nanodiamond rich layer across 3 continents

[The Journal of Geology, 2014, volume 122, p. 000000] 2014 by The University of Chicago.All rights reserved. 0022-1376/2014/12205-0001$15.00. DOI: 10.1086/677046

000

ARTICLES

Nanodiamond-Rich Layer across Three Continents Consistent withMajor Cosmic Impact at 12,800 Cal BP

Charles R. Kinzie,1,* Shane S. Que Hee,2 Adrienne Stich,1 Kevin A. Tague,1

Chris Mercer,3 Joshua J. Razink,4 Douglas J. Kennett,5 Paul S. DeCarli,6, Ted E. Bunch,7

James H. Wittke,7 Isabel Israde-Alcantara,8 James L. Bischoff,9 Albert C. Goodyear,10

Kenneth B. Tankersley,11 David R. Kimbel,12 Brendan J. Culleton,5 Jon M. Erlandson,13

Thomas W. Stafford,14 Johan B. Kloosterman,15 Andrew M. T. Moore,16

Richard B. Firestone,17 J. E. Aura Tortosa,18 J. F. Jorda Pardo,19

Allen West,20, James P. Kennett,21 and Wendy S. Wolbach1

A B S T R A C T

A major cosmic-impact event has been proposed at the onset of the Younger Dryas (YD) cooling episode at 12,800 150 years before present, forming the YD Boundary (YDB) layer, distributed over 150 million km2 on four continents.In 24 dated stratigraphic sections in 10 countries of the Northern Hemisphere, the YDB layer contains a clearlydefined abundance peak in nanodiamonds (NDs), a major cosmic-impact proxy. Observed ND polytypes include cubicdiamonds, lonsdaleite-like crystals, and diamond-like carbon nanoparticles, called n-diamond and i-carbon. The NDabundances in bulk YDB sediments ranged up to 500 ppb (mean: 200 ppb) and that in carbon spherules up to 3700ppb (mean: 750 ppb); 138 of 205 sediment samples (67%) contained no detectable NDs. Isotopic evidence indicatesthat YDB NDs were produced from terrestrial carbon, as with other impact diamonds, and were not derived fromthe impactor itself. The YDB layer is also marked by abundance peaks in other impact-related proxies, includingcosmic-impact spherules, carbon spherules (some containing NDs), iridium, osmium, platinum, charcoal, aciniformcarbon (soot), and high-temperature melt-glass. This contribution reviews the debate about the presence, abundance,and origin of the concentration peak in YDB NDs. We describe an updated protocol for the extraction and concentrationof NDs from sediment, carbon spherules, and ice, and we describe the basis for identification and classification ofYDB ND polytypes, using nine analytical approaches. The large body of evidence now obtained about YDB NDs isstrongly consistent with an origin by cosmic impact at 12,800 cal BP and is inconsistent with formation of YDBNDs by natural terrestrial processes, including wildfires, anthropogenesis, and/or influx of cosmic dust.

Online enhancements: appendixes.

Introduction

The Younger Dryas (YD) impact hypothesis pro-poses that a major cosmic-impact event occurredat the Younger Dryas Boundary (YDB) 10,900 145radiocarbon years before present (RCYBP), a timecorresponding to the onset of the YD cooling re-corded in Greenland Ice Sheet cores and other se-

Manuscript received May 19, 2013; accepted April 18, 2014;electronically published August 26, 2014.

* The authors affiliations can be found at the end of thearticle.

Deceased. Author for correspondence; e-mail: [email protected].

quences (Firestone et al. 2007). The publishedIntCal radiocarbon curve has recently been revised(Reimer et al. 2013) and provides a calibrated agefor this radiocarbon date of 12,830 130 cal BPat 1 standard deviation (j). This differs from earliercalibrated ages for the YDB of 12,900 100 cal BP,used by Firestone et al. (2007), and 12,800 150cal BP, more recently used by Wittke et al. (2013).Because this latest adjustment represents a differ-ence of only 30 yr, we continue to use an age of12,800 150 cal BP for the YDB. We emphasizethat, although the calendar calibration has

This content downloaded from 46.19.139.98 on Thu, 11 Sep 2014 09:26:01 AMAll use subject to JSTOR Terms and Conditions

000 C . R . K I N Z I E E T A L .

changed, the radiocarbon age has remained thesame.

The proposed impact deposited the YDB layer,which contains many cosmic-impact proxies, in-cluding magnetic and glassy impact spherules, irid-ium, fullerenes, carbon spherules, glass-like car-bon, charcoal, and aciniform carbon, a form of soot(Firestone et al. 2007; Wittke et al. 2013). In NorthAmerica and the Middle East, Bunch et al. (2012)identified YDB melt-glass that formed at high tem-peratures (1730 to 12200C), as also reported bythree independent groups, Mahaney et al. (2010) inSouth America and Fayek et al. (2012) and Wu etal. (2013) in North America. This study focusessolely on nanodiamonds (NDs), and so, for inde-pendent discussions of other proxies, see Haynes etal. (2010) and Paquay et al. (2009), who found noevidence for the platinum-group elements iridiumor osmium. Alternately, Wu et al. (2013) found largeYDB anomalies in osmium, as discussed below.Also, in a Greenland ice core, Petaev et al. (2013)found a large YDB abundance peak in the platinum-group element platinum. Surovell et al. (2009)found no YDB peaks in magnetic spherules,whereas LeCompte et al. (2012) found large, well-defined YDB spherule peaks at sites common tothe study by Surovell et al. Also, critical overviewsof the YDB hypothesis are presented in Pinter etal. (2011) and Boslough et al. (2012).

Recently, the YDB cosmic impact was indepen-dently confirmed by Petaev et al. (2013), who re-ported compelling evidence from a well-datedGreenland Ice Core Project (GISP2) ice core exhib-iting a sharp abundance peak in platinum preciselyat the YD onset (12,877 3.4 cal BP). Those au-thors mass-balance calculations indicate that theplatinum peak resulted from a major cosmic-impact event by an impactor estimated to be atleast 1 km in diameter. Similarly, Wittke et al.(2013) estimated that the tonnage of YDB ejecta(spherules and melt-glass) is comparable to thatejected from the 10.5-km-wide Bosumtwi Crater,likely produced by a 1-km-wide impactor. TheGISP2 platinum peak is coeval with the abrupt on-set (1.5 yr) of the atmospheric changes that markthe YD climatic episode in the North GreenlandIce Core Project (NGRIP) ice core at 12,896 cal BP(Steffensen et al. 2008). The discovery of such anunequivocal impact proxy at the YD onset in theGreenland record was predicted by the YDB impacthypothesis when it was initially introduced (Fire-stone et al. 2007).

The comprehensive impact proxy assemblage inthe YDB layer also includes NDs and diamond-like

carbon, which were discovered within carbonspherules, glass-like carbon, and bulk sediment.The polymorphs of carbon extracted from bulk sed-iment and carbon spherules include cubic NDs andhexagonal lonsdaleite-like crystals as well asunique carbon allotropes, called n-diamonds and i-carbon (details in table D1; apps. AD available on-line). These latter two types of nanocrystals, almostas hard as cubic NDs, are frequently used in thin,polycrystalline films for industrial applications re-quiring hardness and abrasion resistance (Wen etal. 2007). Ongoing investigations have been ex-amining whether these polymorphs are simply cu-bic diamonds with atomic substitution of carbonby hydrogen or other elements (Wen et al. 2011) orare new forms of diamond-like carbon (Hu et al.2012). Regardless, the nanoparticles in questionform under exotic temperatures and pressures notpresent naturally at the Earths surface or loweratmosphere but similar to conditions related to cos-mic impact (Wen et al. 2007) and are unlike otherforms of carbon typically found naturally on Earth.For simplicity, we refer to all forms as NDs, eventhough n-diamonds and i-carbon may actually beonly diamond-like. YDB NDs were most likelyformed from terrestrial carbon, based on their car-bon isotopic composition (Tian et al. 2011; Israde-Alcantara et al. 2012b), and are similar to NDsformed during the cosmic impact at the Creta-ceous-Paleogene boundary (K-Pg, formerly referredto as the K-T; Gilmour et al. 1992).

The YDB carbon spherules that contain NDs aremorphologically and compositionally similar toyounger carbon spherules first reported in near-sur-face forest soils of Europe by Rosler et al. (2005),who first suggested an impact-related origin of theparticles. Later, some of the same authors (Yang etal. 2008) stated, Whether this would have occurredduring or before any impact is still unclear for now(p. 943). Carbon spherules have been proven to formin cosmic-impact events, as shown by the discoveryof a !1100-yr-old meteorite crater in Alberta, Can-ada (Newman and Herd 2013). Some carbon spher-ules are fused to fragments of the meteorite, indi-cating that they formed upon impact. Theirmorphology includes an exterior shell around ahighly vesicular interior, identical to YDB spher-ules and carbon spherules found in Europe. Carbonspherules have also been reported from experi-ments using hypervelocity impacts into carbon-rich substrates, duplicating cosmic-impact condi-tions (Heymann et al. 2006). Furthermore, carbonspherules containing NDs have been demonstratedto form from tree sap under laboratory conditions


Journal of Geology N A N O D I A M O N D - R I C H L AY E R 000

that duplicate the temperature, pressure, and redoxvalues within an impact fireball (Israde-Alcantaraet al. 2012b).

Following the identification of NDs by Kennettet al. (2009a, 2009b), Daulton et al. (2010) at-tempted to replicate that discovery at two well-known archaeological sites, Murray Springs, Ari-zona, and Arlington Canyon, California. Daultonet al. (2010) found no YDB NDs and concluded thattheir findings cast doubt on the presence of YDBNDs, although they pointed out that YDB NDsmight occur inhomogeneously and only in someof the YD-boundary carbons and hence are not ob-served in our study (p. 16046). Daulton et al. (2010)also noted that other minerals, including nano-crystalline copper and copper oxide, could be mis-identified as several of the proposed diamond poly-types, because of crystallographic similaritiesbetween copper and diamond.

Later, an independent YDB study by Tian et al.(2011) confirmed the discovery of cubic YDB NDsat Lommel, Belgium, in the charcoal-rich YDBlayer in the upper part of a layer that is knownregionally as the Usselo Horizon. The intersectionbetween the Usselo layer and regional overlyingcover sands has been long recognized as represent-ing the onset of the YD climate change (Van Geelet al. 1989). At Lommel, cubic NDs were embeddedin carbon particles but with no other ND polytypes,and no NDs were observed above or below the YDBlayer. As with previous studies, the authors did notexamine bulk sediment for NDs. Tian et al. (2011)concluded that the NDs alone did not representindisputable evidence for a cosmic impact, but theydid not exclude one.

Israde-Alcantara et al. (2012b) used multiple an-alytical techniques to demonstrate that the YDBNDs from Lake Cuitzeo, Mexico, are cubic NDs,n-diamonds, i-carbon, and lonsdaleite-like crystals.Israde-Alcantara et al. (2012b) also identified sev-eral problems and limitations of the study by Daul-ton et al. (2010), who reported an absence of YDBNDs in carbon spherules at Murray Springs andArlington Canyon. First, Daulton et al. (2010)searched for and failed to find NDs within carbonspherules at Murray Springs, but neither Firestoneet al. (2007) nor Kennett et al. (2009a) reported find-ing carbon spherules at that site, making the relatedabsence of NDs unsurprising. Our investigationsshowed that carbon spherules are most common inregions having conifer trees at 12,800 cal BP, notin scrubby grasslands, as existed at Murray Springsat that time (Haynes and Huckell 2007). Second, atboth sites Daulton et al. (2010) searched for NDs

in charcoal, which has never been reported by anyworkers to contain NDs. Third, Daulton et al.(2010) did not examine bulk sediment, the onlysource of NDs at Murray Springs reported by Ken-nett et al. (2009a).

Kennett et al. (2009b) reported NDs in carbonspherules at Arlington Canyon, California; Daultonet al. (2010) found no NDs there either, but therewas a major flaw in their sample acquisition. Thesame coauthors of Daulton et al. (2010) claimed, inPinter et al. (2011, p. 254), to have acquired theirsamples from a location identical or closely prox-imal to the location examined by Kennett et al.(2009a). Contradicting that statement, Wittke et al.(2013) noted that the Universal Transverse Mer-cator coordinates of their sampling sites show con-clusively that their purported continuous sequencewas actually collected as four separate discontin-uous sections, separated by up to 7000 m horizon-tally from the sampling location of Kennett et al.(2009a, 2009b). Therefore, Scott et al. (2010) did notsample the YDB at the location studied by Kennettet al. (2009a) and did not acquire a dated, contin-uous profile across the YDB at any ArlingtonCanyon location. These mislocated sediment sam-ples collected by Scott et al. (2010) were subse-quently used in several different studies by thesame group of authors (Daulton et al. 2010; Scottet al. 2010; Pinter et al. 2011). Their incorrect strat-igraphic locations apply to all those investigations,explaining their inability to detect YDB NDs, cos-mic-impact spherules, and ND-rich carbon spher-ules at Arlington Canyon.

Daulton (2012) also questioned the identificationof lonsdaleite (hexagonal diamond), suggesting thatsome particles exhibited in Kennett et al. (2009b)appear to be graphene-graphane aggregates. VanHoesel et al. (2012), Madden et al. (2012), and Be-ment et al. (2014) also reported finding graphene-graphane clusters with diffraction patterns similarto those of lonsdaleite. Boslough et al. (2012) sug-gested that some of the reported lonsdaleite fromLake Cuitzeo might instead be other minerals. Wediscuss these points below in Identification ofLonsdaleite-Like Crystals.

Daulton (2012) and Boslough et al. (2012) ques-tioned whether YDB NDs are robust cosmic-impactmarkers. However, cubic NDs are widely acceptedto have formed during the K-Pg impact event andwere not found in sediment before or after the event(Carlisle and Braman 1991; Gilmour et al. 1992;Hough et al. 1997, 1999). Those NDs are found atsix coeval sites across North America: two in Colo-rado and one each in Mexico, Montana, and Al-


000 C . R . K I N Z I E E T A L .

berta, Canada. The K-Pg NDs were reported torange in size from 1 nm to 30 mm, whereas YDBNDs are smaller, spanning a narrower range of 1to 2.9 mm, perhaps because that older impact waslarger and more energetic than the YDB event.

Van Hoesel et al. (2012) observed cubic NDswithin particles of glass-like carbon at the Geldrop-Aalsterhut site in the Netherlands. The NDs werefound in a few-centimeter-thick, charcoal-rich in-terval at the upper boundary of the Usselo layer,the top of which is widely accepted as representingthe onset of the YD cooling episode (Van Geel etal. 1989). They reported NDs only in glass-like car-bon in the bottom 1 cm of that interval and did notexamine bulk sediment for the presence of NDs.

Recently, Bement et al. (2014) discovered anabundance peak in YDB n-diamonds (190 ppm) atBull Creek, Oklahoma, independently confirmingthe discovery there of YDB NDs (100 ppb) by Ken-nett et al. (2009a). They did not observe cubic NDs,as Kennett et al. (2009a) did, and neither group ob-served lonsdaleite at Bull Creek. In addition, Be-ment et al. (2014) observed an ND abundance peakof similar amplitude to their YDB peak in two con-tiguous samples of late Holocene surface sediments(010 and 1020 cm below surface). They suggestedthat this younger ND peak may have been producedby a nearby cosmic-impact event within the pastseveral thousand years. This discovery may corre-late with that of Courty et al. (2008), who discov-ered melt-glass and spherules at widely distributedsites in Syria, Spain, and Peru, localities separatedby up to 13,000 km, as evidence for a 4000-yr-oldNorthern Hemispheric impact event. Bement et al.(2014) concluded from sedimentological evidencethat the peak ND accumulations in the YDB andyounger strata did not result from changes in cli-mate, deposition rates, lag deposits, or human siteusage. Their results refute the hypothesis that theNDs simply resulted from cosmic influx that de-posited them as a lag deposit at the YDB over anextended interval of time (Haynes et al. 2010; Pin-ter et al. 2011; Boslough et al. 2012). Instead, Be-ment et al. (2014) concluded the evidence is con-sistent only with cosmic-impact events.

In summary, abundant NDs within or near theYDB layer have been reported by four independentgroups (Redmond and Tankersley 2011; Tian et al.2011; van Hoesel et al. 2012; Bement et al. 2014).In addition, NDs have been reported independentlyin three conference presentations (at Indian Creek,MT, by Baker et al. 2008; at Newtonville, NJ, byDemitroff et al. 2009; and at Bull Creek, OK, byMadden et al. 2012). These investigations indepen-dently confirm the presence of an ND abundance

peak in the YDB layer, which has also been shownto be associated with a diversity of other cosmic-impact proxies. Research continues into the spe-cific origin of the various YDB ND polytypes andthe presence of lonsdaleite.

Material and Methods

We now present a comprehensive summary of thechemical processing methods that we used to ex-tract and isolate NDs from terrestrial bulk sedi-ments and glacial ice. This is followed by detailsof the characterization, identification, and inter-pretation of YDB NDs. The protocol here super-sedes previous published versions for extractingYDB NDs (Kennett et al. 2009a, 2009b; Kurbatovet al. 2010; Israde-Alcantara et al. 2012b). Furtherdetails are in appendix A.

Our protocol was adapted by one of us (S. S. QueHee) from the extraction procedure developed byHuss and Lewis (1995), who used it to isolate pre-solar NDs from meteorites. We found that the max-imum yield of all types of NDs occurred after theammonium hydroxide extraction step and that sub-sequent oxidation with perchloric acid destroyedmany crystals of n-diamonds and i-carbon and, pos-sibly, some of the lonsdaleite-like crystals. Thiswas an advantage when analyzing cubic NDs buta major disadvantage for the other allotropes, whichwere no longer present. Although the extractionprocess remains difficult, exacting, and labor-intensive, we have successfully extracted NDsfrom hundreds of samples in or adjacent to the YDBlayer on three continents and in the Greenland IceSheet, along with samples from the K-Pg impact,Sudbury Crater, and the Tunguska airburst. Six in-dependent groups have successfully used this pro-tocol or a version of it (Baker et al. 2008; Demitroffet al. 2009; Redmond and Tankersley 2011; Tian etal. 2011; van Hoesel et al. 2012; Bement et al. 2014).

Preparing Sediment and Ice. From each sedimen-tary sample collected, 5001000 g of thoroughlymixed dry bulk sediment was processed through aclean, ultrasonicated !38-mm screen to concentratethe fine, ND-bearing fraction. A minimum of 20150 g of the !38-mm fraction was used for extractionof NDs, whereas the 138-mm-size fraction was notused. Water was removed from bulk ice core sam-ples by freeze-drying or by melting and evaporationbefore chemical processing. This method extractsunattached NDs within the sediment as well as anyNDs that may be contained inside any melt-glassor small mineral aggregates. Great care was takento eliminate contamination by industrial cubicNDs, and any such contamination is highly un-



likely, as indicated by the fact that NDs alwayspeak in the same layer that contains other markers.Peaks of similar magnitude have never been foundoutside the YDB layers.

Materials and Equipment. All solutions were pre-pared with corresponding TraceMetal or Electron-ics-grade chemicals and concentrated acids or ba-ses. A detailed list of standard instrumentation andequipment appears in appendix A.

Extraction and Purification. As many operationsas possible were performed at room temperature,because temperatures above 200C can cause non-cubic types of NDs to gradually convert to otherforms of carbon, such as graphite. Also, adequaterinsing and centrifugation are crucial for successfulpurification of the ND-rich residues.

The YDB NDs contain surface carboxyl groups(COOH) formed in situ either during ND forma-tion or during diagenesis while buried for 12,800yr. These carboxyl groups are a key part of the ex-traction process, because they allow the NDs to gointo suspension in basic solution, thus separatingfrom the non-ND minerals. However, these car-boxyl groups are also subject to decarboxylation un-der strongly acidic conditions and/or at elevatedtemperatures. Therefore, it is vitally important toextract any NDs into room-temperature basic so-lution (pH 1 7) while they still contain the maxi-mum density of carboxyl groups on their surfaces.After the samples were pulverized and massed,therefore, the first chemical step was extraction ofNDs, using room-temperature 0.1 M NaOH. OnceNDs were separated from the remaining sediment,they were consolidated in solution acidified to apH of !2 with 9 M HCl.

Next, acidic dichromate oxidation (K2Cr2O7 andH2SO4) was used to remove the remaining intrac-table organic components that might adhere toNDs. Following dichromate oxidation, sampleswere diluted with deionized water to lower solu-tion density and were centrifuged. Supernatantswere discarded, and residues were rinsed repeatedlywith 0.1 M HCl. Some YDB residues were visible,but most were detectable only by light microscope.For non-YDB samples, often there were no residuesvisible with a light microscope.

At this point, most non-ND minerals were eitherleft behind during basic extraction or oxidized bythe acidic dichromate. Any remaining silicateswere digested with 10 M HF/1 M HCl, and afterrinsing, samples were treated with 9 M HCl to de-stroy fluorides. If necessary, the acidic dichromateand hydrofluoric acid steps were repeated. Finally,the samples were dried and weighed. The typicalresult was a very small amount of whitish-gray res-

idue that contained amorphous carbon and, if pres-ent, an assemblage of several types of NDs (cubic,n-diamonds, i-carbon, and lonsdaleite-like crys-tals). This was the last step performed if we choseto examine all types of NDs.

If we chose to investigate only the cubic NDs,an additional step was added to destroy the n-dia-monds, i-carbon, and possibly lonsdaleite, thusmaking it easier to identify the cubic NDs. To ac-complish this, we added concentrated perchloricacid (HClO4, 70%), heated the mixture, and thenallowed the perchloric acid to evaporate to dryness.After that, samples were rinsed several times with0.1 M HCl and centrifuged. Once dried, the ex-tracted ND residues were ready for further analysis.The acid extraction process commonly yielded verylittle residue that was nearly invisible to the nakedeye inside the centrifuge tubes and often was de-tectable only by light microscope. For non-YDBsamples, there were typically no residues visibleeven with a light microscope.

Early Developmental Studies. To test this protocolat the HF step, we used synthetic cubic NDs fromPlasmaChem (PL-D-G-1g; avg. cluster size 4 nm;NDs usually 2 nm) at a concentration correspond-ing to 1000 ppm in 10 g of sediment. When thenitric acid was substituted for HCl, the HF/HNO3digestion step allowed recovery of 70%80% byweight. The soluble phase contained 20%30% ofNDs by weight, and HCl acid-washing conditionsresulted in quantitative recoveries in the solid res-idue at each such step. The recovery of spiked cu-bics up to the perchloric acid step varied between70% and 80% in these preliminary experiments.The major step responsible for the variation wasthe flocculation step.

To determine the efficacy of the perchloric acidextraction step, we conducted several experiments.When 10 mg of synthetic cubic NDs was subjectedto the perchloric acid processing step, 81% 5%was recovered by weight as unaltered cubic NDs.Transmission electron microscopy (TEM) con-firmed the presence or absence of NDs in all theseexperiments, which confirmed that extraction ofNDs with this protocol has a high success rate.

NDs from Carbon Spherules and AmorphousCarbon. The NDs were not extracted from carbonspherules; instead, the spherules were crushed tofine fragments and placed on a TEM grid, as de-scribed in detail in appendix A. Tian et al. (2011)and van Hoesel et al. (2012) reported YDB NDs inflakes of amorphous carbon or glass-like carbon,and our general protocol for carbon spherules alsoapplies to NDs within these other forms of carbon.

Preparation of TEM Grids for Analyzing NDs. Cur-


000 C . R . K I N Z I E E T A L .

rently, we use 200- to 400-mesh TEM grids of goldor molybdenum with ultrathin carbon film (3 nmthick) over holey carbon. It is important to notethat because ultrathin films are approximately asthick as the NDs and other nanoparticles, theyhave the least effect on energy dispersive X-rayspectrometry (EDS) and other measurements, mak-ing their use essential for obtaining the best anal-yses. We avoided films that have large open holes(holey carbon) because the NDs are much smallerthan the holes.

Some earlier ND work (Kennett et al. 2009a,2009b; Kurbatov et al. 2010) was conducted on cop-per grids, which were discontinued because of thesimilarity between the spacings of crystallographicplanes (d-spacings) of copper and some NDs (Daul-ton et al. 2010). Those early samples on copper gridswere subsequently reanalyzed on gold or molyb-denum grids and with additional analytical tech-niques, such as EDS and energy-filtered TEM(EFTEM) that can differentiate carbon from copperparticles. The results confirmed that the use of cop-per grids, although suboptimal, did not lead to themisidentification of YDB NDs. In addition, al-though silicon films are preferable to carbon filmsfor investigating carbon objects, we have discon-tinued their use because they are less stable undera high-voltage electron beam.

To prepare a TEM grid, we first placed NDs intosuspension by pipetting just 1 or 2 drops of am-monium hydroxide (NH4OH), ethyl alcohol, ordeionized water into the vial and stirring the ND-rich mixture. Then we pipetted the drop onto thegrid and dried it. For further details of preparinggrids, see appendix A.

Experimental Methods: Electron Microscopy andSpectroscopy. Sample residues were examinedwith high-resolution TEM (HRTEM), scanningTEM (STEM), electron energy-loss spectroscopy(EELS), selected-area electron diffraction (SAD), andEDS. To accomplish that, we used an FEI 300-kVfield emission gun Titan equipped with a SuperTwin objective lens, a spherical-aberration-imagecorrector, an EDAX energy-dispersive spectrome-ter, a high-angle annular dark field (HAADF) de-tector, and a Tridiem Gatan imaging filter. For allof the experiments, the instrument was operated at300 kV. The TEM detectors were calibrated for ac-curacy with commercial cubic diamond and goldstandards (Ted Pella #646). Fast Fourier Transform(FFT) analyses of the HRTEM images and the anal-yses of the EELS data were performed with GatanDigital Micrograph. Occasionally, JEOL 1200EX IIand JEOL TEM 1210 transmission electron micro-scopes were operated at 80 kV to acquire images

with higher contrast and to investigate nanoparti-cles that vaporized at temperatures generated bythe higher voltages of the Titan.

Difficulties in Identifying NDs. The extractionprocess detailed in this contribution yields a resi-due that contains amorphous carbon, resistant min-erals, and NDs, when present, but there are tech-nical difficulties in fully characterizing thismaterial. For example, the protocol does not re-move minor amounts of non-ND crystals, includ-ing quartz, rutile, and zircon. Typically, NDs rep-resent !50% of the residue, and the remainingnon-ND residue can mask the NDs, thus makingthem difficult to identify. In addition, there are in-herent difficulties and uncertainties in correctlyidentifying tiny crystals !2 nm in diameter. Fur-thermore, multiple ND polytypes are often inter-mixed, making differentiation of individual poly-types difficult.

Quantification of NDs. Accurate quantification ofND abundances in a sample is difficult, becausethe volume of NDs present is typically minusculeand the NDs are difficult to isolate from amorphouscarbon. We addressed this problem by adaptingmethods used by various researchers to semiquan-tify abundances of other kinds of particles, such asaciniform carbon (a form of soot), charcoal, fora-miniferal species, and various plant microfossils.We developed an 11-point scale (0% to 100%) forestimating the abundances of NDs at parts-per-bil-lion to parts-per-million levels in both the ex-tracted residues and carbon spherules (see Quan-tification of NDs in app. A; fig. A1; app. C). Theabundance values presented here supersede previ-ously published values.

Results and Discussion

Regional Setting. The YDB ejecta field containsa variable assemblage of cosmic-impact markers,including NDs, cosmic-impact spherules, carbonspherules, and high-temperature melt-glass. Thefield spans an area of 50 million km2 across fourcontinents, with no known limits (Wittke et al.2013). For this study, we investigated YDB NDs at22 sites in 10 countries on three continents, andindependent researchers conducted six studies, fora total of 24 sites (fig. 1). Thirteen sites are in theUnited States, two in Canada, two in the Nether-lands, and one each in the Greenland Ice Sheet(Denmark), Belgium, Germany, Mexico, Spain, Sy-ria, and the United Kingdom. These 24 sites occuracross a remarkably diverse range of geologic set-tings, including polar ice, glacial till, mountainlakes, caves, coastal canyons, desert dry washes,



and alluvial terraces, with altitudes varying fromnear sea level to 11800 m. This wide diversity in-dicates that geologic setting has no effect on thepresence of YDB NDs, as also concluded by Bementet al. (2014). Details on site setting, geological in-formation, and dating are in tables D2, D3. Moredetails for most sites are in Bunch et al. (2012) andWittke et al. (2013).

Dating and Age-Depth Models. We present new ra-diocarbon dates for three sites (Lake Cuitzeo, Mex-ico; Lingen, Germany; and Santa Maira, Spain), andwe have generated new age-depth models for Ar-lington Canyon (fig. B1) and Lake Cuitzeo (figs. B2,B3). We present ND data for 15 new sites, for whichsite details, stratigraphic information, and datingare provided in appendix B and tables D2, D3. Ofthe 24 sites investigated, 18 (75%) have either di-rect dates or age-depth models at sufficient chro-nological resolution to confirm correlation withthe YDB. Three others have been indirectly datedvia lithologic and isotopic stratigraphy, archaeolog-ical context, and age-depth modeling, and the re-maining three sites lack dates directly from theYDB but have consistent extrapolated ages basedon dated materials from near the boundary layer.

The dates for a few of these sites have been chal-lenged. For example, Blaauw et al. (2012) ques-tioned the age-depth model for the Lake Cuitzeosediment core in Israde-Alcantara et al. (2012b) andproposed that the YDB layer is up to 2000 yr olderthan the modeled age. Israde-Alcantara et al.(2012a) countered that the modeled age is the onlyone consistent with palynological and climatolog-ical records from this sequence and several siteslocated in Central and South America. To furthertest the age model, we acquired a new acceleratormass spectrometry 14C date (NOSAMS-71325:10,550 35 RCYBP, 12,897 187 cal BP) on or-ganic sedimentary carbon collected above the YDBlayer in a nearby exposed shoreline sediment se-quence, lithologically correlated with the lake core.This helps constrain the age of the ND-rich layerand demonstrates that the model previously pub-lished in Israde-Alcantara et al. (2012b) is correct(table D3; figs. B2, B3).

Boslough et al. (2012) questioned the age deter-mination for the YDB at the Gainey, Michigan, site,on the basis of a modern date acquired for a YDBcarbon spherule. That date replicated a moderndate previously reported for the same stratum ofthe Gainey sequence (Firestone 2009). Also, Ivesand Froese (2013) questioned the inferred YDB agefor carbon spherules at the Chobot site in Canada,also on the basis of a young radiocarbon date fromFirestone (2009). Nevertheless, the archaeological

context argues against a modern age for these near-surface layers. At both the Gainey and Chobotsites, the inferred YDB layers contain glassy andmagnetic cosmic-impact spherules, as well as car-bon spherules, each filled with millions of NDs (seeTEM, SAD, and Scanning Electron Microscopy ofNDs in Carbon Spherules below). This impact evi-dence is associated with large numbers of tempo-rally diagnostic, Clovis-era artifacts that are foundnear the surface but date within a range of 13,25012,800 cal BP (Waters and Stafford 2007). Further-more, the span of an OSL date (12,360 1230 calBP) for the same Gainey layer includes the onsetof the YD at 12,800 cal BP and is not modern inage. Thus, on the basis of available evidence, theseyoung radiocarbon dates do not accurately reflectthe age of the inferred YDB layers at these sites.

Firestone (2009) presented several possibilities toexplain these age discrepancies, and the most likelyis the effects of bioturbation. At some sites, weobserved distinctive root casts, formed from largetaproots of trees that penetrated the YDB layer afterthe impact event occurred. After those roots de-cayed or burned, the resulting cavity filled withsediment containing younger charcoal and carbonspherules that mixed with the older carbon mate-rial. Because this is a common occurrence wherethe YDB is shallow, radiocarbon dating is unreliablefor such sites and OSL dating is preferred (Bunchet al. 2012), as indicated by the older OSL date forthe Gainey site. No matter the cause, the ages ofthese two sites remain poorly constrained. Nev-ertheless, 18 of the 24 sites with the same YDBmarkers are well dated, suggesting that the YDBlayer is correctly identified at Gainey and Chobot(table D3).

Later, van Hoesel et al. (2012, p. 7652) suggestedthat NDs in the Netherlands at the Aalsterhut siteare two centuries younger than the diamonds re-ported by Kennett et al. (2009b) and therefore arefrom an unrelated event. They concluded this onthe basis of an apparent age discrepancy betweenthe mean age of the ND-rich layer at their site andmean age of the Arlington Canyon site in Califor-nia. However, they overlooked the fact that thedate for the Aalsterhut site fully overlaps those formany other YDB sites, including Murray Springs.To test their hypothesis, we performed Bayesiananalysis (Bronk Ramsey 2009) and x2 testing (Wardand Wilson 1978) on the Arlington Canyon radio-carbon dates. Both methods indicate that the Ar-lington Canyon radiocarbon dates have nonnormaldistribution and thus are unsuitable for averaging(fig. B1). Bayesian analysis is particularly useful indetecting outlier dates (nonnormal distribution),


000 C . R . K I N Z I E E T A L .

Figure 1. Map showing 24 sites containing Younger Dryas Boundary (YDB) nanodiamonds. The solid line definesthe current known limits of the YDB field of cosmic-impact proxies, spanning 50 million km2 (Wittke et al. 2013),including the study of Mahaney et al. (2010) in Venezuela (open circle). Numbered sites are from this study: (1) LakeCuitzeo, Mexico (Israde-Alcantara et al. 2012b); (2) Daisy Cave, California; (3) Arlington Canyon, California (Kennettet al. 2009b); (4) Murray Springs, Arizona (Kennett et al. 2009a); (5) Lindenmeier, Colorado; (6) Bull Creek, Oklahoma(Kennett et al. 2009a); (7) Blackville, South Carolina; (8) Topper, South Carolina (Kennett et al. 2009a); (9) KimbelBay, North Carolina; (10) Newtonville, New Jersey; (11) Melrose, Pennsylvania; (12) Sheriden Cave, Ohio; (13) Gainey,Michigan (Kennett et al. 2009a); (14) Chobot site, Alberta, Canada (Kennett et al. 2009a); (15) Lake Hind, Manitoba,Canada (Kennett et al. 2009a); (16) Kangerlussuaq, Greenland (Kurbatov et al. 2010); (17) Watcombe Bottom, Isle ofWight, United Kingdom; (18) Lommel, Belgium; (19) Ommen, Belgium; (20) Lingen, Germany; (21) Santa Maira, Spain;(22) Abu Hureyra, Syria. In addition, independent researchers have reported NDs at six sites, indicated by letters,four of which are in common: (a) Indian Creek, Montana (Baker et al. 2008); (b) Bull Creek, Oklahoma (Madden etal. 2012; Bement et al. 2014); (c) Sheriden Cave, Ohio (Redmond and Tankersley 2011); (d) Newtonville, New Jersey(Demitroff et al. 2009); (e) Lommel, Belgium (Tian et al. 2011); (f) Aalsterhut, Netherlands (van Hoesel et al. 2012).A color version of this figure is available online.

including those that result from the old-wood ef-fect, in which the date for charcoal or wood froma long-lived tree can lead to the erroneous conclu-sion that the stratum in which the charcoal wasfound is much older. Bayesian analysis rejected 14of 16 Arlington Canyon dates as being outliers, con-sistent with the observation that local tree specieshave life spans of up to 1300 yr (see ArlingtonCanyon, California in app. B). After adjusting forthe old-wood effect, OxCal modeled the YDB agefor Arlington Canyon as 12,748 46 cal BP (OxCal,ver. 4.2.3, IntCal-13; Bronk Ramsey 2009). This isstatistically identical to the modeled YDB date forAalsterhut of 12,746 12 cal BP (10,870 15RCYBP; van Hoesel et al. 2012). These results con-tradict the hypothesis that the ND-rich layer atAalsterhut is 200 yr younger than the ND-rich YDBlayer at Arlington Canyon. Furthermore, van

Hoesel et al. based their 200-yr difference on themean ages of the two sites, but the standard de-viation must be considered, and using only meanages is inappropriate. We conclude that van Hoeselet al. (2012) discovered the YDB layer at Aalsterhut.Bayesian analysis shows that the ages of all 18 well-dated YDB sites fall within 1 standard deviation ofthe YDB layer at 12,800 150 cal BP, includingAalsterhut and Arlington Canyon (table D3). Noneof those 18 sites is 200 yr older than Aalsterhut.

Van Hoesel et al. (2014) also questioned whetherthe YDB proxies are synchronous with the onset ofthe YD climatic episode, which is widely acceptedto have occurred abruptly. For example, in theNGRIP ice core, Steffensen et al. (2008) found thatthe YD onset occurred within a span of 1.5 yr, andBrauer et al. (2008) reported a similar narrow spanof 1 yr in varved European lake records. The



Greenland Ice Core Chronology 2005 (GICC05;Rasmussen et al. 2006) found the mean age of theYD onset to be 12,896 138 before 2000 AD (b2k;13,03412,758), and the GISP2 ice model (Meese etal. 1997) placed it at 12,892 260 b2k (13,15212,632). The YDB ages for 20 of 24 sites (83%) intable D3 fall within the ice core age ranges for theonset of the YD climatic episode, suggesting a closerelationship between the YDB and the YD onset.Most importantly, a major peak in impact-relatedplatinum in the GISP2 ice core occurred preciselyat the onset of the YD climatic episode (Petaev etal. 2013), strongly indicating that the YDB cosmic-impact event and the onset of the YD episode aresynchronous.

The YDB hypothesis posits that only one impactoccurred, producing coeval, above-backgroundpeaks in NDs, iridium (Firestone et al. 2007), plat-inum (Petaev et al. 2013), osmium (Wu et al. 2013),high-temperature melt-glass (1730 to 12200C;Bunch et al. 2012), and high-temperature magneticspherules (11500C; Wittke et al. 2013). Othershave proposed various age models for deposition ofthe YDB proxies. The first counterexplanation isthat YDB proxies resulted from multiple, unrelated,natural mechanisms that coincidentally occurrednear 12,800 cal BP (Pinter et al. 2011; Boslough etal. 2012; van Hoesel et al. 2014). To investigate that,our group and others have measured marker abun-dances in several stratigraphic profiles that span asmuch as the past 30,000 yr. These proxies reachedmaximum abundances only in the YDB layer andare not known to peak individually or collectivelyanywhere else in that span, making the YDB highlyunusual. In the second scenario, the YDB proxieswere deposited over several centuries, resultingfrom multiple discrete cosmic-impact events.However, current understanding of impact dynam-ics cannot explain how such a scenario could occurover such a span. In the third scenario, the YDBproxies were deposited during a span of up to sev-eral decades. Such a situation could occur if thedebris field of a fragmented comet or asteroid wasoblique to or wider than Earths diameter upon im-pact. In such a case, some objects would have en-countered Earth at an oblique angle and could haveassumed orbits that decayed over a few years todecades, producing multiple smaller impacts (Faw-cett and Boslough 2002; Petaev et al. 2013). Thefourth and most plausible scenario, the one mostconsistent with our data, is that only one hemi-spheric impact event occurred. This is supportedby the platinum record in GISP2, which forms asingle, brief, coherent abundance peak, the only onewithin the 280-yr interval investigated.

Abundance and Stratigraphic Distribution of NDs.Crystal morphologies vary from angular to roundedand from monocrystalline to twinned, and ND di-ameters average 34 nm (range: 1 nm2.9 mm), withmost measuring between 1 and 20 nm, a typicalsize for detonation-formed NDs (Wen et al. 2007).The quantification method discussed above wasused to estimate abundances of NDs and revealedconcentrations in carbon spherules of 103680 ppb(mean: 755 ppb) and in bulk sediment of 11494ppb (mean: 200 ppb). For carbon spherules, 111 of153 samples investigated (73%) contained no de-tectable NDs, and for sediment, 57 of 87 (66%) sam-ples had no evident NDs, comparable to the nullresults of Bement et al. (2014). Appendix A dis-cusses quantification, figure 2 shows abundancepeaks in NDs, and table D4 lists the stratigraphicabundances of NDs for 22 sites. Table D4 also listsabundances of cosmic-impact spherules and melt-glass for 16 sites; the other six sites have not yetbeen examined for those proxies. All sites exhibitsharp ND abundance peaks at the YDB, with veryfew NDs in the strata above and below (fig. 2). Forsome sites, the peaks are broader, with elevatedabundances of NDs in several contiguous samples.These secondary peaks most likely result from bio-turbation and wind-and-water action that redistrib-uted the NDs upward and/or downward.

Identification and Taxonomy of YDB NDs

Overview. Unknown nanoparticles were inves-tigated with multiple analytical techniques; ananoparticle was conclusively identified as an NDif several basic properties were documented: first,the nanoparticle is composed only of carbon; sec-ond, it has a crystalline structure; and finally, thed-spacings match those of an ND polytype. Al-though it is unnecessary to use all of the analyticalprocedures described below for every particle, at aminimum, we used EDS to determine elementalcomposition and HRTEM, SAD, and/or FFT to de-termine crystalline structure for all nanoparticlesimaged in this contribution. These analyses werestrengthened by use of EELS and EFTEM to assistwith determining elemental compositions and toinvestigate the interatomic bonding typical of car-bon (sp2 and/or sp3).

Electron Microscopy (TEM, HRTEM, and STEM).The STEM mode (HAADF, dark field) was typicallyused to investigate candidate nanoparticles, but thequality of the images was often degraded by scat-tering of electrons by the amorphous residue andby contamination of the grid by vaporized carbon.The TEM and HRTEM modes (bright field) typi-


Figure 2. Abundances of nanodiamonds (NDs; ppb) for 22 Younger Dryas Boundary (YDB) stratigraphic sectionsplotted by depth (cm below surface). Most of the six independent studies did not quantify NDs at or near the YDBand are not represented here. Horizontal bands represent thicknesses of samples containing YDB proxies. Solid linesrepresent ND abundances (ppb), shown on the X-axis. ND abundances were estimated with an 11-point semiquan-tification scale of relative values ranging from 0% to 100% (see Quantification of NDs). CS p NDs extracted from



carbon spherules; SED p NDs from bulk sediment; surface indicates ground surface for eight sites; no NDs wereobserved in these surface layers. Abundances for NDs, carbon spherules, and cosmic-impact spherules are listed intable D4, available online. Data are from Kennett et al. (2009a, 2009b), Kurbatov et al. (2010), and Israde-Alcantaraet al. (2012b). A color version of this figure is available online.

cally produced clearer images of NDs located inresidue, thus strengthening analyses of crystallin-ity and lattice spacings.

Figure 3 illustrates the typical progression nec-essary to identify NDs. First, extracted materialfrom Murray Springs and Lake Cuitzeo producedSTEM images that show thousands of rounded-to-subrounded, nanosized particles (fig. 3A and 3B, re-spectively). At this point, it was unknown whetherthey were amorphous or crystalline. Next, anHRTEM image (fig. 3C) revealed the crystallinestructure of NDs (ND), among a background ofamorphous carbon and strand-like carbon ribbons(CR) that are commonly present in the diamond-rich residue. Finally, an HRTEM image of a carbonnanocrystal (fig. 3D) displayed lattice spacings of2.06 A, consistent with (111) planes for cubic dia-mond (table D1), when viewed along the [110] zoneaxis. Additional testing typically was performed onthese nanocrystals, as discussed below.

SAD and FFT of HRTEM. The SAD image fromMurray Springs (fig. 4A) displays a ring pattern ofcollective d-spacings from multiple crystals, and inthis case, all eight visible reflections match thoseof cubic NDs. Values for graphene and graphane aresimilar to those for six of those reflections, but the(400) and (551) reflections are not present in thoseother crystals (table D1). Their absence in SAD pat-terns for graphene and graphane makes those min-erals easily detectable, thus eliminating the pos-sibility of misidentification. By themselves, SADpatterns are insufficient to identify NDs, and sofurther investigations, such as those using HRTEM,FFT, EDS, and EELS, were performed on these nano-particles to confirm that they are NDs and notsome other mineral. The FFT of an HRTEM imageof multiple nanocrystals (fig. 4B) shows three ringreflections that match the SAD pattern for cubicNDs. The FFT of an HRTEM image of a singlenanocrystal (fig. 4C) displays (111)- and (220)-typespot reflections consistent with a single cubic di-amond viewed along the [110] zone axis. Figures 5and 6 show TEM, HRTEM, and SAD patterns forNDs found in bulk sediment and carbon spherulesfrom sites on three continents. Results from 10other sites are shown in figures C1C6.

EDS. A STEM image from Lake Cuitzeo (fig. 7A)shows a nanoparticle field with the specific area of

investigation boxed near the center. The SAD pat-tern of that boxed area (fig. 7B) exhibits diffractionrings characteristic of i-carbon. The EDS analysisof the boxed area (fig. 7C) indicated a carbon con-centration of 198% as well as low amounts of ox-ygen and a weak signal from the gold grid, but noother elements. Because this analysis encompassesnanocrystals, the grid film (3-nm thick), and sur-rounding amorphous carbon, the elemental per-centages for the nanoparticles are inexact but aredominantly carbon. When other mineral grainswere encountered, e.g., quartz, rutile, and zircon,they were easily identifiable with EDS. AnotherSTEM image (fig. 7D) shows a cluster of angularsynthetic cubic NDs (97% pure 4-nm clusters fromPlasmaChem) with no observable amorphous car-bon; the box indicates the region being analyzed.An SAD pattern (fig. 7E) exhibits diffraction rings,indicating that the nanoparticles are cubic NDs,with no diffraction rings of other minerals. TheEDS spectrum of the commercial diamonds in fig-ure 7F shows 197% carbon, which closely matchesthe EDS of YDB NDs in figure 7C, with a similarlyhigh abundance of carbon.

EELS. The EELS analyses were performed to de-termine whether selected nanoparticles are carbonand whether they display the correct atomic bond-ing for diamond. This is diagnostic for distinguish-ing cubic NDs and lonsdaleite-like crystals fromother forms of carbon. The EELS technique is lessuseful for differentiating n-diamonds, i-carbon,graphite, and other carbon allotropes from eachother, since these produce similar spectra. To dif-ferentiate the various carbon polytypes, it is alsonecessary to acquire SAD patterns and FFTs ofHRTEM images.

The Murray Springs EELS spectrum, known as acore-loss spectrum (fig. 8A), displays the typicalshape for cubic diamond (Peng et al. 2001). The

edge is well above background, indicating thatjthe nanoparticles are carbon. The small edge isprepresentative of lower-order sp2 bonding, as foundin graphite, graphene, graphane, and amorphouscarbon. In this case, the peak most likely representsthe carbon grid film and amorphous carbon inwhich the NDs are embedded. The two peaks at300 and 310 eV with a trough between them rep-resent the characteristic signature of cubic dia-


000 C . R . K I N Z I E E T A L .

Figure 3. Three techniques for identifying candidate nanodiamonds (NDs). A, B, Scanning transmission electronmicroscopy images of clusters of NDs from Murray Springs, Arizona (Younger Dryas Boundary [YDB]: 426 ppb ofNDs at 46.5 cm below surface [cmbs]; A), and Lake Cuitzeo, Mexico (YDB: 493 ppb at 280 cmbs; B). C, Bright-fieldhigh-resolution transmission electron microscopy (HRTEM) of ND-rich residue from Murray Springs. CR p carbonribbon. D, HRTEM image of a rounded cubic ND at Lake Cuitzeo. Parallel lines represent {111}-type lattice planes(2.06-A spacing), as viewed along the [110] zone axis. A color version of this figure is available online.

mond with sp3 bonding. This pattern definitivelyeliminates the possibility that these nanocrystalsare graphite, graphene, and graphane.

The EELS plot in figure 8B is of synthetic com-mercial cubic diamond (PlasmaChem) and closelyresembles the YDB spectrum. Note that the peakpis absent in this case, because of the lack of em-bedding amorphous carbon matrix in the commer-cial diamonds. The Murray Springs EELS spectrum(fig. 8C) indicates a mix of mostly n-diamonds andi-carbon and is significantly different from the EELSspectrum for the synthetic cubic diamond. In thiscase, a peak is present, indicating some sp2 bond-ping, consistent with n-diamond, which is reportedto contain approximately 5% sp2 and 95% sp3 bond-ing (Peng et al. 2001). The absence of a peak-and-trough pattern (arrows in fig. 8A, 8B) indicates thatthe nanoparticle is not a cubic ND. This spectrumis a close match for previously published spectrafor n-diamond and i-carbon (curved lines above thespectra) but is a poor match for graphite, graphene,and amorphous carbon, which typically display

peaks with greater amplitude, indicating pro-pportionately more sp2 bonding. Thus, low-ampli-tude peaks can be used to infer that a particlepis likely a diamond polytype.

EFTEM. We used EFTEM in some cases as an

elemental mapping technique to investigate thespatial distribution of carbon and to examine itsrelative atomic bonding (sp2 and sp3), as was firstused for YDB NDs by Tian et al. (2011). Figure 9A,from the YDB layer in Lake Cuitzeo, is called azero-loss image, exhibiting various lighter nano-crystals embedded in the grayer amorphous carbonresidue, itself superimposed on the darker amor-phous carbon TEM grid film, marked AC. Thisimage is displayed in reverse contrast for clarity.Using HRTEM and FFT, we identified and labeledthe larger nanocrystals by polytype; in this view,n-diamonds, i-carbon, and cubics have a ratio of3 : 1 : 1. There is also one lonsdaleite-like crystal,with a relative abundance that is atypically high inthis case. Next, we generated a jump ratio image(fig. 9B) by comparing postedge energies character-istic of carbon ( edge) with background energiesj(260 eV). The resulting map displays bonding dif-ferences for the larger particles, whose bright gray-to-white contrast indicates the presence of someamount of sp3 bonding that is characteristic of NDsbut not of graphite, graphene, and graphane. Theblack-colored areas (AC) represent the TEM gridfilm, composed of amorphous carbon with no sp3

bonding.The brighter areas between particles indicate the



Figure 4. Two techniques for identifying cubic nanodiamonds (NDs). A, Selected-area electron diffraction patternof cubic NDs from Murray Springs, with d-spacings (Younger Dryas Boundary: 426 ppb at 46.5 cm below surface). B,Fast Fourier transform (FFT) of cubic NDs from Murray Springs; C, FFT of single cubic ND from Murray Springs,viewed along the [110] zone axis. A color version of this figure is available online.

presence of small, sub-nanometer nanoparticlesdisplaying sp3 bonding, and this observation, alongwith the visible whitish-gray color of most resi-dues, indicates that there is very little black, amor-phous carbon present. Instead, the residue betweenNDs appears to consist of diamond-like nanocrys-tals arranged in short-range ordering that causesthem to appear amorphous. It is possible that theseare diamondoids, which are cage-like, ultrastable,saturated hydrocarbons (de Araujo et al. 2012),whose carbon-carbon lattice framework is largelyidentical to a portion of the cubic ND lattice. Dia-mondoids are found in hydrocarbon and coal de-posits; they are nearly as hard as diamonds; eachdiamondoid typically includes from 10 to 30 carbonatoms; they are composed almost entirely of sp3-bonded carbon (de Araujo et al. 2012); and diamon-doid powder can be visibly whitish to clear (Schoelland Carlson 1999). Diamondoids compare favora-bly to most of the crystals in the extracted residue,which also are dominantly carbon, have sp3 bond-ing, produce a diffuse SAD pattern because of theirsmall size, and are optically clear to white. Becauseboth n-diamonds and diamondoids have been foundin petroleum deposits related to the K-Pg, onemight speculate that something similar happenedduring the YDB impact, especially if an impact tookplace in deep, petroleum-rich offshore sediments.

More work is necessary to determine the natureand identity of these small nanoparticles, but theymay be a clue to the YDB ND formation process.

The NDs in figure 9B typically are brighteraround their edges but somewhat darker in theircenters. This variability highlights a disadvantageof using EFTEM, which works best with thin layersof NDs and/or amorphous carbon residue. Becausethe fraction of electrons that undergo a single scat-tering event in a thick area is less than that in athinner area, the jump ratio map shows a strongerEFTEM signal in the thin areas, even though bothareas are equally populated by NDs.

Identification of n-Diamonds and i-Carbon. TheYDB layer contains two diamond-like polytypes, n-diamonds and i-carbon, that were first synthesizedin the laboratory (Wen et al. 2007 and referencestherein). Outside of the laboratory, face-centeredcubic NDs (another name for n-diamonds) werefirst reported within carbon spherules from near-surface sediments of unknown age across northernEurope by Rosler et al. (2006). Later, both n-dia-monds and i-carbon were found in the YDB layer(Kennett et al. 2009a) and the K-Pg layer (Bunch etal. 2008, 2009).

The YDB n-diamonds display the same d-spac-ings as cubic NDs, except for the added presenceof so-called forbidden reflections evident in SAD


000 C . R . K I N Z I E E T A L .

Figure 5. Transmission electron microscopy images (in reverse contrast for clarity; top) and selected-area electrondiffraction patterns (bottom) of nanodiamonds from three continents. A, B, n-Diamonds from sediment (Sed) at AbuHureyra, Syria (Younger Dryas Boundary [YDB]: 443 ppb at 405 cm below surface [cmbs]). C, D, n-Diamonds in carbonspherules (CS) at Santa Maira Cave, Spain (YDB: 38 ppb at 3.5 cmbs). E, F, n-Diamonds from sediment at Lindenmeier,Colorado (YDB: 143 ppb at 101 cmbs). A color version of this figure is available online.

patterns and FFTs of HRTEM images (table D1).The lattice planes that produce these reflections arepresent in both cubic and n-diamonds, but as a re-sult of destructive interference, the reflections aretypically invisible in cubic NDs, hence the termforbidden. These reflections may become visiblein cubics for several reasons: first, because of dou-ble diffraction caused by the twinning; second, asa result of flaws caused by the occasional substi-tution of other elements for carbon atoms; andthird, because of incomplete unit cells at the edgeof the crystal. Thus, it is possible that n-diamondsare actually twinned cubic NDs.

A TEM image from Murray Springs (fig. 10A) ex-hibits more than 100 NDs that are tilted 30 fromnormal in the microscope. A second TEM image(fig. 10B) shows the same NDs, but tilted througha 45 arc to 15. Note that corresponding objectsappear similar in both images (e.g., particles 14),indicating that they all are rounded to subroundedand not planar. An SAD pattern of the same objects(fig. 10C) indicates that these are n-diamonds.

SAD, FFT, and HRTEM for n-Diamonds and i-Carbon. An SAD pattern from Murray Springs (fig.11A) displays seven lattice spacings of n-diamonds(table D1); EDS analyses confirmed these particlesto be composed of carbon. The FFT of an HRTEMimage (fig. 11B) shows five d-spacings of a single n-

diamond, and the HRTEM image of the same n-diamond (fig. 11C) shows three values representingtwo lattice planes, of which the 1.78-A plane is aforbidden reflection in cubic NDs. The SAD patternfor Lake Cuitzeo material (fig. 11D) displays thefirst seven lattice spacings of i-carbon crystals. TheFFT of the HRTEM image of a single i-carbon crys-tal (fig. 11E) shows three values representing twolattice planes, and the HRTEM image of the samei-carbon crystal (fig. 11F) shows three values rep-resenting two lattice planes, as viewed along the[001] zone axis. The EDS, EELS, and EFTEM anal-yses are not shown for these NDs but are similarto the analyses above for NDs.

Twinning in YDB NDs. The YDB NDs larger than2 nm are usually made up of two or more crystalsthat share a common lattice plane (the twin plane)and grow symmetrically in different orientations;twinned NDs were observed at all YDB sites. Twin-ning is also commonly observed in meteorites, cos-mic-impact craters, and commercial NDs (Israde-Alcantara et al. 2012b and references therein).Twins can form in numerous configurations, in-cluding star twins, as observed by Tian et al.(2011) in the YDB layer from Lommel. Figure 12Ashows a multiply twinned ND from Kangerlussuaq,Greenland, composed of 120 individual crystalswith lattice plane spacings and angles character-



Figure 6. Transmission electron microscopy images (top) and selected-area electron diffraction patterns (bottom)used to identify nanodiamonds (NDs) in carbon spherules (CS). A, B, i-Carbon from Kimbel Bay, North Carolina(Younger Dryas Boundary [YDB]: 721 ppb at 351 cm below surface [cmbs]). C, D, n-Diamond from Topper, SouthCarolina (YDB: 108 ppb at 60 cmbs). E, F, n-Diamond from Lingen, Germany (YDB: 431 ppb at 43.5 cmbs); note twinND at upper right. A color version of this figure is available online.

istic of n-diamond, as reported by Yang et al. (2008).Figure 12B is an FFT of the central crystal and dis-plays eight lattice spacings that are consistent withn-diamonds and cubic NDs. Figure 12C is a star-twin n-diamond, so named because of its fivefoldstar-like symmetry. More twinned n-diamonds areshown in figure C6.

TEM, SAD, and Scanning Electron Microscopy of NDsin Carbon Spherules. Figure 13A, from Gainey,Michigan, is a TEM image of a carbon spherulefragment, showing embedded NDs as black dotswithin the amorphous matrix at the arrow. TheEDS analyses indicate that these are carbon parti-cles. Figure 13B is a photomicrograph of a typicalYDB carbon spherule. Figure 13C is an SAD pat-tern, demonstrating that the particles are n-dia-monds with a possible minor admixture of otherNDs. Figure 13D is an HRTEM image from theChobot site of a carbon spherule fragment contain-ing NDs, such as the dark object marked by thearrow. This fragment was removed from inside acarbon spherule with a needle, demonstrating thatsome NDs form throughout the interior matrix ofthe spherules. The photomicrograph (fig. 13E)shows a carbon spherule with a hollow interior.Figure 13F is an SAD pattern demonstrating thatthese particles are n-diamonds. The NDs found incarbon spherules are indistinguishable from the

rounded NDs initially discovered by Rosler et al.(2006) and reported in Yang et al. (2008).

We investigated whether NDs always are dis-tributed throughout the interior matrix of carbonspherules and glass-like carbon, the latter of whichhas been reported to contain NDs (van Hoesel etal. 2012). We used a focused ion beam to mill apiece of glass-like carbon extracted from the YDBlayer at the M33 site, the rim of a Carolina bay inMyrtle Beach, South Carolina (for site details, seeFirestone et al. 2007). The TEM analyses showedthat diamonds were present only from the surfacedown to a depth of 0.75 mm and were not observedin the interior (fig. 14A). The surface layer wassharply demarcated and fused to the interior of thespherule. Figure 14B shows a chip of the glass-likecarbon surface layer removed with a needle; it con-tains hundreds of densely packed NDs. Similarly,examination of carbon spherules from WatcombeBottom, United Kingdom, suggested that someNDs were clustered in a thin layer on the outsideof carbon spherules (fig. 14C). On the other hand,the carbon spherule fragment from the Chobot site(fig. 13D) displays NDs from deep within a spher-ule, and Rosler et al. (2006) reported NDs attachedto the inner surfaces of vesicles in European carbonspherules. These results indicate that at least someNDs form only on surfaces of carbon spherules and


000 C . R . K I N Z I E E T A L .

Figure 7. Scanning transmission electron microscopy (STEM) images (A, D), selected-area electron diffraction patterns(B, E), and energy-dispersive X-ray spectrometry plots (C, F) for elemental abundance of carbon: Younger DryasBoundary nanodiamonds (NDs) from Lake Cuitzeo (493 ppb at 280 cm below surface; AC) and synthetic NDs fromPlasmaChem (DF). Carbon in both is greater than 98%. HAADF p high-angle annular dark field. A color versionof this figure is available online.

Figure 8. Electron energy-loss spectroscopy spectra for differentiating cubic nanodiamonds (NDs) from n-diamonds,i-carbon, and other forms of carbon. A, Younger Dryas Boundary (YDB) NDs from Murray Springs (426 ppb at 46.5cm below surface). B, Synthetic cubic NDs. C, YDB n-diamonds from Murray Springs; solid lines represent spectrafor graphite, amorphous carbon, n-diamond, and i-carbon from Berger et al. (1988); the graphene spectrum is fromDaulton et al. (2010). A color version of this figure is available online.

glass-like carbon, whereas others form throughoutthem. The reason for this difference is unclear, butfinding NDs on spherule surfaces is consistent withone scenario, in which molten carbon spherulesand glass-like carbon in an impact fireball were ex-posed briefly to anoxic conditions and high tem-peratures that caused NDs to form on their surface

layers, but not inside them, while preventing thecarbon from incinerating.

Identification of Lonsdaleite-Like Crystals. Poten-tial YDB lonsdaleite crystals have been identifiedand analyzed with HRTEM, FFT, SAD, and EDS(see Kennett et al. 2009b; Kurbatov et al. 2010;Israde-Alcantara et al. 2012b). On the basis of EDS



Figure 9. Energy-filtered transmission electron microscopy (EFTEM) for identifying carbon bonding in particles. A,Nanodiamonds (NDs) from Lake Cuitzeo (Younger Dryas Boundary: 493 ppb at 280 cm below surface) in a zero-lossEFTEM image (reverse contrast for clarity). B, Jump ratio map, exhibiting NDs with sp3 bonding in the lighter areas(reverse contrast for clarity). N p n-diamond; I p i-carbon; C p cubic diamond; L p lonsdaleite-like crystal; AC pamorphous carbon grid film. A color version of this figure is available online.

and EELS measurements, all the lonsdaleite-likecrystals observed contain only carbon, with noother elements present but oxygen, eliminating thepossibility that they are unidentified, noncarbonmineral. We have observed the crystals along threemajor zone axes ([0001], [ ], and [ ]), and all 0111 1121measured lattice planes are consistent with lons-daleite and no other known carbon allotrope, in-cluding graphite, graphene, and graphane. Never-theless, these crystals are too rare to allowdefinitive identification with all available analyti-cal methods. Because many new, very hard formsof carbon have been discovered within the past fewdecades, these crystals may be some unidentified,diamond-like carbon allotrope. Therefore, we con-sider the identification of lonsdaleite to be provi-sional, pending further work. We include the evi-dence below for the benefit of other researchers.

Daulton (2012) questioned the identification oflonsdaleite in Kennett et al. (2009b), and we agreethat the one cluster of nanoparticles in figure 2D2F of the latter paper appears to consist of graphene-graphane aggregates, which mimic the d-spacingsof lonsdaleite. We thank Daulton (2012) for point-ing this out. He also questioned figure 2A2C ofKennett et al. (2009b). Although the analyses wereinsufficient to conclusively identify the nanocrys-tal shown as lonsdaleite, we find no evidence to

eliminate it as a possibility, as discussed below inAngular Lonsdaleite-Like Crystals.

Nearly all lonsdaleite observed in known impactcraters is angular (Koeberl et al. 1997), and occa-sionally, YDB lonsdaleite-like crystals have beenobserved that are angular (Kennett et al. 2009b).However, in most cases, YDB NDs are rounded tosubrounded in shape. The rounded lonsdaleite-likecrystals may be due to modification by the extrac-tion process, but this seems unlikely because acid-extracted n-diamonds are morphologically identi-cal to nonacidized n-diamonds found in carbonspherules. Alternately, the rounded shapes may re-sult from a different mode of formation. For ex-ample, subrounded to rounded commercial lons-daleite has been produced by microplasmadissociation of ethanol vapor (Kumar et al. 2013),under conditions somewhat similar to those in animpact event, i.e., anoxic atmosphere and a carbonsource. Below, we describe some analyses used tocharacterize the lonsdaleite-like crystals.

Angular Lonsdaleite-Like Crystals. A single YDBsite, Arlington Canyon, California, contains flake-like lonsdaleite-like crystals (Kennett et al. 2009b),a shape similar to that of previously reported cubicNDs (Rosler et al. 2006) and plate-like lonsdaleitefrom known impact craters (Koeberl et al. 1997). ASTEM image shows a tabular lonsdaleite-like crys-


000 C . R . K I N Z I E E T A L .

Figure 10. Tilted transmission electron microscopy (TEM) images of a field of nanodiamonds. TEM images (A, B)and selected-area electron diffraction (C) pattern of n-diamonds from Murray Springs (YDB: 426 ppb at 46.5 cm belowsurface), tilted through 45 (30 [A] through 15 [B]). Comparison of various selected objects, such as 14, dem-onstrates that these particles are three-dimensional, rounded to subrounded crystals. A color version of this figure isavailable online.

tal (2.9 mm long) from the YDB at Arlington Canyon(fig. 15A). This is the same lonsdaleite-like grainshown in Kennett et al. (2009b) as figures 2A2Cand S2B, referred to above. Figure 15B is a TEMimage of the same crystal as in panel A. Figure 15Cis an EDS elemental map of the same crystal andshows the composition to be carbon (lighter con-trast), with no other elements present.

Figure 16A is an HRTEM image showing thesame crystal. The double lines define three sets oflattice planes consistent with { }-type planes of1010lonsdaleite (prism planes) with a d-spacing of 2.18A, as viewed along the [0001] zone axis. Figure 16Bpresents an FFT of an HRTEM image of the samecrystal, displaying a spot pattern consistent withthe d-spacings for lonsdaleite shown in table D1.The spot pattern matches crystallographic simu-lations performed for lonsdaleite. Multiple mea-surements with a calibrated beam (diamond stan-dard) attained an accuracy of approximately 1%,producing a range of 2.162.20 A for the 2.18-Ad-spacing. We also measured d-spacings for com-mercial graphene and were able to easily distin-guish between the d-spacings of 2.18 A for the lons-daleite-like crystal and 2.13 A for graphene,eliminating both graphene or graphane as candi-dates. Although the lonsdaleite-like crystals maybe some other unknown carbon-based mineral,

there is no current evidence that excludes the pos-sibility that it is lonsdaleite.

Rounded Lonsdaleite-Like Crystals. The YDB layerat several sites also contains rounded lonsdaleite-like crystals. Figure 17A is a STEM image from theGreenland Ice Sheet near Kangerlussuaq, exhibitingrounded lonsdaleite-like crystals (arrows) mixedwith n-diamonds and i-carbon, all ranging from 4to 200 nm in diameter. Figure 17B is an HRTEMimage of a rounded 10-nm lonsdaleite-like crystal.The EDS results were presented in Kurbatov et al.(2010), confirming that the crystal is carbon, andan EELS spectrum indicated high sp3 bonding, elim-inating the possibility that it is graphite, graphene,or graphane. Figure 17C is an FFT of an HRTEMimage of the same lonsdaleite-like crystal, showinglattice spacings consistent with lonsdaleite.

We also extracted lonsdaleite-like crystals fromthe YDB layer in several caves. Figure 18A, 18Bshows a 200-nm-long lonsdaleite-like crystal fromSheriden Cave in Ohio. We tilted the TEM stageto confirm that the crystal is three-dimensional androunded. Figure 18C is a STEM image showing a53-nm-wide object from Daisy Cave on San MiguelIsland, one of the Channel Islands located off SantaBarbara, California. Using variable focusing andtilting of the electron beam, we determined that



Figure 11. Selected-area electron diffraction (SAD), fast Fourier transform (FFT), and high-resolution transmissionelectron microscopy (HRTEM) images. AC, SAD pattern (A) and FFT (B) of the HRTEM image (C) of an n-diamondfrom Murray Springs (Younger Dryas Boundary [YDB]: 426 ppb at 46.5 cm below surface [cmbs]). DF, SAD pattern(D) and FFT (E) of the HRTEM image (F) of i-carbon from Lake Cuitzeo (YDB: 493 ppb at 280 cmbs). Images B andC are from Israde-Alcantara et al. (2012b) and are used with permission. A color version of this figure is availableonline.

the object is a three-dimensional ball. The EDSanalysis indicates that the ball is composed almostsolely of carbon, while HRTEM confirms that thematrix is amorphous and studded with a mix ofNDs, including one n-diamond star-twin (2.06 A)and one lonsdaleite-like crystal (2.18- and 1.93-Ad-spacings). We compared that 2.18-A spacing tothe similar 2.13-A d-spacing for graphene and foundthat we were able to distinguish them, making ithighly unlikely that any of these lonsdaleite-likecrystals are graphene. Other than lonsdaleite, noother known carbon allotrope matches all the evi-dence for these crystals.

Origin of YDB NDs

Multiple explanations have been proposed for theorigin of YDB NDs, as follows.

Potential Origin by Cosmic Flux. Cubic NDs arepresent in meteorites and cosmic dust (Hannemanet al. 1967; Grady et al. 1995; Huss and Lewis 1995),and lonsdaleite is present in some meteorites(Daulton et al. 1996). These observations led Pinteret al. (2011) and others to speculate that YDB mi-crospherules and NDs arrived as components of thegradual, noncatastrophic rain of cosmic debris; ifthat speculation were correct, those NDs should

display cosmic chemical signatures. Instead, anal-yses indicate that the isotopic compositions of car-bon and nitrogen (d13C, d15N, and C/N) in YDB NDsare consistent with a terrestrial origin (Tian et al.2011; Israde-Alcantara et al. 2012b). Those resultsare supported by Gilmour et al. (1992), who foundthat d13C and d15N values for K-Pg NDs are consis-tent with formation from terrestrial carbon duringthe impact itself (Belcher et al. 2005).

Potential Origin from Volcanism or in the Mantle.Cubic diamonds occur in terrestrial deposits, suchas kimberlite pipes, which originated from themantle. Boslough et al. (2012) pointed out that YDBlonsdaleite may originate with terrestrial cubic di-amonds because it has been found in a cubic dia-mond deposit in North Kazakhstan, in Ukrainiantitanium placer deposits, in Yakutian diamondplacers, and in metamorphosed basaltic rocks onthe Kola Peninsula and in the Urals. However, itis unclear whether all of those lonsdaleite examplesare terrestrial in origin. For example, Shelkov et al.(1998) presented evidence that lonsdaleite in theYakutian placers eroded from an impact crater.

Daulton (2012) also suggested that YDB NDsmay be derived from mantle material, but isotopicanalyses (Tian et al. 2011) are inconsistent with amantle origin. If such distribution occurred, the


000 C . R . K I N Z I E E T A L .

Figure 12. High-resolution transmission electron microscopy (HRTEM) and fast Fourier transform (FFT) images ofnanodiamonds (NDs) from Kangerlussuaq, Greenland (Younger Dryas Boundary: 174 ppb at 548 cm below surface).A, HRTEM image of an unusually large, multiply twinned n-diamond or flawed cubic ND (53 nm # 39 nm) with20 conjoined crystals. Center diagram shows typical d-spacings (in A) and corresponding angles. B, FFT of HRTEMimage of the central crystal in A shows eight d-spacings that are consistent with n-diamond and cubic NDs, whenviewed along the [011] zone axis. C, HRTEM image of 24-nm-wide n-diamond star-twin. Arrows are at planeboundaries. Parallel lines indicate d-spacing of 2.06 A. A color version of this figure is available online.

geochemical signature of the mantle host materialshould have been detected in more than 700 geo-chemical analyses conducted on YDB materials(Wittke et al. 2013), and instead, no such signatureis apparent. Mantle-derived NDs have never beenfound in any known geological column associatedwith coeval peaks in impact markers, arguingagainst this hypothesis. In any event, terrestriallonsdaleite has never been observed in any depositsof any age in Europe or North America, where YDBlonsdaleite-like crystals are currently found.

We also considered whether NDs might be pro-duced from volcanic eruptions. To test this, we ap-plied our protocol (see Material and Methods) totephra from the Laacher See eruption that occurrednear the time of the YDB event. We observed noNDs and no magnetic spherules, eliminating thepossibility that this eruption deposited those prox-ies in the YDB layer.

Potential Origin in Wildfires. Rosler et al. (2005,2006) and Yang et al. (2008) reported ND-enrichedcarbon spherules of unknown origin at 70 sitesacross western Europe, including Germany, Aus-tria, and Belgium. The ND-enriched carbon spher-

ules were found in upper soils, but recently, N.Schryvers (2014, personal communication) wasmore specific, indicating that their samples werecollected from between 10 and 20 cm deep, afterremoval of topsoil. They wrote that the soils weremodern but reported no dates. Yang et al. (2008,p. 941) stated that some ND-rich carbon spherulesin Germany were found associated with small-scale crater-like structures, estimated to be 1000yr old, but no craters were associated with the ND-rich carbon spherules in other countries. Theyadded that their origin is unclear but that an im-pact related origin ... cannot be ruled out (p. 941),If so, the age is currently unknown, because near-surface sediments can range in age from modern tomillions of years old. As an example, ND-enrichedcarbon spherules from Gainey (14 cm deep) andChobot (33 cm deep) were intermixed with Clovis-age tools dating from 13,25012,800 cal BP (Watersand Stafford 2007). Thus, it is conceivable thatsome or all of the European near-surface, ND-richcarbon spherules date to the YD onset. Determin-ing the age of the surficial sediments at these sitesis necessary to answer these questions.



Figure 13. Transmission electron microscopy images (A, B, D, E) and selected-area electron diffraction patterns (C,F) of carbon spherules from Gainey, Michigan (Younger Dryas Boundary [YDB]: 3933 ppb at 30 cm below surface[cmbs]; AC), and Chobot, Alberta, Canada (YDB: 10 ppb at 13.5 cmbs; DF). A color version of this figure is availableonline.

Van Hoesel et al. (2012, 2014) speculated abouta possible wildfire origin for the ND-rich Europeancarbon spherules, on the basis of a recent discoveryby Su et al. (2011) that NDs form in candle flames.The NDs are produced at high temperatures (11001300C) at the anoxic center of the flames, but be-cause NDs combust at 400600C in oxygen-richatmospheres (Hough et al. 1999), they are rapidlydestroyed as they approach the flames oxidationboundary. To protect the NDs from destruction, Suet al. (2011) developed an elaborate procedure usingporous aluminum foils to capture and extract theNDs before they could combust. While innovative,this elaborate process does not exist in nature. Theparticles were identified as face-centered cubicNDs, more commonly known as n-diamonds, andthere was no evidence of typical, body-centered cu-bic NDs, as found in the YDB layer.

It is well established that carbon spherules canbe produced in intense wildfires involving conifers(Firestone et al. 2007; Israde-Alcantara et al. 2012b).

However, no natural wildfires are known to pro-duce NDs inside carbon spherules or other parti-cles. Similarly, no laboratory experiments havebeen able to produce NDs under conditions thatnormally appear at Earths surface. If NDs could beproduced in natural fires, which typically recur ev-ery 1001000 yr in any given area, they should becommon and ubiquitous in sediments of all ages.Instead, NDs in contiguous stratigraphic horizonsare nonexistent to rare (Tian et al. 2011; Bement etal. 2014) and do not correlate with sedimentary lay-ers with high charcoal abundance (Bement et al.2014). These observations are confirmed by ourown work at 22 YDB sites. Similarly, NDs have notbeen found above or below the K-Pg impact layer(Carlisle and Braman 1991; Gilmour et al. 1992;Bunch et al. 2008), even though biomass burningis accepted as having been broadly pervasive overthe K-Pg boundary interval (Wolbach 1990).

In summary, there is no evidence for and noknown process for production of NDs in natural


000 C . R . K I N Z I E E T A L .

Figure 14. Transmission electron microscopy (TEM) images of nanodiamonds (NDs) in glass-like carbon (GLC) andcarbon spherules (CS). A, TEM image from M33 Bay, Myrtle Beach, South Carolina (40 cm below surface [cmbs]);the sectioned GLC displays only amorphous carbon and no NDs (the dark crust is mounting material). B, TEM imageof a flake from the surface of the GLC in A; selected-area electron diffraction patterns reveal a high abundance of n-diamonds. C, TEM image from Watcombe Bottom, United Kingdom (Younger Dryas Boundary: 130 ppb at 65 cmbs)shows a thin zone of cubic NDs along one edge of a fragment of a carbon spherule. A color version of this figure isavailable online.

wildfires. This argues against biomass burning asthe source of the assemblage of NDs in the YDBor other sedimentary sequences.

Potential Origin within Sclerotia. Scott et al.(2010) and Hardiman et al. (2012) stated that allcarbon spherules from Arlington Canyon and othersites are simply either charred fecal pellets or fun-gal sclerotia. To be viable, the sclerotial hypothesismust account for the presence of millions of NDsentrained within each carbon spherule (Kennett etal. 2009a). There is no credible mechanism bywhich fungi can create NDs in sclerotia, but weconsidered whether NDs might have adhered topreexisting sclerotia while colocated in YDB sedi-ment. For comparison, the average sedimentaryabundance of NDs is 200 ppb, whereas the NDconcentrations in carbon spherules is 135% at threesites, a difference of more than one million times.There is no plausible process by which sclerotiacould extract NDs from surrounding sediment,

concentrate them a million times, and do so onlyat one time during the past 13,000 yr. Thus, thebest explanation is that ND-rich carbon spherulesderive from conifers that were incinerated by theimpact event (Israde-Alcantara et al. 2012b).

Potential Origin from Lightning. We consideredwhether diamonds might form during high-tem-perature lightning strikes. To evaluate this, we ap-plied the protocol to extract potential NDs from acollection of fulgurites but observed not even oneND. Furthermore, Wittke et al. (2013) studied rem-anent magnetism in YDB impact spherules that areclosely associated with NDs. They found no evi-dence for lightning strikes in YDB sediment, thusarguing against this hypothesis.

Potential Origin as Anthropogenic NDs. Rosler etal. (2006) and Bement et al. (2014) discovered NDsin deposits near the ground surface. We investi-gated the possibility that modern anthropogenic ac-



Figure 15. Younger Dryas Boundary lonsdaleite-like crystal. A, Scanning transmission electron microscopy imagefrom Arlington Canyon, California (1760 ppb at 462 cm below surface). B, Transmission electron microscopy imagesof same crystal. C, Corresponding energy dispersive X-ray spectrometry elemental carbon map of the crystal; no otherelements were detected. Images from Kennett et al. (2009b); used with permission. A color version of this figure isavailable online.

Figure 16. Younger Dryas Boundary (YDB) lonsdaleite-like crystal from Arlington Canyon (1760 ppb at 462 cm belowsurface). A, High-resolution transmission electron microscopy of the crystal; B, corresponding fast Fourier transform.A color version of this figure is available online.

tivities might produce synthetic NDs that migrateddownward to the YDB. We examined fly ash residuefrom a modern New Jersey power plant that incin-erates coal at high temperatures under anoxic con-ditions, similar to some laboratory conditions thatproduce NDs. We found abundant graphene but noNDs and no melt-glass containing high-tempera-ture, melted quartz. Furthermore, some YDB NDsare found up to 5 m below surface but not in in-tervening layers, making it unlikely that they mi-grated downward from the surface. In addition, we

found no detectable NDs in surficial sediments in-vestigated at eight sites (fig. 2; table D4). All thesefindings argue against the hypothesis that NDs areproduced through anthropogenic activities and arecommon in surface and other sediments.

Potential Origin by Cosmic Impact. Cubic NDshave been reported at the K-Pg boundary (Gilmouret al. 1992; Hough et al. 1997), and Israde-Alcantaraet al. (2012b) reported that YDB NDs are morpho-logically and compositionally similar to those in


000 C . R . K I N Z I E E T A L .

Figure 17. Scanning transmission electron microscopy (STEM) and high-resolution transmission electron microscopy(HRTEM) images of lonsdaleite-like crystals. A, STEM image of group of lonsdaleite-like crystals (arrows) mixed withother NDs, from Kangerlussuaq (Younger Dryas Boundary [YDB]: 174 ppb at 548 cm below surface [cmbs]). B, HRTEMimage of a 10-nm lonsdaleite-like monocrystal from Lake Cuitzeo (YDB: 493 ppb at 280 cmbs); one ( ) plane with1011a spacing of 1.93 A is visible, along with the ( ) plane at 2.18 A, consistent with lonsdaleite. C, Fast Fourier1010transform of an HRTEM image of the same crystal as in B, with d-spacings along the [ ] zone axis, revealing one0111( ) plane with a lattice spacing of 2.18 A and two ( ) planes with lattice spacings of 1.93 A. A color version 1010 1011of this figure is available online.

the K-Pg (Kennett et al. 2009a, 2009b; Kurbatov etal. 2010). Angular lonsdaleite crystals also formedin some impact events via shock metamorphism ofgraphite in the target rocks (Hough et al. 1997; Koe-berl et al. 1997; Langenhorst et al. 1998; DeCarliet al. 2002; Oleinik et al. 2003). Lonsdaleite grainshave been reported in impact events, e.g., Ries Cra-ter, the K-Pg event (Bunch et al. 2008), and the 1908Tunguska airburst in Siberia (Bunch et al. 2008;Kvasnytsya et al. 2013). Although lonsdaleite isknown to be formed through shock metamorphismduring surface impacts, its presence at the site ofthe Tunguska airburst indicates that it also canform during cosmic airbursts. This is demonstratedin laboratory experiments (Miura and Okamoto1997), in which a high-velocity impact into a lime-stone target produced carbon vapor that condensedinto graphite inside a high-temperature, reducingvapor plume. Subsequently, lonsdaleite and cubicNDs formed when the carbon plume reacted withwater ice, creating oxidizing conditions. This dem-onstrates that lonsdaleite can form through a pro-cess similar to carbon vapor deposition (CVD), es-

pecially if the YDB impactor struck the ice sheetor oceans, as proposed by Firestone et al. (2007).

Criticism of an impact origin for the YDB hasincluded an apparent absence of an impact crater(Boslough et al. 2012). However, this position con-tradicts a broad consensus among impact research-ers, including some of those coauthors, that someimpact events lack known craters or that the cra-ters remain undiscovered (Boslough and Crawford2008). Some examples are the Tunguska airburstdebris field (2 # 103 k

nanodiamond rich layer across 3 continents

Documents

impact diamonds

carbon spherules

major cosmicimpact proxy

yd boundary ydb layer

younger dryas boundary

concentrationof nds

nanodiamonds nds

terrestrial carbon