Chemical evolution of the terrestrial planets

Chemical evolution of the terrestrial planets

aeochlmi~et CosmoaMmka Acta.1906,VoL 30,pp.41&o104. Pixgamon PressLtd. Printedin NorthernIrelad Chemical evolution of the terrestrial planets A. E. D...

7MB Sizes 14 Downloads 102 Views

aeochlmi~et CosmoaMmka Acta.1906,VoL 30,pp.41&o104. Pixgamon PressLtd. Printedin NorthernIrelad

Chemical evolution of the terrestrial planets A. E. Dsparhent

RINCIWOOD

of Geophysics and cheat,

Australia

National U~ve~ty,

Canberra

(Reoeived3 May 1966) A-t-The terrestrial planets are believed to have formed by accretion from an initially cold and chemicallyhomogeneouscloud of dust and gas. The iron occurring in the dust particles of the cloud was present in E completely oxidised form. Either before or during aooretion of dust into planets, partial reduction of oxidised iron to metal ooourred. The role of oxidationreduction equilibria during the formation of terrestrial plan&r is discussed and it is concluded densitiesof the planets are caused dominantly by digering mean that the differingzero-pstates of oxidation which were established during the primary accretionprocesses, This interpretation avoids the neoessity for assuming the occurrenceof physical fractionation of metal from silicates in the solar nebula before accretion. A detailed study is made of the evidence shed by ehondritie meteorites upon oxidationreduotion equilibria occurring early in the history of the solar system. The ~~c~ce of the widely varying oxidation states of chrondrites is discussed. It is concluded that the different olassesof ohondriteshave formed by an autoreductionprocess operating upon primitive material similar in composition to the Type I carbonaceous chondrites. Reduction occurred when this material secreted into parent bodies wbiah were heated internally, perhaps by extinat radioactivities. Under these conditions, trapped carbonaceous material rea&ed with oxidised iron to produce a metallic phase &neitzl. Such a process explains the primary oxidation-reduction relationships in ohondritesestablished by PRIOR. The chemistry of the reduction protlesswhich operated in ohondrites is studied. The evidence strongly indicates that the principal reducing agent was oarbon and not hydrogen. Furthermore, reduction ocourred in a beak environment and 4totin the dispersed solar nebula. The origins and chemical evolution of other terrestrial plsnets are disoussedin the light of evidenceyielded by the chondrites. The hypothesis is advanced that each of the terrestrial planets formed by a single-stage autoreduotion process operating upon primitive material similar to the Type 1 oarbonaeous chondrites. In the case of the earth, autoreduction occurred at higher mean temperatures than in chondrites because of the large gravitational energy source which was involved. Accordingly, it is suggested that selective volatility played a more important part than in chondrites, and that mauy relatively volatile elements were lost from the earth. On the other hand the earth may have retained essentially the primordial abundances of elements which are not readily volatile under high-temperature, reducing conditions. A detailed study of the earth’s chemical compositionsupports this hypothesis. It is possible to construct a self consistentmodel from the primordial abundances of elements which are not readily volatile under high-temperature reducing conditions. The model implies the presence of silicon as an important component of the earth’s core. Independent evidence supporting this implication is cited. The ~~~bution and fractionation of oxyphile non-volatile elements imply that much or all of the mantle hss been subjected to complete or partial melting at some stage in its history. In contrast to the non-volatile elements, it appears that the earth has suffered strong depletion in a large number of volatile elements-Ha, K, Rb, Cs, Zn, Cd, Hg, Bi, TI, Pb, Cl, S and many otbers. It is suggested that loss of these elements by volatilization occurred during the primary accretion of the earth from primitive oxidised material, and that reduotion, complete melting, formation of the core, and fractionation of the mantle occurred during and immediately after the primary accretion process. Studies of the abundance of siderophile elements in the mantle, of the mean oxidation state of the mautle and of the nature of the volatile components which have been degassod from the mantle show that the mantle ie not and never has been in equilibrium with 41

42

A. E. RINGWOOD

the core. This conclusion pl&oes&n important constraint on the core-form&tionprocess. It is shown to be incomp&tibIewith the currently accepted theory th&t the m&terialfrom which the earth &ooretedw&s composed of an intimate mixture of silicate and met&l particles similaf to ordinary chondrites. The formation of the earth by direct accretion &nd autoreduction of primitive material resulted in the generation of an enormous primitive atmosphere composed princip&~y of CO and I-I,, together with the volatile elements mentioned above. It is suggested that this atmosphere subsequently escaped from the earth carrying the volatile elements mentioned abovc~. Possible mechanisms of escape &pediscussed. In the termin& phase of accretion, the temperrtturew&s sufficiently high t.o reduce and vol&tilisemagnesium &nd silicon monoxide from the inflling planetesimals and dust. At this stage, the condensed matter accreting on the earth consisted princip&llyof metallic iron and calcium ctndJuminum siliccttes. When the primitive atmosphere was disrupted and escaped, the ~~rnp~y~g exp&nsion &nd cooling c&used preGipit&tionof the non-volatile silicate oomponents of the atmosphere in the form of planetesimds and smoke. Precipitation of this m&teri&l,m&inly &s iron-poor mctgnesiansilicates occurredin & sediment-ring &roundthe earth. This meteri& becane mixed with primitive planetesimalspossessingthe composition of Type 1 c&rbonaceouschondrites, which h&d not accreted upon the earth. The sediment-rug of highly reduced magnesian silicate planetesimalsand primitive oxidised planetesimalsbecame unst&ble and coagul&tedto form the moon. The properties of the moon &re discussed in terms of its formation from such material. Possible expkmations of the moon’s density, luminescent properties, surface he~rog~ety, therm&lhistory and stresshistory emerge. A possibility that stoney meteorites &rederived from the moon is also discussed. It i8 distinctly possiblethat ordinarrychondritesm&y have formed by autoreduction and ffaotionation which occurred when primitive Type I carbonaceouschondrite pl~e~sim&ls collided with the moon during its terminal period of formation. Other theories of lunar origin sre briefly reviewed. The origins and internal constitution of the other terrestrialplanets &rediscussed. Mercury i8 believed to have accrated from the solar nebula at an initially high temper&ture,msint&ined by &neasly stage of high sofar luminosity. As a result, Mercury suffereddepletion of magnesian silicates which were reduced and volatilised under these conditions. The abundance of iron, which was not volstilised w&s correspondinglyincreased, resulting in a high me&n density for this planet. The evoIution of Venus was very simil&rto that of the earth. Its me&n state of oxidation m&y be slightly higher. The material from which Venus &ccretedpossessed& higher C/IX retie th&n the source m&teri&lof the ectrth. Differing cttmoaphericcompositionsare attribut&ble to this factor. Mars is composed of highly oxidised primordial material, with little or no met&lphase. Lack of reduction is attributed to the sm&llcontent of carbonaceousmaterial in the primordi&lmaterial from which Mars accreted. Physical propertiesand the therm&lhistory of Mars are discussedin terms of the proposed chemical constitution and the possibility of a self consistent solution is demonstrated.

Chemical evolution of the terrestrialplanets

CONTENTS 1. Introduction 2.

Causes of density variation among terrestrial planets

43

44 46

3. Chondritic meteorites (a) Introduction (b) Chemical composition (c) Oxidation-reduction equilibria (d) Nature of reduction process (e) Origin (f) Broader significance of chondrite evolution

49 49 50 54 56 58 59

4. The earth (a) Introduction (b) Mantle-core relationships (c) Composition and differentiation of the mantle (i) Fractionation of non-volatile oxyphile elements (ii) Fractionation of non-volatile siderophile elements (iii) Fractionation of volatile elements (iv) Implications of terrestrial fractionation pattern (d) Disequilibrium between mantle and core (e) Origin of the earth-multistage theories (f) A single stage hypothesis for the origin of the earth (g) Origin of mantle-core disequilibrium (h) Escape of primitive atmosphere

60 60 60 62 63 65 65 67 68 70 72 76 76

6. The moon (a) Introduction (b) Some current theories of lunar origin (i) The fission hypothesis (ii) The binary planet hypothesis (iii) Capture (iv) Coagulation of terrestrial “sediment-ring” (0) A model for the formation of the moon (d) Interpretation of some lunar properties in terms of the proposed model (i) The figure of the moon (ii) Stress history and thermal history (iii) Surface heterogeniety (iv) Emission of gases (v) Luminescence (vi) Meteorites from the moon? 6. Mercury 7. Mare 8. Venus 9. The Galilean Satellites of Jupiter Acknowledgements References Appendix

77 77 78 78 79 80 80 81

87 87 88 91 91 91 92 93 94 96 98 98 98 104

44

A. E.

RIN~WOOD

1. INTRODUCTION IT is widely believed that the stars have formed and are presently forming from initially cold interstellar clouds of dust and gas. Considering the vast size of these clouds and the scale of the processes of nucleosynthesis, it is reasonable to assume that the comparatively small segment of the cloud which was the immediate parent of the solar system was chemically homogeneous, at least, to a first approximation. This assumption is supported by the similarity of the relative abundances of most elements in ohondritic meteorites and in the sun. It is supported also by the general identity of isotopic compositions in meteorites and in the earth with the exception of cases in which isotopic ratios have been altered by subsequent specialized processes such as radioactive decay, spallation, mass fractionation and solar thermonuclear processes. Table 1. Relative abundanoea of some common elements in the sun (H = 1012) (Data mainly from GOLDBERU, M&ER and ALLER,1960) Element

Log abundance

Element

Log abundance -

H He1 C

12.00 11.16

Cl1 Al

6.30 5.60

N 0 B’

Nel Na Mg Al

Si P s

872 7.98 8.96 650 9.00 6.30 7.40 6.20 7.50 5.34 7.30

K Ca SC

4.70 6.15 2.82

Ti V Cr MIn Fe Fe2 Fe3 co

4.68 3.70 5.36 4.90 6.67 7.87 7.16 4.64 5.91

Ni 1 ALLER (1961). 2 PO?cTASCZi( 1963). 3 CLAAS (1961).

During the past 16 years geochemists led by H. C. UREY have established important chemical boundary conditions for the origin of the solar system which must be accommodated in any dynamical theory proposed on astronomical grounds. The chemical state of the parental gas-dust cloud was largely determined by its overall chemical composition as given, in part, in Table 1 and by its temperature, which was probably smaller than 100°K (DUXAY, 1967). It was demonstrated by LATIMER(1960) and UREY (19&J) that all of the common metals in a gas-dust cloud of solar composition (Table I) would occur in the form of ozicEesat temperatures below 300°K, and under equilibrium conditions. This conclusion is of the utmost importance in the case of iron. Since m&a&c iron is an important constituent of meteorites and of the earth, it appears that accretion of the primitive dust into larger bodies was accompanied or preceded by chemical reduction of iron and nickel oxides to the metallic state.

45

Chemical evolution of the terrestrialplanets

The planets fall into two distinct groups with respect to mean density, mass and distance from the sun (Table 2). The terrestrial planets are relatively close to the sun and possess densities between 3.33 and Ii.33 g/cm3 whereas the major planets are much more massive, possess densities between 2.47 and 0.71 g/cm8 and are further away from the sun. The overall differences between the two groups are reasonably well understood in terms of chemical constitution (RAMSAY, 1961, 1963; UREY, 1962,1957a; WILDT, 1961). The major planets appear to have retained more nearly the solar abundances of elements. They are composed dominantly of comparatively volatile components such as H,, CH,, NH,, H,O, Ne and presumably He. The retention of volatile components by the major planets is connected both with their large ultimate masses and with low temperatures in the outer regions of the parental solar nebula. In the inner regions of the nebula, the temperature was apparently higher so that most of these components remained largely in the gas phase and did not accrete with the dust. In consequence, these planets ultimately formed mainly from the metallic oxide component of the primitive dust. Table 2. Physical properties of the terrestrialplanets. Dstte from KUIPER (1952) with supplementary dam by AMEN (1963) and DE V~ucou~~u~s (1964) Planet Mercury Venus Esrth(l’ (Moon) M&m

MsW (relative to earth) 0.0543 0.8136 1*0000 0.0123 0.1069

Asteroids (chondritic)

<0~00013

Jupiter s&lrn Uranus Neptune Pluto

318.35 95.3 14.54 17.2 0.033 4

1 Mass of earth is 5.975

x

Radius (relative to esrth) 0.383 0*9551 1.000 0.273 0.528 SO.058 IO-97 9.03 3.72 3.38 0.45

Density (g/cm? 5.33 5.15 5.52 3.33 4.00 -3.5 l-35 0.71 l-56 2.47 21

Terre&r&l plan&a

Minor planets

Major planets

lOa g (mean radius of earth is 6371.2 km).

In the present discussion, we will consider the moon and chondritic meteorites as members of the terrestrial planet family. The densities of the members of this family, when calculated at a common low pressure (e.g. 10 kb) vary between 3.4 and 6.2 g/cmS (Table 3). This large range of densities implies that important differences in chemical composition exist between the terrestrial planets. The principal objective of this paper is to enquire into the manner in which these differences in chemical composition were established. Although the terrestrial planets were dominantly formed from the non-volatile metallic oxide components of the primitive dust, a small amount of volatile components were also incorporated. Investigations of the amounts and compositions of volatile components retained by the earth and meteorites have provided invaluable information about the nature of the accumulation process. BROWN(1962), CEAMBERLIN(1952) and UREY (1952) have clearly demonstrated that at least part of the

46

A. E. HXKGWOOD

accumulation process proceeded at low temperatures so that some of the primordial water, nitrogen, sulphur, chlorine and carbon might be trapped in chemical cornpounds of comparatively low volatility. If the planets had formed directly from materia1 of solar composition at high temperatures, these elements and compounds would have remained entirely in the gas phase as H,O, N,, H,S, HCI, CH, and CO. The occurrence of substantial quantities of these volatiles on the earth and in meteorites would then be inexplicable. Table 3. Mean densities of terrestiial planets at 10 kb assuming that they are composed of varying proportions of metal phase (pIO= 7.9) and siliaate phase (pxO= 3.3 g/om3) Planet lKercury Venus Earth Moon Mars Chondritic asteroids

Mean density

Mean density at 10kb

Per cent iron-nickel phase

5-33 5.15 5.52 3.33 4.0

5.2 3-91 4.04 3.403 3.71

63 26 31.52 5 19

2.2-3.7

2454-3.955

a-30

l Based on JEFFREYS(1937). 2 BULLEN(1940). a ‘UREY (1960). 4 Type 1 carbonaceouschondritss. 6 Enstatite chondrite.

The above inference that at least part of the ace~ulation of te~strial planets occurred at low temperatures, together with the previous conclusion that majar chemical fractionations of non-volatile components also occurred, together constitute a major boundary condition for all theories of formation of the solar system. 2. CAUSES OF DENSITY VAB~~ON

AMONG TERRESTRIAL PLANETS

JEFZREYS (1937) and UREY (1952, 1957a) have explained the varying densities of the terrestrial planets by assuming that they are composed of varying proportions of silicate (pa N 3.3 g/ems) and nickel-iron (pON 7.9 g/cm3) phases, each phase being of essentially constant composition. The low-pressure densities of planets would therefore inctrease with increasing metal/silicate ratio. Compositions of planets according to this model are given in Table 3. It is clear that the densities can be explained in principle on this basis. There are, however, severe faculties in explaining how the metal/silicate fracttionation was established. UREY (1952,1956, 1967a, 1957b, 1938,1962a, 1963) has extensively investigated this problem and has suggested several related models. Readers of URICY’Epapers will be impressed by the formidable difiiculties which he encountered in trying to reaonoile possible rne~l~~~a~ ~a~~o~at~on me~ha~ms with other irn~~nt boundary aonditions for the formation of the solar s~s~rn-pa~~~ly the reqnirement that at least part of the formation of planets occurred at low temperatures, thus permitting them to trap volatile components. The models evolved by URB~Yto meet these requirements are ingenious. However they are also extremely complex,

Chemical evolution of the terrestrial planets

47

reqniriug severs,1distinct sbges, some of which iu the author’s opinion are lacking both in supporting evidence and intrinsic plausibility. This applies p&icul&y to the physical mechanism by which differential segreg&ion of silicate and metal fragments in the solar nebula wss achieved. According to UREY’Spreferred models, this required

(1) formation

of a series of lunar sized bodies (primary objects) in the nebula accompanied or followed by partial reduction to metal phase.

(2) collisions

between primary objects leading to complete disintegration and production of a fine grained mixture of silicate and metal fragments

(31 ~fferential physical se~~gation of silicates and metal fr&gments by 8 process

dependent upon an assumed smaller grain size of silicate rendering them more susceptible to transport by radiation pressure and moving gases.

(4

Re~ccum~&~on of the resulting i~omogeneous mixture of silicate and metal fragments into planets possessing different net silicate to metal ratios.

The difficulties encountered by UREY with complex physical silicate-metal fr&ctionation me~h~~srns hsve convinced the author of the desirability of exploring alternative approaches to the problem of density varistion and chemical fractionation among terrestrial planets. An attempt is made in this paper to develop a comparatively “simple” single-stage h~othesis in which plsnets form directly by accretion from the parental solsr gas-dust cloud and the varying densities and compositions of the terrestrial planets are established mainly by chemical processes occurring during accretion. The role of oxklation-reduction equilibricz

L~TIMER(1950) and UREY (1952) investigated the oxidation state of iron in the dust phase of the primitive solar nebula. An impotent eq~lib~um is: t l?e,O, + H, = $E’e + H,O

&E&J 2

In the solar nebulrt, R is Axed by the relative ab,undances of hydrogen and oxygen (Table 1) which yield an H,O/H, ratio of 2 x 10-3. The equilibrium constant K is also related to the free energy change AG for the above reaction by the well knowu expression AC, = -RT In K With AG, d8~~~ from the~ochemical data, and K fixed by the H,O/H, ratio of the nebula the equilibrium temperature for the reaction is obtained. It is found to be about 12O’C. At temperatures below this iron would occur oxidised, as magnetite, whilst above 12O”C,magnetite would be reduced to metallic iron. In the dust phase of the cold, solar nebula, a substantial proportion of the oxidised iron occurs as a component of silicate minerals. The temperature required to reduce this iron to metal is higher than for magnetite because of the decrease in Fe0 activity caused by solid solution and compound formation. The relevant equilibria have been discussed by MUELLER(1964) who showed that reduction would occur over a range of temperatures becoming essentially complete around 86O’C 4

A. E. RINGWOO~

48

(for the solar H,O/H, ratio). Thus, in the solar nebula, rather modest changes of temperature (between 120°C and 860%) are capable of causing extreme changes in oxidation state. This conclusion, originally due to LATIMER(1950) and UREY (1952) is of fundamental importance to all considerations of the chemistry of planet formation. It is highly probable that the mean accretion temperatures of the terrestrial planets varied widely, according to the nature of the energy sources, and the local physical conditions. It is also probable that the abundances of the principal reducing agents-carbon and hydrogen, also varied widely according to the local conditions of accretion. Accordingly we would expect from these elementary considerations that substantial variations may exist in the mean oxidation states of different

c

/

41

D

Fig. 1. Showing the increase in den&y of

40 “E s 39 _./ x .G

5 D

C

30

B

3 ?-

0I

3 6O

I

1

I

20 40 60 80

Elemental 9/Toiol

1

60 SI %

primitive oxidised non-volatile material aa iron oxide and eGoa are reduced to metal. The composition of the primitive material is that of Type 1 carbonaceous chondrites on a vol&&+free and sulphur-free b&s. Densitieswerecalculatedto oorrespond to mineral assemblages stable at 10 kb.

100

Elemental Fe/ Total Fe %

terrestrial planets and meteorites. This expectation is in fact, verified by direct investigation of the redox states of meteorites and the earth (sections 3 and 4). RINGWOOD(1959) considered the effect of increasing degrees of reduction upon the density of a primitive oxidised mixture representative of the non volatile oxide component of the dust in the solar nebula. The composition of the dust phase was taken as that of oxidised chondritic metedrites. The common oxides most readily reduced are those of iron and nickel. It is important to note that the next common oxide in terms of “ease of reduction” is silica. Conditions required for the reduction of silicates to elemental silicon were discussed. The silicon appears M a component of the metal phase, with which it forms extensive solid solutions and a complete range of liquid solutions. The effect of varying degrees of reduction on the density of the oxidised primitive chondritic mixture is illustrated in Fig. 1. For convenience of comparison, the densities are
2 MgSiO, = Mg,SiO,

+ Si (metal)

+ 0,

Density increases more rapidly as silioates are reduced. At 44 per cent reduced total silicon, corresponding to 20 per cent (W/W) silicon in the metal phase and an

Chemicelevolutionof the terrestrialplanets

49

MgO/SiOa ratio of 2/l in the silicate phase, the qdeusityis 4.16 g/ems. More inteuse stages of reduction are possible. However, they require ~rn~~t~ above 16OO*C and lead to volatilization of magnesium and silicon monoxide from the residual material. These processes are discussed in section 6 and 6. We have not discussed the nature of the reduction processes. The actual mechanisms may be expected to vary widely in the different planets and will be considered in subsequent sectious. The above discussion is general and simply postulates au increasing degree of reduction (i.e. loss of oxygen) from primitive material according to approp~&~ chemioal equilibria, e.g. Fe0 + H, Fe0 + C SiO, + 2H, SiOz + 2C

= Fe + H,O =Fe+CO = Si + 2H,O = Si + 2CO

We have previously noticed that wide variations in mean oxidation states are likely to occur during formation of planets, and from Fig. 1 we see that these would cause substantial differences in their densities (reduced to a common pressure). RIXUWOOD(1969) showed that the varying “low pressure” densities of the earth, Venus and Mars could be simply explained by assuming that their mean oxidation states differed substantially. Compare Table 3 with Fig. 1. More detailed considerations given in subsequent sections support the previous conclusions. The lowdensity of the moon and the high density of Mercury can also be explained by extensions of the basic hypothesis (sections 6 and 6). It will appear that mechanical fractionation of silicate and metal phases are not required in order to explain the densities of the terrestrial planets. 3. CHONDRITIC

METEORKTES

(a) Intr~~~t~~n Chondrites are by far the most common class of meteorites arriving at the earth, aocounting for approximately 90 per cent of known falls. Detailed studies of their ghetto and mineralogy have convinced almost all workers that chondrites represent the most primitive major class of meteorites. The other classes-irons, stoney-irons and aohondrites-have clearly been subjected to much more complicated processing and physical fractionation. Chondrites are of vital significance for inve~igations beariug on the origin of the solar system sinoe they have preserved a record of the complex physical and chemical processes which occurred during its formation, approximately 4.5 billion years ago. It is widely believed that chondrites are derived from the region between Xars and Jupiter where the asteroids oocur. The asteroids vary in radius from 370 km. downwards. Many are observed to possess irregular shapes suggesting an origin by fragmentation of a larger body. &IK and SINOER(1967) have pointed out that asteroids have high mutual collision probabilities and suggested that collisious will perturb the orbits of some of these bodies so that they intersect that of the earth and fall as meteorites. Eurther studies (ABNOLD,1964 ; ANDERS,1964) indicate that the projeotion of meteorites into earthinterseoting orbits is a more complex process than envisaged by c)~rx and SINUER

A. E. RINQWOOD

50

and probably also requires a strong perturbing influence on the orbits by Mars. As an alternative to asteroidal origin, UEEY (1069) suggested that chondrites are derived from the moon. This would account for their short cosmic ray exposure ages. ANDERS(1964) has extensively discussed the evidence relating to asteroidal and lunar origins for meteorites. He concludes that the evidence strongly points towards an asteroidal origin. ANDERS’arguments are persuasive, however he does not offer a convincing explanation for the low cosmic ray exposure ages of stones. Untif this is forthco~ng, the possibility of a lunar origin for stone meteorites must remain open. ARNOI;D(1964) has provided a stimulating discussion of this possibility. (b) Chemical composition Chondritea as a class are defined by the~possessiou of a characteristic “chondritic” struct~e, one of their outsit and most e~gmatic features. The origin of these structures is controversial. It is clemj however, that they have been caused by extremely specialized physioal processes which were common to nearly all chondrites. Thus, although major differences exist in the mineralogy of different groups of chondrites, and some aspeots of their chemical compositions, their possession of the charaete~stic chondritic texture points ~ambi~o~sly to a common origin. The classi~cation of chondrites is based pr~cipally upon the work of PRIOR (1816b, 1920, 1953) with modifications by UREY and CRAIG (1953), WIIK (1966), MASON(1962 a, b) and Barr; and FREDR~SSON(1964). The classification used in this paper (Table 4) is based on the above work and is further discussed by RINGWOOD (1965a). Table 4. Cl~~~tion

of ohondriteia

Enstatite chondrites Ordinary chondrites Carbonamouechondrites

H group L group

H = high-iron L = low-iron

Type III TYP II Type I

Chemical compositions of chondrites reveal some striking similarities and dissimilarities. The abundances of the major componenfis are remarkably similar in all chondrites (Tables 5 and 6), when considered for example in relation to the compositional variances of a compamble terrestrial petrologic class such as basalts. This similarity is so characteristic that in the few cases where chondritic structures are lacking (e.g. Type 1 carbonaceous chondrites) the chemical composition clearly serves to establish genetic affinity with the chondrite class. Contrasting with the gemwal sim.ihwitiesin abundanws of major components, very large differenms exist in the abuudances of certain minor elements. It is not feasible to discuss the abundances of pilements individually. These have been summar&ed and discussed by MASON(1962b), WOOD (1963b), UREY (1964) and ANDERS (1964) ; see also LOVERINUand MORGAN(1964) for U, Th, SMALESat al. (1964) for Cs and MASON(1963b) for V. An attempt is made to summarize the

51

Chemiosl evolution of the terrestrialplanets

principal abundance trends in Tables 7,8 and 9. The average abundances of elements (relative to silicon) in the different groups of chondrites have been compared with their average abuudances in Type 1 carbonaceous chondrites in these Tables. It is reoognised that this procedure is not ideal. Frequently the abundances of a particular element in a group of chondrites may scatter widely and overlap those of another group. In other cases, an element may be fairly uniformly distributed in a particular group, but may be grossly anomalous in one or two members of this group. However, the primary objective is to define in the simplest manner the overall abundance patterns in chondrites. For this purpose, the use of group averages is both ~onve~ent and adequate. Table 5. Selected analyses of meteorites belonging to different chondrite groups (Mnsow, 1962b, p. 74) 1

2

3

4

6

6

Fe Ni co FeS SiO, TiO, %a

23.70 1.78 0.12 8.09 38.47 0.12 0.02 1.78

15.15 1*88 0.13 6.11 36.55 o-14 0.32 1.91

6.27 1.34 0.0.5 5.89 39.93 o-14 0.33 l-86

4.02 1.43 0.09 6.12 34.82 0.15 0.20 2-18

0.00 0.00 0.00 366(S)* 27.81 0.08 0.21 2.15

0.00 0.00 0.00 6.66* 21.74 0.07 0.18 1.59

Fe0 MgO cao N%O K80

0.23 21.63 1.03 0.64 0.16 trace 0.34 0.23 0.11 0.32

10.21 23.47 2.41 O-78 0.20 o-30 o-21 0.52 -

15.44 24.71 1.70 0.74 0.13 0.31 0.27 0.54 0.03

24.34 23.57 2.17 0.69 0.23 0.20 0.10 0.58 0.00 0.00 0.19

27.34 19.46 1.66 0.63 0*05 0.30 12.86 0.36 1.53 0.07 2.48

22.86 15.24 1.18 0.71 0.07 O-27 19.17 0.35 1.19 0.06 2.99 6.71?

99.89

100.29

99.67

100~08

P206

H& C%% NiO coo C

101.01

100.03

1. Enstatite chondrite (Dan& Kuil; PRIOR, 1916). 2. Ordinesy H chondrite (Oakley; WIIIE, 1950). 3. Ordinary ‘L chondrite (Kywhu; WON and WOE 1961). 4. Type III carbonaceouschondrite (W~n~n; Wxx, 1956). 5. Type II carbonaceouschondrite (~ghei; Wm, 1966). 6. Type I carbonaceouschondrite (Orgueil, Wm, 1966). * Wiik reported all S and FeS but it is given here aa S, and the correspondingFe is reported aa FeO. t Organic matter.

The outsung feature of Tables 7, 8 and 9 is the relative depletion of many elements in all groups of ehondrites compared to Type I carbonaceous chondrites. It would be possible to obtain the compositions of enstatite chondrites, ordinary chondrites and Types II and III carbonaceous chondrites solely from the removal

62

A. E.

RINGWOOD

Table 6. Analyses of meteorites belonging to different ohondritegroups (Table 6), recalcuIatedin atom peroentageson water-, carbon-, and sulphur free basis (~SON, 1962b; p. 76; Column 6 from W=, 1956)

Fe Ni CO

Si Ti Al Mn MS C& Na K P Cr

1

2

28.52 I.65 0.11 34.98 0.08 1.91 0.02 29.43 l-67 1.13 0.18 0.32

26.32 1.74 0.12 33.17 0.10 2.04 0.25 31.72 2.34 1.37 0.23 0.23 0.37

100*00

100*00

3

4

5

21% l-27 O-04 3678 0.09 2.01 0.26 33.92 I.57 1.30 0.17 0.25 0.39

25.72 1.38 0.08 32.77 0.14 242 0.16 33.03 2.19 1.25 0.27 0.16 0.43

26,18 1-41 0.06 31.85 0.09 2.90 0.19 33.19 2.04 1‘40 0.07 0.29 0.33

27.34 l-37 O-07 31.12 0.09 2.68 0.22 32.48 l-81 1.97 0.12 0.33 0.40

100~00

100*00

100~00

100~00

6

(by suibble chemical fr&o~onation processes) of the approprkte amounts of trace and minor elements from Type I carbonaceous chondrites. The abundances of man$klements, e.g. Bi, Pb, Tl, Ha, Te, I, Cd, Zn, Ge, U, Th in ordinary chondrites and-to a lesser ox&t .in e&&tit~~oho&&ite~ de&rt’widely from estimated “cosmic” abundances, derived from nuclide systematics snd theories of nuol~~~thesis (SUESSand UREY, 1956; BURBIDGE et al., 1959). On the other hand, it is extremely significant tha.t the Type I carbonaceous chondrite &bnndanes Table 7. Abundances of elements in ordinary chondrites compared to Type 11 carbonaceouschondrites

Element

Fe, Si, Mg, K, SC Rare earths, Ti, Cr, Mn, Co, Brt Na, Rb, CE Ca, Sr, Y, Al, U, Th P, V, Ta F, Cl, I S, Te Cu, Zn, Sn, Ce, Pb Bi, Tl, Hg C, H, K

Relative abuudauce Atoms y0 element in ordinary ~hond~~~Ato~

y0 Bi

Atoms ye element in Type I carbonaceouschondritesj Atoms y0 Si Not signifleantlydifferent from 1 0+X-1.0 Depletion probably not signiflcsnt 04-0~7 0.3-0.8 0.2-0.4 0.1-0.3 0.1-0.2 0.1-0.3 0~001-0~1 o~oo~o*ol

Abundances of Sn, I and Ta, and 1 Type I oarbonsceousobondriteeare Orgueil and Irma. h8ve not yet been determinedin Type I carbonaceouschondrites. Ratios quoted are based upon determinationsin other carbonaceouschondrites.

63

Chemical evolution of the terrestrialplanets Table 8. Abundances of elements in enstatita chondriteecompared to Type I carbonaceouschondrites Element

Relative abundance (aa in Table 7)

Fe, Ni, Co, Ns, K, Cd, Si, P Mg, Ca, Ba Al, Y, rare earths, SC, Cr, Ti U, Th Cl, I S, Te Sn, Cu, Zn, Ge, Tl, Pb, Bi

Not significantlydifi%rentfrom I.0 O&0*7 06-0.7 0+x-0.4 ox&-o*7 o+0.7 04--0*8 0.001 0~006-0~2

Hg C, H

Table 9. Abundances of elements in carbonaceouschondritcs-Types I, II aud III Element Na, K

ce SC,ram earthe F P S C Zn Ge Cd U Th

Relative abundsncType I

Type II

1 1 1 1 1 1 1 1 1 1 f 1

0.7 0.6 1.6 0.6 0.6 0.6 0.7 0.7 -

-

in Table 7 Type III 0.6

0.2 I.4 0.3 0.6 0.4 0.1 o-2 0.2 o-4 0.6 0.7

are in reasonable agreement with the cosmic abundances for all but the most volatile elements (REED et CL, 1960; GOLES and AXDEIGJ, 1982; So~aa~ et al., 1963; GREEEWD, 1963 ; LCWERIN~ and MOWAX, 1964 ; ANDEW, 1964). This indicates that Type I cltrbonaceous ohondrites have had a simpler chemical history than other olasses of chondrites. This is supported by their highly oxidised state and by the high content of water, carbonaceous compounds and other volatiles which they oontain. ~TXNER (1960) and UREY (1952) have pointed out that in the cold gas-dust aloud believed to be the immediste parent of the solar system, the dust particles would also be highly oxidised, volatile-rich, and composed of most elements in their primordial abundances. Accordingly UREU (1963) suggested that the carbonaceous chondrites might be closely related to the primitive dust of the solar nebula. His suggestion was developed in gre&er detail by MASON(1960) and RINCIWOOD(1961a, 1962). Although this hypothesis became the subject of a spirited controversy rend was tempomrily abandoned by UREY, the &c~urn~~~on of subsequent detailed compositional evidence appears to have plaoed it on a widely accepted basis (Lovunr~o, 1962; GREENLAND, 1663; UREY, 1964; ANDERS, 1964).

54

A. E. RINQWOO~

Although Type I carbonaceous chondrites appear to represent the closest approach to the primordial dust of the solar nebula which we possess, it is probable that important differences exist. The Type I carbonaceous chondrites have had a mild thermal and metamorphic history. Presumably the original dust was rich in condensed water, methane and ammonia. Subsequently the dust accreted into a small parent body which was heated to between 0 and 100°C (Du FRESNE and ANDERS, 1961, 1962) resulting in loss of most of the volatiles, accompanied by reconstitution and recrystallization of the fraction that was not volatile in this temperature range. The present Type I carbonaceous chondrites are not to be regarded as the direct parents of other chondrites. Their importance lies in the fact that they provide us with the closest insight into the probable physical and chemical constitution of the primitive material from which other classes of meteorites and other planets subsequently formed. (c) Oxidation-reduction

equilibria

This is a subject of crucial importance for an understanding of the origin of meteorites and the chemical evolution of the terrestrial planets. It was realized quite early (WAHL, 1910) that the relative amounts of metallic iron and oxidised iron in chondrites were widely variable, despite the fact that total iron was approximately constant. PRIOR (1916b, 1920) made a critical study of the chemistry of chondrites based largely upon his own excellent chemical analyses, and concluded “The less the amount of nickel-iron in chondritic stones the richer it is in nickel and the richer in iron are the magnesium silicates.” This generalization has become known as PRIOR’S rule, or rather “rules” since it expresses two relations, one between total metal present and the MgO/FeO ratio of the silicates and the other between total metal and its Ni/Fe ratio. The analyses in Table 5 illustrate the rule. Confusion concerning the validity of the rule has arisen because of the difficulties in chemically analysing chondrites and correctly partitioning iron between oxide, metal and sulphide. UREY and CRAIG (1953) selected 93 superior analyses of chondrites and plotted oxidised iron against reduced iron. Th.e scatter was large, and these authors expressed reservations about the validity of PRIOR’S rule. Their work disclosed the existence of two subgroups of the ordinary chondrites, characterised by a difference of five per cent in the iron content. These were called “High-iron” (H) and “Low-iron” (L) subgroups. RIN~WOOD (1961a) pointed out that nearly all the oxidised iron in chondrites occurred in solid solution in olivine and pyroxene, the compositions of which could be readily obtained by optical means. This offered a simple method to select conAccordingly he determined the compositions of olivines and sistent analyses. pyroxenes from a collection of chondrites and used these to select another group of superior analyses. These have been plotted in Fig. 2 in a similar manner to UREY and CRAIG. A great reduction in scatter due to analytical error was achieved. MASON (1962a) has extended this work and compiled a similar diagram. Figure 2 shows that there can be no doubt about the validity of PRIOR’s rule if it is taken as implying a general trend rather than a strictly quantitative relationship. As the amount of oxidised iron increases from 1 to 26 per cent, metallic iron falls from 30 to 0 per cent. A wide range in oxidation states (as defined by the

chemical evolution of the terrestrial planets

65

FeO/(E”eO + MgO) ratios of the silicates) is found. However the range is not oontinuous. E’igure 2 shows a large hiatus in oxidation state between enstatite and ordinary H ohondrites. More extensive and precise investigations by MASON(1963) and Kxn, and FR~D~SO~ (1964) indicate the presence of a small but distinct hiatus in FeO/(FeO + MgO) ratios between the H and L groups. MAS~~V (1963) has determined the compositions of olivines from 800 ohondrites by X-ray diffraction. His results emphasize the extreme bimodal distribution of FeO/(FeO + MgO) ratios in H and L ohondrites (Fig. 3).

Wt. % oxidised

iron

Fig. 2. Relationship between metallic iron and oxidised iron in chondrites covered by euperior chemical fmaly 88s. Dotted line in a line of oonstant total-iron composition (after Ring-wood 1961rt).

Mole pw cent Fe2Si04 in oiivine

Fig. 3. Distribution of olivine compositions in ordinacy ohondrites (f&s)Ater MASON,1963.

Figure 2 also supports the group concept of UREY and CRAIG (1963). The L group are clearly defined by their uniform depletion of about 5 per cent of metallic iron. The enstatite chondrites are also seen to form a discrete group in which some independent fractionation of metal has occurred. It is therefore clear as emphasized by UREY and CEUIQthat the relationship between ohondrites cannot be explained by a rigorous application of PRIOR’Srule. In addition to variable redox states at constant total iron composition, some process for independent fractionation of metal must have operated. However this complexity should not be permitted to divert attention from the importance of Prior’s rule viewed as a general trend rather than as a rigorous ~lations~p. KEIL and FREDRIICSSON (1964) have investigated the compositions of olivines and pyroxenes from 90 chondrites with greater precision than previously attained. They conclude that PRIOR’Srule is best used to describe and interpret redox relations between groups of chondrites rather than within individual groups. They make a strong case for this view when applied to the ordinary H chondrites. The metal contents within this group are not related closely to the FeO/(FeO + MgO) ratios

A. E. RINGWOOD

56

of the silicates. It appears that the metal content of this group has become slightly fractionated as with the enstatite chondrites. The situation with regard to the L group is not so clear. There has evidently been an analogous slight independent fractionation of metal for FeO/(FeO + MgO) (olivine) ratios between O-21and O-26. However, a distinct correlation of metal content with FeO/(FeO + MgO) (olivine) ratio is present over the entire O-21 to O-29 interval. KEIL and FBEDRIKSS~N avoid this by establishing a new group with 6 members. This may be premature since this group could also be interpreted as a low density “tail” belonging to the L group as suggested by the more extensive data of MASON(1963) in Fig. 3. Further work should resolve the issue. SUESS(1964) and CRAW ( 1964) have also discussed the significance of KEIL and FREDRIKSSON’S results with regard to PRIOR’Srule. Their conclusions are essentially similar to those of KEIL and FREDRIHSSON, i.e. that PRIOR’Srule should be regarded as a qualitative relationship between several discrete groups of chondrites. SUESSand CRAWdo not, however, appear to appreciate the signifioance of the substantial spread in FeO/(FeO + MgO) ratios of olivines and pyroxene within each group as is shown by the measurements of RINGWOOD(1961a), MASON (1963) and KEIL and FREDRIXSSON(1964). These demonstrate clearly that each group is not characterised by a uniform oxidation state. Figure 1 p. 406, of Craig’s paper shows variations in FeO/(FeO + MgO) ratios within the H, L and LL ordinary chondrite groups which are smaZkerthan the intervals separating the groups. This is in dire& conflict with the experimental measurements referred to above. That part of CRAIG’Sdiscussion which is based upon this diagram is clearly invalid. (d) Nature

of ducti

process

From evidence given previously (LATIMER,1950 ; UREY, 1962) it was concluded that the parental material of the chondrites was highly oxidised. It follows from the relationships discribed by PRIOR and discussed in the previous se&ion, that the oxidised primitive material was subjected to widely varying degrees of chemical reduction. The two most abundant reducing agents are hydrogen and carbon, and it is of vital importance for theories of the origin of the solar system to establish which of these dominated in the chondrite reduction process. The phase assemblages of chondrites provide strong evidence that the metal was produced by a carbon reduction process operating in a condensed environment rather than by a hydrogen reduction prooess in a dispersed system. In ordinary L chondrites containing a substantial amount of oxidised iron, all the silicon, calcium and chromium and most of the phosphorus occur as oxides. As reduction becomes more intense, however, as in ordinary H chondrites, an appreciable amount of phosphorus oocurs as schreibersite (Fe, Ni),P. At the highest stage of reduction, in enstatite chondrites, chromium ooours as daubreelite (FeCr&,), calcium as oldhamite (CaS), phosphorus as schreibersite (Fe, Ni&P, whilst some silicates are reduced to elemental silicon which enters into solid solution in iron. Other minerals occurring in enstatite chondrites are graphite, troiiite and osbornite (TiN). All of these minerals are characteristic of blast furnace assemblages produced when iron ores are strongly reduoed by carbon. The common occurrence of this group of minerals in both environments strongly suggests that the chemical processes which produced meteoritic iron and blast furnace iron have been similar, i e. that the

Chemiod evolution of the &r&rid

plm&

67

iron in meteorites has been produced by reduction in a condensed system in the presence of carbon. Gn the other hand it is very difficult to understand the formation of this mineral assemblage by a procese of hydrogen reduction. Graphite, which is e ailment constituent of most enstatite ohondrites and & minor constituent of many ordin~ cbondrites (RAMDOER,1863) is unstable at elevated temperatures in the presence of excess hydrogen and/or oxygen. With the solar H-O-C abundances, carbon would ocour a,s CO or CH, dependent upon pressure. There is no apparent way in which gmphite *could be formed in a hydrogen reduction process operating at temperatures high enough ( l10~1400°C) to produce the required reduction in ordinary and enstatite chondrites. Enstatite ohondrites are characterised by the presence of elemental silicon in solid solution in the metal phase (RIN~WOOD,1961b). The average Si content in the metal of St. Marls is 6 atomic per cent. Furthermore they are characterkd by very low coning of Fe0 in the ~~c~~s-usually less than 0.1 per cent according to IE'RICDRXKSSON (personal communication). These observations establish that euatatite cthondritesformed under extremely reducing conditions. If they formed by a process of hydrogen reduction or by oondensation from a hydrogen-rich gas phase, a very high hydrogen-to-oxygen ratio is required. WOOD(1963a) has discussed the redox conditions required to form the pyroxenes in the Renazzo ohondrite which contain 1 to 2 weight per cent iron. He found that H/O ratios between 200 and 1000 were needed between 1600 and 21OO’C. The iron content of the dominant pyroxene in enstatite chondrites and achondrites is more than an order of magnitude below the Renazzo values (FREDRXSSON,personal communication). Accordingly, an H/O ratio of at least 1000 would be required according to Woo~‘s calculations, This is the H/O ratio of the solar &tmosphere (GOLDBERG et a&, 1960). The author has carried out analogous calculations to fmd the H/O ratio needed to reduce pure liquid silica to an alloy of the composition Fe,.,&&,,, at 1500°C, using the thermodynamic data of KUBACHEWSILI snd EVANS (1968) and CHIPMANet al. (1954). Au H/O ratio of approximately 300 is required, This is increased if allowance is made for decrease in the activity of SiOz in a silicate melt of the ~mpoaition of an enstatite chondrite. From these caloulations, it is clear, ensrecognized by Sn~ss (1063) that if enstatite chondrites formed near the liquidus (as shown by the chondrules) in a hydrogen-rich environment, then that environment must be close to the unfractionated solar H/O ratio of approximately 1000. It is extremely difficult to understand the occurrence of sulphur, zinc, oadmium and carbon (previously considered) in enstatite chon~~s if they formed under these conditions. Enstatite chondrites contain on the average about 3 per cent of sulphur occurring chiefly aa FeS. At 1500% the equilibrium FeS + H, = Fe + H,S proceeds to the right when the H/S ratio exceeds about ‘70. The solar H/S ratio is 60,000 (GOLDBERG et at.,1964),which is 700 times higher than will permit the st&b~ty of FeS or any other sulphide phase with equivalent aulphur activity. Permissible decreases in aulphur activity due to solution in either silk&e melt or metal will not significautly change this basic result. The presence of abundant sulphur in enstatite chondrites is

.a

A. E. RINC.W~OD

inexplicable in terms of a hy~ogen-~du~tion process. On the other hand, it is completely consistent with a carbon-reduction process in a condensed systemHATCHand (IJHYLPMAN(1949). Analogous arguments apply to zinc and cadmium which are present in some enstatite chondrites in the same propo~ions as in Type I carbo~a~~us chondrites, i.e. in their “cosmic” abundances (GREENLAND,1963; SCHMXTF et al., 1964). It is impossible to condense or retain these volatile elements in the presence of solar H-O-S abundances at temperatures above 1000°C. On the other hand, in a condensed system, in the presence of carbon and an appreciable pressure of sulphur, but at low hydrogen pressure, these elements are stable at high ~mperatur~ as sulphides. It is concluded that the mineral assemblage and chemical composition of enstatite chondrites cannot be accounted for either by reduction of primitive material by hydrogen or by eonde~a~o~ from a hydrogen-rich gas phase. On the other hand, formation by carbon-reduction of primitive oxidised material in a condensed system at high temperatures is consistent with the observations. Further direct observational evidence for the production of nickel-iron and silicon-iron in the Kaba Type III carbonaceous ohondrite and the Grady ordinary &or&rite by carbonreduction has been described by SZTROKAYet a$. (1961) and ~A~~o~R (1963). Finally it should be mentioned that although the preceding arguments point toward carbon as the essential reducing agent, they do not exclude the role of some accompanying hydrogen in the reduction process. In fact it is probable that the reducing agent originally consisted of complex highly polymerized hy~ooarbons, rather than graphite. (e) Origin The evidence discussed in previous sections can be explained by assuming that ohon~ites evolved on small parent bodies which formed by direct accretion from the primitive oxidised dust in the solar nebula. At the low temperatures of the dust cloud, various volatile components, particularly water, ammonia, hydrogen sulphide and methane hydrate would condense as solids on the nuclei of metallic oxides. It is suggested that high energy radiation from radioactive elements present in the oxide nuclei caused loss of hydrogen and polyme~ation of the above simple compounds, leading to the formation of complex carbonaceous compounds of lower volatility (RIN~WOOD,1989). These compounds remained with the non-volatile oxide dust during accretion. It is also possible that graphite was formed directly in the cold solar nebula under eq~ibrium monitions (SUESS,1902) and was incorpora~d in the dust grains. RINUWOOD(1969, 1960, 1961a, 1962, 196Saf discussed the evolution of chondrites and terrestrial planets in terms of the differential chemical autoreduction of such primitive material when it was subjected to high temperature in parent bodies. Under these conditions the carbon reacted with oxidised iron to produce a metal phase in s%s, and in eq~b~~ with the surrounding silicates. In the case of ordinary chondrites, reduction proceeded until almost all of the hydrocarbons had been consumed, The varying oxidation states of the ordinary chondrites are thus explained in terms of varying initial quantities of hydrocarbons trapped in the acereting dust. In the case of enstatite ohondrites, ~~bonaceoua compounds were in

chemioa;ievolution of the t6%TtNtrial planeti

69

excess. Accordingly reduction of iron to metal was complete and reduction proceeded sufllciently far to produce some elemental silicon. The enstatite chondrites contain graphite in accordance with this model. M&ON (1960, 1962b) has advoctated a similar model for the origin of chondrites. The broader physical conditions under which the reduction processes operated are not well known and are the subject of active current debate. RINGWOOD (1965a) considers that the parent bodies were intermediate in size between asteroidal and lunar, and that these were heated internally by short lived radioactivities as suggested by UREY (1956) and Frs~ et al. (1960). During the slow internal heating, volatile components-mai~y H,O, H,S and CO, amounting to about 30 per cent of the initial mass were expelled from the interior in the form of dense, supercritical fluids, which functioned as powerful selective solvents for many minor elementsparticularly the chalcophile group. Extensive chemical fractionations of these elements were thereby caused. Reduction of metal phase was essentially completed under sub-solidus con~tions. With further heating melting occurred in the parent bodies. Rising masses of magma were disrupted by the internal pressure of dissolved gases (mainly CO and H,) near the surface giving rise to a form of volcanism, and resulting in formation of the characteristic chondritic textures. A limited amount of fractionation of metallic iron with respect to the silicate liquids also occurred. As an alternative to the model of internal heating by extinct radioactivities, it is conceivable that collisions of parent bodies of carbonaceous chondrite ~m~sition with the moon or with a large primitive asteroid may have formed a suitable environment in which autoreduction and chondrule formation occurred (RINQWOOD,1965d). Such events could satisfy the chemical requirement that reduction occurred in a relatively condensed environment and that carbon was the principal reducing agent (section 3a). It would also avoid the necessity for assuming the existence of short lived radioactivities as heat sources. The importance of collision phenomena in explaining aspects of meteorite genesis has been repeatedly urged by UREY (1957a, 1963). See also FREDRIKSSON (1963), and section 5d. (f) Broader signi$cance of chondrite ewohtion Regardless of detailed physical models for their origin, the chern~t~ of chond&es provides vital information of broad significance for theories of formation of the terrestrial planets. (1) Type 1 oarbonaceous ohondrites are the most primitive group of meteorites. Their composition suggests that they were formed by accretion of the dust phase of the solar nebula into a small parent body which was subjected to a very mild degree of metamorp~sm. Type I carbonaceous chondrites appear to have retained in most cases the primordial or “cosmic” abundances of elements. Their composition may therefore be used as the initial composition of the parental material from which other terrestrial planets ultimately formed. (2) The metal phase of chondrites was probably formed by anautoreduction process caused by the heating of primitive material similar to Type I carbona~o~ chondrites in a condensed environment. The principal reducing agent was carbon. This immediately suggests the possibility that other terrestrial planets formed in an analogous manner.

60

A. E.

&NGWOOD

(3) Varying degrees of reduction in chondrites were apparently caused by the incorporation of varying amounts of carbonaceous materid in parent bodies during accretion. The ratios of carbonaceous material to silicates and oxides may also have varied in other planets, contributing to differences in redox states. (4) In the enstatite chondrites, silicates have been reduced to silicon which has become a component of the metal phase. This enhances the plausibility of the hypothesis (sections 2, 4) that an analogous process has occurred in other planets. (5) Chondrites are almost certainly derived from a common region in the solar system, yet they display an extremely broad spectrum of oxidation states. This indicates that redox conditions during the formation of the soler system varied widely on a local scale. It therefore appears probable that comparable variations in redox conditions occurred in regions of the solar system where the other terrestrial planets accumulated. This supports a principal contention of this paper (section 2) that substantial differences in mean oxidation states probably exist between the terrestrial planets. It was also demonstrated in section 2 that the differing lowpressure densities of the earth, Mars, and Venus might be interpreted in terms of a common material displaying different mean oxidation states. 4. THE EARTH* (a) Introdwson

In the previous section it wss concluded that ohondrites formed by an autoreduction process operating upon primitive dust similar in composition to the Type I carbonaceous ohondrites. In this section we will explore the hypothesis that the earth formed by direct accretion from aim&r material. According to the model, the primitive material was subjected to sutoreduction at high meen temperatures, resulting in formation of & metal phase and loss of components whioh were vol&ile under these conditions. It follows that similarities in chemical composition between the earth and primitive material are likely to be greatest for elements which possess the lowest volatilities under high-temperature reducing conditions and least for those elements which are readily volatile under these conditions. A classification of elements according to their relative volatilities from silicate melts around 1600°C is given in Table 10. It is suggested as a working hypothesis that the abundances of non-volatile elements (column 1) are similar in Type I carbonaceous chondrites and in the earth. The relative abundsnces of the voletile elements (column 2) may, however, differ considerably. From Table 10 we see that the nonvolatile elements are mostly comprised of oxyphile and siderophile elements. The volatile elements include msny chalcophile elements, alkali metals, and non-metals. The validity and implications of the above hypothesis will be explored in detail in subsequent sections. (b) Mantle-core relationship The two major divisions of the earth are the mantle and core. According to the Bullen Model A, the mantle ( + crust) contains 69 per cent of the earth’s mass and the * Much of this section haa been condensedfrom a longer psper by the author “Composition and origin of the earth” published in Advancea ir, Earth Sciemm (ed. P. M. Hurley). MM’ Press, Boston (1965).

61

Chemical evolution of the terrestrialplsnets

core contains 31 per cent. We must next enquire whether the primordial abundances of “non volatile” elements (Table 10) as obtained from the composition of Type I carbonaceous chondrites are capable of yielding an earth model with the correct core-to-mantle ratio, and acceptable compositions for both these major phases. This has been examined by RIX~WOOD(1968, 1959, L965b), MV&CDOXALD and KNoPon (1958) and MACDONALD (1969) who showed that a self consistent earth model formed by reduction of ohondritic material necessitated the presence of 10 to 20 per cent of elemental silicon in the earth’s core (Table 11). Table 10. Clsssifiostion of some elements according to their relative volatilities from basic silicate melts under high-temperature, reducing oonditions II VOl&t& group

I Non-volatile group (a) Oxyphile elements Be, B, Mg, Al, Si, I?, Ce, Sc, Ti, Sr, Y, Zr, Nb, Bs, Rsre earths, Hf Ta, Th, U

IL C, N F, Cl, Br, I

Probable volatile species H&l, CO, N* halides

s, Se

hydrides

Li( P),Ns, K, Rb, Cs

elements

Zn, Cd, Hg, Tl, Pb, As, Sb, Bi, Te

elements

(6) Siderophatet&mew% Fe, Co, Ni V, Cr, Mn Cue, Ag”, Au+ MO, Sn(?), W Ru(?), Rh, Pd, Re Os( ?), Ir, Pt

Ga, Ge, Sn, In

suboxides, sulphides

* These elements sre siderophilein the absence of sulphur.

Studies of the physical properties of the earth’s core by BIRCH (1952) have indicated that it is about 10 to 20 per cent less dense than nickel-iron and that its seismic velocity is substantially greater than that of nickel-iron under comparable P, T con~tions. These concl~io~ have been confirmed by recent i~vest~atio~ on the densities and seismic velocities of metals at extreme pressure using shock wave techniques (KYNOPOFF and M&DONALD, 1960; BIRCH, 1961, 1963). It therefore appears that the earth’s core contains a substantial amount of an element with a low density which can also increase the elastic ratio and seismic velocity of iron. Limitations upon possible choices are that this element must be reasonably abundant, miscible with liquid iron, and possess chemical properties which would allow it to enter the core. Elements which might be considered as candidates are H, He, C, 0, N, Mg, Si and S, We may reject H, He, C, 0 and N since they are known to form interstitial solid solutions with iron. Additions of these elements do not signifloantly decrease the density of iron since they occupy holes aheady present in the lattice. McNeil is unlikely to be present in substantial amounts since it has a much greater affinity for oxygen than has silicon. Accordingly any chemical conditions which may have led toward the incorporation of magnesium in the core would inevitably have caused the incorporation of much larger amounts of silicon. This leaves us with silicon and sulphur as possibilities. It would require about

62

A. E.

RINQWOOD

16 weight per cent of sulphur in the core to decrease the density by 10 per oent. This would imply that the earth captured almost all of the sulphur originally present in the primitive material. RIN~WOOD(1966b) discussed the occurrence of sulphur in the earth, particularly with respect to its partition between mantle and core. Strong arguments were advanced for concluding that the core does not contain more than a small fraction of the sulphur required to explain the discrepancies in density and seismic velocity. A process of elimination thus points toward silicon as the most likely extra component of the earth’s core. Furthermore, the presence of substantial quantities of silicon in the metal phase of enstatite chondrites shows that chemical conditions during the formation of the solar system were favourable for the reduction of silicates, at least in certain regions. We have seen that an earth model directly constructed from abundances of non-volatile elements in Type I carbonaceous chondrites requires the presence of silicon in the core. Moreover this requirement is supported by independent geophysical and geochemical evidence. The model thus far demonstrates a satisfactory internal consistency. The chemical composition of the earth model obtained by reduction of Type 1 carbonaceous chondrites is given in Table 11, column 2. Table 11. Composition of mrmtle and core as derived by reduction from composition of Type 1 carbonaceous chondrites so as to obtain correct FeO/MgO ratio for mantle and core-to-mantle ratio for earth Type 1 carbonaceous chondrite I SiO, MgO Fe0 Also, cao NR,O NiO

Derived earth model 11

33.32 23.50 35.47 2.41 2.30 1.10 1.90

29.84 26.29 6.38 269 2.57 1.23

100~00

69.00

Mantle

--i

25.87 1.66 Core 3.47I

Fe Ni Si

31.00 (c) Compositiolt and d$ferentiahn

of the mantle*

Inferences concerning the composition of the upper mantle may be drawn from a synthesis of geophysical, geological and geochemical data. The subject has recently been reviewed by CLBR~ and RIN~WOOD(1964), and RINGWOOD(1966b, c). It is believed that the primitive composition of the upper mantle lies between those of peridot&e and basalt, and closer to peridotite. A broadly self consistent model for the upper mantle can be constructed on the assumption that the primitive composition is equivalent to a mixture of approximately 3 parts peridotite to 1 part

* The reader is referred to a more extended discussion of this topic (Ringwood, 1965b) for specific points which are covered only briefly in the present paper.

Chemical evolution of the terrestrialplanets

63

basalt (refs. above). This hypothetical parental upper-mantle rock has been called pyrolite. Fractional melting of pyrolite yields basalt magma, and leaves behind a refracttory residue of dunite or peridotite. There are strong reasons for believing that the continental crust has evolved by fractional melting and differentiation processes from the upper mantle over geologic time (RUBEY, 1951, 1955; BULLARD, 1952; WILSON, 1954; ENOEL, 1963). This implies that the mantle beneath continents is depleted in low melting components and easily fractionated elements, and is probably similar in compo~tion to dun&es and alpine ~~doti~s. According to the model, the zone of residual refractory peridotites grades downwards into parental pyrolite beneath continents. Beneath oceanic areas it appears that the upper mantle has not been subjected to substantial fractionation, and accordingly the primitive pyrolite probably extends upwards to the Mohorovi&? Discontinuity. According to the model, the mean chemical composition of the entire crust-mantle system, when averaged down to depths of several hundred kiolmetres, is approximately the same for continental and oceanic regions, and is given by the pyrolite oom~sition. We will now proceed to explore the consequences of this model with regard to ~fferentiation of the mantle. (i) Fractionation of non-volatile oxyphik elements of Table 10. The major element composition of pyrolite as derived by RINC+WOOD (1965b) is given in Table 12, Table 12. Comparison of pyrolite composition with composition for mantle obtained by reduction of Type I caxbonaoeouschondrite (from Table 11) I

II Maaxtlecomposition derived from Pyrolite carbonaceousehondrite(Table 11, column 2) SiO, AhO, F%O, Fe0 TiO, C%O, C&O WO N&SO KS0 MIIO coo NiO P,O,

45.16 3.54 0.46 8-04I 0.71 0.43 3-08 37.47 O-67 0.13 0.14 0.01 o-20 0.06

43.25 3.90

100~00

100~00

9.25

3.72 38.10 1.78

column 1. This is compared in column 2 with the model mantle Gorn~~tion derived by reduction from the primitive carbonaceous ohondrite composition from Table 11, column 2. It is seen that the two model compositions closely resemble each other. This close match is caused partly by choice of the 3: 1 peridotite/basalt mixture for the composition of pyrolite. Nevertheless there were independent reasons for this choice. Had ratios of 2 : 1 or 4: 1 been chosen the similarity would still have been strong. 5

64

A. E. RINUWOOD

The close correspondence between the model pyrolite composition for the upper mantle with the composition for the entire mantle obtained from chondrites suggests that fractionation of major rock forming elements in the mantle has been relatively small. At first sight, this would not be expected if the mantle had crystallized slowly from a completely or largely molten state. Indeed petrologists have frequently commented that there is much less sialic rock near the surface than might be expected if the whole earth had melted and differentiated. The high MgO/FeO ratios of minerals from the upper mantle combined with the abundance of NiO and Cr,O, (which are characteristically removed at a very early stage from crystallizing basic magmas) likewise point to a lack of strong fractionation of major elements and of trace elements which are able to substitute for major elements in the principal rock forming minerals. In contrast to the above elements there is strong evidence that another group of non-volatile trace elements-U, Th, Ta, Ba, Sr, Zr, Hf, Be, and the rare earths have been strongly concentrated in the upper mantle-crust system (BIRCH, 1958 ; GAST, 1960; RINOWOOD, 1960; TAYLOR, 1964a, b). The abundances of these elements in the upper mantle-crust system require that some process has caused their almost complete removal from the deep mantle accompanied by strong upward concentration. These elements are all characterised by ionic radii and/or charges which do not permit ready entry into the principal phases of the mantle-their upward enrichment is clearly a consequence of these properties. The only physically reasonable explanation known to the author which is capable of explaining the strong upward differentiation of these elements is a process involving crystal-liquid equilibria. This explanation implies that all or most of the mantle has been subjected to complete or partial melting and that the trace elements which could not enter the major crystalline phases (incompatible elements) became strongly concentrated in residual melts, which in turn, migrated to the upper levels of the mantle. This immediately raises the problem of how it was that the incompatible trace elements were so strongly fractionated whereas the major rock forming elements and ions with charges and sizes close to them display a negligible degree of fractionation. Two explanations appear possible During primary crystallization, the (i) The mantle was completely melted. crystal mush was subjected to a few convective overturns. This would mix up and homogenize the major phases, whilst the small quantity of residual liquid containing ions unable to enter major phases would be successively squeezed upwards, leading to strong concentration of incompatible trace elements near the surface (WAGER, 1958; RINGWOOD 1960, 1965b). (ii) Zone melting as suggested by HARRIS (1957) and further developed by VINO~RADOV (1961). As described by its advocates, zone melting is capable of causing extreme fractionation of incompatible elements whilst not affecting the composition of the major phases. Another property is that although the entire mantle must pass through a fractional melting stage, only a relatively small fraction need be molten at any one time. For zone melting to be effective throughout the entire mantle the process must begin at the coremantle boundary. Partial melting commences here, the liquid thus formed

66

Chemioal evolution of the terrestrial planets

moves upwards, solid phases dissolve at the top and precipitate at the bottom whilst incompatible ions are strongly concentrated in the liquid. (ii) Fractionation of non-volatile siderophile elements (Table 13, column 1). In Table 13 (RINGWOOD 196sb), abundances (relative to silicon) of some siderophile elements estimated to be present in pyrolite are compared with their primo~al abundances. It is seen that all of these elements are strongly depleted in the upper mantle (pyrolite) compared to their primordial abundances. It may confidently be assumed that they have entered the metal phase and have been carried into the core. It is somewhat surprising however that the depletions of these elements are not stronger than observed. We will return to this point in section 4d. Table 13.

Comparison of abundances of some siderophib elements in pyrolite with their chondritic abundanoes assumed to be primordial (after RINQWOOD 1966b)

Element

Fe co Ni CU*

l?t Au

Chondritic abundance (per 10s Si atoms)

Pyrolite abundance (per lo6 Si atoms)

8-99 x 106 2300 46,000 490 0.9 0.13

1.79 x 106 280 3670 64 0.01 0.004

Ratio Pyrolite abnndanoe Chondrite abundance 0.20 o-12 0.13 0.08 0.01 o-03

* Cu behaves &Ba siderophile element in the absence of sulphur. (ii;) Fmctimatim of voyage eking in the ear& In section 4rt, Table 10 the elements were somewhat arbitrarily dividedinto two groups-nonvolatile and volatile. It was suggested that the earth probably retained the primordial abundances of the former, but not necessarily of the latter, which may be depleted. The concentrations of volatile elements in the upper mantle and crust will, therefore, be mainly controlled by two faotors-(I) the proportions which were retained by the earth; (2) the behaviour of the elements during the inferred major differentiation of incompatible eIements discussed in section 4c(i). For those volatile elements which are mainly present in the mantle, we may expect that fractionation was governed by ionic radii and charges in a manner similar to the non volatile elements. Strong upward enrichment would be predicted for all ions with radii greater than 1-OA. On the other hand, monovalent and divalent ions possessing radii smaller than 1.0 A would not be expected to fractionate strongly upwards since they are capable of substituting for the Common elements in the p~cipal phases of the mantle, The relative abundances of volatile elements in pyrolite (i.e. in the upper mantle according to the present model) and in Type I carbonaceous chondrites are given in Table 14. The table is incomplete owing to lack of data for many elements. (Table and discussion from Rrxawoor, 1965b ; see also GAST, 1960.) The abundance data from which Table 14 w&s constructed are in many eases sparse and of poor quality. Accordingly it would not be prudent to base a case on any one element. The significant aspect of Table 14 is that it clearly reveals a general pattern, which indicates that the earth is strongly depleted relative to the primordial abundances in many volatile elements pressing widely varying chemical properties.

A. E.

66

RINGWOOD

Furthermore the indicated depletions am in many cases of such a large magnitude that the basic conclusions will not be afIected by substantial modi~cationa in abundance estimates. Table 14. Estimated relative abundances of some volatile elements in pyrolite compared to the primordial abundanoes as given by the Type I carbonaceous chondrites (RINGWOOD,19&b) Element &-a, K, Rb, Csz Zn Cd W Ga 7’1 Pb Cl s 1 Relative abundance =

Relative abundances1 0.3

0.14 0.04 0.003 0.013 O-4 0.8 0.14 0.003 Pyrolite abundance Garb. Chondrite I abundance.

2 Relative abundance for a&&li metals refer to Mean abnndanoe in entire mantle. Garb. chondrite I rsbundance

For elements such as Hg, Tl, Pb, Cl and S, the actual depletions are probably much greater than indicated in Table 14. This is because the ionio radii of these elements are too largeto permit them to enter the minerals of the mantle. Hence they are likely to be as strongly concentrated in the upper mantle a&for example, barium and strontium. Their apparent abundances as given in Table 14 are thus considerably enhanced. It is probable that Tl and Pb have been lost from the earth to a degree comparable with Zn and Cd. An alternative means of eo~de~ng the behaviour of volatile elements is through their crustal abundances. It is believed that the principal factor governing the fractionation and distribution of elements in the mantle and crust are of a crystai chemical nature-principally ionic radius, ionic charge, and bond type. In Table 16, we have compared the abundances of pairs of volatile and non-volatile cations in the crust. The ionic radii of each pair are similar, so that they are able to substitute for each other in minerals, and a~~ornp~y each other in c~s~l-~~~ fractionation prooesses . Moreover the pairs chosen are such that the volatile element has a stronger tendency to concentrate in low-melting fractions than the corresponding non volatile element (beaause of lower ionic charge and/or higher electronegativity-RINQWOOD, 195G). Referring to Table 16, we see that in every case, the volatile element is depleted in the crust, compared to its non-volatile companion, and in most cases the depletion are of a large mag~tude. ~tal~he~~l d.iEerencesbetween the members of pairs should have caused fractionation in the opposite direction. Again it appears that the earth has suffered depletion in a variety of volatile elements, Note that the depletions of Pb and Tl are better reflected in Table 16 than in Table 14.

67

Chemical evolution of the terrestrialplanets

Scnxr~ et al. (1963) have also drawn attention to the deficiencies of Cd, Zn, Bi and Tl in the crust compared to p~mor~al ab~dances. It may be noted that several of the deficient elements are chaloophile. RINGWOOD (196bb) showed that this property was not responsible for the depletions. (&) ~~~~~~~~ of te~$es~~~ract~~~~~ beets_ In section 4a a basic h~othe~s was stated, Elements were classified according to their volatility from basic siIicate melts at temperatures in the vicinity of 1500°C and under reducing conditions. It was suggested that the earth retained the primordial abundances of dements which Table 15. Relative crustal abundances of pairs of volatile (V) and non-volatile (NV) elements possessing similar ionic radii, compared to relative primordialsabundances of the s&mepairs.

Cation

Radius3 (A)

BY BaN

1.33 1-34

Volatile (V) Non-volatile (NV)

O-23

Rb+ Baa+

1.47 l-34

V NV

0.23

cs+ Bas+

167 1.34

V NV

0,097

Tl+ Ba2+

1.47 1.34

V NV

O-019

Zns+ Fea* 4

0.74 074

V NV

0.17

Cd2+ C&2+

0.97 0.99

V NV

0.054

Hg2f Sr2+

1.10 I.12

V NV

0.0005

Pb” Ba2+

I.20 1.34

V NV

0.057

Ins+ Yb”

0.81 O&3

V NV

0.097

Bi*

0.96

Gd3+

0.97

V NV

o-077

Ge” Si4+

0.53 0.42

V NV

0~016~

Volatility

(&)&

+ ( &)*ti~,r&,

l Crustal abundances from TAYLOR(1964c). s Primordial abundances obtained from abundauces in Type I ~bon~eo~ chondrites (UREY, 1964). Primordial abundance of indium was obtained from Tables of SASS and UREY (1956) and CAMERON(1959). s Ionic radii from AERENS(1952). * Iron abundance in the mantle as obtained from Table I1 was used in calculations. 6 Depletion of germanium can be partly attributed to its siderophile properties. However germtium is less siderophilethan iron, cobalt and nickel, and from Table 13 we see that the depletion due to incorporationin core is unlikely to be smaller than 0.15, which is an ordex of magnitude higher than the observed depletion.

68

A. E. RIN~WOOD

were not volatile under these conditions, whereas it may well have lost varying proportions of the volatile elements. Subsequent discussion of the abundances of many elements supported this hypothesis. It was found that a self-consistent earth model could be constructed from the primordial abundances of non-volatile elements (as given by the type I carbonaceous chondrites). On the other hand, the earth appears not to have retained a large proportion of the primordial abundances of many volatile elements of widely varying chemical properties-the alkali metals, Zn, Cd, Hg, Ge, Pb, Tl, Bi, In, Cl and S. It appears probable that this list may be extended when sufficient data become available for many other elements, e.g. As, Sb, Br, I, Se, Te, Ga. From the strontium (GAST,1960) and lead isotopes data it is clear that the loss of these elements from terrestrial matter occurred at or before approximately 4.5 billion years ago. It is important to notice that loss has not been complete, and that a substantial quantity of these volatile elements has nevertheless been retained. It has previously been inferred that the original state of the solar nebula was cold and that the solid phase (dust particles) was highly oxidised. In order to form the metal phase now in the earth’s core, heating and partial reduction of the primitive material occurred. It seems reasonable to assume that loss of the volatile elements discussed above occurred at this stage, and that the heating and reduction must have been strong enough to cause this loss. The construction of a self consistent earth model from the primordial abundances of non-volatile elements requires a strong enrichment of U, Th, Ba, Sr, Ta, Zr, Hf, B, Be and the rare earths in the upper mantle and crust. It was concluded that their upward concentration implied that all or most of the mantle had been subjected to a partial or complete melting process at some stage. The formation of the earth’s core also involves s, major differentiation process. Formation of the core is most readily understood if the mentle passed through a molten or partially molten stage as suggested above. Indeed we will find that if the mantle were not already molten, the gravitational energy liberated by formation of the core would cause it to become so. In this case, differentiation of the mantle and formation of the core were part of a single decisive high temperature event in the earth’s history. It is tempting to identify this decisive event with the high temperature process which caused reduction of primordial material to metal and loss of volatile elements as previously discussed. In this case all these processes occurred during the earth’s formation, about 4.5 billion years ago. (d) Disequilibrium

between core and mantle

Construction of a self-consistent earth model from primordial abundances of non-volatile elements required the presence of some elements,1silicon in the earth’s core. This requirement was supported by direct geophysical evidence. The presence of silicon in the core whereas the mantle oontains substanti&l quantities of oxidised iron, implies that the mantle is not in chemical equilibrium with the core. Several additional lines of evidence support this implication. Consider a hypothetical earth model in which the metal now in the core was originally uniformly distributed in small fragments throughout the mantle. This corresponds to an earth composed of material similar to ordinary ohondrites. The

Chamicd evolution of the tame-&rid pk%mts

69

co-existence of metalliiciron in equilibrium with ferromagnesian silicates establishes the redox ~o~~tio~ in the mantle. queen (1964) pinked out that a gas phase in equilibrium with this assemblage at high temperature would contajn II, greatly in excess of [email protected], and CO grea&lyin excess of CO,. Now the volatile components which have been liberated by degassing of the mantle over geologic time are well known to be composed dominantly of II&l and 60% rather than H, and CO. RUBEY (1951, 1955) estimated that the H,O/H, (rnole~~&r) ratio of degassed volatilea in the earth was of the order of 100, HOLLAXD(1963) pointed out that the gases in eq~b~~ with an average Haw&isn volcanic glass yielded 137 for H,OjH, and 31 for CO&O. It is therefore clear that the volatiles degassed from the e&h were not derived from any part of the mantle containing free iron in equilibrium with silicates. (This is a situation which appears to have prevailed throughout geologic time, and constitutes a serious objection to RUINCIORN’S (1962, 1964) theories.) Upon separation of the metal phase of an or~na~ ohondritic earth into 8 core, the redox s&&eof the mantle was mGnttined by the buEer system Fe”/Fes+. This was initially es~blished by ~q~libri~ with metallic iron before the metal ~p~r&~d. Bowx~ and SCXMRER(1935) observed .Fe~+fFe3fratios of about 70in FeO-MgO-SiOa liquids possessing relevant compositions and in equilibrium with metsllic iron at about 14OO’C. The mean temper~t~e at which sep&r&tionof metal etlld silicate occurred in the mantle was much higher than 14OO’Cand probably exceeded 2500°C. The equilibrium Fe2+/Fes+ ratio of ail&&es in equilibrium with iron in the mantle would be much higher at 25OO’C than at 1400%. Thermod~&mi~ c~lcul&tions indicate a threefold increase in the Fe2+fFea+ ratio over this temperature range. Aooordingly it is probable that the Fe2+/Fes+ ratio established in the mantle after se~&ration of iron into the core under eq~librium monitions would be in the vicinity of 200. The Fee+/Fe8+ ratio of the upper mautle fpyrolite) is 15 cud from the manner in which the pyrolite composition was derived, it is likely that this Sgure is Lbmaximum. In basalt derived by fractional melting of pyrolite, the Fe*+/ Fes+ ratio is approxim&ely 6. Such basalts would be in equilibrium with the volstiles which we observed to be degassed (HOLLAND,1963). To produce vol&iles of the correct compo~tion Gram& molten chondritic earth, it is necessary that the mantle should fractionate in such s, way that the Fe2+JFe3+ratio of the upper mantle is about t0 of the average Fe2*/Fes+ ratio in the entire mantle. It is extremely di& cult to understand how this might be accomplished by crystallization differentiation. We have already observed that the ions which can enter the principal minerals of the mantle have not been strongly fractionated. Both Fe%+ and Fe3+ are readily able to enter major mantle minerals, e.g. pyroxenes, spinels, snd ilmenitetype silicates ~RIX~WOOD,1965~) and hence strong f~~ction&tion would not be expected acoording to previous discussion. If Fe 3f had been strongly concentrated in the upper mantle by crystal fraction&ion, we would expect comparable enrichments of ions possessing similar ionic radii and charge such as Crs+ and V*+. These are not observed. On the Contras the evidence in~~&tes that these ions have been somewhat depleted in the upper mantle compared to primordial abundances (RI~UWOO~,1965b). Accordingly it appears extremely unlikely that t&o &

ratio

in the mantle has been changed by the required amount. We conclude that the

70

A. E. RINGWOOD

mean oxidation state of the mantle is such that it is not now, nor has it ever been, in equilibrium with metallic iron. In section 4c the abundances of some siderophile elements in the upper mantle were discussed. It was observed that Ni, Co, Cu, Au and Pt were strongly depleted in the upper mantle compared to primordial abundances. This is clearly a consequence of their siderophile properties which have caused them to become concentrated in the earth’s core. What is remarkable about Table 13 however is that the depletions of these elements in the upper mantle are not much stronger than observed. The distribution coefficients of Ni, Co and Cu between the olivines and metal phases of meteorites have been measured and show that the concentrations of these elements in the upper mantle are more than an order of magnitude higher than would be expected if equilibrium partitions between metal phase and silicates had been achieved. Distribution coefficients for gold and platinum are not known. However, from the fact that the free energies of formation of the oxides of these metals are positive at high temperatures, it would be expected that they should be relatively concentrated in the metal phase much more strongly than nickel. The substantial amounts of these elements remaining in the mantle thus appear inconsistent with an equilibrium metal-silicate fractionation. It is conceivable that conditions of equilibrium under extremely high temperature and pressure could affect the distribution coefficient sufficiently in certain oases to account for the discrepancies. This, however is not known, and would be an ad hoc assumption. It would be surprising if all of the discrepancies could be plausibly accounted for in this manner. The fact that the proportions of Fe, Ni, Co, Cu and Au, remaining in the mantle are similar within an order of magnitude (Table 13) is most difficult to account for in terms of any equilibrium theory. The combined weight of the evidence relating to the presence of silicon in the core, to the mean oxidation state of the mantle and the nature of the degassing products, and to the abundances of siderophile elements remaining in the mantle points very strongly to the conclusion that the earth’s core and mantle are not in equilibrium. This is an important boundary condition which must be satisfied by any acceptable theory of the formation of the earth. (e) Origin of the earth-multistage

theories

According to our primary hypothesis, the earth was formed from material resembling the Type I carbonaceous chondrites. We have discussed evidence relating to the present chemical composition of the earth and the distribution of elements between mantle and core. From a comparison of the present abundances and distribution of elements in the earth with the primordial abundances, significant conclusions may be drawn regarding the origin of the earth. Current theories of origin may be divided into two classes-multistage and single-stage. We will consider the former first. The multistage class of theories maintains that the primitive oxidised dust in the solar nebula was subjected to a high temperature stage, either in a dispersed state, or in an earlier generation of parent bodies, perhaps of lunar size (UEEY, 1966), before the earth was formed. During this preterrestrial stage, an extensive and complex chemical and physical processing occurred during which a metallic phase

(Ilmnical evolution of the terreatrid planets

71

was produced by reduction, and volatiles were lost. After the high-temperature stage, the mixed silicate and metal phases were cooled and subjected to phy~cal fractionation processes, The earth then formed by accretion of this mixed silicateiron material. Accretion took place sufficiently slowly so that the temperature of the earth remained low. The mean temperature after accretion was perhaps less than 1000°C. Theories of this type are by far the most widely held at present, very largely as a result of Professor UREY’S pioneering ~vestigations. (UREY, 1952,1956, 1957a, 1968, 1962, 1963.) The assumption that the earth formed by accretion under cool conditions of partially degassed silicate-iron material generally similar to ordinary chondrites is widespread as the following recent references indicate (RUNCORN,1962, 1964; ELSASSER, 1963; MUNK and DAVIES, 1964; WOOD, 1962; BIIXCH, 1964; V~O~~DOV,

1961.

If this view were correct we would expect the metal and silicate phases in the earth to be in chemical equilibrium because of their previous pre-terrestrial high temperature history (cf. ohondrites) and because of sustained high temperature conditions within the earth prior to melting of metal phase and separation of the core. As we have seen, the available evidence contradicts these requirements of the model. The ab~dances of Ni, Co, Cu, Au and Pt in the mantle are more than an order of magnitude higher than can be explained by equilibrium partition between metal and silicate. The oxidation state of the mantle and the composition of degassed volatiles are also in conflict with the equilibrium theory. Furthermore the model does not explain the inferred silicon content of the core. Finally, the chemical composition of the earth differs from those of all classes of chondritcs in major respects. The earth is relatively depleted in alkalis, mercury and sulphur compared to ordinary chondrites and in a wide variety of volatile elements compared to Type 1 carbonaceous chondrites. Thus the silicate-iron mixture hypothesis forces us to assume the existence of a material prior to the formation of the earth for which we lack any direct supporting evidence. The nature of this material is such that we are forced to assume the occurrence of most complex chemical processes in the solar nebuIa before formation of the earth for which again we have no direct evidence. The widely held assumption that the earth accreted in a relatively cool and unmelted condition from such material must also be challenged. Some workers (UREY, 1962; RUNCORN 1962, 1964; ELSASSER, 1963; MUNK and DAVIES, 1964) hold that after accretion the earth was heated by long lived radioactivities until the mean ~mperat~e was su~ciently high to melt the dispersed metal phase throughout the earth but not the silicates. The liquid metal aggregated into bodies which were large enough to sink through the solid mantle into the core, accompanied by convection in the mantle. Some advocates of this model argue that the core segregation is still in progress and is the driving force for convection in the mantle. One would expect an essentially homogeneous mantle (apart from iron) on this model. However, this imp~cation is in conflict with the strong upward concentration in the mantle of the incompatible elements U, Th, Ba, Sr, Ta, Be, B and rare earths (section PO(i)which, it was argued herein, implied that the entire mantle has passed through a complete or partially molten stage. Allowing for the effect of pressure on melting point, the mean temperature throughout the earth required to cause

72

A.

E. ~INCJWOOD

melting of metal but not of silicate would be about 2500°C. UREY (1962) showed that formation of the core from an initially uniform state would liberate about 800 Cal/g for the whole earth. Making a correction for strain in the interior, BIRCH (pers. comm.) suggests a smaller value of 600 Cal/g. The liberation of this additional energy after the mean temperature of the earth had reached 2500°C would also cause complete melting. Because the effective “viscosity” of the earth’s solid mantle decreases exponentially with rising temperature, the segregation of the core according to UREY’S model is an inherently unstable process, and the rate of segregation would be expected to increase rapidly with time culminating in complete and perhaps catastrophic melting. We conclude that the presence of a core and differentiated mantle constitute strong evidence for extensive or complete melting of the earth at an early stage of its history. The assumption that the earth would necessarily be solid and cool after accreting from ordinary chondritic material should not pass unnoticed. The mean gravitational energy liberated during formation of the earth is about 10,000 Cal/g. It is often assumed that the gravitational energy is instantaneously radiated and therefore lost when the surface of the growing earth is struck by acureting planetesimals. However analyses of meteoritic impact (SHOEMAKER, 1960, 1962, 1963) show that a substantial portion of the energy of colliding planetesimals is transferred to the earth by shock waves which are dissipated as heat within its interior, If only a small proportion of the energy of the planetesimals were conserved in this manner the earth would have completely melted during accretion. We will also see in the next section that there are additional factors operating during accretion which increase the probability of melting. It may be possible to avoid one or another of the above objections by introducing further restrictive assumptions. Nevertheless, in the author’s opinion, the combined weight of the objections to the hypothesis that the earth formed in a relatively cool, unmelted condition from an intimate silicate-metal mixture which had previously been subjected to highly complex, unrecorded physical and chemical fractionations are so strong that the hypothesis should be abandoned. the origin of the earth (f) A single-stag e h ypothethesisfor Because of the difficulties in multistage theories, the author (1960) proposed that the earth formed directly by accretion from the primitive oxidised dust in the solar nebula, and that reduction to metal, loss of volatiles, melting and differentiation occurred simultaneously and as a direct result of the primary accretion process. The assumptions of intermediate stages of reduction and fractionation were avoided, so that the hypothesis amounted to a single-stage process. It was further argued (RINGWOOD, 1959, lQ61a, 1962, 1965a, b) that chondrites and other terrestrial planets also formed by a single-stage process and did not demand an earlier metalsilicate fractionation. The incorporation of some volatiles within the earth was found not to be inconsistent with an accretion process occurring dominantly at high temperatures. We now consider the development of the earth from the parent gas-dust cloud. The cloud is assumed to be cold ( < 100°K) because of its opacity to radiation (&?lK, 1962a, 1963~). The solid phase in the cloud would consist principally of a mixture

Chemical evolution of the terrestrialplanets

73

of non-volatile silicate-iron oxide dust, ices and complex carbonaceous compounds (section 36). For some reason, as yet unknown, the cloud becomes unstable and a series of condensations or pl~e~simals are formed. Current estimates suggest that these condensations may have had varying sizes up to perhaps 100 km diameter. However, they are continually colliding and disentegrating so there will be a wide range of size distributions within the dust cloud. Eventually a fluctuation arises, such that a large condensation forms. This is sufllciently large to exert substantial ~a~~tional attraction upon other condensation and not be broken up by the resulting collisions. When this stage is reached, the primary accretion process commences, and dust and condensations begin to fall on the central nucleus.

Radius

, km

Fig. 4. ~~tio~p b&wean energy of mm&ion @d/g) end radius during the growth of s temetri& planet.

The mean gravitational energy liberated during formation of a body the size of the earth is about 10,000 Cal/g. The accretion energy increases roughly as the square of the rdius aa shown in Fig. 4. During the early stages of accretion, the ~a~~tional energy is low, and the indentions are not strongly heated when they fall upon the nucleus. As the latter grows, and the energy of accretion increases, the heating on impact becomes sutlicient to vapourise the more volatile components of the dust-principally water in excess of that held by hydrous silicates. It is also possible that some mercury was volatilised at this stage. The earth appears to have sufSered an extreme depletion in this element (Tables 14, Iti). The temperature of the less volatile phases of the condensate remains low because of the buffering effect caused by absorption of latent heat during evaporation of the volatile components. Thus we see that the first stage of accretion results in the development of a cool oxidised nucleus, rich in volatile components, particularly water. Perhaps the size of this nucleus would be about that of Mars. A critical stage of evolution is reached wben the nucleus is large enough to retain (even temporarily) some of the escaping volatiles in an atmosphere. The situation is then radioally changed-instead of dust and planetesimals falling on a aold solid surfaoe, they fall into a blanket of gas, and the interactions are fundamentally

74

A. E.

RINGWOOD

different. As the central body grows and its gravitational potential increases, smaller condensations fall into the atmosphere and become completely evaporated. Ionization and radiation occur, accompanied by chemical recombination of the less volatile components. However the recombination occurs under reducing conditions caused by the presence of carbon and hydrogen, so that the interaction of dust and atmosphere causes reduction of oxidised iron to metal. Larger condensations may possess sufficient coherence to completely penetrate the atmosphere, although they would suffer considerable ablation in so doing. They would explode on hitting the surface of the growing earth, and, in the presence of reducing agents at transient high temperatures, would suffer reduction of iron and rather complete degassing. Ultimately the atmosphere would reach a size after which nearly all of the infalling material was reduced and degassed before reaching the surface of the earth. As the accretion energy increased, the temperatures and intensity of reduction would also increase so that not only iron, but also silicates were reduced to metal. The condensates reaching the surface of the growing earth during the later stages of accretion would thus probably be composed of iron silicide and the non-volatile oxides referred to in Table 10. Virtually all of the elements which were volatile at high temperatures in the presence of excess hydrogen and carbon would enter the atmosphere (Table 10). These include the alkali metals, Zn, Cd, Hg, Ge, Pb, Tl, Bi, In, S, Cl, I? and probably many others, e.g. Br, I, Se, Te, Sb, As, Ga. In this manner the growing earth became depleted in volatile elements. If the above elements are to remain in the atmosphere, the temperature of the atmosphere and of the earth’s surface must have been in the vicinity of 1500°C during the later stages of accretion. We have previously discussed evidence which implied that the earth passed through a completely or partially molten stage very early in its history, and most probably during its formation. Gravitational energy of accretion provides the most plausible source of heating. The surface temperature on the accreting planet is determined by the balance between rate of accretion and the rate at which energy can be radiated away. The two most important parameters are the time scale of the accretion process and the opacity of the medium surrounding the earth. The former is not well known. It appears that accretion may proceed very slowly at first giving rise to a cool central nucleus as previously discussed. However as the mass of the nucleus grows, the accretion rate and hence the surface temperature probably increases rapidly (HOYLE, 1946; TER HAAR and WERUELUND, 1948). HOYLE (1946) found that even assuming a transparent atmosphere, the outer parts of the earth may have melted during the final stages of accretion. However it is improbable that the atmosphere was transparent to radiation

during accretion. It has been suggested here that a large absorbing atmosphere developed. This would increase the effective temperatures developed during accretion. OPIK (1960, and personal communication) has drawn attention to a factor which is probably more important. During accretion, the earth’s environment would be effectively opaque to radiation because of screening by the surrounding dust particles. The energy of accretion is liberated near the earth’s surface and is blanketed by the surrounding dust. Z)PIK has calculated that the time of accretion could be extended over lo6 years and still lead to melting. This time interval need apply only to the later, rapid stages of accretion, so that the total period of accretion may

Chemical evolution of the terrestrial planets

76

have been much longer than IO6years. In view of these considerations, the assumption which is made in this paper, that perhaps two-thirds of the material in the earth accreted under sustained high temperature conditions does not appear unreasonable. During the later stages of accretion when the melting point at the surface ww exceeded, metallic iron would segregate into masses which were large enough to flow directly into the core. We have previously noticed that the gravitational energy liberated during formation of the core would contribute 600-800 Cal/g for the entire earth, sufficient to result in complete melting throughout. On this model, segregation of the core occurred as a continuous process during the primary accretion. A cat~trop~c version of core fo~ation is also a possib~ty (RIN~WOOD, lQ60). (g) Origin of mantle-core

disequili6riu.m

According to the above model, the earth develops in a state which is grossly out of chemical equilibrium. The deep interior is initially highly oxidised and rich in volatiles, whereas the outer regions are progressively more reduced and poor in volatile components. After melting near the surface, the metal phase consisting of iron-nickel-silicon alloy collects into bodies which are large enough to sink into the core, Equilibrium between metal and silicates can only be attained by diffusion across the interface of the sinking bodies of metal. If the rate of sinking of metal is high compa~d to the rate of at~~ent of equilibrium by dif%sion, the core which separates will not be in equilibrium with the mantle. In the author’s opinion the situation is likely to arise, particularly in the later stages after melting, when metal segregation is comparatively rapid. After the core has separated, and the mantle is molten, both regions will become homogenised by convection, but they will remain out of eqnilibrium with one another. The situation has been more fully discussed by RIK~WOOD( 1959,196 1b) . The presence of silicon as an impo~nt component of the earth’s core, whilst the mantle contains oxidised iron, can be understood on these grounds. Also the presence in the mantle of a significant proportion of the primordial nickel, cobalt, copper, gold and platinum is explicable. These elements are derived from the cool oxidised central nucleus, much of which entered the mantle in bulk during core segregation, and was not intimately associated with a metal phase under such con~tions that eq~b~um could be reached. The retention of volatiles, particularly water, nitrogen, carbon dioxide and inert gases, within the earth is also ascribed to the non-attainment of equilibrium. The volatiles are incorporated in the earth at an early stage of accretion as components of the cool oxidised nucleus. Melting and separation of the metal phase take place under such conditions that the water and carbon dioxide in the oxidised nucleus do not reach eq~~b~~ with metal. After the core has segregated the volatiles become distributed throughout the mantle. The only way they can escape is by diffusion processes at the earth’s surface which, considering the vast volume involved, are prohibitively slow. The difficulty of degassing industrial glass melts in large furnaces is well known. Escape of volatiles by formation of bubbles in the upper few kilometers is prevented by the pressure of the atmosphere above the earth’s surface which prevents cavitation. During crystallization of the mantle, some of the volatiles would be expelled, but it is probable that a large proportion would be

76

A. E. RINCSWOOD

trapped within crystals and as separate minor phases, e.g. amphiboles in the upper mantle and various high pressure equivalents in the deeper mantle. (h) Escape of primitive atnzosj&ere The single stage model for the formation of the earth involves the production of an enormous atmosphere amounting perhaps to half of the mass of the earth and composed chiefly of hydrogen and carbon monoxide. There is no evidence from crustal rocks of the presence of such an eno~ous former atmosphere. ~bv~o~ly therefore, a critical requirement of the hypothesis is that complete escape of the primitive atmosphere, together with all its minor components such as volatile metals, occurred at a very early stage of the earth’s history. Geochemical considerations indicate rather strongly that escape of a primitive atmosphere from the earth actually occurred at an early stage. It is generally agreed that the present atmosphere and hy~osphe~ developed by continuous degassing of the earth’s interior throughout geological time (RUBEY, 1961, 1966). This process has not prooeeded to oompletion and large amounts of volatiles remain in the interior. The observed distributions and abundances of nitrogen (RINUWOOD, 1965b) and argon indicate that the volatiles remaining within the earth exceed in amount the degassed volatiles in the hydrosphere and atmosphere by at least a factor of 3. This factor must have been very much higher early in the earth’s history, before the present atmosphere and hydrosphere were formed. The estimates of RUBEY (1951) suggest that early in the earth’s history the quantity of volatile components trapped within the earth was at least 60 times as high as the volatiles occurring at the same time in the atmosphere and hydrosphere. It is most di~~ult to ~derstand how the earth was formed in this condition. Referring to E’ig. 4 we see that when the mass of the earth has grown one tenth the present value (Mars), any solid matter aooreting at the earth’s surface arrives with such a high velocity that it is subjected to intense transient heating during impact, leading to melting and degassing. As the earth grows, and the energy of accretion increases, the transient heating during explosive impact with the solid surface causes almost complete vapou~zation and degassing. This stage occurs before the earth has grown to more than one quarter of its present mass. After this stage, nearly all the incoming material will be extensively degassed. At the oonclusion of accretion the amount of degassed volatiles present in the primitive atmosphere must exceed, on any reasonableassumptions, theamountofvolatilestrappedin theinterior. Sinceit isagreed that most of the present atmosphere and hydrosphere was o~ginally trapped in the interior, it follows that a primitive atmosphere much larger than the present atmosphere and hydrosphere must have been present after the earth was formed. Clearly, some mechanism for the escape of this primitive atmosphere must have existed. The physical conditions under which escape occurred are not known RIN~WO~D (1959, 1960, 1965b) qualitatively mentioned several factors which may have been involved, e.g. intense solar particle radiation when the sun passed through a T-Tam+ stage, and magnetohydrod~amic effects combined with rapid rotation of the newIy formed earth. UREY (1960, 1962a) has argued apparently on the basis of the Jeans-Spitzer theory of selective escape of gases from a gravitational field that the escape of a dense

Chemicalevolutionof the terrest~idplanets

77

primitive atmosphere from the earth is not possible and that accordingly the single stage theory for the origin of the earth as advocated by RINGWOOD (1959,196O) must

be discarded. However the JeansSpitzer theory is not applicable to the physical conditions under co~deration. The JeansSpitzer theory applies to the selective escape of gas molecules from an exosphere where the mean free path of molecules is of the same order as the scale height. In the case under consideration, an exosphere with the required low density is not present, since the atmosphere of the earth is continuous with the hydrogen-rich gas cloud of the solar nebula or, on KUIPER’S model, with the hydrogen-rich atmosphere of the terrestrial protoplanet. ~)PIK (1963a, b) has pointed out that under such con~tions, escape will not be selective with respect to molecular weight. If the temperature is sufficiently high, the atmosphere may “blow off.” The critical parameter for this process is the mean molecular weight of the atmosphere. If this can be lowered sufbciently by introduction of hydrogen, escape by blowing-off becomes correspondingly easier. The atmosphere formed during the autoreduction of the earth consists of a mixture of carbon monoxide, hydrogen and many minor com~nents ~clu~~ easily ionized alkali metals. Presumably this atmosphere will rotate with the same velocity as the earth, particularly if magnetic fields are present. At some height this atmosphere becomes continuous with the hydrogen-rich gases of the solar nebula, or terrestrial protoplanet. The interaction between rapidly moving terrestrial gases and relatively stations solar hydrogen is doubtless highly complex, However it seems possible that it may lead to strong heating and mixing at the interface. The result is a large reduction of the mean molecular weight and it is conceivable that at the high temperatures produced, the terrestrial atmosphere will diffuse into the gases of the solar nebula. The problem of escape of the earth’s atmosphere then becomes identical with the more general problem of dissipation of the solar nebula. KUPER (1962,1957) has discussed this problem and att~bu~s the ~ssipation to solar particle radiation. The entire problem of atmosphere escape urgently requires quantitative investigation. We have seen that geochemical arguments imply that escape has occurred, and that furthermore, the contrary arguments based upon applicability of the JeansSpitzer theory are not valid. Nevertheless we still do not know how this occurred, although several workers have recently expressed the opinion that escape is possible on various grounds, e.g. SUESS, 1949; KUIPER, I967 ; HOYLE and FOWLER, 1964; and CAMERON,1963. The importance of this problem to theories of the origin of the earth can hardly be overestimated. 5. TEE MOON

The density of the moon is 3.33 g/cm a. This implies a major difference in chemical composition compared to the earth. An explanation of the difference is the prime requirement of any theory of lunar origin, The mass ratio of moon to earth is 0.0123. This is much higher than is usual for satellite systems. The moon is clearly exceptional in this respect and is sometimes considered to be more analogous to the smaller component of a binary star system.

78

A. E.

RINGWOOD

Most current theories of lunar origin fall into 4 classes: (i) The fission theory, involving separation of the moon from the earth’s mantle at an early stage (DARWIN, 1880, 1962). (ii) The binary star theory according to which the moon formed by direct accretion close to the earth, and perhaps in the same protoplanet (KUIPER, 1954, 1957, 1959b, lQ63). (iii) The moon was captured by the earth (UREY, 1960, 1962b, 1963, 1965). (iv) The moon is a secondary body, formed by the coagulation of a ring of planetesimals or moonlets, which once surrounded the earth (~PIK, 1955, 1961, 1962b ; MACDONALD,1964). All of these theories in their current forms face serious difficulties and none has won general acceptance. At present, we simply do not possess sufficient data to resolve vital questions. Interpretations of the most straightforward observations vary widely and highly subjective judgements are inescapable. This applies also to the interpretation and theory developed in the present paper. The author’s theory is based upon the belief that an explanation of the moon’s low density is the fundamental problem to be solved and also that the solution must lie within the basic framework already advocated fur the single-stage origin of the earth, chondrites and terrestrial planets. After a brief review of current theories, an attempt will be made to modify one, to explore its consequences, and to attempt to explain in a broadly self consistent manner as many ‘facts’ and inferences about the moon as possible. The choice of the facts and inferences which are to be explained is one of the most subjective parts of the procedure. In the interest of brevity such choices will not be justified. In general, however, the author has accepted the interpretation of lunar surface features offered by BALDWIN(1963), KUIPER (1954, 195Qa,b, 1963) and SHOEMAKER (1962). The present discussion also owes much to papers on broader aspects of lunar evolution and constitution by UREY (1960, 1962b, 1963) and ~PIK (1955, 1961, 1962b), although at many points, the views expressed differ from those of UREY. (b) Some current theories of lunar origin (i) Thafission hypothesis. DARWIN(1880, 1962) developed a theory according to which the moon and the earth were originally combined in a single body which rotated with a period of 4 hours. A resonance effect occurred between the free period of the parent body and the solar tides, leading to the development of an enormous tidal bulge. An instability developed and the tidal bulge was ejected from the earth to form the moon. This theory provided an elegant explanation for the moon’s density, which is similar to the uncompressed density of the earth’s mantle. However, serious objections to DARWIN’Shypothesis were raised by JEFFREYS(1930) and it was generally discarded. RINUWOOD(1960) suggested a new variant of DARWIN’Shypothesis arising from a discussion of the formation of the earth by accretion, as in the preceding section. It was argued that after accretion the extent of reduction and consequently the amount of metal phase increased from the centre of the earth towards its margins. This configuration is highly unstable, and there is the possibility that segregation of metal into the core may have been catastrophic. If, towards the end of the primary accretion process, the rate of rotation of the earth was close to the instability limit,

Chemicalevolutionof the %errestrial planets

7s

a rapid segregation of the core may have decreased the earth’s moment of inertia and accordingly increased its angular velocity suEoiently to cause fission. According to the suggestion the excess angular momentum of the earth-moon system was carried away by the large primitive atmosphere which was also ~s~p~d and escaped during the cataclysm. A similar hypothesis was subsequently suggested and developed by WISE (1963). More recently, CAMERON(1964) has supported this hypothesis and considered it at some length. Interest in the fission hypothesis has been revived partly because of the possibility that tektites come from the moon. The chemical and isotopic compositions of these objects are so similar to terrestrial surface material that if a lunar origin is ever conclusively demonstrated, it will constitute strong evidence for the ultimate derivation of the material from which the moon accumulated from the earth’s outer regions after the core had segregated (O’KEEFE, lQ63). At present the fission hypothesis is faced by severe dynamical difficulties (UREY, 1963 ; ~cDo~~D, 1964) and must be regarded as rather extreme. ~eve~heless, a conclusive demonstration of the lunar origin of tektites would necessitate a revision of this appraisal. (ii) &nary-planet hypothesis. KUIPER (1954, 1957, 195Qb, 1963) suggested that the earth-moon system is essentially analogous to a binary star system. According to this theory, moon and earth formed in close proximity by direct accretion from similar parental material. This immediately raises the problem of explaining the moon’s low density. It has been postulated that in some maimer, the growing earth received a greater proportion of metal phase than the moon, but a plausible mechanism has not been suggested. Alternatively, the moon may possess a similar major element composition to the earth, and may contain in addition 2 or 3 per cent of water andfor several per cent of graphite, which would lower its density to the required value (UREY, 1962, 1960; RIN~WOOD, 1969). Its composition would thus be similar to that of carbonaceous chondrites, and it could therefore have formed from similar parental material to the earth. UREY (1962b) has recognized the chemical and thermal difiiculties in this hypothesis. Thus if the moon were composed of primitive carbonac~us chondrite material, the amounts of radioactive elements present would be sticient to raise the temperature of the interior close to the melting point. It is difficult to understand why carbon was not then consumed by reaction with oxidised iron to produce metal. Furthermore the preservation of the moon’s non-equilibrium shape is difficult to reconcile with chondritic ab~dances of ra~oactive elements. If the moon conks a subs~ntial quantity of water in its interior it might be expected that extensive manifestations of explosive voluanism should be apparent at the surface. Finally, UREY (1963) drew attention to a dynamical difficulty connected with the formation of a binary earthmoon system by accretion. The objections so far raised against the double planet theory are serious but perhaps not decisive. It may well be possible to avoid them by the introduction of special assumptions. Thus, if the moon were formed from hydrous and carbonaceous primitive material resembling the Type 1 carbonaceous chondrites, it is possible that the amount of internal radiogenic heating might have been much smaller than ~ti~ipa~d because of the large amount of energy which must be expended in dehydrating hydrated silicates, heating water to high temperatures and driving it to 6

80

A. E.

RINCJWOOD

the lunar surface. If enormous volumes of water were expelled from the lunar interior at high temperatures, it would dissolve and transport corresponding quantities of soluble components (RIN~WOOD,1965a) which would be deposited at the lunar surfaae after the water had evaporated. Extensive evidence of such deposits should be visible if the composition of the moon is as assumed. The only lunar features of the required size which could conceivably result from such a process are the maria. The author is not arguing that the maria have in fact been formed in this manner, but rather pointing to a probable implication of the double-planet hypothesis. The evidence marshalled by BALDWIN(1963) and KUIPER(1959a, 1965) does not support the interpretation of maria as evaporites. But neither can this implication be conclusively disproved. Until new sources of data become available, it would be imprudent to reject the double planet-theory completely. In the author’s opinion, the theory does not offer the most plausible interpretation of available information and is faced by several difficulties. Nevertheless it should be retained as an alternative working hypothesis. (iii) Capture. UREY (1960, 196213,1963, 1965) favours the hypothesis that the moon was captured by the earth. He regards the moon as a primitive object and one of the few survivors of a generation of parent bodies from which the present terrestrial planets evolved via complex fractionation mechanisms. The moon thus plays a critical role in UREY’S views of the evolution of the solar system. The evidence for UREY’Shypothesis is mainly indirect and will not be considered further. However, there is one important piece of direct evidence. The inferred abundance of iron in the moon is similar to its presumed abundance in the sun (UREY, 1863). The significance of this correlation depends largely upon the reliability of the solar iron abundance.* Recent determinations of the abundance have yielded widely discrepant results for the photosphere (GOLDBERG et al., 1960) and for the corona (POTTASCH, 1963). Even if the lower value for the solar iron abundance should prove correct, its implications for the composition of the moon are not clear. Capture processes are highly improbable events as admitted by UREY (1063). ~PIK (1955,196l) has pointed out that if the moon is to be captured by the earth, it must be moving initially in an orbit very similar to that of the earth. This implies again that the moon was born in a region of the nebula very close to the earth, and it would therefore be expected that it accreted from material of terrestrial composition. Thus the problem of its different composition remains unsolved. There is really little difference between the double-planet and capture hypothesis in their implications concerning lunar composition. However the capture theory is less attractive because of its much smaller intrinsic probability. (iv) Coagzclatio~of a terrestrial “sediment ring.” OPIK (1955, 1961, 1962b) has discussed the formation of lunar craters in connection with the tidal history of the moon. He concluded that the moon probably formed by the aggregation of a cloud of debris and planetesimals which orbited the earth at a distance of 6 to 8 earth radii. Such a cloud was termed a “sediment-ring” by KUIPER(1964,1956). Tidal evolution aommenced only after this material collected into one body. Recently MACDONALD * Note added in press: The lcttest photospheric abundance determinations appear to have removed the disagreement which formerly existed between the solar and chondritioiron abun* dances-see appendix, page 104.

Chemicalevolutionof the terreatrislplanets

81

(1964) has conducted an extensive investigation of tidal evolution in the earth-moon system and has arrived at a somewhat analogous model, although for di%rent reasons. According to MUDON~LD’S model, the moon formed about 1.6 x 10’ years ago by the aggregation of several pre-existing moons. The time scale depends rather critically upon assumptions concerning the nature and constancy of dissipation of tidal friction in the earth. OPIK and MACDONALD do not explain the origin of the parental planetesimal ring or the cause of the moon’s low density, In the following sections an attempt is made to provide such an explanation and to integrate ~PLK’Smodel with the single-stage theory for the formation of the earth as discussed in section 4. (c) A model for the formation of the moon During the final stages of accretion of the earth as discussed in section 4 (see also RINGWOOD,1965b) it was suggested that the ~m~rat~e in the atmosphere exceeded 1500°C leading to melting at the surface, At temperatures between 1600 and 2000°C and depending upon the redox conditions, it is possible that major chemical fractionation occurred, and a large proportion of components previously regarded as non-volatile, e.g. SiO, and MgO, were in fact volatilised and entered the atmosphere. UREY (1952,1964) has shown that in the presence of gases of solar composition there is a substantial P, T field in which silicates are volatile but metallic iron is not. It is suggested that analogous conditions obtained on the earth, and that a fractionation of silicates from metallic iron was thereby caused, with iron continuing to accrete on the earth, and silicates entering the atmosphere in the gas phase. It has previously been argued (section 4, RINQWOOD,1906b) that during or perhaps imme~a~ly after the primary accretion, the massive, primitive, reducing atmosphere escaped from the earth by one of a number of possible processes. The exact mechanism does not concern us here-our objective is to explore some of the possible consequences of such an escape. As the gases recede from the earth, expansion and cooling occur, leading to precipitation of less volatile components in the form of smoke, condensations or plane~simals. According to our previous assumption, the pre~ipita~ would consist dominantly of silicates depleted in iron. Collisions between planetesimals combined with mutual gravitation would cause this debris to collect in a “sediment ring” (KUIPER,1954,1956) surrounding the earth. If the escape of the atmosphere were partly caused by rapid rotation or controlled by magnetio fields, the plane of the ring may have lain approximately in the plane of the earth’s equator. Such a segment-~g has been envisaged by ~PIEC(1955, 1961, 1926b) as the parent of the moon. A viscous interaction between outwardly moving gases and condensate might result in this material moving outwards from the earth and beyond Roohes Limit, where aggregation into larger bodies occurred. Ultimately these bodies collected together to form the moon. We must next examine the chemical conditions under which separation of iron from silicates may have ocourred. According to the model, primitive a~~ret~g material is subjected to autoreduction at very high temperatures when it falls into the terrestrial atmosphere. We have previously seen how this will cause reduction of oxidised iron and some of the silicates to a ferrosilicon alloy. For reduction to proceed further, it is necessary that a sufficient supply of carbon be available. The

82

A. E.

RINUWOOD

amount of carbon available for reduction of silicate depends partly upon the water content of the accreting material. Consider the equilibrium H,O + C = CO + H,

(K20,,00g= ‘~![;’

= 830) 2 It is evident that the equilibrium is driven strongly to the right at high temperatures, and much of the carbon in the primitive accreting material is consumed in reducing water to hydrogen. One way to secure an excess amount of carbon in the accreting material during the final stages of formation of the earth would be loss of most of the water from primordial accreting material before it fell into the primitive atmosphere. It seems quite plausible that the temperature immediately outside the earth and its atmosphere should have been high enough to cause partial dehydration of infalling dust without substantial accompanying loss of carbon during the later stages of formation of the earth. Other sources of excess carbon could be devised (e.g. SUESS, 1962). We will assume in the following discussion that an adequate quantity of carbon was present in the accreting material to cause further reduction of silicates. Since reduction is taking place in the presence of liquid metallic iron, in which carbon is soluble, the activity of carbon in the systems under consideration will be lowered. We will assume that the final carbon content of the ferrosilicon liquid is about two per cent. For a silicon content of 12 per cent, the activity of carbon would then be about O-3 (HOPKINS, 1954). We will use this figure in subsequent calculations. Returning to our model*-we wish to enquire into the conditions under which the entire material now in the moon originally occurred as gaseous components in the primitive atmosphere. The moon now amounts to 0.0123 of the mass of the earth. The mass of the primitive atmosphere is believed to have been about 0.6 of that of the earth, and is assumed to have consisted of carbon monoxide and hydrogen in equal (molecular) proportions. For a first approximation the moon is assumed to consist of MgO, SiO,, A&O, and CaO in the proportions in which these components occur in Type 1 carbonaceous chondrites. From the above considerations the abundances of Mg, Si, Al and Ca in the primitive atmosphere, as required by the present model are established. Consider the equilibria governing reduction and volatilisation of a given silicate, MeSiO,. I MeSiOt(melt) $- C(Fe)= Metgas)+ CO + SiOz(melt)

where

K, P, P,,

= = =

%iOa= %iesiol,= a, =

K, zz p Me*pcO * aSiOa a, - %eSiO, equilibrium constant for above reaction partial pressure of Me atoms in primitive atmosphere partial pressure of carbon monoxide in primitive atmosphere activity of SiO, in silicate phase (melt) activity of MeSi03 in silicate phase (melt) activity of oarbon in coexisting ferrosilicon phase.

II

* In a brief preliminary outline of the model, the author (1966b) used JzZ&,O ratios aa indicators of redox states. This approach has been abandoned in the present paper in favour of a more direct discussionof carbon-reductionequilibria.

a3

Chemical evolution of the terrestrialplanets

Now, H, is obtained from the relation AG = - RT In H, where ACTis the free energy change aseooiated with the above reaction and may be evaluated as a function of temperature from tables of thermochemical data (KUBACHEWSRI and EVANS, 1958). The right-hand side of equation II is obtained as follows Pn, = xP where P is the total pressure in the atmoshpere and x is the molecular fraction of Me atoms in the primitive atmosphere as given by the requirement (previously discussed) that the primitive atmosphere should be capable of precipitating a mass amounting to that of the moon on cooling. PC0 is simply O-5P to an adequate approximation, whilst a, as previously discussed was fixed as 0.3. This leaves the quantity aBIO,/aneBo,to be

OC

1

0 Log

2

P (otm)

Fig. 6. P, T conditionsrequiredfor reductionand volatilisationinto a primitive atmosphereof major componentsfrom primordialmaterial accretingon the earth. Composition of primitive atmosphereis defhwd by the condition that it is capable of one lunar mass of silicate material on cooling. (See text for further details.)

evaluated. There is no firredvalue for this ratio since we are dealing with a complex silicate melt which may be treated as a mixture of MeSiO, components. The activities of all components are interdependent. Accordingly reduction and volatilisation of components will occur continuously over a wide range of temperature and pressure conditions. To simplify the discussion, we have treated components as if they were independent, and assumed asloJu~~slo to be unity. The errors introduced by this assumption are not as serious as might appear. MgSiOs constitutes about 80 per cent of the material being reduced and its activity is not far from unity. It turns out also that Mg and Si are the most volatile components so that when conditions are reached

under

which

Ca and Al become

volatile,

most

of the Mg and Si will

84

A. E. RJXUWOOD

have been previously removed, and accordingly the residual melt is rich in Caf) and Also,. Errors introduced by this ~s~ption do not affect the principal q~litativ~ conclusions which follow. The right-hand side of equation II was evaluated for a series of total pressures using values for individual terms as obtained above. With K, obtained as a function of temperature from thermochemical data, the equilibrium temperatures for reaction I were obtained at the given pressures. Results are given in Fig. 6. The lines show the oon~tions under which particular components would be reduced and volatili~d from primordial material in abundances sufficient to form a lunar mass on cooling of the atmosphere. The calculations are admittedly crude and it would be unwise to attach more than qualitative significance to the position of the boundaries on Fig. 5. Furthermore it should be realized that each equilibrium line in fact represents a broad zone over which ~olat~isation oconrs accompanied by change of composition both of the residual melt and the proportions of volatile components. Kowever, despite these reservations, Fig. 5 yields some highly significant trends, which we will discuss shortly. Also on Fig. 5, the conditions required for complete volatilisation of the metallic iron which originally was present in the accreting material are given. For complete volatilisation, the partial pressure of iron is simply xyP where y is the proportion of iron atoms to Me atoms (e.g. ma~esium) in Type 1 carbonaceous ohon~tes, z and P being already defined. From Fig. 5 we see that silicon and magnesium are much more volatile than iron, calcium and magnesium under the stated conditions. At one atmosphere pressure, the volatilities decrease in the order Si (as SiO) > Mg > Al (as Al,O) > Ca > Fe. A similar order was found by UREY (1964).The general similarity in the ~olatilities of Al, Ca and Fe indicate that fractionation of Al and Ca from Pe would occur only under highly restricted conditions. On the other hand, there is a wide field in which magnesium and silicon would be reduced and volatilised from primordial material whilst iron remained in the condensed state. The P, T conditions required are not extreme and were probably attaiued during the late stages of accretion of the earth. Aocor~ngly it appears possible that a subs~n~al proportion of ~gnesi~ and silicon monoxide entered the atmosphere as gases during the final stages of accretion and were precipitated as magnesium silioates and silica when the atmosphere escaped and cooled. From Fig. 5 we see that silicon monoxide is more volatile than magnesium. InitiaIly therefore, silicon would be selectively volatilised. This would decrease its activity and increase the activity of magnesium in the silicate melt until a composition was reached at which both components volatilise together in comparable proportions. On the simplest assumption the overall Si/Mg ratio in the gas phase would be greater than unity. However allowance must be made for the proportion of silicon (10to 20 per cent) removed in solution in the ferrosilicon phase which continues to accrete upon the earth. All things considered, it appears that the Si/Mg ratio in the gas phase might not be much greater than unity and that the principal phase to condense when the gas cooled would be enstatite, accompanied by lesser quantities of silica materials. The above d&u&on has attempted to cover equilibrium volatilisation under

chemicalevolutionof the tarr&rial planets

86

conditions of comparatively low pressure, as would apply in the outer regions of the primitive terrestrial atmosphere or under conditions such that the primitive atmosphere w&gescaping continuously during the primary aocretion. These eq~b~a will also control the conditions under which precipitation from the escaping atmosphere occurs. On the other hand, conditions would be somewhat different in a more static model, in which a massive primitive terrestrial atmosphere accumulated and did not escape until the earth’s accretion was essentially complete. In this case, the pressure in the deeper levels of the atmosphere would be very high-of the order of 106-IO* atmospheres and the equilibria discussed above, based essentially upon an ideal gas approximation, are not applicable. Under these circumstances it is probable that a non-equilibrium fractionation of silicates from metal would occur. At the enormous pressures involved, the density in the lower levels of the atmosphere would resemble that of a liquid rather than a gw. If the temperature at the base were sticiently high (~200O’C) it would be more realistic to consider components from the outer region of the earth “dissolving” in the dense fluid. A partition equilibrium between mantle and atmosphere would be set up. However iron would not participate, since, aa shown by UREY (1954)it would be reduced to metal at the surface of the mantle and would sink below the surface into the interior of the earth where it would be effectively removed from the system. Thus the results of such a non-equilibrium partition would be generally similar to the eq~librium case discussed above. The major com~nents entering the atmosphere in greatest abundance would probably be those which are most readily reduced-Mg and SiO. However, entry of Ca and Al,0 into the atmosphere may also occur under these circumstances. It will be reoalled that the atmosphere is also believed to contain large quantities of minor elements, e.g. alkali and other volatile metals which were volatilised from infalling terrestrial material during the major phase of accretion, and at much lower temperatures than were required for volatilisation of magnesium and silicon. According to the model, the primitive atmosphere escapes from the earth, expands and cools. Fractionation of components in the reverse sense then occurs, with the least volatile components, principally magnesium and silicon, being precipitated first,probably as a mixture of magnesium silicate and silica. Because of the high temperatures (above the liquidus) the condensate is readily able to collect into substantial planetesimals which are left behind by the outflowing gas. On the other hand, the highly volatile elements including the alkali metals remain in the gas phase at this stage, and are carried further from the earth and from the magnesium silicate planetesimals by the expanding atmosphere. These elements condense from the gas phase at much lower temperatures, under sub-solidus conditions, probably in the form of fine smoke. Because of their lower ~rn~rat~e, this material may not have grown into planetesimals, but would instead have been carried away from the earth’s environment with the outflowing gases. Thus the silicate planetesimals may have been depleted in potassium. Alao, because of the lower volatilities of uranium and thorium compared to calcium and aluminium, to be expected nuder these conditions, the planetesimals may also have been selectively depleted in these elements. Thus, a net depletion of radioactive elements in the planetesimals compared to the primordial abundance might arise from the operation of these processes.

86

A.

E.

RINGWOOD

According to the outline of events proposed above, a large number of planetesimals composed principally of iron-free magnesium silicates and silica were left behind after the primitive terrestrial atmosphere was dissipated. The density of pure enstatite and forsterite is close to 3.20. If substantial quantities of free silica were present in the planetesimals, their density would be correspondingly lowered. The density of the moon is 3.33, accordingly we are required to assume the presence in the moon of a denser material capable of increasing the mean density to the correct value. We might assume that the fractionation processes discussed above did not lead to efficient separation of material according to volatility. Accordingly, some calcium, aluminium and iron were also volatilised into the primitive atmosphere and precipitated in the planetesimals. The presence of some metallic iron or iron silicide in the planetesimals might increase their density to the observed lunar value. Within the framework of the model, the assumption is probably at least partly correct. It does not appear likely that a perfect separation of magnesium silicates from other components would be achieved in such a complex natural process as that envisaged. If we accept this model, it is possible to account for many of the properties of the moon as in the following sections (i) to (v). On the other hand we could make an additional assumption concerning the origin of the high density material. It is possible that this material was derived from primitive planetesimals similar in composition to the Type I carbonaceous chondrites, from which the earth accreted. It does not appear reasonable to conclude that the earth collected al2 the primitive material within its sphere of influence. It seems more likely that an appreciable proportion of this material might have remained behind in satellite orbits. These primitive planetesimals might then have become mixed with the iron-poor magnesium silicate planetesimals precipitated by the escaping atmosphere and the moon subsequently formed from the mixture of planetesimals. The primitive planetesimals would become heated and metamorphosed within the moon resulting in a loss of volatiles, partial reduction of iron, and would probably reach a density about 3.7 to 3.8. Thus the presence of a substantial amount of this material mixed with the less dense magnesium silicate planetesimals might yield the required lunar density. It might well appear that this additional assumption is a needless complication to an already complex theory. Nevertheless within the broad frame-work of the model the assumption is not implausible. Its chief virtue is that it offers the possibility of a more detailed and specific explanation of lunar properties, than does the simpler model according to which the moon is formed from planetesimals composed dominantly of pure magnesium silicates, and lesser quantities of free silica, calciumaluminium silicates, and metallic iron. Without wishing to discard the latter hypothesis, it is interesting to follow up some of the consequences of the more specific model. Accordingly, it is proposed that immediately after its formation, the earth was surrounded by a swarm of two types of planetesimals, (i) planetesimals composed dominantly of iron-poor magnesium silicates, and (ii) planetesimals of primitive origin, left over from the accretion of the earth. Under the influence of collisions and mutual gravitation, it is proposed that the mixture of planetesimals evolved into a “sediment-ring” surrounding the earth, as suggested by dpik and that the moon finally formed by coagulation of this sediment-ring.

Chemical evolution of the terreskid planets

81

Lacking more detailed knowledge of probable ~om~sitio~ of pla~e~s~als, it is not possible to be ape&& about the relative proportions required in the model. It is probable that iron-poor planetesimals would be predominant. For the sake of setting up a specific model, we will assume that the mass ratio of iron-poor to iron rich (primitive) planetesimals was 3 to 1. Whatever the correct ratio might be, it is clear that the model is capable of explaining in principle, the observed density of the moon. This was a principal objective of the present hypothesis. (d) [email protected]~p~~~~~o~ of &%?&ar ~~~T~~~~ in ter?nsof ~~~o~~~ ~~~~~ The basio h~thesis proposed is that much or most of the material now residing in the moon was precipitated during the escape of a dense p~~tive atmosphere. It has been shown that this process would lead to depletion of iron in the precipitate compared to the primordial ~orn~o~tion and oonseque~tly an explanation of the moon’s low density becomes possible. The detailed development of this hypothesis is rather flexible because of the range of unknown physical and chemical conditions under which fractionation may have occurred. The author has suggested a fairly specific chemical model for the nature of the material from which the moon formed but it is not claimed that the model is unique. Numerous variants could be devised. We will now attempt to explain some properties of the moon in terms of the specific model proposed above. (i) F&eagile of the ~0~. Neither the dynamic- figure of the moon as determined from its oblations nor the optically observed figure agree with the eq~~b~~ figure. UBEP (1852,1957a, 1962b) has ~~a~~y emphasized the im~~ance of the moon’s non~qu~b~~ figure as a property to be explained by any acceptable theory of lunar origin. It requires either that the lunar interior possesses subs~ntial long-term strength or that significant deep-seated variations in density are present (UR;EY, ELSASSER and ROCHESTER, 1959). The former condition implies that the temperature in the deep interior of the moon is now, and always has been, far below the melting point. Such a temperature distribution would be inconsistent with chondritic abundances of radioactive elements being present in the moon (UREY, 1962b). The latter condition implies that the moon has never been extensively melted, otherwise it would be expected that the primary density i~omoge~eties presumably estab~shed during accretion would have been eli~na~d. Urey stro~ly favours the view that the moon accreted initially at low temperat~s and has not been generally melted since. The present model is in aeeord with these views. The moon is believed to have acereted from a sediment-ring of cold planetesimaIs. The mean radioactivity of the planetesimals was much less than that of chondrites so that subsequent radioactive heating has not caused a close approach to the melting point except in limited regions (Fig. 6). If the moon contained one half of the primordial abundances of radioactive elements (26 per cent from the primitive planetesimals, and the remhder from the rna~~ia~ silicate plane~~mals} the temperature at its eentre may not now exceed 1000°C This is sufficiently far below the melting point to permit the retention of long-term strength. It is also usable that the he~rogeneoug Constitution of the moon, as implied in the model, cont~bu~s to the non-eq~lib~um bulge.

88

A. E. RIX~WOOD

(id)&rese hktorg ad tlt-mnd hi&my. BAL~FVIN (1063) has discussed the shapes of lunar craters as a function of age. He showed that the older pre-mare craters were systematically shallower than younger pre-mare craters, and possessed smaller rims. BALDWINinterpreted these observations as implying that isostatic adjustment was operative in the lunar “crust” during formatian of the earlier craters, but the isostatic: response of the crust decreased with time, so that little orno adjustment occurred after the younger craters had formed. Accordingly he suggested that a substantial increase in the meanstrength of the lunar crust had occurred over the time interval represented by the visible pre-mare craters. The increase in strength was attribu~d to a decrease in the mean temperature of the crust over this time interval. In view of the long time constant for thermalconduction processesin large depthsof rock, he concluded that the formation of the visible craters of the moon had occupied an extended time interval, This conclusion does not appear unreasonable-presumably the rate of aocretion from the parental sediment ring deoreased substantially during the terminal period of the moon’s formation because of depletion of planetesimds in the neighbourhood and consequent lower collision probabilities. Accordingly, following BAL~~ZN, we establish our first boundary condition for the thermal history of the moon: during the earliest “recorded” phase of lunar history as represented by the pre-mare craters, the outer crustal regions of the moon were cooZ~~g,and as a result, the crust developed long-term strength which permitted it to support until today, the observed large differences in surface elevation. Most workers regard the lunar maria as of volcanic origin. Detailed studies of the lunar surface have revealed many other features presumed to be of volcanic origin (KUIPER,106Oa; SHOEMBXJER, 1062). The extensive internal melting implied by the maria is difficult to reconcile with UREY’S strong argument that the moon’s noneq~librium bulge implies a cool interior. UREY attempts to avoid this difIiculty by assuming that the maria were formed by impact melting but the observations of BALDWIN(1063) and SHOEMLBEER (pers. comm.) argue against this. BALDWINshowed that the maria were formed by the flooding of pre-existing craters and that in many cases, a large time interval had elapsed between formation of the crater and its filling by lava from below. Thus to account for the maria we are forced to conclude that the temperat~es of extensive regions within the moon were ~~cre#~~g from the time of its formation until, at a later period, large scale melting occurred and the maria were erupted. If we are to reconcile cooling of the outer regions of the moon soon after its formation with heating of the deep interior on a longer time scale sapiently to cause melting and at the same time to retain sufficient strength in the deep interior to support a non-equilibrium bulge, it is clear that a highly specific thermal history must be co~t~c~d, We should not be surprised if the results appear contrived. The alternative is to discard enough of the apparently self-contradictory observations and inferences to permit a ‘simple’ thermal history to be oonstructod. We shall attempt to pursue the former course. It is suggested that the gravitational energy of accretion played an important role in the moon’s thermal evolution. (See also BALD~XN, 1063, p. 306.) The mean gravitational energy dissipated during formation of the moon is approximately 400 Cal/g, baaed upon a free fall model. It is possible that the mean energy was

Chemical evolution of the terrestrial planets

89

snb~~~y higher than this if collision velocities exceeded escape velocities during coagulation of the sediment-ring. As iu the case of the earth, the gravitational energy liberated during the early stages of the moon’s accretion would be relatively small so that a cool central nucleus would form. However, during the later stages of accretion, the energy liberated may be in the range 60~1000 Cal/g, de~n~g upon the additional orbital kinetic energy which may be contributed according to the location and velocity of the planetesimal in the sediment ring. The initial temperature ~st~bution within the moon will depend upon the proportion of gravitational energy which is retained thermally. Detailed discussions of meteoritic impact mechanics by SHOEMAXER (1960, 1962, 1963) indicate that a substantial portion of the energy of infalling bodies is converted into shock waves which are dissipated t~oughout a relatively large volume of the parent body. This energy is thus effectively conserved and contributes to heating. It is suggested as a working hypothesis that on the average perhaps 20 to 30 per cent of the energy of accreting planetesimals is retained in this way. The temperature of the moon after accretion may therefore increase from zero at the centre to approxima~Iy 1200°C near the surface. * Following BALDWIN(1963) the final stages of accretion of the moon are believed to have occurred over an extended time interval, so that ~~ifi~ant cooling of the near surface region by conduction occurred. Thus a temperature distribution qualitatively similar to Curve 1 of Fig. 6 might be produced. This is characterised by a sharp maximum near the surface caused by the effect of conduotion on the initial temperature distribution established by retention of ~a~tational energy during accretion. The mean near-surface temperatures are initially high and the crustal region is correspondingly weak. This stage coinoides with the formation of the earliest visible craters on the moon, whilst the crust was still weak enough to deform isostatically. However the sharp temperature maximum decays by conduction comparatively rapidly, and it is possible that over a time scale of 108years the crustal regions became sufficiently strong to prevent the latest craters regaining isostatic equilibrium. This would account quali~tively for BALD~IN’Sobservations. The temperature maximum thus established would then decay very slowIy and flatten. Superimposed on this pattern would be a general heating caused by radioactivity. We have previously sugges~d that the moon may contain onlyone half of the primordial abundance of radioactive elements. Over a time scale of IO*years, this would cause a general heating throughout the moon, except in near surfaoe regions which would continue to cool. It is oonceivable that this heating may be sufficient to push the ~rn~rat~e rna~~ through the melting point ~st~bution (Curve II, Fig. 6). This would cause extensive partial melting, mixing, and differentiation at depths of several hundred kilometers. The maria may have resulted from this magmatic phase. At the same time, heating in the deep interior would not have raised the temperature more than 600°C. This region would therefore retain its strength, and primary density and compositional inhomogeneities would remain. The succeeding thermal evolution would consist of further slow cooling of the surface * The proportion of gravit&ionrtI energy which was captured thermally may have bet+u greater in the early stages of accretion when the acmretionenergies were low, than in the later strtges.

90

A.

E.

RINI~WOOD

regions and slow heating of the deep interior, which however is still well below the melting point (Curve III, Fig. 6). The broad temperature maximum would lie mainly below the melting point. However primary inhomogeneities in dist~bution of radioactive elements implicit in the model Gould result in localised hot spots and intermittent volcanic activity persisting until the present day.

0

1000

500 DEPTH,

1500

km

Fig. 6. Suggeeted temperature distributions within the moon at differentstages of ita history. I. Immediately after accretion. II. lo5 years after formation III. Present day. Curve MP is the probable melting point distribution with depth.

The above thermal history is admittedly of ad hoc nature and carefully tailored to produce the desired effeots. Nevertheless, it appears capable of qualitatively sat~f~g the evidence produced by BALDWET. The major a~~~tio~ conveys the possible role of gravitational energy in the lunar thermal history and this is open to quantitative investigation. The implication that the maria are one or two billion years yormger than the uplands is supported by the latest crater oounts of BALDWM (1963) and KUIPER(1966). One further property of the model may be noted. If the primitive planetesimals in the deep interior have not been strongly heated they may have retained their volatile components. On the other hand, in the outer regions of the moon, subsequent heating in most cases would have led to loss of volatiles and partial reduction. Accordingly the density of the primitive planetesimal component of the moon is likely to &cpeme with depth. If this effect is not oouuterbalaneed by a substantial increase iu the density of the iron-poor pl~e~im~ oorn~~nt with depth, which would not nomaHy be expeoted, then the mean density of the moon should decrease with depth. The observed moment of inertia of the moon is in acoord with this inference (JBUTFREYS, 1961; MACDONALD,1962) although the uncertainties in determination of the lunar moment of inertia render this agreement of doubtful signifioance.

Chemicalevolutionof the tmmetria-l planets

91

(iii)Burfuce heterogeniety.FIELDER (1961) and BALDWIN(1963) have reviewed the extxmsive evidence for widespread variations in colour and albedo of the lunar surface (see also KUIPER, 1966). BALDWIN concluded that the material from which the moon is now constituted is non-homogeneous. According to the present model, this may refleot the original compositional heterogeneity of the material from which the moon accreted, i.e. 4ron-poor plane~simals and primitive iron-rich planetesimals which have been subjected to varying degrees of autoreduction. Because of the net depletion of the moon in radioactivity according to the model, complete melting of the moon has not occurred and hence much of the original comp~itional heterogeneity has been preserved. (iv) Emission of gases. KOZYREV (1962) has reviewed the evidence for emission of gases conning complex moleoules from the crater Alphonsus. According to the present model, the moon is a heterogeneous mixture varying between primitive, highly oxidised material similar to Type 1 carbonaceous chondrites (subsequently metamorphosed) and highly reduced material dominantly consisting of magnesium silicates, but probably also containing small amounts of other minerals, e.g. calcium sulphide and iron silicide. On a large scale, the moon is thus composed of material which is grossly out of chemical equilibrium and any subsequent heating may well cause the formation of complex volatile compounds, (v) Luminescence. LINK (1662) has reviewed evidence for the emission of excess light from the moon during eclipses and suggested that it is caused by a general ~u~nescen~e of the lunar surface excited by the solar corona. Further evidence in favour of widespread lunar luminescence was obtained by DUBOIS(1956,1957,1959) and KOZYREV (1956). See also GRAINIERand Rnva (1962). Very reoently, decisive evidence of lunar luminescence probably excited by solar particle radiation was obtained by KOPU and RACKHAM(1963, 1964). Further cases of luminescence were reported by GREENACREand BARR (1963) and SPINRAD(1966). These observations may provide a vital clue to the chemical composition of the moon’s surface. KOZYREV (1956) pointed out that the lunar phosphor contained negligible amounts of combined iron because iron was an active extinguisher of luminescence. Furthermore KOZYREVargued that the luminescent material in the rays of Aristarchus had a large albedo similar to that of white sand. DEREAX and GEAKE (1964) examined many meteorites and found that only 3 displayed luminescence. These were all enstatite achondrites, composed dominantly of iron-free enstatite. RENDet a& (1964) aIso found that natural con-contai~ng o~hop~oxene did not luminesce, whereas orthopyroxenes containing no iron showed marked luminescence. These observations strongly support KOZYREV’S conclusion that the lunar

phosphor does not contain chemically combined iron. Bearing in mind DEEHAY and GEA~E’s observations of luminescence in enstatite

achondrites, KOPAL and RACKIUMsuggested that the crater which they observed (Kepler) was formed by the impact of an enstatite achondrite. GBAICE(1964) however suggested the more plausible hypothesis that the primary lunar materiai which was exposed by the impact resembled enstatite aohondrites. According to the present model, iron-free enstatite and to a lesser degree quartz, are probably the most abundant minerals on the moon. Obse~ations of lunar luminescence so far thus appear to provide support for the compositional model. The

92

A. E. RIN~WOOD

results of future investigations of lunar luminescence will be awaited with great interest. According to the heterogeneous moon model, under discussion, primitive carbonaceous chondrite-like material may also be present on the moon. The hydrooarbon component of this material may also be capable of luminescence. However the fact that the observed lunar luminescence (KOPG and RAC~M, 1963, 1964) occurred at a wavelength identical to that at which meteoritic enstatite luminesces supports the hypothesis that the latter mineral was the phosphor. (vi) Jfeteorites from #he moo12 ? To account for the low cosmic ray exposure ages of chondrites, UREY (1959) suggested that they are derived from the moon, Whilst it is probable that most workers still favour an asteroidal origin for chondrites, the possibility of a lunar origin cannot be dismissed in the present state of our knowledge (ARNOLD,1964). A~cor~ngly it is worth while exploring the extent to which a lunar origin for chondrites might be ~on~s~nt with the origin of the moon as postulated in this paper. The author has previously argued that ordinary ohondrites were formed when primitive material similar in composition to Type 1 carbonaceous chondrites were subjected to autoreduction at high temperatures, resulting in the production of a metal phase in s&u,and loss of volatiles. In section 3, it was suggested that the reduction process occurred in parent bodies heated by extinct ra~oaoti~t~s, Now in the model for the formation of the moon proposed herein, it has been suggested that the moon accreted from an approximate 1 to 3 mixture of primitive carbonaceous chondrite material with magnesium silicates. Could ordinary chondrites have evolved from the parental carbonaceous chondritic material on the moon? To the author this appears ~st~ctly po~ible. Two models are conceivable. The parent bodies of the chondrites might be identified with the primitive Type 1 carbonaceous chondrite planetesimals which accreted on the moon. If these bodies were sufficiently large-with dimensions of the order of 100 km, they might have umlergone an independent thermal evolution if they were not completely dispersed when they were incorporated within the moon. The evolution of these included bodies would, in fact, be similar to that proposed in section 3 for the independent parent bodies of chondrites. It would be necessary to assume the presence of short-lived radioactivities as the heat source. It may be desirable to attempt to construct a model which avoids the use of extinct radioactivities. This leads us to reconsider earlier suggestions by UBEY(1966, 196?‘b, 1958,1964) that collisions played an important part in the formation of ehondrites (see also FREDRIKSSON, 1963). It may be possible to combine key aspects of the hypothesis of UREY and the author by suggesting that planetesimals of Type 1 carbonaceous chondrite material were converted into ordinary and enstatite chondrites by impact phenomena when they fell on the moon’s surface during the final stages of accretion. The impact caused shock melting of the plan&esimals aeeompanied by reduction and loss of volatiles. The ~s~ption caused by escaping gases might have caused the formation of c-hondrules in an analogous mszmer to that suggested by FREDRIKSSOPI and RIN~WOOD(1963). This hypothesis appears to satisfy the major requirement that ordinary choni by t-qmmss of oarbon reduction in a oondemd drites formed from primitive

Chemioal evolution of the femestrial planets

93

environment (section 3d). The problems of chemical fractionation among the chondritegroupsare much more difficult. Previously the author (196fia) has suggested that these ~aetionatio~ were caused by selective solubility of various trace elements in supercritical, dense solutions of H,O, H&Sand CO, which were expelled during heating in the parent body. In the present model, about 30 per cent of the mass of the carbonaceous ohondrite must also be lost as a gas phase consisting of similar components together with CO and H,, during the heating caused by impact with the moon. It is possible that this gas phase, which would initially be compressed to high deusities acted as a carrier for the elements which were lost from the primitive material. 6. MERCIURY KUIPER(1962) gives the density of Mercury as 6.46 g/cm*. The slightly smaller value of 5.33 is obtained if the radius recently selected by DE VAUCOULEURS (1964) is used. Because of the relatively small mass of Mercury, the: density is not greatly affected by self compression and the uncompressed density is probably close to 5.2 g/cmS. UREY (1952, 1957a, 1963) has repeatedly emphasized that this high density implies the presence in Mercury of a larger proportion of iron than occurs on other planets. There does not appear to be any escape from this oonclusion. The high density and inferred abundanoe of iron is probably connected with the fact that Mercury is the closest planet to the sun. HAYASHI(1961) and EZER and CAMERON (1963) have shown that the sun passed through a fully convective stage during its contraction, resulting in a greatly increased luminosity, Accordingly it is possible that the temperature in the vicinity of the orbit of Mercury may have been as high as 1500°C. Thus the condensed materials from which Mercury later formed would have separated from p~mor~al solar material at an initially high ~rn~rat~e, whereas all of the other planets formed from solid mat8rial which separated from the nebula at a very low temperature It is this difference which is probably responsible for the high density of Mercury. The condensed phases in equilibrium with gases of solar composition around 1500°C would be highly reduced, containing no oxidised iron. Furthermore, 8s shown by UREY (1952, 1984) selective reduction and volat~ation of silicate components would probably occur under these conditions, leading to an increased conoentration of iron in the non-volatile, condensed phase in the nebula. Thus the density of Mercury may be explained by selective volatilisation operating in the opposite sense to that invoked (section 5) in the case of the moon. From Fig. 5 we see that on this h~thesis Mercury would be depleted in silicon and magnesium and co~spon~gly etiehed in iron, calcium, aluminium and less volatile elements such as uranium and thorium. Thus it is possible that Mercury is composed principally of iron (containing some silicon in solution) and calcium and aluminium silicates. Elements with volatilities greater than silicon and magnesium, e.g. the alkali metals, would have been completely lost. Certain implication of this model might be mentioned. Frm (1963) has pointed out that the side of Mercury facing away from the sun is warmer than would normally be expected and has suggest8d that heat is convected from the hot sunlit side to the cold side via a tenuous argon atmosphere. The present model is not consistent with

94

A. E.

&NCtWOOD

this hypothe& Potassium which is the source of the argon is believed not to be present in kfercury in any sig~ficant quantity because of its high volatility, Loss of po~ssi~ would affect the thermal history of Mercury. Its loss would be partially offset by an increased relative concentration of uranium and thorium, caused by depletion in magnesium and silicon. Aocording to the model, the accretion temperature was high and the amount of uranium and thorium would probably have been s&Gent to cause widespread melting. This would be followed by separation of a core, as in the earth, with an outer mantle of al~nium and calcium silicates. It is probable that strong upward concentration of uranium and thorium occurred, as a result of crystallization differentiation. Because of this, together with the relatively small size of Mercury, its high core-to-mantle ratio, and the high thermal conductivity of the core, it is probable that the planet has been cooling slowly at all depths since melting and differentiation, and that the interior is now well below the melting point at all places. Appreciable internal long-term strength would thus be expected. 7. &tARS The diameter of the solid surface of Mars is subject to some unoertainty because of atmospheric obscuration. DEVAWC~UL~~S (1964) has recently reviewed all available data and selected a mean value of 3366 km for the radius. This value, together with the mass of 0*1069 earth masses, yields a mean den8ity of 4.00 g/cm3. When the effects of self compression and phase change8 are considered, it is found that on any reasonable assumptions, the mean uncompressed density of Mars is much smaller than for the earth. JEFFREY~(1937) and UREY (1962) have argued that the smaller density of Mars is due to a smaller content of metal phase compared to silicates. According to JEFFREY’Sand UREY’S model, the mean compositions of metal and silicate phases are similar in the earth and in Mars, but the proportions of these phases differ. According to the author’s model also, Mars contains less metal than the earth. However this is a consequence of a higher mean state of oxidation. According to this model, the in~~dual com~tio~ of metal and silicate phases are di~erent on the earth and Mars, but the overall ratios of iron to silicon and magnesium are the same (RIN~WOOD,1969, section 2 this paper). We will discuss the implications of both of these hypotheses separately. (i) S&ate-metal fractionation. An important property of Mars is the dynamical elliptic&y as obtained from the motions of its satellites. This yield8 a value of 0.0052 for the flattening (~GD~N~D, 1962) assuming hy~ostati~ eq~~b~urn. UREY (1952,1957a) pointed out that when the most probable value of the radius was used, the observed flattening could only be satisfied if Mars wa8 approximately homogeneous, and did not possess a core. Accordingly UREY s.uggested that Mars consisted of a uniform mixture ofiron and silicate phases. ~%UDOXALD(1962) has investigated the internal structures of a large number of Martian models characterised by difIerent assumed values for the radius and flattening. The effects on the models of major phase transformations in silicates, which are known to occur at pressures between 100 and 300 kb were also examined. ~~ACDONALD’S results sub&anti&e the conclusions of UREY. For preferred values of the radius and flattening and for models

~e~~~l

evolution of the terrestrialplm&s

95

containing phase transitions, the surface density is between 35 and 3.8 g/ems. This v&e is higher then the uncompressed density of the earth’s mantle (~3935 gjcm3, CLARK and Rrnawoo~, 1964). Furthermore, in the acceptable M&ian models, a dense core is either absent, or very small (less than 5 per cent of the total mass). Using the radius preferred by DE VA~COULE~S, and in~~ol~t~g between MACDONALD’EI models, a surface density between 3.5 and 3.6 g/ems is obtained. Thus, if UREY’S expl&n~tion is to be employed, a substitntial amount of free metallic iron must be uniformly dispersed thoughout Mars. This, however, raises another problem. MACDONALD (1962,1963) has also investigated the thermal constitution of Mars. He finds that if the radioactivity of ordinary chon~i~s is used, the melting tempe~t~e of iron is exoeeded at shallow depths. This conclusion is reinforced if the uranium and thorium contents of Type 1 osrbonaceous chondrites sre used instead of the values for or~n~ry chon~~s. It now appears that the sbundanees of U and Th in Type I carbonaceous chondrites sre at least twice as high ELS in ordinary chondrites (LOVERIBCJ and MORGAN,1964), Accordingly, on this model it is impossible to understand why the melted iron has not separated into a core. ~~~o~AL~ suggests that the mean abundance of radioactive elements is substantially smaller in Mars than in chondrites. T.he csnse of this depletion of ra~oactivity is not readily explicable on the basis of URBY’S model for the interpret&ion of the densities and compositions of the terrestrial planets and chondrites. (ii) Oxidation hpoths&s. RINGWOOD(1969) suggested that Mars is composed of the ~rirnor~~l abundsnces of the common metals which are, however, present in a completely oxidized state. From Fig. 1, we see that the density of such material would be about 3.7 g~cm3which is eonsis~nt with the range of ne&r-s~&~e densities estimated by UREY and MaoDo~&n. Furthermore, RXXWWOOD estimated a value of 4.09 g/cm3 for the mean density of a Martian model constructed of this material. The effect of probable high-pressure phase changes in silicates wss incorporated in this estimate. Co~ide~ng the various sources of un~e~ai~ty, this estimate is in s&tisf~etory agreement with the most probable observed density of 4-00 g/cmz. This hypothesis avoids the diEculty noted by MCDONALD for the metallic iron-silic&te mixture model of UREY. Melting of such completely oxidised material would not result in the formation of a dense iron core and an accompanying strong decrease in density of the outer regions of the planet. The physical inferences of high near surfaoe density and lack of a subst~nti&l core sre thus ~nsis~~t with the melting and diffe~nti~tion of oxidised Martian material, It follows that Mars may contain the primordial abundances of radioactive elements and additional fr~ction&ti~n mechanisms for these elements do not have to be postulated. The oxidised state of Mars is presumably a consequence of the composition of the primordial dust from which the planet acoreted. Apparently the dust contained a rather small prounion of ~~bona~eous material (Rmowoon, 1959). We have afready considered the evidence provided by ~hon~tes (section 3) that wide v~ations occurred in the amounts of carbonaceous compounds which were trapped in the primitive material. The oxidation state of Mars may be about similsr to that of the chondrite Karoondrt, which is composed dominantly of magnetite and olivine ((Mg,.,,Fe,.,&SiO,). Karoonda contains no metal and only L trace of carbon.

96

A.

E.

kNCi\VOOD

Although it is suggested that the parental dust from which Mars formed was comparatively poor in hydrocarbons, it is unlikely that they were completely absent. Mars contains an atmosphere of CO, and a trace of water vapour. These gases have presumably been evolved from the interior, and would be formed by reaction of hydrocarbons with excess oxidised iron under equilibrium conditions. The reactions may have resulted either in the reduction of ferric to ferrous iron, or in the production of a little metallic iron. In the latter case, this is probably present as a very small core. Optical studies of the Martian surface also suggest a highly oxidised state. DOLLFUS (1961) interpreted his polarization measurements as indicating the presence of limo&e as an abundant phase. Furthermore, SHARANOV (1961) has found that the brightness and colour-index of the Martian surface is best matched by limonite. The yellow colour of the clouds of Mars is also most plausibly explained by dust particles rich in ferric iron. These observations are entirely consistent with the model suggested according to which oxidised iron is a major component of the outer regions of Mars. Most probably magnetite would be the dominant iron mineral at substantial depths. At shallow levels and low temperatures, however, magnetite might be oxidised by water being degassed from the interior to form hydrated ferric oxides such as limonite. The hydrogen produced by the oxidation reaction would escape from the planet. 8. VENUS The mass of Venus is 0.8136 of the earth’s mass and its density based upon the optical diameter is 5.06 g/cma (KUIPER, 1952). DE VAUCOULEURS(1964) has estimated the diameter of the solid surface of Venus to be 12,170 km, which implies a mean density for the solid globe of 5.15 g/cm3 compared to 5.52 for the earth. UREY (1952) has remarked that the two planets are so similar that practically any hypothesis of density distribution and origin would fit both equally well. This is certainly true to a large extent. Nevertheless significant differences may exist. It is not certain that the differences in density can be completely explained by differing degrees of self compression of similar material, arising out of the rather small mass difference. It is possible that the low-pressure density of Venus may be slightly smaller than that of the earth. If so, this would imply on the present hypothesis that Venus was slightly more oxidised than the earth with a smaller core to mantle ratio. One important difference between earth and Venus is the absence of a satellite to Venus. This may be a consequence of the mass difference. According to the present model, the moon was formed from material which was fractionated by selective volatilisation in the primitive terrestrial atmosphere during the final stages of accretion of the earth. During this final stage, the accretion rate was rapid and the atmosphere was maintained at a high temprature. It is possible that this stage was not reached in the case of Venus. Although the difference in mass is only about 20 per cent, it might have resulted in very different accretion conditions in the terminal stages. Volatilisation of magnesium and silicon into the terrestrial atmosphere required temperatures which may not have been reached on Venus. Apart from this, we might expect that the formation of Venus from primitive oxidised dust followed a course rather similar to that of the earth. A substantial

Chemicalevolutionof the terrestrialplanets

97

proportion of volatile metals were lost, including alkali metals. Formation of a core probably occurred under non-equilibrium conditions, otherwise it would not be easy to understand the high abundance of carbon dioxide in the Venusian atmo~he~ in preference to carbon monoxide (section 5a). With a similar initial temperature distribution to the earth, and similar amounts of radioactive elements, the subsequent thermal histories of the planets have been similar. It is believed that the surface temperatures of Venus are high (~600°K) because of a greenhouse effect and MACDONALD (1963) has pointed out that the mean temperature in the Venusian upper mantle may be substantially higher than in the earth, and that extensive regions may be at or near the melting point. MACDOXUD suggests that accordingly, solar tidal dissipation in the Venusian upper mantle may have been relatively greater than on earth. As a result, the rate of rotation of Venus has been retarded much more than in the case of the earth. This suggestion is certainly plausible and may be correct. However it is doubtful whether extensive volumes of the Venusian upper mantle would remain stable for long periods in a partially melted condition. Presumably the liquid would collect into bodies of magma which would rise to the surface. Radioactive elements would be selectively concentrated in the magmas and hence the net effect of such a situation over a long period might be a much stronger upward concentration of radioactivity and low melting components in Venus than in the earth. This would permit an internal temperature distribution in equilibrium with the high surface temperature which was below the melting point throughout the upper mantle. For these reasons, the author believes that other possible causes of retardation should be considerede.g. the transfer of angular momentum by magnetic fields to a primitive massive atmosphere which sub~quently escaped. The composition of the Ven~ian atmosphere has excited much interest during recent years. It is known to contain large quantities of carbon dioxide. Nitrogen also is probably an important component. Most discussion has centred on the role of water vapour in the atmosphere. Reports are contradictory, but recent work appears to have established its presence in very low concentrations (DOLLFUS, 1963). It is nevertheless clear that the overall H,OjCO, ratio on Venus is very much smaller than on the earth. Since the origins of the atmospheres on both planets are probably similar, both being of secondary origin, the existence of the large compositional differences have caused much speculation. As in the case of Mars (Section 7) the difference is most simply explained in terms of the nature of the primitive material from which the planet accreted. The varying redox states of the chon~~s were taken to imply that varying amounts of carbonaceous material became trapped in the accreting dust owing to the local conditions of accretion (section 3). Furthermore there is evidence suggesting varying carbonto-hydrogen ratios in the accreting material (RINGWOOD, 1966a). It is therefore quite possible that the material from which Venus accreted possessed a higher ratio of carbon to Ha0 than the earth, This may have been due to a higher mean ~m~rature in the dust cloud before accretion (HOLLAND,1963). Gases evolved subsequently would be expected to reflect this difference.

A. E. RIN~WOOD

98

9. THE GALILEANSATELLITESOF JUPITER The densities and masses of these satellites are given in Table 16 (after KUIPER, 1952). KZTIPER(1956) has plausibly suggested that the low densities of J III and J IV are caused by the presence of substantial quantities of water in these objects. According to the present model these satellites may be very similar to Type 1 carbonaceous chondrites in constitution. Internal radiogenie heating may have driven most of the water into the surface regions but because of the low temperature, the water has not escaped and has remained as ice. Table IS. Densities and masses of G&lean satellites Relative Mass ZM,

Density g/cm”

J1

0.0121

4.03

J II J III J IV

0.0079 0.0260 0.0162

3.78 2.36 2.06

The densities of J I and J II are readily interpreted in terms of different degrees of reduction of primordial material (Fig. 1). On this model, most of the iron in J I would be reduced to metal, whereas only a small amount of reduction has occurred in JII. The systematic decrease in density outwards from Jupiter is an interesting feature and suggests that the reduction and loss of volatiles were controlled by the thermal field es~blished by liberation of gravitational energy during abortion of Jupiter. Thus it is possible that J I and J II secreted at fairly high tem~rat~es from material of solar composition and that reduction was caused by hydrogen rather than carbon. The maintenance of elevated temperatures in the neighbourhood of Jupiter during accretion would be aided by the opacity of the surrounding dust cloud (OPIR, 1962a, 1963c). A~k~~wZ~g~~ent-The author wishes to t.hankRALPEC3. BALDWIN for constructivecommentson t.he manuscript.

REFERENCES AHRENSL. H. (1952) The use of ionizationpotentials I. @xx&~. et Comwchim. Acta 8, I&3-169. ALLEN C. W. (1963) A~~~~~~~~ Q~nt~t~ (2nd ed.), Chapter 7, Athlone Press, London. AXLERL. H. (1961) !Ph A~~~~e of theEking. Interscienee,New York. ANDERSE. (1964) Origin, age and composition of meteorites. Space Sci. Rev. 8, 583-714. .bNOLD J. R. (1964) The origin ofmeteorites as small bodies. Chapter 23 in Ieotopic and G’oamic Chemisary(dedicated to H. C. Urey, edited by H. Craig, S. L. Miller and G. J. Wasserburg), North Holland, Amsterdam. BALDWINR. B. (1963) The Measure of the Moon. University of Chicago Press. BALDWIN R. B. (1964)Lunar crater counts. A&r. J. 68, 377-392. BIRCHF. (1962) Elasticity and co~itution of the earth’s interior. J. ~~~~~3. RM. 62’,227-2116. BIMH F. (1968) Differentiationof the mantle. I&&$.Beol. Sac. Am. 68, 483-486. BIRCH F. (1961) Composition of the earth’s mantle. (feophys.J. 4, 295-311. BIRCHF. (1963) Some geophysicalapplications of high-pressureresearch. Solid8 W&W Pre8mre, Chapter 6, pp. 137-162 (ed. W. Paul and D. M. Warschauer). McGraw-Hill, New York.

Chemical evolution of the terrestrialplanets

90

Baaon F. (1964) Density and composition of mantle and aore. J. aeophy8. R&I. [email protected],4377-4388. Bowzr~ N. L. and Sc~mt J. F. (1935) The system MgG-Fe&SiO,. Amer. J. i&i. I%, 151-217. I&own H. (1952) Rare gases and the formation of the earths i%tmosphsre.ThaAt+-mwphm-eof tlaaEarth and Pkme& (Znd, ed.) Chapter 9, pp, 253-266 (ed. G. P. KUIPER). Chicago University Press. Bw E. C. (lQ52) Discussion of paper by Revelle and Maxwell. Nature, Loswl170, ZOO. B~LL~NK. E. (IB4O)~~problemoftheea~h’sde~ity variation. BwZZ.Se%, Sot. Am. 80,23fl-260. BURXDCJE E. M., BU~BID~E, G. R., FO\NLER W. A. and HOYLE F. (1957) Synthesis of the elements in the stars. Rev. Mod. Phye. [email protected], 546650. C~E~ON A. G. W, (1959) A revised table of abundance of the elements, A~t~~hy~~ 6. m, 076-699. Cahzrc~o~A. G. W. (1963) The origin of the atmospheres of Venus and the earth. 1cum 2, 249-257. CH&.XBERLIX R. T. (1952) Geological evidence on the evolution of the earth’s atm~sph~e, T& Atnw8dmm of the Earth and Pkmets (2nd ed.), Chapter 8, pp. 246-257 (ed. G. P. Kuiper). Chioago Umversity, Press. a J., Tm.~o~ F. C., Goox~w N. and CASKET Q. R. (1954) Activity of silicon in liquid Fe-& and Fe-C-% alloys. Ada Met. 2, 434-460. Calves W. J. (1951) The composition of the solar atmosphere. Bechewhe8 A&r. Obsarv. Utrwht, XII, Part I, l-52. C%AXXS. P. and R~~w~o~ A. E. (I964) Density distribution and ~o~titution of the mantle. Rev. &op?qj8. 2, 36-38. CRAIG H. (1964) Petxologieal and composition&l relationships in meteorites. Isotq& ona! Go8&8 ~~~~~ chapter 26, pp. 401-451 (dedicated to H. C. UREY, edited by H. Craig, S. L. Miller and G. J. Wasserburgf. North Holland, Amsterdam. DARWING. H. (1880) Gn the precessionof a viscous spheroid and on the remote history of the earth. Phil. Trcsns.Boy. Sot. I;ondom171, 713-891. DAR~WNG. H. (1962) Tiae T&&8. W. 11. Friedman, San Francisco. DERXZAX C. J. and GE~KE J. E. (1964) The luminescenceof meteorites. Nature, Lord. 201,62. DoLm~s A. (1961) Polarizationstudies of planets in Plm& mad~~~~~: TlaeSolar #y&em (ed. G. P. Kuiper and B. M. ~~d~~~t~, Vol. 3, p. 379. U%versiQ of Cbioago Press. Do~~rua A. (1963) Observationsof water vapour on Mars and Venus. The Ori& and ~~~~i~ of Atnto8pheresa& Oceana,Chapter 12, pp. 267-268 (edited by P. J. Branoazioand A. G. 717. Cameron). John Wiley, New York. Dmtoxs J. (1956) Peut-on observer sur la lune des phenomen& de l~~nescence? A~r~~~ i+M, 297. Dtmom J. (1957) Sur I”existenoede la luminescencelunaire. J. P&j8. RQdizlna18,135. Dwsors J. (1959) Contribution $ l’etude de la luminescencelunaire. Rozpravg ~~h~~ove~~~. Akad. Ved. 69, l-44. DWAY J. (1957) G&&c Nebulae and Interste&w Matter. Hutchinson, London. Du FRESNEE. F. and ANDERS E. (1961) The record in the meteorites, V, A thermometer mineral in the Mighei carbonaceouschondrite. ~we~~~~.et Co8?n&&m, A&a %3, 200-208, Do- FRESNEE. F. and Armnns E. (1962) On the chemical evolution of the carbonaceouschond&es. Ge~~~~. et C~~eh~~. Acta z2fi,1085-1114. ELSAS~ERW. M. (1963) Early b&tory of the earth. _I&& Science ati ~e~~~~, Chapter 1, pp. l-30 (dedicated to F. G. Hou~a~, edited by J. Geiss and E. D. Goldberg). North Holland, Amsterdam. ENonL A. E. J. (1963) Geologic evolution of North Americe. Science 140, 143-152. Eznn D. and CAZ&ERON A. G. W. (1963) The early evolution of the sun. .Ieam I,422-441. FAIZXNQTON0. C. (1916) Xeteoritea. Published by the author, Chicago. FXElX3G. (1963) The atmosphere of Mercury. T&8 Origin and ~~~~~0~ of Atmwphmm ati Oceans,Chapter 13, pp. 269-278 (editedby P. Brancazio and A. Camexon). Wiley, New York. FXEL~ERG. (1961) Share of the ~oo~‘8 [email protected] Pergamon Press, London. Fxs~ R. A., GALESG. G. and ANDERBE. (1960) The reoord in the meteorites. III, On the development of meteorites in asteroidal bodies. Astrophy8. J. l%$, 243-258.

100

A.

IL

RINGWOOD

FREDRIKESON K. (1963) Chondrulesand the meteoritic parent bodies. Tra?ze.N.Y. Acad. Sk., ser. 2 aS, 756-769. FREDRIKSSON K. and RINUWOODA. E. (1963) Origin of meteoritic chondrules. Cfeochim.et Coemochim.Acta 27, 639-641. FULTONJ. C. and CHIPMAN J. (1954) Slag-metal-graphite reactionsand the activity of silica in lime-alumintiilica slags. J. Melds 11361147. CAST P. W. (1960) Limitations on the composition of the upper mantle. J. [email protected] Res. 65, 1287-1297. GEAKEJ. F. (1964) Lunar luminescence. Nature, Lond. m, 866-867. GOLDBERG L., MILLER E. A. and ALLERL. H. (1960) The abundances of the elements in the solar atmosphere. Astrophya. J. SuppL 45, 5, 1-138. GOLESG. G. and ANDER~E. (1962) Abundance of iodine, tellurium and uranium in meteorites. Geochim. et Coemochim.Acta 25, 723-737. GRAINIERJ. F. and RING J. (1962) The luminescenceof the lunar surfaoe. In Phy& and Astronomy of the Moon, Chapter 10, pp. 385-406 (edited by Z. Kopal). Aoademic Press, New York. GREENACRE J. A. (1963) A recent observation of lunar colour phenomena. Sky S TeZeac. B, 316-317. GREENLAND L. (1963) Fractionation of chlorine, germanium and zinc in chondritic meteorites. J. Geophy8. Ra. 68, 6507-6514. HARRISP. G. (1957) Zone refining and the origin of potassic basalts. Geo&m. et Co8mochim. Acta 12,195-208. HATCHC. G. and CHIPBUNJ. (1949) Sulphur equilibria between iron blast furnace slags and metal. J. Met&e I, 274-284. HAYASHIL. (1961) Stellar evolution in early phases of gravitational contraction. Publ. Astr. Soc.Japan l&460-452. HERBIG G. H. (1962) The properties and problems of T-Tauri stars and related objects. Advances in Astronomy and Astrophysics, (Z. Kopal ed.), pp. 47-103. Academic Press, New York. HOLLANDH. D. (1963) On the chemical evolution of the terrestrialand Cytherean atmospheres. The origin and evolutionof atmoapherea and oceans, Chapter 5, pp. 86-101 (edited by P. J. Brancazio and A. G. W. Cameron). Wiley, New York. HOPKINSD. W. (1954) Physical &em&try and Metal Extraction. Garnet Miller, London. HOYLE F. (1946) On the condensation of the planets. Mon. Not. Roy. Aetron. Sot. 106,pp. 406414. HOYLEF. and FOWLERW. A. (1964) On the abundancesof uranium and thorium in solar system material. [email protected] and CosrrJic Chemktry, Chapter 30, pp. 616-529 (dedicated to H. C. Urey. eds. H. Craig, S. L. Miller and G. J. Wasserburg). North Holland, Amsterdam. HULTGREN R. (1963) Selected valuesof thermodynamic propertie of meti and 02oy8. Wiley, New York. JEFBREYSH. (1930) The resonant theory of the origin of the moon 2. Mon Not. Roy. Astr. Sot. 91,169-173. JEFFREYSH. (1961) On the figure of the moon. Mon. Not. Roy. Astr. Sot. [email protected]& 421-432. JEFFREYSH. (1937) The density distributionsin the inner planets. Mon. Not. Roy. Astr. SOL, Gwphys. Suppl. 4, 62-71. KEIL K. and FREDRIESSON K. (1964) The Fe, Mg and Ca distributionin coexisting olivines and rhombic pyroxenes in chondrites. J. Geophye. Res. 69, 3487-3515. KNOPOFFL. and MACDONALDG. J. F. (1960). An equation of state for the core of the earth. Geophye. J. 8, 68-77. KOPAL Z. and RAOKHAMT. W. (1963) Excitation of lunar luminescence by solar activity. Icarus 2,481-600. KOPALZ. and RACY T. W. (1964) Excitation of lunar luminescenceby solar flares. N&ure Lond. 201,239-241. KOZYREVN. A. (1956) Luminescence of the Moon and the intensity of corpuscular radiation from the sun (in Russian). 1zv. O&m&anA8trc&/8. ob8. 16,148.

Chemical evolution of the terrestrialplanets

101

KOZYR~V N. A. (1962) Physiosl observations of the lunar surface. Phy& and A8trmonzy 01 tfse Moon, Chapter 9, pp. 361383 (ed. 2. Kopial). Academia Press, New York. KWACCHEWSKI 0. and EVANS E. L. (1968) .i%feti~rgti Thermaochemietry. Pergamon Press, London. KXJXPER G. P. (1952) Th.eAtmospheres of the Earth and Pkmets, Chapter 12, pp. 306406 (2nd ed., ed. G. P. Kuiper). Chicago University press. K~PER G. P. (1954) On the origin of the lunar surfaoe features. Proc. Nat. Aeud. Sci. Wash. 40, 1096-1112. KTJI~ERG. P. (1966) The origin of the s&&l&es and the Trojans. V&m in Astronomy, Vol. 2, pp. 1631-1666. (edited by A. Beer). Pergamon Press, London. KUIPER G. P. (1957) Origin, age and possible ultimate fate of the earth. The Pktnet Earth Chapter 2, pp. 12-30, (edited by D. R. Bates). Pergamon Press, London. Vol. 2, pp. 273-312. KWER G. P. (1969~) Exploration of the moon. V&,tasin A8t~o~~t~, (2nd annual astronautical symposium New York.) Pergamon Press, London. KUIPERG. P. (195Bb) The moon. J. Beophya. Raearch 64, 1713-1719. KUIPERG. P. (1963) The surface of the Moon. Space Science, Chapter 15, pp. 630-649 (edited by D. P. Le Galley). John Wiley, New York. KUIPERG. P. (1965) Interpretation of Ranger VII records, Preprint 1965. LATIMERW. M. (1960) Astrochemical problems in the formation of the earth. Science 118, 101-104. Lmx F. (1962) Lunar eclipses. Physics ca-nd A8t~o~~y of the Moort, Chapter 6, pp. 161-229. (ed. 2. Kopal). Academia Press, New York. LOVERINU J. F. (1962) The evolution of the meteorites---evidencefor the coexistenceof ohondritic, achondriticand iron meteorites in a typical meteoritioparent body. Researches on Meteori& pp. 179-198, fed. C. B. Moore). John Wiley, New York. LOWERING J. F. and MOR~AXJ. W. (1964) Ursnium and thorium abundancesin stoney meteorites I. The chondritic meteorites. J. Qeophys. Rea. 62, 1979-1988. MACDONALD G. J. F. (1959) Chondritesand the chemical composition of the earth. Researches in ~eo~~~~~, pp. 476-494, (ed. P. H. Abelson). John Wiley, New York. MACDONALDG. J. F. (1962) On the internal constitution of the inner planets. J. #eophye. Research 67, 2945-2974. MACDONALD G. J. F. (1963) The internalconstitutionsof the innerplanets and the moon. Spam Sci. Rev. 2, 473557. MACDONALDG. J. F. (1964) Tidal friction. Rev. Geophy8. 2, 467-541. MACDONALDG. J. F. and KNOPO~AF L. (1958) The chemical composition of the outer core. @eophy8. J. 1, 284,297. MABONB. (1960) The origin of meteorites. J. (%eophy8. Res. 65,[email protected][email protected] MASONB. (1962s) The clsssitlcationof chondritic meteorites. Amer. Mm. No&t. No. 2085. MASONB. (1962b) Meteorites. John Wiley, New York. MASON B. (1953a) Olivine composition in chondrites. ~wchirn. et Comehim. Aeta 27, 1011-1023. MASONB. (1963b) The carbonaoeouschondrites. Space Sci. Rev. 1,621-646. &bSON B. and Wrrx H. B. (1961) The Kyushu, Japan ohondrite. Gwchim. et Cornchina. Aota 21,272275, MTJELLI~R R. F. (1964) Phase equilibriaand the crys~~~ation of chondriticmeteorites. ~eoeh~~. et Comochim. Acta 22, 189-207. MTME W. H. and DAVIESD. (1964) The relationship between oore accretion and the rotation rate of the earth. f8otrupic and Co.3mict&?&&y, Chapter 22, (dedioatedto H. C. Urey, ed. by H. Craig, S. L. Miller and G. J. Wasserburg). North Holland, Amsterdam. O’KEEFE J. A. (1963) Two avenues from astronomy to geology. The Earth Scierzce.~,pp. 43-68, (ed. T. W. Donnelly). University of Chicago Press. ozprxE. J. (1965) The origin of the moon. I&h Ask. J. 3,24&248 QPIIIE. J. (lQ60) T& Os~~~ng Un&er8e,p. 16. A Mentor book, New Americtm Library, New York. ~)PIKE. J. (1961) Tidal deformations and the origin of the moon. As&. J. 66, 60-6’7.

102

ii.

E.

&NGWOOD

&IK E. J. (1962a) Jupiter: Chemical composition, structure and origin of a giant planet,. Icamls I, 221-223,200-257. ijPIK E. J. (1962b) Surface properties of the moon. [email protected]%+8 its tl&eA~~~~a~~~~ S~~r~~~, Chapter 5, pp. 219-260 (ed. 8. F. Singer). North Holland, Amsterdam. iiPIK E. J. (1963a) Dissipation of the solar nebula. Origin of the Solar &y9&3m, pp. 73-75 (od. R. Jastrow and A. G. L. Cameron). Academic Press, New York. iiPIK E. J. (1963b) Selective escape of gases. Geophys. ,J. 7, 490-509. OPIK E. 5. (19630) Jupiter. 1&h A&. J. 6, 135-149. ~~PIK E. J. and SINGER 8. F. (1957) Reinterpretation of the ur~ium-heIium ages of iron meteorites. !C%a?i8. Amer. Geophys. Un. 88, 566-568. Pon\asc~ S. It. (1963) Tf-lc lower solar corona--t,he abundance of iron. MoY~.Not. Roy. &LZ. sot. 125,543-556. PRIOR G. T. (1916a) The meteoric stones of Launton, Warbreccan, Cronstad, Daniels Kuil, Khairpur and Soko-Banja. &fin. Mw. 18, l-25. PRIOR G. ‘1’. (1916b) On t.he genetic relat,ionship and classification of meteorites. &fin. &fag. 18, 26-44. PRIOR G. T. (1920) Cla~i~e~tion of meteorites. &f&. Bug. 12,51-63. PRIOR G. T. (1953) Cuti~ogue of Meteorites (second edition revised by M. H. Hey). British Museum, London. RAMDOHR P. (1963) The opaque minerals in stony meteorites. J. Geophys. Res. 88, 20112036. RAMSAY W. H. (1951) On the constitutions of the major planets. Mm. Not. Roy. A&. Sot. 111, 427-447. RAMSAY W. H. (1963) On the densities of methane, metallic a~onium, water and neon at planetary pressures. &fo%, Not. Roy. A&r. See. U&, 468-485. REID A. M., BUNCH T. E., COHEN A. J. and POLLACK 8. S. (1964) Luminescence of orthopyroxenes. Nature, Lcmd. [email protected]& 1292-1293. REED G. W., KI~O~III K. and TUKKEVICH A. (1960) Determinations of concentrations of heavy elements in meteorites by activation analysis. Geochim. e8 Comehim. Acta 80, 122-140. RINCWOOD A. E. (1955) The principles governing trace element distribution during ma~matic crystallization I. The influence of electronegativity. Geochim. et Co~~rn. Acta 7,18Q-202. RINGWOOD A. E. (1958) Constitution of the mantle 3. Consequences of the olivine-spine1 transition. Geochim. et Camehim. Acta 15, 195-212. RINGWOOD A. E. (1959) On the chemical evolution and daneitios of the planets. Geochim. el Cosmochim. .&a 15,257-283. RINGWOOD A. E. (1960) Some aspects of t,he thermal ovohnion of the earth. Geoc?$&n.et ~os?no~~~~rn. Acta 20, 241-259. RINGWOOD A. E. (1961a) Chemical and genetic relationships among meteorites. ffwchim.. et Cosmoc?&n. Acta 24, 159-197. RINGWOODA. E. (196lb) Silicon in the metal phase of enstatita ohondrites and some geochemical implications. Geochim. et Coemochim. Acta 25, I-13. RIKGWOODA. E. (1962) Present status of the chondritic earth model. In Ruearches on Meteor&es, pp. 198-216 (ed. C. B. Moore). John Wiley, New York. RI~GWOO~ A. E. (196Sa) Genesis of chondritic meteorites. TO be published. RINC+WOODA. E. (1965b) Composition and origin of the earth. Advances i9+ Earth Sciences (edited by P. M. Hurley). M.I.T. Press, Boston, To be published. RINGWOOD A, E. (1965~) Phases of the mantle. Advances in Earth S&ewes (edited by P. M. Hurley). M.I.T. Press, Boston, To be published. RINGWOODA. E. (1965d) Origin of chondrites. Nature, Losd. [email protected]‘, 701-704. RUBEY W. W. (1951) Geologic history of sea water. Geob.Soo. Am. B&X Sa, Zltl-1147. RUBEY W. W. (1955) Development of the hydrosphere and atrn~pha~ with special reference to the probable composition of the early atmosphere, in ‘“Crust of the Earth” (ed. A. Poldervaart). cfeol. Sot. Am. Spec. Paper 82, 631-650. RUNCORN R. K. (1962) Palaeomagnetic evidence for continental shift and its geophysical C~USB~ ContilzedaE Drift, Chapter 1, p. l-39 (ed. 8. K. Runcorn). Academic Press, New York.

Chemical evolution of the terrestrialplanets

103

RUNCORNS. K. (1964) A growing core and a convecting mantle. I~otq~ic ati Got&c Chemistry, Chapter 21 (dedicated to H. C. Urey. Ed. by H. Craig, S. L. Miller and G. J. Wasserburg). North Holland, Amsterdam. Scwa~ R. A., Sr,rrrnR. H. and OLEW D. A. (1963) Cadmium abundances in meteoritic and terrestrial matter. Owoibim.et Co8mochim. Acta 2’7, 1077-1088. SHARANOV V. V. (1961) A lithologicalinterpretationof the photometric and calorimetricstudies of Mars. A8tr. Zh. 88, 267. SEIOEXAEER E. M. (1960) Penetration mechanics of high velocity meteorites as illustrated by Meteor Crater, Arizona. Rept. 2% SessionInternat. Geol. CongressNorden, Part l&418-434. SEOE~~ER E. M. (1962) ~te~re~tion of hmar craters. P&/Sk% and A8tPorumzy of the [email protected], Chapter 8, pp. 283-359, (edited by Z Kopal). Academic Press, New York. SHOEMAXERE. M. (1963) Impact mechanics at Meteor Crater, Arizona. !I%e Solar System, Vol. IV, Chapter 11, pp. 301-336 (edited by G. P. Kuiper and B. Middlehurst). Univ. of Chicago Press, Chicago. SCALPSA. A., HUGHEST. C., MAPPERD., MCINNESC. A. and WEBSTERR. K. (1964) The deviation of rubidium and eaesium in stony meteorites by neutron activation analysis and by mass spectrometry. ~w~~~. et Co~h~m. Acta a8,209-233. SPINRAII H. (1964) Lunar lmninescencein the near ultraviolet. Icow 3, 500-501. SUESSH. E. (1949) Die Haufigkeit der Edelgas auf der Erde und im Kosmos. J. Cfeol.57, 600-607. SUESSH. E. (1962) Thermodynamicdata on the formationof solid carbon and organiccompounds in primitive planetary atmospheres. J. Geophy8. Res. 67, 2029-2034. SUESSH. E. (1964) The Urey-Craig groups of chondrites and their states of oxidation. ~80~0~ and CosmicCornets, Chapter 25, pp. 385-400 (dedicatedto H. C. Urey, edited by H. Craig, S. L. Miller and G. J. Wasserburg), North Holland, Amsterdam. SUESSH. E. and UREY H. C. (1956) Abundances of the elements. Rev. Mod. Phys. 28, 53-74. SZTROKAY K. I., TOLNAYV. and FOLDVARI-VOOL M. (1961) Mineralogicaland ohemicalproperties of the carbonaceousmeteorite from Kaba, Hungary. Acta Geol. 7, Fl-2, 57-103. TAYLORS. R. (1964a) Trace element abundances and the chondritic earth model. C7eochim. et Co8~oc~~rn.A&a 28, 1989-1999. TAYLORS. R. (1964b) Chondriticearth model. Nature, Land. 209,281-282. TAYLOR S. R. (19640) Abundance of chemical elements in the earth’s crust-a new table. Geochim. et Ooemochim. Acta 28, 1273-1285. TER HAAR D. and WERGELANDH. (1948) On the temperature of the earth’s crust. KgZ Nor8ke Videvak Se&k Forh. 20, 52. UREY H. C. (1952) The PEccnets. Yale University Press, ?;TewHaven. UREY H. C. { 1953) ~~c~~n on Near Processes in Geology S~t~~ge, p. 49. National Research Council Publ. 400, Washington, D.C. UREY H. C. (1954) On the dissipation of gas and volatilized elements from protoplanets. Astrophy8. J. Suppl. I, 147-173. UREY H. C. (1956) Diamonds, meteorites and the origin of the solar system. Astrophy.% J. 124,623-637. UREY H. C. (1957a) Boundary constitutions for the origin of the solar system. P!~yaiceand Cornets of the Barth, Vol. 2, pp. 46-76 (ed. L. H. Ahrene, F. Press, K. Rank&ma and S. K. Runcorn). Pergamon Press, London. UREY H. C. (1957b) Meteorite8 and the Origin of the Solar Sy8tem, pp. 14-29. 41st Guthrie Lecture, Yearbook of the Physical Society, London. UREY H. C. (March, 1958) The early history of the solar system as indicated by the meteorites. Proc. Chem. Sot. 67-78. UREY H. C. (1959) Primary and secondary objects. J. Geophye. Res. 64, 1721-1737. UREY H. C. (1960a) Lines of evidence in regard to the composition of the moon, In &XXX Research,pp. 1114-1120 (ed. H. K. Kalhnan-Bijl). North Holland, Amsterdam. UREY H. C. (1960b) On the chemical evolution and densities of the planets. Geochim. et Cosmochim. Acta 18, 151-153.

104

-4. B. RINOWOOU

UREY H. C. (1962a) Evidence regarding the origin of the earth. Beochim. et Coemochim. .4ctn 26, 1-13. UREY H. C. (196213) Origin and history of the moon. l%y&cs and Aetron.omy of the MOOH. Chapt,er 13, pp. 481-523, (edited by Z. Kopal). Academic Press, New York. ~JEEY H. C. (1963) The origin and evolution of the solar system. Space &Xerze, Chapter 4, pp. 123-168, (ed. D. P. LeGalley). John Wiley, New York. UREY H. C. (1964) A review of atomic abundances in chondrites and the origin of meteorites. IZe?;. Geophys. 2, l-34. UREY H. C. (1965) Meteorites and the moon. Science 147,1262-1265. UREY H. C. and CRAIO H. (1953) The composition of the stone meteorites and the origin of the meteorites. Geochim. et Cosmochim. Acta 4, 36-82. UREY H. C., ELASSER W. M. and ROCHESTER M. G. (1959) Xote on the internal structure of the moon. Astrophys. J. 129, 842-848. DE VAUCOULEURS 0. (1964) Geometric and photomet,ric parameters of the terrestrial planets. Icarw 3, 187-235. VINO~RADOV A. P. (1961) ‘The origin of the material of the earth’s crust,. Communication I, Geochemistry, U.S.S.R. (English transl.), l-32. WAGER L. 12. (1958) Beneath the earth’s crust. Aduanc. Sci. 58, 1-14. (Presidential address to Section C, British association for the Advancement of Science.) WAIIL W. (1910) Beitrage zur Chemie der Meteoriten. 2. Anorg. Chem. 89, 52-96. WAHL W. (1952) The brecciated stony meteorites and meteorites containing foreign fragments. Geochim. et Cosmochim. Acta 2, 91-l 17. WIIK H. B. (1956) The chemical composition of some stony meteorites. Oeochim. et Cosmochim. Acta 9, 279-289. Planets and Satellites, Chapter 5, pp. 159-212, Vol. 3 of WILDT R. (1961) Planetary interiors. The Solar System (ed. G. P. Kuiper and P. M. Middlehurst). Univ. of Chicago Press. WILSON J. T. (1954) The development and structure of the crust. The Earth a8 a Planet (ed. (G. P. Kuiper), pp. 138-214. Chicago University Press. WISE D. U. (1963) An origin of the moon by rotational fission during formation of the earth’s core. J. Geophys. Res. 68, 1547-1554. WOOD J. A. (1962) Chondrites and t)he origin of the terrest,rial planets. Nature, Lond. 194, 127-130. WOOD J. A. (1963a) On t,he origin of chondrules and chondrites. Icarus 2, 152-180. WOOD J. A. (1963b) Physics and Chemistry of meteorites. The Moon, Meteorites, and Comets, The Solar System IV, Chapter 12, pp. 337-401, (ed. P. M. Middlehurst and G. P. Kuiper). University of Chicago Press. APPENDIX The solar iron abundance ALLER~ has recently revised the abundances of many elements in the sun on the basis of new f-value determinations. The abundances of several elements have been decreased whereas the are compared with the abundance of iron has slightly increased. r--3 When the revisedabundances Type I carbonaceous chondrite abundances, excellent agreement (within the limits of error of abundance determinations) is now secured for most elements incZud&g iron. One of the few remaining discrepant elements (by a factor of 3) is silicon, the abundance of which is poorly determined.l Others are Sn, Yb, Y and S, which are also poorly determined. It appears that silicon may be a most unsatisfactory element for normalization in abundance calculations. The removal of the discrepancy in the iron abundance between chondrites and the inner planets on the one hand, and the sun on the other, is clearly of great importance and reinforces the basic thesis of this paper. I am grateful to Dr. ALLER for sending me a copy of his revised abundances in advance of’ publication. 1 ALLER L. H. in Advances in Astronomy and Astrophysics, (ed. Z. Kopal), in press. 2 ALLER L. H., O'MARA B. J. and LITTLE S. (1964) Proc. Nat. Acad. Sci. 61,1238-1243. 3 MULLER E. and MUTSCETLECNER J. P. (1964) Astrophys. J. Suppl. 9, No. 85.