SP-345 Evolution of the Solar System



Physical and Chemical Structure of the Solar System





20.1. SURVEY

[339] In the theories derived from the Laplacian concept of planet formation it is usually postulated that both the Sun and the planets- satellites are often not even mentioned derive from a solar nebula with a chemical composition assumed to be uniform and characterized by "cosmic abundances" of elements. The Sun and the giant planets are supposed to have condensed directly from the solar nebula and are thought to have the same composition as this nebula. The solar photosphere has been proposed as the closest available approximation to this composition (Suess and Urey, 1956). The terrestrial planets should consist of the refractory ingredients of the nebula condensing in the inner regions of the solar system.

We have summarized earlier a number of objections to Laplacian-type theories, including the difficulty that not even bodies as large as Jupiter can condense directly from a nebula (sec. 11.2). The only reasonable alternative was found to be the planetesimal approach. To the objections discussed earlier we should add that the composition of the solar system appears far from uniform. It is well known that densities derived from mass and size indicate substantial differences in chemical composition among the different outer planets, among the terrestrial planets, and among the small bodies in the solar system. The notable variability in surface composition of asteroids supports this conclusion.

The marked differences in composition among the various groups of meteorites and comets also point at fractionation processes operating on matter in the solar system, before or during the formative stage. The observational evidence for the chemically fractionated state of the solar system will be discussed in this and the next chapter.



The empirical knowledge we have about the chemical composition of the solar system may be categorized with regard to level of certainty:

[340] (1) Surface layers and atmospheres. The surface layers of the Earth and Moon have been analyzed under well-defined conditions. The data refer to less than 10-3 of the total mass of these bodies. Fragmentary information from landed instruments has been obtained from Venus.

The surface layers of the Sun have been analyzed by remote spectroscopy; however, the error limits are generally large in comparison to elemental fractionation factors characteristic of planetary processes (Aller, 1967; Urey, 1972; Worrall and Wilson, 1972). Independent indications derive from analysis of corpuscular radiation from the Sun (Price, 1973), but they clearly represent material fractionated at the source. The composition of solar wind is found to vary up to a factor of 3 (S) to 10 (He) from assumed solar abundances.

Emission, absorption, and polarization of electromagnetic radiation by planets, satellites, and asteroids give some qualitative information about the structure and chemical composition of their surface layers (of the order of a fraction of a millimeter up to a few centimeters in depth) and of their atmospheres (Dollfus, 1971b; Gehrels, 1972a; Chapman, 1972a; Newburn and Gulkis, 1973).

(2) Bulk composition. Our knowledge of the bulk composition of the planets and satellites is extremely uncertain. Parameters that yield information on this question are

(a) Mass and radius, from which average density can be calculated.
(b) Moment of inertia, which allows conclusions about the density distribution.
(c) Seismic wave propagation, electric conductivity, heat flow, magnetic properties, and free oscillations, which have been studied in the case of Earth and Moon. The resulting data can be inverted to model internal structure and, indirectly, composition, but generally with a wide latitude of uncertainty.


In the case of Jupiter, the observation of a net energy flux from the interior also places a limit on the internal state.

Extrapolation of bulk composition from chemical surface properties of Earth, Moon, and Sun has been attempted but is necessarily uncertain.

Several hundred meteorites have been analyzed. These are of particularly great interest since they are likely to approximate the bulk composition of both the bodies from which they came and the parent streams of particles from which these bodies accumulated. A major limitation of this material as a record of the formative processes in the solar system comes from the fact that the regions of origin and the genetic interrelation of different types of meteorites are uncertain (chs. 22 and 19).



The solar system is generally considered to have formed by emplacement of gas and possibly solid dust in some specific configuration in space and time. Regardless of the details assumed with regard to this configuration and the state of its component matter, it appears highly unlikely that such an emplacement, which by definition involves the release of an enormous amount of energy, would proceed without accompanying chemical separation effects (see secs. 21.11 and 21.12). It is also improbable that the subsequent thermal evolution of each emplaced portion of matter would take place without some degree of chemical separation of the components. Hence the solid condensates, forming in the solar system in different regions and at different times as precursor material of the subsequently accreting bodies, were probably chemically different from each other.

If we could precisely determine the chemical differences among bodies from known and widely separated regions in the solar system, planets, satellites, comets, and the Sun, it should be possible to study in detail the effects of fractionation processes active in the hetegonic era. However, direct chemical measurements of the bulk composition of large celestial bodies do not exist; in the course of accretion and subsequent thermal evolution all such bodies must have become stratified, and we are unable to obtain samples deeper than a thin outer layer. We know with certainty that even a body as small as the Moon has thoroughly altered the primordial material from which it accreted. Consideration of accretional heating as a function of terminal velocity of the source particles at impact suggests that the effect of the accretional hot-spot front would be considerable for bodies larger than a few hundred kilometers (sec. 12.12). Hence even the surfaces of the largest asteroids would not be representative of the bulk composition of these bodies. For this same reason we have no certain knowledge of the deep interior chemical composition of any planet, not even our own (secs. 20.4(1) and 20.5.1).

On the other hand, bodies with sizes of tens of kilometers and smaller are likely not to have been subject to accretional and post-accretional differentiation of this kind, and it should be possible to determine their bulk composition from samples of surface material. The small asteroids and comets and the matter trapped in the Lagrangian points of the larger planets (such as the Trojan asteroids of Jupiter) are possible sources for such samples. The meteorites (ch. 22) in all likelihood constitute samples of such small bodies.



As stated in sec. 20.2, in most cases the measurement of the average density is our main source of information about the bulk chemical composition of a body. However, interpretation of the mean density in terms of chemical composition is often difficult because we know so little about the state of matter at high pressure. Nor do we have satisfactory information about the properties of solid bodies aggregated in low gravitational fields.

(1) Matter at high pressure. Static pressure experiments with satisfactory calibration extend into the range of a few hundred kilobars (Drickamer, 1965), corresponding to pressures in the upper mantle of the Earth. In transient pressure experiments using shock waves, pressures in the megabar range can be reached (see, e.g., McQueen and Marsh, 1960). Although such experiments are useful in studying elastic compression effects, their general applicability is more questionable in studies of materials undergoing high-pressure phase transformations. The reason is that the material in the shock front is strongly heated, and the relaxation time for phase transformations may be long compared to the duration of the pressure pulse.

Under these circumstances it is difficult to predict with certainty the structure and composition of matter in the deep interior of the planets. The interpretation of the nature of the cores of the Earth and Venus, for example, has important consequences with regard to the inferred chemical composition of these planets. Lodochnikov (1939) and Ramsey (1948, 1949) proposed that the high density of the core of the Earth and the high bulk density of Venus could be due to pressure-induced transformation of magnesium-iron silicate into a high-density phase. If this were the case, the Earth's core and mantle could have the same chemical composition. Although the formation of an unknown high-density phase may possibly have escaped detection in transient compression experiments, it has been considered unlikely (see, e.g., Samara, 1967) that such a density change could assume the magnitude required, about 70 percent, at the core-mantle boundary. Recent experiments (Simakov et al., 1973) suggest, however, that minerals which already possess close-packed structures before the shock experiment undergo phase transitions of this kind at shock pressures in the megabar range.

The alternative explanation is that the Earth's core consists of material with higher mean atomic mass than that of the mantle; for example, nickel-iron with some lighter elements such as silicon or sulfur (Birch, 1964; Ringwood, 1966; Murthy and Hall, 1970; Lewis, 1971a). Much of the uncertainty concerning the properties of materials in the pressure range typical of the terrestrial planets could probably be clarified in the near future due to progress in high-pressure experimental studies. This would, however, not solve the problems of the state of matter in the giant planets.

[343] (2) Grain aggregates. According to ch. 11, planets and satellites must have formed from smaller bodies (planetesimals) and ultimately from small condensed particles. Such particles can accumulate to form larger bodies only if they are held together by an attractive force. Since gravity is negligible in the incipient growth stages, the main initial cohesive effect is likely to have been provided by electric charge and vapor deposition, as exemplified by the lunar soil. The nature of such aggregates and the dynamic conditions of their formation are discussed in secs. 7.4 and 22.6-22.7. High porosity and hence low bulk density may thus have been common in the initial stages of planetesimal accretion and still occur today in bodies that have remained at a small size. A major portion of the solid matter inter-...


FIGURE 20.4.1.- Impact cratering of the Martian satellite Phobos.

FIGURE 20.4.1.- Impact cratering of the Martian satellite Phobos. In suitable illumination craters such as A above can be seen to have rims of substantial height above the surrounding terrain. Since ejecta with velocities exceeding a few meters per second will leave the satellite, the crater cones cannot be generated by fallout from the impact as is the case on Earth, Mars, and the Moon. The dimensions of the cones also appear larger than the elevation of crater rims observed on the Earth as a result of shock rebound. A possible explanation of this phenomenon is that Phobos, or at least its outer regions, consists of aggregate material with low bulk density, and that impacting projectiles dissipate their energy largely below the target surface. (NASA Photograph 71-H-1832.)


[344] ..-cepted by Earth appears to have fluffy texture with mean bulk densities below 1.0 g/cm3. Such materials are destroyed during passage through the atmosphere (Verniani, 1969, 1973; McCrosky, 1970).

Although gravitational compaction would be practically absent in bodies of small size, shock compaction of the original texture would be expected as a result of collisions leading to repeated breakup and reaccumulation during the evolution of jet-stream assemblages of such bodies. Evidence of a wide variety of such effects is given by textures in meteorites ranging from complete melting (achondrite parent rocks formed from melts; e.g., Duke and Silver, 1967) and shock-induced reactions and phase changes (Neuvonen et al., 1972) to less dense packing (10-20 percent porosity) without fusion bonding between the particles, such as in carbonaceous meteorites (see fig. 7.1.1) and some chondrites.

The largely unexplored fluffy state in some small bodies in the solar system could have important consequences for their response to collision and hence for the processes of disruption, accretion, and chemical fractionation (sees. 7.4 and 22.6).

The Martian satellites are the first small objects in space studied with sufficient resolution to record discrete surface features. The two satellites are saturated with impact craters and these have characteristics that suggest the possibility of porous target material (fig. 20.4.1).



Physical data available for the planets and satellites are listed in table 20.5.1, together with estimated uncertainties.

In many respects the information from the Earth is most reliable. For this reason we shall begin with the data and theories relating to the composition of our own planet.


20.5.1. Earth

In the Earth and Moon, the only sampled terrestrial planets, the surface composition implies that oxygen is the most abundant element to considerable depth. At a depth of a few hundred kilometers in the Earth, the density is likely to be controlled essentially by close-packed oxygen ions.

The steep increase in density indicated at the core-mantle boundary has been interpreted in different ways:

(1) One suggestion is that the boundary represents a pressure-induced phase transformation associated with a substantial decrease in specific volume and with band gap closure resulting in metallic conductivity. The [345] general background of this proposition has been discussed in sec. 20.4(1) Objections against it are partly based on the results of model experiments which have failed to produce the high-density silicate phase. These results are, however, not entirely conclusive since the experiments employ transient shock rather than static pressure; hence, transformation with relaxation times longer than the shock duration would not necessarily be reproduced.

(2) To avoid the assumption of a hypothetical high-density silicate phase, the other current interpretation assumes that the core differs distinctly from the mantle in chemical composition and consists mostly of nickel-iron alloyed with 10-20 percent of light elements such as silicon or sulfur. This hypothesis requires a mechanism to explain the heterogeneous structure of the Earth. It also implies a high concentration of iron in the source material from which the Earth was formed.

Four types of mechanisms have been suggested to account for the proposed separation of an oxygen-free metal core from a mantle consisting mainly of silicates:

(1) A metallic core developed as a result of accretional heating. The progression of the accretional hot-spot front has been discussed in secs .12.11-12.12; this analysis shows that (a) the Earth's inner core should have accreted at low temperature; (b) runaway exhaustion of the source material in the terrestrial region of space would have coincided roughly with the formation of the outer core; and (c) the mantle accreted at a low mean temperature but with local heating at each impact causing light melts to migrate outward with the surface of the growing planetary embryo. Hence heavy differentiates including metal would not be able to sink further than to the bottom of locally melted pools. Large-scale simultaneous melting and sinking of metal over large radial distances would be limited to the still-liquid outer core, entirely melted in the runaway phase of accretion.

Complete melting of the entire planet at catastrophic accretion has been proposed by Hanks and Anderson (1969) as a means for gravitational separation of a metallic core. This approach, however, does not take into account a distribution of matter preceding accretion, which satisfies the boundary conditions for obtaining the present structure of the planet and satellite systems. Furthermore, it meets with the same objection as any scheme involving complete melting of the Earth, further discussed in (2) below.

(2) The Earth' core developed during or after the accretion of the planet. This type of theory has been developed in detail by Elsasser (1963, and Birch (1965). Elsasser suggested that the Earth accreted as a homogeneous body consisting of a mixture of metal, silicates, and sulfides, similar to meteorite material. The interior of the planet heated up gradually due to radioactivity decay, reaching the melting point of iron (or the eutectic point in the iron-sulfide system at about 44 atomic percent S (Murthy and Hall 1970; Lewis 1971a)) at a depth below the surface determined by the pressure.....


346] TABLE 20.5.1. Physical Data for the Planets, Former Planets (Moon and Triton), and Asteroids


Mean orbital radius rorb from Sun (1012 cm)

Radius (108 cm)

Mass (1027 g)

Average density (g/cm3)

Best estimate

Upper limit

Lower limit



a 5.791

b 2.434 ± 0.002

b 0.3299 ± 0.0029

h 5.46

h 5.53

h 5.40


a 10.821

b 6.050 ± 0.005

b,f 4.870

b 5.23




a 14.960

a 6.378

a 5.976

b 5.52




j ~14-20

a 1.738

b 0.0735

b 3.35




a 22.794

b 3.400

b 0.6424

b 3.92




i 35.331

d 0.285 ± 0.015

e 0.00024 ± 0.00003

h 2.5

h 3.3

h 1.9


i 41.402

d 0.567 ± 0.042

e 0.00119 ± 0.00014

h 1.6

h 2.2

h 1.1


c 77.837

c 71.60

c 1899.

c 1.31




c 142.70

c 60.00

c 568.

c 0.70




c 286.96

c 25.40

c 87.2

c ~1.3




c 449.67

c 24.75 ± 0.06

c 102.

c 1.66




j ~ 450

c 1.88 ± 0.65

c 0.135 ± 0.024

c ~4.8

h 20.1

h 1.6


c 590.00

c,g 3.20 ± 0.20

k 0.66 ± 0.018

c 4.9

h 7.4

h 2.9

a Allen, 1963.
b Lyttleton, 1969.
c Newburn and Gulkis, 1973.
d Morrison, 1973
e Schubart, 1971.
f Howard et al., 1974.
g Upper limit of radius obtained from near occultation.
h Average density is calculated from the mean values of mass and radius given in the table. The upper density limit is calculated by combining the lower estimated error limit for the radius and the upper estimated error limit for the mass, and vice versa for the lower density limit. Spherical shape is assumed for all calculations.
i Ephemerides of Minor Planets for 1969.
j The distances of interest in the present discussion are those at the time of formation. Since Moon and Triton are considered to be captured planets, their original orbital radius can only be approximated.
k Seidelmann et al., 1971; these authors suggest a mass error of 16-17 percent. We have here arbitrarily used a higher value ( +25 percent) in order to, if anything, exaggerate the uncertainty margin of the density estimate.


[347] TABLE 20.5.1 (Continued) Physical Data for the Regular Satellites With Radii > 108 cm




Mean orbital radius rorb from Sun (1010 cm)

Radius (108 cm)

Mass (1024 g)

Average density (g/cm3)

Best estimate

Upper limit

Lower limit




c 4.22

d 1.83 ± 0.01

d 89.2 ± 1.1

d 3.5




c 6.71

d 1.50 ± 0.05

d 48.7 ± 1.1

d 3.45

d 3.75

d 3.1


c 10.71

d 2.64 +0.01

- 0.10

d 149. ± 1.5

d 1.9

d 2.2

d 1.9


c 18.84

d 2.50 ± 0.08

d 106. ± 3.2

d 1.6

d 1.75

d 1.5




c 2.38

c 0.27 ± 0.15

c 0.085 ± 0.028

c ~1.0

b 16.

b 0.2


c 2.94

c 0.60 ± 0.10

c 0.648 ± 0.017

c 0.7

b 1.3

b 0.4


c 3.77

a 0.58 ± 0.10

c 1.05 ± 0.03

a 1.4

a 2.0

b 0.8


c 5.27

a 0.80 ± 0.13

e 1.8 ± 2.2

a 0.9

b 3.2

b 0.0


c 12.22

c 2.42 ± 0.15

c 137. ± 1.

c 2.3

b 2.8

b 1.9


c 35.62

e 0.85 ± 0.10

e 2.24 ± 0.74

b 0.9

b 1.7

b 0.4

a Morrison, 1974.
b Average density is calculated from the mean values of mass and radius given in the table. The upper density limit is calculated by combining the lower estimated error limit for the radius and the upper estimated error limit for the mass, and vice versa for the lower density limit. Spherical shape is assumed for all calculations.
c Newburn and Gulkis, 1973.
d Anderson et al., 1974.
e Murphy et al., 1972.


[348] ....effect on melting. At further heating the point would be reached where the strength of the supporting silicate material became insufficient to sustain the gravitational instability due to the higher density of the iron (or ironsulfide) liquid. At this point the liquid would drain toward the center of the Earth, releasing potential energy. The energy release would be sufficient to completely melt the entire planet.

This scheme encounters difficulties from the time constraints in the Earth's thermal evolution. On one hand the core formation process is not allowed to begin until radioactive heating has raised the initiating material to the melting-point range and the supporting silicate material to its yield temperature. On the other hand, preserved segments of the crust are found which are as old as 3.6 Gyr. It is questionable if these limitations would allow complete melting of the planet to occur at any time in its early history, even as early as the time of accretion (Majeva, 1971; Levin, 1972). Such an event would also generate a heavy atmosphere containing the major fraction of the planet's accreted volatiles (ch. 26). This would be likely to prevent cooling to such a temperature that an ocean could form even today; nonetheless evidence for condensed water and development of life are found in the earliest preserved sediments, exceeding 3 Gyr in age (see ch. 26).

Another observation of importance in connection with the question of core formation is the consistently high content (~0.2 percent) of Ni+2 in the magnesium silicates from the upper mantle. If the metallic iron, now assumed to form the core, at one time was homogeneously distributed as small particles throughout the protoplanet, such as in stone meteorites, the melting, migrating droplets of iron would be expected to reduce nickel ion in the silicate phase and to remove the resulting metallic nickel into solution in the melt (Ringwood, 1966); hence, a metallic core is generally thought of as consisting mainly of nickel-iron (see, e.g., Birch, 1964). Accretional melting indeed leads to such extraction of nickel, as demonstrated by the conditions in the lunar surface rocks. These are low in metallic nickel iron, and have an order of magnitude less nickel ion in the magnesium silicates than do terrestrial mantle rocks. Generation of core metal by accretional or postaccretional reduction of iron silicates with carbon (Ringwood, 1959) would doubtless be a still more efficient way to remove nickel from the silicate phases. Hence the presence of substantial concentrations of oxidized nickel in the Earth's mantle also speaks against melt extraction of a metallic core from an originally homogeneous planet.

(3) The differentiation, ultimately leading to the formation of an iron core, is due to a solid grain interaction process in the Earth's jet stream. It has been suggested that condensed nickel-iron metal particles would aggregate together at higher relative velocities, and hence at an earlier [349] time in the evolution of the jet stream than silicate grains. This would be due to the plastic properties of the metal (Orowan, 1969) or to a high accretion cross section caused by magnetization of the grains (Harris and Tozer, 1967). Such a selective accretion of metal grains, if possible at all, could only occur when relative velocities had been brought down to the subsonic range since hypervelocity impact invariably leads to breakup and vaporization in the metal grains (Gault et al., 1968; Neukum et al., 1970).

Observations in meteorites do not provide support for this type of mechanism as far as preferential accretion of metal by collisional or magnetic processes is concerned. Studies of the state of metal grains in chondrites such as those by Urey and Mayeda (1959) do not indicate collision-induced welding. Nor do any observations appear to exist of clustering of metal grains, characteristic of magnetic accretion. In contrast, such clustering is indeed observed for ferromagnetic iron-oxide crystals (magnetite) accreted in space and subsequently aggregated into carbonaceous chondrites (fig. 22.7.1; Jedwab, 1967; Kerridge, 1970; Brecher, 1972a). Arguments have been given by Banerjee (1967) against magnetostatic accretion of multidomain grains of nickel-iron. Finally, runaway accretion in the Earth's jet stream would take place at about 1/10 of the present mass of the planet, corresponding to the mass of the core. Even if it had been possible to selectively accrete metal and leave silicate material behind in the jet stream during the formation of the inner core, all the material orbiting in the source region of the Earth, regardless of composition, would be swept up during the runaway accretion coinciding with the formation of the outer core (sec. 12.6).

(4) The differentiation took place in conjunction with the gas emplacement and condensation processes. A suggestion of this kind, now mainly of historical interest, was made by Eucken (1944a). It has recently been revived in modified form by Turekian and Clark (1969) but without application of the physical constraints of condensation (Arrhenius and De, 1973) or accretion dynamics (secs. 12.1-12.7). This type of hypothesis could in principle be made physically and chemically consistent if it is assumed ad hoc that the composition of condensable impurities in the region of the inner terrestrial planets changed with time, having higher iron content during the first ~3 X 107 yr of infall (the order of magnitude of time required for accretion of the Earth's core; see secs. 12.8-12.9).

If it were conclusively demonstrated that the high densities of the Earth and Venus are due to a high content of iron, this fact would lend observational support to an assumption of a change with time of the composition of the source materials of these planets. At the present time such an assumption, although speculative, receives some support from the relationships discussed in sec. 21.12.2.


[350] 20.5.2. Mercury

Mercury with a radius of 0.38 mathematical symbolhas a pressure at the center which is as low as that in the Earth's upper mantle (Lyttleton, 1969). In spite of this Mercury has a density as high as 5.46 g/cm3. This can be understood in terms of the general mechanism for fractionation in the inner solar system discussed in sec. 20.5.1.


20.5.3. Venus

The discussion of the composition of the Earth in sec. 20.5.1 applies also to Venus, which has 85.5 percent of the volume and 81.6 percent of the mass of the Earth. Its density, estimated at 5.25 g/cm3, is only 5 percent less than that of the Earth. With the assumption of a core of densified silicate, Venus could have the same composition as the Earth, the Moon, and Mars. If, on the other hand, as is likely, excess iron is needed to account for the high bulk density in both Earth and Venus, these two planets, together with Mercury, would be distinctly different from the Mars-Moon group (fig. 20.7.1a).


20.5.4. Moon and Mars

Since there are strong indications that the Moon is a captured planet (Alfvén, 1942, 1943a, 1946, 1954;Urey, 1952;Gerstenkorn, 1969;Alfvén and Arrhenius, 1972), it is here included in the discussion of planetary compositions.

The observed chemical composition of the lunar surface cannot be characteristic of the interior. If the high thorium and uranium contents of the surface rocks persisted at depth, the lunar interior would be extensively pelted, but seismic observations indicate possible partial melting only in :he central region below 103 km (Toksoz et al., 1972).

Furthermore, rocks of the observed surface composition of the Moon would, in the interior and in a limited zone in the lower crust, seem to transform to high-density assemblages (seismic data may indeed indicate such a transformation in the lower crust (Toksoz et al., 1972)). If these high-density phases prevailed throughout the interior of the Moon its average density would be considerably higher than the observed value, 3.35 g/cm3 Wetherill, 1968). Therefore, the higher content of radioactive elements in he outer crust as well as its basaltic-anorthositic composition suggests either that the Moon accreted sequentially from materials of different chemical compositions (Arrhenius, 1969; Arrhenius et al., 1970; Gast, 1971) or that a differentiation process selectively removed the critical components from the interior to the surface.

[351] The latter explanation would appear possible since it is difficult to escape the conclusion that an accretional front of hot spots has swept through the mantles of the terrestrial planets including that of the Moon (sees. 12.9-12.12). Such a progressive zone melting would be likely to cause removal to the planetary surface region of components with low melting temperature range, low density, or large ionic radius (Vinogradov, 1962; Vinogradov et al., 1971). The crusts of the Earth and the Moon consist of such materials except that much of the volatile components appear to have escaped thermally in the low gravitational field of the Moon (see ch. 26).

The former suggestion, namely that the source material for the lunar interior differed in composition from the material that formed the outer layer of the Moon, may seem more ad hoc. However, support for such an assumption can be drawn from the closeness and possible overlap of the A and B regions where the source materials of the terrestrial planets condensed. These relationships are discussed in sec. 21.12.2.

Regardless of the cause of the lunar differentiation, the low mean density of the Moon (table 20.5.1) makes it clear that it differs chemically from Mercury, most likely by having a lower iron content. It is also possible that the Moon differs substantially from the Earth and Venus in bulk chemical composition. This possibility becomes certainty if it can be verified that the latter two planets owe their high densities to a high content of iron (see sec. 20.5.1).

The bulk density of Mars, 3.92 g/cm3, suggests that the bulk proportion of heavy to light elements is similar to that of the Moon, and hence lower than those of Venus and the Earth (see fig. 20.7.1a).


20.5.5 Asteroids

These bodies are of sufficiently small size that pressure-induced phase changes can be neglected. On the other hand, asteroids of a size larger than about 100 km have gravitation that is probably large enough to effectively compact fluffy material. Hence some of the uncertainties in data interpretation discussed in sec. 20.4 do not apply to such large asteroids. Their densities, in the few cases where they are known at all, furnish suggestive information on gross chemical composition.

Mass determinations from gravitational perturbation of the orbits of other asteroids exist only for Vesta and Ceres (Schubert, 1971). These values, combined with the most accurate measurements of radii (Morrison, 1973) give a density of 1.6±0.5 g/cm3 for Ceres and 2.5±0.7 g/cm3 for Vestal In bodies like these, several hundred kilometers in size, porosity can probably only be maintained in a small surficial region. The low densities, if correct, therefore suggest the presence of hydrous minerals or ice in the interior or (less likely) rocks virtually free of iron. Optical measurements by Chapman [352] et al. (1971) indicate that the surface layer of Vesta consists of material with absorption properties closely similar to the meteorites known as calcium-rich eucrites (density 3.4-3.7 g/cm3) which are also similar to some common lunar rocks. Ceres, in contrast, has a lower albedo and more bluish color than Vesta and lacks diagnostic absorption bands; it does not bear close resemblance to any known type of meteorite (Chapman, 1972a).

The optical properties of the dusty surface material of the near-Earth object 1685 Toro (Gehrels, 1972b) are similar to those of the most common type of chondritic meteorite (Chapman, 1972b). In general, however, the asteroids show widely differing optical surface properties (Chapman et al., 1971). We do not yet know to what extent, if any, there is a corresponding variation in their bulk composition.


20.5.6. Jupiter and Saturn

These planets are so massive that our lack of knowledge of matter at high pressures precludes any detailed speculation about their chemical composition. Not even a meaningful comparison between Jupiter and Saturn can be carried out in view of the large difference in size between them.

Attempts have been made to construct models for the different giant planets (DeMarcus, 1958: DeMarcus and Reynolds, 1963; Reynolds and Summers, 1965), but the assumptions used are necessarily highly uncertain. Existing calculations are generally based on the arbitrary assumption that the composition of the source material of all planets and satellites is the same and, specifically, is that of the solar photosphere. Such assumptions are in conflict with the wide variation in bulk densities observed among small bodies in the solar system.

Furthermore, in order to draw conclusions about the chemical composition from the average density of a body it is necessary to know the internal temperature distribution. However, attempts to estimate interior temperatures are highly sensitive to the assumed composition and to the unknown properties of the elements in question at high pressure. If the interior of Jupiter is assumed to be at relatively low temperature and to consist of solid metallic hydrogen and helium, accretional heat could then effectively be removed by conduction. The discovery of excess energy emission from Jupiter (Hubbard, 1969; Bishop and DeMarcus, 1970) has, however, shown that this commonly accepted picture is unrealistic, and leaves us with a wide range of uncertainty regarding interior temperature and chemical composition. It should be noted that during planetesimal accretion the primordial heat distribution probably differed substantially for the individual planets (sec. 12.11; fig. 12.11.1). This distribution is likely to have affected the present-day internal temperature profile.

Finally, a strong magnetic field, as existing in Jupiter and possibly also in [353] the other giant planets, could profoundly affect the heat transfer in a liquid or gaseous interior by inhibiting convection. Hence, the interior temperature of the giant planets may well be much higher than existing models have indicated, and the average atomic mass could also be correspondingly higher.

Although space missions to the giant planets will certainly provide additional information with direct or indirect bearing on the problem of the interior state, this problem is likely to remain in a speculative state for a long time. Suggestive information on the completely unknown composition of the nonvolatile material in the giant planets could perhaps be obtained from residues of the source material in their regions of formation. Small bodies in the Lagrangian points L4 and L5 (Trojan asteroids in the case of Jupiter) may consist of such material.


20.5.7. Uranus and Neptune

The uncertainties in chemical composition, further complicated by the unknown internal thermal states of Jupiter and Saturn, apply also to Uranus and Neptune. However, because of the close similarity in size (Neptune possibly being slightly smaller but definitely more massive than Uranus; see table 20.5.1), comparison of physical properties of this pair is, to some extent, meaningful. It is interesting to note that the density of Neptune (1.66 g/cm3) at a solar distance of 30 AU is larger than that of Uranus (1.3 g/cm3) at 19 AU, and both are much denser than Saturn at 9 AU (see fig. 20.7.1a).


20.5.8. Triton

The retrograde orbit of this body, now a satellite of Neptune, indicates that it was captured from a planetary orbit (McCord, 1966) and underwent an evolution partially similar to that suggested for the Moon (Gerstenkorn ,1969; Alfvén and Arrhenius, 1972). Mass and radius for Triton have been measured with estimated errors of ±18 percent and ±30 percent, respectively. A combination of the extremes would give a lower density limit of 1.6 g/cm3 and an upper exceeding 8 g/cm3; the "best" value is around 5 g/cm3.


20.5.9. Pluto

Considering even the largest estimated "possible" errors in the values for the mass and diameter of Pluto, it is difficult to escape the conclusion that its density considerably exceeds 2 g/cm3. A density of 4.8 g/cm3 (calculated assuming a radius of 3200 km, a value close to the definitive upper limit of 3400 km set by occultation measurements) is regarded as the best estimate [354] (Newburn and Gulkis, 1973, p. 253; and Seidelmann et al., 1971). Combining the occultation volume limit with a negative mass error of 25 percent of the best estimate gives a "minimum" density of 2.9 g/cm3. To bring the mass estimate into lower values it would be necessary to assume much larger errors in the mass estimates for Neptune and Saturn (Halliday, 1969) than are presently believed to be feasible (Newburn and Gulkis, 1973). The lower limit for the radius is less precisely defined than the upper limit, but it cannot be much different from the estimated best value since lowering of the radius rapidly results in unreasonably high densities (table 20.5.1).

Pluto, like Triton, is sufficiently small to rule out the possibility of unknown high-density phases in its interior. The relatively large bulk density of Pluto consequently indicates a substantial fraction of rocky material, and, if the best present estimate is close to reality, also a significant proportion of iron.


20.5.10.Bulk Density in Relation to Planetary Mass

The densities of the terrestrial planets, discussed above and summarized in table 20.5.1, have been plotted against planetary mass in fig. 20.5.1. A regular increase in density with increasing mass is found in the series Moon-....


FIGURE 20.5.1.- Density of the terrestrial planets as a function of their mass.

FIGURE 20.5.1.- Density of the terrestrial planets as a function of their mass. A smooth curve could be drawn through Moon-Mars-Venus-Earth indicating that all may have a similar composition. This would require the assumption that Moon-Mars-like material can be compressed to the high core densities indicated (~15-17 g/cm3) at the core pressures of Venus and the Earth (~1.5 Mb). But it is also possible that Moon and Mars have a heavy element content entirely different from that of Earth-Venus. The composition of Mercury must in any case be different from all the other bodies.


[355] ...-Mars-Venus-Earth. This density increase could possibly be due to compression including pressure-induced phase transformations; if this were the case, the chemical composition of all these bodies might be the same.

On the other hand, arguments can be made for a higher content of heavy elements in Venus and Earth than in the Moon and Mars (secs. 20.4(1) and 20.5.1). However, for Mercury it is in any case necessary to assume a difference in chemical composition, presumably a higher iron content.

The densities of the outer planets have been plotted as a function of their masses in fig. 20.5.2. Also in this group it is obvious that factors other than the mass determine the densities of the planets.


20.5.11. Compositions of Satellites

Except for the Moon, which is here considered as a planet, satellite mass and radius values are most reliable for the Galilean satellites of Jupiter. The reported values of their densities display marked differences, the two smaller inner satellites (Io and Europa) consisting of more dense material (3.1-3.75 g/cm3) than the outer ones (Ganymede and Callisto) (1.5-2.2 g/cm3) (table 20.5.1). This density variance probably indicates differences in the proportion of light elements in icy or liquid compounds to the heavier elements as....


FIGURE 20.5.2.- Density of the outer planets as a function of their mass.

FIGURE 20.5.2.- Density of the outer planets as a function of their mass. It is difficult to believe that the density variation can be due to only the difference in mass.


[356] ...found in earthy components (Lewis, 1971b), and demonstrates again the nonuniformity in composition of the source materials and bodies in the solar system.

The densities of the Saturnian satellites are poorly known except perhaps for Titan with a reported density of 2.3 g/cm3. The estimated densities for the other satellites (table 20.5.1), to the extent they can be relied upon, would suggest variations by a factor of four.

The densities of the Uranian satellites are completely unknown.



20.6.1. Spectrometric Analysis

In principle, the composition of the solar photosphere, the chromosphere (including prominences), and the corona can be found by spectrometric analysis. This involves two steps; namely, measurement of line intensity profile, etc., which can be made with a high degree of accuracy, and, secondly, calculations of abundances from the spectrometric data based on models of the solar atmosphere. The models are usually homogeneous in the sense that they assume that light received by the spectrograph emanates from a region with density and temperature which are functions of only the height in the atmosphere.

As pointed out in sec. 15.3, homogeneous models are often misleading in astrophysics. In the case of the Sun, a homogeneous model is unrealistic, since we know that the solar atmosphere has a fine structure with elements of a size down to the limits of resolution and presumably still smaller. The differences in temperature and density between such elements are so large that the averaging introduced by the homogeneous model may cause gross errors. It is well known that solar magnetograph measurements are seriously in error, and in many cases it is even doubtful whether solar magnetograms can be interpreted at all. This is suggested by the fact that the "magnetic field" derived from solar magnetograms does not obey Maxwell's equations (Wilcox, 1972). It is possible that the major uncertainties in chemical analysis by means of spectral analysis (Worrall and Wilson, 1972) are due to the same inhomogeneity effects. This must be clarified before we can rely on spectrometric results for abundance estimates for more than an order-of magnitude accuracy.


20.6.2. Analysis of Corpuscular Radiation From the Sun

Space measurements of solar wind composition and of solar cosmic rays have provided quantitative information on the chemical composition of the material emitted from the upper corona and of the flare regions (Price,...



FIGURE 20.6.1- Coronal streamers, visible at solar eclipse.

FIGURE 20.6.1- Coronal streamers, visible at solar eclipse. The photograph illustrates the inhomogeneous nature of emission of solar material. Homogeneous models of the Sun are often completely misleading.


...1973). The abundances obtained from these measurements have no simple relationship to the chemical composition of the regions from which they derive because of selective processes during emission (see fig. 20.6.1). We know very little about the fractionation processes themselves; however, fluctuations in them are manifest by variations of two orders of magnitude in the helium content of the solar wind (Hirschberg, 1973), and also by variations in the heavier elements (Price, 1973; Price et al., 1973).

Long-term integration of the corpuscular flux may eliminate the effects of short-term fluctuations in selective emission processes and give clues to their nature. However, they leave unknown any permanent differences between the composition of the Sun and the material that leaves it.


20.6.3. Significance of Solar Photospheric Abundance Data

For the reasons outlined above, elemental abundances in the accessible layers of the Sun are known with much less accuracy than in samples of the Earth, Moon, and meteorites analyzed under controlled conditions, and it is difficult to assign a probable error to any individual elemental abundance determinations (Urey, 1972).

[358] It is often assumed that the bulk composition of the Sun is identical to some undifferentiated matter that was conjectured to be the source of other bodies in the solar system. This assumption derives from the Laplacian concept that all the matter of the solar system taken together once formed a dense solar nebula. It was further assumed that throughout the presumed process of contraction and dynamic differentiation of such a nebula, the chemical composition somehow remained uniform.

As has been discussed in detail in other sections of this work, theories of this type are unrealistic since they ignore many of the important facts concerning the observed present state of the solar system and do not incorporate modern knowledge of the behavior of particles and fields in space. Hence there is no reason to believe a priori that the Sun has a composition which accurately corresponds to that of the bulk of any satellite, planet, or group of meteorites. Indeed, this is demonstrated already by the observed variability in composition of rocky components among various bodies in the solar system (secs. 20.5 and 20.7). Furthermore, we do not know whether the surface composition of the Sun is representative of its bulk composition. Theories of the solar interior are not very useful since they seem to be seriously out of line with observation (Fowler, 1972).

The range of actual variation in chemical composition is hard to specify because we have sampled only a few of the relevant bodies, and most of these are strongly differentiated. An indication of the variations in composition is given by the range of densities of the small bodies in the solar system (sec. 20.5.11) and, on a smaller scale, by the differences in composition between unmodified primordial condensate components in meteorites from different parent jet streams.

In order to place limits on the differences between accurately measurable materials such as meteorites and approximately measurable materials such as solar photosphere, comparisons such as those in fig. 20.6.2a are useful. Carbonaceous chondrites of Type I (Wiik, 1956) have been chosen for the comparison since there is general agreement that they consist of primary condensate material (one of the necessarily many different types) which does not seem to have been significantly modified with regard to elemental composition after condensation.

Elemental abundance data on this type of meteorite were obtained from a methodologically critical review compiled from work by a number of analytical experts (Mason, ed., 1971). To avoid bias in selection of analyses, all reported measurements accepted in that review have been included without preferential selection. The solar abundances are taken from the evaluations by Muller (1968) and Grevesse et al. (1968). In the case of the solar abundances a potential bias may be caused by the presumption that the solar and meteoritic abundances ought to converge on a value, referred to as the "cosmic abundance." The literature indicates that marked deviations [359] from such agreement become subject to more extensive scrutiny, revision, rejection, and exclusion than do the abundance ratio estimates which fall close to 1.0. The distribution shown in fig. 20.6.2a therefore probably represents a minimum dispersion.


FIGURE 20.6.2a.- Comparison of solar photospheric abundance estimates with measurements on carbonaceous meteorites of Type I.

FIGURE 20.6.2a.- Comparison of solar photospheric abundance estimates with measurements on carbonaceous meteorites of Type I. Each analytical chondrite value, normalized to silicon, has been divided by each of the several current photospheric values. Four of the ratio values for mercury (Z = 80) exceed 20 and are not shown in the diagram. Data compiled by L. Shawl It has commonly been assumed that these two materials can be regarded as splits from a chemically homogeneous body "the solar nebula" having "cosmic abundances" of elements. Except for components with high vapor pressures or nuclear instabilities the compositions of these meteorites and of the solar photosphere then ought to approach identity, and the elemental abundance ratios should be close to 1. The strong scatter of data in the figure shows, however, that they do not provide a basis for the assumption of a close agreement between the solar photosphere and this group of meteorites (see also fig. 20.6.2b).


[360] As shown by fig. 20.6.2b, for about 50 percent of all abundance pairs determined, the solar and meteoritic values are within a factor five of each other. About 10 percent of all elements deviate by more than a factor of 60. The most extreme cases are the relative concentrations of the noble gases (measured only in meteorites and not included in fig. 20.6.2b), mercury, thorium, uranium, and the rare earth elements. Particularly in the latter three cases it is difficult to tell what fraction of these deviations reflect real differences; the oscillator strengths are very poorly known and the solar data for these elements may have large experimental errors. The noble gas anomalies, on the other hand, are based on implanted vs. occluded components in meteorites and implanted solar emissions in lunar materials. These anomalies would consequently seem to reflect real fractionation of the kind expected in the emplacement and condensation process of solids (Signer and Suess, 1963; Jokipii, 1964; Arrhenius and Alfvén, 1971; Arrhenius, 1972).

It is clear from the comparison that observational uncertainties leave room for considerable differences in composition between the solar photosphere on one hand and various condensates such as the one represented....


FIGURE 20.6.2b. Frequency distribution of abundance ratios from fig. 20.6.2a.

FIGURE 20.6.2b. Frequency distribution of abundance ratios from fig. 20.6.2a. The diagram shows that on the average there is about 50 percent probability for solar photospheric observations to agree within a factor of five with their meteorite counterparts, and a 90 percent probability for agreement within a factor of 60. Ratios for elements with atomic number smaller or equal to10 are not included in this diagram since they are affected by preferential nuclear instabilities or are highly volatile. Neither are the noble gases included because their abundances in solids are strongly permuted due to volatility and other factors; furthermore their photospheric abundances are not known. Two abundance ratios exceed 128 and are not shown in the graph. Data compiled by L. Shaw.


[361] ....by carbonaceous meteorites, Type I, on the other. As indicated above there is no particular a priori reason why there should be any close agrement in composition between these materials. The differences in bulk densities among the individual planets and satellites discussed in sec. 20.5 are related to differences in abundances of the elements of which the bodies consist. Abundance differences of a factor of about four in the major condensable elements appear sufficient to explain the density differences among the small bodies in the solar system.



Our analysis of the solar system is based on the "hetegonic principle" implying that we should investigate to what extent the same relationships hold for all bodies formed in orbit around a primary body. From this point of view it is important to compare the chemical composition of the satellite systems and the planetary system. This is admittedly difficult because we know little about the chemical compositions of the planets and still less about those of the satellites. The only comparison we can make is between their densities.


20.7.1. Density As a Function of Potential Energy

As we shall see in ch. 21, there are reasons to believe that the emplacement of plasma in different regions around a central body is regulated by the critical velocity for ionization of the neutral gas falling toward the body. This implies that we should expect the abundances of elements in a system to vary with the gravitational potential energy. For this reason, it is useful to plot densities of the celestial bodies as a function of this gravitational potential energy (the ratio of the mass Mc of the central body to the orbital radius rorb of the body in question). In this way planets and satellites can be compared. Figure 20.7.1a shows gravitational potential energy as a function of density for the planets (including asteroids, Moon, and Triton), fig. 20.7.1b shows the satellite systems of Jupiter and Saturn, and fig. 20.7.1c shows a composite of planets and satellites. The parameter Mc/rorb allows a direct comparison of the planetary system and the different satellite systems.

Looking at figs. 20.7.1a, b, and c, we can conclude that the bulk densities decrease from the high value for Mercury, Venus, and Earth (at Mc/rorb = 3 X 1020 g/cm) to a minimum at a gravitational potential energy of about1019 g/cm (the region of Saturn in the planetary system) and then rise again to higher values with decreasing gravitational potential energy.



FIGURE 20.7.1a.- Average density of planets and former planets as a function of orbital distance rorb from the Sun.

FIGURE 20.7.1a.- Average density of planets and former planets as a function of orbital distance rorb from the Sun. The guideline through the population of density points is intended for intercomparison of this figure with figs. 20.7.1b and c. The ordinate is also given in terms of gravitational potential energy (mass, Mc, divided by orbital radius, rorb); this makes it possible to directly compare the distribution of satellites with that of the planets The gravitational potential energy is also a parameter which enters in an important manner in the discussion of the critical velocity phenomenon (see chs. 21 and 23). Since the Moon and Triton are captured planets, the Sun is regarded as their central body. Hence the Moon and Triton have the approximate gravitational potential energy of the Earth and Neptune, respectively. The horizontal lines through the points for Ceres, Vesta, Triton, and Pluto indicate the estimated range of uncertainty, with the vertical bar designating the lower limit for the density of Pluto as discussed in the text. Data from table 20.5.1.



FIGURE 20.7.1b.- Average density as a function of gravitational potential energy, Mc/rorb, for the regular satellite system of Jupiter and the two best known Saturnian satellites, Titan and Tethys.

FIGURE 20.7.1b.- Average density as a function of gravitational potential energy, Mc/rorb, for the regular satellite system of Jupiter and the two best known Saturnian satellites, Titan and Tethys. Solid circles denote density values based on the best estimates of radius and mass horizontal lines indicate the estimated range of uncertainty. Data from table 20.5.1.



FIGURE 20.7.1c.- Average density as a function of gravitational potential energy, Mc/rorb for the planets and better known satellites. The distribution indicates that heavy substances accumulated both in the inner and outermost regions of the systems, whereas light substances dominate in the intermediate region. Symbols are those used in figs. 20.7.1a and b. Data from table 20.5.1.


[365] 20.7.2. Chemical Meaning of Bulk Densities

The chemical meaning of the bulk densities of the large planets is rather uncertain. Because of the insignificance of pressure effects, the values for Mercury, Mars, Moon, Triton, Pluto, the asteroids, and the satellites are in principle more reliable, although possible measurement error is high in several cases.

The interpretation of the densities of Uranus and Neptune also suffers from the uncertainties related to compression and temperature in the large planets but they can be better intercompared because of the closely similar size of these two planets.

In the case of the least dense objects, namely Ganymede, Callisto, Tethys, and the giant planets, it is likely that substantial amounts of volatile light elements in unknown proportions contribute significantly to the low density. This indicates that heavy substances were accumulated both in the inner and the outermost regions of the systems, whereas light substances dominate in the intermediate region.


20.7.3. Density as Influenced by Solar Radiation

There is a common notion that the density of a body in the solar system is an inverse function of solar distance; this decrease in density is thought to be due to the decrease in radiation temperature at greater solar distances, which enhances capability for retaining lower density volatile elements and compounds. The fact that Neptune's density is higher than that of Uranus (which, in turn, is higher than that of Saturn) proves that this view is not correct. Together with the suggestive densities of Triton and Pluto this indicates that the chemical composition changes such as to give increasing density with increasing solar distance in this part of the solar system.


20.7.4. Theoretical Implications of Bulk Densities in the Solar System

We have seen above that bulk densities vary among the bodies of the solar system. This variation substantiates that the solar system did not form from a homogeneous medium. Hence it does not make sense to refer to any specific body in the solar system as representative of an average "cosmic" composition of the source materials, and the Sun is no exception. Furthermore, we know very little about the bulk composition of the Sun (see sec. 20.6).

Other conclusions to be drawn from our survey of the bulk densities in the solar system are that the density of a given body is not a function of [366] mass (see sec. 20.5.10) nor is it a monotonic function of the distance from the central body (see secs. 20.7.1 and 20.7.3).

Consequently, an explanation is needed for these variations in density, and presumably composition, in regions of different gravitational potential. A theory making detailed predictions of composition, however, cannot be verified because such detailed data are not yet available. An explanation of the variation of densities and compositions throughout the solar system, however, follows from consideration of plausible courses of the primordial emplacement of matter around the central bodies, such as discussed in ch. 21 and 23.