Cooling rates of 27 iron and stony-iron meteorites

Geochimica et Cosmochimica Acta,1967,Vol. 31,pp. 1001 to 1023. Pergamon PressLtd. Printedin Northern Ireland

Cooling rates of 27 iron and stony-iron meteorites JOSEPH I. GOLDSTEIK Geochemistry Laboratory, Laboratory for Theoretical Goddard

Space Flight

Center, Maryland

Studies, 20771, U.S.A.

and JAMES M. Planetology

SHORT*

Branch, Space Sciences Division, Ames Moffett Field, California, U.S.A. (Received

25 July

Research

Center,

1966)

Abstract-A method for the precise determination of the cooling rat,cs of iron and stony-iron meteorites has been developed and applied to 18 irons, 8 pallasites, and the siderophyre Steinbach. The cooling rates are determined by comparing the measured Ni gradients in several kamacitetaenite areas which result from the growth of the Widmanstiitten structure between 700 and 300°C with gradients calculated by a theoretical growth analysis. The theoretical diffusion-growth analysis model shows qantitatively that the imperfecti correlation between kamacite band width and bulk Xi content is due to variations in cooling rates. It is suggested that a classification system based on both kamacite band width and bulk Ni content should be used so that variations in cooling rates from one meteorite to another can be clearly seen. A very wide range of cooling rates was found for the 2’7 meteorites studied (0.4-40°/106 yr). Three approximate groupings were found: (1) fine octahedriees, Ge (IVa), 9-40”/106 yr, (2) fine and medium octahedrites, l-5”/106 yr, (3) pallasites, 0.4-1.0”/106 yr. It can be seen from these groupings that various iron and stony-iron meteorites developed their Widmanstatten structure in different thermal environments, although each meteorite was insulated within a parent body to a temperature less than 350°C. The differences in cooling rates may be due to either varying depth of burial within one or several parent bodies or within the core of several parent bodies. However, not all of the iron meteorites could have cooled within the core of one parent body. 1. INTRODUCTION

SIWCE von WidmanstBtten

discovered the remarkable crystalline structure present in iron meteorites, which has come to be known by his name, much effort has been devoted to learning more about this structure. The Widmanstgtten pattern is formed by the nucleation of the kamacite phase preferentially along the oct’ahedral planes of the parent taenite phase. The size of the pattern, that is the average width of the kamacite phase, differs from one meteorite to another. Iron meteorites are classified texturally into three main groups: hexahedrites, Hexahedrites are one phase kamacite, retaining no octahedrites, and ataxites. Widmanstgtten structure. Octahedrites show a WidmanstBtten structure and are further classified according to the size of the kamacite plates which make up the Widmanstgtten structure (coarse to fine). Ataxites with less than 30 wt.% Ni normally have microscopic kamacite plates in an octahedral orientation, although * Present address: U.S. Geological Slwq-,

Branch of Analytical

California.

1001

Chemistry,

Menlo

Par:;,

1002

and J. M. Smm

S. 1. GOLDSTEIN

the term Widman8titten struoture is not normally applied to this Joealized texture, The Widma~~t~n structure is al80 found in the sub-group of stony-iron meteorites called the pa&&es, especially those meteorites with large metallic areas. Generally, the band width of the kamscite phase decreases with increasing Ni eontent of the meteorite. Thus, the hexahedrites and coarse octahedrites can also be identified as separate group8 based on their Ni contents. However, the Ni

r

I

I

/

I

I

I

900

Ni-Fe

PHASE DMRAM

IGOLOSTtlN

8

OQltV1,

19651)

8C@-

700y

TAEIIITE

600-

JOO) 400-

m ~COARSE

/

0

OCTAHLORITES MEDIUM

Co@

OCTAl+EDRITES

al

]

flNE

(om)

OCTAWEORITES

fofJ

ZT

TAENRIC PALLASITES

0

tP8

I

I

I

I

i

I

5

10

I3

20

30

40

WT. Fig.

1.

BTi-Fe

phase

diagrwn

Jso

YZNi

and

meteorite

&WS~S.

contents of the medium and fine octahedrites and the ataxities overlap (Fig. l), as do the Ni aontents of the metallic phruse8of ths pallasites. Thus, a meteorite of 10 wt.% Ni may have a structure of a medium or tie octahedrite, an ataxite, or a pallasite. The imperfect correlation between Ni content and structure was pointid out by GOLDESERG et a& (19H) and di8uussed more recently by essay (1962), SHORT and ANDE~ESOX (1965) and REED (1966). All these authors suggest that the imperfect correlation can be best explained in terms of the cooling rates of these meteorites. For example, the faster a meteorite of a given composition cools, the finer its Widmallstiitten pattern will be, since 14388time is available for the growth of the kamacite phase. In other word8, the cooling rate, as well as the Ni content,

I MI LAC,

0 Fig.

40 2. Structure

and composition

80

PALLASITE

120 DISTANCE, p

of the metallic

160

phases of the pallasite,

240 Imilac.

1002

Cooling rates of 27 iron and stony-iron meteorites

1003

In addition, the cooling rate is the determines the &ructure of a meteorite. important factorin determining under what conditions meteorites formed. It is therefore the purpose of this paper to show how cooling rates of meteorites can be uniquely determined, to report our findings on 27 iron and stony-iron meteorites, and to make use of these cooling rates to explain some of the features and relations we find among meteorites. 2. METHODS FOR DETERJXIKING COOLING RATES (a) I~tro~~c~~o~~ To determine a cooling rate for an iron or stony-iron meteorite, we first measure with the electron probe the Ni concentration gradients in kamacite and the surrounding taenite for several oriented bands of the Widmanst~tten pattern. Secondly, we calculate Ni concentration gradients in kamacite and taenite for the meteorite of interest using a computer-growth analysis program. This is done for several proposed cooling rates. Thirdly, we compare the measured and calculated The cooling rate is determined when a unique fit is concentration profiles. established between the measured and calculated concentration gradients. The cooling rate determined is then established for the temperature range 700-300”C, the temperature range during which the Widmanstatten pattern forms. The individual steps in the analysis of cooling rates are discussed in detail in the following sections.

The selection of meteorites was partially limited by the capability of the methods used. The Widmanstatten structure develops to varying degrees of perfection in different classes of meteorites. In fine and medium octahedrites, the structure is best developed, often with excellent planefront regularity in the octahedrally oriented kamacite band systems. Individual kamacite platelets can also be found in octahedral orientation in many ataxities. In coarse octahedrites, impingement of adjacent kamacite phases on the intermediate taenite phase makes diffusion analysis difficult. All pallasites have kamacite rimming the non-metallic phases, and the larger interior metallic areas sometimes show a Widmanst~tten texture or isolated kamacite plates (Fig. 2). We also attempted to select meteorites whose Xi-content lay toward the extremities of their classification group (e.g. Tazewell and Huizopa) and meteorites with similar N-contents but different structures. In addition, the siderophyre Steinbach, which contains the low pressure mineral tridymite (MASON, 1962) was also selected. Thus the 27 irons and stony-irons investigated are not a random sample. The meteorites studied are listed in Table 1. (c) Sample preparation Unweathered surfaces of etched sections of the meteorites were examined and a section was made perpendicular to one band system. This section was polished through 2 p diamond and the original band system was identified with microhardness indentations. Then electron microprobe scans were made perpendicular to these kamacite bands; i.e. in the direction of diEusion. Straight bands were selected and profiles were taken perpendicular to the bands and away from interfering band systems or inclusions, such as schreibersite (Fig. 2). In the case of the pallasites, care was taken to stay away from the rimming kamacite

Spearmarl

7.81 8.6 X-J

12.3

X2.66

Elualapai (Wdlapaij Huizopa

Off

Edmonton

10.0 9.63

9.6

Of

Duchesne

18.1

GlY.&ht;

Off-D

Ik@on

10.5 12.68

7.96

Om Off

Garbo CEbrhlton

16.0

Gibeon

Off-D

Butler

8.3 8.32

9.7

Of

Bristol

843 8.56

Four Corners

Of

structura

Altonah

Irons

Meteorite

Bulk Ni content ql w* %t

E. P.

W;.

J.F. et al. (1957) ‘VI;ASSON J.T. (PC.) and M.P.A. M.P.A. LOVERIN

P,OVE.WNG

J. P. ebal. (1957)

GOLDBER~E. et aE. (1961)

(1958)

NIOHIP~RVK

lxsxD~S0N (P.C.) M.P.A.

CRESSY P.(P.c.)2 GOLDBBRUE. et al. (1951)

\vASSON

J.T. (1966s)

(1958)

~IOEIPORUK

Ti'v.

F. et al?.(1957) M.P.A. hWl%ItINU~.

M.P.A.1

Reference

Table 1. Meteorites selected for analysis

0.122 I\-;1 44 ma

76 III

36 IIT

0.111 IVa

141 IL

34 III

0.126 IVa

3.5

85 III 8.6

2000

0.125 xva

0.123 IVa

J. T. (1966b)

,J. P. et al.

J. T. (PC.)

J. k’. et d. (1957) \V:‘ASSOX J. T. (1966b) \\‘assow J. T. (P.C.) hVE&INO,

\~‘ASSO,U CT.T. (P.C.)

JVASSON

(1957)

T,OVERING

WASSON

\-vASSON


J. T. (1966b)

Wtisox

\vASSON

J. T. (PC.) J. T. (P.C.)

J. T. (1966s)

J-. T. (1966b)

J. T. (1966b)

tYASSO??

WASSOW

WASSON

\vASSOK

Reference

Of (silicate inclusions)

Woodbine

L0VElUNG J. F. et al. (1957 M.P.A. LOWERING et al. (1957) M.P.A. MASON B. (1963)

11.32

Tho first Ni value is used for cooling rate ealculstions. 1 M.P.A. = Plessite analysis measured with the microprobe. 2 P.C.-Private Communioation.

Steinbach

9.8

13.8 13.65

Springwater

Xiderophyre

128 10.83

13.1

136 11.8 11.79

M.P.A.

LOVERING J. F. et al. (1957) M.P.A. LEVERING J. ‘I?. et al. (1957) M.P.A. M.P.A. HENDERSON E. P. (P.c!.) M.P.A.

10.43 118 10.98

M.P.A.

M.P.A. HENDERSON E. 1’. (P.C.) M.P.A.

CULDBERG E. cf. al. (1951) (:OLDBERG F. Pt al. (1951)

138

12.6

12.4 11.71

8.31

17.1

Mt. Vernon Newport

Off

Om

Giroux Glorieta Mt.

Imilac

Om

Brenham

Pa&&es Albiu

Wiley

Om (silicate inclusions) Off-D

Off

Toluoa

Tazewell

28 38

29

41

9 11

69

30

-

-

-

170 II

4.0 IV

-

LOVERING J. F. et al. (1957) WASSON J. T. (P.C.) LOVERING J. F. et al. (1957)

-

LOVERING J. F. et al. (1957)

CLARK R. (P.C.) WASSON J. T. (P.C.)

WASSON J. T. (PeC’.)

LOVERING J. F. et czt. (1957)

-

TLOv~t~~?;OJ. F. (1957)

W-ASRON .J. T. (1966b)

8

3 ?t 8 z.

k z g ” K’ 0 =

g

K

2

0 ct,

B

z

Gz

0 8 i=

1006

3. I.

GOLWTEIX

and

J. M.

SHORT

borders around divine. Isolated kamacite bands were found in the meteorites Imilac, Springwater and Giroux. The other pallasites with larger metal areas have Widmanstatten structures with typical o&ah&al arrangement. The ideal trace crosses a symmetrical kamacite band into a wide area of fine plessite. Tracks into coarse plessite are undesirable because of the high and often unsymmetrical buildup of Ni in the center of the taenite. If the plessite forms wide lamellae or large droplets in the middle of the taenite area next to the kamacite, the Ni profile can be seriously altered (MASSALSKI et al. 1966). However, careful selection of areas for microanalysis avoided these difficulties. Applied Research Laboratories instruments were used for all electron probe microanalysis. Seven chemically analyzed, annealed Fe-Ni alloys ranging from 5--M% Xi were used as calibration standards. Calibration curves permitted accurate interpolation. Corrections for drift, background, and counting system dead-time were also made. The data were taken using fixed time counting and the Fe and Ni always added to within -+2% of 100%. Counting steps as short as I ,u were taken in regions near the kamacite-taenite border. Because the electron probe has a finite diameter (1.0-1.5 p at 20 kV), the maximum Ni aontent in taenite at the a/r boundary cannot be measured. The exact position of the a/r boundary is usually obscured, since at least one electron probe data point lies between the minimum kamacite composition and the maximum taenite composition (see Fig. 2). The position of the a[~ interfaoe relative to the electron probe data was de~rmined by validating the point where the average between the maximum and minimum interface values occurred. This procedure was checked by using the same method on non-diffused Fe-Ni couples. The interface position calculated by the above procedure is estimated to be within 50.2 p of the correct value. The bulk composition of the meteorite was established by analysis of several large plessite areas while scanning across the plessite with a large probe size of 20-100 ,u. The standards were also measured with a large beam. Usually agreament with bulk chemical analysis was found (Table 1). When the two disagreed, the microprobe analysis was accepted as the local bulk composition. The formation of the Widma~t~t~n pattern in meteorites can be explained in terms of a nucleation and growth process, This process has been described most recently by ~KAS~ALSKI (1962a), SHORT and ANDERSON(1965) and REED (1965). Quantitative analyses using computer techniques have also been described (Woou, 1964; GOLDSTEIN, 1964; GOLDSTEIN and OQI[LVIE,1965b). The nucleation and growth process Occurs in the following manner. During cooling of the meteorites, the kamacite phase (a) will nucleate along the octahedral planes of the parent taenite (y) phase. The kamacite growth is controlled by solid-state diffusion. As the meteorite continues to cool, both taenite and kamaoite must increase in nicka content, but kamacite grows at the expense of &mite in accordance with the level rule (Pig. I). However, the movement of Ni from the kamacite (~)~tae~te (~1 boundary becomes restricted because of the low diffusion rates. Therefore, Xi builds up in the taenite, especially near the a/r interface.

cooling rates of 27 iron and stony-iron meteorites

1007

The nucleation and growth process is influenced by five independent factors: overall Ni concentration (C,) of the meteorite, undercooling below the equilibrium precipitation temperature (A!.!‘), pressure (I’), impingement of the neighboring kamacite plates (X,), and the cooling rate (T) of the meteorite. The Ni content determines what phases form. For a given cooling rate, as the Ni content of the meteorite increases, the width of kamacite phase decreases since Ni stabilizes the taenite phase (Fig. 1). The undercooling determines the temperature at which the As the amount of undercooling increases for a given kamacite phase nucleates. meteorite, the size of the kamacite phase decreases. Pressure lowers the diffusion rates at any temperature and stabilized the y phase at low temperatures. It has been shown quantitatively (GOLDSTEIN and OGILVIE, 1965b) that the pressure effect is not an important factor in the growth of the Widmanstatten structure. Other arguments have also been given (REED ; ANDERS, 1964) showing that pressure is minimal as the pattern develops. Therefore, the pressure effect will not be The interference of growing kamacite platelets is more considered further. significant for meteorites with low Ni contents. A build-up of Ni in the taenite phase which separates two impinging kamacite plates occurs. The cooling rate and the bulk Ni content are the most important factors controlling the size of the Widmanstatten structure. For example, faster cooling produces smaller bands Any analysis of the growth of the for a meteorite with the same Ni content. Widmanstatten structure must take into account all of these factors. Because of the complexity of such a problem, a computer program to simulate the growth was developed and employed. A model for the isobaric diffusion-controlled plane-front growth of a precipitate with time dependent boundary conditions. In formulating a growth model for kamacite growth, we made the following assumptions: 1. The growth of kamacite occurs along a plane front (one-dimensional growth). 2. The interface compositions between the kamacite and taenite phases are given by the equilibrium phase diagram. 3. The volume differences between the kamacite and taenite phases can be neglected. 4. No material flow occurs across the centers of the kamacite and taenite phases, i.e. a closed system. The first three factors have been discussed previously (GOLDSTEIN and OGILVIE, 1965b). The fourth assumption requires that if impingement occurs, &,/ax, &,/ax = 0; no material flow occurs across the centers of kamacite and taenite. This assumption allows us to calculate the Ni concentration build-up in taenite. However, this assumption demands that symmetric taenite composition gradients be developed in the meteorite, since each impinging kamacite band must nucleate In this calculation C,, the bulk Ni; AT, the underat the same temperature. cooling; X,, the distance between impinging kamacite plates; and 7, the cooling rate are the variables which determine the final width of the kamacite phase 14’ and the concentration graduents in kamacite and taenite C,, C,. Therefore C,, C,, W = f (C,, AT, 20,~) The dependent

variables, the phase boundary

compositions

(1) CY1,C,, and the diffusion

1008

3. I.

GOLDSTEIN

and J. M. SEIORT

coefficients in kamacite be and in taenite 6,, are a funct&on.of C,, C, and 7, Cyt’ c.1, 13,) 8, = f(C,, Q,, 7).

(2)

The dependent variables are a function of what we wish to calculate. Therefore, the calculation of W, C,, 0, must be a reiterative one. The calculation of the concentration gradients and the width of the kamacite phase is determined by assuming that the kamaoite growth occurs by N successive isothermal growth steps. If each step is sufficiently small, we can assume that this is equivalent to continuous growth. The general procedure for calculating AlV,, during one growth step is shown in Fig. 3.

COMPOUTION CHAMX

GROWTH IN At

IN At

DIFFA - CAWlA

(DIFFA - CAMA) = (Cyl - Cal)dWn N W=C AW, n=l Fig. 3. Calculation of the growth of krtmacite.

At time t, the Ni concentration gradients in kamacite (a) and taenite (y) is given by C,* and C,” respectively. W is the width of kamacite at time t and X, is the half-width between the centers of impinging kamacite plates. The phase boundary compositions are given by Cal* and Cyln. We then calculate the change in the Ni concentration that occurs in kamacite and taenite if growth occurs isothermally for a time period At. These are C;+l, C;+l. The central taenite composition will also inorease. The amount of Ni entering the kamaoite during At is CAMA and the amount of Ni entering the taenits during At is DIFFA. Since no material can be created or destroyed during the growth period within X0,

(DIFFA

-

OAMA)

= (C;l+l -

C:,+l) AWn

(3)

where 0;;’ and C:zl are phase boundary oompositions at t’ = t + At. Therefore, A W,, the growth of kamacite in At, can be calculated. To calculate the total growth

Cooling rates of 27 iron and stony-iron meteorites

w: we

SUM fhe N isothermal

steps so that;

$ AW,.

W = To calculate

material

flow within

(4)

each phase in the isothermal

growth

time

At, we must solve Pick’s second law:

ac

a

(

at = ax i3(c);

1

where D = f(e) for the taenite phase. Analytical solutions to the above equations can be obtained if the di~usioll In the cases we are interested in, there ~oe~cient is independent of composition.

--__

W

F--.-.- _____

I x m -I__--

--__---x-IA Cl Fig. 4. Calculation of the composition change of taenitc.

are no known solutions for the general problem, and numerical methods are needed. To accomplish this, we use the method of finite differences (see Fig. 4). Jie divide each phase into a grid of NX points All: apart. The change in compositioi~ C, at each point in the grid is calculated for each time i~l~rernent At. Let Ax, and At be increments of the variables x and t. The set of points in the x-t plane for phases o! and y given by: 12 = 0, 1, . . . , N;

t = n At;

y=W-SUME-(p-2)Ay; x = W + (Ax - SUMD)

t’ = (n -+ 1) At . . , NY m = 0, 1, . . . , NX

p = 0, l,,

+ (m - 2) Ax;

(6)

determines a grid whose mesh size is fixed by At, Ay in the cc phase, and Ax in the y phase. W is the width of the tc phase and SUME and Ax - SUMD are the Using the relations Lieven first grid points in the cc and y phases respectively.

ac -=zz

en-t1 m

At

at az,

------_

ax

cn??z

_

D:+l

- DZ-1

2 Ax

(7)

1010

J. I. &%~s’mIrr

and

J. %I. kh%?RT

we obtain the governing difference equations;

D”, + -

ma.+1- x-d 4

- Dh-l) _

c”

1

rnfl

- 2 D,“C,”

D

4

By using the forward time differences, we employ the current values of the composition at given grid points to predict ahead in time. After calculating C+l for all values of p, p = 1, NY fox a and Qz+‘” for all values of m, ?n = 1, NX f& y;

After A W, is calculated acaording to equation (3), Ax - SUJVD is decreased by AW and SUb4E increased by A W, and the temperature is decreased for the next isothermal step and the process of calstulating A W, is repeated. Growth is considered to stop at that temperature for which material flow only occurs in regions 80%) occurs within the first 50°C after nucleation of kamacite begins. Below 45O”C, the only change in the Ni concentration gradient in taenite occurs in the first 10 p beyond the a/y interface. Here the Ni oontent increases greatly approaching 50% Ni at 350°C. Diffiaion in the kamacite is infinite for the first 60°C below the nucleation temperature. After this period, a gradient in kamacite oocurs. The gradient increases as cooling proceeds reaching 0.4 at.% Ni at 35O“C. The shape of the kamacite gradient is qualitatively the same as predicted by GOLDSTEIN(1965~). It is inte~sti~ to note that at no time during the growth process is equilibrium present except at the a/y boundary and possibly in the kamacite above 550 1!.

Cooling

rates of 27 iron and stony-iron

1011

meteorites

Since kamacite and taenite are inhomogeneous, and therefore in a non-equilibrium state, it is not valid to correlate the relative proportions of kamacite and taenite present in a meteorite with the Fe-Ni phase diagram, which represents the equilibrium case. There observations preclude the use of any lowest intermediate temperature of equilibrium T, (MASSALSKI and PARK, 1962b, 1964) for determining the time scale associated with the formation of the Widmanstatten pattern. This growth program represents a significant upgrading of the growth program of GOLDSTEIWand OGILVIE (196513). In this program we avoid the assumption that fi, is infinite in the kamacite phase and that growth of kamacite stops above 350°C.

X0=650/,, COOLING

0

50

C, =9.0, TN = 605’C

RATE-5*1'/106YEARS

100

150 DISTANCE

200

250

300

350

400

IN MICRONS

Fig. 5. Growth of kamacite as a function of time-concentration graidents in kamacite and taenite. AC, indicates amount of concentration gradient in kamacite. The program makes use of the newly determined Fe-Ni phase and OGILVIE, 1965a), not an assumed diagram as used by OWEN and Lm diagram as used by GOLDSTEIN and OGILVIE One of the input variables is the cooling rate or the value function of time. We assumed a logarithmic temperature simple equation: T = T,ecrt where

T T, T t

= = = =

cooling temperature, constant, variable which is changed time.

according

diagram (GOLDSTEIT WOOD (1964) or the (1965b). of temperature as a decay following the (10)

to the cooling rat,e desired,

This equation has the same functionality, T = f(t) as the cooling curves proposed by ALLAN and JACOBS (1956) for central cooling with long-lived radioactivity or for cooling as a function of distance from the center of spherical bodies over the limited temperature range 700-350°C. This functionality makes the equation useful in our analysis since it is not tied to any particular cooling model, parent body radius, or thermal diffusivity. It should be noted that the equation produced a cooling rate dT/dt = (-T)T. (11) The cooling rates calculated

in this study are normalized

to 5OO”C, the temperature

1012

J. I.GOLDSTEINand J. 35.SHORT

at which, for most octahedritos, the Widma~~tten of its growth.

structure has completed most

(0) CalcuEcltionof molhg ratee To determine a oooling rate for a meteorite, use is made of both the eleotron microprobe data and the theoretical kamaoite growth analysis. If no build-up of Ni is observed in the taenite of a measured kamacite-taenite area, then no impingement with neighboring kamacite platelets has occurred (X, N a3). The value of W for the kamacite band oan be obtained by a one dimensional mass balance around C,, the average meteorite composition. If Hi did build-up in the center of taenite because of impingement of kamacite platelets, it is necessary to calculate the value of X,, the distance between two impinging kamacite platelets. One must check first, however, to make sure that the Ni gradients in the taenite between two impinging kamacite plates are symmetrical (see Fig. 8). If this is not the case, then one of the plates nuoleated at a higher temperature than the other. Dif%rent nucleation temperatures allows material flow to occur across the center of the tasnite phase (the position of the minimum Ni concentration). Therefore, if the Ni gradients are not symmetrical, assumption 4 of the growth analysis, that is no material flow occurs across the midpoint of the taenite does not hold and the kamaeite-taco area is rejected for the cooling rate analysis. If the Ni gradients are symmetrical, X, is measured by a mass balance computed on the basis of the Ni accumulated in the taenite from the R/Y boundary to the center of the taenite. To calculate the cooling rate of each meteorite, we first measure the concentration gradients C,, C, in several kamacite-taenite areas. After this, we calculate X, and W for each kamacite band. The average composition may also be measured by plessite analysis if necessary. The next step is to select one of the measured bands for comparison with the growth analysis. Usually this band has the least amount of Ni build-up in taenite. This selection then allows us to enter a measured X, and C, as two of the four input variables. The eal~~atio~ procedure is shown in the form of a flow chart in Fig. 6. In the computer analysis, then X, and C, are entered as measured variables, and an assumed cooling rate (T) and the amount of undercooling AT are entered as independent variables. As a result of the computer calculation, a kamacite band width W, in this case the half width of the kamacite band, and the Ni conoentration in kamacite C, and taenite C, are computed. To oalculate the cooling rate: (1) The assumed oooling rate is held constant and AT is varied until WCaO.equals W mew 3 * (2) The caloulated and measured taenite composition gradients are then compared. If agreement is not found then, (3) A new aooling rate is assumed and steps 1-2 repeated. Generally if the assumed cooling rate is too low, the oaIculated Ni values fall above the measured values close to the ~/y interface and below the measured values farther on. If the assumed cooling rate is too high, the oalculated Ni values fall below the measured values close to the a/r interface and above the

Cooling rates of 27 iron and stony-iron meteorites

1013

measured values farther on. An example of this comparison procedure is shown in Fig. 7 for a kamacite-taenite area in the Grant meteorite. Agreement was found at a cooling rate of 5*l”/106 yr. Since each meteorite has only one cooling rate, the comparison method can be repeated for all the other kama~ite-taenite bands measured for the meteorite.

MEASURED-

c,,

CALCULATED-

1NPUTS

cI

, C,

W, X,

TO COMPUTER

SPECIFIED DEPENDENT

X

ONE

KAMACITE

Co,

X,

=

VARIABLES-

6T, ‘I’

VARIABLES-O,,

Dr,

COMPARISON

OF W (CALCULATED)

w MEASURED)

IF WC = W,

BAND

PROGRAM

VARIABLES-

INDEPENDENT

WITH

VS.

FOR

IF

C,,

[zziiq

t

, C 11

,

4

,

---IvF:“,~Ls.x /

W,#W,,,

ADJUST

AT

I COMPARISON wiTi

cv

OF vs.

cI

vs.

x (CALCULATED)

x (MEASURED)

I

Fig. 6. Flow diagram for the calculation of meteorite cooling sates.

The only variable that is changed is AT, the undercooling, and the only measured variable that is changed is x0, the impingement factor. Figure 8 shows the calculation made for an area in the Grant meteorite where impingement has taken place. 3. RESULTS The calculated cooling rates are given in Table 2 and shown in Fig. 9 with their precision, based primarily on estimates of the agreement between the theoretical and experimental Ni profiles in the several bands measured. The average precision is &2Oo/o although some meteorites were assigned precisions of 140%. The cooling rates are normalized to 500°C, but essentially apply from the temperature at nucleation 700 to 300°C at which growth of the WidmanstBtten structure becomes negligible. The most important result is the wide range of cooling rates found-from O-4

J. I.

1014

5q-

and J. M.

GOLDSTEIN

SHORT

7

GRANT METEORITE *ELECTRON MICROPROBE

DATA

!I

40 c ,

/

z g

30

,COOl.ING RATE 3*3?106 YEARS, W =245/i

1

DISTANCE, &CRONS Fig. 7. Determination of the cooling rate for the Grant meteorite.

+o

as

I

-

-

ORANT METEORITE

l

a2

-

2a

-

-

ELECTRON MICROPRD8E DATA

CALCULATED CURVE

.;

z

CODLIWO RATE-

ip

W = 222 P.

5*1*/ IO6

TN c 601.

‘(F’S.

C

0

s

E

24

-

20

-

I‘

-

IL

X0 =SOC&

I

I

CENTER TAENITE

OF -

-

I

1

I

I

I

I

I

I

I

I

I

I

I

I

I-law4

DISTANCE IM MICRONS

Fig. 8. Calculation of taenite composition grtiient in the Grant meteorite when impingement occurs.

Cooling rates of 27 iron and stony-iron Table 2. Summary

1015

meteorites

of cooling rates for iron and stony-iron meteorites

Meteorite

Structure

Huizopa Gibeon Bristol altonah Duchesne Toluca Spearman Four Corners Carbo Grant Hualapai Woodbine Carlton Edmonton Tazewell Wiley Butler Dayton

Of (IVa) Of (IVa) Of (IVa) Of (IVa) Of (IVa) Om Om Om Om Of Of Of Off Off Off Off-D Off-D Off-D

Bulk Ni

Cooling rate

Precision

(wt. %)

(yr)

(9;)

ITOW

7.81 7.96 X.3 8.8 10.0 8.31 8.6 9.7 10.6 9.6 12.3 12.6 12.68 12.66 17.1 12.4 16.0 18.1

W/10” 2O”/lO” %O”/lO” 20”/106 9”/106 1.6”/10G 4.0”/106 1.9”/10” 1.0/“106 5.1”/106 1.1”/106 ‘.‘i’/l()”

13.8 11.8 13.8 11.8 13.1 12.8 10.83 13.8

05”/10” 0.5”/10~ 1.3”/106 0.5”/106 0~5”/106 0,8”/1OC lW/lO” 0~4”/106

k-30 _L40 125 ,40 :z 20

9.8

(i.?~~/lO~

*35

W/10” 1.9”/10” 2.2”/10” 15”/10” 0.5”/106 6+/l@

$25 :k25 :/ 1.5 $30 :_2 5 I40 :Cl5 f2O *15 z- 30 120 -&25 i: 15 *15 +20 k35 -_k3O _L30

Pallmites

Albin Brenham Giroux Glorieta Mt. Imilac Mt. Vernon Newport hpringwater

Om Om -

-

_-_20 _ 15 + 40

Siderophyre

Steinbach

Off

to 40’/10” yr-a range which is much greater than any imprecision in the measurements and calculations. In terms of meteorite class, three approximate groupings can be made : low Ni contents (8-lo?;,) (a) A group of fine octahedrites with exceptionally ranging from 9 to 40°/106 yr, and which because of their low Ge content belong to a group WASSON (1966b) has designated IVa; (S-17 wt.?, Ni) with cooling rates (b) Typical medium and fine octahedrites ranging from 1 to 5’/106 yr; 0*4-l.O”/lO” (c) A group of pallasites which have lower cooling rates-about yr. The meteorite Butler, whose texture is transitional between a fine octahedrite and an ataxite, has a very low cooling rate (0*5’/10 yr) and also an unusually high Ge content, 2000 ppm (WASSO~;, 1966a). The Siderophyre Steinbach has an intermediate cooling rate of 7”/106 yr.

J. I. GOLDSTEINand J. M. SHORT

1016

Correlations between bulk Ni content, bend width and cooling rate which wc have calculated ere seen by comparison of the photomicrographs in Fig. 10. The meteorites Toluc~, Edmonton, and Tszewell have about the same cooling rate: but show, quite normally, progressively finer structure with increasing Ni content. The meteorites Toluca and Bristol have almost the same Ni content but Bristol is much finer and has a correspondingly higher cooling rate. The same effect was

I

6

0.1

1

I

I,,,,

I

IO

o-5

CbOLlNG

RATE

&O’

,,,,a

50

100

YRS.)

Fig. 9. Compilation of the oooling rates of 2’7 iron and stony-iron meteorites. Cooling rate preoisionsare indicated by the bars.

seen in the other meteorites of similar Ni content but varying cooling rates. The combination of the two efXects is seen in the series Butler, Edmonton and Bristol. The increasing cooling rates offset the increasing nucleation temperatures, and tho resulting kamacite band widths are nearly the s&me. 4. DBCUSSION

(a) Accuracy

mui precision

of method

The method of determining the cooling rates is based on a rather complex model for nucleation and &f&&on-controlled growth of kamactie. Therefore, arly estimate of the accuracy of the results must be conserva$ive, even though the internal precision is cert&nly f IO-40%. The most obvious source of error would be the diffusion coefbcients for Ni in y. These have been determined by GOLDSTEI~V et al. as a function of temperature, composition, and pressure cbbove 8OO”C,and these may be extrapolated to the temperature range in question with an estimated accuracy of better than -&50%. Grain boundary diffusion will not contribute

~00~1~~

TOLUCA, C, CR

= 8.3 = 1.6”/

RATES OF IRON METEORITES

EDMONTON,

Om

Off

TAZEWELL,

Co : 12.7 Ni.

Ni. IO6

c A. = 1.9”/

YRS

BRISTOL, Co= 8.3 C.R. = 20”/

Of Ni. IO6

‘0%.

Co= 106

YRS

I7

C R = 2,2’/

BUTLER,

Off-D

c,

Ni

= 16.0

C R = O-5”/

IO6

Off Ni. IO6

YRS.

YRS.

Fig. 10. Textnral differences in iron meteorites as a function of ;“u’icontent and cooling r&e. The arrow indicates the direction of the kamacite bands oriented perpondiculzlr to the surface.

1016

Cooling rates of 27 iron and stony-iron meteorites

1015

significantly to mass transfer between kamacite and taenite because of the extremely large crystal sizes. The +X + y) b oundary of the Fe-Ni phase diagram is poorly known below 5OO”C, although it seems clear from the decrease in Ni in kamacite near the taenite interface and from theoretical calculations (GOLDSTEIR, 1965a) that the boundary inverts back to decreasing Ni content with decreasing temperature below about 45O’C. Any error in this phase boundary would be reflected in the calculation of W and AT especially in meteorites with low bulk Ni. Cobalt is also present in meteorites to the extent of about 0.1 wt.*,& in taenite and O-6 wt.“;b in kamacite (SHORT and AWDERSON). This amount of Co does not shift the E/(X + y) or Y/(K I_ JJ)phase boundaries enough to prevent the use of the binary Fe-Xi diagram (REED, 1965). The effect of phosphorus in meteorites is more significant than that of Co. However, most of the phosphorus is contained in schreibersite (Fe-Ni),P. The measured phosphorus content in kamacite is less than 0.1 wt.*/* (GOLDSTEIS and OGILVIE, 1963). This amount of phorphorus might change the E/(E + y) phase boundary up to 1% Ni (REED). The effect of Co and P on the difXusion coefficients for Ni in y is certainly less than the estimated accuracy in the diffusion coefficients of &50%. It is to be emphasized, however, that while the calculated cooling rates may be inaccurate by a factor of 2, their precision and the variations in cooling rates between different meteorites will not be significantly affected by moderate changes in the diffusion coefficients or phase diagram. SHORT and ANDERSON found that if the di~usion gradients from two growing kamacite plates extend to the center of the taenite area between them, the gradients were often unsymmetrical. They suggested that the unsymmetrical gradients were caused by different nucleation temperatures for the two kamacite plates. The development of these gradients can be explained qualitatively using Fig. 11. Kamacite-taenite area t, has grown during the time period for 30” of cooling before kamacite-taenite area t, even nucleates. During this period the Ni diffusion gradient oft, extends to and crosses the center of the taenite area. After kamacitetnenite area t, nucleates the Ni diEusion gradient extending from t, is superimposed on that gradient already developed from t,. An unsymmetrical gradient around the minimum Ni concentration point in the taenite is then developed. A taenite area such as shown in Fig. 11 is unsuitable to an analysis of the type we have developed. This is because a closed system for Xi diffusion cannot be assumed (Assumption 4; (~c~~~)~ at the center of the taenite - 0). If we do assume a closed system for both t, and t, the Ni concentration gradients can be calculated. The results of these calculations are shown in Fig. 11. The agreement near the u/y interface is good, but the agreement is poor near the center of the taenite area because the calculated gradient does not extend to the center of taenite and the calculated minimum Ni concentration is not in agreement with the measured value. The net effect of different nucleation temperatures is to lengthen the size of the taenite area and produce unsymmetrical concentration gradients. In this work, we used Ni concentration data only from taenite areas with symmetrical gradients (Fig. 8). Therefore, material flow can be considered to occur in a closed system and the two kamacite plates associated with the taenite

1018

J. 1.

&XDSTEIN

and J. M. S.EORT

nucleate at the same- tem~rat~e. The growth analysis as formulated for this paper is therefore applicable. Calculations of cooling .rates have also been made by WOOD. To determine the cooiing rate he measured the central Ni oontent in taenite for a set of kamacite plates and compared his measured value with central Ni values oalculated by a

SP&AIWM

R(EfEOAITE

TN * 593, C TAENITE

Fig. 11. ~on-~~~t~ic~~

taenite gradients in the Spearman rnet%~rit0.

theoretical analysis. Implicit in his analysis is the assumption of a closed system and that symmetrioal J$i gradients develop in the taenite. However, many of his measured gradients are ~sy~met~oa~. This elect contributes toward his rather large error limits on cooling rates ( j, 100% or more) and it precludes the use of that type of (unsymmetrical~ analysis for the precise determination of cooling rates, that is an average precision of f25%. The fact that the central Ni contents of taenite in coarse o~tahe~ites, < 8 wt. % Ni are usually above 15 wt.% Ni, the eo~~entration gradients in y are sometimes unsymmetrical, and pksite is composed of a/r particles which have ohanged the Ni gradients in taenite, has not akiwed us to use the growth analysis to determine the cooling rate of these nieteorites. This fact has biased our selection of samples toward those with finer structure.

Cooling

(b) Possible injhence

rates of 27 iron and stony-iron

meteorites

1019

of martensite

The previous discussion and calculations of the growth of the kamacite phase and the cooling rates has not taken into account any change of the diffusion coefficients of Ni in the y phase upon cooling below the M, temperature where bee cr, forms (KAUFMAN and COHEN, 1956). That is because the formation of the LYidmanstatten pattern and the Ni profile in taenite develops above the M, temperature. Thus the theoretical curves match the experimental data down to the composit,ion where plessite forms. While a2 production may effect the growth of plessite (MASSALSKI, 1966), it is simply not significant in altering the taenite profile from which the cooling rates are calculated. In all cases, the cut-off points where measured and calculated values were compared was taken at the place where the first rise in Ni occurs (see Fig. 7). The impingement effect of the high Ni taenite paralleling the kamacite within the plessite on the main taenite profiles could only extend a few microns, and t*herefore any modification of the theory used was found to be unnecessary. (c) Correlations among meteorites The 18 iron meteorites studied represent only about 3% of the total known irons; the eight pallasites represent about 20% of known pallasites. The siderophyre Steinbach, a unique stony-iron, contains the mineral tridymite (MASON, 1962) which indicates that the meteorite cooled under conditions of low pressure. Its c+ooling rate is similar to that of the other meteorites studied. Even though the size of the sample studied is relatively small, it is difficult to believe that the groupings of meteorites with respect to cooling rates (pallasites, Group IVa, and t(he irons) will disappear, although new ones could well appear as more meteorites are analyzed. This is because of the strong interaction of CO, cooling rate, and W. In the case of Group IVa,, all these meteorites have a very fine texture, and a restricted range of Ge and Ni content (WASSOY, 1966b). Their taenite diffusion gradients are very steep ( ~10 ,u) and they represent fine octahedrites with the lowest Ni content. Tazewell has the highest Ni content of any known octahedrite. WYley, Dayton, and Butler are transitional toward fine octahedrites. Carbo is at the extreme end of the Ni-range for medium octahedrites. The pallasites chosen are typical although Brenham and Glorietta have unusually large areas of metal. Further analyses of Ni-rich or Ni-poor pallasites may resolve more groups or produce a further spread in cooling rates. The cooling rate and Ge contents of meteorites are also correlated in certain ways. We have found that Ga-Ge Group IVa meteorites (~0.1 ppm Ge) cooled between 9-40”/106 yr. The cooling rate increases with increasing Ni content within the group. There is little doubt that Group IVa meteorites has a unique thermal environment with respect to other meteorites. Spearman is our only sample in Group IIIa (-40 ppm Ge) (WASSON, 1966c) and has a cooling rate of 4”/106 yr. Possibly members of Group IIIa and IIIb may have a small range of cooling rates as found for Group IVa. The quantization of the distribution of Ga and Ge has been taken as evidence for multiple parent bodies. WASSON (1966b) has suggested that the degree of reducing condit’ions present at the time of formation of the meteorites determined the Ge/Ni,

1020

J. I. GOLIXWEIN aad J. M. &IOBT

G#Ni and Fe/X ratios of the parent body. Stronger reducing conditions fatvor higher Ge/Ni, G@i and Fe[R ratios. Fnrthermore, if different iron meteorite groups heve arisen in digerent asteroids, then the l&rger of these objects should have attained higher interior temper&urea (stronger reducing conditions) than the smaller ones. Therefore, WASSON(1966b) suggests that meteorites with high Ge and Ga contents are associated with large asteroids and that meteorites with low Ge and Ga contents s,re associated with smaller asteroids. One could then argue that the cooling rate of a meteorite is a function of its Ge content; the lower the Ge content, the faster the cooling rate. This trend is gener&lly true for the irons. ~eteori~s in Ge Group II (139-230 ppm) which we have studied &ve a cooling r&e of 1-2”f106 yr, in Ge Group III (N-80 ppm)-I-So/lo6 yr snd in Ge Group IVa (O-10-0*14 p~m)-9-40°~106 yr. However, we find variances in cooling rates for meteorites cont&ing the same Ge content (Grant, 36 ppm Ge- 5-1”/106 yr vs. Edmonton, 34 ppm Ge-I*S”JIO” yr). This variance within Group III may be real or may be due to the presence of sever&l subgroups such s,s Groups IIIa, and IIfb (WASSON,1966~)not yet identified. It seems evident that correlations between groupings found by the independent methods of cooling rate determinations and trace element analysis will permit genetic and geographic resolution of iron and stony-iron meteorites. It has also been pointed out by VOSHA~E (1966) that a correlation exists between cosmic ray r~~~tio~ Ebgtts and the Ga-Ge groups. V~SIIAOE measured K@-K*l exposure ages of 65 irons snd found B peak of 600 x 108 yr for Group II& (WASHES, 19~6c), 920 x IO8yr for Group II and 400 x IO6yr for Group IVa (WASSON,ISSrSb). These correlations point out again tha&we are obtaining tkvery biased sampling of meteorites, possibly coming from only a minimum number of parent bodies. (d) Im#caitorts for meteorite premt

bodies

!lXe cooling rates derived by the diflFusiongrowth-analysis theory do not require the assumption of any thermal model. However, a model um be formulated which may represent the cooling history of the meteorite parent bodies. From the order of magnitude of the cooling r&es, it is obvious that the irons and pa&sites must have been i~~~~ within a parent body during the time in which the Widm~nstitten structure formed. The essentially mmive single crystaJ structure of the original taenite phase implies that the metal probably cooled slowly from the melt. Below 300% break-up of the prartnt body and more rapid cooling would not affect the difFusion gradients in the Widmanst;litten structure. However, long-time reheating above this temperature or shorter beating at higher temperatures would &ect the di&sion gradients. The absence of any change shows that this did not happen. The most reasonable insul&or around the metal in the parent body would be silicate m&erial having the thermal dif&sivity (0.007 cmslsec) and concentration of long-lived radioactivities present in chondrites (GO~;DSTELN and O~ILVIE;, 196&b). Assuming slow cooling commenced upon soli~~c~tiou of met& temperature time curves and cooling r&es csn be ~~lc~~d, using the standard equations of ALLANand ~AUUBS. Using such n,model, it is also seen that the centers of large

Cooling rates of 27 iron and stony-iron meteorites

1021

bodies (radius > 800 km) will heat up and never attain a cooling rate less than @lo/IO6 yr in 4.5 x IO9yr. The cooling rate is dependent upon both the size of the parent body and the location of the meteorite within it. The position of the meteorites in the parent body is still a question producing much controversy. The hypothesis that the iron meteorites cooled in the metallic core of a differentiated parent body was supported by calculations of Frsn et al. (1960) and arguments by ANDERS (1964). HENDERSON

100

/

i-

x=.g

0.1

--

i

!~

70 = 2100-K, a-.007

L-.-___ 1UU 200 300 400 RADIUS OF METEORITE PARENT BODY IN Km.

500

Fig. 12. Cooling rate for different sized meteorite parent bodies. T, = initial tompes&ure, 2100°K; c( = thermal diffusivity, 0.007 cm2/sec.

and PERRY (1958) and UREY (1956) have argued that iron meteori~s existed as small objects in the parent bodies and that pallasites come from boundary regions between metallic and silicate areas. The relation between cooling rates and parent body radius for different distances from the center is seen in Fig. 12. The assumption of an initial temperature of solidification or temperature distribution has only a small effect on the cooling rates calculated below 700°C. Near the surface the model breaks down and the surface temperature and thermal diffusivity become crucial. However, comparison of these curves to the range of cooling rates found in this study places some limits on the sizes of the meteorite parent bodies. The Group IVa irons could not have cooled at the core of bodies with chondritic mantles with a radius less than 50 km, nor could any of the others. It is quite possible that different meteorites or groups cooled at the cores of several bodies of different size. It is also possible that they cooled at varying depths (inclu~ng the upper third of the mantle) of one (or more) bodies (e.g. 300 km radius). However,

1022

J. I. GOLDSTEIN and J. MM.SHORT

it is clear that many individual meteorites developed WidmanstBtten structure ill different thermal environments. It is not possible at this time to distinguish between these two alternatives 1,) cooling rate determinations alone. However, additional information such as abundances of elements in the ppm range (Ga, Ge, etc.) and cosmic ray exposure ages coupled with a larger sample of cooling rate values may indeed allow us tz) correlate various meteorites with their parent bodies. .5. CONCLUSIONS 1. The kamacite

band width and Ni content are related quantitatively through difFerences in cooling rates. 2. A very wide range of cooling rates exists for the iron and stony-iroll meteorites (04-40°/106 yr). Three approximate groupings as to cooling rate were found (1) low Ni-filkcb octahedrites (IVa)-9-40”/106 yr, (2) fine, and medium octahedrites: l-3”/10” >-I’. (3) pallasites, 0*4--la/lo6 yr. These groupings which may increase in number LS more cooling rates are determined suggest that many meteorites developed ill different thermal environments. Acknowledgements--We wish to thank .F. WOOU and P. SOULESfor their assistance wit,11t/113 experimental part of the investigation, W. PUTNEY for his assistance with the computer prepgramming, and K. KEIL and J. WASSON for their valuable suggestions and criticisms. 1Vt.ja~‘,* very grateful to E. P. HENDERSON and R. CLARKE of the National Museum, and C. B. JIoo~r: of the Nininger collection at Arizona State University for the samples we have ohtainrtl f’ror~~ them. REFERENCES ALLAN D. W. and tJ~~~~~ J. A. (1956) The melting of asteroids and the origin of lur~t~~~orir~~; Geochim. Cosmochim. Acta 9, 256-272. ANDERS E. (1964) Origin, age, and composition of meteorites. Space Sci. IZec. 3, 383 -7 14. BORG:R. J. and Lila D. Y. F. (1963) The diffusion of gold, nickel, and cobalt in alpha iroil. \ study of the effect of ferromagnetism upon diffusion. Acta Met. 11,861-866. FISH R. A., GOLES G. G. and ANDERS E. (1960) The record in the meteorites. Ss~~~~,phys. ./. 132, 243-258. GOLDBEXQE., UCHIY~MA A. and BROWN H. (1951) The distribution of nickel, c:obalt, gallilirlr, palladium, and gold in iron meteorites. Geoohim. Comoehim. Acta 2, l-25. GOLDSTEINJ. I. and OQILVIE R. E. (1963) Electron microanalysis of metallic meteorites. I’:II.I I-phosphides and sulfides. Geochim. Cogmochim. Acta 27, 623-637. GOLDSTEINJ. I. (1964). Sc.D. Thesis, Massachusetts Institute of Technology. GOLDSTEINJ. I., HANNEMAN R. E. and OCILVIE R. E. (1964) Diffusion in the Fe--Si systenl ;II ! atmosphere and 40 kb pressure. Trans. Met. Sot. AIME 285, 812-820. GOLDSTEINJ. I. and OUILVIE R. E. (1965a) A re-evaluation of the Fe-rich portion of the Ft. S f system. Trans. Met. Sot. AIME $$3& 2083-2087. GOLDSTEINJ. I. and O~ILVIE R. E. (1965b) The growth of the Widmanstiltten pattern in mct~a.Ilir~ meteorites. Geochim. Cosmochim. Acta aS, 893-920. GOLDSTEINJ. I. (19650) The formation of the kamacite phwe in metallic meteorites. ./. Geo#!/s. Res.70,6223-6231. HENDERSON E. P. and PERRY S. H. (1968) Studies of seven siderites. Proc. I.7.S.S. AVlrxrltr,l 107, 339-403. KAUFMAN L. and COHEN M. (1966) The Martensitic transformation in the iron-nickel s@enl. Trans. Met. Sot. AIME m, 1393-1401.

Cooling rates of 27 iron and stony-iron

meteorites

1023

LOVERING J. F., NICHIPORUK W., CHODOS A. and BROWN H. (1957) The distribution of gallium, germanium, cobalt, chromium, and copper in iron and stony-iron meteorites in relation to nickel content and structure. Geochim. Cosmochim. Acta 11,263-278. MASON B. (1962) Meteorites. Wiley. MASON B. (1963) The pallasites. Am. Museum Novitates 2163, 1-19. MASSALSKI T. B. (1962a) Researches on Meteorites. Wiley. MASSALSKI T. B. and PARK F. R. (1962b) A quantitative metallographic study of five octahedrite meteorites. J. Geophys. Res. 67, 2925-2934. MASSALSKI T. B. and PARK F. R. (1964) A study of four pallasites using metallographic, microhardness and microprobe techniques. Geochim. Cosmochim. Acta 28, 1165-1175. MASSALSKI T. B., PARK F. R. and VASSAMILLET L. F. (1966) Speculations about plessite.

Geochim. Cosmochim. Acta 30, 649-662. NICHIPORUK W. (1958) Variations in the content of nickel, gallium, germanium, cobalt, copper and chromium in the kamacite and taenito phases of iron meteorites. Geochim. Cosmochim. Acta 13,233-247. REED S. J. B. (1965) Electron-probe microanalysis of the metallic phases in iron meteorit.es.

Geochim. Cosmochim. Acta 29, 535-549. SHORT J. M. and ANDERSON C. A. (1965) Electron microprobe analyses of the Widmanstgtten structure of nine iron meteorites. J. Geophys. Res. 70, 3745-3759. UREY H. C. (1956) Diamonds, meteorites, and the origin of the solar system. Astrophys. J, 124,

623-637. VOSHAGE H. (1966) Private communication. WASSOX J. T. (1966a) Butler, Missouri-an iron meteorite with extremely high germanium content. To be published in Science. WASSON J. T. (196610) Iron meteorites with low concentrations of gallium and germanium and the Ga-Ge classification of iron meteorites. Submitted to Geochim. Cosmochim. Acta. WASSON J. T. (1966c) Private communication. WOOD J. A. (1964) The cooling rates and parent planets of several iron meteorites. Icarus 3, 429-459.

8