6 June 1980, Volume 208, Number 4448

Extraterrestrial Cause forCretaceous-Tertiary Extinct

Experimental results and theoretical interpret.

Luis W. Alvarez, Walter Alvarez, Frank Asaro, Helen V. A

In the 570-million-year period forwhich abundant fossil remains are avail-able, there have been five great biologi-cal crises, during which many groups oforganisms died out. The most recent ofthe great extinctions is used to define theboundary between the Cretaceous andTertiary periods, about 65 million years

microscopic floating aiboth the calcareous pnifera and the calcarecwere nearly extermin.few species survivingother hand, some groifected, including the 1;diles, snakes, mammal

Summary. Platinum metals are depleted in the earth's crust relatabundance; concentrations of these elements in deep-sea sedimdicate influxes of extraterrestrial material. Deep-sea limestones expmark, and New Zealand show iridium increases of about 30, 160,spectively, above the background level at precisely the time of thtiary extinctions, 65 million years ago. Reasons are given to indicatEof extraterrestrial origin, but did not come from a nearby supernovsuggested which accounts for the extinctions and the iridium obseia large earth-crossing asteroid would inject about 60 times the obj4atmosphere as pulverized rock; a fraction of this dust would stay iifor several years and be distributed worldwide. The resulting darpress photosynthesis, and the expected biological consequences nthe extinctions observed in the paleontological record. One predictisis has been verified: the chemical composition of the boundary clato come from the stratospheric dust, is markedly different from thatthe Cretaceous and Tertiary limestones, which are chemically sirrFour different independent estimates of the diameter of the asterolie in the range 10 ± 4 kilometers.

SCIENCE

extinctions (3, 4), and two recent meet-ings on the topic (5, 6) produced no signof a consensus. Suggested causes in-clude gradual or rapid changes in ocean-

ographic, atmospheric, or climatic con-

ditions (7) due to a random (8) or a cy-

clical (9) coincidence of causative fac-

the tors; magnetic reversal (10); nearby

supernova (11); and the flooding of thetion ocean surface by fresh water from a pos-

tulated arctic lake (12).A major obstacle to determining the

ation cause of the extinction is that virtually allthe available information on events atthe time of the crisis deals with biological

4ichel changes seen in the paleontological rec-

ord and is therefore inherently indirect.Little physical evidence is available, andit also is indirect. This includes varia-

nimals and plants; tions in stable oxygen and carbon isotop-Alanktonic forami- ic ratios across the boundary in pelagicus nannoplankton sediments, which may reflect changes inated, with only a temperature, salinity, oxygenation, andthe crisis. On the organic productivity of the ocean water,ups were little af- and which are not easy to interpret (13,and plants, croco- 14). These isotopic changes are not par-Is, and many kinds ticularly striking and, taken by them-

selves, would not suggest a dramaticcrisis. Small changes in minor and trace

tive to their cosmic element levels at the C-T boundary haveents may thus in- been noted from limestone sections inosed in Italy, Den- Denmark and Italy (15), but these dataand 20 times, re- also present interpretational difficulties.e Cretaceous-Ter- It is noteworthy that in pelagic marineD that this iridium is sequences, where nearly continuousfa. A hypothesis is deposition is to be expected, the C-Trvations. Impact of boundary is commonly marked by a hia-act's mass into the tus (3, 16).n the stratosphere In this article we present direct phys-rkness would sup- ical evidence for an unusual event at ex-

natch quite closely actly the time of the extinctions in theon of this hypothe- planktonic realm. None of the currentLy, which is thought hypotheses adequately accounts for this:of clay mixed with evidence, but we have developed a hy-iilar to each other. pothesis that appears to offer a satisfac-id give values that tory explanation for nearly all the avail-

able paleontological and physical evi-dence.

ago. At this time, the marine reptiles, thfeflying reptiles, and both orders of dino-saurs died out (1), and extinctions oc-curred at various taxonomic levelsamong the marine invertebrates. Dra-matic extinctions occurred among theSCIENCE, VOL. 208, 6 JUNE 1980

of invertebrates. Russell (2) concludesthat about half of the genera living at thattime perished during the extinctionevent.Many hypotheses have been proposed

to explain the Cretaceous-Tertiary (C-T)

Luis Alvarez is professor emeritus of physics atLawrence Berkeley Laboratory, University of Cali-fornia, Berkeley 94720. Walter Alvarez is an associ-ate professor in the Department of Geology andGeophysics, University of California, Berkeley.Frank Asaro is a senior scientist and Helen Michel isa staff scientist in the Energy and Environment Divi-sion of Lawrence Berkeley Laboratory.

0036-8075/80/0606-1095$02.00/0 Copyright © 1980 AAAS 1095

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Identification of Extraterrestrial

Platinum Metals in Deep-Sea Sediments

This study began with the realizationthat the platinum group elements (plati-num, iridium, osmium, and rhodium)are much less abundant in the earth'scrust and upper mantle than they are inchondritic meteorites and average solarsystem material. Depletion of the plati-num group elements in the earth's crustand upper mantle is probably the resultof concentration of these elements in theearth's core.

Pettersson and Rotschi (17) and Gold-schmidt (18) suggested that the low con-

centrations of platinum group elementsin sedimentary rocks might come largelyfrom meteoritic dust formed by ablationwhen meteorites passed through the at-mosphere. Barker and Anders (19)showed that there was a correlation be-tween sedimentation rate and iridiumconcentration, confirming the earliersuggestions. Subsequently, the methodwas used by Ganapathy, Brownlee, andHodge (20) to demonstrate an extra-terrestrial origin for silicate spherules indeep-sea sediments. Sarna-Wojcicki etal. (21) suggested that meteoritic dust ac-

cumulation in soil layers might enhancethe abundance of iridium sufficiently topermit its use as a dating tool. Recently,Crocket and Kuo (22) reported iridi-um abundances in deep-sea sediments

and summarized other previous work.Considerations of this type (23)

prompted us to measure the iridium con-

centration in the 1-centimeter-thick claylayer that marks the C-T boundary insome sections in the Umbrian Apen-nines, in the hope of determining thelength of time represented by that layer.Iridium can easily be determined at lowlevels by neutron activation analysis(NAA) (24) because of its large capturecross section for slow neutrons, and be-cause some of the gamma rays given offduring de-excitation of the decay prod-uct are not masked by other gamma rays.

The other platinum group elements are

more difficult to determine by NAA.

Italian Stratigraphic Sections

Many aspects of earth history are bestrecorded in pelagic sedimentary rocks,which gradually accumulate in the rela-tively quiet waters of the deep sea as in-dividual grains settle to the bottom. Inthe Umbrian Apennines of northern pe-

ninsular Italy there are exposures of pe-

lagic sedimentary rocks representing thetime from Early Jurassic to Oligocene,around 185 to 30 million years ago (25).The C-T boundary occurs within a por-

tion of the sequence formed by pinklimestone containing a variable amountof clay. This limestone, the Scaglia ros-

Fig. 1. Photomicro-graphs of (a) the basal

bed of the Tertiary,showing Globigerinaeugubina, and (b) thetop bed of the Cre-taceous, in which thelargest foraminifer is

Globotruncana con-

tusa. Both sectionsare from the Bottac-

cione section at Gub-bio; they are shown atthe same scale and

the bar in (a) is 1 mm

long.

1096

sa, has a matrix of coccoliths and cocco-lith fragments (calcite platelets, on theorder of 1 micrometer in size, secretedby algae living in the surface waters) anda rich assemblage of foraminiferal tests(calcite shells, generally in the size range0.1 to 2.0 millimeters, produced bysingle-celled animals that float in the sur-face waters).

In some Umbrian sections there is ahiatus in the sedimentary record acrossthe C-T boundary, sometimes with signsof soft-sediment slumping. Where the se-quence is apparently complete, forami-nifera typical of the Upper Cretaceous(notably the genus Globotruncana) dis-appear abruptly and are replaced by thebasal Tertiary foraminifer Globigerinaeugubina (16, 26). This change is easy torecognize because G. eugubina, unlikethe globotruncanids, is too small to seewith the naked eye or the hand lens (Fig.1). The coccoliths also show an abruptchange, with disappearance of Cre-taceous forms, at exactly the same levelas the foraminiferal change, althoughthis was not recognized until more re-cently (27).

In well-exposed, complete sectionsthere is a bed of clay about 1 cm thickbetween the highest Cretaceous and thelowest Tertiary limestone beds (28). Thisbed is free of primary CaCO3, so there isno record of the biological changes dur-ing the time interval represented by theclay. The boundary is further marked bya zone in the uppermost Cretaceous inwhich the normally pink limestone iswhite in color. This zone is 0.3 to 1.0 me-ter thick, varying from section to sec-tion. Its lower boundary is a gradationalcolor change; its upper boundary isabrupt and coincides with the faunal andfloral extinctions. In one section (Con-tessa) we can see that the lower 5 mm ofthe boundary clay is gray and the upper 5mm is red, thus placing the upper bound-ary of the zone in the middle of the claylayer.The best known of the Umbrian sec-

tions is in the Bottaccione Gorge nearGubbio. Here some of the first work onthe identification of foraminifera in thinsection was carried out (29); the oldestknown Tertiary foraminifer, G. eu-gubina, was recognized, named, andused to define the basal Tertiary biozone(16, 26); the geomagnetic reversalstratigraphy of the Upper Cretaceousand Paleocene was established, corre-lated to the marine magnetic anomaly se-quence, and dated with foraminifera(30); and the extinction of most of thenannoplankton was shown to be syn-chronous with the disappearance of thegenus Globotruncana (27).

SCIENCE, VOL. 208

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Results from the Italian Sections

Our first experiments involved NAAof nine samples from the Bottaccionesection (two limestone samples from im-mediately above and below the boundaryplus seven limestone samples spacedover 325 m of the Cretaceous). This wassupplemented by three samples from thenearby Contessa section (two from theboundary clay and one from the basalTertiary bed). Stratigraphic positions ofthese samples are shown in Fig. 2.

Twenty-eight elements were selectedfor study because of their favorable nu-clear properties, especially neutron cap-ture cross sections, half-lives, and gam-ma-ray energies. The results of theseanalyses are presented in Fig. 3 on a log-arithmic plot to facilitate comparison ofthe relative changes in elements over awide range of concentrations. The onlypreparation given to these samples wasremoval of the CaCO3 fraction by dis-solution in dilute nitric acid. Figure 3shows elemental abundances as gram ofelement per gram of insoluble clay resi-due. The limestones generally containabout 5 percent clay. The boundary claylayer contains about 50 percent CaCO3,but this is coarse-grained calcite thatprobably crystallized during deformationlong after deposition. Chemical yields ofiridium in the acid-insoluble fraction av-eraged 44 percent for the red and grayContessa boundary clays, and this valuewas assumed for all the other samples.Twenty-seven of the 28 elements show

very similar patterns of abundance varia-tion, but iridium shows a grossly dif-ferent behavior; it increases by a factorof about 30 in coincidence with the C-Tboundary, whereas none of the other ele-ments as much as doubles with respectto an "average behavior" shown in thelower right panel of Fig. 3. Figure 4shows a typical gamma-ray spectrumused to measure the Ir abundance, 5.5parts per billion (ppb).

In follow-up experiments we analyzedfive more samples from the Bottaccionesection, eight from Gorgo a Cerbara (28kilometers north of Gubbio), and fourlarge samples of the boundary clay fromthe two sections near Gubbio and twosections about 30 km to the north (31).The chemical yield of iridium in the acid-insoluble fraction was 95 + 5 percent forthe Contessa boundary clay, and a 100percent yield was assumed for all theother samples.

but changes to logarithmic to show re-sults from 350 m below to 50m above theboundary. It is also important to notethat analyses from five stratigraphic sec-tions are plotted on the same diagram onthe basis of their stratigraphic positionabove or below the boundary. Becauseslight differences in sedimentation rateprobably exist from one section to thenext, the chronologic sequence of sam-ples from different sections may not beexactly correct. Nevertheless, Fig. 5

a b c400-_

350-

300-

Fig. 2. Stratigraphic sec-tion at the BottaccioneGorge, Gubbio (30). (a)Meter levels. (b) Systems.(c) Stages. (d) Magneticpolarity zones (black isnormal, white is reversedpolarity, letters give Gub-bio polarity zonation, num-bers are equivalent marinemagnetic anomalies). (e)Lithology. (f) Samplesused in first NAA study(samples I, J, and L arefrom equivalent positionsin the Contessa section, 2km to the northwest). (g)Formation names.

250-

200-

150-

100-

50-

Figure 5 shows the results of 29 Iranalyses completed on Italian samples.Note that the section is enlarged and thatthe scale is linear in the vicinity of the C-T boundary, where details are important,6 JUNE 1980

0-

0

I-

a)

a)

a)

0

01)

0u

a

a)

a)

0)

3.0

01)C01)00

u-i

0c

1-

U1)

0

c0

c

a0L.E0o

c_O

C0

Oa

Cf

0

0

C~

E0

0

H

0

0)o.

C

0

.0

Apt.

gives a clear picture of the general trendof iridium concentrations as a function ofstratigraphic level.The pattern, based especially on the

samples from the Bottaccione Gorge andGorgo a Cerbara, shows a steady back-ground level of - 0.3 ppb throughout theUpper Cretaceous, continuing into theuppermost bed of the Cretaceous. Thebackground level in the acid-insolubleresidues is roughly comparable to theiridium abundance measured by other

d

K-

I-

G-

f2-

c-

A-

e f g

N. (26)

L. (27)A (29)

F3+ (30)

Fl. (31)

D34

Dl1. (32)

B+

(33)

(34)

c0

N

0

E

0

z

c0

0

u

0

0

U)U,0

0..0'_o C_

0

C-)

U)

-Bonarelli level

a

0

000

_B

C-)

VI

0.- A E

01.0

1097

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

workers (19, 22, 32) in deep-sea claysediments. This level increases abruptly,by a factor of more than 30, to 9.1 ppb,the Ir abundance in the red clay from theContessa section. Iridium levels are highin clay residues from the first few beds ofTertiary limestone, but fall off to back-ground levels by 1 m above the bound-ary. For comparison, the upper dashedline in Fig. 5 shows an exponential decayfrom the boundary clay Ir level with ahalf-height of 4.6 cm.To test the possibility that iridium

might somehow be concentrated in claylayers, we subsequently analyzed twored clay samples from a short distancebelow the C-T boundary in the Bot-taccione section. One is from a distinc-tive clay layer 5 to 6 mm thick, 1.73 mbelow the boundary; the other is from a1- to 2-mm bedding-plane clay seam 0.85m below the boundary. The whole-rockanalyses of these clays showed no de-tectable Ir with limits of 0.5 and 0.24ppb, respectively. Thus neither clay lay-ers from below the C-T boundary norclay components in the limestone showevidence of Ir above the background lev-el.

The Danish Section

To test whether the iridium anomaly isa local Italian feature, it was desirable toanalyze sediments of similar age fromanother region. The sea cliff of StevnsKlint, about 50 km south of Copenhagen,is a classical area for the C-T boundaryand for the Danian or basal stage of theTertiary. A collection of up-to-date pa-pers on this and nearby areas has re-cently been published, which includesa full bibliography of earlier works(6, vol. 1).Our samples were taken at Hojerup

Church (33). At this locality the Maast-richtian, or uppermost Cretaceous, isrepresented by white chalk containingblack chert nodules in undulating layerswith amplitudes of a few meters andwavelengths of 10 to 50 m (14). These un-dulations are considered to representbryozoan banks (34). The C-T boundaryis marked by the Fiskeler, or fish clay,which is up to 35 cm thick in the deepestparts of the basins between bryozoanbanks (14) but commonly only a few cen-timeters thick, thinning or disappearing,over the tops of the banks. The fish clay

at Hojerup Church was studied in detailby Christensen et al. (14), who sub-divided it into four thin layers; we ana-lyzed a sample mixing the two internallayers (units III and IV of Christensen etal.). These layers are black or dark gray,and the lower one contains pyrite con-cretions; the layers below and above (IIand V) are light gray in color. Undis-turbed lamination in bed IV suggests thatno bottom fauna was present during itsdeposition (14). Above the fish clay, theCerithium limestone is present to a thick-ness of about 50 cm in the small basins,disappearing over the banks. It is hard,yellowish in color, and cut by abundantburrows. Above this is a thick bryozoanlimestone.The presence of a thin clay layer at the

C-T boundary in both the Italian andDanish sections is quite striking. How-ever, there are notable differences aswell. The Danish sequence was clearlydeposited in shallower water (35), andthe Danish limestones preserve an exten-sive bottom-dwelling fauna of bivalves(36), echinoderms (37), bryozoans (38),and corals (39).

Foraminiferal (40) and coccolith (41)

FTert.1 L-T/C K-

boundary 1 Clay jregion [ rt._

Lower portof UpperCretaceous

Fig. 3. Abundance variations of 28 elements in 12 samples from two Gubbio sections. Flags on "average rare earth" diagram are + 30 percentand include all rare earth aata.

1098 SCIENCE, VOL. 208

Elemental abundancesin 12 samples from

two sections at Gubbio

I

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

zonation indicates that the C-T bounda-ries at Gubbio and Stevns Klint are atleast approximately contemporaneous,and they may well be exactly synchro-nous. However, no paleomagnetic re-sults are available from Stevns Klint, sosynchroneity cannot be tested by revers-al stratigraphy.

the atomic number position of iridium.Iridium has been detected (45) in a warmspring on Mount Hood in northern Cali-fornia at a level of 7 x 10-12 g per gramof water, and in two cold-water sourcesat levels of 3 x 10-13 to 4 x 10-13 g pergram of water. Many other cold-watersources in this area had Ir levels lessthan 1 x l0-13 g/g.

The Boundary Layers

The whole-rock composition of theContessa boundary layer (a mixture ofred and gray clay) is shown in Table 3.There are two recognizable sublayers,each about 0.5 cm thick, the upper beingred in color and the lower gray. The ele-mental iron content, which may explain

Results from the Danish Section

Seven samples were taken from nearthe C-T boundary (Fig. 6). Fractions ofeach sample were treated with dilute ni-tric acid, and the residues were filtered,washed, and heated to 800°C. The yieldof acid-insoluble residue was 44.5 per-cent for the boundary fish clay and var-ied from 0.62 to 3.3 percent for the pelag-ic limestones (Table 1).Neutron activation analysis (24) and x-

ray fluorescence (XRF) (42, 43) measure-ments were made on all seven samplesboth before and after the acid treatment.This measurement regime was more so-phisticated than that used for the Italiansections studied earlier, and 48 elementswere determined.The Cretaceous and Tertiary acid-in-

soluble residues were each rather ho-mogeneous in all of the measured ele-mnents, and the two groups were onlyslightly different from each other. Theresidue from the clay boundary layerwas much different in composition(Figs. 7 and 8 and Table 2), and this sug-gested a different source for the bound-ary clay.As shown in Table 1, the Ir in the

boundary layer residue rises by about afactor of 160 over the background level(- 0.26 ppb). A 1-cm thickness of thislayer would have about 72 x 10-9 gramof Ir per square centimeter. To testwhether there is enough Ir in the sea-water to contribute to this value, wemade a measurement of the Ir in theocean off the central California coast. Inwater passed through a 0.45-,tm filter Irwas undetected, giving an upper limit of4 x 10-13 g of Ir per gram of seawater. Ifthe depth of the shallow ancient Danishsea is assumed to be less than 100 m andour limit for Ir in seawater is applicable,then the maximum Ir in the 100im col-umn of water should be 4 x 10-9 g/cm2,almost a factor of 20 lower than the ob-served value. So there was probably notenough Ir stored in the seawater to ex-plain the amount observed in the Danishboundary. Iridium has apparently notbeen detected in seawater. One tabulat-ed result (44) contains a typographicalerror that places the value for indium in6 JUNE 1980

128,000 134 Cs 1056-0

z 260o0 [ A | | | 181Hf 980 mi.

12,000

126000a19aI-r 39.8-day decay

124,000

u122,o0 0 131BatoEE0 120,000

118,000 I460 470 480 490

Gamma-ray energy (keV)

Fig. 4. Typical gamma-ray spectrum used to determine Ir abundance (5.5 ppb) in nitric acid-insoluble residues without further chemistry. Note that the entire spectrum rests on a back-ground of 118,000 counts. Detector volume was 128 CM3; length of count was 980 minutes.Count began 39.8 days after the end of the irradiation. Residue is from a Tertiary limestonesample taken 2.5 cm above the boundary at Gorgo a Cerbara (see Fig. 5).

Fig. 5. Iridium abun-dances per unitweight of 2N HNO3acid-insoluble resi-dues from Italianlimestones near theTertiary - Cretaceousboundary. Error barson abundances arethe standard devia-tions in counting ra-dioactivity. Errorbars on stratigraphicposition indicate thestratigraphic thick-ness of the sample.The dashed lineabove the boundary isan "eyeball fit" ex-ponential with a half-height of 4.6 cm. Thedashed line below theboundary is a best fitexponential (twopoints) with a half-height of 0.43 cm. Thefilled circle and errorbar are the mean andstandard deviation ofIr abundances in fourlarge samples ofboundary clay fromdifferent locations.

Age(m.y.)

1053.5- u

E

Z

: -i

o1I

8Z.

IC;7

0

Large samples ofmixed red and gray T/C clay

7 1$>1l ~ Bottaccione-o Contessa

e Acqualagna* Petriccio

Meanvalue

Gray clay Red clay -

E4.

i

I

Legend

Gubbio0 Bottaccione° Contessa

27 km north of GubbioIs AcqualagnaO Petriccioo Gorgo a Cerbaral

2 4 6 8 10

Iridium abundance (ppb)

6

70-78-

1 00-

Height aboveor below

T/C- (cm)

1 04

1 03

Gubbiomeasuredsection

(m)

-400

-360

102

20 -347.8

10

0

10 -347.5

-347.6

1 02

103

1 04 -300

- 0

1099

ac

Da

C-

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

the color, is significantly higher in theresidue of the red layer (7.7 versus 4.5percent) than in the gray, and so is theIr (9.1 + 0.6 versus 4.0 + 0.6 ppb).Boundary samples were analyzed fromthe Bottaccione Gorge nearby and twoother areas about 30 km to the north.

In samples taken near the Italianboundary layer, the chemical composi-tions of all clay fractions were roughlythe same except for the element Ir. How-ever, there are discernible differences, asshown in Fig. 9, which suggest that atleast part of the boundary layer clay hada different origin than the Cretaceousand Tertiary clays.The Danish C-T boundary clay is

somewhat thicker than 1 cm and is di-vided into four layers, as mentioned ear-lier. Only a single mixed sample from thetwo middle layers was measured, so noinformation is available on the chemicalvariations within the boundary. The av-erage Ir abundance is 29 ppb in the wholerock or 65 ppb based on the weight ofacid-insoluble residue.The whole-rock abundances and min-

eral composition of the Danish boundaryclay are shown in Table 3, and the abun-dances of pertinent trace elements areshown in Table 2. The major silicate min-erals that must be present were not de-tected, so the other mineral abundanceswere normalized to give the amount of

Table 1. Abundance of iridium in acid-in-soluble residues in the Danish section.

Abun-Abun- dancetdance of acid-

Sample* of insol-iridium uble(ppb) resi-

dues(%)

SK, +2.7 m < 0.3 3.27SK, + 1.2 m < 0.3 1.08SK, +0.7 m 0.36 ± 0.06 0.836Boundary 41.6 ± 1.88t 44.5SK, -0.5 m 0.73 ± 0.08 0.654SK, -2.2 m 0.25 + 0.08 0.621SK, -5.4m 0.30 ± 0.16 0.774

*Numeencal values are the distances above (+) orbelow (-) the boundary layer; SK, StevnsKlint. tThe boundary layer has a much higherproportion of clay than the pelagic limestones aboveand below. *Some iridium dissolved in the nitricacid. The whole-rock abundance was 28.6 ± 1.3ppb.

calcite expected from the calcium mea-surement. The boundary clay fraction isfar different chemically from the lime-stone clay fractions above and below,which are similar to each other. Pyrite ispresent in the boundary clay, and ele-ments that form water-insoluble sulfidesare greatly enhanced in this layer. Thetrace elements that are depleted arethose that often appear as clay com-ponents. The element magnesium is an

exception. Its enhancement may be dueto replacement of iron in the clay latticesin the sulfide environment or to a dif-ferent, more mafic source for the bound-ary layer clay than for the Tertiary andCretaceous clays.Recent unpublished work by D. A.

Russell of the National Museums of Can-ada and by the present authors hasshown that the boundary layer whole-rock concentration of Ir in a section nearWoodside Creek, New Zealand, is ap-proximately 20 times the average con-centration in the adjacent Cretaceousand Tertiary limestones.

A Sudden Influx of ExtraterrestrialMaterial

To test whether the anomalous iridiumat the C-T boundary in the Gubbio sec-tions is of extraterrestrial origin, we con-sidered the increases in 27 of the 28 ele-ments measured by NAA that would beexpected if the iridium in excess of thebackground level came from a sourcewith the average composition of theearth's crust. The crustal Ir abundance,less than 0.1 ppb (19, 22), is too small tobe a worldwide source for material withan Ir abundance of 6.3 ppb, as foundnear Gubbio. Extraterrestrial sourceswith Ir levels of hundreds of parts per

a b c d

- 2.7+3

*- +1.2

- +0.7

T/C

- -05

- -2.2

__ -EC)a

0 D

_ )

EIVw

--3

Mg Ti

0.3

0.2 _

0,1 _

0.0LL_Na

3

2

0L

I I Fig. 6 (left). Stratigraphic section at H0jerup Church, Stevns Klint, Denmark. (a) Lithology (C,__r~~~0 --5.4 Cerithium limestone; F, fish clay). (b) Stages. (c) Samples analyzed in this study. (d) Meter

levels. Analytical results are given in Tables I to 3. Fig. 7 (right). Major element abundancesL L Li in acid-insoluble fractions from Danish rocks near the Tertiary-Cretaceous boundary. The

crosshatched areas for the Cretaceous and Tertiary values each represent root-mean-squaredeviations for three samples. (Only two measurements of magnesium and silicon were included in the Cretaceous values.) For the boundary layersample the crosshatched areas are the standard deviations associated with counting errors. Measurements of silicon and magnesium were doneby XRF (42), all others were by NAA.

1100 qr(Fwr'P_ vnT. )n

K

at1n1m,i vv -. 4vo

MA

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

billion or higher are more likely to haveproduced the Ir anomaly. Figure 10shows that if the source had an averageearth's crust composition (46), increasessignificantly above those observedwould be expected in all 27 elements.However, for a source with average car-

bonaceous chondrite composition (46),only nickel should show an elemental in-crease greater than that observed. Asshown in Fig. 11, such an increase innickel was not observed, but the predict-ed effect is small and, given appropriateconditions, nickel oxide would dissolvein seawater (47). We conclude that thepattern of elemental abundances in theGubbio sections is compatible with an

extraterrestrial source for the anomalousiridium and incompatible with a crustalsource.The Danish boundary layer, which has

much more Ir than the Italian C-T clay,is even less likely to have had a crustalorigin. Rocks from the upper mantle(which has more Ir than the crust) haveless than 20 ppb (48) and are therefore an

unlikely worldwide source. There are,however, localized terrestrial sources

with much higher Ir abundances; for ex-

ample, nickel sulfide and chromite ores

(48) have Ir levels of hundreds and thou-sands of parts per billion, respectively.The Danish boundary layer, however,does not have enough nickel [506 partsper million (ppm)] or chromium (165

C

-a

c

.-1

w

6

4

2

0

Th

-m~

Cs

60

401

20

O0Co

0.06

0.04-

0.024

0.00

c

0

4o

0,0

Ir

Ta Hf

Fig. 8. Selected trace element abundances of Danish acid-insoluble residues. First bar is themean value [root-mean-square deviation (RMSD) is shown by the crosshatched areas] for thegiven element in the three Cretaceous residues. Second bar is the abundance (counting error isshown by solid areas) for the given element in the boundary layer residue. Third bar is the meanvalue and RMSD for the given element in the three Tertiary residues. Measurements were byNAA, except for zinc, which was measured by XRF (43). Significant amounts of nickel, zinc,cobalt, iridium, and thorium in all samples dissolved in the 2N HNO3. Very little cesium, tan-talum, or hafnium in any of the samples dissolved in the acid.

ppm) to explain its Ir in this fashion, un-

less the marine chemistry concentratedIr preferentially (even in the sulfide envi-ronment) and disposed of the other ele-ments elsewhere. The probability of

these effects occurring worldwide seems

less likely than an extraterrestrial originfor the Ir.We next consider whether the Ir

anomaly is due to an abnormal influx of

Table 2. Abundance of trace elements in the Danish boundary layer (parts per million).

(1) (1)Ele- Abundance in (2) El Abundance in (2)ment whole rock/ Abundance mee-t wholerock/ Abundance

abundance of in residue* abundance of in residueresidue* residue

Enhanced elementst Depleted elementsV 391 ±27 330 ± 31 Mn 102.0 ± 1.3 21.3 ±0.5Cr 371 ±13 358 ± 9 Rb 27 ± 7 35 ±4Co 141.6 ±1.8 57.2 ±0.7 Yt 79 ± 6 6.3 ±1.8Ni 1137 ±31 479 ±14 Zrt 144 ±11 125 ±6Cut 167 ±14 93 ± 6 Nbt 8 ± 4 6.1 ±1.8Znt 1027 + 49 378 ± 18 Cs 1.87 ± 0.19 1.51 ± 0.14As 96 ±8 68 ±4 La 61.1 ±1.6 6.8 ± 0.4Se§ 46.5 ± 0.6 12.1 ± 0.3 Ce 57.0 ± 1.2 9.7 ± 0.6Mo 29.0 ± 2.5 20.3 ± 1.4 Nd 63.4 ± 2.7 5.4 ± 0.6Ag§ 2.6 ± 0.9 3.5 ± 0.7 Sm 11.93 ± 0.08 0.781 ± 0.008In 0.245 ± 0.022 0.086 ± 0.019 Eu 2.76 ± 0.11 0.121 ± 0.010Sb 8.0 ± 0.4 6.7 ± 0.4 Th 1.84 ± 0.04 0.148 ± 0.014Ba 1175 ± 16 747 ± 11 Dy 11.24 ± 0.12 0.908 ± 0.033Ir 0.0643 ± 0.0029 0.0416 ± 0.0018 Yb 5.02 + 0.09 0.56 ± 0.05Pb* 64 + 14 28 ± 7 Lu 0.553 ± 0.031 0.083 0.004

Hf 4.34 ± 0.16 3.88 + 0.07Other elementst Ta 0.508 ± 0.011 0.500 + 0.005

Sc 20.74 + 0.16 14.30 ± 0.14 Th 7.1 ± 0.4 1.28 ± 0.06Gat 30 ± 6 19.8 ± 3.0 U 8.63 ± 0.09 0.918 + 0.024SE 1465 ±72 48.1 + 2.4Au < 0.12 0.027 ± 0.007

*Column I minus column 2 is the amount ofan element that dissolved in the acid or was lost in the firing; abundance of residue = 44.5 percent. tElements V, Ag,and In are at least 20 percent and all other "enhanced elements" are at least a factor of 3 more abundant in the boundary residue than in the other residues. All"depleted elements" are at least 20 percent less abundant in the boundary residue than in the other residues. "Other elements" do not show a consistent pattern ofboundary residue abundances relative to the others. tMeasured by hard XRF (43). §Flux monitors were used in the NAA measurements of these elements.The indicated errors are applicable for comparing the two entries for a given element, but calibration uncertainties of possibly 10 to 20 percent must be consideredwhen the values are used for other purposes.

00Lo 0

6 JUNE 1980 1101

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

extraterrestrial material at the time of theextinctions, or whether it was formed bythe normal, slow accumulation of mete-oritic material (19), followed by concen-tration in the boundary rocks by someidentifiable mechanism.There is prima facie evidence for an

abnormal influx in the observations thatthe excess iridium occurs exactly at thetime of one of the extinctions; that theextinctions were extraordinary events,which may well indicate an extraordi-nary cause; that the extinctions wereclearly worldwide; and that the iridiumanomaly is now known from two dif-ferent areas in western Europe and inNew Zealand. Furthermore, we will showin a later section that impact of a 10-kmearth-crossing asteroid, an event thatprobably occurs with about the same fre-quency as major extinctions, may haveproduced the observed physical and bio-logical effects. Nevertheless, one can in-vent two other scenarios that might lead

50

Expected increases in elemental abun

at T/C boundary if Ir anomaly source40 average"earth's crust" composition

C0

30E 0

Ecn

20

.2~~~~~~~~~~~~~~

10

Observedi

0

to concentration of normal backgroundiridium at the boundary. These appear tobe much less likely than the sudden-in-flux model, but we cannot definitely ruleout either one at present.The first scenario requires a physical

or chemical change in the ocean watersat the time of the extinctions, leading toextraction of iridium resident in the sea-water. This would require iridium con-centrations in seawater that are higherthan those presently observed. In addi-tion, it suggests that the positive iridiumanomaly should be accompanied by acompensating negative anomaly immedi-ately above, but this is not seen.The second scenario postulates a re-

duction in the deposition rate of all com-ponents of the pelagic sediment exceptfor the meteoritic dust that carries theconcentrated iridium. This scenario re-quires removal of clay but not of iridium-bearing particles, perhaps by currents ofexactly the right velocity. These currents

4

3

2

0

must have affected both the Italian andDanish areas at exactly the time of theC-T extinctions, but at none of the othertimes represented by our samples. Wefeel that this scenario is too contrived, aconclusion that is justified in more detailelsewhere (23).

In summary, we conclude that theanomalous iridium concentration at theC-T boundary is best interpreted as in-dicating an abnormal influx of extra-terrestrial material.

Negative Results of Tests for the

Supernova Hypothesis

Considerable attention has been givento the hypothesis that the C-T ex-tinctions were the result of a nearby su-pernova (I1). A rough calculation of thedistance from the assumed supernova tothe solar system, using the measuredsurface density of iridium in the Gubbio

be0

+10to

or

-C

z _ Z._ - a O- Do- c.a n-toa-u * oE* E D£ - az<CnI. >U U Uz"wMA U accmo-j U 3=FLa I_Q._ 0 = ZZ@E.,,0s ~~~~~z e-Cw o1 z i Co La I.- ) mX

R EE REE

Fig. 10 (left). Comparison of observed elemental abundance patterns in the Gubbio section samples with average patterns expected for crustalmaterial (46). Fig. 11 (right). Comparison of observed elemental abundance patterns in the Gubbio section samples with patterns expected forcarbonaceous chondrites (46).

1102 SCIENCE, VOL. 208

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

boundary layer and the amount of iridi-um expected to be blown off in the super-nova explosion, gives about 0.1 Ilight-year. The probability is about 10-9 (49)that, during the last 100 million years, asupernova occurred within this distancefrom the sun. Any mechanism with sucha low a priori probability is obviously aone-time-only theory. Nevertheless, be-cause the theory could be subjected todirect experimental tests, it was treatedas a real possibility until we obtained twoother independent pieces of evidencethat forced us to reject it.

Elements heavier than nickel can beproduced in stars only by neutron cap-ture followed by beta decay. The mostintense source of neutrons so far postu-lated is that produced by the gravitation-al collapse of the core of a star that leadsimmediately to a supernova explosion.In this environment the rapid capture ofneutrons ("r process") leads to the for-mation of the heaviest known isotopes.The slower capture of neutrons by heavyisotopes in highly evolved stars ("s pro-cess") leads to a different mix of iso-topes (50).One heavy isotope in particular of-

fered the possibility of testing the super-nova hypothesis; this is 244Pu, with ahalf-life of 80.5 x 106 years. The ex-plosion of a supernova should send outan expanding shell of newly createdheavy elements, with a ratio of Ir atomsto 244Pu atoms equal to about 103. Thisvalue is inferred from the existence of ananomaly in the meteoritic abundance,kifheavy xenon isotopes that is interpeldas being due to the fission of 244Pu (51).Any 244PU incorporated in the earth at thetime of the creation of the solar system,about 4.7 billion years ago, would havedecayed by 58 half-lives, or by a factor of1017, which would make it quite un-

detectable in the Gubbio section bythe most sensitive techniques available.If the C-T extinctions were due to a su-pernova, and if this were the source ofthe anomalous Ir, each Ir atom shouldhave been accompanied by about 10-3244Pu atom, and this 244PU would have de-cayed by only a factor of 2.

Plutonium-244 is easily detected bothby mass spectrometry and by NAA. Theformer is more sensitive, but the latterwas immediately available. In NAA,which we utilized, 244pU is converted to245N1, which has a half-life of 10 hoursand emits many characteristic gammarays and x-rays. Plutonium was chem-ically separated from 25- and 50-g batch-es of boundary clay and from a 50-gbatch of bedding clay from below theC-T boundary, and nearly "mass-free"6 JUNE 1980

samples were obtained-no carrierswere added. Chemical separations werealso performed on the plutonium fractionafter the neutron irradiation. No signifi-cant gamma radiation was observed, oth-er than that associated with the pluto-nium isotopes. In order to measure ourchemical yields, Gubbio acid-soluble andacid-insoluble residues were spiked withsmall amounts of 238Pu tracer. This pluto-nium isotope is easily detectable throughits alpha decay, as its half-life is only87.7 years. In addition, one of the sam-

ples was spiked with 244P. Figure 12ashows the gamma-ray spectrum of the

ppmIFig. 9. Some of theelement abundancesmeasured in acid-in-soluble residues ofCretaceous, boundarylayer, and Tertiaryrocks near Gub-bio. Data include allsamples from thatarea measured within19 m of the boundary.There were four sam-ples from each of thethree layers; thecrosshatched areasare the standard de-viations. The abun-dance patterns forsamples from 27km north of Gubbioare similar to thoseshown.

12

8

4

0

ppm

6

4

2

0 L

Th

Hf

I1nThi

sample spiked with about 20 picogramsof 2"Pu; it indicates both the sensitivityofNAA for the detection of 244Pu and thefreedom of the purified sample from oth-er elements that might interfere with thedetection of 244Pu. The plutonium isotop-ic ratios in this sample and in the tracerwere also measured with a single-direc-tion-focusing mass spectrometer 5 feet inradius.No 244PU was detected in the Gubbio

samples (Fig. 12b), with a detection limitof less than 10 percent of the amount thatwould be expected to accompany themeasured iridium if a supernova were re-

ppm

30

20

10

I0

ppm

60

40

20

0

SC

Co

1.2

0.8

0.4

0.0

ppm

12

8

4

0

Ti

- i, >1o Oa.m

= IV0

Cs

Table 3. Whole-rock composition of the Gubbio and Danish boundary layers (percent).

Abundance in boundary layertElement* or Gubbio Denmark

mineral (Contessa)measured Measured Normalized

SiO2 27.7 ± 0.6 29.0 ± 0.6A1203 12.19 ± 0.15 8.01 ± 0.17FeOt 4.53 ± 0.05 4.35 ± 0.04MgO 1.10 ± 0.07 3.07 ± 0.10CaO 22.6 ± 0.4 23.1 ± 0.4Na2O 0.1806 ± 0.0036 0.0888 ± 0.0018K20 2.46 ± 0.20 0.38 ± 0.04TiO2 0.521 ± 0.022 0.324 ± 0.016S2- Not detected - 1.1P043- Not detected 0.92 ± 0.09C02§ 17.7 ± 0.3 18.4 ± 0.3X Trace elements - 0.2 - 0.3Sum 89.2 ± 0.8 90.3 ± 1.0Differencell 10.8 ± 0.8 9.7 ± 1.0

Calcite - 90 41.5 (norm)Quartz 5-7 - 3Pyrite - 5 - 2Illite 2-3 - I

*Abundance values are for element expressed as form shown. tElements Si, Ca, Mg, S, P, and Gubbio Tiwere measured by soft XRF (42). Some S may be lost in this sample preparation procedure. The Denmark Tiwas measured by hard XRF (43). All other measurements were by NAA. Mineral analyses were done by M.Ghiorso and I. S. E. Carmichael by x-ray diffraction. *Total Fe expressed as FeO. §The CO2 abun-dance was calculated from the Ca abundance by assuming all Ca was present as the carbonate. Irrhedifference is mainly water and organic material.

1103

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

sponsible for the latter. The ocean, how-ever, can produce chemical and physicalchanges in depositing materials as wellas diagenetic alterations in the depositedsediments, so the absence of measurable244Pu is not an absolutely conclusive ar-gument.The second method that was used to

b Expected Expected

100 position of position of

Am Ka2 ; V Am Ka 1

50

*% . @09% 0 @0

0 0Il .

I590 100 110 120

Gamma-ray energy (keV)

test whether a supernova was respon-sible for the iridium anomaly involved ameasurement of the isotopic ratio of irid-ium in the boundary material. Iridiumhas two stable isotopes, 191 and 193,which would be expected to occur inabout the same relative abundances, 37.3to 62.7 percent, in all solar system mate-

Fig. 12. Gamma-rayspectra of Pu frac-tions from acid-in-soluble residues of ir-radiated boundarylayer clay samples1103 from Gubbio. (a)

K13 Sample had beenspiked with 2,"Pu and238Pu containing rela-

___ tively small amountsof 239PU, 240PU, and242pu. Dashed linesshow expected ener-gies and abundancesof 24sPu and equil-ibrated daughter rad-iations normalized tothe 327.2-keV gam-ma ray of 245PU (notshown). (b) Samplehad been spiked with238Pu containing rela-tively small amountsof 239Pu, 240PU, and242pU. No 244PU was

130 detected.

n,ion

Fig. 13. (a) Gamma-s81Hf ray spectrum of irra-\ diated acid-insoluble

residue from bound-ary layer clay at Gub-bio without chemistrybefore or after irradia-tion. (b) Same asabove with chemistrybefore and after irra-diation used in isotop-ic ratio determina-tions. Counting peri-ods, decay periods,

A and chemical yieldsare different for thetwo spectra.

290 300 310 320 330 340 350

Gamma-ray energy (keV)

1104

rial because of mixing the protosolar gascloud. However, different supernovasshould produce iridium with differentisotopic ratios because of differences inthe contributions of the r and s processesoccasioned by variations in neutronfluxes, reaction times, and so on, fromone supernova to the next. According tothis generally accepted picture, solarsystem iridium is a mixture of that ele-ment produced by all the supernovasthat ejected material into the gaseousnebula that eventually condensed toform the sun and its planets. A particularsupernova would produce Ir with an iso-topic ratio that might differ from that ofsolar system material by as much as afactor of 2 (52).We therefore compared the isotopic

ratio of Ir from the C-T boundary claywith that of ordinary Ir, using NAA. Thisis a new technique (23), which we devel-oped because of the extreme difficulty ofdetermining Ir isotope ratios by massspectrometry. In our earlier analyticalwork we used only the 74-day 19Ir, madefrom 191kr by neutron capture. But in thisnew work we also measured the 18-hour19Ir made from the heavier Ir isotope,and extensive chemical separations be-fore and after the neutron irradiationswere necessary. Figure 13 contrasts atypical gamma-ray spectrum of the kindused in the isotopic ratio measurementwith one used in an Ir abundance deter-mination. This comparison demonstratesthe need for chemical purification of theiridium fraction as well as the lack of ma-jor interfering radiations.The final result is that the isotopic ra-

tio of the boundary Ir differs by only0.03 ± 0.65 percent (mean + 1 standarddeviation) from that of the standard.From this, we conclude that the 191Ir/193Irratio in the boundary layer and the stan-dard do not differ significantly by morethan 1.5 percent. Therefore the anoma-lous Ir is very likely of solar system ori-gin, and did not come from a supernovaor other source outside the solar system(53)-for example, during passage of theearth through the galactic arms. [In avery recent paper, Napier and Clubesuggest that catastrophic events couldarise from the latter (54).]

The Asteroid Impact Hypothesis

After obtaining negative results in ourtests of the supernova hypothesis, wewere left with the question of what extra-terrestrial source within the solar systemcould supply the observed iridium andalso cause the extinctions. We consid-

SCIENCE, VOL. 208

0

cc

cL

0

EE

Co

4-

c

c

0

to

fo

E

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

ered and rejected a number of hypothe-ses (23); finally, we found that an exten-sion of the meteorite impact hypothesis(55, 56) provided a scenario that explainsmost or all of the biological and physicalevidence. In brief, our hypothesis sug-gests that an asteroid struck the earth,formed an impact crater, and some of thedust-sized material ejected from the cra-ter reached the stratosphere and wasspread around the globe. This dust ef-fectively prevented sunlight from reach-ing the surface for a period of severalyears, inmththe dust settled to earth.Loss of sunlight suppressed photosyn-thesis, and as a result most food chainscollapsed and the extinctions resulted.Several lines of evidence support this hy-pothesis, as discussed in the next fewsections. The size of the impacting ob-ject can be calculated from four inde-pendent sets of observations, with goodagreement among the four different di-ameter estimates.

Earth-Crossing Asteroids and

Earth Craters

Two quite different data bases showthat for the last billion years the earthhas been bombarded by a nearly con-stant flux of asteroids that cross theearth's orbit. One data base comes fromastronomical observations of such aster-oids and a tabulation of their orbital pa-rameters and their distribution of diame-ters (57). Opik (58) computed that themean time to collision with the earth fora given earth-crossing asteroid is about200 million years. To a first approxima-tion, the number of these objects with di-ameters greater than d drops roughly asthe inverse square of d. E. M. Shoe-maker [cited in (59, 60)] and Wetherill(60) independently estimated that thereare at present about 700 earth-crossingasteroids with diameters greater than 1km (Apollo objects), so there should beabout seven with diameters greater than10 km. This assumes that the power lawwith exponent -2 extends from the ac-cessible 1-km-diameter range into the 10-km-diameter range. If one accepts thenumbers given above, the mean time tocollision for an earth-crossing asteroidwith a diameter of 10 km or more wouldbe 200 million years divided by 7, or- 30 million years. In a more sophisti-cated calculation, Shoemaker (61) esti-mates that a mean collision time of 100million years is consistent with a diame-ter of 10 km, which is the value we willadopt. A discussion of cratering data,which leads to similar estimates, is given6 JUNE 1980

in Grieve and Robertson's review article(62) on the size and age distribution oflarge impact craters on the earth. Ratherthan present our lengthy justification (23)for the estimates based on the crateringdata, we will simply report the evalua-tion of Grieve (63), who wrote: "I canfind nothing in your data that is at oddswith your premise." Grieve also esti-mates that the diameter of the craterformed by the impact of a 10-km asteroidwould be about 200 km (63). This sectionof our article has thus been greatly con-densed now that we have heard from ex-perienced students of the two data basesinvolved.

Krakatoa

The largest well-studied terrestrial ex-plosion in historical times was that of theisland volcano, Krakatoa, in the SundaStrait, between Java and Sumatra (64).Since this event provides the best avail-able data on injection of dust into thestratosphere, we give here a brief sum-mary of relevant information.On 26 and 27 August 1883, Krakatoa

underwent volcanic eruptions that shotan estimated 18 km3 of material into theatmosphere, of which about 4 km3 endedup in the stratosphere, where it stayedfor 2 to 2.5 years. Dust from the ex-plosion circled the globe, quickly givingrise to brilliant sunsets seen worldwide.Recent measurements of the 14C injectedinto the atmosphere by nuclear bombtests confirm the rapid mixing (about 1year) between hemispheres (65). If wetake the estimated dust mass in thestratosphere (4 km3 times the assumedlow density of 2 g/cm3) and spread it uni-formly over the globe, it amounts to1.6 x 10-3 g/cm2. This layer did not ab-sorb much of the incident radiation on a"straight-through" basis. However, if itwere increased by a factor of about 103 (arough prediction of our theory), it ismost probable that the sunlight would beattenuated to a high degree.

Since the time for the colored sunsetsto disappear after Krakatoa is frequentlygiven as 2 to 2.5 years, we have assumedthat the asteroid impact material in thestratosphere settled in a few years. Thus,65 million years ago, day could havebeen turned into night for a period ofseveral years, after which time the atmo-sphere would return relatively quickly toits normal transparent state.What happened during the Krakatoa

explosions can be expected to happen toa much greater extent during the impactof a large asteroid. An interesting dif-

ference is that extreme atmospheric tur-bulence would follow the impact. The as-teroid would enter the atmosphere atroughly 25 km/sec and would "punch ahole" in the atmosphere about 10 kmacross. The kinetic energy of the aster-oid is approximately equivalent to that of108 megatons of TNT.

Size of the Impacting Object

If we are correct in our hypothesis thatthe C-T extinctions were due to the im-pact of an earth-crossing asteroid, thereare four independent ways to calculatethe size of the object. The four ways andthe results obtained are outlined below.

1) The postulated size of the incomingasteroid was first computed from theiridium measurements in the Italian sec-tions, the tabulated Ir abundances (66) intype I carbonaceous chondrites (CI),which are considered to be typical solarsystem material, and the fraction oferupted material estimated to end up inthe stratosphere. If we neglect the latterfraction for the moment, the asteroidmass is given by M = sAlf, where s isthe surface density of Ir (measured atGubbio to be 8 x 10-9 g/cm2), A is thesurface area of the earth, andf is the Irmass fractional abundance in CI meteor-ites (0.5 x 10-6). This preliminary valueof the asteroid mass, 7.4 x 1016g, is thendivided by the estimated fraction stayingin the stratosphere, 0.22, to giveM = 3.4 x 10'7 g. The "Krakatoa frac-tion," 0.22, is used simply because it isthe only relevant number available. Itcould differ seriously from the correctvalue, however, as the two explosionsare of quite different character. At a den-sity of 2.2 g/cm3 (67), the diameter of theasteroid would be 6.6 km.

2) The second estimate comes fromdata on earth-crossing asteroids and thecraters they have made on the earth'ssurface. In a sense, the second estimatecomes from two quite different databases-one from geology and the otherfrom astronomy. Calculations of the as-teroid diameter can be made from bothdata bases, but they will not really be in-dependent since the two data bases areknown to be consistent with each other.As shown in an earlier section, the mostbelievable calculation of the mean timebetween collisions of the earth and aster-oids equal to or larger than 10 km in di-ameter is about 100 million years. Thesmaller the diameter the more frequentare the collisions, so our desire to fit notonly the C-T extinction, but earlier onesas well, sets the mean time between ex-

1105

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

tinctions at about 100 million years andthe diameter at about 10 km.

3) The third method of estimating thesize of the asteroid comes from the pos-sibility that the 1-cm boundary layer atGubbio and Copenhagen is composed ofmaterial that fell out of the stratosphere,and is not related to the clay that ismixed in with the limestone above andbelow it. This is quite a surprising pre-diction of the hypothesis, since the mostobvious explanation for the origin of theclay is that it had the same source as theclay impurity in the rest of the Cre-taceous and Tertiary limestone, and thatit is nearly free of primary CaCO3 be-cause the extinction temporarily de-stroyed the calcite-producing planktonfor about 5000 years. But as discussedearlier, the material in the boundary lay-er is of a different character from the clayabove and below it, whereas the lattertwo clays are very similar. To estimatethe diameter of the asteroid, one can usethe surface density of the boundary layer(about 2.5 g/cm2), together with an esti-mate of the fraction of that materialwhich is of asteroidal origin. The aster-oid diameter is then calculated to be 7.5km. The numbers used in this calculationare the following: clay fraction in theboundary layer, 0.5; density of the aster-oid, 2.2 g/cm3; mass of crustal materialthrown up per unit mass of asteroid,- 60 (63); fraction of excavated materialdelivered to the stratosphere, 0.22 (fromthe Krakatoa measurements). If one usesdifferent numbers, the diameter changesonly by the cube root of the ratio of inputvalues.The first and the third methods are in-

dependent, even though they both de-pend on measurements made on theboundary material. This can best be ap-preciated by noting that if the Ir abun-dance were about the same in the earth'scrust as it is in meteorites, the iridiumanomaly seen in Fig. 5 would not exist.Therefore, method 1 would not exist ei-ther. The fact that method 3 could still beused is the indicator of the relative inde-pendence of the two methods.

4) The fourth method is not yet able toset close limits on the mass of the in-coming asteroid, but it leads to consist-ent results. This method derives from theneed to make the sky much more opaquethan it was in the years following theKrakatoa explosion. If it is assumed thatthe Krakatoa dust cloud attenuated thevertically incident sunlight by about 3percent, then an explosion involving 33times as much material would reduce thelight intensity to lie. The stratosphericmass due to an explosion of the magni-tude calculated in the three earlier meth-

1106

ods-about 1000 times that of Kra-katoa-would then be expected to re-duce the sunlight to exp(-30) = 10-'3.This is, of course, much more light atten-uation than is needed to stop photosyn-thesis. But the model used in this sim-plistic calculation assumes that the dustis a perfect absorber of the incident light.A reasonable albedo coupled with aslight reduction in the mass of dust canraise the light intensity under the as-sumed "optical depth" to 10-7 of normalsunlight, corresponding to 10 percent offull moonlight.Although it is impossible to make an

accurate estimate of the asteroid's sizefrom the Krakatoa extrapolation, itwould have been necessary to abandonthe hypothesis had a serious discrepancybeen apparent. In the absence of goodmeasurements of the solar constant inthe 1880's, it can only be said that thefourth method leads to asteroid sizes thatare consistent with the other three.

Until we understand the reasons forthe factor of 10 difference in Ir content ofthe boundary clay between Denmark andItaly, we will be faced with different val-ues for the asteroid diameter based onthe first method. The "Danish diameter"is then 6.6 km x 101/3 = 14 km. The sec-ond and third estimates are unchanged;the second does not involve measure-ments made on the boundary layer, andthe third uses the thickness of the clay,which is only slightly greater in Denmarkthan in Italy. The fourth method is basedon such an uncertain attenuation value,from Krakatoa, that it is not worth re-calculating. We conclude that the dataare consistent with an impacting asteroidwith a diameter of about 10 + 4 km.

Biological Effects

A temporary absence of sunlightwould effectively shut off photosynthesisand thus attack food chains at their ori-gins. In a general way the effects to beexpected from such an event are whatone sees in the paleontological record ofthe extinction.The food chain in the open ocean is

based on microscopic floating plants,such as the coccolith-producing algae,which show a nearly complete ex-tinction. The animals at successivelyhigher levels in this food chain were alsovery strongly affected, with nearly totalextinction of the foraminifera and com-plete disappearance of the belemnites,ammonites, and marine reptiles.A second food chain is based on land

plants. Among these plants, existing in-dividuals would die, or at least stop pro-

ducing new growth, during an interval ofdarkness, but after light returned theywould regenerate from seeds, spores,and existing root systems. However, thelarge herbivorous and carnivorous ani-mals that were directly or indirectly de-pendent on this vegetation would be-come extinct. Russell (2) states that "noterrestrial vertebrate heavier than about25 kg is known to have survived the ex-tinctions." Many smaller terrestrial ver-tebrates did survive, including the ances-tral mammals, and they may have beenable to do this by feeding on insects anddecaying vegetation.The situation among shallow marine

bottom-dwelling invertebrates is lessclear; some groups became extinct andothers survived. A possible base for atemporary food chain in this environ-ment is nutrients originating from decay-ing land plants and animals and broughtby rivers to the shallow marine waters.We will not go further into this matter,

but we refer the reader to the pro-ceedings of the 1976 Ottawa meeting onthe C-T extinctions. This volume repro-duces an extensive discussion among theparticipants of what would happen if thesunlight were temporarily "turned off"(5, pp. 144-149). Those involved in thediscussion seemed to agree that many as-pects of the extinction pattern could beexplained by this mechanism, although anumber of puzzles remained.We must note, finally, an aspect of the

biological record that does not appear tobe in accord with the asteroid impact hy-pothesis or with any sudden, violentmechanism. Extinction of the foraminif-era and nannoplankton occurs within re-versed geomagnetic polarity zone Gub-bio G- in the Gubbio section (30). Butlerand co-workers (68, 69) have studied thenonmarine sequence of the San Juan Ba-sin of New Mexico and have found apolarity sequence that appears to be cor-related with the reversal sequence atGubbio. In the San Juan Basin, the high-est dinosaur fossils are found in the nor-mal polarity zorqe (anomaly 29) that fol-lows what is identified as the Gubbio G-zone. It would thus appear that the dino-saur and foram-nannoplankton ex-tinctions were not synchronous. (Ex-tinctions occurring in the same polarityzone in distant sections would not estab-lish either synchroneity or diachroneity.)Three commpnts on the San Juan Basinwork have been published (70) calling at-tention to the possibility of an uncon-formity at the boundary, in which casethe correlation of the magnetic polarityzones could be in error and the ex-tinctions might still be synchronous.Lindsay et al. (69) argue strongly against

SCIENCE, VOL. 208

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

a major hiatus, but admit that "the case

is not completely closed." Russell (71)has noted stratigraphic evidence againsta diachronous extinction in the continen-tal and marine realm.

Resolution of the question of whetherthe extinctions could have been synchro-nous will depend on further paleomag-netic studies. In the meantime we canstate that the asteroid impact hypothesispredicts that the apparently diachronoustiming of the foram-nannofossil and di-nosaur extinctions will eventually beshown to be incorrect.

Problems in Boundary Clay Composition

One would expect from the simplestconsiderations of our hypothesis that theboundary layer resulted from crustal ma-

terial (enriched in certain elements bythe asteroidal matter) that was distrib-uted worldwide in the stratosphere andthen fell into the ocean. This materialwould be subjected to chemical andphysical processes in the atmosphereand then in the ocean, which would alterthe composition. The enhancements ofmetals having water-insoluble sulfides inthe Danish C-T boundary compared tothe Italian might be related to an anaero-

bic environment during deposition of theformer and an aerobic one for the latter.Hydrogen sulfide can be produced bybacteria in oxygen-deficient waters, andthis would precipitate those metals ifthey were available. This would not,however, explain the striking depletionof some trace elements in the Danishboundary or its very high Ir abundance.If chondritic Ir with an abundance of

500 ppb were diluted 60-fold withcrustal material, the Ir abundance shouldbe 8 ppb rather than the 65 ppb ob-served. Possible solutions to these diffi-culties may arise when better estimatesof the extent of mixing of asteroidal andterrestrial material in the atmosphere are

made, and when the boundary layerchemistry is studied at additional loca-tions and a better understanding of themarine chemistry is achieved..

Implications

Among the many implications of theasteroid impact hypothesis, if it is cor-

rect, two stand out prominently. First, ifthe C-T extinctions were caused by an

impact event, the same could be true ofthe earlier major extinctions as well.There have been five such extinctionssince the end of the Precambrian, 570million years ago, which matches well

6 JUNE 1980

the probable interval of about 100 millionyears between collisions with 10-km-di-ameter objects. Discussions of these ex-tinction events generally list the orga-

nisms affected according to taxonomicgroupings; it would be more useful tohave this information given in terms ofinterpreted ecological or food-chaingroupings. It will also be important tocarry out iridium analyses in completestratigraphic sections across these otherboundaries. However, E. Shoemaker(private communication) predicts that ifsome of the extinctions were caused bythe collision of a "fresh" comet (mostlyice), the Ir anomaly would not be seen

even though the extinction mechanismwas via the same dust cloud of crustalmaterial, so the absence of a higher Irconcentration at, for example, the Per-mian-Triassic boundary would not inval-idate our hypothesis. According to Shoe-maker, cometary collisions in this sizerange could be twice as frequent as aster-oidal collisions.

Second, we would like to find the cra-ter produced by the impacting object.Only three craters 100 km or more in di-ameter are known (62). Two of these(Sudbury and Vredefort) are of Pre-cambrian age. For the other, PopigayCrater in Siberia, a stratigraphic age ofLate Cretaceous to Quaternary and a po-

tassium-argon date of 28.8 million years(no further details given) have been re-

ported (72, 73). Thus, Popigay Crater isprobably too young, and at 100-km-di-ameter probably also too small, to be theC-T impact site. There is about a 2/3probability that the object fell in theocean. Since the probable diameter ofthe object, 10 km, is twice the typicaloceanic depth, a crater would be pro-duced on the ocean bottom and pulver-ized rock could be ejected. However, inthis event we are unlikely to find the cra-

ter, since bathymetric information is notsufficiently detailed and since a sub-stantial portion of the pre-Tertiary oceanhas been subducted.

References and Notes

1. D. A. Russell, Geol. Assoc. Can. Spec. Rep. 13(1975), p. 119.

2. ,in(5),p. 11.

3. M. B. Cita and 1. Premoli Silva, Riv. Ital. Pa-leontol. Stratigr. Mem. 14 (1974), p. 193.

4. D. A. Russell, Annu. Rev. Earth Planet. Sci. 7,163 (1979).

5. K-TEC group (P. Beland et al.), Cretaceous-Tertiary Extinctions and Possible Terrestrialand Extraterrestrial Causes (Proceedings ofWorkshop, National Museum of Natural Sci-ences, Ottawa, 16 and 17 Nov. 1976).

6. T. Birkelund and R. G. Bromley, Eds., Cre-taceous-Tertiary Boundary Events, vol. 1, TheMaastrichtian and Danian of Denmark (Sym-posium, University of Copenhagen, Copenha-gen, 1979); W. K. Christiansen and T. Birke-lund, Eds., ibid., vol. 2, Proceedings.

7. H. Tappan, Palaeogeogr. Palaeoclimatol.Palaeoecol. 4, 187 (1968); T. R. Worsley, Na-ture (London) 230, 318 (1971); W. T. Holser,ibid. 267, 403 (1977); D. M. McLean, Science

200, 1060 (1978); ibid. 201, 401 (1978); S. Gart-ner and J. Keany, Geology 6, 708 (1978).

8. E. G. Kauffman, in (6), vol. 2, p. 29.9. A. G. Fischer, in (6), vol.2, p. 11;__ and M.

A. Arthur, Soc. Econ. Paleontol. Mineral.Spec. Publ. 25 (1977), p. 19.

10. J. F. Simpson, Geol. Soc. Am. Bull. 77, 197(1966); J. D. Hays, ibid. 82, 2433 (1971); C. G.A. Harrison and J. M. Prospero, Nature (Lon-don) 250, 563 (1974).

11. 0. H. Schindewolf, Neues Jahrb. Geol. Pa-laeontol. Monatsh. 1954, 451 (1954); ibid. 1958,270 (1958); A. R. Leoblich, Jr., and H. Tappan,Geol. Soc. Am. Bull. 75, 367 (1964); V. I. Kra-sovski and I. S. Shklovsky, Dokl. Akad. NaukSSSR 116, 197 (1957); K. D. Terry and W. H.Tucker, Science I59, 421 (1968); H. Laster, ibid.160, 1138 (1968); W. H. Tucker and K. D. Terry,ibid., p. 1138; D. Russell and W. H. Tucker, Na-ture (London) 229, 553 (1971); M. A. Ruderman,Science 184, 1079 (1974); R. C. Whitten, J. Cuz-zi, W. J. Borucki, J. H. Wolfe, Nature (London)263, 398 (1976).

12. S. Gartner and J. P. McGuirk, Science 206, 1272(1979).

13. A. Boersma and N. Schackleton, in (6), vol. 2,p. 50; B. Buchardt and N. 0. Jorgensen, in (6),vol. 2, p. 54.

14. L. Christensen, S. Fregerslev, A. Simonsen, J.Thiede, Bull. Geol. Soc. Den. 22, 193 (1973).

15. N. 0. Jorgensen, in (6), vol. 1, p. 33, vol. 2, p.62; M. Renard, in (6), vol. 2, p. 70.

16. H. P. Luterbacher and I. Premoli Silva, Riv.Ital. Paleontol. Stratigr. 70, 67 (1964).

17. H. Pettersson and H. Rotschi, Geochim. Cos-mochim. Acta 2, 81 (1952).

18. V. M. Goldschmidt, Geochemistry (OxfordUniv. Press, New York, 1954).

19. J. L. Barker, Jr., and E. Anders, Geochim. Cos-mochim. Acta 32, 627 (1968).

20. R. Ganapathy, D. E. Brownlee, P. W. Hodge,Science 201, 1119 (1978).

21. A. M. Sarna-Wojcicki, H. R. Bowman, D.Marchand, E. Helley, private communication.

22. J. H. Crocket and H. Y. Kuo, Geochim. Cosmo-chim. Acta 43, 831 (1979).

23. These are briefly discussed in L. W. Alvarez,W. Alvarez, F. Asaro, H. V. Michel, Univ. Cal-if. Lawrence Berkeley Lab. Rep. LBL-9666(1979).

24. A description of the NAA techniques is given inAlvarez et al. (23), appendix II; I. Perlman andF. Asaro, in Science and Archaeology, R. H.Brill, Ed. (MIT Press, Cambridge, Mass., 1971),p. 182.

25. These limestones belong to the Umbrian se-quence, of Jurassic to Miocene age, which hasbeen described in V. Bortolotti, P. Passerini, M.Sagri, G. Sestini, Sediment. Geol. 4, 341 (1970);A. Jacobacci, E. Centamore, M. Chiocchini, N.Malferrari, G. Martelli, A. Micarelli, Note Es-plicative Carta Geologica d'Italia (1:50,000),Foglio 190: "Cagli" (Rome, 1974).

26. H. P. Luterbacher and I. Premoli Silva, Riv.Ital. Paleontol. Stratigr. 68, 253 (1962); I. Pre-moli Silva, L. Paggi, S. Monechi, Mem. Soc.Geol. Ital. 15, 21(1976).

27. S. Monechi, in (6), vol. 2, p. 164.28. D. V. Kent, Geology 5, 769 (1977); M. A. Ar-

thur, thesis, Princeton University (1979).29. 0. Renz, Eclogae Geol. Helv. 29, 1 (1936); Serv.

Geol. Ital. Mem. Descr. Carta Geol. Ital. 29, 1(1936).

30. M. A. Arthur and A. G. Fischer, Geol. Soc. Am.Bull. 88, 367 (1977); I. Premoli Silva, ibid., p.371; W. Lowrie and W. Alvarez, ibid., p. 374;W. M. Roggenthen and G. Napoleone, ibid., p.378; W. Alvarez, M. A. Arthur, A. G. Fischer,W. Lowrie, G. Napoleone, I. Premoli Silva, W.M. Roggenthen, ibid., p. 383; W. Lowrie and W.Alvarez, Geophys. J. R. Astron. Soc. 51, 561(1977); W. Alvarez and W. Lowrie, ibid. 55, 1(1978).

31. Locations of the sections studied are: (i) Bot-taccione Gorge at Gubbio: 43°21.9'N, 1r35.0'E(0'7.9' east of Rome); (ii) Contessa Valley, 3 kmnorthwest of Gubbio: 43°22.6'N, 12°33.7'E(0'6.6' east of Rome); (iii) Petriccio suspensionbridge, 2.3 km west-southwest of Acqualagna:43°36.7'N, 12°38.7'E (0lI1.6' east of Rome); (iv)Acqualagna, road cut 0.8 km southeast of town:43°36.7'N, 12°40.8'E (0°13.7' east of Rome); and(v) Gorgo a Cerbara: 43°36.1'N, 12°33.6'E(0°6.5' east of Rome). We thank E. Sannipoli,W. S. Leith, and S. Marshak for help in sam-pling these sections.

32. J. H. Crocket, J. D. McDougall, R. C. Harriss,Geochim. Cosmochim. Acta 37, 2547 (1973).

33. Location: 5516.7'N, 12°26.5'E. We thank I.Bank and S. Gregersen for taking W.A. to thislocality.

34. A. Rosenkrantz and H. W. Rasmussen, Guide toExcursions A42 and C37 (21st International

1107

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Geological Congress, Copenhagen, 1960), part1, pp. 1-17.

35. F. Surlyk, in (6), vol. 1, p. 164.36. C. Heinberg, in (6), vol. 1, p. 58.37. H. W. Rasmussen, in (6), vol. 1, p. 65; P. Grave-

sen, in (6), vol. 1, p. 72; U. Asgaard, in (6), vol.1, p. 74.

38. E. Hakansson and E. Thomsen, in (6), vol. 1, p.78.

39. S. Floris, in (6), vol. 1, p. 92.40. I. Bang, in (6), vol. 1, p. 108.41. K. Perch-Nielsen, in (6), vol. 1, p. 115.42. Soft x-ray fluorescence measurements for major

element determinations were made by S. Flex-ser and M. Sturz of Lawrence Berkeley Labora-tory.

43. Hard x-ray fluorescence measurements for traceelement determinations were made by R. D.Giauque of Lawrence Berkeley Laboratory.

44. F. G. Walton Smith, Ed. CRC Handbook ofMa-rine Science (CRC Press. Cleveland, 1974), vol.1, p. 11.

45. H. A. Wollenberg et al., Univ. Calif. LawrenceBerkeley Lab. Rep. LBL-7092, revised 1980.

46. Encyclopaedia Britannica (Benton, Chicago,ed. 15, 1974), vol. 6, p. 702.

47. K. K. Turekian, Oceans (Prentice-Hall, Engle-wood Cliffs, N.J., 1976), p. 122.

48. J. H. Crocket, Can. Mineral. 17, 391 (1979); J.R. Ross and R. Keays, ibid., p. 417.

49. I. S. Shklovsky, Supernovae (Wiley, New York,1968), p. 377.

50. D. D. Clayton, Principles of Stellar Evolutionand Nucleosynthesis (McGraw-Hill, New York,1968), pp. 546-606.

51. D. N. Schramm, Annu. Rev. Astron. Astrophys.12, 389 (1974).

52. C. F. McKee, personal communication.53. These observations were reported at the Ameri-

can Geophysical Union meeting in May 1979and at the Geological Society of America meet-

ing in November 1979 [W. Alvarez, L. W. Alva-rez, F. Asaro, H. V. Michel, Eos 60, 734 (1979);Geol. Soc. Am. Abstr. Programs 11, 350 (1979)].

54. W. M. Napier and S. V. M. Clube, Nature (Lon-don) 282, 455 (1979).

55. H. C. Urey, ibid. 242, 32 (1973).56. E. J. Opik, Ir. Astron. J. 5 (No. 1), 34 (1958).57. E. M. Shoemaker, J. G. Williams, E. F. Helin,

R. F. Wolfe, in Asteroids, T. Gehrels, Ed.(Univ. of Arizona Press, Tucson, 1979), pp. 253-282.

58. E. J. Opik, Adv. Astron. Astrophys. 2, 220(1963); ibid. 4, 302 (1964); ibid. 8, 108 (1971).These review articles give references to Opik'sextensive bibliography on meteorites, Apolloobjects, and -asteroids.

59. C. R. Chapman, J. G. Williams, W. K. Hart-mann, Annu. Rev. Astron. Astrophys. 16, 33(1978).

60. G. W. Wetherill, Sci. Am. 240 (No. 3), 54 (1979).61. E. M. Shoemaker, personal communication.62. R. A. F. Grieve and P. B. Robertson, Icarus 38,

212 (1979).63. R. A. F. Grieve, personal communication.64. G. J. Symons, Ed., The Eruption of Krakatoa

and Subsequent Phenomena (Report of the Kra-katoa Committee of the Royal Society, Harri-son, London, 1888).

65. I. U. Olson and I. Karlen, Am. J. Sci. Radio-carbon Suppl. 7 (1965), p. 331; T. A. Rafter andB. J. O'Brien, Proc. 8th Int. Conf. RadiocarbonDating 1, 241 (1972).

66. U. Krtihenbuhl, Geochim. Cosmochim. Acta37, 1353 (1973).

67. B. Mason, Space Sci. Rev. 1, 621 (1962-1963).68. R. F. Butler, E. H. Lindsay, L. L. Jacobs, N.

M. Johnson, Nature (London) 267, 318 (1977);E. H. Lindsay, L. L. Jacobs, R. F. Butler, Geol-ogy 6, 425 (1978).

69. E. H. Lindsay, R. F. Butler, N. M. Johnson, inpreparation.

70. W. Alvarez and D. W. Vann, Geology 7, 66(1979); J. E. Fassett, ibid., p. 69; S. G. Lucasand J. K. Rigby, Jr., ibid., p. 323.

71. D. A. Russell, Episodes 1979 No. 4 (1979), p. 21.72. V. L. Masaytis, M. V. Mikhaylov, T. V. Seliva-

novskaya, Sov. Geol. No. 6 (1971), pp. 143-147;translated in Geol. Rev. 14, 327 (1972).

73. V. L. Masaytis, Sov. Geol. No. 11 (1975), pp.52-64; translated in Int. Geol. Rev. 18, 1249(1976).

74. It will be obvious to anyone reading this articlethat we have benefited enormously from conver-sations and correspondence with many friendsand colleagues throughout the scientific commu-nity. We would particularly like to acknowledgethe help we have received from E. Anders, J. R.Arnold, M. A. Arthur, A. Buffington, I. S. E.Carmichael, G. Curtis, P. Eberhard, S. Gartner,R. L. Garwin, R. A. F. Grieve, E. K. Hyde, W.Lowrie, C. McKee, M. C. Michel (who was re-sponsible for the mass spectrometric measure-ments), J. Neil, B. M. Oliver, C. Orth, B. Par-doe, I. Perlman, D. A. Russell, A. M. Sessler,and E. Shoemaker. One of us (W.A.) thanks theNational Science Foundation for support, theother three authors thank the Department of En-ergy for support, and one of us (L.W.A.) thanksthe National Aeronautics and Space Administra-tion for support. The x-ray fluorescence mea-surements of trace elements Fe and Ti by R. D.Giauque and of major elements by S. Flexserand M. Sturz were most appreciated. We appre-ciate the assistance of D. Jackson and C.Nguyen in the sample preparation procedures.We are grateful to T. Lim and the staff of theBerkeley Research Reactor for many neutronirradiations used in this work. We also ap-preciate the efforts of G. Pefley and the staffof the Livermore Pool Type Reactor for theirradiations used for the Ir isotopic ratiomeasurements.

No-Tillage AgricultureRonald E. Phillips, Robert L. Blevins, Grant W. Thomas

Wilbur W. Frye, Shirley H. Phillips

For over 100 years, agriculture has re-

lied upon the moldboard plow and diskharrow to prepare soil to produce food.Without the moldboard plow and disk itwould not have been possible to controlweeds and to obtain the yields necessary

to provide favorable economic returnsfrom agriculture. Weeds are strong com-petitors with food crops for water andplant nutrients, and it was not until plantgrowth regulators were introduced in thelate 1940's that attention was turned tono-tillage agriculture. From plant growthregulators selective herbicides were de-veloped, and these increased the feasibil-

R. E. Phillips and G. W. Thomas are professorsand R. L. Blevins and W. W. Frye are associate pro-fessors in the Department of Agronomy, College ofAgriculture, University of Kentucky, Lexington40546. S. H. Phillips is assistant director of the Co-operative Extension Service for Agriculture, Lex-ington 40546.

ity of growing many different crops with-out tilling the soil (1-3).

In this article, we define conventionaltillage as moldboard plowing followed bydisking one or more times. By this meth-od one obtains a loose, friable seedbed inthe surface 10 centimeters of soil. We de-fine the no-tillage system (4) as one inwhich the crop is planted either entirelywithout tillage or with just sufficient till-age to allow placement and coverage ofthe seed with soil to allow it to germinateand emerge. Usually no further cultiva-tion is done before harvesting. Weedsand other competing vegetation are con-

trolled by chemical herbicides. Soilamendments, such as lime and fertilizer,are applied to the soil surface.

In pasture management, chemicals are

used as a substitute for tillage, herbicidesbeing used to restrict growth and com-

0036-8075/80/0606-1108$01.50/0 Copyright K) 1980 AAAS

petition of undesirable plants during theestablishment of the newly seeded cropor to supress growth of grasses and allowestablishment of legumes. Row crop pro-duction with the no-tillage system is al-most always carried out by planting thecrop into soil covered by a chemicallykilled grass sod or with dead plant resi-dues of a previous crop. For example, incontinuous no-tillage corn (Zea mays)production, the soil surface at the timeof planting is covered with corn stalkresidues of the previous corn crop. Indouble-cropped soybeans, the soil at thetime of planting is covered with residuesof a recently harvested small-grain cropsuch as barley or wheat.The land area used for row crops and

forage crops grown by the no-tillage sys-tem has increased rapidly during the past15 years. In 1974, the U.S. Departmentof Agriculture (5) estimated that theamount of cropland in the United Statesunder no-tillage cultivation was 2.23 mil-lion hectares, and that 62 million hec-tares or 45 percent of the total U.S. crop-land (6) will be under the no-tillage sys-tem by 2000. An estimated 65 percent ofthe seven major annual crops (corn, soy-beans, sorghum, wheat, oats, barley,and rye) will be grown by the no-tillagesystem by the year 2000 and 78 percentby the year 2010 (5). In Kentucky therewere 44,000, 160,400, and 220,000 haof no-tillage corn and soybeans grownin 1969, 1972, and 1978, respectively.

SCIENCE, VOL. 208, 6 JUNE 19801108

on

Aug

ust 3

1, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om