The most evident component of the distal and regional debris layers is spherical particles, some of which are large enough to be seen with the naked eye. Due to their spherical shape it is assumed that they are part of the melt debris from the impact or the condensed vapor from the impact (Johnson and Melosh, 2012b, 2014). The particles are not thought to have melted on reentry into the atmosphere since debris launched above the atmosphere by the impact should not reach high enough velocities to melt when it reenters the atmosphere. According to Bohor and Glass (1995), there are two types of spherules, with differing composition and distribution. They identify Type 1 splash-form spherules (tektites or microtektites) that occur in the melt-ejecta (basal or lower) layer of the regional debris layer where it has a two-layered structure. These spherules are found as far from the Chicxulub site as Wyoming, but generally do not extend beyond about 4000 km away from Chicxulub. While the Type 1 particles are derived from silicic rocks, they are also mixed with sulfur-rich carbonates from the upper sediments in the Yucatan. The Type 1 spherules are poor in Ni and Ir, and the lower layer is poor in shocked quartz, consistent with their origin from the lower energy impact ejecta from the crater. Generally, the debris layer within about 4000 km of the crater is almost entirely composed of target material, rather than material from the impactor itself. Type 2 spherules, on the other hand, are found in the distal debris layer, and presumably formed primarily from the condensation of rock vapor from the impactor and target (O'Keefe and Ahrens, 1982; Johnson and Melosh, 2012b). There are subtypes of Type 2 spherules that correspond to varying composition of the original source material. Type 2 spherules occur in the upper layer in impact sites near Chicxulub, which merges into the fireball layer at distal sites. The Type 2 spherules are rich in Ni and Ir, while the fireball layer is rich in shocked quartz.

The formation of the spherical particles may depend on two different processes. Melosh and Vickery (1991) describe one formation mechanism, probably occurring in less heavily shocked portions of the target, when molten material decompresses until it reaches a critical line at which it starts to boil. The gas drag from the rock vapor on the molten rock spheres then tears apart the molten material, just as water droplets break apart when they fall through air. The relative velocities of water drops in air and the melt in vapor are similar, as are the surface tensions. As a result, melt droplets are similar in size to drizzle drops in light rain near 250 µm. According to Johnson and Melosh (2012b), these spherical particles are most likely to be found within 4000 km of the impact site, and to be chemically related to the target material, and not to the impactor. Such materials are reported across North America as Type 1 spherules (Bohor et al., 1987), sometimes referred to as microtektites. Since these spherules are not global, they likely were not as relevant to climate as the Type 2 spherules.

Melt droplets can also form in heavily shocked parts of the impact debris as rock vapor condenses to form melt in the fireball, which rises thousands of kilometers above the Earth's surface. These melt droplets form the Type 2 spherules. O'Keefe and Ahrens (1982) first modeled this process, and deduced that particles near a few hundred microns in size would form, as is observed. They also pointed out that the size of the spheres would be proportional to the size of the impactor. Johnson and Melosh (2012b) recently reconsidered this process for forming melt particles. They point out that the large spherules contain iridium (e.g., Smit, 1999), which is consistent with them being composed partially of the vaporized impactor. Their model of the formation and distribution of these particles suggests the particles have a size that varies spatially over the plume. Averaging over the simulated plume yields a mean size of 217 µm with a standard deviation of about 47 µm for a 10 km diameter impactor hitting at 21 km s-1. From the two examples given by Johnson and Melosh (2012b) it appears that the standard deviation is consistently 22 % of the mean radius for asteroids of different sizes. The initial values for the various properties of Type 2 spherules described above are summarized in Table 1 for the K-Pg impactor.

Smit (1999), who refers to the Type 2 spherules in the distal layer as microkrystites, estimated that these particles typically have a diameter near 250 µm, and a surface concentration of about 20 000 particles cm-2 over the Earth. Unfortunately, we are not aware of studies that measure the dispersion of the size distribution, or the spatial variation of the abundance of these particles. We assume that the particles have the density of CM2 asteroids, since Cr isotope ratios suggest that is the composition of the K-Pg impactor (Trinquier et al., 2006). Assuming this density, ∼ 2.7 g cm-3, the mass of spherules per unit area of the Earth is about 0.4 g cm-2, and the initial optical depth is about 20, as noted in Table 1. These spherules compose about half of the mass of the distal layer. We assume the particles were initially distributed uniformly around the globe, with the initial mixing ratio in the atmosphere varying only in altitude. Some theoretical studies, such as Kring and Durda (2002) and Morgan et al. (2013), suggest that these particles were not uniformly deposited in latitude and longitude, but had focusing points such as the antipodes of the impact site. Unfortunately, we are not aware of quantitative data on the global distribution of the spherules. The study by Morgan et al. (2013) may also be more applicable to the Type 1 spherules since their numerical model does not produce vaporized material from the asteroid impact.

According to the simulations of Goldin and Melosh (2009), the infalling spherical particles reached terminal fall velocity near 70 km altitude, at which point they begin to behave like individual airborne particles. Kalashnikova et al. (2000) investigated incoming micrometeorites in the present atmosphere, which generally ablate near 85 km. Kalashnikova et al. (2000) found material entering from space stops in the atmosphere after it encountered a mass of air approximately equal to its own mass. Therefore, the altitude distribution is taken to be Gaussian, centered at 70 km and with a half width of one atmospheric scale height (about 6.6 km based on the US standard atmosphere). A scale height is chosen as the half width of the injection profile since it is a natural measure of the density of the atmosphere. Figure 1 illustrates the vertical injection profile of the spherules (green curve). As discussed below, we expect several materials with origins similar to those of the spherules to be injected in this same altitude range, but others with origins unrelated to the impact generated plume, such as soot from fires, to be injected at lower altitudes.

The 70 km injection altitude refers to the level at which the large spherical particles reached terminal velocity. However, as is evident from the optical depth, many spherules entered through the same air mass. The column mass of the distal layer is ∼ 1 g cm-2 so the air pressure needs to about 1 hPa for the air mass above the altitude in question and the particle mass to be comparable. A pressure of 1 hPa occurs at about 48 km. Therefore, if the entire distal layer mass is placed into a model above 48 km, its mass mixing ratio will be greater than 1, and the atmosphere will be significantly out of hydrostatic balance. We are not aware of any simulations of the first few hours after the impact, but significant turbulence and mixing must have occurred as the atmosphere adjusted to the large mass imbalance. Model initialization should be checked to determine if the planned simulations start out of hydrostatic balance. If so, the injection altitude should be lowered below 70 km.

The energy release from the reentry of the large spherical particles into the atmosphere was likely responsible for setting most of the aboveground terrestrial biosphere on fire. However, due to their size, the spherules could not have remained in the atmosphere for more than a few days. Hence, they likely did not have a significant direct impact on the climate, but fell to Earth like a gentle rain.

Type 1 spherules, melt droplets, will form from impacts by 1 km diameter asteroids, and produce millimeter-sized particles in the ejecta curtain layer located near the crater (Johnson and Melosh, 2014). We do not expect an impact by a 1 km diameter asteroid to create a global layer of Type 2 spherules (Toon et al., 1997). Like O'Keefe and Ahrens (1982), Johnson and Melosh (2012b) conclude that the particle size will vary in proportion to the impactor diameter and the impactor velocity. For a 1 km diameter impactor hitting the land at 20 km s-1, they suggest that the mean diameter of the spherical particles will be about 15 µm, with somewhat larger sizes as the impact velocity increases to 30 km s-1. Table 3 provides our assumed properties of the spherules from a hypothetical 1 km diameter impactor hitting the land. It is likely that spherules would not be distributed over all of the globe for the 1 km diameter impact. Johnson and Melosh (2012a) as well as Glass and Simonson (2012) report a spherule layer associated with the Popigai impact in the late Eocene which Johnson and Bowling (2014) suggest was global in extent. This layer contains spherules similar in size or even larger than those associated with the Chicxulub impact. However, this layer is only about 10 % as thick as the distal layer from the Chicxulub impact. A 1 km impactor hitting the deep oceans may not produce a layer of spherules.

Spherical soot (also referred to as black carbon, or elemental carbon) particles were discovered in the boundary layer debris at sites including Denmark, Italy, Spain, Austria, Tunisia, Turkmenistan, the United States, and New Zealand (among others) by Wolbach et al. (1985, 1988, 1990a, b). Soot was also found in anaerobic deep-sea cores from the mid-Pacific (Wolbach et al., 2003). Soot was apparently lost by oxidation in aerobic deep-water sites in the 66 million years since emplacement. There is debate about whether these particles originated from global wildfires, or from the impact itself (Belcher et al., 2003, 2004, 2005, 2009; Belcher, 2009; Harvey et al., 2008; Robertson et al., 2013a; Pierazzo and Artemieva, 2012; Premović, 2012; Morgan et al., 2013; Kaiho et al., 2016). Robertson et al. (2013a), Pierazzo and Artemieva (2012), Premović (2012) and Morgan et al. (2013) argue that it is implausible that there was enough carbon at the impact site to produce the amount of soot observed by Wolbach et al. (1988). This debate about the origin of the particles does not greatly affect the impact these particles would have had on the climate when they were suspended in the atmosphere. The particles are small and widely distributed. They are numerous and so must have produced a very large optical depth and, being composed of carbon, they would have been excellent absorbers of sunlight. Whether the soot particles originated from global fires and were deposited in the upper troposphere, or they originated at the impact site and were deposited in the mesosphere, the climate effect of the observed soot would have been very great. Some have suggested that the soot resulted from wildfires in dead and dying trees that occurred well after the impact. However, Wolbach et al. (1988, 1990b) show that soot and iridium are tightly correlated and collocated. Indeed, Wolbach et al. (1990b) suggest the soot and iridium may have coagulated in the atmosphere. The soot and iridium in the distal layer must have been deposited within a few years of the impact, since small particles will not stay in the air much longer. Therefore, any fires must have been very close in time to the impact, and were likely contemporaneous.

Wolbach et al. (1988) estimated the global mass of elemental carbon (including aciniform soot, charcoal and any unreactive aromatic kerogen) in the debris layer as 7 ± 4 × 104 Tg of C or equivalently 13 ± 7 mg C cm-2 based on data from five sites. Wolbach et al. (1990b) updated these mass determinations to 5.6 ± 1.5 × 104 Tg or 11 ± 3 mg C cm-2 based on data from 11 sites. This mass of elemental carbon would require that the bulk of the aboveground biomass burned and was partially converted to elemental carbon with an efficiency of about 3 %, assuming the biomass is 1.5 g C cm-2 of aboveground, dry organic mass per square centimeter over the land area of Earth. This biomass density is typical of current tropical forests. This inferred 3 % emission factor is about 60 times greater than that suggested by Andreae and Merlet (2001) for current wildfires, but agrees with laboratory and other observations from burning wood under conditions consistent with mass fires (Crutzen et al., 1984; Turco et al., 1990). Mass fires are more intense than forest fires, and consume all the fuel available, possibly including that in the near-surface soil. Ivany and Salawitch (1993) argued independently from oceanic carbon isotope ratios that at least 25 % of the aboveground biomass must have burned at the K-Pg boundary.

Wolbach et al. (1990b) distinguish several forms of elemental carbon. Aciniform carbon is composed of grape-like clusters of 0.01 to 0.1 µm spherules. On average, this type of soot is 26.6 % of the elemental carbon, yielding a global mass abundance of 1.5 × 104 Tg of aciniform carbon. Charcoal is estimated at 3.3 to 4.1 × 104 Tg, and unreactive kerogen at 0 to 0.8 × 104 Tg. Wolbach et al. (2003) discuss a data set from the mid-Pacific that suggests aciniform soot is 9 × 103 Tg, and charcoal is also 9 × 103 Tg. Wolbach et al. directly measure the carbon content of their samples. The aciniform soot to charcoal ratio is determined by using an electron microscope to distinguish small and large particles.

There are several uncertainties in determining the amount of soot to use in a model. An upper limit of the amount injected into the stratosphere is 7.1 × 104 Tg based on the upper error bar of the Wolbach et al. (1990b) elemental carbon values. An important assumption in this upper limit is that the larger particles found by Wolbach et al. (1990b), are either aggregates of smaller ones, or of the same general size as the aggregates of the smaller ones that occur after coagulation. A lower limit of 1.1 × 104 Tg is obtained using the lower error bar of the elemental carbon from Wolbach et al. (1990b), and assuming 26.6 % is aciniform soot. Alternatively, one could argue that this lower limit of aciniform soot should be injected into the stratosphere, along with 3.3 × 104 Tg of charcoal using different size distributions. The most likely value of the aciniform soot in the stratosphere is 1.5 × 104 Tg, and of elemental carbon 5.6 × 104 Tg. We use these most likely values in Table 1.

Kaiho et al. (2016) argue that the soot came from burning hydrocarbons in the crater and that the total mass emitted was either 5 × 102, 15 × 102, or 26 × 102 Tg. If we reduce these values by the authors' factor of 2.6 to represent the stratospheric emissions, they are 0.4, 1.0, and 1.7 % of the globally distributed elemental carbon reported by Wolbach et al. (1990b).

Kaiho et al. (2016) measured several polycyclic aromatic hydrocarbons (PAHs) that are minor components of soot from one distal site in Caravaca de la Cruz, Spain, and another site at Beloc, Haiti, that is about 700 km from the crater. Since the PAHs measured are minor constituents of soot, Kaiho et al. (2016) need to use a large correction factor to determine the amount of soot. They first multiply by factors of 2, 5.9, or 10 to account for possible loss of PAH concentrations over time. They present no data to justify these factors. They then multiply by 3.3 × 103 citing this as the ratio of their measured PAHs to soot in diesel soot. No error bars were presented for this factor, and no values were given for the ratio in biomass soot. The origin of this correction factor is not evident in the cited reference. They then multiplied by another factor of 2.6 to represent the fraction of their soot estimate that they suspect reached the stratosphere. Their overall correction factors were therefore 17 × 103, 50 × 103, and 86 × 103. Given these large correction factors, and the lack of information about their uncertainty, it is difficult to compare them with the direct determinations done by Wolbach et al. (1990b), which do not require any correction factors.

As noted in Table 1, the mass of soot found by Wolbach et al. (1988) would produce an optical depth near 100 if the particles coagulated to spheres with a radius of 1 µm while they were in the atmosphere. Toon et al. (1997) pointed out that soot clouds with such a large optical depth would reduce light levels at the Earth's surface effectively to zero. The optical and chemical evolution of the particles once in the atmosphere may be influenced by the presence of liquid organics on the soot particles. Bare soot particles coagulate into chains and sheets, while particles that are coated by liquids may form balls. Chains, sheets, and coated balls have very different optical properties than spheres (Wolf and Toon, 2010; Ackerman and Toon, 1981; Bond and Bergstrom, 2006; Mikhailov et al., 2006). Particulate organic matter can be absorbing, and soot coated with organics can have enhanced absorption relative to soot that is uncoated (Lack et al., 2012; Mikhailov et al., 2006). These fractal shapes and organic coatings might not be preserved in samples in the distal layer since all the particles have been consolidated in a layer, and even in the current atmosphere the organics have short lifetimes due to rapid oxidation.

Wolbach et al. (1985) fit the size of the particles they observed, after exposing them to ultrasound to break up agglomerates, to a lognormal size distribution, described by dNdln⁡r=Ntln⁡σ2πexp⁡-ln⁡2rrm/2ln⁡2σ. Here, r is the particle radius, Nt is the total number of particles per unit volume of air, rm is the mode radius, and σ is the width of the distribution. Wolbach et al. (1985) found rm=0.11 µm and σ=1.6 for the soot in the K-Pg boundary layer. We assume this distribution represents the initial sizes of the soot particles. The final size, which would be determined by coagulation while in the atmosphere, might not be preserved in the sediments, and loosely bound clumps of particles would have been destroyed by the ultrasound treatment of the samples.

The size distribution of soot from the K-Pg boundary is similar to that of smoke near present-day biomass fires as indicated in Fig. 2 (e.g., Matichuk et al., 2008). This similarity in sizes is somewhat surprising because the present-day smoke size distribution includes organic carbon, which is present in addition to the elemental carbon (soot). Generally, in wildfire smoke, organic carbon has 5–10 times the mass of soot, so one might anticipate that the K-Pg soot would be about half the size of the present-day smoke rather than of similar size since the organic coatings are no longer present, or were never present, on the K-Pg soot. The organics might never have been present because mass fires are very intense and tend to consume all the available fuel, which might include the organic coatings. Aggregation in the hot fires may have caused this slightly larger than expected size in the K-Pg sediments. Wolbach et al. (1985) suspended their samples in water and subjected them to ultrasound for 15 min in a failed attempt to completely break up agglomerates. This failure indicated that the remaining agglomerates might have been flame welded. Therefore, the K-Pg size distribution from Wolbach et al. (1985) does not represent the monomers in the aggregate soot fractal structures. Rather the K-Pg size distributions represent a combination of monomers and aggregates that may have formed at high temperatures. Possibly the smallest-sized particles measured by Wolbach et al. (1985), which have radii of 30–60 nm, represent the soot monomers. These are in the same general range as monomer sizes observed in soot from conventional fires (Bond and Bergstrom, 2006).

The injection altitude of the soot depends on its source. In a series of papers, Belcher et al. (2003, 2004, 2005, 2009) and Belcher (2009) argue from multiple points of view that there were no global forest fires. Harvey et al. (2008) and Kaiho et al. (2016) argue that the soot originated from oil, coal, and other organic deposits at the location of the impact. If correct, the soot might have been injected at high altitude along with the large spherules. Recently, Robertson et al. (2013a) reconsidered each of the arguments presented by Belcher et al. (2003, 2004, 2005, 2009) and Belcher (2009) and came to the conclusion that global wildfires did indeed occur. Pierazzo and Artemieva (2012), Premović (2012), Morgan et al. (2013), as well as Robertson et al. (2013a) have independently argued that oil and other biomass in the crater is quantitatively insufficient to be the source of the soot. Therefore, we assume that the soot indeed originated from burning biomass distributed over the globe. The soot is clearly present in the distal layer material, and therefore was once in the atmosphere where it could cause significant changes to the climate.

Toon et al. (2007) have outlined the altitudes where one expects large mass fires to inject their smoke. Numerical simulations have shown that mass fires larger than about 5 km in diameter have smoke cloud tops well into the stratosphere. The smoke itself is distributed over a range of heights, however. The details of the injection profiles depend on the rate of fuel burning, the size of the fires, and the meteorological conditions among other factors. In addition, some smoke is quickly removed from the atmosphere by precipitation in pyrocumulus. However, it is thought that overseeding of the clouds by smoke prevents precipitation, and that only 20 % or so of the smoke injected into the upper troposphere is promptly rained out (Toon et al., 2007). Smoke that is injected near the ground, on the other hand, will be removed by rainfall within days or weeks.

The K-Pg impact occurred at a time when average biomass density likely was higher than now. Following Small and Heikes (1988; their Fig. 3f) and Pittock et al. (1989) one would expect smoke from large-area fires burning in high biomass density areas to show a bi-modal smoke injection profile. The smoke at higher levels is injected in the pyrocumulus and other regions with strong vertical motions. However, once the fires die-down smoke will be emitted in the boundary layer. There are also downdrafts, as well as entrainment and mixing with the environment, that occur in all cumulus and these will carry some smoke into the boundary layer. We simulate this with injections whose vertical distributions are Gaussian functions centered at the tropopause and at the surface, as illustrated in Fig. 1. The injection at the tropopause (Eq. 2) has a half width of 3 km, but nothing is injected above about 25 km. We set this upper altitude limit based on the heights of the stratospheric sulfate clouds from explosive volcanic eruptions, which rise buoyantly as do smoke plumes. The Gaussian distribution at the ground (Eq. 3) has a half width of 1 km, assuming that the local boundary layer is relatively shallow. We assume 50 % of the soot is contained in each of these distributions (Eqs. 2 and 3) for the general case, and for the 1 km impact. For the K-Pg, we assume the soot observed in the distal layer by Wolbach et al. (1988, 1990b) was all in the portion of the Gaussian distribution at the tropopause (Eq. 2).

Therefore, the injection profiles are given by I(gs-1km-1)=IT1η√2πe-0.5z-ztropη2I(gs-1km-1)=IT2μ√2πe-0.5zμ2 Here, I is the mass emission rate per kilometer of altitude, IT1 and IT2 are the total mass emitted per second into the upper (Eq. 2) or lower (Eq. 3) altitude range after correcting for the emission altitude range (0–25 km) and grid spacing, μ is 1 km, η is 3 km, and ztrop is the altitude of the tropopause.

Geographically, we assume for the K-Pg event that all the surface biomass is set on fire. For the 1 km diameter impact, however, only the region near the impact site would burn as discussed further below.

There is also an issue of how long it takes to inject the smoke. Forest fires often burn for days, advancing along a fire front as winds blow embers far beyond the flames and onto unburned terrain. Mass fires may not spread because powerful converging winds restrict the spread. However, little is known observationally about mass fires, and fires can spread by intense infrared radiation lighting adjacent material. If mass fires are restricted then they will burn only as long as they have fuel. The present aboveground global biomass in tropical forests is in the range of 0.6–1.2 g C cm-2 (Houghton, 2005). The energy content of biomass is on the order of 3 × 104 J (gC)-1 or, given the biomass concentration just mentioned, about 3 × 108 J m-2. Penner et al. (1986) and Small and Heikes (1988) found that large area mass fires with energy release rates of 0.1 MW m-2 would have plumes reaching the lower stratosphere. Hence, it would be necessary to assume that the fuel burned in an hour or so to achieve these energy releases. Of course, it might take some time for fires in different places to start fully burning, so considering the entire region of the mass fire, as opposed to a small individual part of the fires, might prolong the energy release considerably. For example, it took several hours for the mass fire in Hiroshima to develop after the explosion of the atom bomb (Toon et al., 2007).

It should be noted that in simulations of stratospheric injections of soot from nuclear conflicts, soot is self-lofted by sunlight heating the smoke (Robock et al., 2007b). However, in the case of the K-Pg impact, if there are other types of particles injected above the soot, which then block sunlight, the soot may not be self-lofted, which will limit its lifetime. The initial soot distribution that is estimated here does not include the effects of self-lofting, which would continue after the initial injection and should be part of the climate simulation.

The final property to specify for soot is the optical constants. This issue is complicated by the possible presence of organic material on the soot (Lack et al., 2012). However, it is known that many of these organics are quickly oxidized by ozone, which is plentiful in the ambient stratosphere. The stratosphere after the impact however, may have become depleted in ozone very quickly, so that the organic coatings might have survived. It is also possible that intense fires, such as mass fires, will consume the organic coatings, which may explain why the production of soot in the fires seems to have been so much more efficient than for normal fires. It may therefore be sufficient to treat the soot as fractal agglomerates of elemental carbon (Bond and Bergstrom, 2006). It is known that the optical properties of the agglomerates will not obey Mie theory. However, one may treat their optical properties as well as their microphysical properties using the fractal optics approach described by Wolf and Toon (2010). The optical constants for elemental carbon may then be used for the monomers. Alternatively, one may add the organic mass to the particles, and treat them using core–shell theory (Toon and Ackerman, 1981; Mikhailov et al., 2006).

Bond and Bergstrom (2006) have thoroughly reviewed the literature on the optical properties of elemental carbon. They conclude that the optical constants are most likely independent of wavelength across the visible, with a value that depends on the bulk density of the particles. Following their range of values for refractive index vs. particle density, we suggest using a wavelength-independent real index of refraction n=1.80 and an imaginary index k=0.67. We also use these values in the infrared as shown in Fig. 3. For the monomers in Tables 1 and 3, we adopt the density suggested by Bond and Bergstrom (2006) for light-absorbing material, 1.8 g cm-3.

Extrapolations of the soot injection parameters to smaller impactors than the one defining the K-Pg boundary should only involve changes to the mass of soot injected, since the basic properties of the soot at the K-Pg boundary are similar to those of forest fire soot. Therefore, the particle sizes, injection heights, and optical constants recommended in Table 3 for the smaller impact are the same as listed in Table 1 for the Chicxulub impact. The mass of soot injected is estimated from the extrapolations in Toon et al. (1997). For an impactor as small as 1 km diameter, debris from the impact site would not provide sufficient energy to ignite the global biota since the energy of the 1 km impactor is about 1000 times less than that of the Chicxulub impactor. Instead, radiation from the ablation of the incoming object and from the rising fireball at the impact site would ignite material that is within visible range of the entering object and the fireball. This ignition mechanism is well understood from nuclear weapons tests (Turco et al., 1990). Hence, for a 1 km diameter impactor the fuel load at the site of the impact becomes critical to evaluate the soot release. No soot would be produced from an impact in the ocean, an ice sheet, or a desert. In Table 3, to compute the smoke emitted (28 Tg), we use Eq. (12) from Toon et al. (1997) to obtain an area of 4.1 × 104 km2 for the expected area exposed to high thermal radiation density from the fireball for a 1 km diameter impactor with an assumed energy of 6.8 × 104 Mt. We then multiply that area by 3 % (the fraction of C in the burned fuel that is converted to smoke) and by 2.25 g C cm-2 (the assumed carbon content per unit area of the dry biomass that burns). The user of Table 3 can choose alternate values of the injected soot by scaling linearly to the biomass concentration they chose.

Ivany and Salawitch (1993) suggest that the land average aboveground biomass was about 1 × 1018 g (about 0.7 g C cm-2) at the end of the Cretaceous. The current land average, aboveground biomass is about 0.3 to 0.44 g C cm-2 (Ciais et al., 2013). An additional 1 to 1.6 g C cm-2 is currently present in the soil, while Ivany and Salawitch suggest 1 g C cm-2 in the soil in the Cretaceous. Some of the soil biomass may burn in a mass fire. Tropical and boreal forests currently have average biomass concentrations (aboveground and in soil) of about 2.4 g C cm-2, while temperate forests have about 1.6 g C cm-2, including soil carbon (Pan et al., 2011). Soil carbon is 30 % of carbon in tropical forests and 60 % in boreal forests. Together, tropical and boreal forests cover 6 % of the Earth's surface, and temperate forests 1.5 %. These forests cover 26 % of Earth's land area. In Table 3, we assume that the biomass that burns is typical of a tropical or boreal forest, assuming the soil carbon burns. The reader can make other choices for the biomass by scaling from the fuel load that the reader prefers.

Another modeling issue of concern is the ability of models to follow the initial evolution of the plume. If we assume that half of the 28 Mt of smoke from the 1 km impact is injected over an area of 4 × 104 km2 and over a depth of 6 km near the tropopause (Eq. 2) as 0.1 µm radius smoke particles, the smoke will have an initial optical depth near 4000, and the number density of particles will be about 107 cm-3. (The other half of the smoke mass injected near the ground (Eq. 3) will likely be removed quickly and have little impact on climate.) Intense solar heating at the top of the smoke cloud near the tropopause will loft it, while coagulation will reduce the number of particles by a factor of 2 and increase their size proportionately in only 1 min. Hence, one needs to model this evolution on sub-minute time scales to accurately follow the initial evolution. Alternatively, but less accurately, one might spread out the injection in time and space, so that the climate model can track the evolving smoke cloud using typical model time steps.

Johnson and Melosh (2012b) found that, at the end of their simulations of the rising fireball, about 44 % of the rock vapor that was created from the K-Pg asteroid impact remained as vapor rather than condensing to form large spherules. This vapor is about an equal mixture of impactor and asteroid, so the 44 % mass fraction is approximately equal to the mass of the impactor. This 44 % vapor fraction depends on the pressures reached in the impact, the equation of state of the materials, as well as the detailed evolution of the debris in the fireball. The fate of this vapor-phase material is not well understood and has been little studied. It may simply have condensed on the spherules, or it may have remained as vapor.

Presently, 100 µm and larger-sized micrometeoroids ablate to vapor in the upper atmosphere. Hunten et al. (1980), following earlier suggestions, modeled the condensation of these rock vapors as they form nanometer-sized particles in the mesosphere and stratosphere. Bardeen et al. (2008) produced modern models of their distribution based on injection calculations from Kalashnikova et al. (2000). Hervig et al. (2009) and Neely et al. (2011) showed that these tiny particles are observed as they deposit about 40 t of very fine-grained material on Earth's surface per day. It is possible that a similar process occurred after the Chicxulub impact. However, in the Chicxulub case the vaporization occurred during the initial asteroid impact at Chicxulub rather than on reentry of the material after the fireball rose thousands of kilometers into space and dispersed over the globe.

The presence of 15–25 nm diameter iron-rich material has been recognized in the fireball layer at a variety of sites by Wdowiak et al. (2001), Verma et al. (2002), Bhandari et al. (2002), Ferrow et al. (2011), and Vajda et al. (2015), among others. The nanophase iron correlates with iridium and is found worldwide, and therefore is likely a product of the impact process. Unfortunately, these authors have not quantified the amount of this material that is present. Berndt et al. (2011) were able to perform very high-resolution chemical analyses, and also report a component of the platinum group elements that arrived later than the bulk of the ejecta, and was probably the result of submicron-sized particles. However, they were not able to size the particles, nor quantify their abundance.

In Table 1, we take the upper limit of the injected mass of nanoparticles to be 2 × 1018 g. The lower limit is zero. This choice for the upper limit is consistent with the vapor mass left at the end of the simulations by Johnson and Melosh (2012b). We assume an initial diameter of 20 nm, following Wdowiak et al. (2001). We assume the particles are initially injected over the same altitude range as the Type 2 spherules, because we speculate that the small particles would not separate from the bulk of the ejecta in the fireball until the ejecta entered the atmosphere and reached terminal velocity. The mass injected would lead to an optical depth of particles larger than 1000 even if they coagulated into the 1 µm size range. Goldin and Melosh (2009) point out that such an optically thick layer of small particles left behind by the falling large spheres might also be important for determining whether the infrared radiation from the atmosphere heated by the Type 2 spherules is sufficient to start large-scale fires.

The optical properties of the nanoparticles are not known. We suggest using the optical properties of the small vaporized particles currently entering the atmosphere from Hervig et al. (2009). These optical constants are plotted in Fig. 3. We also assume that the particles have the density of CM2 asteroids, since Cr isotope ratios suggest that is the composition of the K-Pg impactor (Trinquier et al., 2006). This density is 2.7 g cm-3. A significant fraction of the vaporized material may be from the impact site, so using an asteroidal composition to determine the density is an approximation.

Johnson and Melosh (2012b) did not comment on the amount of vapor that would be expected to not condense as spherules from a 1 km diameter impact. From the theory of impacts, it is expected that an amount of impactor plus target that is about twice the mass of the impactor would be converted into vapor from a 1 km diameter impact, just as it is for a 10 km diameter impact. In Table 3, we assume that as an upper limit 35 % of the impactor mass plus an equivalent amount of target material would be left as vapor after spherules form. We chose this mass fraction, which is lower than that for the K-Pg object, because the 1 km impact will have a smaller fireball, and be more confined by the atmosphere. We also assume the injected particles will have a diameter of 20 nm. From simple energy balance along a ballistic trajectory, we would expect that the vaporized ejecta in the fireball from a 1 km impact would rise about a thousand kilometers above the Earth's surface. This altitude is consistent with limited numerical calculations for large energy releases, which indicate that the vertical velocity of the fireball is not significantly reduced in passing through the atmosphere (Jones and Kodis, 1982). As the material reenters the atmosphere, the particles will come to rest when they encounter an atmospheric mass comparable to their own mass. Hence, it is likely that the altitude distribution of the nanoparticles from the 1 km impact will be the same as we have assumed for the K-Pg impactor in Table 1, which is also similar to but slightly lower in altitude than the vertical distribution of micrometeorites on present-day Earth, as discussed by Bardeen et al. (2008). It is difficult to determine precisely the area that will be covered by this material as it reenters the atmosphere. If we assume that it takes about 30 min for the debris to reach peak altitude and return to the Earth and that the plume is spreading horizontally at about 4 km s-1 then the debris would enter the atmosphere over an area of about half that of the Earth. These estimates of area covered are consistent with the observations of the SL-9 impact collisions with Jupiter, and the plume from the much less energetic impact at Tunguska, though these are not perfect analogs (Boslough and Crawford, 1997). The optical depth of the nanoparticles from the 1 km diameter impact averaged over the Earth is estimated for comparison with the estimates of other types of particles to be relatively large (1.5) as noted in Table 3.

Another clear component of the K-Pg debris layer is pulverized target material. This clastic material was first recognized from shocked quartz grains (Bohor, 1990), but there are also shocked carbonate particles from the Yucatan Peninsula in the K-Pg boundary layer material (Yancy and Guillemette, 2008; Schulte et al., 2008). Because of chemical alteration of much of this material in the past 65 million years, it is difficult to determine the mass and size distribution directly except for the shocked quartz, which is readily identified. The shocked quartz grains generally are large and would not have remained long in the atmosphere. However, the shocked quartz is probably not directly related to the bulk of the clastics. For instance, within 4000 km of Chicxulub, the shocked quartz is primarily in the few millimeter-thick fireball layer, which is distinct from the several centimeter or thicker ejecta layer that is dominated by clastics. The shocked quartz likely came from basement rock, reached higher shock pressures than the bulk of the pulverized ejecta, and therefore was distributed globally in the impact fireball along with the melted and vaporized material from the target and impactor. The other pulverized material, in contrast, came mainly from the upper portions of the target along with basement rocks toward the exterior of the crater, and the fragments were distributed locally (within about 4000 km of Chicxulub) in the impact ejecta debris.

The submicron fraction of the clastics is of interest because particles of such size might remain in the atmosphere for months or years and perturb the climate, unlike larger particles that would be removed quickly by sedimentation. For instance, Pueschel et al. (1994) found 3–8 months after the 1991 eruption of Mt. Pinatubo in the Philippines that volcanic dust particles with a mean diameter near 1.5 µm were optically important in the lower stratosphere in the Arctic.

The optical constants for the injected clastics are suggested from their composition. For the Chicxulub impact, the clastic material is largely carbonate evaporates. We suggest using the optical constants of limestone from Orofino et al. (1998). Unfortunately, the values need to be generated from a table of oscillator strengths. They also need to be interpolated into the visible wavelength range. We suggest extending the oscillator predictions into the visible range as done by Querry et al. (1978). The density of limestone is in the range of 2.1–2.6 g cm-3, while dolomite and anhydrite have densities near 2.9 g cm-3. Granite has a density near 2.6–2.8. While each of these materials contribute to the clastic debris, for convenience, we assume the pulverized ejecta have a density of 2.7 g cm-3.

Pope (2002) and Toon et al. (1997) used two different methods to determine the amount of the submicron-clastic material from the Chicxulub impact. Unfortunately, these estimates disagree by about 4 orders of magnitude, as indicated in Table 5, third row, columns 1 and 2. Toon et al. (1997) used arguments based mainly on impact models to estimate that more than 10 % of the mass of the distal layer (> 7 × 1017 g) is submicron diameter clastics, which would be significant to climate. Pope (2002) estimated that the clastics in the distal layer have a mass that is < 1014 g. Pope (2002) used data on shocked quartz to constrain the amount of clastics, which in principle is a better approach than using estimates based on a model as in Toon et al. (1997). The amount of clastics of all sizes in the Pope (2002) model (1016 g) is only 12–30 times larger than the clastics of all sizes emitted in the relatively small 1980 Mt. St. Helens eruption. Therefore, based on the Pope (2002) analysis, the submicron fraction would not be of significance to climate. Below, we attempt to reconcile these two approaches to better determine the amount of submicron clastics.

Toon et al. (1997) estimated the amount of submicron clastics starting from analytical models of the mass of material injected into the atmosphere by a 45∘ impact. They estimated the mass of melt plus vapor per megaton of impact energy (∼ 0.2 Tg Mt-1) and the mass of pulverized material per megaton of impact energy (about 4.5 Tg Mt-1). Assuming a 1.5 × 108 Mt impact, these formulae suggest a melt plus vapor amount of 3 × 1019 g (∼ 1 × 104 km3, assuming a density of 2.7 g cm-3) and a pulverized amount of 7 × 1020 g (∼ 2.5 × 105 km3). While sophisticated impact calculations generally agree with the amount of melt plus vapor, not all of it is found to reach high enough velocity to be ejected from the crater. For example, Artemieva and Morgan (2009) investigated a number of impact scenarios that created transient craters with diameters of 90–100 km, which they thought to be consistent with the transient diameter of the Chicxulub crater. Considering those cases with oblique impacts from 30–45∘ with energies of 1.5–2 × 108 Mt, they found that the melt was in the range of 2.6 × 104 to 3.8 × 104 km3. However, the amount that reached high enough speed to be ejected from the crater was in the range of 5 × 103 to 6 × 103 km3 (average 5.6 × 103 km3, 1.4 × 1019 g, about 2–10 impactor masses). On average, only about 20 % of the melt and vapor amount escapes from the crater. Therefore, Toon et al. (1997) may have overestimated the amount of melt escaping from the crater by about a factor of 2. It should be noted that in Artemieva and Morgan (2009) the melt exceeds the mass of the distal layer, which is about 4 × 1018 g, by about a factor of 5 because much of the melt is deposited as part of the ejecta curtain and never reaches the distal region.

Artemieva and Morgan (2009) found that the total mass ejected from the crater is 1.3 × 104 km3 (2.9 × 1019 g). Assuming that 90 % of this material is pulverized rock, their results imply that Toon et al. (2007) overestimated the amount of clastic debris ejected from the crater by a factor of about 25. In column 3 of Table 5, we correct the amount of pulverized material to agree with the Artemieva and Morgan (2009) value of 2.9 × 1019 g of clastics escaping the crater. It is interesting to note that the clastic mass from Chicxulub is only a factor of about 10 larger than the minimal estimated mass of clastics ejected in the Toba volcanic eruption about 70 000 years ago (Matthews et al., 2012).

Another issue is the fraction of the pulverized debris that is submicron. Toon et al. (1997) computed the amount of pulverized debris whose diameter is smaller than 1 µm from size distributions measured in nuclear debris clouds originating from nuclear tests that were many orders of magnitude lower in energy than the K-Pg impact, and from impact crater studies cited by O'Keefe and Ahrens (1982) based on grain size measurements from craters. Toon et al. (1997) assume that 0.1 % of the total clastic material would be submicron. Pope (2002) cited studies of volcanic clouds to conclude that 1 % by mass of the pulverized material would be submicron.

Rose and Durant (2009) examined the total grain size distribution (TGSD) from a number of volcanic eruptions and concluded that the amount of fine ash is related to increasing explosivity of the event. The TGSD is supposed to represent the size distribution as the clastics left the crater. Mt. St. Helens is the most likely of the volcanic eruptions they considered to be relevant to the extreme energy release in a large impact. About 2 % of the total ejecta from Mt. St. Helens had a diameter smaller than 1 µm. Since the erupted mass was about 3–8 × 1014 g, the submicron mass emitted by Mt. St. Helens was about 6–16 × 1012 g. Matthews et al. (2012) considered the Toba eruption, whose clastics are within an order of magnitude of those from Chicxulub. Their data show that 1–2 % of the mass of the clastics is in particles smaller than 1 µm and 2–6 % in clastics smaller than 2.5 µm.

In Table 5, we use 2 % of the pulverized material as a revised estimate for the fraction of the clastic material that is released as submicron ejecta. This fraction is a factor of 20 larger than the one used in Toon et al. (1997). Hence, our revised submicron mass estimate for the Chicxulub impact (column 3 row 3) is very similar to the one Toon et al. (2007) estimated (column 2 row 3) because, although we lowered the estimate of the clastic mass exiting the crater to agree with Artemieva and Morgan (2009), we increased the estimate of the fraction that is submicron.

A confounding issue is the amount of submicron and other clastics that escapes from the near-crater region and is distributed globally. A large fraction of the pulverized debris in the ejecta curtain was removed within 4000 km of the impact crater (Bohor and Glass, 1995), and volcanic ejecta is likewise largely removed near the volcanic caldera. For example, there is 4–8 cm of ash 3000 km from the Toba crater, which is not too different from the thickness of the Chicxulub deposits at a similar distance from the crater. If the removal occurred only by individual particle sedimentation, one could simply take the mass in the smaller ranges of the size distribution and assume it spread to the rest of the globe. However, it is clear from volcanic eruption data that a significant fraction of the submicron debris is removed near the volcano by processes other than direct sedimentation (Durant et al., 2009; Rose and Durant, 2009). These processes include rainout of material from water that condenses in the volcanic plume, and also agglomeration possibly enhanced by electrical charges on the particles. It is likewise clear that such localized removal occurred after the K-Pg impact. Yancy and Guillemette (2008) describe accretionary particles that make up a large fraction of the debris layer as far as 2500 km from the Chicxulub crater. These agglomerated particles, which range in size from tens to hundreds of micrometers, are composed mainly of particles with a radius of 1–4 µm. While largely composed of carbonate, the particles are enriched in sulfur.

One can use the size distributions from volcanic data, along with the total clastic mass ejected from Chicxulub to compute the particle agglomeration, and thereby follow the particles as they spread across the Earth. Such work is now being done for volcanic events, for example, by Folch et al. (2010). They found that they can successfully reproduce mass deposited on the surface from the Mt. St. Helens eruption by including agglomeration. However, such calculations for Chicxulub are difficult for several reasons: the large clastic masses involved exceed the mass of the atmosphere for a considerable distance from the crater, so the debris flows cannot be reproduced in standard climate models; the complexity of the distribution of material in the plume with some material reaching escape velocity and other parts being hurled over a substantial fraction of the planet make it difficult to determine the spatial distribution of the material, and some material is likely lofted well above the tops of most climate models; and the presence of clastics, melt, and rock vapor together with sulfur and water produces a chemically complex plume.

Eventually, it will be necessary to use detailed nonhydrostatic, multiphase plume models including agglomeration to better understand the distribution of Chicxulub ejecta. In the meantime, for climate modeling, we suggest placing the clastic mass in Table 5 (2.9 × 1019 g) in a circular area with radius of 4000 km, which is 22.4 % of the area of Earth. This will result in a column density of 25 g cm-2, or a layer thickness of about 10 cm. The mass density of the atmosphere is about 1000 g cm-2, so this is about a 2.5 % perturbation to the mass of the atmosphere. In reality the mass is concentrated near the crater as shown by Hildebrand (1993). However, the observed mass density is relatively constant between 1000 and 4000 km. The initial vertical distribution of this material may be very complex due to density flows within several hundred kilometers of the crater. We suggest initializing models assuming an injection with an altitude independent mass mixing ratio of about 2.5 %. Given our suggested vertical distribution, 90 % of the material will initially lie in the troposphere. Tropospheric material is unlikely to become globally distributed even if it escapes agglomeration, because it will quickly be removed by rainfall.

As an alternative to the complexity of modeling the loss of this material in the troposphere and considering the entire size distribution, we suggest simply placing an appropriate mass into the stratosphere. The values for a stratospheric injection are given in the bottom row of Table 5 and the first row of Table 1. For illustration, we have estimated the final optical depth assuming that 10 % of the submicron material (the amount placed into the stratosphere) will escape removal. For a size distribution, we suggest using the smaller size mode measured in the stratosphere after the Mt. St. Helens eruption, as summarized by Turco et al. (1983). This size distribution is log-normal (Eq. 1), with a mode radius of 0.5 µm and a standard deviation of 1.65. The estimated optical depth of 88 is very large, even though the submicron clastic material in this estimate is only about 1 % of the mass of the distal layer.

Pope (2002) determined the amount of clastics by modeling the amount of quartz in the distal layer. He found that he needed an initial injection of about 5 × 1015 g of quartz to match the distribution of quartz mass with distance from the impact site. It is not clear how good this estimate is because the removal rate of material in large volcanic clouds, a possible impact analog, does not occur by individual particle sedimentation, but rather by settling of agglomerates (Folch et al., 2010). Hence, removal in the region near the impact site may have been larger than Pope estimated, requiring a larger volume of quartz, or the removal of clastics may be different than that of quartz. The value in Artemieva and Morgan (2009) for the pulverized material ejected from the crater is 3 orders of magnitude larger than the estimate of Pope (2002). Most of this material is in the ejecta curtain, not in the impact fireball, and so is deposited close to the impact crater. The shocked quartz is primarily associated with the impact fireball, so the bulk of the pulverized material may not be seen in Pope's analysis.

Pope assumed that quartz composed 50 % of all the clastic debris, so that all of the clastics injected weighed about 1016 g. This number is about 2 orders of magnitude less than the clastics from the Toba eruption (Matthews et al., 2012), and more than 3 orders of magnitude less than the Artemieva and Morgan (2009) estimate for clastics from the Chicxulub impact.

The assumption by Pope (2002) that quartz is 50 % of all the clastics is likely in error. There is no reason to think there is much quartz in the upper layers of sediment at the Chicxulub site. In the stratigraphic columns shown by Ward et al. (1995) the pre-impact sediments at Chicxulub consist of approximately 3 km of Mesozoic carbonates and evaporites with ∼ 3–4 % shale and sandstone. Therefore, it is more likely the quartz originates from the basement rocks. There is also not a strong connection between the physical processes that distributed the quartz (the impact fireball with high ejection velocity), and those that distributed the pulverized material (the ejecta curtain with low velocity).

It is possible that the quartz to clastics ratio is determined by the ratio of quartz to total debris in the samples closest to Chicxulub, since these may have suffered the least removal by sedimentation. Pope suggests these intermediate distance layers contain about 1 % quartz, but only considers the fireball layer, which is less than 10 % of the total ejecta layer within 1000 km of the crater. The remainder of the intermediate distance layer contains little quartz, so the clastics could be more than 1000 times the mass of the quartz. It is not clear that 1000 is an upper limit to the ratio of clastics to quartz because the quartz and pulverized material move along different paths in the debris cloud. If we accept this ratio of 1000 for the ratio of clastics to quartz, the mass of clastics from Pope's analysis would be 5 × 1018 g, which is within a factor of 6 of the Artemieva and Morgan (2009) value. If 1 % of this mass is submicron then 5 × 1016 g of submicron clastics would have been injected into the upper atmosphere.

Table 5 shows that the new estimate of submicron mass following the procedure of Toon et al. (1997) agrees with the new estimate following the procedure of Pope (2002) within 20 %. The new estimate is about 12 times less than the Toon et al. (1997) value mainly because Toon et al. (1997) did not consider that most of the pulverized mass would not be ejected from the crater. The new application of the Pope (2002) approach leads to estimated submicron dust emissions that are about 500 times larger than the one originally derived by Pope (2002). The major difference is that we have assumed the ratio of quartz to clastics is about 1000, rather than 1 as assumed by Pope (2002). Despite the perhaps coincidental agreement of these two estimates, there is substantial uncertainly in the true mass of submicron clastic particles in the K-Pg distal layer. Observations of the submicron material in the distal layer are needed.

In order to determine the properties of the pulverized ejecta from a 1 km impactor, we use the pulverized mass injection per teragram of impact energy from Toon et al. (1997), but reduce it by the factor of 25 discussed earlier to account for the fraction of the clastic mass with enough velocity to escape the crater. This procedure yields a clastic mass of 1.3 × 1016 g. For reference, the volume of clastics from the eruption of Mt. Tambora in 1815 is estimated to have been about 150 km3, which is a mass of about 3 × 1017 g. Hence, the Tambora eruption likely surpasses the clastics from the hypothetical 1 km diameter impactor by more than a factor of 10. The same size distribution for the clastics is recommended for the 1 km impact and the Chicxulub impact, since it seems to hold for a range of volcanic events from Mt. St. Helens to Toba, which span the 1 km diameter impactor in terms of clastics. We also suggest that the mass be initially mixed uniformly in the vertical above the tropopause. According to Stothers (1984), the Tambora clastics were deposited in layers that are centimeters in thickness at distances 500 km from the volcano. Accounting for the drift of the ash downwind, the area of significant ash fall was about 4.5 × 105 km2. If this same area is used for the initial injection of the clastics for the 1 km impact, then the column mass concentration is about 8.7 g cm-2, which in turn is slightly less than 1 % of the atmospheric column mass. The estimated optical depth of the clastics in Table 3 is about 25 % of the optical depth from nanoparticles originating from vaporized rock. Given that these materials are much less absorbing than soot, and lower in optical depth than nanoparticles, they can probably be neglected in estimates of the climate changes due to a 1 km diameter impact on land.

Kring et al. (1996) summarized the S, C, and water contents of a large number of types of asteroids. Trinquier et al. (2006) found from chromium isotopes that the Chicxulub impactor was most likely a carbonaceous chondrite of CM2 type. Such asteroids have 3.1 wt % S, 1.98 wt % C, 11.9 wt % water, and a density of 2.71 g cm-3. Over the range of chondrites, which constitute 85% of meteorite falls, S varies from 1.57 to 5.67 wt %, C from 0.04 to 3.2 wt %, and water from 0.2 to 16.9 wt %. Kallemeyn and Wasson (1981) report that by mass CM carbonaceous chondrites contain about 4 ppm Br. Goles et al. (1967) report that Cl ranges from 190–840 ppmm of carbonaceous chondrites, Br ranges from 0.25 to 5.1 ppmm, and iodine ranges from 170 to 480 ppbm. Table 6 summarizes the composition of asteroids using values for CM2 type carbonaceous chondrites from Kring et al. (1996) for S, C, and water, and for the Mighei (the CM2 type example) from Goles et al. (1967) for Cl, Br, and I.

Tables 2 and 4 indicate the direct contributions from 1 and 10 km impactors of a number of chemicals, as discussed further below. We assume that the entire 10 or 1 km diameter impactor melted or vaporized so that all of the gases are released. For the 10 km impactor, these gases would have been distributed globally in the hot plume along with the melt spherules within hours. They would reenter with the same vertical distribution as the Type 2 spherules. For the 1 km diameter impactor, the initial injection may have only covered half the Earth, with global distribution over days via wind, after reentry into the upper atmosphere.

We further assume that the vapors under consideration do not react with the hot mineral grains either in the plume or in the hot layer at the reentry site. In fact, given the large particle surface areas in the atmosphere over the globe it is possible that there was a significant transfer of material from the gas phase to the surfaces of the mineral grains in a short period of time.

As pointed out by Kring et al. (1996) and Toon et al. (1997), the S in a 10 km diameter impactor would exceed that from the Mt. Pinatubo volcanic injection by a factor above 1000. Even a 1 km diameter carbonaceous chondrite could deliver several times as much sulfur to the atmosphere as the Mt. Pinatubo eruption in 1991. Stratospheric water could be enhanced by a factor of more than 100 from the water in a 10 km impactor. Cl could be enhanced by factors above 500, Br by almost 500, and I by more than 50 000. However, there is not enough C in a 10 km asteroid to affect the global carbon cycle significantly.

Many investigators have pointed to sulfate as an important aerosol following the Chicxulub impact. Tables 1 and 3 compare the mass of sulfur from the impactor with the mass of the spherules and nanoparticles. The optical depth, which controls the climate change following the impact, and the particle surface area, which likely controls chemistry, are approximately linear with the mass. In our estimates, the sulfate coming directly from the asteroid could have a large optical depth assuming it was not removed on the spherules or large clastics.

The composition of seawater is given in Table 6 (Millero et al., 2008). It is thought that injections of water into the upper atmosphere will lead to droplet evaporation, with small crystals of salt left behind. If liquid water is left after a massive injection of water, the droplets will likely freeze leaving salt behind as particles embedded in ice crystals. Vaporization of water during the impact may leave behind salt crystals, or the salts may decompose into their components. As discussed by Birks et al. (2007), complex simulations are needed to determine how much material is freed from the salt particles to enter the gas phase where it might destroy ozone. In Tables 2 and 4, we list the total amounts of several interesting chemicals that might be inserted into the stratosphere. However, all of them except water vapor are likely to be in the form of a particulate until photochemical reactions liberate them.

A significant uncertainty related to any oceanic contribution to atmospheric composition is the depth of the ocean in relation to the size of the impactor, and the water content of sediments at the crater site. The depth of the ocean at Chicxulub at the time of the impact is not known. Many investigators have referred to it as a shallow sea. However, Gulick et al. (2008) estimates that the water depth averaged over the impact site was 650 m, which is considerably deeper than earlier estimates. We use a water depth of 650 m in Table 2 to estimate the amounts of material injected by Chicxulub. A 1 km diameter impactor is smaller than the average depth of the world oceans, which is about 3.7 km.

For the Chicxulub impact, we follow Pope et al. (1997) and assume that the 650 m depth of seawater within the diameter of the impactor (10 km) will be vaporized, follow the path of the Type 2 spherules, and reenter the atmosphere globally. In Table 2 we compute the water vaporized following the equations in Toon et al. (1997). These equations, assuming an impact velocity of 20 km s-1, led to an order of magnitude greater injection of water than using Pope's estimate. The vaporized water is 0.4 times the impactor mass. During the vaporization of the seawater we assume the water will be present as water vapor, and that the materials in the water will be released as vapors. Some of these materials likely would react quickly with the hot minerals in the fireball or later with the hot minerals in the reentry layer.

It is also likely that a considerable amount of water was splashed into the upper atmosphere. Ahrens and O'Keefe (1983) estimated that the water splashed above the tropopause from a 10 km diameter impact into a 5 km deep ocean would be 30 times the mass of the impactor. We assume that the amount of water splashed above the tropopause will scale linearly with the depth of the ocean. Therefore, about 4 times the impactor mass of water may have been splashed into the upper atmosphere. Much of this water may immediately condense and rainout, as discussed in Toon et al. (1997). However, some of the dissolved salts may be released if some of the water evaporates. The assumed injection of gases, and particulates that might become gases, from the ocean is summarized in Table 2 for the Chicxulub impact.

No seawater is injected by the 1 km diameter asteroid impact on land. If a comet hit the land there would be a water injection.

Pierazzo et al. (2010) estimated that 43 Tg of water would be injected above 15 km by a 1 km asteroid impact into the deep ocean. Of this water, 25 % is in the form of vapor and 75 % in the form of liquid water. In their modeling the water was assumed to be distributed with a uniform mixing ratio from the tropopause to the model top. It was also spread uniformly over an area 6200 × 6200 km in latitude and longitude. Using the equations in Toon et al. (1997) for the vaporized water produces a value which is 60 % of the vaporized water from the detailed modeling used in Pierazzo et al. (2010). Given these water injections we use the composition of seawater to determine the injections of the various species. Pierazzo et al. (2010) estimate injections of Cl and Br that are more than an order of magnitude smaller than ours because they consider the amounts that have been converted into gas-phase Cl and Br by photochemical reactions in the atmosphere, while we estimate the total injections, which initially are likely to be in the particulate phase.

The sea floor at the Chicxulub impact site, like the modern Yucatan, contained abundant carbonate and sulfate-rich deposits. Ward et al. (1995) conclude that 2.5–3 km of sedimentary rock were present at Chicxulub, composed of 35–40 % dolomite, 25–30 % limestone, 25–30 % anhydrite, and 3–4 % sandstone and shale. The dolomite and limestone are no doubt porous. Pope et al. (1997) estimate the carbonates in the Yucatan have a porosity of 20 %. The pores would have been filled by seawater since the sediments were submerged. This groundwater produces an equivalent water depth of about 400 m. The carbon content of limestone is 12 % by weight and of dolomite 15 % by weight. The sulfur content of anhydrite is 23.5 % by weight. To our knowledge, trace species such as Br, Cl, and I have not been reported for these sedimentary rocks, but would be present in the seawater in the pores.

For the 10 km Chicxulub impact, we follow Pope et al. (1997) for the abundances of S and C assuming 30 % anhydrite, 30 % limestone, and 40 % dolomite. The composition of the impact site is given in Table 6. We ignored species other than S and C that might be in the target material. It is difficult to follow the target debris since some of it is vaporized, and some melted. We follow Pope et al. (1997) and assume that the upper 3 km of the target is vaporized within the diameter of the impactor. The gases within this volume of vaporized material are assumed to be released, and to follow the trajectories of the Type 2 spherules. Pope et al. (1997) estimated the amount of material that would be degassed from target material that was melted or crushed in a large impact. We use the values from Table 3 of Pope et al. (1997) for out of footprint vapors in our Table 2 for the degassed impact site emissions. We also assume that the granite underlying the impact site does not contribute.

The source gases from a 1 km land impact would depend on the composition of the impact site, so we do not list values in Table 4. We assume nothing would be liberated from the sea floor in a 1 km impact in the deep ocean.

It is well known that forest fires emit a wide variety of vapors into the atmosphere. Andreae and Merlet (2001) provide emission ratios (gram of material emitted per gram of dry biomass burned) for many vapors expected to be important in the atmosphere, as listed in Table 6. As discussed in Sect. 2.2.1, the soot emission may have been enhanced relative to wildfire estimates by Andreae and Merlet (2001) after the Chicxulub impact because the impact-generated fires were mass fires. We do not consider any enhancements of the gas-phase emission ratios, but they may also be impacted by fire intensity or the types of plants making up the biomass.

In Tables 2 and 4, we computed the burned mass from Chicxulub assuming that 1.5 g cm-2 of dry biomass burns over the entire land surface area of the Earth, and then used the emission factors from Andrea and Merlet (2001) to obtain the gas-phase emissions. For a 1 km impact, we assume the area burned is 4.1 × 104 km2 (Toon et al., 1997), and the dry biomass is 2.25 g C cm-2. We then used the emission ratios from Andreae and Merlet (2001) to compute the gas-phase emissions. Comparing the gas-phase emissions from fires in Tables 2 and 4 with ambient values indicates that there would be large perturbations for all gases for the 10 km diameter impact. Only iodine is significantly perturbed for the 1 km impact. For the gas-phase emissions, we suggest using the same vertical profile as suggested for soot earlier. The emissions would only occur over the region near the impact site for the 1 km impact.