We characterize in a novel manner the physical properties of the low temperature Fermi gas in the degenerate domain as a function of temperature and chemical potential. For the first time we obtain low temperature T results in the domain where several fermions are found within a de Broglie spatial cell. In this regime, the usual high degeneracy Sommerfeld expansion fails. The other known semi-classical Boltzmann domain applies when fewer than one particle is found in the de Broglie cell. We also improve on the understanding of the Sommerfeld expansion in the regime where the chemical potential is close to the mass and also in the high temperature regime. In these calculcations we use a novel characterization of the Fermi distribution allowing the separation of the finite and zero temperature phenomena. The relative errors of the three approximate methods (Boltzmann limit, Sommerfeld expansion, and the new domain of several particles in the de Broglie cell) are quantified.
Dedication
We dedicate this work to Louis de Broglie.
100 years ago Louis de Broglie recognized the wave character of matter, the linchpin in the physical interpretation of our results.
Introduction
Motivation
Many interesting effects related to the Pauli principle appear at finite, but low temperatures, in vastly different physical systems [1, 2, 3, 4]. The limit T→0 of the Fermi-Dirac (FD) quantum system is subtle. We show analytical results which depend on particle number density within the de Broglie volume VdB, in D-dimensions defined by
VdB=LdBD,LdB=⟨p⟩T2πℏ,⟨p⟩T=⟨E2−m2⟩T,(1)
where angle brackets denote thermal averages. Any physical systems maintaining a prescribed particle number density in the T→0 limit requires a simultaneous chemical potential μ limit. For this reason, key physical features of low-temperature FD quantum gasses can be missed if the appropriate relation between μ and T is not maintained in the low temperature limit. In this work, we show that a careful treatment of the simultaneous limit leads to novel asymptotic expansions of FD thermal averages. While in principle this is not an unexpected finding, we have found no mention of such mathematical considerations in literature.
The asymptotic limit of integrals of the FD distribution [5, 6] are of interest to physicists studying many diverse statistical quantum systems [7, 8, 9, 10, 11, 12] and behavior around the Fermi surface [13, 14]. Other use cases include compact astrophysical systems (white dwarfs, neutron stars, quark stars) [3, 15, 16], Early Universe phenomenology [17, 18, 19, 20], quark-gluon plasmas (QGP) [2, 21, 22, 23], and in any situation when finite temperature Fermi effects are important. When investigating the low temperature limit of a Fermi gas, a systematic approach to approximating the physical integral involves employs the Sommerfeld expansion. It is worth noting that the Sommerfeld expansion, as described in Ref. [24], is asymptotic in nature. It is within this context that we often see matter-of-fact statements that the “Sommerfeld expansion fails” in various regimes. A textbook example is seen on page 38 of Ref. [4] (for the Sommerfeld expansion in this text, see p. 24). However, there appears to have been no progress towards a wholly distinct expansion which could rectify this issue. This situation seems to have lasted now for nearly a century.
After several in-depth reviews of literature, we conclude that a full understanding of the domain of validity of the Sommerfeld expansion in the (T,Δμ)-plane
Δμ≡μ−m,(2)
is lacking. The Sommerfeld expansion fails for regimes where the corrections to the T→0 limit cannot be taken to be small. This makes it difficult to search for alternative expansions that are valid in other regions of the (T,Δμ)-plane when dealing with multiple scales. One generally requires specialized asymptotic expansions that are adapted to specific regimes under consideration. This is particularly relevant when Δμ is small, a regime that the traditional Sommerfeld expansion was not designed for and which we consider here. Specifically, expanding the Sommerfeld result around the small chemical potential may lead to encountering singularities which underscore the necessity for a better tool to study this scenario.
In this work we explore the different domains of low temperature FD gas, characterize the domain of validity of Sommerfeld expansion in physical terms, and obtain a novel approximation applicable to domains of statistical parameters where the “expansion fails”. As a side-line result, we also provide a novel form of FD distribution that manifestly separates the zero and finite temperature contributions so that the finite temperature correction can be studied on its own in more complex environments than considered in our current method-oriented presentation.
Our own interest and motivations to explore FD systems originated in cosmological computations involving both high and low temperature physics, such as cosmic neutrino background [25, 26] and cosmic magnetism [20]. In the context of magnetism an added difficulty is the Landau discrete sum over Landau orbitals. In the regime of neutrino cosmology, considering their likely mass, the present-day free-streaming neutrinos exhibit characteristics of being cold and nonrelativistic fermions, with a small chemical potential reflecting the baryon asymmetry in the universe. During the Universe expansion neutrino abundance per quantum correlation de Broglie volume decreases when the ambient temperature scale and neutrino mass scale coincide. Therefore the free-streaming neutrinos might fall into the novel regime described here for quite a long period of Universe expansion. We note that a related technique to ours, for the case of number density applied to study a degenerate gas in a freely expanding universe, can be found in [27] in the limit in which free-streaming phenomena do not matter.
Another natural area of physical relevance involves ‘warm’ nuclear matter (temperature well below transformation to quark-gluon plasma). An estimation of the number of indistinguishable nucleons in the de Broglie volume at a normal nuclear matter density produces a value ≃5, consistent with the general understanding of atomic nuclei as being quantum objects, rather than assembly of classical interacting balls. This means that neutron stars in their formation history, and during neutron star mergers evolve through the warm nuclear matter non-Sommerfeld and non-Boltzmann regime we characterize. For in depth discussion of the current status of this field see for example Refs. [28–30] and references therein. We further note recent interest in the understanding of properties of warm nuclear matter [31], which is at present limited to the Sommerfeld expansion. Our newly identified regime and associated asymptotic expansion present an opportunity to further refine this work.
The newly identified regime is transient, however, it could influence the dynamical evolution or radial dependence of dense nuclear systems which impact the physical outcome concerning the formation of neutron stars, their size and mass, or the detailed questions concerning dynamical observations involving neutron star mergers [32–35]. Similarly, the results we present could influence the interpretation of laboratory experiments involving the study of dense warm nuclear matter in relativistic heavy ion collisions, see Ref. [36], and references therein.
Our method can be applied to general observables in these diverse cases. We intend to use the techniques developed here in future application-oriented work. This work could address applications in cosmology, in the study of the case of free-streaming neutrinos, which is of profound relevance for better understanding the evolution of the Universe. Furthermore our techniques are applicable to dense Fermi systems in e.g. stellar environments, and to dense Bose systems.
Domains of interest and their characteristic particle density
In this work we derive and contrast expansions of FD matter in the following regimes which, to the best of the authors’ knowledge, contain several novel insights. In Figure 1 we show the domains of interest in the (T,Δμ)-plane (in units of m). The darker background shows regions with relative error less than 10%, lighter background shows error less 20%, when computing ⟨G⟩T for the illustrative choice G(p)=1. We emphasize that other choices of G will lead to different domains; see Section 1.3 for further examples.
Figure 1. Domains of interest spanned in the (T/m,Δμ/m≡(μ−m)/m)-plane, top for three dimensional and bottom frame for two dimensional systems: Sommerfeld expansion (blue), Boltzmann approximation (lined purple), and the new Theorem 5 domain (dotted green). Solid black contours show the number of particles per de Broglie volume Eq. (1). See text for a more detailed discussion.
Errors here, and elsewhere in this work, were obtained by comparing the approximations with the result obtained from standard numerical quadrature; given the form of the FD integrand, such methods can be reliably used as long as the values of μ and T are not too extreme. We also show using solid black contours lines in Figure 1 the number of particles per de Broglie volume. From these one can recognize
The traditional Sommerfeld asymptotic expansion applies when statistical parameters allow for a quantum degenerate density of Fermi particles, with about 5 or more particles found in the corresponding de Broglie volume. As will be considered here the validity of this regime can be extended to the dense relativistic quantum domain.
The domain of validity of the newly characterized dilute quantum region (Theorem 5) is found when the combination of statistical parameters allows a moderate particle density of a few (2 to 3) particles per de Broglie volume, where quantum effects are expected to be nontrivial but different from properties of dense FD matter.
The Boltzmann limit applies to systems with about 1 or fewer particles per de Broglie volume.
Particle abundance described and shown in Figure 1 does not account for spin and other degeneracies. Finally we note that Figure 1 focuses on the non-relativistic domain; for a similar figure that shows the relativistic regime see Figure 8.
Figure 8. Comparison of two and three dimensional cases of domains of validity in the (T/m,Δμ/m)-surface (Δμ≡μ−m), including relativistic parameter range, of the Sommerfeld expansion (solid blue) and the expansion from Theorem 5 (dotted green), where validity is defined as the computation of ⟨G⟩T having relative error less that 10% for the choice G(p)=1.
blackSpecific examples independent of chemical potential
When analyzing physical properties of a quantum gas, it is often of interest to express all results in terms of temperature, and particle number density rather than the grand canonical chemical potential. This approach aligns more closely with experimental conditions, as temperature and particle number density are directly measurable quantities, and particle density is often created by physical circumstance. In this section, we present results that are independent of the chemical potential providing a more practical understanding of the Fermi gas in different physical regimes.
In Figure 2 we present the domains of validity as defined by the number density on the plane parameterized by T/m and particle number per de Broglie volume for D=2 (bottom frame) and D=3 (top frame) dimensional systems. The black solid contours indicate the kinetic energy per particle in units of T. For D=3 we see that the traditional Sommerfeld asymptotic expansion is applicable for T≪m, when particle number per de Broglie volume is greater than 4.5, which is for example the case for normal and gravitationally compressed nuclear matter.
Figure 2. Study of energy per particle regimes of an ideal Fermi gas in the plane spanned by the number of particles in de Broglie volume, and temperature in units of m, T/m; top frame for the three dimensional D=3, and bottom frame for the two dimensional D=2 systems: Sommerfeld expansion (blue), Boltzmann approximation (lined purple), and the new Theorem 5 domain (dotted green). Solid black contours show the kinetic energy per particle in units of T. The darker color applies to domains with less than 10% error, while the lighter color allows up to 20% error.
For the case T≫m the domain of validity of the Sommerfeld regime decreases to be as low as 1.5 particles per de Broglie volume. The Boltzmann limit is valid for all temperatures when the number of particles per de Broglie volume is near to, or below one. The domain of validity of the newly characterized semi-dilute quantum region (Theorem 5) occurs at low temperature when the number of distinguishable quantum particles per de Broglie volume is between 2 and 3 approximately. We see furthermore that in this study of energy content of a Fermi gas there remains at a low temperature a significant (white domain) regime in which numerical quadrature needs to be employed to obtain reasonable precision. On the other hand we see domains where two different asymptotic expansions overlap.
In Figure 3 we illustrate the regimes of validity for the energy density to pressure ratio ϵ/P, again in the plane spanned by the number of particles per de Broglie volume, and T/m, for D=2,3 dimensional systems, respectively (compare Figure 2). At high temperature the equation of state approaches ρ=DP and this is properly captured by the Boltzmann approximation, which is therefore valid at high temperature even for systems with a larger number of particles in the de Broglie volume.
Figure 3. Study of energy density to pressure ratio ε/P regimes of an ideal Fermi gas in the plane spanned by the number of particles in de Broglie volume and temperature in units of m, T/m; top frame for the three dimensional D=3, and bottom frame for the two dimensional D=2 systems: Sommerfeld expansion (blue), Boltzmann approximation (lined purple), and the new Theorem 5 domain (dotted green). Solid black contours show the energy to pressure ratio. The darker color applies to domains with less than 5% error, while the lighter color allows up to 10% error.
For low temperature we see in Figure 3 that the Sommerfeld regime in D=3 case applies when there are approximately 3.5 or more particle per de Broglie volume while the new Theorem 5 domain applies in the entire gap with Boltzmann regime when the energy density to pressure ratio ε/P is larger than around 10.0, with 2∼3 particles per de Broglie volume.
Novel insights
Our novel contributions include consideration of:
Nonrelativistic degenerate Fermi gas region (Theorem 2): Application of the traditional Sommerfeld expansion in this regime requires care, as the coefficients in the expansion can become singular when μ→m. In § 3.3 we derive a formula that resolves those singularities so as to avoid such numerical difficulties and obtain an expansion up to a specified order. In this derivation we utilize a novel decomposition of the FD distribution into finite and zero temperature components, see § 3.1, which we find convenient for guiding this and other derivations.
Degenerate relativistic Fermi gas Sommerfeld expansion (Theorem 3): We prove that the traditional Sommerfeld expansion, which applies to the regime T≪m with Δμ not small, also applies to the high-temperature regime μ≫T≫m under appropriate assumptions.
Dilute quantum region (DQR) (Theorem 5): We demonstrate that the Sommerfeld expansion can fail completely when ∣μ−m∣≪T≪m; see the example in Figure 7 below. Accurate results in this regime require the use of the new expansion that we derive in § 4. In Figure 1 we preview the new expansion stated in Theorem 5, comparing its domain of validity to that of the Sommerfeld expansion in dimensions D=2,3. In particular, note that the new expansion derived here covers a distinct new region in the (T/m,Δμ/m)-plane where just a few particles are present in a de Broglie volume.
Figure 7. Left: Comparison of asymptotic expansions of ⟨G⟩T, G(p)=1, via the Sommerfeld expansion Eq. (26) (red dot-dashed line) and the new expansion Eq. (81) (green dashed line) where μ=m(1+T3/2). Under this relationship between μ and T, Theorem 3 and Theorem 5 respectively prove that Eq. (53) is valid at high temperature and Eq. (81) is valid at low temperature; we see this reflected in the figure. Right: The relative error of these asymptotic expansions, also compared with that of the Boltzmann approximation.
The energy density to pressure ratio ε/P is an important physical quantity; the results shown in Figure 3 demonstrate that our new result Theorem 3 provides nontrivial insights into the working of equations of state in a regime neither covered by the Sommerfeld nor by the Boltzmann approximations.
Sommerfeld-Type Asymptotic Characterization of Thermal Properties
A new FD distribution decomposition
In this section we will explore the Sommerfeld-type asymptotic expansions of thermal averages in various physics-relevant chemical potential and temperature regimes. As a convenient tool, we will employ the following novel way to write FD distribution in the zero temperature limit T→0.
At T=0 the FD distribution reduces to a step function where a state Ei is either filled or empty. For given chemical potential μ(T), we have
At T=0 the energy of the last filled state is the Fermi energy EF=μ(T→0), while at finite temperature one speaks of chemical potential instead.
For the calculations in this section we find it useful to employ a novel reformulation which allows exact separation of the finite temperature behavior from the singular zero temperature limit, replacing the smooth FD distribution by the sum of three singular functions,
ex+11=Θ(−x)+21e−∣x∣[sgn(x)+tanh(x/2)],(4)
the first term (Heaviside function) being the T=0 limit and the other two describing the finite temperature residual.
In this form the zero temperature part is explicitly separated from the finite temperature contributions.
Lemma 1. The Fermi-Dirac distribution can be decomposed into three components:
where Θ(x) is the Heaviside step function and sgn(x) is the sign function and the equality is in the sense of distributions. In particular the values at x=0 are convention-dependent and do not affect our results.
Remark 1. To show the equivalency between the two distinct forms of the FD, we look at the three relevant regions of x>0, the origin x=0, and x<0. For x>0, Eq. (5) evaluates as
fFD(x>0)=0+21e−x[1+tanh(x/2)]=(ex+1)−1.(8)
This can be seen by using the hyperbolic formula tanh(x/2)=(ex−1)/(ex+1). In a similar manner, the x<0 region evaluates as
fFD(x<0)=1+21ex[−1+tanh(x/2)]=(ex+1)−1.(9)
As we are considering the expressions in the sense of distributions, this completes the proof. The values at x=0 are convention-dependent and irrelevant for our purposes, as the point x=0 is a set of measure zero with respect to integration.
In Figure 4 the solid line shows the FD distribution as a function of energy for chemical potential μ=0.461 MeV, and electron mass m=0.511 MeV, in the top frame for temperature T=0.02 MeV, and in bottom frame for T=0.2 MeV. Purple dashed line is the corresponding T=0 limit. As the temperature increases, the distribution becomes broader as a wide range of states become thermally populated. As $T decreases the fT=0 (blue dotted) and fT=0 (red dash-dotted) components, see Eq. (6), which characterize finite T contributions become progressively more singular.
References
[1]Bludman, S.A., Van Riper, K.A.: Equation of state of an ideal Fermi gas. Astrophysical Journal 212, 859–872 (1977)DOI
[3]Ferrer, E.J., Hackebill, A.: Thermodynamics of Neutrons in a Magnetic Field and its Implications for Neutron Stars. Phys. Rev. C 99(6), 065803 (2019) https://doi.org/10.1103/PhysRevC.99.065803 arXiv:1903.08224 [nucl-th]
[7]Garoni, T.M., Frankel, N.E., Glasser, M.L.: Complete asymptotic expansions of the Fermi–Dirac integrals Fₚ(η) = 1/γ(p + 1) ∫₀∞εᵖ/(1 + eᵋ⁻η)dε. Journal of Mathematical Physics 42(4), 1860–1868 (2001) https://doi.org/10.1063/1.1350634 https://pubs.aip.org/aip/jmp/article-pdf/42/4/1860/1931445/1860_1-online.pdf
[8]Lourenco, O., Dutra, M., Delfino, A., Sa Martins, J.S.: Relativistic Sommerfeld low temperature expansion. Int. J. Mod. Phys. D 16, 285–289 (2007) https://doi.org/10.1142/S021827180701002X
[9]Wasserman, A., Buckholtz, T.J., DeWitt, H.E.: Evaluation of Some Fermi-Dirac Integrals. Journal of Mathematical Physics 11(2), 477–482 (1970) https://doi.org/10.1063/1.1665160 https://pubs.aip.org/aip/jmp/article-pdf/11/2/477/19230647/477_1-online.pdf
[10]Fukushima, T.: Analytical computation of generalized Fermi–Dirac integrals by truncated Sommerfeld expansions. Applied Mathematics and Computation 234, 417–433 (2014) https://doi.org/10.1016/j.amc.2014.02.053
[11]Gil, A., Segura, J., Temme, N.M.: Complete asymptotic expansions for the relativistic Fermi-Dirac integral. Applied Mathematics and Computation 412, 126618 (2022) https://doi.org/10.1016/j.amc.2021.126618
[12]Gil, A., Odrzywołek, A., Segura, J., Temme, N.M.: Evaluation of the generalized Fermi-Dirac integral and its derivatives for moderate/large values of the parameters. Computer Physics Communications 283, 108563 (2023) https://doi.org/10.1016/j.cpc.2022.108563
[13]Kim, R., Wang, X., Lundstrom, M.: Notes on Fermi-Dirac integrals. arXiv preprint arXiv:0811.0116 (2008)arxiv.org/abs/0811.0116
[14]Buterakos, D., Vu, D., Yu, J., Das Sarma, S.: Presence versus absence of two-dimensional Fermi surface anomalies. Phys. Rev. B 103, 205154 (2021) https://doi.org/10.1103/PhysRevB.103.205154
[16]Ferrer, E.J., Hackebill, A.: The Importance of the Pressure Anisotropy Induced by Strong Magnetic Fields on Neutron Star Physics. J. Phys. Conf. Ser. 2536(1), 012007 (2023) https://doi.org/10.1088/1742-6596/2536/1/012007
[17]Rafelski, J., Yang, C.T.: The muon abundance in the primordial Universe. Acta Phys. Polon. B 52, 277 (2021) https://doi.org/10.5506/APhysPolB.52.277 arXiv:2103.07812 [hep-ph]
[18]Rafelski, J., Birrell, J., Steinmetz, A., Yang, C.T.: A Short Survey of Matter-Antimatter Evolution in the Primordial Universe. Universe 9(7), 309 (2023) https://doi.org/10.3390/universe9070309 arXiv:2305.09055 [hep-th]