Evidence of Cosmic Impact at Abu Hureyra, Syria at the Younger Dryas Onset (~12.8 ka): High-temperature melting at >2200 °C

At Abu Hureyra (AH), Syria, the 12,800-year-old Younger Dryas boundary layer (YDB) contains peak abundances in meltglass, nanodiamonds, microspherules, and charcoal. AH meltglass comprises 1.6 wt.% of bulk sediment, and crossed polarizers indicate that the meltglass is isotropic. High YDB concentrations of iridium, platinum, nickel, and cobalt suggest mixing of melted local sediment with small quantities of meteoritic material. Approximately 40% of AH glass display carbon-infused, siliceous plant imprints that laboratory experiments show formed at a minimum of 1200°–1300 °C; however, reflectance-inferred temperatures for the encapsulated carbon were lower by up to 1000 °C. Alternately, melted grains of quartz, chromferide, and magnetite in AH glass suggest exposure to minimum temperatures of 1720 °C ranging to >2200 °C. This argues against formation of AH meltglass in thatched hut fires at 1100°–1200 °C, and low values of remanent magnetism indicate the meltglass was not created by lightning. Low meltglass water content (0.02–0.05% H2O) is consistent with a formation process similar to that of tektites and inconsistent with volcanism and anthropogenesis. The wide range of evidence supports the hypothesis that a cosmic event occurred at Abu Hureyra ~12,800 years ago, coeval with impacts that deposited high-temperature meltglass, melted microspherules, and/or platinum at other YDB sites on four continents.

. Photograph of Abu Hureyra occupational surface, dating to ~12,825 ± 55 cal BP 4 . Wattle-and-daub huts typically enclosed multiple small, hand-dug round pits, each less than a few meters across. The main interior room of a pit-house (red arrow) and charcoal-rich outside work area (orange arrow) contained abundance peaks in spherules, nanodiamonds, carbon spherules, meltglass, and platinum. Note that the surface contains numerous small holes, some having been dug by the villagers, suggesting considerable reworking of sediment. Orange lines along walls indicate the strata that were deposited during the onset of YD climate change. Figure By 1300°C, the fine-grained, clayey sediment melted and many small grains began to melt. (c) By 1400°C and continuing to 1700°C, progressively more detrital grains melted, and at 1700°C, some larger grains still survived. Heating experiments used Abu Hureyra bulk sediment from level 435, sample ES15, 395 cm depth. Figure S4. Charcoal temperature comparisons from reflectance measurements for different organic materials: #1 = loose Abu Hureyra charcoal extracted from AH sediment; #2 = charcoal embedded within Abu Hureyra meltglass; bars #3 through #5 = glass made from reeds, oak, and pine; bar #6 = charcoal within trinitite; bars #8 through #9 = Calgon activated carbon and source charcoal; bars #9 through #11 = YDB carbon spherules; and bar #12 = charcoal from Tunguska sediment. Gray rectangles represent the range of temperatures inferred from our reflectance percentages. Orange bars represent measured/inferred minimum formation temperatures. Figure S5. Reflectance results for (a) Abu Hureyra charred carbon that was fully encapsulated in AH meltglass. Inferred temperature of >1250°C, the approximate melting point of local sediment. Reflectance-inferred temperature was ~421°C, a difference of >829°C. Samples are from YDB level 445, sample E301, 405 cm; (b) Charcoal splattered with trinitite melted at >1250°C, the approximate melting point of local sediment. Reflectance-inferred temperature was ~406°C, a difference of >844°C.           (c) At 1400°C, a 250-μm-long zircon grain showed more thermal alteration but still maintained its original shape. (d) At 1500°C, 3 zircon grains ~120 µm wide show thermal alteration, while retaining their original shapes. Most but not all zircon <50 µm melted at ≤1500°C. (e) Photomicrograph of zircon grains (labeled Zr) heated to ~1700°C shows that the largest, white zircon grains generally retained their original shapes. Laboratory-melted bulk sediment from level 435, sample ES15, 395 cm depth. Figure S17. Heating experiments with chromite-spiked AH sediment. We conducted laboratory heating experiments to determine the response of these minerals to high temperatures, using ~1 g of AH sediment, mixed with ~10 wt.% crushed grains of chromite and zircon. At 1200ºC, the fine-grained, clayey sediment melted and enclosed numerous unmelted grains of black chromite, white zircon, clear/white quartz, and other detrital grains. At 1400ºC, a larger percentage of the bulk sediment melted, but there were still large numbers of unmelted grains visible in the dark, transparent glass. (a) At 1500°C, no detectable melting occurred for the added grains of chromite ((Fe,Mg)Cr2O4), labeled Cr. Limited edge melting occurred in zircon, labeled Zr. (b)-(c) At 1600° and 1700°C, no detectable melting occurred in chromite grains. For zircon, smaller grains melted completely and diffused into the aluminosilicate matrix at ~1600°C, but zircon grains approximately >100 µm showed moderate melting at ~1700°C. Under normal conditions, chromite crystals typically melt at equilibrium temperatures of ~2265°C 7 . Laboratory-melted bulk sediment from level 435, sample ES15, 395 cm depth. Figure S18. Images of chromite grains in heating experiments. (a) SEM image of an unheated, 235-μm-long grain of chromite ((Fe,Mg)Cr2O4). (b) At 1500°C, a 250-μm-long chromite grain embedded in the surface of laboratory-melted AH sediment showed minimal thermal alteration. (c) At 1600°C, a 165-μm-long chromite grain displayed moderate surface melting, but temperatures were insufficient to cause diffusion into AH sediment matrix. (d) At 1700°C, a 240μm-long chromite grain displayed moderate surface melting with limited diffusion into the matrix. (e) Photomicrograph of partially melted chromite grains, labeled Cr, after heating to 1700°C. Some euhedral edges remain visible. Laboratory-melted bulk sediment from level 435, sample ES15, 395 cm depth. Figure S19. Fe-rich inclusions (globules) in AH glass from heating experiments. Equipment used included a Spark Plasma Sintering (SPS) furnace capable of reaching 1850°C. Graphite crucibles were used to mimic the hypothesized impact environment at Abu Hureyra that was carbon-rich because of vaporized vegetation. As with previous experiments, ~1 g of AH sediment was heated in ~100°C steps from 1000° to 1850°C. (a) SEM image of Fe-rich globules on the surface of the laboratory-melted glass. The maximum temperature was ~1850°C for several minutes. SEM-EDS indicates a composition of native Fe (Fe 0 ), reduced iron (FeO) and iron silicide (Fe3Si), all of which reflect very lowƒO2 and do not normally exist in nature. (b)-(f) Single-element EDS maps with intensity scales showing variations in the abundances of Fe, Ca, Al, Si, and O. Note that oxygen in these blebs has very low abundance in blue-black areas, indicating native Fe and native Si in some cases. The laboratory-melted glass is enriched in Ca, Al, and Si, the same as excavated AH glass. Laboratory-melted bulk sediment from level 435, sample ES15, 395 cm depth.   The nuclear detonation occurred ~30 m above the ground with a TNT energy equivalent of ~21 kilotons, reaching an average plume temperature of 8000°C after three seconds 9 . The detonation formed a shallow crater 1.4 m deep and 80 m in diameter and melted the top 1-3 cm of the surface sediment, mostly composed of quartz, feldspar, muscovite, actinolite, and iron oxides. Molten material from the nuclear test, referred to as trinitite, fell back onto the surface up to distances of ~600 m, sometimes forming molten pools of glass. Some of the melted material was drawn into the rapidly rising plume and as the plume drifted north-eastward ~30 km from ground zero, trinitite rained out of the cloud as melted spherules and aerodynamically shaped glass. The Trinity ejecta includes irregularly shaped fragments, as well as melted teardrops, beads, and dumbbell shapes, many of which show collisional and accretional features. This evidence is morphologically similar to melted material recovered from Abu Hureyra.  Figure S25. Iron silicide in trinitite from Trinity atomic testing. (a) Transmitted light photomicrograph of trinitite glass droplets; the longest ones are ~2.5 mm long. The green color is due to small amounts of Fe. Note dark inclusions that usually are partially-to fully-melted hightemperature minerals. (b) Photomicrograph of two fused spherules of trinitite; largest is ~1400 µm in diameter. Arrow points to a dark globule of iron silicide, Fe2Si, that created a high-velocity impact crater with a raised rim. (c) SEM-EDS analyses indicate spherules and globules of three types of iron minerals: iron oxide (FeO) and two types of iron silicide (FeSi and Fe2Si), coexisting with native Si. The largest globule is ~50 µm wide. From the inner wall of trinitite vesicle. The ultra-high temperatures in the Trinity atomic bomb test caused thermal dissociation of the melt into the elemental species followed by condensation of the various phases. Because of fast-reaction kinetics, non-equilibrium conditions, and low ƒO2, highly reduced compounds (metals and silicides) formed side-by-side with oxidized magnetite. Analyses of this grain suggest instantaneously melting and instantaneous quenching after flash heating that reached temperatures above the boiling point of quartz at 2230°C. The process was so rapid that it did not allow for the complete incorporation of the molten quartz grains into the bulk melt. This melted quartz grain is nearly identical to melted quartz grains in AH glass, suggesting that the latter could have resulted from a similar formation process.

Supporting Information: Text Text S1. Previous Heating Experiments
Thy et al. 10 conducted heating experiments to estimate likely maximum temperatures for AH meltglass. Thermal analyses and modeling suggested maximum temperatures in the range of 1000-1200°C, rather than temperatures of >1700°C proposed by Bunch et al. 1 . Although this study found that ~1175°C is approximately the minimum temperature required to melt the local sediment, melting at minimum temperatures does not preclude reaching higher maximum temperatures. The presence of high-temperature, melted minerals, such as monazite, chromite, and chromferide, confirm that temperatures much higher than 1175°C under highly reducing conditions were reached during the melting of the glass. Thatched hut fires could have produced low-temperature glass at Abu Hureyra, but such fires could not have melted grains of quartz, monazite, chromite, chromferide, and suessite (1720° to 2300°C), as discussed in the main manuscript.

Text S2. Reflectance as a Temperature Indicator.
Reflectance Sample Analyses. See Methods below for preparation details. Reflectance was measured on fusinite (carbonized wood with clear cellular structure) in the case of charcoal particles, char (carbonized organic material of unknown origin with random degassing pores), and cell walls of vesicles (carbon spherule samples). In some cases, multiple reflectance populations were combined into a single reported measurement value (values commented as highly variable or multiple populations included). This was done where insufficient sample volume prevented a significant number of sample fragments from polishing into the examination surface, or where a dominant population was not clearly identified. Where sufficient numbers of fragments from a multi-population reflectance sample permitted the identification of the dominant population, that population was selected for the reported value. However, a second preparation of the same sample can result in a different dominant population occurring as the dominantly exposed fragments in the examination surface. Therefore, data from samples with multiple populations need to be interpreted cautiously.
General types of samples. Samples were divided into three types based on visual appearance: i) carbon spherules, ii) charcoal, and iii) char or amorphous carbon, as in sample ABU GLASS+CHAR, which does not fit into either category. See Appendix, Fig. S4-S6.
General type: Carbon spherules. These samples were subdivided into two categories: i) YDB carbon spherules, as extracted from sediment and ii) YDB carbon spherules that have been reheated to known temperatures. Carbon spherules generally have shapes that range from rounded to elongated to flattened ovoid. Sometimes, they are fragmented, consisting of a hardened, high relief rind surrounding a softer, lower relief interior. Generally, the carbon spherule rinds have higher reflectance than the carbon spherule interiors, reflecting lower inner temperatures. Extending from the rind inward are spherical to ovoid cells, usually filled with lower reflecting nonfluorescent material. Sometimes cells are empty or the lower reflecting cell-filling material contains an empty hole at its core. Proceeding further into the core of original carbon spherules, a collapsed spongy cellular structure is present, wherein individual open cells are poorly preserved.
Reheated carbon spherules are higher in reflectance than the original carbon spherules found in YDB sediment, and any retained cell filling material is of generally equivalent reflectance to cell walls. However, in general, most cells are open in reheated carbon spherule samples. The rinds of reheated carbon spherules contain more cell wall material than the interior of reheated spherules but rind reflectance and hardness are qualitatively identical to that of the interior material. Fe oxides occurring as discrete fragments also are present in some samples.
Carbon spherules are presumed to derive from most wildfires, including those associated with extraterrestrial impact events 11,12 . Alternately, they have been claimed to be modern fungal sclerotia and/or fecal pellets from insects 13,14 . Here, we discuss only the reflectance evidence; origin is not considered. The observations presented here are consistent with the prior descriptions given in the literature for carbon spherules, except that the occasional low-reflectance cell-filling material is reported here for the first time.
In an independent study, Scott et al. 13 and van Hoesel et al. 15 investigated YDB carbon spherules, such as those found at Abu Hureyra, which were inferred by Firestone et al. 11 to have formed at high temperatures during the impact event. Scott et al. 13 used charcoal reflectance, the same techniques used here 16 . Scott et al. 13 reported that all YDB carbon spherules show reflectance values indicating maximum temperatures of <450°C, thus precluding high-temperature impact fires. Similarly, van Hoesel et al. 15 reported charcoal reflectance for carbonaceous particles from the YDB layer in Europe. The results suggest maximum temperatures of ~420 ± 10°C, assuming a charring period of one hour, also precluding high-temperature impact fires. In contrast, Bunch et al. 1 reported charcoal and carbon spherules embedded in high-temperature melted glass from Abu Hureyra, Melrose (Pennsylvania), and Blackville (South Carolina), suggesting much higher temperatures for the carbon. These differences were explored by testing the hypothesis that, while reflectance is an adequate analytical technique for normal wildfires, it provides erroneous values for extremely brief, high-temperature events, such as a cosmic impact.
General type: Charcoal. These samples contain the maceral fusinite, which is the carbonized remains of charred wood with well-preserved cellular structure 17 . Charcoal samples are highly variable in reflectance, possibly indicating that incomplete or partial charring occurred in the same sample, or that heating was unevenly applied. Temperatures of combustion can be interpreted by a comparison of reflectance values to various experimentally derived chars [18][19][20] . However, given the high variability in reflectance present in the majority of samples evaluated herein, interpretations of combustion temperatures are preliminary and suspect. In addition, published calibrations of the reflectance of inertinite (a common organic component of coal and oil shale) to temperature are far from universally applied. Instead, they are restricted to isolated and specialized research projects and are apparently dependent on precursor material and charring time.
General type: Char. Char is high reflectance (1.3%) organic material with randomlyoriented, randomly-sized elongate ovoid to spherical degassing pores, as well as some dense regions lacking pores.
Sample: carbon in AH glass: ABU GLASS+CHAR contains 3 types of material: i) modern organic material, ii) char (as per Kwiecińska and Petersen 21 ), and iii) meltglass). Some observed organic material is a low reflectance (0.2%) humic gel containing foraminifera and sponge spicules with dispersed mineral fragments. This is most likely post-depositional contamination. The char in sample ABU GLASS+CHAR appears to derive from the burning of gelified organic material, such as peat (e.g., Petersen 22 ), possibly represented by the humic gel fragments also present in this sample. The AH glass sample also contains layered Fe-oxides, which are also present as discrete fragments. Taken at face value, the uneven distribution of degassing porosity and dense areas suggests that the char formed at relatively low temperatures. However, its association with high-temperature meltglass argues for a more complex heating history.
Sample: modern reeds, oak, and pine. This experiment was designed to replicate impact conditions, where a high-temperature combustion source is applied unidirectionally at a short duration. The combusted portion of the reed sample was converted to 96.3 wt.% gases, 0.9 wt.% ash, 0.4 wt.% charcoal, and 2.4 wt.% glassy spherules 23 . For fragments of oak and pine, ~97 wt.% was transformed into gases, <1 wt.% into charcoal and ash, and ~2 wt.% into spherules 23 . Ash and glassy spherules were produced from where the wood was exposed to the highest heat, grading into charred wood, and then into unburned wood farthest from the heat source. Charcoal samples were collected from the end of the section that remained after combustion. Very little of the material exposed to the flame survived (<4 wt.%), and even less survived as charcoal (<1 wt.%). Samples of reed (Phragmites australis), oak (Quercus turbinella), and pine (Pinus ponderosa) were heated using an oxygen/propylene torch. Portions of short sections of each material (~6 × 0.5 cm) were exposed to a direct flame for ~30 seconds at temperatures of ~1700 to ~2600°C, the highest temperature measurable by a thermocouple.
The Reed Char sample contains some modern un-charred wood (fluorescent) fragments as well as fragments that exhibit a transition from semifusinite to fusinite (less to more carbonized woody material in the same fragment, i.e., less reflective to more reflective). Reflectance measurements did not confirm the measured high temperatures of >1700°C for the reed, oak, or pine. Instead, %Ro values indicated average temperatures of ~632°, 631°, and 648°C, respectively, far below the measured temperatures (Appendix , Table S5).
Sample: trinitite. Reflectance was measured for a charred twig partially covered with trinitite, recovered from the blast zone of the Trinity atomic bomb test at Alamogordo Bombing Range, NM in 1945. The average plume temperature was ~8,000°C at 3 seconds 9 , falling after ~3 seconds to ~1720°C 8 , the melting point of quartz. In addition to fusinite, the Trinitite Lump Char sample contains some relatively low reflecting semifusinite with preserved cellular structure. These semifusinite fragments in Trinitite Lump Char contain inclusions of fluorescing liptinite coal macerals (possible remains of corky material) as well as areas of mesophase. In this sample, the presence of fluorescing liptinite material preserved in structured semifusinite immediately adjacent to what is interpreted as mesophase indicates a transient heating event applied to a modern wood sample. Preserved cell structure indicates no humification occurred prior to the carbonization event and the presence of fluorescing material indicates transient heat as well as low levels of postcharring oxidation. The mesophase areas indicate molecular order precipitating from a plastic phase (boiling fluids and remnant solids). Reflectance measurements did not yield the correct temperature of the trinitite (minimum temperature: 1250°C), but rather, average %Ro values yield a temperature of ~406°C (Appendix , Table S5).
Sample: activated carbon and charcoal feedstock. Reflectance was measured for samples of commercially-made activated carbon and samples of the charred feedstock used for making activated carbon, provided by Calgon Carbon Corporation. According to the company, the raw material (coconut shells) had been subjected to temperatures of ~450°C for ~8 hours under conditions that restricted but did not eliminate oxygen, similar to ambient atmospheres during a cosmic impact event. This carbon feedstock is then used to produce activated carbon by processing it with steam for ~8 hours at ≥1100°C to create anoxic conditions, as in a cosmic impact event). Reflectance-derived inferred temperatures closely matched the maximum temperature of the charred coconut (511°C; Appendix, Table S5). On the other hand, reflectance values for the activated carbon incorrectly indicated an average temperature of ~585°C instead of the recorded temperature of ≥1100°C (Appendix , Table S5).
Sample: YDB carbon spherules. We investigated whether reflectance measurements can determine the accurate temperatures for nanodiamond-rich carbon spherules that had been reheated to known temperatures measured with a thermocouple. Nanodiamond-rich carbon spherules from the YDB layer in Gainey, Michigan; Kimbel Bay, North Carolina; and Indian Creek, Montana were placed separately in a tube furnace filled with a CO2 atmosphere, ramped up for ~10 minutes to the maximum temperature of 650-850°C, and held there for ~5 min. The average %Ro for Gainey indicated 537°C, compared to a measured temperature of 730°C; the average %Ro for Kimbel Bay indicated 515°C, compared to a measured temperature of 700°C; and the average for Indian Creek was 567°C, compared to a measured temperature of 650°C (Appendix , Table S5). Thus, reflectance values are relatively close to but lower than measured temperatures for all three YDB carbon samples.
In similar reflectance experiments, Scott et al. 13 investigated carbon spherules from several YDB sites and found reflectance values of <2%Ro, consistent with charring temperatures of <450°C. Similarly, van Hoesel et al. 15 reported that charcoal particles from the YDB-age Usselo horizon show a reflectance of 0.96 ± 0.06% Ro, indicating a charring temperature of approximately 420 ± 10°C, assuming a charring period of 1 h. However, those workers reached their conclusions based on experimental charring of organic material for a duration of one hour or more. Guo and Bustin 24 found that the duration of heating for charcoal must be considered when inferring fire temperatures. The experimental conditions in those studies are much different than those in cosmic impacts, and so those results are inapplicable to high-temperature events of extremely short duration, ranging from a few seconds to a few minutes, as detailed in Schultz et al. 25 Sample: Tunguska charcoal. To compare a known impact fire to the wildfire charcoal at Abu Hureyra, we measured reflectance on charcoal from the Tunguska airburst, which felled 80 million trees 26 and triggered wildfires across 2150 km². The charcoal samples were extracted from the peaty impact layer that dates to 1908, the time of the airburst. The average reflectance-derived temperature for the Tunguska charcoal is 413°C. The original formation temperatures for the charcoal are unclear, and so, temperatures can only be inferred based on other lines of evidence. Nanodiamond-rich carbon found in nearby peat samples was estimated to have been exposed to temperatures from 700 to 1475°C, and the charcoal-rich layer contained Fe-rich spherules that melted at ~1500°C 23 . Thus, the charcoal from Tunguska most likely formed at ≥1500°C (Appendix , Table S5). Reflectance values suggest temperatures for Tunguska charcoal that are much lower than actual temperatures. 24 found that the duration of heating has a major effect on charcoal reflectance, and so, a key question is whether charring for one hour is comparable to charring for a few seconds to a few minutes. In one experiment with charring wood, Guo and Bustin 24 found that after exposure to 600°C for 40 minutes, reflectance values were ~8.5× higher than for 6 minutes. In other words, the lower value at 6 minutes erroneously indicated a heating temperature of ~350°C, instead of 600°C. Their results conclusively demonstrated that the duration of heating for charcoal must be considered when inferring fire temperatures.

Text S3. Possible Vapor Deposition.
Epitaxial films may grow from gaseous or liquid precursors in a process that deposits a crystalline layer over a crystalline substrate, where there is structural continuity between the overlayer and the substrate. However, here the process is only analogous and not epitaxial in the strictest sense because epitaxia entails growth of crystals on a crystal substrate, not deposition of amorphous glass on glass. Lechatelierite cannot be produced volcanically but is found in fulgurites, the tubular meltglass formed during a lightning strike 1,27 . No piece of AH glass investigated resembles the distinctive tubular shape of a fulgurite. Lechatelierite is also common during cosmic impact events, such as at Meteor Crater, AZ 28 , Haughton Crater, Canada 29 , Australasian tektites 30 , Libyan Desert glass 30 , and Dakhleh glass 31 .

Text S4. Oxygen fugacity (oxygen deficiency).
Low oxygen fugacity (ƒO2) occurs within an airburst/impact fireball 32 . Native Fe spherules within trinitite glass indicate that regions of extremely low ƒO2 (within the metal) coexisted with high ƒO2 (within the glass matrix) over very short distances. Native Fe is common in extraterrestrial material but extremely rare in terrestrial rocks, mostly limited to trinitite and to rocks in contact with underground coal fires 1 , which are not known to have occurred near Abu Hureyra. In the absence of sufficient oxygen, Fe sometimes combined with silica (yielding silicides, e.g., Fe3Si), sulfur (yielding sulfides, e.g., FeS), carbon (yielding carbides, Fe3C), or phosphorus (yielding phosphides, Fe3P). Several variants of these minerals are known only from meteorites, and others were discovered in meteorites first, and only later found at a terrestrial location Such low-ƒO2 minerals have been documented in YDB meltglass from Blackville, South Carolina, and Melrose, Pennsylvania 1 , in trinitite 1 , and now, in meltglass from Abu Hureyra.
One question is how these oxygen-reduced, high ƒO2 Fe and Ni phases formed in the highly restricted space inside AH glass vesicles. Carbon concentrations in some fragments of AH glass are estimated to have reached 15 to 25 wt.%, based on EDS spectral peak heights, and therefore, it is proposed that the high carbon content in the high-SiO2 glass yielded highly reducing local conditions that promoted the formation of native Fe and NiFe, rather than magnetite. AH glass and spherules probably resulted from nearly instantaneous melting of both carbonate-rich sediment and significant amounts of carbon-rich biomass and the vesicles in AH glass trapped carbon in both solid and gaseous states, with carbon acting as a reducing agent. Oxygen fugacities were highly variable, some vesicles, contain coexisting oxidized Ni-bearing magnetite and native Fe and NiFe metals within sub-millimeter distances of each other.
The likely formation mechanism for awaruite is different. These grains occur on the outside of spherules and AH glass, not within vesicles, and thus, were not produced in a closed system. In this case, the NiFe-bearing spherule melts were also subjected to changing atmospheric conditions and low ƒO2 in the hypervelocity impact cloud. The morphologies and compositions of some AH glass fragments imply that they formed in the impact cloud similar to the way that trinitite formed in the ground-hugging base surge produced by atomic bomb tests. Other AH glass fragments resemble trinitite droplets that were drawn up into the rising atomic bomb plume and then fell out as melt droplets. The spiral morphology of some AH glass droplets is very similar to some trinitite droplets that show evidence of cooling while rapidly rotating in the impact cloud.
Some of the Ni-rich material may contain a small percentage (<1%) of Ni from the impactor. Pechersky et al. 33 , who analyzed samples from 25 different types of meteorites, found that metallic inclusions in meteorites clustered in three different groups: i) pure or native iron, ii) kamacite containing 3-6% Ni, and iii) taenite containing ~50% Ni. AH glass displays the same three groups, in which Fe and Ni concentrations overlap or are close to concentrations in meteorites, suggesting that the Ni in AH glass may derive from meteoritic material, most likely during an airburst, possibly by a small dust-rich comet or a rubble-pile asteroid 34 .
In summary, the temperatures involved in forming AH glass are important, but more important is the rapid, non-equilibrium cooling under highly variable ƒO2. Such fast-reaction, isentropic conditions (i.e., having equal entropy) are known to exist only in lunar materials, fulgurites, aerial nuclear detonations, and cosmic impact events. Only the latter is a plausible source for AH glass.

Text S5. Transmission FTIR and water content.
Other results. A layered sample from the Australasian tektite field (Muong Nong, Laos) showed H2O content ranging from 182-227 ppm (n=6), which falls within the published FTIR range of 90-300 ppm 35 . Another example, a splash-form Australasian tektite, had a low H2O content ranging from 78-89 ppm (n=7), falling within with the published range of 40-120 ppm for various splash-form tektites 35 . Both tektite glass samples analyzed yielded results within the range of water content in other tektites from various fields (~20 to 500 ppm) 35 .
A sample of known cosmic impact glass from the Darwin crater in Western Tasmania, Australia had an H2O concentration of 611 ppm (n=7), similar to the previously published value of 470 ppm (n=2) 36 . One sample of impact glass from the Zhamanshin crater in western Kazakhstan yielded a range of 705-1036 ppm (n=10), which is higher than the reported range of 50-630 ppm for the crater 35 but overlaps published values for H2O in various impact glasses (80-1300 ppm). It also overlaps previously reported Zhamanshin glass values of ≥1000 ppm 37,38 .
FTIR analysis of a sample of trinitite from the Trinity airburst at the Alamogordo Bombing Range, New Mexico yielded water contents of 283-510 ppm, somewhat higher than previous reports of 70-100 ppm 39 .
Lightning-produced glass (fulgurites) from near Socorro, New Mexico yielded an H2O concentration of 159 ppm (n=7), which is lower than some published values of 500-1400 ppm for some fulgurites 37,40 .
Chaiten volcanic glass samples (rhyolite) from Chile had an H2O content of 1497 to 1769 ppm, consistent with previous values of 1200 ppm reported for this volcano 41,42 . These values are in the range of H2O contents (500-4000 ppm) reported for various other volcanic glasses [43][44][45][46][47] , excluding volcanic glass samples that appear to have absorbed additional water after cooling.
Biomass glass, a natural glassy slag from Botswana and other sites in Africa, was not analyzed but has been reported to have an H2O concentration of 1000 to 9100 ppm 48,49 . Medieval anthropogenic glass from an archaeological site has been reported to contain from 500 ppm to 7,600 ppm H2O 50 . Other examples of modern human-produced glasses range from ~1000 ppm to 12 wt.% H2O 51,52 .

Text S6. Remanent Magnetism.
Results for other YDB sites. One sample from Blackville, South Carolina displayed magnetization along one vector that was moderately higher than the normal geomagnetic field. This result is inconsistent with normal terrestrial rocks but consistent with impact-related material, possibly due to thermal and mechanical shock, as seen in laboratory remanent magnetism experiments 53 . The Melrose samples fall into 2 groups; the first with 3 samples showed low levels of initial magnetization, meaning that this group could not have formed by lightning. This group had soft magnetic carriers and showed dramatic, continuous changes of the magnetization vectors, consistent with rotation of the meltglass during cooling, as could have occurred during either a surface impact or a cosmic airburst that lofted molten material into the air. The second group of 2 samples showed maximum levels of magnetization, consistent with having been melted either by natural lightning strikes or by impact-induced lightning 54 , which occurs during both volcanic eruptions and cosmic impacts. Thus, two of the Melrose samples of meltglass could have formed by lightning, but three other Melrose samples could not have formed that way.
Text S7. Discussion of potential meltglass formation mechanisms Building fires. Thy et al. 10 proposed that thatched hut fires created AH glass at temperatures in the range of 1100-1200°C. However, based on our laboratory heating experience, we conclude that the presence of melted chromite and monazite in AH glass indicates flux-adjusted temperatures of ~1872-2065°C, well above the highest known temperatures for building fires. For comparison, the phosphorus-induced firebombing of Dresden, Germany, during World War II produced updrafts of ~275 km/h, forming fire tornados that produced maximum temperatures of 1000°C 55 , sufficient to soften but not melt glass and too low to melt iron structural materials. It seems unlikely that small thatched huts could burn at temperatures as high as modern buildings, and therefore, the AH huts are unlikely to have reached temperatures of ~1200°C.
Thy et al. 10 also investigated meltglass fragments at other excavated archaeological sites in northern Syria, including Mureybet, ~32 km north of Abu Hureyra, and Jerf el Ahmar, ~76 km north, from which the lead author kindly provided samples for this study. Our analyses by light microscopy show that they are morphologically identical to those from Abu Hureyra, displaying both plant imprints and aerodynamic shaping. However, our SEM-EDS analyses of only a few available glass samples identified no high-temperature melted minerals in the samples, precluding a direct comparison with AH glass. Thy et al. 10 did not report sedimentary concentrations of meltglass, so it is unknown whether or not the meltglass is rare or common at those sites. Thus, those investigators may be correct that meltglass from these other sites formed during hut fires, unlike AH glass, and therefore, lacks high-temperature melted minerals.
The sediments investigated by Thy et al. 10 span 1500 years, a wide range that led them to conclude that the presence of meltglass at the sites in northern Syria is inconsistent with a single cosmic impact. However, Bunch et al. 1 and this study found very small amounts of AH glass in Holocene sediments above the YDB layer and attribute this "young" meltglass not to additional impacts but rather to the reworking (redeposition) of deeper YDB-age sediment. In support of that, Moore et al. 5 documented extensive reworking of 12,800-year-old meltglass as the most likely explanation for apparently younger meltglass.
One interesting possibility deserves future study. The question is whether the meltglass at Mureybet and Jerf el Ahmar was created at the time of the YDB impact event and then was later reworked into younger sediment. If so, YDB-age meltglass covers a much wider regional area.
Biomass or "haystack" fires. Thy et al. 48 reported that biomass glass or slag is sometimes found in midden piles of prehistoric settlements in Africa with estimated formation temperatures of 1155-1290°C, well below the melting points of high-temperature minerals in age glass. The lead author kindly provided us with samples of biomass glass, which we found to be morphologically different from AH glass and to contain no high-temperature minerals, making it easily distinguishable.
The lead author of Thy et al. 10 provided samples of biomass slag from Africa that are morphologically somewhat similar to AH glass but typically rougher and more heterogeneous. Analyses of the slag using SEM-EDS found that its composition is somewhat different from AH glass; the main component is SiO2 at 66.9 wt.% for biomass slag compared to 50.9 wt.% for Abu Hureyra. Investigations of the outer surfaces of biomass slag samples and interior material on sectioned slides revealed no melted, high-temperature mineral grains like those on AH glass. Instead, there are only low-temperature melted grains, including plagioclase and feldspar with melting points of ~1200°C, consistent with a temperature range of 1155-1290°C, as estimated for biomass glass in Thy et al. 48 . The biomass glass investigated formed at much lower temperatures than AH glass. In addition, biomass glass typically cools slowly, so that the pieces of glass we analyzed show no flow marks that are evident in AH glass.
Anthropogenic contamination. AH glass production by Abu Hureyra villagers can be ruled out because they were unable to achieve the requisite temperatures. Pottery-making began ~14,000 years ago, but maximum temperatures were <1050°C 48 ; copper smelting began ~7000 years ago but only reached temperatures of ~1100°C 1 ; glass-making at ~5000 years ago only reached ~1100°C 1 . These temperatures are too low to have melted quartz, chromite, and/or zircon grains. In addition, contamination from modern human activities can be eliminated because AH glass was buried nearly 4 m below the modern surface and sealed beneath undisturbed living floors.
Coal-seam fires. Subsurface burning of coal seams can produce a glasslike material, called clinker, or scoria 1 , at low-pressure formation temperatures of 1000-1400°C. However, these temperatures are too low to melt zircon or chromite, as observed at Abu Hureyra. In addition, no coal deposits are known near Abu Hureyra; the nearest are in Turkey.
Lightning-induced melting. Temperatures in fulgurites can far exceed 1720°C, hot enough to fully melt quartz, making lightning a potential source of AH glass. However, magnetic measurements, as discussed in the section below, indicate that lightning did not produce AH glass, which lacks the high remanent magnetism characteristic of fulgurites. In addition, meltglass containing high-temperature minerals are found concentrated at Abu Hureyra in 12,800-year-old sediment across Trenches D, E, and G 5 , while almost none are found above and below. This evidence is widely distributed. Trench E is ~122 m from Trench D and ~110 m from Trench G; Trenches D and G are ~175 m apart, an area of about 6700 m 2 , or ~0.67 hectares.

Text S8. Astronomical Environment
Astronomical discoveries over the last few decades demonstrate that the mass distribution of comets is biased towards larger bodies, up to 300 km in diameter. A single giant comet (diameter ≳100 km) may contain >100 times the mass of all the asteroids that currently threaten Earth, and if thrown into a short-period, Earth-crossing orbit, it would disintegrate, like most comets do 56 . The hierarchic fragmentation of a large comet in a short-period orbit may yield many hundreds of short-lived debris streams comprising dust, boulders, and cometary fragments, expanding along the orbital track. Over the lifetime of the comet, such co-moving fragments constitute a significant terrestrial hazard 57,58 .
Such large comets drift into the near-Earth environment frequently in relatively short geological timescales. In fact, the fragmented remains of two such bodies are present in the inner Solar System today. One of them, the Taurid Complex, is composed of debris from an ~100-kmwide comet that arrived at least 20,000 to 30,000 years ago from the Centaur system of large comets, after which it disintegrated hierarchically into short-period, Earth-crossing orbits 59,60 . One or two intersections with such material, sufficiently massive to yield YDB-like catastrophes, are reasonably probable over the course of the 20,000-year-long breakup of this Taurid progenitor, thus providing a plausible impact mechanism for the YDB Impact Hypothesis.

METHODS
SEM-EDS analyses used a JEOL JSM 6010PLUS/LA at Elizabeth City State University and a Hitachi S3200N variable pressure scanning electron microscope (VPSEM) at North Carolina State University. All SEM imagery was acquired at a resolution of 2560 × 1920 pixels. Images were uniformly post-processed for contrast and brightness, if necessary, using Adobe Photoshop CC2014. SEM-EDS images for all analytical tests were uniformly post-processed for contrast and brightness, if necessary, using Adobe Photoshop CC2014. Colorized phase maps were manually constructed with Adobe Photoshop CC2014, based on multiple single-element EDS maps with rainbow-colored intensity scales. M.A.L. and A.V.A. performed the analyses.
Reflectance. Samples were received into the laboratory in glass vials. Eight of seventeen samples were prepared in 1-inch thermoplastic mounts heated at 360 °F and 4000 psi pressure for 10 minutes, with no additional processing. The examination surfaces were ground and polished following ASTM D2797 (ASTM 2012a). Nine low-volume samples were poured into pre-drilled holes in otherwise identical 1-inch molds, mixed with thermoplastic binder powder and mounted at the conditions given above. Examination surfaces of the low volume samples were slabbed with a wafer saw prior to grinding and polishing via ASTM D2797. All samples were desiccated overnight prior to reflectance measurements.
Select sample mounts were mapped using a Leica DM4000 microscope equipped with customized LED illumination and image mosaic software from Hilgers, Inc. A Leica DMRX Pol microscope equipped with a J&M photomultiplier (PMT) and Zeiss MRc digital camera was used for reflectance analysis and imaging. Reflectance was determined according to ASTM D7708 61 . Reflectance values were checked on the Hilgers system, which employs a camera as the detector. Reflectance measurements were calibrated using a K&B cubic zirconia standard (3.13%; used with PMT system) and a K&B glass standard (1.31%; used with a camera system).
Samples were examined dry at 50x with white LED illumination for mapping and at 500x under oil immersion with tungsten halogen incident light (PMT system) for reflectance and with white and blue LED (camera system). Sample and material identifications are noted in the photographs. P.H. performed all analyses and interpreted the results.
Transmission FTIR and water content. The glasses were prepared as doubly polished wafers for analysis with a Nicolet iN10 MX infrared imaging microscope with an attached liquid-N2-cooled MCT-B detector at the USGS in Menlo Park, CA USA. The iN10 MX is purged with low-CO2, dry air and is equipped with a collar that, for transmission experiments, can be lowered around the sample stage to maximize purge during sample and reference collection 41 . A reference spectrum (R) was collected away from the hydrous glass sample through the BaF2 window, transparent to both visible and infrared light. The sample surface was brought into focus with reflected light prior to the collection of the sample spectrum (S). Sufficient scans were collected to minimize noise (usually 256) at 4 cm -1 resolution. A square aperture was used to define the precise area of analysis, which could be sized as small as 20µm to avoid bubbles. Measured absorbance (a) was calculated as follows.
A = log (IR/IS) [Eq. 1] where IR = the radiation transmitted in the reference spectrum and IS = radiation transmitted through the sample plus reference.
We analyzed the 3570 cm -1 peak, which measures total water (OH-and molecular water; i.e., H2Om) 41 . The samples had very low H2O, and in such cases for silicate glass, the water is expected to be found entirely as OH-. The peak height for the principle OHstretching band (located near 3600 cm -1 ) was measured from a flat baseline to the maximum peak height. Concentrations were calculated assuming Beer's Law: wt.% H2O = 1802A/(εdρ) [Eq. 2] where A is the absorption for the peak of interest, ε is the extinction coefficient for the peak in L mol-cm -1 , ρ is the density in g L -1 , and d is the thickness of the doubly polished sample in cm. An ε value of 75 was chosen as reasonable for high-silica tektite glasses 35 , though H2O is quantified with ε ranging from 63 (basaltic glass, Dixon et al., 1988) to 150 (cf. Paterson 62 ) to provide a maximum possible range of H2O contents within the sample suite). All glass densities were assumed to be 2350 g cm -3 (anhydrous rhyolite), though a few of the more mafic samples may have had densities 10-15% greater. An increase in the assumed density would result in a proportional decrease in the final calculated H2O concentration. Wafer thicknesses were measured with a digital pin micrometer. Jake Lowenstern, USGS, Menlo Park, CA performed all analyses and interpreted the results.
Electron microprobe analyses were conducted using a JEOL 8900 at the U.S. Geological Survey in Menlo Park, California (USA). Concentrations in samples and standards are reported as oxides, except for Cl. Each spot was analyzed two times; once for major elements (Si, Al, Ca, K, Na, Mg, and Fe) and once for minor elements (F, Cl, Mg, P, Mn, S, Ti). The first analysis used a 10-nA, 2-µm diameter beam, and count times of 20 seconds for all elements except for 10 seconds on easily volatilized Na. Standards included a variety of glasses (VG2, RLS-132), minerals (Tiburon albite, OR1, sodalite, barite,) and simple elemental oxides (TiO2, Mn2O3) in standard use at the Menlo Park facility 47 . Jake Lowenstern, USGS, Menlo Park, CA performed all analyses and interpreted the results. Tables  Table S1. Fourteen Abu Hureyra radiocarbon dates from Moore et al. 5 were used to develop a Bayesian age-depth model by Kennett et al. 4 . The most abundant YDB Abu Hureyra meltglass was found in level 445 (green), directly dated to 12,933 ± 68 cal BP (UCIAMS-105429, 11,070 ± 40 14 C BP) with smaller amounts of glass in the other green-highlighted levels with a Bayesianmodeled calibrated radiocarbon age of 12,825 ± 55 cal BP at 68% Confidence Interval (roughly equivalent to one sigma standard deviation), overlapping the previously published YDB age of 12,835 to 12735 cal BP for ~40 sites 4 . The green-highlighted age overlaps the published YDB age range of 12,835 to 12,735 4 , indicating that identification of the YDB layer is robust. Anomalously old and young dates were excluded by Bayesian analysis. Dates based on humic fractions of bones were considered unreliable and were excluded 5 . "Extracted" column lists field numbers of samples. "Elevations" for each level (meters above sea level) vary across the trench and are averaged.     Table S3. Representative SEM-EDS analyses of glass from Abu Hureyra. Shows keyed ms figure #s, elemental abundances, and total wt.%. Dashes indicate not present or not measured.   Table S5. Time and temperature for furnace experiments. "Sample" = # assigned by Moore et al. 2000. "Set T" = experimental target temperatures. "Insertion T" = actual temperature at start of experiment. "Overshot T" = actual maximum temperature of experiment. "Time to target T" = ramp-up to maximum temperature. Table S7. Elemental results of Abu Hureyra sediment and magnetic grains, using INAA, fire assay, and ICP-MS (ActLabs). Samples with the suffix of "-m" represents magnetic fractions; "glass" represent pieces of excavated AH glass; other samples are AH bulk sediment. Depths and abundances of Abu Hureyra impact proxies: cobalt = Co; chromium = Cr; nickel = Ni; and iron = Fe, with values in ppm. Palladium = Pd; platinum = Pt; and gold =Au, with values in ppb. Dark green highlighting indicates peak concentrations in the YDB layer, sample E301 at 405 cm.