li ) exp(−βV (Li , Lj )), N ! i=1 i

where fugacity z(Li ) and two-body potential V (Li , Lj ) are defined below. A loop L is characterized by its position X, species α = p, e and number q of exchanged particles, while its shape is defined by a closed Brownian path η(s) with s ∈ [0, q] and η(0) = η(q) = 0. Genuine particle positions in matrix elements of exp(−β(HNp ,Ne ) reduce to x(k) = X + λα η(k) with k integer, k = 0, . . . , q − 1 (x(0) = x(q) = X) (see Fig. 3). Phasespace measure D(L) is the product of discrete summations over α and q, spatial integration over X and functional integration over η(s) with normalized Gaussian measure D(η) defined by its covariance  D(η)ημ (s)ην (t) = δμν q inf(s/q, t/q)(1 − sup(s/q, t/q)). (2.34) Fugacity z(L) reads [14, 19] zαq 2 q (2πqλ2α )3/2     q βe2 q ˜ − t)v(|λα η(s) − λα η(t)|) (2.35) × exp − ds dt (1 − δ[s],[t] )δ(s 2 0 0 ∞ ˜ − t) = n=−∞ δ(s − t − n) is Dirac comb, while [s] ([t]) denotes the integer part where δ(s of s (t ). In (2.35), factor (1 − δ[s],[t] ) avoids counting point particle self-energy contributions, while Dirac comb ensures that only loop elements with equal times (modulo an integer) interact, an essential feature specific to quantum mechanics. Eventually, two-body potential V (Li , Lj ) reduces to  qi  qj ˜ − t)v(|Xi + λαi ηi (s) − Xj − λαj ηj (t)|). (2.36) ds dt δ(s V (Li , Lj ) = eαi eαj z(L) = (−1)q−1



At large distances, V behaves as the Coulomb potential between point charges qi eαi and qj eαj , i.e. V (Li , Lj ) ∼ qi qj eαi eαj /|Xi − Xj |. Therefore, usual Mayer diagrammatics for

Exact Results for Thermodynamics of the Hydrogen Plasma

Fig. 4 Order of magnitude of effective potential φ at various distances. In region λ < r < lQ , interactions are exponentially screened according to Debye potential. For r > lQ , φ is dominated by (unscreened) multipolar interactions

loops are plagued with long-range divergences. As in the case of classical Coulomb fluids, they are removed by summing infinite chains built with V . This amounts to replace V by screened potential φ which can be viewed as the quantum analog of Debye potential [13]. The explicit formula for the Fourier transform of φ is recalled in Appendix A:. Its spatial behaviors, according to the hierarchy of scales displayed in Fig. 1, are roughly summarized in Fig. 4, where only orders of magnitude of φ are given (we set r = |Xi − Xj | and omit all shape dependences which occur for r < λ or lQ < r). Notice that familiar exponential decay of φ breaks down at large distances r  lQ . The asymptotic dipolar behavior of φ is sufficient for ensuring that every graph built with φ is finite [2, 3]. As detailed in [8], so-called screened cluster expansion for protonic density ρp follows from an exact transformation of formal Mayer diagrammatics for loop density ρ(La ) which provides ρp =


1 S(G)

 D(Ca )ZφT (Ca )qa


D(Ci )ZφT (Ci )


(2.37) G

(a similar expression holds for ρe ). In (2.37), bare potential V is replaced by screened potential φ. Graphs G are identical to usual Mayer graphs, where points are now particle clusters, except for some specific rules (arising from the replacement of V by φ) which are (p) described below. Each cluster Ci (i = 0, . . . , n) contains Ni protons and Ni(e) electrons. The internal state of a cluster C(Np , Ne ) (C ∈ {Ci , i = 0, . . . , n}) is determined by Lp and (α) Le loops (L(α) 1 , . . . , LLα ) in which the Np protons and Ne electrons are distributed (in root (p) cluster C0 = Ca , L1 is identified to La which contains the root proton). Integration within phase space measure D(C) reduces to the sum over all possible distributions of particles (p) into loops combined with integrations over loop positions and shapes (with X1 = Xa fixed (p) T at the origin for loop L1 = La ). Statistical weight Zφ (C) for a cluster C(Np , Ne ) reads Lp ZφT (C)

(p) Le (e) (α) k=1 zφ (Lk ) k=1 zφ (Lk ) T Bφ ({Lk }), Np Ne n (q)! n (q)! p e q=1 q=1



where nα (q) is the number of loops containing q particles of species α (for Ca , np (qa )! is replaced by (np (qa ) − 1)!). Weight zφ (L) reduces to zφ (L) = z(L) exp[IR (L)]


with ring sum IR (L) given by IR (L) =

1 2

 D(L1 )z(L1 )βV (L, L1 )βφ(L1 , L).


A. Alastuey et al. T Truncated Mayer coefficient Bφ,N is defined by a suitable truncation of usual Mayer coeffiT remains cient Bφ,N for N loops with pair interactions φ. This truncation ensures that Bφ,N integrable over relative distances between loops when φ is replaced by V . First truncated Mayer coefficients are

T Bφ,1 = 1,

T Bφ,2 = exp(−βφ) − 1 + βφ −

β 2φ2 β 3φ3 + , 2! 3!



Bond Fφ (Ci , Cj ) can be either −βΦ, β 2 Φ 2 /2!, −β 3 Φ 3 /3!, where total potential Φ(Ci , Cj ) is the sum of pairwise interactions φ(L, L ) over loops L and L defining internal states of Ci and Cj respectively. As for ordinary Mayer diagrams, two clusters are connected by at most one bond, and graph G is connected. Here, symmetry factor S(G) is computed by permuting only clusters with identical numbers of protons and electrons. Moreover, for a cluster C different from Ca , the internal state of which is determined by a single loop L(α) 1 , when C is either, the intermediate cluster of a convolution (−βΦ) ∗ (−βΦ), or connected to the rest of the graph by a single bond β 2 Φ 2 /2!, expression (2.38) of its statistical weight must be replaced by (α) ZφT (C) = zφ (L(α) 1 ) − z(L1 ).


Eventually, summation in (2.37) involves only graphs G which are no longer integrable over relative distances between clusters {Ci , i = 0, . . . , n} when φ is replaced by V . Screened cluster expansion for the pressure is inferred from use of (2.37) in thermodynamics identities, as described in Sect. 4.

3 Estimations of Ideal and Non-Ideal Contributions to Fugacity Expansions of Particle Densities Now, we proceed to asymptotic estimations, in the Saha regime, of all contributions to ρ = ρp in screened cluster expansion (2.37). Every contribution to ρp is expressed, similarly to (2.15) and (2.16), as ρ ∗ times a power of γ , and times a dimensionless temperaturedependent function. This provides a formal representation of ρ/ρ ∗ in powers of γ , where the coefficients depend only on temperature (see (4.2)). At low temperatures, every coefficient decays exponentially fast. In Sects. 3.1–3.6, we select all contributions which are smaller than leading terms (2.15) and (2.16) (divided by ρ ∗ ) by exponentially decaying factors of maximum order exp(βEH ) (β → ∞). In Sect. 3.7, we show that all other contributions decay faster by factors exponentially smaller than exp(βEH ). Beyond leading ideal contributions of ionized protons (2.15) and hydrogen atoms (2.16) (which are recovered in Sects. 3.1 and 3.2), we determine first corrections arising from their mutual interactions id (Sects. 3.1, 3.4, 3.5 and 3.6). Such corrections are at most of order (ρp,e,at )2 , so they are smaller than leading terms by exponential factor exp(βEH ). We also study ideal-like contributions of recombined entities, molecules H2 , ions H − and H2+ (Sects. 3.3 and 3.6) which must be accounted for at that order. In the following, a graph with Np protons and Ne electrons will be obviously denoted GNp ,Ne (for Np + Ne > 1, there are several graphs with identical particle numbers).

Exact Results for Thermodynamics of the Hydrogen Plasma Fig. 5 Graphs representing simple entities: a ionized proton and b hydrogen atom

3.1 Ionized Proton and Plasma Polarization An ionized proton appears in graph G1,0 (see Fig. 5a) made with the sole root cluster Ca (p) containing a single proton. The internal state of Ca is defined by the sole protonic loop La with qa = 1. The contribution of G1,0 to (2.37) then reads   2zp D(ξ a )zφ (L(p) D(ξ a ) exp(IR (L(p) (3.1) a )= a )). (2πλ2p )3/2 (p)

We stress that collective effects are embedded in ring sum IR (La ). Thus, strictly speaking, G1,0 describes an ionized proton dressed by the surrounding plasma of ionized protons and electrons. Within the present framework, that dressing mechanism accounts for the familiar plasma polarization induced by an immersed charge. (p) In the Saha regime, ring sum IR (La ) can be evaluated by using the exact expression of φ (see Appendix A:). The corresponding asymptotic behavior can be easily recovered via the following simple estimation of convolution integral (2.40). At leading order, only terms q1 = 1 (α1 = p, e) need to be retained into D(L1 ). Moreover the integration over position (p) X1 is controlled by relative distances |X1 − Xa | of order κ −1 . At such distances, φ(L1 , La ) (p) can be replaced by its Debye form, while V (La , L1 ) merely reduces to eeα1 /|X1 − Xa |. This gives   βe2 κ 2 exp(−κ|X1 − Xa |) βe2 κ (p) dX1 D(ξ 1 ) , (3.2) ∼ IR (La ) ∼ 8π |X1 − Xa |2 2 in perfect agreement with the detailed analysis of Appendix A:. Since βe2 κ is small (see Fig. 1), dressing effects in (3.1) can be treated perturbatively by (p) expanding exp(IR (La )) in powers of IR . The resulting leading contribution of G1,0 reads  2zp 2zp D(ξ a ) = = ρpid = ρ ∗ γ , (3.3) (2πλ2p )3/2 (2πλ2p )3/2 where functional integration over ξ a merely reduces to 1 by normalization of Gaussian measure D(ξ a ). That leading term reduces to ideal Maxwell-Boltzmann density of ionized protons (2.15) (i.e. bare contribution zp B1,0 as it should). Taking into account (3.2), we find that first correction to (3.3) is rewritten as   2zp 2zp βe2 κ (p) 2 . (3.4) D (ξ )I ( L ) ∼ βe κ D(ξ a ) = ρpid a R a 2 3/2 2 3/2 (2πλp ) (2πλp ) 2 Contribution (3.4) involves a factor z3/2 , and hence a factor γ 3/2 times exp(3βEH /2). One factor exp(βEH ) may be absorbed into the prefactor ρ ∗ . It remains a factor exp(βEH /2), which multiplies the remaining part of the contribution. Therefore, we rewrite (3.4) as ρpid

βe2 κ = ρ ∗ γ 3/2 S3/2 (1, 0) exp(βEH /2), 2


A. Alastuey et al.

where we define screening function S3/2 (1, 0) =

(β|EH |)3/4 . π 1/4


Index 3/2 of screening function refers to the power of γ and (1, 0) to the single proton cluster. All subsequent contributions will be written according to the same prescription. First correction (3.4) accounts for familiar plasma polarization induced by a single proton (i.e. cluster (1, 0)) at lowest order. Its simple structure results from the almost classical and weakly coupled nature of the plasma mentioned in Sect. 2.3. Higher order collective corrections proportional to γ p (with p integer or half-integer), can be rewritten similarly to (3.4) via the definition of screening functions Sp (1, 0) which depend only on β. For instance, next correction to (3.4) merely reduces to ρ ∗ γ 2 S2 (1, 0) exp(βEH ), with S2 (1, 0) =


    2m 1/2 β|EH | (β|EH |)3/2 . − 1+ √ mp 8 2 π


The first contribution in the r.h.s. of (3.8) arises from the quadratic term in the expansion (p) of exp(IR (La )), while the second one arises from the linear term where loop-shape depen(p) dence of IR (La ) beyond classical form (3.3) is taken into account (see (A.11)). Further corrections are exponentially smaller than (3.7) as shown in Appendix A:. The dressing mechanism associated with plasma polarization occurs for any particle in all other graphs GNp ,Ne . At lowest order, every ring factor exp(IR ) can be replaced by 1, and first corrections are obtained by using (3.2). 3.2 Hydrogen Atom: Recombination and Dissociation Contributions A hydrogen atom is expected to appear in graph G1,1 made with single root cluster Ca (see Fig. 5b). The contribution of G1,1 reads    (e) T (p) D(ξ a )zφ (L(p) dX1 D(ξ 1 )zφ (L(e) a ) 1 )Bφ (La , L1 ) =

   4zp ze D (ξ ) exp(I (a)) dX D(ξ 1 ) exp(IR (1)) R 1 a (2πλ2 )3   β 2 φ 2 (a, 1) β 3 φ 3 (a, 1) + × exp(−βφ(a, 1)) − 1 + βφ(a, 1) − 2! 3! (p)


(with obvious simplified notations for the dependence of IR and φ on loops La and L(e) 1 ). (p) In (3.9), protonic loop La and electronic loop L(e) contain one proton (q = 1) and one a 1 electron (q1 = 1) respectively. Each of those particles are dressed like the ionized proton in G1,0 . Furthermore, their mutual interaction φ involves screening effects, which are also due to the surrounding plasma of ionized protons and electrons. At leading order, since φ(a, 1) reduces to V at finite distances |X1 − Xa | (see Appendix A:), BφT (a, 1) can be replaced by BT (a, 1) defined by (2.41) with V in place of φ. The resulting bare contribution of G1,1 , as well as first corrections due to collective effects, are successively estimated as follows.

Exact Results for Thermodynamics of the Hydrogen Plasma

3.2.1 Bare Contribution in the Vacuum The bare contribution of G1,1 reads   4zp ze dX1 D(ξ a )D(ξ 1 ) (2πλ2 )3   β 2 V 2 (a, 1) β 3 V 3 (a, 1) × exp(−βV (a, 1)) − 1 + βV (a, 1) − + . 2! 3!


In (3.10), functional integrations over shapes ξ a and ξ 1 can be exactly rewritten in terms of matrix elements of suitable operators by applying backwards Feynman-Kac formula (2.32). For the exponential factor in BT (a, 1), we obviously obtain  1 D(ξ a )D(ξ 1 ) exp(−βV (a, 1)) (2πλ2p )3/2 (2πλ2e )3/2 = Ra r1 | exp(−βH1,1 )|Ra r1 


with Ra = Xa and r1 = X1 . Functional integrations of powers of V (a, 1) in BT (a, 1), are related to the corresponding terms arising in Dyson expansion of Ra r1 | exp(−βH1,1 )|Ra r1  with respect to interaction part V1,1 of H1,1 . Moreover, let us introduce position R∗ = (mp Ra + me r1 )/M of the atom mass center, and one-body Hamiltonian Hpe of relative particle with position r∗ = r1 − Ra . Then, bare contribution (3.10) becomes ρ∗

γ2 Z(1, 1) exp(βEH ), 8


with (2πλ2H )3/2 Tr[exp(−βH1,1 )]TMayer Λ  = 4 dr∗ r∗ |[exp(−βHpe )]TMayer |r∗ ,

Z(1, 1) =


where [exp(−βHpe )]TMayer stands for truncated Mayer operator [exp(−βHpe )]TMayer = exp(−βHpe ) − exp(−βKpe )  β + dτ1 exp[−(β − τ1 )Kpe ]Vpe exp[−τ1 Kpe ] 








dτ1 0

dτ2 exp[−(β − τ1 )Kpe ]Vpe exp[−(τ1 − τ2 )Kpe ]Vpe exp[−τ2 Kpe ]



dτ2 0


dτ3 exp[−(β − τ1 )Kpe ]Vpe exp[−(τ1 − τ2 )Kpe ]


× Vpe exp[−(τ2 − τ3 )Kpe ]Vpe exp[−τ3 Kpe ] (Kpe = −2 Δ/(2m) and Vpe = −e2 /r).


A. Alastuey et al.

Partition function (3.13) is similar to the so-called direct quantum virial function first introduced by Ebeling [23] (see Sect. 4.1). It incorporates contributions from both bound states (recombination of proton and electron into an hydrogen atom) and diffusion states (dissociation of an hydrogen atom into ionized proton and electron). Contrary to the trace of [exp(−βHpe )]Mayer , Z(1, 1) is finite because r∗ |[exp(−βHpe )]TMayer |r∗  decays as 1/(r ∗ )4 at large distances. Though truncation in [exp(−βHpe )]TMayer can be traced back to collective screening effects, Z(1, 1) depends only on temperature, and no longer on density. In order to estimate (3.13) at low temperatures, we can heuristically extend the very simple argument used in Section 2.3 for estimating B1,1 . For r ∗ ∼ aB , contribution of ground state ψ0 (r ∗ ) of Hpe to r∗ | exp(−βHpe )|r∗  exponentially dominates all other contributions because of the finite gap between EH and the rest of the spectrum. Moreover, truncated terms in [exp(−βHpe )]TMayer , which are crucial for ensuring the finiteness of the trace, do not generate exponentially growing terms at low temperatures, because they only involve Gibbs operators associated with kinetic Hamiltonian Kpe . Therefore, the leading behavior of (3.13) when β → ∞, obtained by replacing r∗ |[exp(−βHpe )]T |r∗  by |ψ0 (r∗ )|2 exp(−βEH ), merely is Z(1, 1) ∼ 4 exp(−βEH ).


Beyond the previous heuristic argument, we present in Appendix B: a non-perturbative derivation of (3.15), which is quite useful for further purposes (see Sects. 3.3 and 3.7) since it provides convincing low-temperature estimations of quantities similar to (3.13) involving three or more particles. Eventually, according to formula (3.12), the leading bare contribution of G1,1 reads ρ∗

γ2 , 2


which is nothing but ideal contribution (2.16) of hydrogen atoms in their groundstate. Beyond leading term (3.16), the rest of the bare contribution of G1,1 can be rewritten as ρ∗

γ2 Zexc (1, 1) exp(βEH ), 8


with Zexc (1, 1) = Z(1, 1) − 4 exp(−βEH ). At low temperatures, leading contribution to (3.17) arises from the first excited level (EH(1) = EH /4) of the hydrogen atom and reads 

 3βEH 2ρ γ exp . 4 ∗



It can be viewed as the ideal density of hydrogen atoms in their first excited state. As expected, that level is less populated than the ground state by exponentially decaying Boltzmann factor exp(3βEH /4) associated with energy difference EH(1) − EH = −3EH /4 (apart from the trivial factor 4 arising from orbital degeneracy of the first excited state). If the identification of atomic states contributions (like (3.16) or (3.18)) makes sense in the zero-temperature limit defining Saha regime, at finite temperatures the definition of an atomic part ZH in Z(1, 1) is arbitrary, as it has been noticed for a long time (see Ref. [39] and references quoted therein). That ambiguity is related to the fact that contributions of (p) bound states with |EH | ≤ kB T cannot be disentangled from that of diffusion states since they have the same order of magnitude. A possible definition of ZH is a finite sum of terms

Exact Results for Thermodynamics of the Hydrogen Plasma (p


analogous to (3.18) up to pmax such that |EH max |  kB T : that procedure accounts for expected thermal ionization which prevents the existence of highly excited hydrogen atoms in so-called Rydberg states. As emphasized in Ref. [39], only the full contribution embedded in Z(1, 1), obviously independent of above arbitrariness, is relevant for thermodynamics. Notice that diffusion state contributions describe (unscreened) short-distance interactions between ionized proton and electron. Such contributions are similar to that involved in G2,0 (see Sect. 3.3), and they are smaller than ideal contribution (3.16) by exponential factor exp(βEH ) apart from powers of β. 3.2.2 Collective Corrections The first contributions of G1,1 due to collective effects are obtained by expanding, in (3.9), ring factors exp(IR (a)) and exp(IR (1)) in powers of IR (a) and IR (1), and Mayer coefficient BφT (a, 1) in powers of (φ − V )(a, 1). At lowest order, IR (a) and IR (1) behave as βe2 κ/2, while (φ − V )(a, 1) behaves as e2 κ at distances r < βe2 . Therefore, first polarization corrections, which are smaller than leading bare contribution (3.16) by an extra factor βe2 κ, cancel out: an Hydrogen atom, which is a neutral entity, does not polarize its surrounding plasma at lowest order. Collective corrections to (3.16) are then determined by the behavior of IR and (φ − V ) beyond the previous simple constants. In other words, the bare proton-electron Coulomb potential is modified, beyond the familiar Debye shift, by a coupling between quantum fluctuations of both particles and the surrounding plasma. That effect cannot be incorporated into an effective potential. The corresponding calculation, performed in Appendix A:, gives at lowest order, ρ ∗ γ 3 S3 (1, 1) exp(2βEH ),


where screening function S3 (1, 1) for cluster (1, 1) is given by (A.16). Contribution (3.19) is exponentially smaller than (3.16) by factor exp(βEH ) and must be retained at that order, because S3 (1, 1) behaves as a power of β times exp(−βEH ) (see (A.17)). As shown in Appendix A:, higher order collective corrections decay exponentially faster than (3.19). 3.3 Other Complex Entities 3.3.1 Two-Proton Cluster A two-proton cluster is described by graph G2,0 made with single root cluster Ca (see Fig. 6a). There are two possible loop configurations for the internal state of root cluster (p) (p) Ca : either the two protons belong to two different loops La and L1 , or they belong to a (p) single loop La . The corresponding contribution reads     (p) (p) D(ξ a )zφ (L(p) dX1 D(ξ 1 )zφ (L1 )BφT (L(p) D(ηa )2zφ (L(p) a ) a , L1 ) + a ) =


D(ξ a )D(ξ 1 ) exp(IR (a)) exp(IR (1)) (2πλ2p )3   β 2 φ 2 (a, 1) β 3 φ 3 (a, 1) + × exp(−βφ(a, 1)) − 1 + βφ(a, 1) − 2! 3!  2zp2 D(ηa ) exp(IR (a)) exp(−βU (a)). − (4πλ2p )3/2



A. Alastuey et al.

Fig. 6 Graphs representing various complex entities: a two-proton cluster; b molecule H2 ; c ion H2+ ; d ion H −

Like (3.9), (3.20) incorporates collective effects, i.e. dressing of each proton and screening of their mutual interactions. At leading order, after applying backwards Feynman-Kac formula (2.32), we find that the bare contribution of Fig. 6a reduces to   γ 2 m 3/2 Z(2, 0) exp(βEH ), ρ∗ √ 2 mp


with Z(2, 0) =

(πλ2p )3/2 



Tr[exp(−βH2,0 )]TMayer

dr∗ {2 r∗ |[exp(−βHpp )]TMayer |r∗  − −r∗ | exp(−βHpp )|r∗ }.


In (3.22), Hpp is the one-body Hamiltonian of relative particle with position r∗ = R1 − Ra and mass mpp = mp /2, Hpp = Kpp + Vpp with Kpp = −2 Δ/(2mpp ) and Vpp = e2 /r. Moreover, [exp(−βHpp )]TMayer is defined as (3.14) with Kpp and Vpp in place of Kpe and Vpe respectively. Like Z(1, 1), Z(2, 0) is also merely related to Ebeling quantum virial functions (see Sect. 4.1). Thanks to truncation in [exp(−βHpp )]TMayer , the integral over r∗ does converge contrary to the integral in (2.25) that formally defines B2,0 . Because of the continuous nature of the spectrum of Hpp which starts at zero, Z(2, 0) behaves as a power law at low temperatures. Contribution (3.21) then decays faster than ρ ∗ by exponential factor exp(βEH ) (discarding powers of β). Collective corrections to (3.21) arise from expansions of ring factors and of Mayer coefficient in (3.20). At lowest order, we can use IR (a) ∼ IR (1) ∼ βe2 κ/2 and (φ − V )(a, 1) ∼ −e2 κ (r < βe2 ) for qa = q1 = 1, while IR (a) ∼ 2βe2 κ for qa = 2. Therefore, the first polarization correction to (3.21), which can be treated at a purely classical level, is smaller than ρ ∗ by factor exp(3βEH /2). 3.3.2 Molecule H2 Contribution of a molecule H2 is embedded in graph G2,2 made with the single root cluster Ca containing two protons and two electrons (Fig. 6b). Again, dressing of particles as well as screening of their mutual interactions can be treated perturbatively in the Saha regime. At leading order, the resulting bare contribution of G2,2 is then transformed into √  3/2 2 m ρ γ Z(2, 2) exp(3βEH ), 32 M ∗



Exact Results for Thermodynamics of the Hydrogen Plasma

with Z(2, 2) =

(2πλ2H2 )3/2

Tr[exp(−βH2,2 )]TMayer  = (2πλ2H2 )3/2 dR1 dr1 dr2 {4 Ra R1 r1 r2 | exp(−βH2,2 )|Ra R1 r1 r2  Λ

− 2 R1 Ra r1 r2 | exp(−βH2,2 )|Ra R1 r1 r2  − 2 Ra R1 r2 r1 | exp(−βH2,2 )|Ra R1 r1 r2  + R1 Ra r2 r1 | exp(−βH2,2 )|Ra R1 r1 r2  + · · ·}


(λH2 = (β2 /(2M))1/2 ). Like (3.14), truncated Mayer operator [exp(−βH2,2 )]TMayer is defined as a suitable truncation of [exp(−βH2,2 )]Mayer inherited from the structure of coeffiT (N = 1, 2, 3, 4). In addition to the terms already present in [exp(−βH2,2 )]Mayer , cients Bφ,N that truncation involves products of imaginary-time evolutions of interaction potentials between subsets of two protons and two electrons (for our purpose, it is not necessary to detail here all the numerous terms involved in that truncation). This ensures that [exp(−βH2,2 )]TMayer has a finite trace contrary to [exp(−βH2,2 )]Mayer . Similarly to (3.13), partition function (3.24) incorporates contributions from both recombination into molecules H2 , and dissociation (interactions at short distances between atoms H , ions H2+ , ions H − , ionized protons and ionized electrons). At low temperatures, the leading behavior of Z(2, 2) is determined by applying the method described in Appendix B:. A key ingredient is the discrete nature of the spectrum of H2,2 (discarding the trivial contribution of the center of mass) near its infimum. Moreover, we assume quite weak bounds for three- and four-body Coulomb Green functions, inspired in part from their known exact two-body counterparts [36]. Then, we show that leading contribution to Z(2, 2) arises from the first four terms in the r.h.s of (3.24) evaluated for the ground state of mole(0) . Thus, despite truncated terms beyond matrix elements of cule H2 with energy EH2 = E2,2 exp(−βH2,2 ) not written explicitly in the r.h.s. of (3.24), are crucial for ensuring finiteness of Z(2, 2), they do not affect its leading low-temperature behavior which merely reads Z(2, 2) ∼ exp(−βEH2 )


when β → ∞. Since H2 contains two protons, the resulting contribution (3.23) is twice ideal density ρHid2 of molecules H2 in their para-groundstate where the two protons, as well as the two electrons, have opposite spin orientations, while the total angular momentum is zero. First thermal corrections to (3.25) arise from molecular excited states. Contrarily to the atomic case, such states are not exactly known. However, according to the usual phenomenology, they are expected to be well described by para-states and ortho-states (the two protons have the same spin orientation) with non-zero angular momenta describing global rotations of the molecule [42]. Moreover, excited states with still higher energies can be associated with proton vibrations and ultimately electronic excitations [42]. Beyond above purely molecular terms, Z(2, 2) also incorporates short-range contributions which account for interactions between products of molecular dissociation, as well as the corresponding exchange effects. Similarly to the case of Z(1, 1) where atomic contributions are mixed to those of interactions between ionized-charges, the extraction of either a molecular part ZH2 or an atom-atom contribution in Z(2, 2), remains arbitrary. Again, that arbitrariness does not cause any trouble for thermodynamics which depend only on the full contribution Z(2, 2).

A. Alastuey et al.

Collective corrections to (3.23) embedded in G2,2 can be studied as above (see Sect. 3.2). Like atom H , molecule H2 is neutral so it does not polarize (at lowest order) the surrounding plasma. First collective corrections are then smaller than (3.25) by an extra factor (βe2 κ)2 of order exp(βEH ). Therefore, they are smaller than ρ ∗ by a factor exp[β(4EH − EH2 )], which is itself exponentially smaller than exp(βEH ) by virtue of inequality 3EH < EH2 . 3.3.3 Ions H − and H2+ Ions H2+ and H − appear in graphs G2,1 (Fig. 6c) and G1,2 (Fig. 6d) respectively. The corresponding bare contributions are rewritten as 3

me (M + mp ) ρ 16 M2 ∗γ

and ρ∗

3/2 Z(2, 1) exp(2βEH )


  γ 3 mp (M + me ) 3/2 Z(1, 2) exp(2βEH ), 32 M2


with Z(2, 1) =

(2πλ2H + )3/2

Tr[exp(−βH2,1 )]TMayer


(2πλ2H − )3/2 Tr[exp(−βH1,2 )]TMayer Λ




and Z(1, 2) =

(λH + = (β2 /(M + mp ))1/2 and λH − = (β2 /(M + me ))1/2 ). Truncated Mayer operators 2 [exp(−βH2,1 )]TMayer and [exp(−βH1,2 )]TMayer are defined similarly to [exp(−βH2,2 )]TMayer and [exp(−βH1,1 )]TMayer . The low-temperature behaviors of (3.26) and (3.27) are determined by applying the method described in Appendix B:. As for (3.13) and (3.24), truncated terms beyond exp(−βH2,1 ) or exp(−βH1,2 ) do not contribute at leading order. Therefore, we find that (3.26) behaves as ρ∗

  γ 3 me (M + mp ) 3/2 exp[β(2EH − EH + )] = 2ρHid+ , 2 2 8 M2


(0) where ρHid+ is the ideal density of ions H2+ in their groundstate with energy EH + = E2,1 , 2


which is doubly degenerated because of electron spin. Similarly, we obtain leading behavior of (3.27), i.e. 3

mp (M + me ) ρ 16 M2 ∗γ

3/2 exp[β(2EH − EH − )] = ρHid− ,


(0) where ρHid− is the ideal density of ions H − in their groundstate with energy EH − = E1,2 , which is doubly degenerated because of proton spin. Like (3.23), those ideal contributions decay exponentially faster than ρ ∗ in the Saha regime. Density effects embedded in G2,1 and G1,2 are similar to those encountered above for an ionized proton. They provide contributions which are smaller than ρ ∗ by factors exp(β(5EH /2 − EH + )) and 2 exp(β(5EH /2 − EH − )), while such factors are themselves exponentially small compared

Exact Results for Thermodynamics of the Hydrogen Plasma

to exp(βEH ) by virtue of inequalities 3EH /2 < EH + and 3EH /2 < EH − (see numerical 2 values given in Sect. 2.3). 3.4 Interactions between Ionized Charges beyond Polarization Effects Since the Saha regime is quite diluted and weakly coupled (see Sect. 2.3), leading contributions of screened interactions are embedded in the polarization mechanism described in Sect. 3.1 for a graph with a single particle. This provides well-known Debye correction (3.4). Beyond that mean-field contribution, next contributions of interactions between ionized charges arise from graphs involving two particles, namely G1,1 and G2,0 shown in Figs. 5b, 6a and 7a–c. As quoted above, graphs made with one cluster (Figs. 5b and 6a) involve contributions of unscreened interactions at short distances. Graphs made with two clusters, Ca (one proton) and C1 (one proton or one electron), connected by a single bond Fφ (Ca , C1 ) which can be either −βΦ (Fig. 7a), β 2 Φ 2 /2! (Fig. 7b), or −β 3 Φ 3 /3! (Fig. 7c), account for large-distance screened contributions which are estimated as follows. Graphs shown in Fig. 7a (with α = p, e) provide contribution   (p) (p) (e) (e) (p) (p) −β dX1 D(ξ a )D(ξ 1 )zφ (L(p) a )[zφ (L1 )φ(La , L1 ) + zφ (L1 )φ(La , L1 )]   4βzp dX D(ξ a )D(ξ 1 ) exp(IR (L(p) 1 a )) (2πλ2p )3/2  zp (p) (p) × exp(IR (L1 ))φ(L(p) a , L1 ) (2πλ2p )3/2  ze (e) (e) (p) + exp(IR (L1 ))φ(La , L1 ) . (2πλ2e )3/2





The expansion of ring factors exp(IR (La )) and exp(IR (L1 )) provides a first contribution which vanishes by virtue of identity (A.3) derived in Appendix A:. The first non-vanishing (p) contribution arises from linear terms IR (L1 ) and IR (L(e) 1 ) where loop-shape dependences beyond classical behavior (3.2) are included. At lowest order, φ can then be replaced by its classical Debye form, and the resulting leading contribution of Fig. 7a is −ρ ∗ with S2 (0, 1) =

γ2 [S2 (1, 0) − S2 (0, 1)] exp(βEH ) 2


    (β|EH |)3/2 2m 1/2 β|EH | , − 1+ √ me 8 2 π


Fig. 7 Graphs describing screened interactions between one proton and one electron (α = e), or between two ionized protons (α = p)

A. Alastuey et al.

which follows from (A.11). Further contributions decay exponentially faster than ρ ∗ exp(βEH ). Because weight of cluster C1 has specific form (2.42), contribution of Fig. 7b reads β2 2





2 (p) D(ξ a )D(ξ 1 )zφ (L(p) a )[(zφ (L1 ) − z(L1 ))φ (La , L1 )

(e) (e) 2 (p) + (zφ (L(e) 1 ) − z(L1 ))φ (La , L1 )]   2β 2 zp dX1 D(ξ a )D(ξ 1 ) exp(IR (L(p) = a )) (2πλ2p )3/2  zp (p) (p) × (exp(IR (L1 )) − 1)φ 2 (L(p) a , L1 ) (2πλ2p )3/2  ze (e) (e) 2 (p) + (exp(I ( L )) − 1)φ ( L , L ) . R a 1 1 (2πλ2e )3/2



At lowest order, we can replace factors (exp(IR (L1 )) − 1) and (exp(IR (L(e) 1 )) − 1) by βe2 κ/2 on one hand, and φ by its classical Debye form on another hand. This provides the leading contribution of (3.35) (ρpid )2

β 3 e6 κ 2


exp(−2κ|X1 − Xa |) |X1 − Xa |2

= π(ρpid )2 β 3 e6 = ρ ∗ γ 2 [W (1, 0|1, 0) + W (1, 0|0, 1)] exp(βEH )


with W (1, 0|1, 0) = W (1, 0|0, 1) =

(β|EH |)3/2 , √ 4 π


in agreement with asymptotic formula (A.6) derived in Appendix A:. Functions W can be interpreted as resulting from effective interactions between ionized charges generated by quadratic fluctuations of φ. Next corrections to (3.36) decay exponentially faster than ρ ∗ exp(βEH ), as inferred from (A.6) and (A.11). Eventually, contribution of Fig. 7c is −β 3 3!






3 (p) 3 (p) D(ξ a )D(ξ 1 )zφ (L(p) a ) zφ (L1 )φ (La , L1 ) + zφ (L1 )φ (La , L1 )

  2β 3 zp dX D(ξ a )D(ξ 1 ) exp(IR (L(p) 1 a )) 3(2πλ2p )3/2  zp (p) (p) × exp(IR (L1 ))φ 3 (L(p) a , L1 ) (2πλ2p )3/2  ze (e) (e) 3 (p) + exp(IR (L1 ))φ (La , L1 ) , (2πλ2e )3/2



Exact Results for Thermodynamics of the Hydrogen Plasma (p)


with qa = q1 = 1, ηa = ξ a and η1 = ξ 1 . All collective effects can be omitted in (3.38) at lowest order, so leading contribution reads −ρ ∗ γ 2

cp (β|EH |)3/2 exp(βEH ) 12π 3/2


with numerical constant cp given by (A.8). Next corrections to (3.39) decay exponentially faster than ρ ∗ exp(βEH ). 3.5 Interactions between an Atom and an Atom or an Ionized Charge As argued in Sect. 2.3, atoms H are expected to be weakly coupled under Saha conditions, like ionized charges (see Sects. 3.1 and 3.4). Leading contributions of interactions between atoms and ionized charges should then involve either two atoms or a single one. As quoted in Sect. 3.3, short-range parts of those contributions are embedded in Figs. 6b–d made with a single cluster. Here, we consider other graphs made with two clusters which account for complementary parts including long-range effects. 3.5.1 Atom-Atom Interactions Figures 8a–c describe interactions between two atoms. Contrary to the case of ionized charges, screening effects can now be omitted at leading order, because each atom is neutral. In other words, potential Φ(Ca , C1 ) between clusters Ca and C1 can be replaced by its bare counterpart V (Ca , C1 ), which decays as a dipolar interaction (the corresponding 1/R 3 decay is sufficient for ensuring integrability in Fig. 8a for symmetry reasons). Of course, in statistical weights defining internal states of Ca and C1 , collective effects can be also ignored at leading order. Then contribution of Fig. 8a vanishes by symmetry, while the resulting bare contributions of Figs. 8b and 8c can be rewritten in terms of matrix elements of Gibbs operators by applying backwards Feynman-Kac formula (2.32). Leading contribution of Fig. 8b reads

 zp2 ze2 dR1 dr1 dr2 16 Ra R1 r1 r2 | 



dτ1 0


dτ2 exp[−(β − τ1 )(Hat + Hat )]Vat,at


× exp[−(τ1 − τ2 )(Hat + Hat )]Vat,at

 × exp[−τ2 (Hat + Hat )]|Ra R1 r1 r2  + · · · ,


where Hat = H1,1 is the Hamiltonian of a single atom, while Vat,at is the interaction potential between two atoms. Terms · · · in (3.40) have a structure analogous to those subtracted

Fig. 8 Graphs accounting for interactions between two hydrogen atoms

A. Alastuey et al.

from exp(−βHpe ) in (3.14). The corresponding truncation, inherited from that in the BT ’s, is analogous to that defining individual atomic partition functions: it ensures that spatial integration over R1 , r1 , r2 does converge. An expression similar to (3.40) can be obtained for Fig. 8c. Full bare contribution of Figs. 8b and 8c takes the form (see Appendix C:) ρ ∗ γ 4 W (1, 1|1, 1) exp(3βEH ),


discarding terms which decay exponentially faster than ρ ∗ exp(βEH ). When β → ∞, W (1, 1|1, 1) behaves as W (1, 1|1, 1) ∼

cat,at exp(−2βEH ) 32π 3/2 (β|EH |)1/2


where cat,at is the pure numerical coefficient (C.1). Function W (1, 1|1, 1) accounts for unscreened interactions between two hydrogen atoms in their groundstate. Contributions from both short and large separations R are involved. In particular, contributions from familiar van der Waals interactions UH −H (R) = −AH −H /R 6 (with positive constant AH −H computed from quantum perturbation theory at zero temperature [42]), do emerge through the large-distance (R = |R∗2 − R∗1 |  λH ) behavior 


Ra R1 r1 r2 |

dτ1 0


dτ2 exp[−(β − τ1 )(Hat + Hat )]Vat,at


× exp[−(τ1 − τ2 )(Hat + Hat )]Vat,at exp[−τ2 (Hat + Hat )]|Ra R1 r1 r2  ∼−

exp(−2βEH ) |ψ0 (r1∗ )|2 |ψ0 (r2∗ )|2 βUH −H (|R∗1 − R∗2 |), (2πλ2H )3


for spatial configurations r1∗ ∼ r2∗ ∼ aB and sufficiently low temperatures. Collective corrections to (3.41) are exponentially smaller than its leading behavior. Notice that they arise from various effects: plasma polarization associated with ring factors exp(IR ), Debye exponential screening of interactions at scales κ −1 , and also modification of 1/R 6 -tails at distances larger than lQ as detailed elsewhere [9]. 3.5.2 Atom-Proton and Atom-Electron Interactions Figs. 9a–f account for interactions between one atom H and a single ionized charge. Leading contribution of Fig. 9a (obtained by replacing ring factors by 1) vanishes by virtue of

Fig. 9 Graphs accounting for interactions between a hydrogen atom and an ionized charge (α = p or e)

Exact Results for Thermodynamics of the Hydrogen Plasma

identity (A.3). Like Fig. 8a, bare contribution of Fig. 9d also vanishes for symmetry reasons. Therefore contributions of Figs. 9a and 9d decay faster than ρ ∗ exp(βEH ). Figs. 9e and 9f provide contributions obviously identical to that of Figs. 9b and 9c with α = p. In Figs. 9b and 9c, all collective effects can be neglected at leading order, in particular φ(Ca , C1 ) can be replaced by bare potential V (Ca , C1 ). Within that substitution, integrability at large distances R between Ca and C1 , is obviously ensured thanks to dipole-charge 1/R 2 decay of V (Ca , C1 ). The resulting bare contributions of above graphs are rewritten in terms of matrix elements of Gibbs operators similarly to (3.40). For instance, bare contribution of Fig. 9b with α = p reads  zp2 ze


dR1 dr1 {8 Ra R1 r1 |


dτ1 0

dτ2 exp[−(β − τ1 )(Hat + Hp )]Vat,p


× exp[−(τ1 − τ2 )(Hat + Hp )]Vat,p exp[−τ2 (Hat + Hp )]|Ra R1 r1  + · · ·} (3.44) where Hp = H1,0 is the Hamiltonian of a single proton, while Vat,p is the total interaction potential between an atom and a proton. Like in (3.40), terms · · · in (3.44) have a structure analogous to those subtracted from exp(−βHpe ) in (3.14), which ensures spatial integrability over R1 and r1 . Bare contributions of the other considered graphs can be expressed similarly to (3.44). As shown in Appendix C:, the full bare contribution of Figs. 9a–f behaves as ρ ∗ γ 3 [2W (1, 1|1, 0) + W (1, 1|0, 1)] exp(2βEH ),


plus terms which decay exponentially faster than ρ ∗ exp(βEH ) when β → ∞. Functions W (1, 1|1, 0) and W (1, 1|0, 1) account for unscreened interactions between an atom in its groundstate and an ionized charge. Their low-temperature behaviors are cat,p exp(−βEH ), 16π 3/2 (β|EH |)1/2 cat,e W (1, 1|0, 1) ∼ exp(−βEH ) 16π 3/2 (β|EH |)1/2 W (1, 1|1, 0) ∼


where cat,α are pure numerical constants (C.3). Long-range contributions to W (1, 1|1, 0) and W (1, 1|0, 1) do reduce to that of the attractive interactions UH −α (R) = −AH −α /R 4 between an atom H and an ionized charge, with positive constant AH −p = AH −e computed within quantum perturbation theory at zero temperature. First collective corrections result from plasma polarization by the considered ionized charges, and they reduce to (3.45) multiplied by simple factor βe2 κ/2 of order exp(βEH /2). As for atom-atom interactions, part of further density corrections result from screening of atom-proton or atom-electron interactions at large distances. 3.6 Interactions between an Ionized Proton and Charged Clusters Screened interactions between an ionized proton and charged clusters are embedded in any graph made with a root cluster Ca containing a single proton connected to a charged cluster C1 via a bond −βΦ. As shown below, such a graph provides a contribution which behaves, at leading order, as that of the part connected to Ca through C1 . Moreover, that mechanism enforces charge neutrality (ρp = ρe ) by symmetrizing protonic and electronic contributions to SCE of ρp .

A. Alastuey et al.

3.6.1 Two-Proton and Two-Electron Clusters In Figs. 10a and 10b, C1 contains either two protons or two electrons. Leading contribution of Fig. 10a arises from relative distances between clusters Ca and C1 of order κ −1 , while relative distances between particles inside cluster C1 are of order βe2 . For such configurations, Φ(Ca , C1 ) can be replaced by its Debye classical form −2βe2 exp(−κX)/X, where X is the relative distance between Ca and C1 (cluster C1 carries a total charge 2e). At the same time, statistical weights ZφT can be replaced by their bare forms. Then, integration over internal degrees of freedom of C1 merely provides half contribution (3.21) of Fig. 6a made with a single root cluster identical to C1 : that factor 1/2 arises from the combinatorics specific to root cluster of any graph (see comment after formula (2.38) and factor qa in the corresponding contribution). Integration over internal degrees of freedom of Ca obviously provides ρpid , while the remaining spatial integration over X reduces to 

  8πβe2 1 exp(−κX) dX −2βe2 =− = − id . 2 X κ ρp


Eventually, leading contribution of Fig. 10a is   γ2 m 3/2 −ρ ∗ √ Z(2, 0) exp(βEH ), 2 2 mp


i.e. minus half bare contribution (3.21) of Fig. 6a. Next corrections to (3.47) decay exponentially faster than ρ ∗ exp(βEH ). A similar calculation provides leading contribution of Fig. 10b  3/2 γ2 m ρ∗ √ Z(0, 2) exp(βEH ), (3.49) 2 2 me where we have used that C1 carries a charge −2e. Next corrections to (3.49) also decay exponentially faster than ρ ∗ exp(βEH ). 3.6.2 Ions Leading contributions of Figs. 10c and 10d can be treated as above. Taking into account that ion H2+ carries a charge e, we find for Fig. 10c −ρ ∗

  γ 3 me (M + mp ) 3/2 Z(2, 1) exp(2βEH ), 64 M2

Fig. 10 Graphs accounting for interactions between an ionized proton and a charged cluster


Exact Results for Thermodynamics of the Hydrogen Plasma

i.e. minus one fourth bare contribution (3.26) of Fig. 6c. For Fig. 10d, no combinatorics factor 1/2 appears when integrating over internal degrees of freedom of C1 because C1 contains a single proton. Since C1 carries a charge −e, leading contribution of Fig. 10d becomes   γ 3 mp (M + me ) 3/2 ρ∗ Z(1, 2) exp(2βEH ), (3.51) 64 M2 i.e. half bare contribution (3.27) of Fig. 6d. Contributions (3.50) and (3.51) can be interpreted as the modification of density of ionized protons due to their coupling with ions H2+ and H − respectively. As mentioned above, those contributions added to that of Figs. 6c and 6d provide a full contribution to ρp which is indeed identical to that relative to ρe . Thus, charge neutrality is indeed enforced by the structure of SCE (2.37). Next corrections to (3.50) and (3.51) decay exponentially faster than ρ ∗ exp(βEH ), as well as all other non-ideal contributions of ions H2+ and H − . Part of such contributions may be related to modifications of screening length, which are taken into account by summing suitable chain graphs (we have checked that this does provide the screening Debye length for a mixture of ionized charges and ions). 3.6.3 Other Charged Clusters Eventually, Figs. 11a–d made with three clusters Ca , C1 and C2 , also provide leading contributions of order ρ ∗ exp(βEH ) via the same mechanism as above. At leading order, Φ(Ca , C1 ) can be replaced by its Debye classical form. Then, integrations over internal degrees of freedom of clusters C1 and C2 , and over relative distance X2 − X1 between those clusters, are identical (apart from obvious substitutions p → e) to those relative to Figs. 7b (for 11a), 7c (for 11b), 9e (for 11c) and 9f (for 11d). Using again identity (3.47) for integration over X = X1 − Xa , and (A.7) for the integral of φ 3 , we obtain for Fig. 11b (C1 is made with a single particle and carries a charge ±e) ρ∗γ 2

(cp − ce )(β|EH |)3/2 exp(βEH ), 24π 3/2


and for Figs. 11c and 11d ρ∗

γ3 [W (1, 1|0, 1) − W (1, 1|1, 0)] exp(2βEH ). 2


Total leading contribution of Fig. 11a vanishes by charge neutrality constraint (2.8). Next corrections to (3.52) and (3.53) decay exponentially faster than ρ ∗ exp(βEH ).

Fig. 11 Graphs of order ρ ∗ exp(βEH ) accounting for interactions between three clusters

A. Alastuey et al.

3.7 Contributions with Arbitrary Particle Numbers The evaluation of any contribution arising from a graph GNp ,Ne made with at least three particles (Np + Ne ≥ 3), can be carried out by extending the methods described above for graphs with few particles. The outlines of the analysis are briefly sketched below. We first proceed to an estimation of the leading contribution. The behaviors of further collective corrections are discussed afterwards. 3.7.1 Leading Contributions ∗ At leading order, we make the substitutions exp(IR ) → 1 and BφT → BT in any weight ZφT (Ci ), and (exp(IR ) − 1) → IR in specific weight (2.42). Moreover, we explicit each D(Ci ) in terms of spatial integrations over particle positions and of functional integrations over Brownian bridges. ∗ Let consider two clusters Ci and Cj connected by a bond Fφ (Ci , Cj ). If one of them is electrically neutral, i.e. it contains the same number of protons and electrons, φ can be replaced by V in Fφ (Ci , Cj ). If both carry a net charge, φ must be replaced by its Debye classical form φD . ∗ By virtue of Feynman-Kac formula, functional integrations over Brownian bridges reduce to matrix elements of either exp(−βHMp ,Me ), or interactions V evolved according to exp(−τ HMp ,Me ) (0 ≤ τ ≤ β). ∗ In graphs only made with bare bonds, integrations over positions of particles provide a function W accounting for bare interactions between clusters. If GNp ,Ne contains a single cluster, such integrations give raise to partition function Z(Np , Ne ). ∗ When one or more bonds involve φD , positions of particles which belong to charged clusters connected by such bonds, are rewritten in terms of relative positions inside a given cluster and cluster position (arbitrarily defined as the position of a given particle). Let consider a charged cluster C(Mp , Me ) (Mp + Me = 0), not connected to any neutral cluster. Integration over its position X can be disentangled from integrations over internal relative positions, since its internal weight BT decays on a scale βe2 much smaller than κ −1 which controls the decay of φD . Integration over its internal relative positions provide partition function Z(Mp , Me ). Integration over X is performed by rescaling X in units of κ −1 . This provides multiplicative inverse powers of κ, with possible logarithmic terms ln(κλ) arising from integrands built with φD3 . ∗ According to above analysis and prescriptions, the leading contribution of GNp ,Ne can be rewritten as (apart from a pure numerical coefficient which depends on ratio me /mp ) ρ ∗ γ Np +Ne −P /2 exp[β(Np + Ne − 1 − P /2)EH ]




where each Z and each W depends only on temperature, while P is a positive integer. Term γ −P /2 exp(−PβEH /2) arises from contribution 1/κ P , which is generated by both integrations over positions of charged clusters and specific weights IR proportional to κ (P = 0 when GNp ,Ne contains only neutral clusters). ∗ The low-temperature behaviors of functions Z and W can be inferred from the methods exposed in Appendix B:. If there exists a bound state made with Mp protons and Me electrons, partition function Z(Mp , Me ) then behaves as (0) ), exp(−βEM p ,Me


Exact Results for Thermodynamics of the Hydrogen Plasma

apart from a multiplicative integer which accounts for groundstate degeneracy. In the other case, asymptotic behavior (3.55) has to be multiplied by some power of β. A given interaction function W behaves as the product of Boltzmann factors (3.55) associated with each interacting cluster times a power of β. ∗ According to the above low-temperature behaviors of Z and W , leading contribution (3.54) of GNp ,Ne reduces to ρ ∗ γ Np +Ne −P /2 times a power of β times exp[β(Np + Ne − 1 − P /2)EH ]

(0) exp(−βEM ) p ,Me


when β → ∞. The precise form of factor (3.56) has been studied above for several graphs GNp ,Ne . For all other graphs, we have checked that (3.56) is exponentially smaller than exp(βEH ). In particular, ideal contributions of complex entities made with more than four particles can be omitted at considered order. The analysis is achieved by using the known values of EH + , EH − and EH2 given in Sect. 2.2, as well as inequality (2.31) for Mp + 2 Me ≥ 5. Neutrality constraint (2.8) ensures the cancellation of the leading contributions of graphs which differ only by ending clusters made with either a single proton or a single electron: above statement then applies, strictly speaking, to the leading contribution of the sum of those graphs (for instance, see graphs G1,1 and G2,0 shown in Fig. 7a). 3.7.2 Collective Corrections ∗ Collective corrections are obtained by expanding ring factors exp(IR ) in powers of IR , and truncated Mayer coefficients BφT in powers of (φ − V ). At the same time, both IR and (φ − V ) are expanded in positive integer powers of κλ as described in Appendix A:. Then, integration over cluster degrees of freedom involved in previous expansions, provide screening functions S. For a given graph GNp ,Ne , the resulting corrections take the general form ρ ∗ γ Np +Ne −P /2+L/2 exp[β(Np + Ne − 1 − P /2 + L/2)EH ]





with L a positive integer. ∗ The low-temperature behavior of S is analogous to those of Z and W , and it reduces to the product of a power of β times groundstate Boltzmann factors (3.55) associated with each cluster involved in S (for instance, see the calculation of S3 (1, 1) detailed in Appendix A:). Thus, and as expected from the weakly-coupled conditions enforced in the Saha regime, any correction (3.57) arising from GNp ,Ne becomes exponentially smaller than its leading contribution (3.54) when β → ∞. Collective corrections arising from graphs considered in Sects. 3.1–3.6 have been explicitly computed up to order ρ ∗ exp(βEH ) included. All other corrections, in particular those arising from other graphs, decay exponentially faster than ρ ∗ exp(βEH ).

4 Scaled Low-Temperature Expansions According to the analysis of Sect. 3, we derive the structure of the asymptotic expansion of ρ/ρ ∗ (Sect. 4.1). Then, we proceed to the calculation of the pressure as a function of ρ (EOS), by using thermodynamical identities (Sect. 4.2). We derive the corresponding expansion around ideal Saha pressure (1.3), and the first four corrections are explicitly computed.

A. Alastuey et al.

4.1 Structure of the Asymptotic Expansion of Particle Density According to Sect. 3.7, every contribution arising from any graph GNp ,Ne can be rewritten as ρ ∗ times γ n times a temperature-dependent function. Power n is integer or half-integer, n ≥ 1, while γ n may be multiplied by integer powers of ln γ (it is not necessary to write explicitly such logarithmic terms since they do not play any role in the following). For a given n, there is a finite number of contributions proportional to γ n , i.e. such that Np + Ne − P /2 + L/2 = n. Their sum can be recast as ρ ∗ γ n gn (β) exp(β(n − 1)EH ).


Functions gn (β) are expressed in terms of bare partition functions Z(Mp , Me ) of clusters (Mp , Me ), bare interactions W between clusters, and screening functions S which may involve either a single or various clusters. Roughly speaking, the number of involved graphs, as well as the maximum total particle number Np + Ne , increase with n. Taking into account the results derived in Sect. 3, screened cluster expansion of common particle density ρ = ρp = ρe can be formally rewritten as γ2 + γ 3/2 g3/2 (β) exp(βEH /2) + γ 2 g2,exc (β) exp(βEH ) 2 ∞  + γ n gn (β) exp(β(n − 1)EH )

ρ/ρ ∗ = γ +



where the sum runs over integer and half-integer values of n. In (4.2), we have extracted in γ 2 g2 (β) exp(βEH ) contribution (3.16) of atoms H in their groundstate, while the remaining part defines g2,exc (β). First two functions g3/2 and g2,exc are g3/2 (β) = S3/2 (1, 0)


according to (3.4), and g2,exc (β) =

1 [S2 (1, 0) + S2 (0, 1)] + W (1, 0|1, 0) + W (1, 0|0, 1) 2 (cp + ce )(β|EH |)3/2 24π 3/2       1 2m 3/2 2m 3/2 + Zexc (1, 1) + Z(2, 0) + Z(0, 2) 8 mp me −


by summing (3.7), (3.17), (3.21), (3.33), (3.36), (3.39), (3.48), (3.49) and (3.52). Notice that (4.4) is symmetric with respect to permutations of species indexes p and e in agreement with ρp = ρe . We stress that, in the Saha regime, γ is a fixed parameter not necessarily small, while β → ∞. Then, functions g2,exc (β) exp(βEH ) and gn (β) exp(β(n − 1)EH ) with n ≥ 3/2 and n = 2, decay exponentially fast. Thus, the whole sum over n in (4.2) can be reordered according to the corresponding decay rates. Each term γ n gn (β) exp(β(n − 1)EH ) is then rewritten as γ n hk (β) where k = k(n) is some integer. Functions hk decay exponentially fast, i.e. hk (β) ∼ exp(−βδk ) (apart from powers of β), with decay rates δk ranked as 0 < δ1 < δ2 < · · ·: hk+1 decays exponentially faster than hk when β → ∞. According to the

Exact Results for Thermodynamics of the Hydrogen Plasma

analytic results derived in Sects. 3.1–3.7 on the one hand, and to the numerical values of EH , EH − , EH + , EH2 (see Sect. 2.3) on the other hand, we find 2

h1 (β) = g3/2 (β) exp(βEH /2),

n1 = 3/2,

h2 (β) = g4 (β) exp(3βEH ),

n2 = 4,

h3 (β) = g2,exc (β) exp(βEH ),

n3 = 2,

h4 (β) = g3 (β) exp(2βEH ),

n4 = 3.


Their corresponding decay rates δk can be found in the table (1.5) given in the Introduction, while all higher-order functions hk (β) with k ≥ 5 decay exponentially faster than exp(βEH ), i.e. their decay rates δk are larger than |EH |  13.6. Notice that both γ 5/2 g5/2 (β) exp(3βEH /2) and γ 7/2 g7/2 (β) exp(5βEH /2) decay faster than exp(βEH ), so both k(5/2) and k(7/2) are strictly larger than 4. Within previous reordering, (4.2) becomes ρ/ρ ∗ = γ +

γ 2  nk + γ hk (β). 2 k=1


At order exp(βEH ) included, all terms with k ≥ 5 can be omitted in (4.6). Moreover, for the sake of consistency, it is sufficient to compute functions hk with 1 ≤ k ≤ 4 at the same order (beyond its leading behavior exp(−βδk ), hk involves other exponentially small contributions). This gives (β|EH |)3/4 exp(βEH /2), π 1/4


  1 2m 3/2 Z(2, 2) exp(3βEH ) + W (1, 1|1, 1) exp(3βEH ), 64 M


h1 (β) =

h2 (β) =

   1 4m (β|EH |)3/2 1 h3 (β) = − + 1 + ln exp(βEH ) 2 12 M π 1/2  

  1 2m 3/2 1 + E(−x ) + ) − ) 2Q(x Q(−x pe pp pp 8π 1/2 mp 2     2m 3/2 1 + Q(−xee ) − E(−xee ) exp(βEH ), me 2


and 3 h4 (β) = 64

me (M + mp ) M2


mp (M + me ) Z(2, 1) + M2


 Z(1, 2) exp(2βEH )

3 + S3 (1, 1) exp(2βEH ) + [W (1, 1|1, 0) + W (1, 1|0, 1)] exp(2βEH ). 2


In (4.7) and (4.9), full contributions of respectively g1 and g2,exc , are kept, while analytic expressions (3.3), (3.8), (3.34), (3.37) and (3.39) have been used. Moreover, and according to formula (A.12) derived in Appendix A:, Zexc (1, 1), Z(2, 0) and Z(0, 2) have been expressed in terms of Ebeling quantum virial functions Q (direct part) and E (exchange

A. Alastuey et al.

part) defined in Ref. [39], with arguments xpe = 2(β|EH |)1/2 , xpp = (2mp /m)1/2 (β|EH |)1/2 and xee = (2me /m)1/2 (β|EH |)1/2 . Term −1/2 in h3 (β) subtracts the ground-state contribution included in function Q(xpe ). In (4.8) and (4.10), contributions of g4 and g3 which decay exponentially faster than exp(βEH ) have been omitted. The resulting expression for h2 is obtained by summing (3.23) and (3.41). Similarly, expression (4.10) for h4 follows by summing (3.26), (3.27), (3.50), (3.51), (3.19), (3.45) and (3.53). As a conclusion, it is useful to summarize the main features and ingredients of expansion (4.6). The hk -functions are ordered, at sufficiently low temperatures, according to |h1 (β)| > |h2 (β)| > |h3 (β)| > |h4 (β)| > · · ·. They incorporate corrections to ideal Saha equation which arise from different physical phenomena, as listed in the Introduction. Explicit expressions for h1 (β) and h3 (β) are known, see (4.7) and Ref. [39], while h2 (β) and h4 (β) involve integrals associated with 3-body and 4-body problems which cannot be expressed in closed analytical forms. In h2 (β), the internal partition function Z(2, 2) of a hydrogen molecule is defined in (3.24), and its low-temperature form is determined in Appendix B:. The function W (1, 1|1, 1), which accounts for atom-atom interactions, is defined in (3.41), and its low-temperature form is computed in Appendix C:. In h4 (β), the internal partition functions Z(2, 1) and Z(1, 2) of ions H2+ and H − are defined in (3.28) and (3.29) respectively, while their asymptotic expressions at low temperatures are derived in Appendix B:. The interactions W (1, 1|1, 0) and W (1, 1|0, 1) between an atom and an ionized proton or electron, are defined in (3.45) and their low-temperature expressions are given in Appendix C:. Eventually, the screening function S3 (1, 1) of a hydrogen atom accounts for collective corrections to the bare proton-electron Coulomb potential beyond the familiar Debye shift, and it is given by formula (A.16) at low temperatures. 4.2 Equation of State In order to compute the pressure, we consider identities ρp = zp

∂βP , ∂zp

ρe = ze

∂βP . ∂ze


Taking into account that P depends only on z and β, and using parametrization of z in terms of β and γ , z = (m/M)3/4 γ exp(βEH )/4, we rewrite such identities as 2ρ ∂βP = , ∂γ γ


where partial derivative of βP with respect to γ is taken at fixed β. After inserting expansion (4.6) of ρ into the r.h.s. of (4.12), a straightforward term by term integration with respect to γ provides βP /ρ ∗ = 2γ +

γ 2  2γ nk + hk (β). 2 nk k=1


The required equation of state follows by inserting into (4.13) the expression of γ in terms of ρ obtained from the inversion of (4.6). That inversion can be performed perturbatively

Exact Results for Thermodynamics of the Hydrogen Plasma

around the simple expression γSaha = (1 + 2ρ/ρ ∗ )1/2 − 1


obtained by retaining only the first two terms of (4.6). The resulting SLT expansion of the pressure takes the form (1.1) presented in the Introduction where βPSaha /ρ ∗ is given by (1.3). The general structure of βPk /ρ ∗ reduces to a function of ρ/ρ ∗ times a polynomial in the hl (β)’s with l ≤ k. Therefore, for a fixed ratio ρ/ρ ∗ , corrections βPk /ρ ∗ decay exponentially fast when β → ∞. Moreover, each βPk+1 /ρ ∗ decays faster than βPk /ρ ∗ for k ≥ 0 (with P0 = PSaha ). First corrections in (1.1) read βP1 /ρ ∗ =

[(1 + 2ρ/ρ ∗ )1/2 − 3][(1 + 2ρ/ρ ∗ )1/2 − 1]3/2 h1 (β), 3(1 + 2ρ/ρ ∗ )1/2


βP2 /ρ ∗ =

−[(1 + 2ρ/ρ ∗ )1/2 + 2][(1 + 2ρ/ρ ∗ )1/2 − 1]4 h2 (β), 2(1 + 2ρ/ρ ∗ )1/2


βP3 /ρ ∗ =

−[(1 + 2ρ/ρ ∗ )1/2 − 1]2 h3 (β), (1 + 2ρ/ρ ∗ )1/2


βP4 /ρ ∗ =

−[(1 + 2ρ/ρ ∗ )1/2 + 3][(1 + 2ρ/ρ ∗ )1/2 − 1]3 h4 (β), 3(1 + 2ρ/ρ ∗ )1/2


[(1 + 2ρ/ρ ∗ )1/2 − ρ/ρ ∗ ][(1 + 2ρ/ρ ∗ )1/2 − 1]2 [h1 (β)]2 . (1 + 2ρ/ρ ∗ )3/2


and βP5 /ρ ∗ =

Next correction βP6 /ρ ∗ decays faster than exp(βEH ). In previous corrections βPk /ρ ∗ , functions h1 (β) and h3 (β) can be expressed in closed analytical forms according to (4.7) and (4.9) respectively. Similar analytical expressions for h2 (β) and h4 (β) are not available. Nevertheless, the low-temperature behaviors of those functions are exactly known, i.e. h2 (β) ∼ and

  1 2m 3/2 exp(β(3EH − EH2 )) 64 M

  3 me (M + mp ) 3/2 exp(β(2EH − EH + )) h4 (β) ∼ 2 64 M2



when β → ∞. Eventually, the various terms in (1.1) display interesting behaviors with respect to ratio ρ/ρ ∗ , at fixed β sufficiently large: • For ρ much smaller than ρ ∗ , each βPk /ρ ∗ , as well as βPSaha /ρ ∗ , can be expanded in powers of ρ/ρ ∗ . This leads to the virial expansion of βP in powers of ρ. Since all βPk /ρ ∗ ’s for k ≥ 6 are at least of order ρ 5/2 , the full contribution of terms with k ≤ 5 in (1.1) provides the expansion of βP up to order ρ 2 , i.e. βP = 2ρ −

23/2 (2π)3/4 (λpe )3/2 exp(−βEH /2)h1 (β)ρ 3/2 3

A. Alastuey et al.

− (2π)3/2 (λpe )3 exp(−βEH )[2h3 (β) + 1 − 2(h1 (β))2 ]ρ 2 + O(ρ 5/2 ) 

 π 1 (8πβe2 ρ)3/2 − √ 2λ3pe Q(xpe ) + λ3pp Q(−xpp ) − E(−xpp ) = 2ρ − 24π 2 2     4m 3 6 2 1 π + λ3ee Q(−xee ) − E(−xee ) ρ 2 − ln β e ρ + O(ρ 5/2 ) (4.22) 2 6 M which does coincide with the well known expression derived previously by other methods [4, 23, 39] (λpe = (β2 /m)1/2 , λpp = (β2 /mpp )1/2 and λee = (β2 /mee )1/2 ). Notice that contribution of βP4 is of order ρ 3 , while that of βP2 is of order ρ 4 . • For ρ of order ρ ∗ , leading term βPSaha /ρ ∗ , as well as each correction βPk /ρ ∗ , can be viewed as infinite resummations of terms with arbitrary high orders in the above lowdensity expansion. Such resummations account, in a non-perturbative way with respect to density, of recombination processes for any value of the ionization rate. The relative orders of magnitude of the various corrections to Saha pressure are mainly controlled by their decay rates δk . Therefore the larger correction indeed is βP1 , which results from plasma polarization around a given ionized charge, evaluated within Debye classical mean-field theory. That result is equivalent to the modified Saha condition which determines the ionization rate [39, 41]. • For ρ much larger than ρ ∗ , βPSaha behaves as βPSaha ∼ ρ,


which illustrates the almost full atomic recombination of the plasma. The larger correction to Saha pressure is now βP2 which behaves as βP2 ∼ −2h2 (β)ρ ∗

ρ ρ∗

2 ,


so it overcomes βP1 which grows only as (ρ/ρ ∗ )3/4 , as well as further corrections βP3 ∼ (ρ/ρ ∗ )1/2 , βP4 ∼ (ρ/ρ ∗ )3/2 and βP5 ∼ (ρ/ρ ∗ )1/2 . Therefore molecular recombination prevails over plasma polarization. Of course, expansion (1.1) is no longer appropriate for too large values of ratio (ρ/ρ ∗ ), since some corrections βPk /ρ ∗ become much larger than Saha pressure. 4.3 Numerical Estimations and Validity Domain of SLT Expansions Quantitative estimations of corrections βP1 /ρ ∗ , βP3 /ρ ∗ and βP5 /ρ ∗ are easy, because functions h1 (β) and h3 (β) can be represented by simple analytical expressions at finite temperature. For functions h2 (β) and h4 (β) which involve 3 and 4-body contributions, no explicit finite-T representations are available beyond their low-temperature asymptotic forms determined in Appendices B and C. In order to obtain reliable values for those functions at moderate temperatures, we have used a simple approach in which important finite temperature effects (such as atomic vibrations and rotations) are phenomenologically taken into account. As mentioned in the Introduction, the corresponding numerical evaluations of the various corrections to Saha pressure (and internal energy), together with a comparison of our predicted isotherms with the results of PIMC simulations, will be presented in a forthcoming paper [1]. Here, we exhibit the validity domain of SLT expansion (1.1). A rigorous analysis of the convergence of that expansion is a tremendous mathematical task, much beyond the scope

Exact Results for Thermodynamics of the Hydrogen Plasma

Fig. 12 Phase diagram showing the validity domains of SLT expansion (1.1) (hatched region) and of the virial expansion (shaded region). Atomic recombination density ρ ∗ (β) (1.2) is a straight line at low temperatures in the (β, log ρ)-plane. The validity domain is delimited at high densities, and for temperatures below 10000 K, by critical density ρc (β) at which molecular recombination occurs. Crosses indicate state points where simulation results are available [48]. State points of astrophysical systems (Sun photosphere and Brown dwarfs) are also shown in the diagram

of the present work. We estimated a quite plausible validity domain, by employing the semiempirical criterion |Pk | < PSaha /10 for all five corrections (1 ≤ k ≤ 5): it covers the region hatched in the temperature-density plane shown in Fig. 12. The shaded region at low densities and high temperatures corresponds to the validity domain of the virial expansion (i.e. low-density expansion at fixed temperature) determined from a similar criterion. Obviously, SLT expansion improves widely upon virial expansion, by providing reliable results in the atomic phase, including the temperature and density ranges around ρ  (β) which correspond to partially ionized hydrogen gases. In Fig. 12, we have also shown state points, symbolized by crosses, for which PIMC simulation results have been obtained [48]. It turns out that some of them lie within the validity domain of the SLT expansion. We have checked that our calculations, both for pressure and internal energy, are in agreement with PIMC results within statistical errors [1]. This confirms the reliability of SLT expansions in the domain inferred from the above semi-empirical criterion. Notice that such domain extends to rather high densities, up to a/aB = 6 at 15000 K, which corresponds to a mean inter-particle distance of the order of twice the size of a hydrogen atom. For temperatures below 10000 K, the validity domain is limited at high densities by molecular recombination which occurs around densities ρc (β) shown as a dashed line in the phase diagram. That limitation is not intrinsic to the theory, and an SLT expansion applicable in the molecular regime can be derived as well. Such a generalization requires replacing the scaling (2.12) of the chemical potential by a similar scaling corresponding to a molec-

A. Alastuey et al.

ular regime (see Fig. 2), and performing the inversion μ = μ(ρ) in the appropriate density range. Notice that if ρ is not too high above ρc , expansion (4.6) of particle density in terms of chemical potential should remain valid. Performing a non-perturbative inversion of the chemical potential in favor of the density, should then provide accurate thermodynamical functions which account not only for atomic recombination at ρ ∼ ρ ∗ , but also for molecular recombination above ρc . For temperatures above 10000 K, the borderline of the validity domain has a complicated shape determined by correction term P3 which accounts for atomic excitations and for interactions between ionized electrons and protons. At high temperatures, typically above 30000 K, thermal ionization prevents recombination of protons and electrons into atoms, and our validity criterion is then equivalent to a weak coupling condition for ionized charges, i.e. coupling parameter Γ = βe2 /a smaller than some value, which also determines the validity of the virial expansion. Eventually, state points of two astrophysical systems of interest, Sun photosphere and a typical brown dwarf atmosphere, are also shown in Fig. 12. Our equation of state (1.1) clearly holds at the temperature and density of Sun photosphere. In order to be applicable to brown dwarf atmospheres, the SLT expansion would need to be generalized to the molecular regime, as discussed above. Acknowledgements We are very indebted to Vincent ROBERT (Laboratoire de Chimie, ENS Lyon, CNRS) for his efficient and kind support. We thank him in particular for his spectroscopic calculations (ground state energies and excitations energies of a few compounds) which we used in Sect. 2.3 [52].

Appendix A: Screened Potential and Related Integrals A.1 Expression and Behavior of φ The Fourier transform of φ(Li , Lj ) with respect to Xi − Xj reads  qi  qj ˜ χi , χj ) = eαi eαj φ(k, ds dt exp[ik · (λαi ηi (s) − λαj ηj (t))] 0



∞  n=−∞


4π exp[−2nπi(s − t)], + κ 2 (k, n)


where χ = (α, q, η(·)) denotes the loop internal degrees of freedom, while function κ 2 (k, n) is defined in Ref. [13]. Functions κ 2 (k, n) are analytical in k 2 near k = 0, while κ 2 (0, n) = 0 for n = 0 and κ 2 (0, 0) = 0 is of order κ 2 . For large values of k, κ 2 (k, n) remains bounded by a constant independent of n (of order κ 2 ). For k ∼ κ, κ 2 (k, n) for n = 0 is smaller than κ 2 (k, 0) by a factor of order κ 2 λ2 , while κ 2 (k, 0) can be replaced by κ 2 . The behaviors of φ with respect to relative distance r = |Xi − Xj | (roughly described in Fig. 4), can be readily derived from those of φ˜ with respect to k, as detailed in Ref. [13]. Here, we briefly summarize that analysis. For k  κ, each fraction 4π/(k 2 + κ 2 (k, n)) can be replaced by 4π/k 2 in (A.1), so φ(Li , Lj ) behaves as V (Li , Lj ) at short distances r  κ −1 . At distances r ∼ κ −1 , we recover the Debye classical form φD (Li , Lj ) = qi eαi qj eαj exp(−κr)/r , by noting that terms n = 0 in (A.1) provide contributions smaller than the one of n = 0 by a factor of order κ 2 λ2 . Eventually, terms n = 0 in (A.1) provide ˜ χi , χj ), which in turn induces a dipolar-like a singularity in the small-k expansion of φ(k, decay of φ(Li , Lj ) at large distances r  lQ .

Exact Results for Thermodynamics of the Hydrogen Plasma

A.2 Integrals of Powers of φ (p)




We consider φ(La , L1 ) where loops La and loop L1 contain, respectively, one proton (qa = 1) and either one proton or one electron (q1 = 1). According to the definition of (p,e) ˜ χa , χ1 ), the integral of φ(L(p) ) over X1 and ξ 1 = η1 is nothing but φ(k, a , L1    (p,e) ˜ χa , χ1 ). dX1 D(ξ 1 )φ(L(p) ) = D(ξ 1 )φ(0, (A.2) a , L1 The r.h.s. of (A.2) is computed by taking the limit k → 0 of expression (A.1). The corresponding contribution of a term n = 0 is obtained by expanding phase factor exp[ik · (λp ξ a (s) − λp,e ξ 1 (t))] in powers of k. Since all odd moments of measure D(ξ 1 ) van1 ish, as well as 0 dt exp(2nπit), the first non-vanishing term in that expansion is at least of order k 3 . It has to be multiplied by a factor of order 1/k 2 which arises from fraction 4π/(k 2 + κ 2 (k, n)), so the resulting contribution to the r.h.s. of (A.2) vanishes. Therefore, the sole contribution arises from term n = 0, i.e.   4πe2 (p,e) , (A.3) )=± 2 dX1 D(ξ 1 )φ(L(p) a , L1 κ (0, 0) (p)

with a positive sign for L1 and a negative one for L(e) 1 . (p) (p,e) According to Fourier-Plancherel formula, the integral of [φ(La , L1 )]2 over X1 , ξ a = ηa and ξ 1 = η1 , is rewritten as   (p,e) 2 dX1 D(ξ a )D(ξ 1 )[φ(L(p) )] a , L1   1 ˜ χa , χ1 )|2 dk D(ξ a )D(ξ 1 )|φ(k, (2π)3   1  1  1  1 2e4 dk = ds1 dt1 ds2 dt2 π 0 0 0 0 =


exp[−2n1 πi(s1 − t1 )] exp[−2n2 πi(s2 − t2 )] k 2 + κ 2 (k, n1 ) k 2 + κ 2 (k, n2 ) n1 ,n2 =−∞   × D(ξ a ) exp[iλp k · ξ a (s1 − s2 )] D(ξ 1 ) exp[−iλp,e k · ξ 1 (t1 − t2 )]. (A.4) ×

In the last equality of (A.4), we have used that the average over shape ξ of any function f (ξ a (s1 ) − ξ a (s2 )) is identical to the average of f (ξ a (s1 − s2 )), provided that ξ (s) for s outside [0, 1] is defined as equal to ξ (s − [s]) [13]. Within variable change k = κq, we can replace κ 2 (κq, ni ) by either κ 2 for ni = 0, or 0 for ni = 0, discarding terms which provide contributions smaller by a factor (κλ)2 at least. Summations over ni = 0 are then performed according to identity n =0 exp[−2nπi(s − t)] = δ(s − t) − 1. Since measure D(ξ ) is Gaussian with covariance (2.34) (for q = 1), we transform (A.4) into   1  ∞ q2 8e4 1 ds dt dq 2 exp[−κ 2 λ2p q 2 s(1 − s)/2 − κ 2 λ2p,e q 2 t (1 − t)/2] κ 0 (q + 1)2 0 0   ∞ 8e4 1 1 + ds dq 2 {exp[−κ 2 (λ2p + λ2p,e )q 2 s(1 − s)/2] − 1} κ 0 q 0

A. Alastuey et al.

8e4 κ









1 q2

× {exp[−κ 2 λ2p q 2 s(1 − s)/2 − κ 2 λ2p,e q 2 t (1 − t)/2] − 1}


discarding terms of order κ −1 O((κλ)2 ). The integrals over q in (A.5) are computed in terms of elementary functions of arguments [κ 2 λ2p s(1 − s)/2 + κ 2 λ2p,e t (1 − t)/2]1/2 and [κ 2 (λ2p + λ2p,e )s(1−s)/2]1/2 , which can be expanded in Taylor series since κλ is small. For the leading (order κ −1 ) and first subleading (order κ −1 O(κλ)) contributions, the remaining integrals over s and t are readily calculated (some complicated double integrals over s and t arising from respectively first and third terms in (A.5) cancel out). Eventually, we obtain √     2πe4 π (α) 2 (p) 2 dX1 D(ξ a )D(ξ 1 )[φ(La , L1 )] = (A.6) 1 − √ κλpα + O((κλ) ) . κ 2 2 (p)

3 The integral of [φ(La , L(α) 1 )] over X1 , ξ a and ξ 1 , can be evaluated within similar techniques and tricks. Discarding terms of order O(κλ), its leading behavior reduces to a constant times ln(κλpα ) plus another constant. When the two integrals corresponding respectively to α = p and α = e are summed together, logarithmic terms in κ cancel out. Therefore, we obtain    (p) 3 (e) 3 (p) 6 (A.7) dX1 D(ξ a ) D(ξ 1 )[(φ(L(p) a , L1 )) + (φ(La , L1 )) ] = cp e + O(κλ)

where cp is the constant cp =

2 π3

1 0

dt 0





1 q12 q22 |q1 + q2 |2

× {exp[−(q12 s(1 − s) + q22 t (1 − t) + 2q1 · q2 t (1 − s))] − exp[−(q12 s(1 − s) + q22 t (1 − t) + 2q1 · q2 t (1 − s))mp /(2m)]}


entirely determined by ratio mp /m. As it should, leading contribution cp e6 in (A.7) is nothing but the value of the considered integral with bare potential V in place of φ (that bare (p) (p) (p) 3 integral does converge thanks to the 1/|X1 |4 -decay of [V (La , L1 )]3 + [V (La , L(e) 1 )] ). (p) (e) When the root proton is replaced by a root electron (La → La ), the resulting integral behaves similarly to (A.7) where constant ce is given by (A.8) with me in place of mp . A.3 Behavior of IR We consider a loop L containing a single particle of species α. Convolution formula (2.40) ˜ χa , χ1 ) for IR (L) is first transformed according to Fourier-Plancherel identity, in which φ(k, is replaced by (A.1). Discarding terms smaller by a factor O((κλ)2 ), only the contributions (p,e) associated with a single proton or a single electron, are retained. Moreover, of loops L1 at the same order, after making variable change k = κq, we can replace κ 2 (κq, n) by either κ 2 for n = 0, or 0 for n = 0. Using again identity n =0 exp[−2nπi(s − t)] = δ(s − t) − 1, we then obtain   1   1  βe2 κ  1 βe2 κ 1 + ds ds1 dt1 D(ξ 1 ) dq 2 2 IR (L) = 2 4π 2 γ 0 q (q + 1) 0 0

Exact Results for Thermodynamics of the Hydrogen Plasma

× [exp(iκq · (λα ξ (s) − λγ ξ 1 (s) + λγ ξ 1 (s1 ) − λα ξ (t1 ))) − 1]   1   1  βe2 κ  1 1 + ds ds dt D (ξ ) dq 4 1 1 1 4π 2 γ 0 q 0 0 × exp(iκq · (λα ξ (s) − λγ ξ 1 (s) + λγ ξ 1 (s1 ))) × [exp(−iκq · λα ξ (s1 )) − exp(−iκq · λα ξ (t1 ))] + βe2 κO((κλ)2 ). (A.9) The leading behavior of IR (L) reduces to the first term in the r.h.s. of (A.9). In the second term of (A.9), we can first perform the integration over q thanks to Cauchy’s theorem. The resulting elementary functions of the argument κ|λα ξ (s) − λγ ξ 1 (s) + λγ ξ 1 (s1 ) − λα ξ (t1 )| are then expanded in Taylor series since κλ is small. The remaining integrations over times and shape ξ 1 provide a contribution of order βe2 κO(κλ) which depends on ξ . The third term in the r.h.s. of (A.9) has the same order and a similar shape-dependence, as shown by variable changes q = κ|λα ξ (s) − λγ ξ 1 (s) + λγ ξ 1 (s1 ) − λα ξ (t)|u with t = s1 or t = t1 (the integral over q is splitted as the sum of two integrals by adding and subtracting 1 to [exp(−iκq · λα ξ (s1 )) − exp(−iκq · λα ξ (t1 ))]). The integration of IR (L) over shape ξ readily follows from (A.9). Now in the second and third terms of (A.9), it is convenient to first perform integration over shapes ξ and ξ 1 , using the previous trick relative to differences ξ (s) − ξ (t1 ) and ξ 1 (s) − ξ 1 (s1 ), as well as the Gaussian nature of measures D(ξ ) and D(ξ 1 ). This leads to 

  1  ∞ βe2 κ  1 βe2 κ 1 + D(ξ )IR (L) = ds1 dt1 dq 2 2 π (q + 1) 0 0 0 γ × {exp[−κ 2 λ2γ q 2 s1 (1 − s1 )/2 − κ 2 λ2α q 2 t1 (1 − t1 )/2] − 1}   ∞ βe2 κ  1 1 + ds1 dq 2 {exp[−κ 2 (λ2α + λ2γ )q 2 s1 (1 − s1 )/2] − 1} π q 0 0 γ −

  1  ∞ 1 βe2 κ  1 ds1 dt1 dq 2 π q 0 0 0 γ

× {exp[−κ 2 λ2γ q 2 s1 (1 − s1 )/2 − κ 2 λ2α q 2 t1 (1 − t1 )/2] − 1} + βe2 κO((κλ)2 ).


The integrals over q in the second, third and fourth terms of (A.10) are computed, similarly to that arising in (A.5), in terms of elementary functions which are afterwards expanded in powers of κλ. Contributions of second and fourth terms with order βe2 κO(κλ) cancel out, so it remains √     βe2 κ π D(ξ )IR (L) = κλαγ + O((κλ)2 ) . (A.11) 1− √ 2 8 2 γ A.4 Truncated Integrals of Powers of V Quantum virial functions Q(±xαγ ) are defined [39] through a truncation similar to that arising in r|[exp(−βHαγ )]TMayer |r, where matrix elements of time-evolved operators Vαγ and [Vαγ ]2 are replaced by βeα eγ /r and β 2 e4 /r 2 respectively, while [Vαγ ]3 -term is omitted.

A. Alastuey et al.

Within such truncation, convergence at large distances is ensured by taking the limit R → ∞ of the corresponding spatial integral inside a sphere with radius R plus logarithmic counter terms [39]. When partition functions Zexc (1, 1), Z(2, 0) and Z(0, 2) are expressed in terms 2 − e4 /r 2 and [Vαγ ]3 are of the Q’s and E’s, the integrals arising from Vαγ − eα eγ /r, Vαγ computed within previous methods applied to similar integrals of powers of φ. The sum of contributions due to Vαγ − eα eγ /r vanishes by virtue of identity λ2pp + λ2ee − 2λ2pe = 0. We then find       2m 3/2 2m 3/2 1 Z(2, 0) + Z(0, 2) Zexc (1, 1) + 8 mp me  

  1 2m 3/2 1 = 2Q(xpe ) + Q(−xpp ) − E(−xpp ) 8π 1/2 mp 2   3/2  2m 1 + Q(−xee ) − E(−xee ) me 2 −

1 β 2 e4 β 3 e6 exp(−βEH ) + (2λpe + λpp + λee ) + (cp + ce ) 3 2 32λpe 24(2πλ2pe )3/2


β 3 e6 ln(λpp λee /λ2pe ). 12(2π)1/2 λ3pe


In the r.h.s of (A.12), the first additional term to the linear combination of the Q’s and E’s merely arises from the substraction of groundstate contribution in Zexc (1, 1), while the last one is due to the logarithmic counter terms introduced in the definitions of the Q’s. A.5 Integral Mixing φ, IR and V At lowest order, effects of atom polarization are entirely embedded in the integral   4z2 dRdr D(ξ a )D(ξ 1 )BT (a, 1) (2πλ2 )3 Λ × [IR (Lpa ) + IR (Le1 ) − β(φ(Lpa , Le1 ) − V (Lpa , Le1 ))]


where we set R = Ra and r = r1 . Similarly to the case of bare integral (3.10), the presp ence of Boltzmann factor exp(−βV (La , Le1 )) in BT (a, 1) implies that leading contributions in (A.13) arise from configurations where loop sizes are at most of order λ, while relative proton-electron distance |r − R| is of order the extension aB of the atom groundstate. The IR ’s are then expanded with respect to small parameter κλ as above, while a similar expansion is derived for (φ − V ) by starting from a convolution relation analogous to (2.40) and by noting that κaB is also a small parameter. This provides IR (Lpa ) + IR (Le1 ) − β(φ(Lpa , Le1 ) − V (Lpa , Le1 ))   1 βe2 κ 2 1 ds dt[2|r + λe ξ 1 (t) − R − λp ξ a (s)| − λe |ξ 1 (t) − ξ 1 (s)| = 4 0 0 − λp |ξ a (t) − ξ a (s)|] + βe2 κO((κλ)2 ).


In (A.14), terms proportional to βe2 κ cancel out, as well as terms proportional to βe2 κ 2 |r − R| which do not depend on loop shapes.

Exact Results for Thermodynamics of the Hydrogen Plasma

Using (A.14) in (A.13), the functional integrations over loop shapes can be rewritten in terms of matrix elements of Gibbs operators. For instance, the integral associated with the p first term exp(−βV (La , Le1 )) in truncated Mayer coefficient BT (a, 1) can be rewritten as  2z2 e2 κ 2 dRdrdR1 dR2 dr1 dr2 βΛ  τ1  β × dτ1 dτ2 Rr| exp[−(β − τ1 )H1,1 ]|R1 r1  0


× R1 r1 | exp[−(τ1 − τ2 )H1,1 ]|R2 r2  R2 r2 | exp[−τ2 H1,1 ]|Rr × [2|r2 − R1 | − |r2 − r1 | − |R2 − R1 |]


discarding terms of order βe2 κO((κλ)2 ). Next subtracted terms in BT (a, 1) are rewritten similarly to (A.15) where imaginary-time evolution operators are associated with purely kinetic Hamiltonian H1,0 + H0,1 . At low temperatures, such terms become exponentially smaller than (A.15), the behavior of which is controlled by atomic groundstate contributions. That behavior is determined by starting with decomposition H1,1 = −2 ΔR∗ /(2M) + Hpe . An eigenstate of H1,1 reduces then to the product of a plane wave exp(iK · R∗ )/Λ1/2 for position R∗ = (mp R + me r)/M of the atom mass center, times an internal wave function (p) ψp (r∗ ) for relative position r∗ = r − R, while its energy reads 2 K2 /(2M) + EH . For bound (p) states, p → (n, l, m) where n is the usual quantum principal number (EH = EH /n2 , 1 ≤ n), l is the orbital number (0 ≤ l ≤ n − 1) and m is the azimuthal number −l ≤ m ≤ l (0 → (1, 0, 0) denotes the ground state); for diffusive states, p → (k, l, m) where k parametrizes (p) the positive energy EH = 2 k 2 /(2m) while l and m are again the orbital and azimuthal numbers with 0 ≤ l (the precise forms of the corresponding ψp ’s are given in Ref. [42] for instance). After changing proton and electron positions in favor of their mass center and relative particle counterparts in (A.15), the matrix elements are evaluated by suitable insertions of closure relation for the basis made with previous eigenstates. The resulting integrals over τ1 and τ2 are readily performed for each set of involved eigenstates. According to the scaling prescriptions defined in Sect. 3, integral (A.13) is then rewritten as (3.19) plus terms which decay exponentially faster than ρ ∗ exp(βEH ), while screening function S3 (1, 1) reads √

 sinh(K · Q) 2(β|EH |)1/2 exp(−(K 2 + Q2 )/2) exp(−βE ) 4 dKdQ S3 (1, 1) = H 5 64π K·Q  × dXdr∗1 dr∗2 exp(−2iK · X)|ψ0 (r∗1 )|2 |ψ0 (r∗2 )|2      2β|EH |m 1/2 me ∗ mp ∗  r1 + r2  × 2 X+ M M M     2β|EH |m 1/2 mp ∗ mp ∗  r1 + r2  X− −  M M M   1/2  2β|EH |m me ∗ me ∗   r − r X+ − M M 1 M 2

A. Alastuey et al.



p =0


exp(−K 2 /2) (p) β(EH

− EH ) + Q2 /2 − K 2 /2

dXdr∗1 dr∗2 exp(i(K − Q) · X)

     2β|EH |m 1/2 me ∗ mp ∗   r + r 2 X+ M M 1 M 2     2β|EH |m 1/2 mp ∗ mp ∗   r + r X− − M M 1 M 2     2β|EH |m 1/2 me ∗ me ∗  r1 − r2  . X+ −  (A.16) M M M × ψ 0 (r∗1 )ψ0 (r∗2 )ψp (r∗1 )ψ p (r∗2 )

In (A.16), all integration variables are dimensionless, in particular r∗i is expressed in units of aB . Moreover, the sum over diffusive states must be understood as an integral over k from 0 to ∞ and a discrete sum over l and m. When β → ∞, S3 (1, 1) behaves as S3 (1, 1) ∼

cat exp(−βEH ) 8π 3/2 (β|EH |)1/2


with pure numerical constant cat =

|EH |

(p) p =0 EH

   × 2

− EH

dYdr∗1 dr∗2

m|EH |

exp(−|Y|) ψ 0 (r∗1 )ψ0 (r∗2 )ψp (r∗1 )ψ p (r∗2 ) |Y|


me ∗ mp ∗ r + r M 1 M 2 − EH )   1/2  mp ∗ mp ∗  m|EH |  r + r − Y− (p) M 1 M 2 M(EH − EH )   1/2  me ∗ me ∗  m|EH | r r Y + − −  . (p) M 1 M 2 M(EH − EH ) (p) M(EH



Notice that only the second term ( p =0 · · ·) in (A.16) contributes to asymptotic behavior (A.17) (the first term provide contributions smaller by a factor ln(β|EH |)/(β|EH |) as a consequence of the spherical symmetry of groundstate wavefunction ψ0 (r∗ ) = ψ0 (r ∗ )). An accurate simplified expression for cat can be derived by setting m/M = me /M = 0 and mp /M = 1 in agreement with numerical value of ratio me /mp  1/1850, i.e. cat  −4π

|EH |

p =0



which provides cat  10.1.

dr∗1 dr∗2 ψ 0 (r∗1 )ψ0 (r∗2 )ψp (r∗1 )ψ p (r∗2 )|r∗1 − r∗2 |


Exact Results for Thermodynamics of the Hydrogen Plasma

Appendix B: Low-Temperature Behavior of Few-Body Partition Functions B.1 Behavior of Z(1, 1) Two-body proton-electron partition function Z(1, 1) reads  Z(1, 1) = 4 dx x|[exp(−βHpe )]TMayer |x,


where reduced Hamiltonian Hpe describes a single particle with mass m submitted to attractive Coulomb potential Vpe (x) = −e2 /|x|. Here, for the sake of notational convenience, we set x = r∗ , while z will denote a complex number. The integral over x can be splitted into two parts, |x| < βe2 and x > βe2 , the contributions of which are determined as follows. For |x| < βe2 , truncated Mayer operator [exp(−βHpe )]TMayer is replaced by its definition (3.14) in (B.1). The contribution of second term in (3.14), as well as those of next terms involving imaginary-time evolutions of Vpe with kinetic Hamiltonian Kpe , are readily computed in terms of elementary functions by exploiting the Gaussian nature of matrix elements of exp(−βKpe ). Such contributions are bounded by a power of β. The contribution of first  y; z) defined as the materm in (3.14) is analyzed by introducing the Green function G(x, −1 trix elements of resolvent (z + Hpe ) . That function is the solution of partial differential equation   2 e2  y; z) = δ(x − y), Δx − + z G(x, (B.2) − 2m |x| with suitable boundary conditions [36]. Its exact expression reads [36]  y; z) = mΓ (1 − iν) [Wiν,1/2 (−ikd+ )M˙ iν,1/2 (−ikd− ) G(x, 2π2 |x − y| − W˙ iν,1/2 (−ikd+ )Miν,1/2 (−ikd− )],


with k = (−2mz/2 )1/2 ((k) > 0), ν = 1/(kaB ), d+ = |x| + |y| + |x − y| and d− = |x| + |y| − |x − y|, while Γ (u) is the familiar gamma-function and Wiν,1/2 (u), Miν,1/2 (u) are Whittaker functions [31] (W˙ iν,1/2 (u) = ∂Wiν,1/2 (u)/∂u, M˙ iν,1/2 (u) = ∂Miν,1/2 (u)/∂u).  y; z) is analytical in z in the whole complex plane, except on the negaGreen function G(x, tive real axis ((z) ≤ 0, (z) = 0) which is a cut line, and also at z = zn = −EH /n2 (n ≥ 1) which are simple poles (of course, such singularities are controlled by the spectrum of Hpe ).  y; z) diverges as m/(2π2 |x − y|), as shown by expanding Whittaker When x → y, G(x, functions for arguments close to −2ik|x|. That 1/|x − y|-behavior, is also displayed by free 0 (x, y; z) = m exp(ik|x − y|)/(2π2 |x − y|), and it can be interpreted by Green function G quoting that (B.2) reduces to Poisson equation for |x − y| small. It is convenient to define  (x, y; z) = G(x,  y; z) − G 0 (x, y; z) + drG 0 (x, r; z)Vpe (|r|)G 0 (r, y; z) H  =

0 (x, r1 ; z)Vpe (|r1 |)G 0 (r1 , r2 ; z)Vpe (|r2 |)G(r  2 , y; z) dr1 dr2 G


where the second equality follows from a standard identity for interacting and free Green (x, y; z) has the same analytical properties as G(x,  y; z). Using functions. That function H

A. Alastuey et al.

(x, x; z) reads above expansions of Whittaker functions, we find that H  (x, x; z) = −imkΓ (1 − iν) 2W˙ iν,1/2 M˙ iν,1/2 − W¨ iν,1/2 Miν,1/2 H 2π2  imk − Wiν,1/2 M¨ iν,1/2 (−2ik|x|) − 2π2  exp(2ik|x − r|) m2 e 2 dr − . 2 4 4π  r|x − r|2


−G 0 which diverges as −2m2 e2 ln(|k||x|)/(π4 ) when x → 0, Notice that, contrarily to G 0  0 Vpe G H (x, x; z) remains finite at x = 0 thanks to the addition of the integral of G in (B.4).  y; z) is nothing but the Laplace transform with respect to β of density matrix Since G(x,

x| exp(−βHpe )|y, the standard inversion formula provides  1  y; z)

x| exp(−βHpe )|y = dz exp(βz)G(x, (B.6) 2iπ Dσ where Dσ is any straight line parallel to imaginary axis and which cuts real axis at σ > −EH .  into (B.6), we find that the contribu in terms of G 0 , Vpe and H Inserting decomposition of G 0 give raise to free density matrix x| exp(−βKpe )|y = exp(−|x − tion of terms involving G y|2 /(2λ2pe ))/(2πλ2pe )3/2 . The resulting expression of x| exp(−βHpe )|y can be specified to (x, x; z) is finite. The integral of H (x, x; z) exp(βz) diagonal elements x = y because H along Dσ is then transformed by applying Cauchy’s theorem with contour Cσ,δ shown in (x, x; z) exp(βz) is analytical inside Cσ,δ exFig. 13 (−EH /4 < δ < −EH ). Function H cept at z = z1 = −EH which is a simple pole with residue |ψ0 (x)|2 exp(−βEH ). Moreover, it satisfies Jordan’s lemma on the circular parts of Cσ,δ which connect Dδ to Dσ , so the corresponding parts of the contour integral vanish when the radius is sent to infinity. This provides

x| exp(−βHpe )|x = |ψ0 (x)|2 exp(−βEH ) −  1 βe2 + 1 + (2πλ2pe )3/2 |x|

1 2iπ  1

(x, x; z) dz exp(βz)H Dδ

 √ dsΦ(|x|/( 2s(1 − s)λpe ))



where Φ is the familiar Error function [31]. Last term in (B.7) is bounded by a power of β. Along Dδ , index iν of Whittaker functions follows a closed curve in complex plane which cuts real axis at non-strictly positive integers. At the same time, their argument u = −2ik|x| (x, x; z)| remains bounded remains inside sector −3π/8 < arg(u) < 3π/8. Consequently, |H by a constant when z runs along Dδ . The modulus of the corresponding integral in (B.7) is then bounded by a power of β times exp(βδ). Taking into account above power-law bounds for the contributions of truncated terms in [exp(−βHpe )]TMayer , and noting that ψ0 (x) decays exponentially fast for |x| > βe2 , we eventually obtain  dx x|[exp(−βHpe )]TMayer |x = exp(−βEH ), (B.8) |x|<βe2

Exact Results for Thermodynamics of the Hydrogen Plasma Fig. 13 Path in complex plane

discarding additional terms which are exponentially smaller than exp(−βEH ) when β → ∞. For |x| > βe2 , it is convenient to use Feynman-Kac expression of x|[exp(−β × Hpe )]TMayer |x. Within the variable change x = βe2 v, the corresponding integral is rewritten as  dx x|[exp(−βHpe )]TMayer |x |x|>βe2

 2 3   1 βe = dv D(ξ ) (2π)3/2 λpe |v|>1 −1  −1   1   1      λpe λpe × exp ds v + 2 ξ (s) ds v + 2 ξ (s) −1− βe βe 0 0 −1 2 −1 3   1   1      1 1 λpe λpe − ds v + 2 ξ (s) − ds v + 2 ξ (s) . 2 0 βe 6 0 βe


When β → ∞, ratio λpe /βe2 vanishes. Then, potential |v + βepe2 ξ (s)|−1 can be merely replaced by |v| because the mean extension of ξ (s) is of order 1. The corresponding functional integral over ξ (s) reduces to 1 by normalization of Wiener measure D(ξ ), and the remaining λ

A. Alastuey et al.

integral over v is a pure number. Thus, the integral in the l.h.s. of (B.9) behaves as a power of β, and it becomes exponentially smaller than (B.8) when β → ∞. This implies Z(1, 1) = 4 exp(−βEH )


discarding terms which are exponentially smaller when β → ∞. Next corrections to that leading behavior arising from excited atomic states, can be readily obtained within a similar approach by adjusting the position δ between two successive poles, i.e. zn+1 < δ < zn with n > 1. B.2 Behavior of Z(1, 2) The low-temperature behavior of Z(1, 2), which involves one proton and two electrons, is determined by a straightforward extension of previous methods applied to Z(1, 1). If an exact expression of three-body Green function is not available, its main properties of interest can be guessed, under rather weak assumptions, by exploiting relations and analogies with its free counterpart, as well as known results on the spectrum of H1,2 . Such properties appear to be quite natural extensions of the exact behaviors observed in the two-body case. − − Let R = (mp Ra + me r1 + √ me r2 )/MH be the position of center of mass √ (MH = mp + 2me ), while x1 = (r1 − r2 )/ 2 and x2 = (mp /MH − )1/2 (r1 + r2 − 2Ra )/ 2 are the reduced variables. After expressing H1,2 in terms of those variables, we find that three-body partition function Z(1, 2) can be rewritten as  ∗ )|x1 x2  Z(1, 2) = 2 dx1 dx2 {2 x1 x2 | exp(−βH1,2 ∗ )|x1 x2  + · · ·} − −x1 x2 | exp(−βH1,2


with reduced Hamiltonian √ 2 2 2 e2 2e Δx1 − Δx2 + √ − 2me 2me 2|x1 | |x1 + (MH − /mp )1/2 x2 | √ 2 2e − . |x1 − (MH − /mp )1/2 x2 |

∗ =− H1,2


Off-diagonal matrix elements in the r.h.s. of (B.11) are associated with the exchange of the electrons. First potential term in (B.12) describes Coulomb repulsion between those electrons, while the second and third ones account for Coulomb attractions between the proton and each electron. In the double integral involved in (B.11), we make a partition of space integration into three domains Ω (i) (i = 0, 1, 2), such that i, and only i, sides of triangle (0, x1 , x2 ) are smaller than βe2 in Ω (i) . ∗ ) as inverse Laplace For x1 , x2 inside Ω (2) , we express matrix elements of exp(−βH1,2  transforms of Green function G(x1 , x2 ; y1 , y2 ; z) solution of ∗  1 , x2 ; y1 , y2 ; z) = δ(x1 − y1 )δ(x2 − y2 ) + z)G(x (H1,2


with suitable boundary conditions. That Green function is analytical in z in the whole complex plane, except on a part of real axis with (z) < −EH − , while z1 = −EH − is a simple (isolated) pole with residue ψ0 (x1 , x2 )ψ 0 (y1 , y2 ) (ψ0 is the groundstate wavefunction of

Exact Results for Thermodynamics of the Hydrogen Plasma ∗  1 , x2 ; y1 , y2 ; z) disH1,2 with energy EH − ). For a given z not close to its singularities, G(x 0 (x1 , x2 ; y1 , y2 ; z) plays a position-dependence analogous to that of free Green function G solution of Helmholtz equation in six dimensions. In particular, for xi close to yi , potential  1 , x2 ; y1 , y2 ; z) also diverges as 1/[(x1 − y1 )2 + (x2 − terms in (B.13) can be omitted and G(x 2 2 y2 ) ] , i.e. the Coulomb potential in six dimensions. In order to handle such divergences for diagonal matrix elements, we consider identity

 1 , x2 ; y1 , y2 ; z) G(x 0 (x1 , x2 ; y1 , y2 ; z) =G  ∗  1 , r2 ; y1 , y2 ; z) 0 (x1 , x2 ; r1 , r2 ; z)V1,2 − dr1 dr2 G (r1 , r2 )G(r


∗ ∗ where V1,2 is the potential part of H1,2 . Successive iterations of formula (B.14) gen∗  1 , x2 ; y1 , y2 ; z) in powers of V1,2 erate the perturbative expansion of G(x . We define ∗ 5 (x1 , x2 ; y1 , y2 ; z) as G(x  1 , x2 ; y1 , y2 ; z) minus the first six terms (up to order (V1,2 H ) in  cluded) of that expansion. Similarly to (B.4), H has the same analytical properties as G, (x1 , x2 ; x1 , x2 ; z) is now finite. After inserting the expression of G  in terms of G 0 , while H ∗ ∗  into the inversion formula for diagonal matrix element of exp(−βH1,2 V1,2 and H ), we ∗  find that terms built with G0 and V1,2 are bounded by powers of β. For dealing with the , we introduce a closed contour Cσ,δ in complex plane similar to that contribution of H  below z1 . Along shown in Fig. 13, with z2 < δ < z1 and −z2 the first singularity of H  that contour, z stays at a finite distance at any singularity of H . Also, for |z| large, in ∗  ∗   as a spatial integral of G 0 V1,2 G0 · · · V1,2 G (similar to that involved the expression of H    in (B.4)), G can be replaced by G0 . Therefore, H goes to zero as a power of k, as shown by variable changes ri → |k|ri (integrals over the ri ’s do converge thanks to exponen0 ’s). Hence, we conclude that tially decaying factors exp(ik|ri − rj |) arising from the G  |H (x1 , x2 ; x1 , x2 ; z)| remains bounded along Dδ , while Jordan’s lemma applies to the inte(x1 , x2 ; x1 , x2 ; z) exp(βz) upon the circular parts of Cσ,δ . This provides gral of H ∗ 2 x1 x2 | exp(−βH1,2 )|x1 x2  = 2|ψ0 (x1 , x2 )|2 exp(−βEH − )


discarding terms which are exponentially smaller. Within similar methods, we can also es∗ timate the off-diagonal matrix element −x1 x2 | exp(−βH1,2 )|x1 x2 . Contributions of terms ∗  involving G0 ’s and V1,2 ’s are readily bounded by powers of β by rescaling positions in units  is treated as above since H (−x1 , x2 ; x1 , x2 ; z) is bounded along of λe . Contribution of H ∗ is Dδ and decays sufficiently fast for |z| large. After using ψ0 (−x1 , x2 ) = ψ0 (x1 , x2 ) (H1,2 invariant under transformation x1 → −x1 at fixed x2 ), we find ∗

−x1 x2 | exp(−βH1,2 )|x1 x2  = |ψ0 (x1 , x2 )|2 exp(−βEH − ),


discarding terms which are exponentially smaller. Next terms · · · in the r.h.s. of (B.11), which arise from truncation in [exp(−βH1,2 )]TMayer , can be also estimated by similar techniques. For instance, term 


dτ exp[−(β − τ )(H1,1 + H0,1 )]Vat,e exp[−τ (H1,1 + H0,1 )] 0


A. Alastuey et al.

provides a contribution which can be rewritten as the inverse Laplace transform of 

∗ ∗ )−1 |y1 y2 Vat,e (y1 , y2 ) y1 y2 |(z + Hpe,e )−1 |x1 x2  dy1 dy2 x1 x2 |(z + Hpe,e

with ∗ Hpe,e

√ 2 2 2 2e =− Δx − Δx − 2me 1 2me 2 |x1 + (MH − /mp )1/2 x2 |

and e2



2e2 Vat,e (x1 , x2 ) = √ . − 2|x1 | |x1 − (MH − /mp )1/2 x2 |


∗ Green function defined as matrix elements of (z + Hpe,e )−1 , is analytical with respect to z in the whole complex plane, except on part (z) ≤ −EH of the real axis. In the Laplace inversion formula, we introduce a contour analogous to that of Fig. 13 with δ > −EH . We ∗ )−1 |y1 y2  which remains finite at xi = yi . When also define a regular part of x1 x2 |(z + Hpe,e z follows Dδ , (k) remains larger than a given positive constant, so previous regular part is bounded by an exponentially decaying function of [(x1 − y1 )2 + (x2 − y2 )2 ]1/2 (for large separations of the arguments, Coulomb potential terms vanish so Green functions behave as their free counterparts which decay exponentially on a scale ((k))−1 ). This implies that (B.18) remains bounded by a constant along Dδ . Contribution of (B.17) is then found to be bounded by a power of β times exp(βδ), with δ arbitrarily close to −EH . Since −EH < −EH − , that contribution is exponentially smaller than (B.15) and (B.16). Contributions of all the other truncated terms in [exp(−βH1,2 )]TMayer behave similarly because groundstate energies of Hamiltonians (H1,0 + H0,2 ) and (H1,0 + H0,1 + H0,1 ) (which both vanish) are strictly larger than EH − . Since volume of Ω (2) is bounded by a constant times (βe2 )6 on one hand, while ψ0 (x1 , x2 ) decays exponentially fast for |xi | large on another hand, we eventually obtain

 Ω (2)

∗ ∗ dx1 dx2 {2 x1 x2 | exp(−βH1,2 )|x1 x2  − −x1 x2 | exp(−βH1,2 )|x1 x2  + · · ·}

= exp(−βEH − )


discarding terms which are exponentially smaller. For x1 , x2 inside Ω (1) , two of three distances |x1 |, |x1 − x2 |, |x2 | are larger than βe2 . For instance, we may have both |x1 | and |x1 − x2 | larger than βe2 , while |x2 | is smaller than βe2 . For such configurations, both distances |x1 | and |x1 − (MH − /mp )1/2 x2 | are larger than βe2 . √ 1 ∗ )|x1 x2 , potentials 0 dse2 / 2|x1 + In the Feynman-Kac formula for x1 x2 | exp(−βH1,2 1 √ 2 λe ξ 1 (s)| and − 0 ds 2e /|x1 + λe ξ 1 (s) − (MH − /mp )1/2 (x2 + λe ξ 2 (s))| can then be re√ √ placed by their classical counterparts e2 / 2|x1 | and − 2e2 /|x1 − (MH − /mp )1/2 x2 | re∗ )|x1 x2  spectively, because λe /βe2 goes to zero when β diverges. Thus, x1 x2 | exp(−βH1,2 behaves as ∗

x1 x2 | exp(−βHpe,e )|x1 x2  exp(−βVat,e (x1 , x2 ))


∗ at leading order, where Hpe,e and Vat,e (x1 , x2 ) are given by (B.19) and (B.20) respectively. A similar estimation holds for truncated terms in [exp(−βH1,2 )]TMayer built with powers of

Exact Results for Thermodynamics of the Hydrogen Plasma

Vat,e , so the corresponding full contribution integrated upon considered configurations behaves as  ∗ 2 dx1 dx2 x1 x2 | exp(−βHpe,e )|x1 x2  |x1 |,|x1 −x2 |>βe2 ,|x2 |<βe2

 1 × exp(−βVat,e (x1 , x2 )) + βVat,e (x1 , x2 ) − (βVat,e (x1 , x2 ))2 2  1 3 + (βVat,e (x1 , x2 )) . 6


∗ )|x1 x2  is bounded According to above properties of Green functions, x1 x2 | exp(−βHpe,e by some power of β times exp(βδ) with −EH < δ < −EH − . Therefore, integral (B.23) is also bounded by a power of β times exp(βδ), because purely classical integral

 |x1 |,|x1 −x2 |>βe2 ,|x2 |<βe2

dx1 dx2 exp(−βVat,e (x1 , x2 )) + βVat,e (x1 , x2 )

1 1 − (βVat,e (x1 , x2 ))2 + (βVat,e (x1 , x2 ))3 2 6


is proportional to (βe2 )6 as shown by rescaling xi in units of βe2 . A similar analysis applies to the contributions of the other terms in [exp(−βH1,2 )]TMayer . When off-diagonal matrix elements are involved, we use bounds inferred from properties of Green functions, which decay exponentially fast with respect to relative distances between differ∗ ent arguments. For instance, | −x1 x2 | exp(−βH1,2 )|x1 x2 | is bounded by a constant times exp(−βEH − ) exp(−c|x1 |/aB ) with c > 0, so  −

|x1 |,|x1 −x2 |>βe2 ,|x2 |<βe2

∗ dx1 dx2 −x1 x2 | exp(−βH1,2 )|x1 x2 


decays exponentially faster than exp(−βEH − ). Previous analysis can be also repeated for the other configurations belonging to Ω (1) , i.e. {|x2 |, |x1 − x2 | > βe2 , |x1 | < βe2 } and {|x1 |, |x2 | > βe2 , |x1 − x2 | < βe2 }. We eventually find that  Ω (1)

∗ ∗ dx1 dx2 {2 x1 x2 | exp(−βH1,2 )|x1 x2  − −x1 x2 | exp(−βH1,2 )|x1 x2  + · · ·}


decays exponentially faster than exp(−βEH − ). For x1 , x2 inside Ω (0) , all distances |x1 |, |x1 − x2 |, and |x2 | are larger than βe2 . For diagonal matrix elements, potential parts can be treated classically at leading order, as immediately seen from Feynman-Kac formula by noting that λe /βe2 vanishes. Such matrix elements then behave as their free counterparts times classical Boltzmann factors. The corresponding full contribution integrated upon Ω (0) is shown to be proportional to (βe2 /λe )6 , as shown by variable changes xi = βe2 vi . The contribution of remaining terms with offdiagonal matrix elements decays exponentially faster than exp(−βEH − ), thanks to the existence of bounds which are proportional to exp(−βEH − ) and decay exponentially fast for large separations (over a finite length scale proportional to aB ). Thus, and like (B.26), contribution of Ω (0) to Z(1, 2) also decays exponentially faster than exp(−βEH − ), so we even-

A. Alastuey et al.

tually obtain Z(1, 2) = 2 exp(−βEH − )


discarding terms which are exponentially smaller when β → ∞. B.3 Behaviors of Z(2, 1), Z(2, 2), . . . The low-temperature behaviors of Z(2, 1) and Z(2, 2) can be also determined by previous methods introduced for studying Z(1, 1) and Z(1, 2). Again, analytic properties of Green ∗ −1 ∗ −1 ) and (z + H2,2 ) play a crucial role in the functions associated with resolvents (z + H2,1 derivations. Such functions are analytical in the whole complex plane, except on a part of real axis with (z) < −EH + or (z) < −EH2 , while z1 = −EH − or z1 = −EH2 is a simple 2 (isolated) pole. Along integration contours analogous to that described in Fig. 13, they can be bounded as above, within quite plausible arguments based on the properties of their free counterparts, solutions of Helmholtz equation in six or nine dimensions. Integration space upon reduced positions is splitted into several parts according to the values of relative distances compared to βe2 . The parts inside which all relative distances are smaller than βe2 , provide the leading contributions, i.e. Z(2, 1) = 2 exp(−βEH + )


Z(2, 2) = exp(−βEH2 )




discarding terms which are exponentially smaller when β → ∞. The analysis can be applied to any partition function Z(Np , Ne ). However, when the infimum of reduced Hamiltonian HN∗p ,Ne is not separated from the rest of the spectrum (i.e. in the corresponding groundstate, the Np protons and the Ne electrons are not binded together), the first singularity (with the largest real part) z1 = −EN(0)p ,Ne of Green function associated with (z + HN∗p ,Ne )−1 , is a branching point which is not isolated from other singularities. Then, contour Cσ,δ cuts real axis at δ > z1 , so the previous methods show that |Z(Np , Ne )| is bounded by a power of β times exp(βδ). A more detailed analysis of the behavior of Green function for z close to z1 is then required for determining the precise leading behavior of Z(Np , Ne ). Nevertheless, since previous bound is valid for δ arbitrarily close to z1 , it is quite reasonable to assume that Z(Np , Ne ) then behaves as a power of β times exp(−βEN(0)p ,Ne ) (0) (such a behavior is indeed observed for Z(2, 0) with E2,0 = 0).

Appendix C: Leading Contributions of Interactions between Atoms and Ionized Charges C.1 Expression of W (1, 1|1, 1) The low-temperature behavior of bare contributions of Figs. 8b and 8c, is determined along similar lines as that of polarization contribution (A.13). The integrals of interest are again expressed in terms of the atom mass centers and of the reduced variables. Matrix elements are also evaluated via insertions of the closure relation for a suitable basis. Each eigenstate in that basis, is the product of plane waves describing mass center motions, times atomic

Exact Results for Thermodynamics of the Hydrogen Plasma

internal wavefunctions. The resulting integrals over times τi are readily performed for each set of involved eigenstates. According to the scaling prescriptions defined in Sect. 3, full bare contribution of Figs. 8b and 8c is then rewritten as (3.41) plus terms which decay exponentially faster than ρ ∗ exp(βEH ). Like Z(1, 1) or S3 (1, 1), function W (1, 1|1, 1) is determined by atomic groundstate contributions, and further contributions of excited states might be bounded by methods similar to that exposed in Appendix B:. Similarly to expression (A.16) for S3 (1, 1), W (1, 1|1, 1) reduces to the product exp(−2βEH ) of atomic groundstate Boltzmann factors, times a function of β|EH | which remains bounded by a power law at low temperatures. Its asymptotic form when β → ∞ is given by (3.42), where cat,at is the pure numerical coefficient cat,at =

 (at,at) |D00,00 (K)|2 2 M dK π m K6   (at,at)  |D0p (K)|2 2|EH | 1 ,0p2 dK (p ) + (p ) K4 (EH 1 + EH 2 − 2EH + 2 K 2 /(MaB2 )) (p1 ,p2 ) =(0,0)   (at,at) (at,at) (at,at) D00,00 (K)D00,00 (Q − K)D00,00 (−Q) M 2 dKdQ m K 4 Q4 |K − Q|2   2|EH | dKdQ (p ) + (p2 ) 1 (EH + EH − 2EH + 2 K 2 /(MaB2 )) (p1 ,p2 ,p3 ,p4 ) =(0,0,0,0) −1 × 3 π

× ×

2|EH | (p )

(p )

(EH 3 + EH 4 − 2EH + 2 Q2 /(MaB2 )) (at,at) D0p (K)Dp(at,at) (Q − K)Dp(at,at) (−Q) 1 p3 ,p2 p4 1 ,0p2 3 0,p4 0

K 2 Q2 |K − Q|2



(at,at) (K) In (C.1), K and Q are dimensionless (units aB−1 ) wavenumbers, while function D0p 1 ,0p2 reduces to     m m (at,at) ∗ ∗ (K) =

ψ | exp −i K · r  ψ | exp i K · r |ψ |ψp2  D0p 0 p 0 1 1 ,0p2 me me     m m ∗ ∗ + ψ0 | exp i K · r |ψp1  ψ0 | exp −i K · r |ψp2  mp mp     m m K · r∗ |ψp2  − ψ0 | exp −i K · r∗ |ψp1  ψ0 | exp −i me mp     m m ∗ ∗ K · r |ψp1  ψ0 | exp i K · r |ψp2 . (C.2) − ψ0 | exp i mp me

C.2 Expressions of W (1, 1|1, 0) and W (1, 1|0, 1) A straightforward extension of previous methods provides the low-temperature behaviors of bare contributions of Figs. 9a–f. Using again the scaling prescriptions defined in Sect. 3, the corresponding full bare contribution is then rewritten as (3.45) plus terms which decay exponentially faster than ρ ∗ exp(βEH ). Functions W (1, 1|1, 0) and W (1, 1|0, 1) are also determined by atomic groundstate contributions. They behave as exp(−βEH ) times functions

A. Alastuey et al.

of β|EH | which remain bounded by power laws. Their asymptotic forms when β → ∞ are given by 3.46, where pure numerical constants reduce to (2) = cat,α

 2mα M |D (at,α) (K)|2 2 dK 00 6 π m(mα + M) K   (at,α)  |D0p (K)|2 2|EH | 1 dK (p ) + K4 (EH 1 − EH + 2 K 2 /(2MaB2 ) + 2 K 2 /(2mα aB2 )) p1 =0

 2  (at,α) (at,α) D (at,α) (K)D00 (Q − K)D00 (−Q) −1 2mα M dKdQ 00 3 4 4 2 π m(mα + M) K Q |K − Q|  2|EH | + dKdQ (p ) 1 2 2 (EH − EH +  K /(2MaB2 ) + 2 K 2 /(2mα aB2 )) (p1 ,p2 ) =(0,0) ×


2|EH | (p )

(EH 2 − EH + 2 Q2 /(2MaB2 ) + 2 Q2 /(2mα aB2 ))  (at,α) D0p (K)Dp(at,α) (Q − K)Dp(at,α) (−Q) 1 p2 1 20 × , K 2 Q2 |K − Q|2


with (at,p)

(at,e) (K) D0p1 (K) = −D0p 1     m m = ψ0 | exp −i K · r∗ |ψp1  − ψ0 | exp i K · r∗ |ψp1 . me mp


References 1. Alastuey, A., Ballenegger, V., Cornu, F.: Low temperature isotherms of the hydrogen plasma: comparisons between analytical and Monte Carlo results. ENS Lyon preprint 2. Alastuey, A., Martin, Ph.A.: Absence of exponential clustering for static quantum correlations and timedisplaced correlations in charged fluids. Eur. Phys. Lett. 6, 385–390 (1988) 3. Alastuey, A., Martin, Ph.A.: Absence of exponential clustering in quantum Coulomb fluids. Phys. Rev. A 40, 6485–6520 (1989) 4. Alastuey, A., Perez, A.: Virial expansion of the equation of state of a quantum plasma. Europhys. Lett. 20, 19–24 (1992) 5. Alastuey, A., Perez, A.: Virial expansion for quantum plasmas: Fermi–Bose statistics. Phys. Rev. E 53, 5714–5728 (1996) 6. Alastuey, A., Cornu, F., Perez, A.: Virial expansion for quantum plasmas: diagrammatic resummations. Phys. Rev. E 49, 1077–1093 (1994) 7. Alastuey, A., Cornu, F., Perez, A.: Virial expansion for quantum plasmas: Maxwell–Boltzmann statistics. Phys. Rev. E 51, 1725–1744 (1995) 8. Alastuey, A., Ballenegger, V., Cornu, F., Martin, Ph.A.: Screened cluster expansions for partially ionized gases. J. Stat. Phys. 113, 455–503 (2003) 9. Alastuey, A., Cornu, F., Martin, Ph.A.: Van der Waals forces in presence of free charges: an exact derivation from equilibrium quantum correlations. J. Chem. Phys. 127, 054506 (2007) 10. Ballenegger, V.: Étude des phénomènes d’écran et de polarisation dans un plasma quantique par la méthode des graphes de Mayer. PhD thesis, École Polytechnique et Fédérale de Lausanne (2002) 11. Ballenegger, V., Martin, Ph.A.: Quantum Coulomb systems: some exact results in the atomic limit. Physica A 306, 59–67 (2002) 12. Ballenegger, V., Martin, Ph.A.: Dielectric versus conductive behavior in quantum gases: Exact results for the Hydrogen plasma. Physica A 328, 97–144 (2003)

Exact Results for Thermodynamics of the Hydrogen Plasma 13. Ballenegger, V., Martin, Ph.A., Alastuey, A.: Quantum Mayer graphs for Coulomb systems and the analog of the Debye potential. J. Stat. Phys. 108, 169–211 (2002) 14. Brydges, D.C., Martin, Ph.A.: Coulomb systems at low density: a review. J. Stat. Phys. 96, 1163–1330 (1999) 15. Ceperley, D.M.: Fermion nodes. J. Stat. Phys. 63, 1237–1267 (1991) 16. Ceperley, D.M.: Path integrals in the theory of condensed helium. Rev. Mod. Phys. 67, 279–355 (1995) 17. Ceperley, D.M.: In: Binder, E.K., Ciccotti, G. (eds.) Monte Carlo and Molecular Dynamics of Condensed Matter Systems. Editrice Compositori, Bologna (1996) 18. Conlon, J.G., Lieb, E.H., Yau, H.T.: The Coulomb gas at low temperature and low density. Commun. Math. Phys. 125, 153–180 (1989) 19. Cornu, F.: Correlations in quantum plasmas, I: resummations in Mayer-like diagrammatics. Phys. Rev. E 53, 4562 (1996) 20. DeWitt, H.E., Schlanges, M., Sakakura, A.Y., Kraeft, W.D.: Low density expansion of the equation of state for a quantum electron gas. Phys. Lett. A 197, 326–329 (1995) 21. Dyson, F., Lenard, A.: Stability of matter I. J. Math. Phys. 8, 423–434 (1967) 22. Dyson, F., Lenard, A.: Stability of matter II. J. Math. Phys. 9, 698–711 (1968) 23. Ebeling, W.: Ann. Phys. Leipz. 19, 104 (1967) 24. Ebeling, W.: Coulomb interaction and ionization equilibrium in partially ionized plasmas. Physica 63, 293 (1969) 25. Ebeling, W.: Statistical derivation of the mass-action law for interacting gases and plasmas. Physica 73, 573–584 (1974) 26. Ebeling, W., Kraeft, W.D., Kremp, D.: Theory of Bound States and Ionization Equilibrium in Plasmas and Solids. Akademie, Berlin (1976) 27. Fefferman, C.: The atomic and molecular nature of matter. Rev. Math. Iberoam. 1, 1–44 (1985) 28. Fefferman, C.: The n-body problem in quantum mechanics. Commun. Pure Appl. Math. 39(S1), 67–109 (1986) 29. Gell-Mann, M., Brueckner, K.A.: Correlation energy of an electron gas at high density. Phys. Rev. 106, 364–368 (1957) 30. Ginibre, J.: Some applications of functional integration in statistical mechanics. In: DeWitt, C., Stora, R. (eds.) Statistical Mechanics and Quantum Field Theory. Gordon and Breach, les Houches (1971) 31. Gradsteyn, I.S., Ryzhik, I.M.: In: Jeffrey, A. (ed.) Tables of Integrals, Series and Products. Academic Press, London (1994) 32. Gruter, P., Laloe, F.: Ursell operators in statistical physics I. J. Phys. I. (France) 5, 1255 (1995) 33. Gruter, P., Laloe, F.: Ursell operators in statistical physics II. J. Phys. I. (France) 5, 181 (1995) 34. Gruter, P., Laloe, F.: Ursell operators in statistical physics III. J. Phys. I. (France) 7, 485 (1997) 35. Hill, L.T.: Statistical Mechanics. McGraw-Hill, New York (1956) 36. Hostler, L.: Coulomb’s Green functions and the Furry’s approximation. J. Math. Phys. 5, 591 (1964) 37. Kahlbaum, T.: The quantum-diffraction term in the free energy for Coulomb plasma and the effectivepotential approach. J. Phys. IV 10(P5), 455 (2000) 38. Kleinert, H.: Path Integrals in Quantum Mechanics, Statistics, Polymer Physics and Financial Markets. World Scientific, Singapore (2004) 39. Kraeft, W.D., Kremp, D., Ebeling, W., Ropke, G.: Quantum Statistics of Charged Particle Systems. Plenum, New York (1986) 40. Krauth, W.: Statistical Mechanics: Algorithms and Computations. Oxford University Press, Oxford (2006) 41. Kremp, D., Kraeft, W.D., Lambert, A.J.M.D.: Equation of state and ionization equilibrium for non-ideal plasmas. Physica A 127, 72–86 (1984) 42. Landau, L.D., Lifchitz, E.M.: Quantum Mechanics, 3rd edn. Course of Theoretical Physics, vol. 3. Pergamon, Oxford (1977) 43. Lebowitz, J., Pena, R.: Low density form of the free energy of real matter. J. Chem. Phys. 59, 1362–1364 (1973) 44. Lieb, E.H., Lebowitz, J.: The constitution of matter: existence of thermodynamics for systems composed of electrons and nuclei. Adv. Math. 9, 316–398 (1972) 45. Macris, N., Martin, Ph.A.: Ionization equilibrium in the proton-electron gas. J. Stat. Phys. 60, 619–637 (1990) 46. Martin, Ph.A.: Quantum Mayer graphs: applications to Bose and Coulomb gases. Acta Phys. Pol. B 34, 3629 (2003) 47. Militzer, B., Ceperley, D.M.: Path integral Monte Carlo calculation of the deuterium hugoniot. Phys. Rev. Lett. 85, 1890–1893 (2000) 48. Militzer, B., Ceperley, D.M.: Path integral Monte Carlo simulation of the low-density hydrogen plasma. Phys. Rev. E 63, 066404 (2001)

A. Alastuey et al. 49. Militzer, B., Pollock, E.L.: Variational density matrix method for warm, condensed matter: application to dense hydrogen. Phys. Rev. E 61, 3470–3482 (2000) 50. Montroll, E.W., Ward, J.C.: Quantum statistics of interacting particles: General theory and some remarks on properties of an electron gas. Phys. Fluid 1, 55 (1958) 51. Morita, T.: Equation of state of high temperature plasma. Prog. Theor. Phys. 22, 757 (1959) 52. All spectroscopic calculations were performed by V. Robert, who used a coupled cluster (CCSD(T)) approach including an extended basis set for hydrogen atoms (4s3p2d1f ) 53. Roepstorff, G.: Path Integral Approach to Quantum Physics. Springer, Berlin (1994) 54. Rogers, F.J.: Statistical mechanics of Coulomb gases of arbitrary charge. Phys. Rev. A 10, 2441 (1974) 55. Rogers, F.J.: Equation of state of dense, partially degenerate, reacting plasmas. Phys. Rev. A 24, 1531 (1981) 56. Rogers, F.J.: Occupation numbers for reacting plasmas—the role of the Planck–Larkin function. Astrophys. J. 310, 723 (1986) 57. Rogers, F.J.: A distribution function approach for effective occupation numbers and the equation of state of hydrogen plasmas. Astrophys. J. 352, 689 (1990) 58. Rogers, F.J.: In: Chabrier, G., Schatzman, E. (eds.) The Equation of State in Astrophysics. Cambridge University Press, New York (1994) 59. Rogers, F.J., Young, D.A.: Validation of the activity expansion method with ultrahigh pressure shock equations of state. Phys. Rev. E 56, 5876 (1997) 60. Saha, M.: Philos. Mag. 40, 472 (1920) 61. Schulman, L.S.: Techniques and Applications of Path Integrals. Wiley, New York (1981) 62. Simon, B.: Functional Integration and Quantum Physics. Academic, New York (1979)