WorksPhysics and Astronomy
Quarkyonic stars with isospin-flavor asymmetry
Abstract
We suggest an extension to isospin asymmetric matter of the quarkyonic model from McLerran and Reddy. This extension allows us to construct the -equilibrium between quarks, nucleons and leptons. The concept of the quarkyonic matter originates from the large number of color limit for which nucleons are the correct degrees of freedom near the Fermi surface – reflecting the confining forces – while deep inside the Fermi sea quarks naturally appear. In isospin asymmetric matter, we suggest that this new concept can be implemented within a global isoscalar relation between the shell gaps differentiating the nucleon and the quark sectors. In addition, we impose the conservation of the isospin-flavor asymmetry in the nucleon and the quark phases. Within this model, several quarkyonic stars are constructed on top of the SLy4 model for the nucleon sector, producing a bump in the sound speed, which implies that quarkyonic stars are systematically bigger and have a larger maximum mass than the associated neutron stars. They also predict lower proton fraction at -equilibrium, which potentially quenches fast cooling in massive compact stars.
I Quarkyonic model in asymmetric matter
The concept of quarkyonic matter has emerged in the large number of color, , limit of QCD [20]. In this limit and when the nucleon density is much larger than the QCD scale, , the confining potential of QCD is dominant even though the nucleonic Fermi momentum is large, . The concept of quarkyonic matter has been introduced in order to resolve this apparent paradox: the ground state of dense matter is composed of dressed quarks (with mass ) that are freely moving deep inside the Fermi sea, and of a shell of baryons generated by the strong confining force, which lies close to the Fermi level [20]. Baryons occupies a momentum shell whose width is taken to be , while deepest states are occupied by quarks.
Even if matter is composed of nucleons and quarks, low energy excitations around the Fermi level involve only quasi-particles of nucleonic type. The excitation of quarks require the expense of a momentum of the order of the shell width . In this prospect, nucleons and quarks are not independent particles, but there are realization of matter excitations in well separated energy regions. This picture is clearly illustrated on the left panel of Figure 1 for SM.
Figure 1. Schematic views of the different Fermi seas at stake in quarkyonic matter for SM (left) and AM (right). Quarks occupy deep states inside their Fermi spheres while nucleons occupy the shells close to their Fermi levels.
In SM, this picture is simple since it is sufficient to fix a condition between the nucleon and quark single particle energies and to separate the deep quark states from the nucleons ones located around the Fermi level. This condition imposes that the last quark occupied state coincides with the first nucleon one [1],
where and are isoscalar nucleon and quark Fermi momenta. The factor in Eq. (1) stands for the number of quarks forming each nucleon in its ground state. With this condition, the nucleon and quark degrees of freedom are reduced to only one free variable, that we fix to be , the nucleon Fermi momentum for convenience. Note that in Eq. (1) the nucleon single particle energy is the non-interacting (NI) one, . This choice is performed in order to avoid unnecessary complication of the model. We further assume that chiral symmetry remains broken to set the quark mass () with , as in the constituent quark model.
Finally a prescription for the thickness of Fermi layer where nucleons reside has to be taken. We adopt the same relation from the original paper [1],
with MeV and . defines the energy scale below which nucleons reside.
The concept of quarkyonic matter however leads to a fundamentally new way to represent the nucleon and quark densities and and their associated Fermi seas. As in a superfluid, there is only one chemical potential that enters into the thermodynamic equilibrium, and which is associated to the last nucleon occupied state where is the total baryon energy in the ground state and stands for the number of baryons which is a partition of nucleons and quarks states. By adding or removing a baryon, the entire Fermi sea is reorganized leading to a new partition between quark and nucleon states, with the condition on the number densities . This new concept is first applied to symmetric and neutron matter [1] and we are now suggesting an extension for AM.
Global isoscalar relation between nucleon and quark Fermi seas
By breaking the isospin symmetry in AM, the nucleon and quark states are replaced by four other states: neutrons, protons, as well as and quarks are the natural components in AM, where they are represented by their associated densities, , , and . The four Fermi seas are schematically represented in the right panel of Figure 1. The baryon charge density controlling SM is completed in AM by the isospin asymmetry . Since there are only two charges and four particles, the concept of quarkyonic matter in asymmetric matter requires two additional relations.
At variance with SM where nucleons and quark Fermi seas could be defined by imposing a relation between their single particle energies, see Eq. (1), nothing similar can be done in AM between the four particles. The concept of quarkyonic matter suggests however that the relation between nucleons and quarks in SM, see Eq. (1), remains globally valid in AM. The nucleon and quark Fermi momenta in AM, and , are represented in Figure 1 by the red dashed circles.
Expanding the single particle energies in Eq. (1), e.g. , one gets the following relation between the nucleon and quark isoscalar Fermi momenta,
Note that in Eq. (3) we assumed that the nucleon shell gap is also an isoscalar quantity. Eq. (3) allows quark to appear as soon as independently of the isospin asymmetry. This is supported, as we explain previously, by the idea that exciting quarks in AM requires an energy of the order of irrespectively of the partition of matter between neutrons and protons. Without an actual solution of QCD, this assumption is the simplest one which can be done.
The limit of NM was also explored in the original paper by McLerran and Reddy [1]. The prescription taken there imposes that one has to choose which among and quarks are connected to the neutron states. In addition, the value for MeV considered in SM was changed to 380 MeV in NM. In our model, is taken constant, but the change of the Fermi momentum between SM and NM, see Eqs. (3) and (17), induces an effective modification of between SM and NM with the ratio , which is exactly the same ratio considered by McLerran and Reddy [1]. In practice, the two approaches lead to similar results in NM. The isoscalar relation (3) presents however the advantage to describe AM and to recover the limit of SM, where the concept of the quarkyonic matter is simple.
In summary, we remark that the isoscalar Fermi momentum controls both the isoscalar quark Fermi momentum , see Eq. (3), as well as the nucleon gap (3) from the prescription (2). The isoscalar nucleon and quark densities can be safely determined as,
and the quark density,
The total baryon density is built upon the nucleon and quark contributions as,
While the isoscalar densities and can now be calculated from , the connection to the densities of the four particles , , and are yet unknown. They are related to the isoscalar densities from the following relation:
In the following, we suggest that the densities , , and can be obtained in AM by imposing the isospin/flavor asymmetry in the nucleon and quark phases.
Isospin/flavor asymmetry
We will now determine the particle densities. In the quarkyonic model, there is a partition between nucleons and quarks which changes as the density evolves. It reveals the dynamical process converting quarks into nucleons or breaking the nucleons into their constituents, as it shall be in compound systems. A simple way to translate this symmetry into the nucleon and quark phases is to impose that the two phases conserve the isospin/flavor asymmetry. Since and , we obtain the following relations for the quark density in the nucleon phase, and , which leads to the following simple connection between the isospin asymmetry parameter in the nucleon phase and the flavor asymmetry parameter in the quark phase:
From SM to NM, goes from 0 to 1, while goes from 0 to . The dynamics of the phase transition thus imposes that the and flavor ratio reflects the isospin asymmetry of the nucleon phase.
As a side note, we are aware that the isospin/flavor asymmetry relation (9) can possibly be violated by the two phases if the energy minimization is injected. Such a refinement of the quarkyonic model is indeed very interesting but it complexifies the quarkyonic approach, which nice feature remains in its simplicity. Further extensions of the present model will be explored in the future, especially to analyze their role in the predictions presented here.
Knowing and which fixes and one can deduce all particle densities as
The and quark Fermi momenta are simply related to their densities as
since and quarks occupy their Fermi sphere, see Figure 1.
The neutron and proton Fermi layers can be calculated from the difference of two Fermi spheres with different radii defined as,
where and are the lower bound of the nucleon shell, see Figure 1. Injecting Eq. (4) into () and identifying with Eqs. (16), we obtain
as well as
Note that and can be re-expressed as , for . At low densities, in the absence of quarks , so neutrons and protons occupy entirely their Fermi spheres with radii given by Eqs. (17). It is interesting to note that Eqs. (17) are identical in the presence or the absence of quarks.
Now that the Fermi spheres and the Fermi shells for , , and particles are well defined, all thermodynamical quantities can be determined, e.g. the energy of the ground state, the pressure, the chemical potentials, or the sound speed.
Energy density and derivatives
The energy density of quarkyonic matter is given by
where the nucleon and quark terms are given by:
with the nuclear residual interaction given by the meta-model [19],
where , and . The coefficients and are related to the empirical parameters, e.g. MeV, fm and MeV in nuclear matter, considering the relativistic extension of the meta-model (for the kinetic term only) [33]. The present calculation is based on the nucleon Skyrme interaction SLy4, on which the meta-model is adjusted. We consider the parameters given in table I.
The binding energy density is defined as
and the binding energy per baryon number is
The other quantities as the chemical potentials, pressure and sound velocities are computed using the usual definitions, see Ref. [19, 34] for more details.
Results
In this section we compare a pure nucleon model for matter properties against quarkyonic models constructed on top of the same nucleon model. The choice of SLy4 is influenced by the fact that this pure nucleon model reproduces most of the recent observational data, such as the tidal deformability from GW170817 or the NICER x-ray observation of PSR J0030+0451. We then explore the influence of the quarkyonic model parameters and .
The neutron chemical potential , the energy per particle , the baryon pressure and baryon sound speed are shown in Figure 2 for SM (solid lines, ), AM (dashed lines, ) and NM (dotted lines, ). The quarkyonic model parameters are fixed to be MeV and . The predictions for the quarkyonic phase (green lines) are confronted to the ones for the pure nucleon phase (magenta lines). The model predictions are stopped when causality is violated. The sound velocity in quarkyonic matter has a peak at around fm, as shown in Ref. [1] for SM and NM, and confirmed here for AM. The position of the peak is almost independent of the isospin asymmetry, but the peak is a bit more pronounced in NM compared to SM. For the chosen parameters, the sound speed predicted by the quarkyonic model at high density reaches a value close to the conformal limit, i.e. 1/3. The bump in the sound speed density profile present in quarkyonic matter impacts the pressure, the chemical potential and the binding energy. These thermodynamical quantities are strongly increased for densities where the sound speed is maximal, and they are softer at higher densities. The softening is such that the pressure of quarkyonic matter crosses the pure nucleon one at high density, see Figure 2. The softening of the EoS is also predicted by usual construction of first order phase transitions from nucleon to quark matter, and the interesting feature of the quarkyonic model is the stiffening of the EoS at low densities, where it really matters for NS, before the softening at high density. The region of importance for NS properties coincides mostly with the densities where the pressure is stiff. This is the reason why this model is of particular interest for the phenomenology of compact stars.
Figure 2. Baryon chemical potential , energy per particle , pressure and sound speed in symmetric (SM, ), asymmetric (AM, ), and neutron matter (NM, ) for MeV and .
The increase of the chemical potential in the crossover region can also be explained from the behavior of the nucleon Fermi momentum, which can be traced down from Eq. (4) and re-written as,
showing that the quark contribution to the nucleon Fermi momentum is strongly enhanced by the factor . As the density increases however, the quark component of matter becomes more and more dominant and the softening actually occurs.
The conclusion of this first set of results is that the generic features of the quarkyonic model predicted by McLerran and Reddy [1] are preserved in our extension of the quarkyonic model for AM, and we can predict similar features in NM with the same parameters as the one in SM. Our results are also in qualitative agreement with
| Model | |||||||||||||
| MeV | MeV | fm | MeV | MeV | MeV | MeV | MeV | MeV | MeV | ||||
caption: Parameters of the SLy4 meta-model used in the non-relativistic (NR) case for the description of nuclear matter and in the relativistic case (RL), where only relativistic kinematic is considered, for the quarkyonic matter.
:::
the ones found by Zhao and Lattimer [27] as well as Jeong McLerran and Sen [29] where different nuclear potential were used.
II Quarkyonic model at -equilibrium
We are now constructing the beta-equilibrium solutions, which describe the ground state of dense matter existing in the core of compact stars. In cold catalyzed NS, matter components are determined from the -equilibrium equations,
and charge neutrality,
At fixed total density, these three equations allows the determination of three variables: the isospin asymmetry and the electron and muon densities, and .
Note that the charge neutrality condition (29) in NM becomes , which coincides with the relation between and quark Fermi momenta, , imposed in NM in [1].
The particle fractions in dense matter are shown in Figure 3 for the SLy4 nucleon model (magenta) and the quarkyonic model taking MeV and (green). On the left panel of Figure 3 are shown only the baryon contributions, , , and , while on the right panel, the contributions of the nucleons, quarks, electrons and muons are represented. In the crossover region, neutron and proton densities are reduced compare to the original nucleon model while the amount of quarks increases. In particular, we observe that the fraction of protons is reduced in the quarkyonic model such that it remains below the dURCA threshold ( in the presence of muons [28]).
Figure 3. Particle fractions for the SLy4 nucleon model and the quarkyonic model taking MeV and .
In order to investigate the role of the parameters of the quarkyonic model on the proton fraction, we show in Figure 4 a comparison for different choices for the parameters : (250, 0.3), (270, 0.3) and (250, 0.2). Increasing from 250 MeV to 270 MeV induces an increase of the proton fraction at high density, which passes the dURCA threshold, but at higher density. In terms of energy density , right panel in Figure 4, the dURCA threshold is pushed even higher, since the quarkyonic model predicts larger energy densities than the pure nucleon one.
Figure 4. Comparison of the proton fraction in the pure nucleon model with the prediction of the quarkyonic model for various sets of the parameters : (250, 0.3), (270, 0.3) and (250, 0.2). Left: function of the particle density in , right: function of the energy density in MeV .
Finally, the thermodynamical properties of quarkyonic matter at -equilibrium are shown in Figure 5: the neutron chemical potential , the total energy per particle , the equation of state and the total sound speed. The effect of varying the parameters of the quarkyonic matter is also shown. The largest impact is observed when the parameter in increased from
Figure 5. Neutron chemical potential , total energy per particle , total pressure and total sound speed at -equilibrium for the parameters , same choice as in Figure 4.
250 MeV to 270 MeV. Increasing has the effect of pushing the onset of the first quarks at higher density, as it can also be observed in Figure 5. As a consequence, increasing makes quarkyonic matter more and more repulsive, except at low density where the larger the softer the EoS, as discussed previously. The effect of changing is smaller. It was tuned in the original paper by McLerran and Reddy [1] to the conformal limit for the sound speed.
Quarkyonic stars
The structure of non-rotating neutron stars is provided by the solution of the spherical hydrostatic equations in general relativity, also named the Tolmann-Oppenheimer-Volkof equations [35],
where is the gravitational constant, the speed of light, the total pressure, the enclosed mass, is the total mass-energy density and the gravitational field. and have contributions from both the baryons and the leptons .
The four variables are obtained from the solution of the three TOV equations (30) and the EoS for the quarkyonic matter. In the present calculation, a crust EoS is smoothly connected to the core EoS as described in Ref. [34]. The tidal deformability induced by an external quadrupole field is expressed in terms of the Love number as , where the compactness is , and is calculated from the pulsation equation at the surface of NS [11, 12],
where is the value of the function at radius , , and is the solution of the following differential equation,
with the boundary condition and the functions
and defined as,
The NS moment of inertia is obtained from the slow rotation approximation [36, 37] as
where is the local spin frequency, which represents the general relativistic correction to the asymptotic angular momentum and is defined as .
As usual, for a given EoS the family of solutions is parameterized by the central density or pressure or enthalpy. The EoS are characterized by their evolution in the mass-radius diagram, where both masses and radii of compact stars could in principle be measured, as discussed in our introduction, see also Ref. [38].
We show in Figure 6 the predictions for the mass , radius , tidal deformability , central density , binding energy and moment of inertia associated to various quarkyonic EoS with : ,
Figure 6. Neutron star properties (mass , radius , tidal deformability , central density , binding energy , where is the baryon mass, and the moment of inertia ) for quarkyonic matter with and fixed as in Figure 4 (green curves) and nucleon matter (solid magenta curve).
and (green lines). These predictions are confronted to the ones for a nucleon EoS (solid magenta line).
The impact of quarkyonic matter on the mass-radius relation is huge, as already noticed in Refs. [1, 29]. While the maximum mass for SLy4 is reached at , the quarkyonic stars almost reach . These is also a large impact on the radius: the radius of the pure nucleon model is about km, while it is pushed up to 13-14 km in the quarkyonic model. Quarkyonic stars can therefore be much more massive than pure nucleon ones, and are also bigger in size. Quarkyonic stars have also different tidal deformabilities compared to the pure nucleon case. For the same mass, the quarkyonic stars have larger , and at fixed , they have larger radii. At a fixed central density (in ), quarkyonic stars are much more massive than the pure nucleon model we considered. This is an effect of the repulsion observed for the pressure in Figure 5. For the same mass, quarkyonic stars have a slightly lower compared to the associated nucleonic star, they however have a larger moment of inertia.
One can estimate the influence of the parameters and . As observed in previous figures, the parameter is almost not influential at all, while is much more critical. By increasing , the realization of quarkyonic matter is pushed up and the EoS get closer to the pure nucleon case. The quarkyonic star therefore also get a bit closer to the neutron star as increases. The journey in the mass-radius diagram is therefore very much controlled by the parameter . By decreasing the quarkyonic star gets bigger and bigger compared to the NS, as well as it gets more and more massive. In the future, the observation of several points in mass and radius, e.g. from NICER observations, will thus be very useful for the determination of the parameter .
Finally we discuss the dURCA threshold, represented in the curves by the solid circle. Only the pure nucleon case reaches the dURCA threshold, and it happens close to . Even the quarkyonic model with , which satisfies the dURCA condition at high density, see Figure 4, reaches the unstable branch before it gets to the dURCA density. It is thus more difficult to reach the dURCA condition with quarkyonic stars. The same conclusion was also obtained by Zhao and Lattimer [27] with their version of quarkyonic matter.
We show in Figure 7 the NS compactness defined as , where is expressed in km, function of the mass (left panel) and of the radius (right panel). The compactness of the isolated NS RX J0720.4-3135 has been extracted from observations and estimated to be [39]. Reporting this value in the left panel of Figure 7 we observe that the nucleonic EoS SLy4 suggests that the mass of the pulsar is , compatible with observed masses but close to their lower limit [40], while the quarkyonic model suggest higher masses, up to , which is compatible with the canonical NS mass. Note that a Bayesian exploration of nucleonic models has predicted a centroid of about [34].
Figure 7. Compactness as function of the mass (left) and radius (right) for various sets of the parameters and fixed to be as in Figure 4.
This value is slightly larger than the one suggested by SLy4 EoS, but is still lower that the canonical mass. On the right panel of Figure 7 the radii associated to the observed compactness are also reported. While the SLy4 EoS favors 11.9 km, the quarkyonic stars point towards larger radii, up to 13.6 km in the upper case. So we conclude that for a fixed value of the compactness, Figure 7 shows that quarkyonic stars have larger masses and radii than NS.
The gravitational redshift associated to the radial emission of photons from the surface, which is detected by a distant observer is defined as . In Figure 8 we show versus the stellar mass (left panel) and versus the stellar radius (right panel). The emission line feature of the gamma ray burst GB790305, assumed to originate from the annihilation (observed peak at 430 keV, line width keV), and assuming thermal nature of line broadening and taking due account of the thermal blueshift, leads to the observational constraint [41, 42], which is reported in Figure 8. We also deduce from this observational data that SLy4 would favor typical masses of the order of , low masses, while the quarkyonic star built on the same nucleonic star would point towards , closer to the canonical NS mass. The radius of extracted from SLy4 would be 11.9 km, while quarkyonic star would point towards larger radii, up to about 13.8 km.
Figure 8. Gravitational redshift as function of the mass (left) and radius (right) for various sets of the parameters and fixed to be as in Fig. 4.
Finally, we construct the quarkyonic model on top of a nucleonic model which does not reach the observation constraint of about . To do so, we reduce the value of from the SLy4 nucleonic model by 300 MeV, see table I. The nucleonic model is shown in Figure 9 under the label SLy4-soft (solid magenta line) while the quarkyonic models are shown for the same three cases as before. With this example, we show that the crossover transition to quark matter, as described by the quarkyonic approach, can bring enough repulsion to reach large maximum mass, even if the model for the nucleonic part cannot satisfy the observed requirement that the maximal mass of NS should be above about .
Figure 9. Mass-radius relation for SLy4-soft nuclear interaction and for the quarkyonic model with various sets of the parameters and fixed to be as in Fig. 4.
Conclusions
We have proposed an extension of the original quarkyonic model from Ref. [1] to AM, where the original quarkyonic model for SM is recovered as a limit. Our extension assumes (i) that the description of the quark Fermi sea and nucleon shell is globally isoscalar and (ii) that the isospin-flavor asymmetry in the quark and nucleon phases is fixed. These assumptions root into the concept of the quarkyonic model where nucleons result from the strong confining force, which strength is large close to the Fermi level. The assumption (i) allows us to smoothly connect to the quarkyonic model in SM, and suggests a description of NM quite comparable – at least qualitatively – to the original one suggested by McLerran and Reddy [1]. By fixing the isospin/flavor asymmetry in the nucleon and quark phases, assumption (ii), the properties of isospin asymmetric quarkyonic matter can be entirely determined from the nucleon Fermi momentum and the isospin asymmetry .
NS matter at -equilibrium is then been calculated and it provides qualitatively similar results as in the original model [1]. It is also in agreement with other extensions in asymmetric matter [27, 29], while being based on different assumptions. In our model, quarkyonic stars are larger and heavier than the associated NS, and the parameter is playing a dominant role in changing the radius and the mass of quarkyonic stars. This result is valid even if the nucleonic component is soft, e.g. too soft to reach . The proton fraction at -equilibrium is found to be reduced in the quarkyonic matter, compared to the related pure nucleonic phase, which potentially quench fast cooling – based on dUrcA process – in massive compact stars. The confrontation to a set of masses and radii, potentially obtained in future observations like NICER or gravitational wave detections of in-spiral binary NS, will potentially constrain , as well as cooling scenarios.
In the future, we aim at incorporating the quarkyonic model in systematical comparisons to observational data in order to better understand the properties of dense matter. Extension of the present model to finite temperature is also on our map for the near future, as well as improving the isospin/flavor asymmetry relation. Also, adding chiral symmetry consideration in the quarkyonic model, taking into account the two most striking features of QCD, will certainly be an interesting extension to study.
Acknowledgments
We thank K. S. Jeong, L. McLerran, S. Reddy, and A. Schmitt for fruitful discussions, as well as R. Stiele, R. Somasundaram, and A. Pfaff for careful checks of our formalism. J.M., H.H. and G.C. are supported by the CNRS/IN2P3 NewMAC project, and benefit from PHAROS COST Action MP16214. All authors are grateful to the LABEX Lyon Institute of Origins (ANR-10-LABX-0066) of the Université de Lyon for its financial support within the program Investissements d’Avenir (ANR-11-IDEX-0007) of the French government operated by the National Research Agency (ANR).
References
References
- [1]L. McLerran and S. Reddy, Phys. Rev. Lett. 122, 122701 (2019).
- [11]T. Hinderer, Astrophys. J. 677, 1216 (2008).
- [12]É.É. Flanagan and T. Hinderer, Phys. Rev. D 77, 021502 (2008).
- [19]J. Margueron, R. Casali Hoffmann, F. Gulminelli, Phys. Rev. C 97, 025806 (2018).
- [20]L. McLerran and R.D. Pisarski, Nucl. Phys. A 796, 83 (2007).
- [27]T. Zhao and J. L. Lattimer, Phys. Rev. D 102, 023021 (2020).
- [28]J. L. Lattimer, C. Pethick, M. Prakash, and P. Haensel, Phys. Rev. Lett. 66, 2701 (1991).
- [29]K. S. Jeong, L. McLerran, and S. Sen, Phys. Rev. C 101, 035201 (2020).
- [33]J. Margueron, unpublished.
- [34]J. Margueron, R. Casali Hoffmann, F. Gulminelli, Phys. Rev. C 97, 025806 (2018).
- [35]R. C. Tolman, Phys. Rev. 55, 364 (1939); J. R. Oppenheimer and G. M. Volkoff, Phys. Rev. 55, 374 (1939).
- [36]J. B. Hartle and D. H. Sharp, Astrophys. J. 147, 317 (1967).
- [37]I. A. Morrison, T.W. Baumgarte, S. L. Shapiro, and V. R. Pandharipande, Astrophys. J. 617, L135 (2004).
- [38]A. Watts et al., PoS 215 (AASKA14), 043 (2015).
- [39]V. Hambaryan, V. Suleimanov, F. Haberl et al., Astron. Astrophys. 601, A108 (2017).
- [40]F. Özel and P. Freire, Ann. Rev. Astron. Astrophys. 54, 401 (2016).
- [41]E. P. Mazets, S. V. Golenetskii, Yu. A. Guryan, and V. N. Il'inskii, Astrophys. Space Science 84, 173 (1982).
- [42]J. C. Higdon, and R. E. Lingenfelter, Ann. Rev. Astron. Astrophys. 28, 401 (1990).








