We present the results of numerical modeling of the formation of the Ries crater utilizing the two hydrocodes SOVA and iSALE. These standard models allow us to reproduce crater shape, size, and morphology, and composition and extension of the continuous ejecta blanket. Some of these results cannot, however, be readily reconciled with observations: the impact plume above the crater consists mainly of molten and vaporized sedimentary rocks, containing very little material in comparison with the ejecta curtain; at the end of the modification stage, the crater floor is covered by a thick layer of impact melt with a total volume of 6–11 km3; the thickness of true fallback material from the plume inside the crater does not exceed a couple of meters; ejecta from all stratigraphic units of the target are transported ballistically; no separation of sedimentary and crystalline rocks—as observed between suevites and Bunte Breccia at Ries—is noted. We also present numerical results quantifying the existing geological hypotheses of Ries ejecta emplacement from an impact plume, by melt flow, or by a pyroclastic density current. The results show that none of these mechanisms is consistent with physical constraints and/or observations. Finally, we suggest a new hypothesis of suevite formation and emplacement by postimpact interaction of hot impact melt with water or volatile-rich sedimentary rocks.
At the 26 km-diameter Ries impact structure (Germany), suevite occurs in three different geological settings (see Stöffler et al. 2013): (1) a thick and likely continuous layer of “crater suevite” (CS) in the central crater cavity inside the inner ring, (2) thin isolated patches of “outer suevite” (OS) on top of the continuous ejecta blanket of polymict lithic breccia (“Bunte Breccia” = BB) outside of the crater rim, and (3) dikes of suevite in the crater basement and in displaced megablocks. Furthermore, crater suevite as observed in the Nördlingen 1973 drill core can be subdivided into three major units: (1) graded suevite from 314 to 331.5 m depth, (2) melt-rich suevite from 331.5 to about 525 m, and (3) melt-poor suevite from 525 to 602 m as dike-like intercalations within the brecciated crystalline basement (Pohl et al. 1977; Engelhardt and Graup 1984). A detailed description of the crater and its impactites, as well as a review of the history of Ries research, is presented in the companion paper by Stöffler et al. (2013).
Despite decades of research and a wide variety of detailed observations, the formation and emplacement of suevite have remained enigmatic. Furthermore, the closely related problems of melt deficiency and lack of a coherent impact melt sheet inside the crater also remain unexplained. The knife-sharp contact between BB and OS clearly suggests two different deposition regimes, presumably separated in time. In principle, three different hypotheses have been suggested to explain the emplacement of OS; however, not all of them account for the initial formation process and explain the paucity of melt inside the crater:
OS components are involved in the ejecta plume and deposited with a free-fall velocity after a substantial time lag (i.e., well after deposition of the BB; Stöffler 1977; Engelhardt and Graup 1984; Engelhardt 1997);
Ejecta plume collapse causes “pyroclastic-like” flows (Newsom et al. 1986, 1990); and
Impact melts (or melt-rich materials) flow off the central uplift during the modification stage and may be deposited far beyond the crater rim (Osinski 2004; Osinski et al. 2004).
All three hypotheses are based on observations, but have never been rigorously tested whether they satisfy principal physical laws or are quantitatively in agreement with, for instance, the estimated amount of total suevite. Therefore, we prefer to address them with the term hypothesis, although the above-mentioned concepts have often been referred to as models.
The terms “hypothesis” and “model” are widely used in modern geology and have often caused misconceptions. Therefore, we would like to begin this article with a few simple definitions explaining their scientific meaning and applicability to geological processes. These definitions are mainly extracted from the Oxford English Dictionary (www.oed.com) and popular literature on scientific methods. A hypothesis is a proposed explanation for a phenomenon, e.g., the presence of a double ejecta layer near the Ries crater. In other words, a hypothesis is based on observations that cannot be explained adequately with the present knowledge and scientific theories. Indeed, the formation of double ejecta layers cannot be explained by standard impact crater mechanics including the ballistic motion of ejected materials. Additional dynamic processes (see 1–3 above) have been suggested to explain the given observations at the Ries. Obviously, these processes cannot be observed in situ (during a Ries-like impact) and, hence, these hypotheses are based exclusively on nonimpact analogs that can be studied by direct observations such as volcanic or nuclear explosion plumes or pyroclastic or melt flows after volcanic eruptions.
A basic prerequisite for a hypothesis to become a scientific hypothesis is that it can be tested. The most common approach to test a hypothesis for a complex phenomenon such as the double ejecta layer at Ries is to suggest a model. The term model may be defined as a simplified or idealized description or conception of a particular system, situation, or process. This model may be used in laboratory experiments and, in this case, has to include proper rules to scale laboratory results to natural events. The above-mentioned hypotheses for the origin of suevite are difficult to test in this way, as impact plumes and impact melt sheets are rarely observed in laboratory experiments where the impact velocity is usually significantly below the average velocity of natural impacts on earth. Another way to test a hypothesis is to create a numerical model or simulation. A dynamic simulation provides information over time and shows how a particular object or phenomenon will behave. These simulations are based on first principles and have been widely used as an important tool in impact research since the early 1960s. To ensure reliability of results produced by numerical modeling, thorough validation against laboratory experiments, explosion tests, and benchmarking of different codes is crucial (e.g., Pierazzo et al. 2008). The major shortcoming lies in an appropriate description of material behavior under the extreme conditions during the formation of an ejecta plume and impact-induced melt. Theoretical assumptions have to be complemented by empirical (based on experiments) or semi-empirical (experiments plus theory) equations of state and constitutive equations.
Despite all the efforts over the last decades, numerical models are not perfect. Computer simulations mimic the natural processes only to the extent of our knowledge of the physics that governs the processes we are interested in. If the model does not account for a specific process or misses an important parameter describing the given properties of a system, the dynamic simulation fails to reproduce the system's behavior. For example, if the atmosphere and vaporization of involved materials are neglected (a common simplification often applied to planetary bodies with no or a very thin atmosphere and a low impact velocity), we cannot expect nonballistic ejecta transport. In these cases, hydrocodes in use produce accurate results, but our starting models are incomplete.
If the simulations confirm the hypothesis, it may then be regarded as a theory or law of nature. If the simulations do not confirm the hypothesis, it must be rejected or modified. Moreover, there is always the possibility that a new observation or a new experiment may conflict with a hypothesis that has been regarded as well tested.
We started this project 5 yr ago with one simple goal—to test the plume hypothesis of suevite origin and emplacement. The Ries crater and its ejecta as an excellent test case due to its unique state of preservation did not allow us to accomplish this mission and we had to consider all other hypotheses and, finally, suggest a new one. In this article, first, we review existing hypotheses of the Ries ejecta formation and summarize the questions, which may be crucial and which cannot be addressed without numerical modeling. Then, we present the results of a standard cratering model of the Ries crater including crater formation, deposition of ballistic ejecta, and ejecta plume collapse. In the third section, we quantitatively evaluate the existing hypotheses of suevite formation and conclude that none of them can be reproduced in the numerical simulations. Finally and consequently, we suggest an additional postimpact mechanism to modify the Ries crater melt sheet and to produce the second layer of its ejecta. This process does not contradict observations, has observable analogs in nature, and may be simulated, albeit with substantial simplifications.
Existing Hypotheses of the Ries Crater Suevites and Related Allochthonous Impact Breccias
On the basis of field observations and laboratory studies, the Ries proximal impactites may be divided into four groups: Bunte Breccia, megablocks, crater suevite, and outer suevite. Note, the latter two terms for suevite deviate from the traditionally used terminology in the literature. As we intend to test previously suggested hypotheses, we prefer to avoid the use of any genetic terms (see also Stöffler et al. 2013). Genetically, CS comprises several types of suevites that differ, inter alia, by their position inside the crater, melt content, and sorting. A detailed description of these impactites is presented in Stöffler et al. (2013). The five most important features of the impact rocks at the Ries crater, which have to be explained by models, are:
The precursor rocks of suevite are mainly crystalline basement rocks, whereas Bunte Breccia is mainly derived from sedimentary rocks;
The components of suevite represent a wide range of shock metamorphism including whole rock melting, whereas Bunte Breccia contains only weakly shocked or unshocked clasts;
Outer suevite is separated from the underlying Bunte Breccia by a knife-sharp boundary and does not show any substantial mixing with the substrate Bunte Breccia;
Outer suevite extends to only 1.8 crater radii, whereas the radial extension of Bunte Breccia is up to 3.3 crater radii;
Both OS and CS are unsorted on a meter-sized scale.
Bunte Breccia and Megablock Emplacement
BB has been considered by all authors as ballistic ejecta from the sedimentary target rocks (for a section on the preimpact target, see Stöffler et al. 2013). Although the theory of ballistic sedimentation was first described by Oberbeck et al. (1975) for craters on the airless Moon, it is perfectly applicable to BB emplacement: massive, mostly unshocked fragments from the sedimentary cover are ejected; they move along parabolic trajectories without substantial interaction with the atmosphere and with each other, fall back to the surface with approximately the same high velocity as they were ejected with (increasing at greater distance from the crater), and incorporate a substantial amount of local material into the ejecta blanket (Pohl et al. 1977; Hörz et al. 1983).
Megablocks also traveled ballistically after ejection or slumped outward from the inner ring. From the peripheral and azimuthal distribution of ejected rock types around the crater, Graup (1978) developed a model of the original positions of different rock types in the basement—a predominantly granitic unit overlying a metamorphic series. In the same paper, he showed that some blocks were simply moved radially and elevated, and some were rotated slightly or considerably.
However, the problem of ballistic ejecta is still not fully resolved. Simple estimates based on experiments (Stöffler et al. 1975), as well as observations (Grieve et al. 1981), have shown that excavation depth at the Ries was at least 1.2 km, i.e., the projectile penetrated substantially into the basement. Recent numerical modeling (Wünnemann et al. 2005) confirmed this value for the excavation depth. What happened with these basement-derived ejecta? Why are the basement rocks underrepresented in BB?
Crater Suevite (CS) and Suevitic Dikes
The interpretation of the CS is largely based on studies of the Nördlingen 1973 core. There is general agreement that the CS is a mixture of impact-modified rocks within a transient cavity without substantial influence of an atmosphere and/or plume (i.e., the main body of crater suevite is not fallback material). Even in the first detailed description of the Nördlingen 1973 drill core (Stöffler 1977), it was already suggested that the lower melt-poor layers were intruded, in the mode of a “ground surge,” into the fractured crystalline basement (505–602 m and 642 m). The suevite below about 378 m was interpreted as a fallback formation, in which the melt content is highest at the top and decreases continuously downward (material with a higher proportion of strongly shocked rocks was ejected first and relatively faster). The suevites above 378 m that have a variable content of melt might have slumped into the crater from a position near the inner crystalline ring, because it is separated from the suevite below by a thin layer of sorted suevite produced by fluvial action. Although the word “ejected” was used, later in the same paper, it was stated that “the velocity (of the CS) did not result in an ejection beyond the rim of the transient crater cavity.”
The scarcity of sedimentary rocks in the crater suevite was explained either by fusion/vaporization of these rocks to produce vesiculation in the CS or by ejection of these rocks with extremely high velocity far beyond the crater rim and onto the continuous ejecta blanket (Stöffler 1977). This may be, in particular, true for the uppermost Neogene and Jurassic strata, whereas a small Triassic component is verifiably contained in the CS.
This 30 yr-old interpretation is far from being comprehensive (or quantitative), but it provides satisfactory explanations of the features of CS and is, in general, consistent with the results of analytical and general numerical models: the surficial material is ejected with higher velocity than the materials from deeper layers; these materials are less shocked than deep-seated materials with similar ejection velocity. To explain the same features, Engelhardt and Graup (1984) applied the deep-burst model of Kieffer and Simonds (1980) and claimed that melting/vaporization occurred mainly within the crystalline basement, not in the sedimentary cover. This idea appears unrealistic and does not agree with recent numerical models of impact melt production (e.g., Pierazzo et al. 1997): although shock pressures decay quickly near the free surface, a lot of sedimentary rocks at the Ries crater were affected by high shock pressures up to melting or degassing (Stöffler et al. 2002).
Graup (1981) claimed that the lower CS unit (melt-poor suevite) could not be a fallback formation because it does not contain glass bombs or glass spherules, only lithic “chondrules” (rounding of shocked clasts and mineral grains by abrasion in an impact-generated base surge). He also studied accretionary lapilli in the upper sorted layer of the CS and found that they resemble volcanic lapilli, formed in an “impact cloud laden with water vapor and gases from vaporized meteorite and target rocks.” Engelhardt and Graup (1984) and Engelhardt (1990, 1997) confirmed Stöffler's (1977) description, pointing out that the CS “was not lifted to great altitudes and did not leave the crater cavity,” except of “the uppermost graded unit settled from the dust-rich plume” (see Jankowski [1977a] and Stöffler et al.  for more details about graded suevite).
Newsom et al. (1990) suggested two possibilities for formation of CS: the lower part had been deposited as a coating of the transient cavity and was never ejected from the crater, or, alternatively, the CS was ejected and immediately fell back from a very low-density, but high-temperature region of the plume above the crater. The authors used the results of numerical modeling (Roddy et al. 1987) to show that this region had been formed by the passage of the projectile and the resultant strong, outward radial flow of air. Its low density prevented aerodynamic shaping of ejecta, while its high temperature prevented solidification of glass spherules and water condensation. The upper sorted suevite precipitated from an impact cloud. The increasing thickness of the CS toward the crater center (based on the analysis of the reworked CS from three drill cores) suggested that a cloud of debris was localized above the crater center. The authors compared this cloud with a cloud produced by the fireball from a nuclear explosion, although the importance of this cloud may have been limited compared with the effects of a nuclear bomb, because atmospheric heating in an impact is much less efficient than in a nuclear explosion (Schultz and Gault 1982).
However, none of the models described above resolves another crucial “enigma” of the Ries crater—what happened to the unequivocally generated impact melt formed due to high shock compression in the crystalline basement? Why do we have an apparent melt deficiency at Ries (e.g., Grieve and Cintala 1992)? Engelhardt and Graup (1984) discussed the matrix composition of suevite and suggested that, if this matrix consisted initially of molten particles, then the CS was equivalent (both with regard to its volume and its stratigraphic position) to coherent melt sheets developed by impact into crystalline targets. They wrote: “…it seems evident that the anomaly of the Ries does not consist in an extremely low amount of melt, but in the fact that the melt did not form a coherent melt sheet, but was dispersed as small particles.” Osinski et al. (2004) further supported this idea by finding much more melt in the OS than had been previously considered, and assumed that the CS may be similar to the OS (see also Stöffler et al. 2013).
A double ejecta blanket at the Ries crater is a unique feature in terrestrial impact craters (in which ejecta blankets are rarely preserved at all). There was broad agreement that OS (the upper layer of the blanket) was derived from the ejecta plume and was deposited “non-ballistically” as fallback material from the plume. Stöffler (1977) suggested that the lower part of the impact melt zone (mainly crystalline basement) was ejected at the very end of the excavation stage, i.e., substantially later than the Bunte Breccia. Melt was disrupted into melt particles of various sizes by the upward acceleration and became mixed with nonmolten fragments. These particles represent the observed glassy or crystallized melt inclusions of the suevite (“glass bombs”). This dense cloud left the crater as “a sheet-like conical plume, which is irregular and discontinuous, resulting in a patchy distribution of the OS.”
Engelhardt and Graup (1980) compared this cloud with a nuclear fireball and/or with experimental plumes observed by Piekutowski (1977) at experimental explosion craters in layered targets. Later, the same authors (Engelhardt and Graup 1984) presented the pressure–temperature history of the OS: (1) the initial melt-vapor mixture had a temperature of at least 2000 °C and was formed from basement gneisses compressed above 80 GPa; (2) rising from the crater, this cloud incorporated cooler fragments of crystalline basement rocks together with very few unshocked sedimentary clasts; (3) the mixture cooled rapidly by adiabatic expansion in the fast rising and slowly falling suevite cloud; (4) viscous pieces of undercooled melt were deformed by aerodynamic forces during turbulent movement in the suspension; (5) broken fragments of glass bombs show that before deposition on the ground, the temperature of the suspension had dropped below the transformation temperature of siliceous glass, i.e., below 750 °C, whereas the lower limit of temperature at deposition was defined by the Curie temperature of magnetite (580 °C, Pohl et al. 1977). Engelhardt et al. (1995) showed that the suevite glasses crystallized at elevated water pressure because the formation of plagioclase requires water pressures >10 bar. Although the OS was mainly transported as a turbulent cloud, an influence of initial ballistic ejection is indicated by increasing degrees of shock (on average) with increasing radial distances from the crater (Engelhardt 1997). Engelhardt (1990) discussed a possible source of hot gases, sweeping initially ballistic ejecta into the plume. He argued that not vaporized sedimentary rocks (as suggested by Kieffer and Simonds 1980), but vaporized crystalline basement (3.5 km3) provided the propellant.
Osinski et al. (2004) criticized the plume model described above, keeping in mind true fallout volcanic deposits, which (1) are well-sorted and normally graded; (2) do not show preferential horizontal orientation; (3) do not have chilled margins, i.e., they are deposited over a prolonged time interval; and (4) have a uniform thickness over large distances draping the underlying topography. Indeed, all these features are incompatible with the OS, which is certainly not a volcanic fallout material. However, this criticism is not quite relevant to impact “fallout,” which may differ substantially from its volcanic “counterpart” due to different energies, solid/gas ratios, geometry of the flows, etc. All above-mentioned properties of volcanic fallout deposits are directly connected to an extremely diluted character of a volcanic cloud and, as a consequence, to a low surface density of deposits (2 × 103 kg m−2 at most) in comparison with atmospheric surface density (1.3 × 104 kg m−2). At the same time, unsorted (and thick) fallout materials routinely occur near volcanic vents.
The really crucial questions about the plume hypothesis are: how much material can be involved in the plume? How much gas is needed to sweep out the observed amount of the outer suevite (taking into account possible substantial erosion)? What is the source of this gas? And why are sediments extremely rare in the “plume” deposits? These questions cannot be answered without detailed physical models, which are the subject of this article.
Pyroclastic Flow Hypothesis
Kieffer and Simonds (1980) suggested that impacts into volatile-rich material could result in the dispersion of impact melt and suevite such that the volume of impact melt deposited in and around a crater could be as much as a factor 100 less than observed in impact deposits at craters formed in crystalline rocks. Following up on this idea, Newsom et al. (1986) referred to the chimney-like degassing pipes traversing the suevites (first described by Hörz 1965 and by Engelhardt et al. 1969), pointed out that these pipes are analogous to features occasionally observed in ignimbrites, and suggested a new mechanism for OS emplacement, similar to a pyroclastic flow. A pyroclastic flow is a relatively dense and hot, topographically controlled, surge of a mixture of poorly sorted, fluidized particulates. It is a typical consequence of a volcanic plume collapse. Similarly, an impact plume (described above) may collapse and may form a flow. Fluidization is an important mechanism allowing pyroclastic flows to extend to large distances (Wilson 1980). The amount of gas (essentially, the gas velocity) needed for fluidization depends on the properties of the flowing materials and, in particular, on the size distribution of particles and their sorting (Wilson 1984). Newsom et al. (1986) used the experimental calibration curve of Wilson (1984) and estimated that the fluidized gas velocity for poorly sorted suevitic material at the Ries should be of the order of 30–200 cm s−1, i.e., one to two orders of magnitude higher than that for ignimbrites. The authors also suggested several possible sources of the gas that provided the fluidization pressure. A possible internal source could be the release of steam from shock-heated and melted lithic clasts. They argued that basement rocks contained up to 0.8 wt% of water and sedimentary rocks up to 2 wt%. Their qualitative estimates lead to approximately 715 L of steam in each degassing pipe, i.e., degassing could have been sustained for about 2 h (if the flow rate was approximately 100 cm s−1). External sources could have included (1) gas trapped during the formation of suevite, (2) gas incorporated at the front of a moving suevitic flow, (3) gases released by heating of water in Bunte Breccia (see also Wittmann and Kenkmann 2009). The first source may be available if suevite was formed by collapse of an eruption-like column. The fluidization as well as the pipes' growth occurred mainly after suevite deposition.
Newsom et al. (1990) developed this model further, trying to explain the delay in deposition of OS until after deposition of Bunte Breccia by an influence of the atmosphere. Namely, a rising fireball caused a strong air flow toward the crater center, which, in turn, swept suevitic material from the surface and into the atmosphere. Later, a base surge may be formed due to the fallout of high-angle ejecta and collapse of the debris cloud (as observed in large weapon tests at Bikini atoll, e.g., Glasstone and Dolan 1977). In the same study, they suggested an alternative explanation involving the layered nature of the target and two separate phases of the “impact explosion”: the initial phase produced the BB, which had been ejected at low ejection angles; and the second occurred during the excavation of the crystalline basement. Indeed, such double ejecta curtains had been observed in experiments (Quaide and Oberbeck 1968) and have been modeled (Lindström et al. 2005; Senft and Stewart 2007) in layered targets with a strong contrast in physical properties (e.g., water/ice overlying crystalline basement). However, it is not clear (i.e., without numerical modeling of the cratering process for the Ries case) why sedimentary rocks differ from the basement rocks so strongly to explain the observed hiatus in deposition. Even if one assumes such a strong contrast, the morphology of the resulting crater would be significantly different from the surface expression of the Ries.
Impact Melt-Flow Hypothesis
More recently, it was proposed (Osinski 2004; Osinski et al. 2004) that the Ries impact melt rocks and outer suevites had been emplaced during the modification stage of crater formation as ground-hugging impact melt flows. This model is similar to the model of a pyroclastic flow except for one substantial detail. Namely, the authors claim that the OS was deposited at extremely high temperatures, i.e., not as a pyroclastic flow, but as a melt flow. Their arguments include (1) well-preserved nonfractured glass bombs (in contrast, “numerous brittle fragments” mentioned by earlier publications) showing evidence of flow after deposition; (2) intimate mixing of different melt phases during transport; (3) decarbonization of some limestone clasts after deposition requiring temperature >1150 K. The authors argued that similar melt flows routinely occur on the Moon and on Venus. They invoked an old idea of Hawke and Head (1977) that the exterior melt flows on the Moon were emplaced during the modification stage of complex crater formation, allowing a temporal hiatus between deposition of ballistic ejecta and the overlying melt. “During the modification stage, the floor of the transient crater at Ries was uplifted, resulting in the so-called inner ring of uplifted crystalline basement. Movements associated with this uplift could have imparted an outward-directed vector, resulting in the transportation of some of the melt phases still within the trace of the original transient crater outward as flows and toward and beyond the final crater rim” (Osinski 2004). The idea seems questionable as extremely high temperature (and, hence, extremely low viscosity) of the flow is not verified and is in a serious contradiction with matrix texture (Stöffler et al. 2013) and with previous temperature estimates (Engelhardt and Graup 1984). Moreover, properties of melt flows routinely observed by volcanologists as lava flows are not in favor of this model: (1) substantial displacement of lava flows against local topography (i.e., uphill to distances of tens of km) has never been observed in volcanology and seems to be in contradiction with the fundamental physical law of energy conservation; (2) lava flows leave distinct flow patterns on the surface, not a patchy deposition; (3) small-scale flows may occur after ballistic (or nonballistic) deposition known in volcanology as rootless lava flows (Head and Wilson 1989); (4) 10 m thick ignimbrites deposited at elevated temperatures reveal substantial welding (Riehle 1973). As the similarity between lava flows and impact melt flows may be restricted due to their different physical properties, the hypothesis has to be verified by numerical modeling of an impact melt flow from the central uplift.
Numerical Models and Initial Conditions
We used two different hydrocodes, SOVA (Shuvalov 1999) and iSALE (Amsden et al. 1980), coupled with the ANEOS (Thompson and Lauson 1972) equation of state, to describe the kinematics and thermodynamic behavior of materials during crater formation, plume expansion, and ejecta deposition. In a recently performed suite of benchmark and validation tests, it was demonstrated that SOVA and iSALE produce results comparable to those of other hydrocodes and are capable of reproducing laboratory shock and crater formation experiments (Pierazzo et al. 2008). With both codes, we used a tracer (massless) particle technique to reconstruct dynamic (trajectories, velocities), thermodynamic (pressure, temperature), and disruption (strain, strain rate) histories in all parts of the material flow. The values of maximum shock compression recorded by these tracers allow us to calculate an amount of melt produced in the Ries event, while their ejection velocities combined with ballistic equations provide information on the ejecta distribution around the crater. We used two numerical codes for the following reasons: (1) as the mechanical and physical processes during crater modification substantially differ from the dynamics of the generation and expansion of the ejecta plume, we applied the iSALE code, which is optimized for simulating crater formation, and the SOVA code, which is optimized for simulating ejecta plume and nonballistic ejecta; and (2) shock melting, vaporization, and ballistic ejection can be calculated with both codes, which allows direct comparison of the models to check the reliability of the results.
To model the formation of the Ries crater, we used the two-dimensional cylinder-symmetric “hydrocode” iSALE (for a brief description of the history of development of the code by different authors, see Wünnemann et al. 2006). The model is based on the original Simplified Arbitrary Lagrangian Eulerian (SALE) hydrocode (Amsden et al. 1980) that was developed to model fluid and gas flow at all speeds. To simulate hypervelocity impact processes in solid targets consisting of layers with different material properties (in the case of the Ries, at least one sedimentary layer needs to be distinguished from crystalline basement), iSALE includes a sophisticated elasto-plastic constitutive model, a fragmentation model (Collins et al. 2004), and an interface tracking routine to separate different materials to avoid artificial numerical diffusion across material boundaries. We assume temporary weakening of the target rocks during crater collapse by acoustic fluidization (Melosh 1979) to explain the crater collapse. The implementation of the acoustic fluidization model (block model) is described in Melosh and Ivanov (1999) and Wünnemann and Ivanov (2003). The cylindrical geometry only allows modeling of vertical impacts into layered targets. The grid resolution in all iSALE models is approximately 20–40 cells per projectile radius (CPPR), with cell size varying between 20 and 40 m.
SOVA (Shuvalov 1999) is a two-step Eulerian code that can model three-dimensional, multimaterial, large deformation, strong shock wave physics. It is based on the same principles utilized in the well-known CTH code (McGlaun et al. 1990). It includes a general treatment of viscosity for modeling viscous flow with Newtonian or Bingham rheology, while the implementation of the rigid-plastic model (Dienes and Walsh 1970) allows us to model material strength (Shuvalov and Dypvik 2004). The initial spatial resolution is usually 20 CPPR near the impact point. By doubling the length of the spatial increment a couple of times, we increase the size of the computational domain to cover the expanding impact plume and the growing crater, while keeping the number of computational cells constant. Apart from modeling crater formation and plume expansion (vapor phases in the plume are generated due to shock-induced partial vaporization of the involved materials), SOVA is also capable of describing the interaction of solid/molten ejecta with vapor/atmosphere in the context of multiphase hydrodynamics (Valentine and Wohletz 1989). Materials that are ejected from the crater and experience a certain amount of tension (i.e., with densities substantially lower than normal density for a given temperature) are no longer treated as a continuum (as in standard hydrocode models), but turned into a cloud of representative particles with a specified size-frequency distribution (SFD). The size-frequency distribution of these particles is assumed to be a function of the peak shock pressure and covers the size range from a few micrometers to meters in size (for details, see Artemieva and Ivanov 2004). Each particle exchanges energy/momentum with the surrounding vapor phase; however, direct particle–particle interaction is neglected in our model assuming that the solid/gas volume ratio is low (i.e., the solid/vapor mass ratio is <1000). The method has been successfully applied to model dusty flows (Shuvalov 1999), tektite formation and deposition (Stöffler et al. 2002), and the development of Chicxulub distal ejecta (Artemieva and Morgan 2009).
For the Ries model calculations, we simplify the target to a 600 m thick layer of calcite, representing the sedimentary cover, and a granitic layer underneath, representing the crystalline basement. To mimic possible water-saturated pores in the sedimentary rocks and to maximize vapor production, we also use EOS of a calcite/water mixture, which was constructed in a way similar to a basalt/water mixture (Pierazzo et al. 2005). We assume a stony nonporous projectile (granite EOS) impacting the target at 18 km s−1 (average asteroid impact velocity on Earth, e.g., Ivanov 2001). Granite is not a good analog of asteroidal material; it was chosen because both numerical codes are restricted to three different materials and, hence, the same materials have to be used to describe the projectile and the basement (two other materials are the atmosphere and calcite = sediments). Specific properties of the projectile are not of crucial importance in crater modeling as long as we do not consider specific chemistry of impact rocks and their contamination by a projectile. No unequivocal projectile signature has been found in the Ries crater so far, although an aubrite-like (achondritic) asteroid has been proposed (Morgan et al. 1979; Horn et al. 1983; Pernicka et al. 1987; Tagle and Hecht 2006). The projectile diameter varies from 1.1 km (vertical impact) to 1.5 km (30° impact) to keep the transient cavity volume constant (assuming that only the vertical component of the impact velocity contributes to the size of the transient cavity, Chapman and McKinnon 1986). These projectile parameters correspond approximately to the parameters used in previous models of Ries crater formation (e.g., Stöffler et al. 2002; Wünnemann et al. 2005; Collins et al. 2008a).
Results—Ries Crater Formation
In this section, we summarize the principal results related to the formation of the Ries crater as obtained in our 2-D and 3-D models.
Shock Compression and Melting of the Target Rocks
Table 1 summarizes the results regarding melt and vapor production for various impact scenarios. In the ANEOS package, incipient melting occurs at 46 GPa and melting is complete at 56 GPa (Pierazzo et al. 1997). However, in this article, we use a value of 60 GPa for melting of the crystalline basement as defined by experiment-based shock classification of crystalline rocks (Stöffler 1971; Stöffler and Grieve 2007). Vaporization of dry nonporous granite starts at 150 GPa. According to the EOS in use, the vapor content in the melt-vapor mixture is about 2.6 wt% at a shock compression of 180 GPa, and 6.2% at 250 GPa. The updated EOS (Melosh 2007) takes molecular clusters in the vapor phase into account. As a result, the critical shock compression allowing vaporization decreases, and the mass of vapor for a given shock compression is approximately doubled. New experiments (Kraus et al. 2012) revealed an even lower pressure value for incipient vaporization in silica; however, this updated EOS is not ready yet to be used in numerical models.
Table 1. Modeling results for the Ries impact event
In this column, numbers in parentheses are for a lower resolution of 20 CPPR, numbers without parentheses for a resolution of 40 CPPR. The former underestimates melt production in the iSALE (first-order accuracy) code. All SOVA models have a resolution of 20 CPPR; however, this method has second-order accuracy and gives reasonable results even at 20 CPPR. In other words, comparison should be made between 40 CPPR iSALE runs and 20 CPPR SOVA runs.
CB (crystalline basement) is represented by the ANEOS equation of state for granite. Figures in parentheses show the net vapor production obtained when an updated version of the ANEOS (Melosh 2007) is used.
Although the volume of melt relative to the projectile volume (Vpr) decreases with decreasing impact angle (from 15 × Vpr for a vertical impact to 9 × Vpr for a 30º impact, see Table 1), the absolute melt volume increases from 10 to 16 km3. The volume of the melt pool within the crater (total melt minus ejected melt) varies between 6.6 and 11 km3 (see Table 1 and Fig. 1). Almost all melt generated from sedimentary precursors is ejected from the crater; its fraction in the melt pool never exceeds 9%, i.e., intracrater melt should be mainly of crystalline basement (CB) composition with minor addition from the lowest sedimentary units. This is in excellent agreement with observations on drill cores in crater suevite, where the sedimentary fraction of lithic clasts is negligible in comparison with crystalline basement-derived clast content.
An extremely small fraction of basement rock is vaporized upon impact. The fraction of this vapor in the ejecta is extremely low (<0.3%) and cannot modify the ejecta pattern (see below). Also, its fraction in the melt pool (<0.5%) neither can change the physical state of the melt, nor can it lead to a substantial redistribution of the intracrater melt.
The SOVA model produces 10–15% more melt than the iSALE model, whereby even less sedimentary rocks are found to contribute to the melt in the SOVA models. The reason may be sought in the target structure (water-saturated porous sediments in the SOVA versus nonporous sediments in the iSALE computations), as well as in the second-order accuracy of the SOVA code versus first-order accuracy of the iSALE (at the highest resolution of 40 CPPR, the code produces 10% less melt than in the ideal “infinite” resolution, see Wünnemann et al. 2008 for detail). We also calculated an amount of crystalline rocks subjected to various degrees of shock compression—see Table 2.
Table 2. Shock compression of crystalline rocks
Shock pressure, GPa
Within inner ring, km3
2D, 90° (iSALE)
3D, 45° (SOVA)
2D, 90° (iSALE)
3D, 45° (SOVA)
These numbers are in a good agreement with analytical estimates: assuming pressure decay P(r) ~ r−2.3 (e.g., Pierazzo et al. 2008) and an approximately hemispherical shape of isobars after a 45° impact at 20 km s−1, shock-compressed volumes may be estimated as follows VI = 8VIV, VII = 0.67VIV, VIII = 0.338VIV (i.e., 68, 5.7, 2.8 km3, respectively, if VIV = 8.5 km3).
Early Ejecta: Tektites and Reuter Blocks
Tektites (moldavites) in the Czech Republic (e.g., Cohen 1961), East Germany (Lange 1990), and Austria (Koeberl et al. 1988) and so-called Reuter blocks found in the Molasse (Gall and Müller 1975) have been identified as the most distal ejecta of the Ries crater. Ejection and deposition of tektites have been studied and modeled previously (e.g., Stöffler et al. 2002). It was shown (Engelhardt et al. 1987) that tektites originate from the uppermost Cenozoic sands that have been shocked to >30 GPa, which corresponds to the melting pressure of porous silica (Kieffer 1971; Stöffler et al. 1975). The distribution of tektites restricts the direction of impact toward the southeast. Here, we present models for the early high-velocity ejecta from the uppermost part of the target to reproduce Reuter blocks (RB). According to observations and petrographic analysis, Reuter blocks originate from the Malmian limestone layer and experienced relatively weak shock compression (although there is no satisfactory petrographic evidence of shock compression in limestone in contrast to siliceous rocks). As these blocks often contain shatter cones (Hofmann and Gnos 2006), the maximum shock compression for their formation may be estimated in the range of 2–6 GPa on the basis of experimentally produced shatter cones (Roddy and Davis 1977). Early ejecta (future tektites and Reuter blocks) are shown in Fig. 2.
To reproduce the deposition of these blocks at a distance of 100–200 km, we first extracted ejection velocities and trajectories for Malmian rock fragments ejected to an altitude above 30 km from the SOVA model. In a second step, we calculated their trajectories assuming an undisturbed atmosphere above this altitude to determine the final deposition. The distribution of RB-like ejecta is shown in Fig. 3. The maximum distance of 300–400 km from the crater is reached in downrange direction (sedimentary rocks with higher shock compression of 6–10 GPa are deposited even further). At an azimuth of 90°, RB may be found as far as 150 km from the crater center. There are no RBs deposited in uprange direction (southwest). After a 30º impact Reuter blocks form a distinct ray pattern (Fig. 3, right plate). All RB originate from the uppermost 50 m of the target, i.e., from the same level as tektites. Tektites were generated closer to the impact point (shock pressure was substantially higher to melt the target) and had sandy precursors. There are no contradictions to these results, as Neogene sands had a patchy distribution near the impact site (Hüttner and Schmidt-Kaler 1999).
Excavation Depth and Ballistic Ejecta
The excavation depth is approximately 1.4 km for a 45° impact, i.e., substantially deeper than the top of the crystalline basement (600 m in our models). The values of shock compression in ejected materials vary over a wide range from a few hundred MPa to several hundred GPa in the sedimentary unit (dashed lines, Fig. 4, left panel). The amount of sediments compressed above 55 GPa (threshold of calcite decomposition, Agrinier et al. 2001) varies from 2 km3 after a vertical impact to 6 km3 after a 30° impact (these volumes were calculated for the initial calcite density of 2.6 g cm−3). Almost all highly shocked sedimentary rocks are ejected as a mixture of CO2 and CaO (see Ivanov et al. 2002 for discussion). These ejecta form an extensive impact plume above the growing crater.
Shock compression in the basement rocks varies in a comparatively narrower interval and never exceeds 150 GPa. This value is insufficient to allow substantial vaporization of granite, which occurs at approximate pressures of 100–150 GPa (Pierazzo et al. 1997; Melosh 2007); however, a significant amount, 7–11 km3, of crystalline material is molten (peak pressures >60 GPa, see Table 1). Thus, basement material is ejected as a mixture of solid and molten materials with minor vapor. Moreover, there is a substantial delay between plume formation and basement entrainment into the ejecta curtain. The former consists mainly of partially vaporized projectile material and surficial sediments.
The maximum excavation depth varies between 1.3 and 1.5 km. The SOVA code produces a larger excavation depth than the iSALE code for the same initial conditions (1.1 km diameter projectile impacting vertically at 18 km s−1). This discrepancy may be explained by two factors: first, the upper sedimentary layer in the SOVA model was less dense (water-saturated calcite); second, the strength model in the SOVA code is more primitive than in the iSALE code and does not work well at the late stage of crater excavation.
The ejection velocity for basement material does not exceed 1.2 km s−1, whereas sedimentary rocks are ejected with velocities up to 10 km s−1 (Fig. 4, right panel), which is required to explain the formation of tektites (Stöffler et al. 2002) and Reuter blocks (Gall and Müller 1975). Nonetheless, maximum ejection velocity from the basement by far exceeds the velocity of 0.3 km s−1 needed to explain the deposition of the Amerbach and Polsingen melt lumps at 10–11.5 km from the crater center, approximately 4–5.5 km outside the inner crater basin. As the maximum ejection velocity of molten basement is 1.2 km s−1 (for a 45° impact, and approximately half that for a vertical impact), i.e., theoretically, such lumps could have been deposited as far as 60–150 km from the crater (neglecting atmospheric drag).
The average shock compression in the ejecta increases with ejection velocity (Fig. 5, left plate). Note, at the same ejection velocity, ejecta from the crystalline basement have been compressed to substantially higher pressure than sedimentary rocks, i.e., high-velocity (approximately 1 km s−1) basement ejecta are mainly molten, whereas sedimentary ejecta have been compressed below 25 GPa. It has to be mentioned that these average values do not represent the whole range of shock compression at a given ejection velocity. For example, fragments of the Malmian sedimentary layer are ejected with velocities above 3 km s−1, but have been compressed below 10 GPa (see the Early Ejecta section).
The total volume of excavated sedimentary rocks is approximately 70 km3, of crystalline rocks approximately 20 km3. Generally, for ejection velocities above 0.3 km s−1, sedimentary rocks dominate in the ejecta at least by an order of magnitude compared with crystalline rocks (Fig. 5, central plate). Although the total amount of basement ejecta reaches 20 km3, less than 2% are deposited beyond the crater rim in a downrange direction. Most probably, these “distal” crystalline rocks are deposited in a mixture with voluminous sedimentary rocks to form the Bunte Breccia. Massive ejecta from the crystalline basement may be found between the inner ring and the tectonic rim. Our models do not allow for separating these ejecta into megablocks and the rocks of the inner ring (some of the observed megablocks certainly slumped off the inner ring). Recent mapping of 25% of the megablock zone resulted in an estimate of 3.2 km3 of crystalline megablocks (Sturm 2011; Sturm et al. 2011). Estimates by Stöffler et al. (2013, table 3), taking into account megablocks within so-called rays, further increase the total volume of megablocks up to 24 km3—a bit larger than the calculated volume of excavated crystalline rocks.
We can also estimate the percentage of vapor in the ejecta, which may potentially lead to nonballistic distribution of ejected materials (see Fig. 5, right plate). Assuming that all vapor originates from decomposed sediment (i.e., limestone subjected to shock compression above 55 GPa), the gas content is >10 wt% at ejection velocities above 1 km s−1 and it is below 1 wt% at velocities below 0.4 km s−1. As the lowest section of the sediment cover (Triassic) is composed mainly of shale and sandstones (not limestone), the real amount of vapor above the growing crater may be smaller than our approximation based on 100% calcite in the sediment column. On the other hand, if some vapor is derived from vaporized water (e.g., pore water in sediments), then gas may be important at ejection velocities above 0.6 km s−1. Even this lower ejection velocity corresponds to deposition at distances close to, or beyond, the outer margin of the continuous ejecta blanket. This estimate confirms that the concept of ballistic continuation is an appropriate simplification to describe near-crater ejecta.
Using this simplification for all ejecta tracers, we calculate the ballistic ejecta thickness around the crater (Figs. 6, 7) that may be compared with the corresponding thickness of Bunte Breccia deposits. Sedimentary rocks prevail at all distances from the crater. Thus, we can conclude that the modeled continuous ejecta blanket is mainly of sedimentary origin and is more than 10 m thick up to 30 km from the crater center. Note, this estimate does not include incorporated local materials due to secondary mass wasting as observed by Hörz et al. (1983). A thinner, but still continuous layer may be found up to a distance of 100 km. Beyond this limit, ejecta are patchy. Moreover, the proximal ejecta were highly compressed (middle plate, Fig. 6) and the ballistic continuation may not be a good approximation. In particular, decomposition of limestone may lead to a substantial amount of vapor and nonballistic deposition. Basement-derived ejecta occupy a much smaller area that is delimited by the final crater rim uprange and stretches up to 30 km from the crater center downrange (Fig. 7). On average, basement ejecta are compressed to substantially higher pressures than sedimentary ejecta (Fig. 7), although within the megablock zone shock compression is <10 GPa (and even less along the inner ring).
Transient Cavity, Final Crater, and Internal Melt Pool
The evolution of a transient cavity for the Ries crater is shown in Fig. 8. Ten seconds after the impact, the cavity reaches its maximum depth of 4.3 km and its maximum volume of 158.5 km3; 8 s later, it reaches its maximum diameter of 9 km and the volume decreases to 142 km3. The inner ring consisting mainly of weakly shocked uplifted crystalline rocks is formed at 30 s after the impact and the crater reaches its final shape at 60 s. Note the structural uplift of about 2 km of stratigraphically deeper seated rock units in the central part of the crater down to a depth of approximately 6 km. Between 60 and 160 s, the topographically not very prominent central uplift collapses (the surface is practically flat), although underlying rocks are certainly strongly deformed. The central part of the crater (within the inner ring) is overlain by a melt pool and highly shocked crystalline rocks.
The total impact melt volume depends on the impact scenario, including, in particular, impact velocity, impact angle, and target porosity. In our nominal impact scenario (impact velocity of 18 km s−1, impact angle 45°, water-saturated sediment), approximately 11 km3 of impact melt has been produced (12 times the projectile volume). Approximately 25% of this melt originates from the Triassic (sandstones and shales, lower 400 m of the sedimentary cover) and another 6% from the Jurassic formations (mainly limestones, upper 200 m of the sedimentary cover). For simplicity, we do not distinguish melting from decomposition here, although a more accurate application of the EOS does allow this. During the excavation stage, approximately 1/3 of the melt is ejected from the crater. Ejected melts have mainly sedimentary precursors, in particular, all “limestone melts” are removed from the melt pool. Consequently, concentration of sedimentary rocks (sandstones and shales) in the final melt pool is as low as 5 wt%—see Fig. 1 and Table 1. At a shallow impact angle of 30°, all basement-derived melt remains within the crater. As we have impact melt lumps with crystalline precursors outside the crater (Polsingen and Amerbach), a very shallow impact angle (30° or less) may be excluded. A vertical (or nearly vertical) impact may be excluded as well—these impacts cannot provide an efficient mechanism to melt surficial layers and, hence, to produce tektites (Stöffler et al. 2002).
The amount of melt in the crater ranges from 6.6 to 11 km3, depending on the different impact scenarios (see Table 1). The maximum value is higher by a factor of 1.4 than the estimated maximum total volume of impact melt contained in suevite and impact melt rocks (Stöffler et al. 2013, table 10). However, the mechanism of melt transformation into suevite is not clear yet and is certainly different from any process included in our standard numerical model. Moreover, the melt volume in suevite is still debated mainly because of the unknown melt content in the suevite matrix (Stöffler et al. 1977; Engelhardt and Graup 1984; Osinski et al. 2004; Reimold et al. 2011; Stöffler et al. 2013) and because of imprecise knowledge of the volume of CS. The recent study of the Enkingen drill core (Reimold et al. 2011) revealed that approximately 48% of melt could occur in suevite, i.e., much higher than any other previous estimates based on the analysis of suevites from the Nördlingen drill core (7–21 vol%, Meyer 2011, 2012). A new assessment of the volume of crater suevite and of the volume of the melt in it has been made by Stöffler et al. (2013). Their maximum best estimate of the melt content of suevite (approximately 5.1 km3 in crater suevite and 2.2 km3 for outer suevite, assuming conservatively that approximately 20% of the melt is contained in the matrix and is now altered) is within the range of the volume of the melt pool within the crater as calculated by numerical modeling (6–11 km3), if we assume that the melt of both types of suevite (approximately 7.3 km3) is derived from the melt pool (see later section and Stöffler et al. 2013).
The amount of volatiles due to impact vaporization within the melt pool is negligible (see Table 1). Some gas may be produced by thermal degassing of weakly shocked sedimentary rocks (Malmian limestone, shales) within the hot melt. However, the analysis of tracers within the inner crater ring revealed an absence of any nonmolten (i.e., rich in volatiles) sedimentary rocks within the melt pool after crater formation. However, weakly shocked sedimentary material deposited proximal to the inner ring may still contain its original water content. Such rocks could have slumped into the melt pool during collapse of the TC and thereby added more volatiles to the melt pool (see below).
Projectile Material Within the Crater
We use tracers to follow the projectile fate—i.e., its shock compression and ejection from the growing transient cavity. There is an extremely small amount of the projectile compressed below 60 GPa (i.e., remaining solid after the impact) or above 300 GPa (i.e., substantially vaporized after decompression). In all impact scenarios, summarized in Table 1, approximately 75% of the projectile is compressed to 100–250 GPa, i.e., was subjected to partial (and minor) vaporization after release. At an impact angle of 30°, all projectile materials are ejected from a growing transient cavity within the first few seconds after contact with the target in a mixture with sediments. At a 45° impact, 11% of the initial projectile mass remains within the transient cavity at the end of its formation. Our models for intermediate impact angles of 35° and 40° (we run the models only for the first 5 s after the impact) reveal 0% and 2% of the initial projectile mass remain within the TC, respectively. If all this projectile melt/vapor is mixed homogeneously with the melt pool, we may expect up to 1.3 wt% of the projectile material within the melt at a 45° impact. This amount is high enough to be identified by geochemical analysis (see, for example, fig. 2 in Tagle and Hecht 2006). However, it is much lower after a 40° impact and vanishes at lower impact angles. As the projectile material has not been identified in the Ries so far (see Stöffler et al. 2013), the impact angle could have been ≤40° or the projectile could have been of achondritic composition.
Plume Dynamics (3D) and Nonballistic Ejecta
Figure 9 shows a series of snapshots of crater formation stages, the expansion and collapse of the impact plume, and the particle distribution derived from the two lithologies in the model (yellow particles originate from the sedimentary unit, magenta particles are solid fragments from the crystalline basement, and red particles represent molten basement material). As can be seen from Fig. 2, the plume is generated at an early stage of crater formation (before the transient crater is fully developed) and consists of partially vaporized and molten projectile material (light green color) and sediment (light gray color). At 18 s after the impact (Fig. 9), the well-developed plume is restricted by the ejecta curtain, consisting initially of sedimentary fragments (yellow dots in the figure). After 20–30 s, the first crystalline material (magenta and red dots in the figure) emerges above the pre-impact surface primarily in downrange direction. It is a mixture of solid fragments and molten particles. The smallest modeled fragments (micrometers in size) are incorporated into the expanding plume and move nonballistically. However, the majority of crystalline rocks follows ballistic trajectories, as they are large enough, and the high-velocity, high-energy gas has already expanded into the upper atmosphere. Within approximately 90 s after impact, the ejecta curtain, consisting of material from both lithologies, is deposited around the crater. Two minutes after impact (approximately the same time is needed to form the final crater), the plume becomes buoyant and leaves the crater site. Some particles are still suspended in the restored atmosphere and may be further redistributed by atmospheric flows (this process is not modeled here). The total volume of these particles is extremely small: approximately 4 km3 of sedimentary material and 0.02 km3 of basement material. In principle, these particles could form “true” fallback ejecta (as observed in the Bosumtwi crater at the very top of the impact breccia crater fill—Koeberl et al. 2007); however, a similar unit has not been identified in the Ries crater so far.
Ejecta deposited during the first 2 min after impact are discussed with Figs. 10-12. For these figures, we used representative particles interacting with the atmosphere and the plume, not ballistic continuation as in the previous section. Ejecta thickness derived from the calculations concerning plume dynamics is substantially less than in the ballistic case (Figs. 10 and 7, respectively). There are several reasons for this discrepancy: (1) only ejecta above 2 km altitude have been transformed into particles, whereas low-velocity ejecta below an altitude of 2 km have been modeled as a continuum and their thickness cannot be resolved in the model (these ejecta have to be deposited at distances larger than 8 km from the crater center, approximately four times the maximum altitude); (2) sediments compressed above calcite decomposition pressure of 55 GPa have been excluded; (3) small particles comprise a substantial part of these ejecta and have been additionally redistributed by atmospheric flows. However, similar to the ballistic calculations, ejecta from the crystalline basement are deposited mainly in downrange direction to a distance of 40 km, i.e., a bit further than in the case of ballistic deposition.
Figure 11 shows ejecta properties along three azimuths—downrange (0°), at 45º, and at 135º to the trajectory of the impactor (averaged over 15º-sectors). Ejecta thicknesses are almost identical for all three directions, as it is expected for low-velocity ejecta surrounding the crater. Deposition velocities are similar for two downrange directions and lower for the uprange direction. In contrast to the ballistic results (Fig. 6, bottom plate), velocities do not increase with increasing distance from the crater, i.e., they slightly increase up to a distance of 20–25 km, and then decrease again. This means that the nonballistic component of ejecta deposition is substantial at large distances, where the ejecta thickness is of the order of 1 m. This result seems to be in contradiction to the theory of ballistic sedimentation (Oberbeck 1975), claiming that due to high deposition velocity, the proportion of local material entrained in the ejecta deposits increases with increasing distance from the crater. However, it is consistent with the observations of Hörz et al. (1977, 1983)—the proportion of locally derived clasts increases in the range from 16 to 25 km and remains constant beyond this limit (see fig. 9 in Stöffler et al. 2013). As we did not model the process of secondary mass wasting during the landing of ejecta on the surface, we cannot reproduce these observations in more detail.
Finally, the bottom plate of Fig. 11 shows the value of average shock compression in the fragments derived from the crystalline basement. At least for a downrange direction, where the total amount of these fragments is high (20–30%, without taking local materials into account), it increases with increasing distance. Moreover, some clasts have been shocked above 60 GPa (not shown in the figure, as averaged values are below maximum values). For a 45º azimuth, fragments from the basement may be found up to a distance of 20 km (<10% of the total mass) and the value of maximum shock compression does not depend on distance. There are very few (<1%) basement-derived fragments deposited in uprange direction. Although there have not been any observations of crystalline clasts of shock stage higher than II (>45 GPa) in the Bunte Breccia, in our models, we do have highly shocked crystalline particles incorporated into these deposits (mainly downrange). This result may be explained by a reduced thickness of the sedimentary sequence applied in the model (600 m in the model versus 700–800 m, which was assumed for most parts of the impacted area). Obviously, a thicker layer of sediments leads to a decrease in the mass of basement ejecta in the BB. The presumed relief of the Ries target with an escarpment of several hundred meters approximately parallel to the direction of impact may have affected the distribution of highly shocked ejecta from the crystalline basement.
More detail on ejecta deposits may be found using so-called numerical outcrops: In our numerical models, particles representative of ejected material (the total number of these particles is usually in the range from 100 to 10,000, depending on a distance from the crater), landing in a sample area of 1 × 1 km at a given distance and direction from the point of impact, contain all information about the source region (lithology), maximum shock compression, final velocity, and time of deposition, which is equivalent to a specific depth within the ejecta blanket. Thus, the sample region can be considered as a vertical profile through the ejecta blanket similar to a geological outcrop or a drill core. However, the incorporation of local material (secondary mass wasting) is not taken into account. The results are presented in Fig. 12. At any distance from the crater, ejecta are deposited quickly, within 10–20 s after impact, although some small particles arrive much later and may create a thin graded upper layer. These late particles are excluded from the analysis presented in Fig. 12. Up to a distance of 15 km from the crater center, there is a lot of basement material incorporated into ejecta deposits, which we interpret here as Bunte Breccia. Moreover, at some horizons, basement fragments are the main component of the sequence. This result seems to be inconsistent with all known observations, showing that the percentage of crystalline clasts in Bunte Breccia is only 3–10 vol% (Ackermann 1958; Pohl et al. 1977). However, there are no good data for crystalline rocks in Bunte Breccia deposits for distances smaller than 15 km. Moreover, 99% of the modeled crystalline fragments originate from a depth shallower than 800 m, and 45% from less than 700 m. In other words, if the real sedimentary sequence at the time of impact was 800 m thick, the total amount of basement material in the numerical outcrops would have decreased to a negligible value, and if it was 700 m, basement-derived material would be less than half of the currently computed values. On average, crystalline fragments in these outcrops have been compressed during the impact to 10–30 GPa, and sedimentary rocks to much lower pressures of 2–3 GPa. Median radii of the fragments vary with depth from 1 mm to 10 cm without any gradation, and deposition velocities do not correlate with fragment sizes (not shown in the figure). Both results favor ejecta deposition en masse, in a turbulent regime. We also do not observe any separation between basement-derived materials and sedimentary clasts, although the concentration of the basement material is higher at the bottom of numerical outcrops and nearly zero at the top (Fig. 12, left plate).
According to our estimates, the maximum thickness of nonballistic ejecta (plume deposits) inside the crater is no more than 1–2 m. This “true fallback” component is deposited during the first few hours (hence, is not shown in the figure) after impact and consists mainly of small solid particles and melt droplets derived from the sedimentary sequence. Particles that are suspended in the atmosphere somewhat longer than 2 min (final time in the model) may add another 0.5 m to the layer (and only a few mm of this originates from basement rock). This ejecta layer may be preserved within the crater if there would not be any intense postimpact processes active after crater formation, besides quiet sedimentation. For example, such a layer has been found in the Bosumtwi crater drill core (Koeberl et al. 2007), of less than 30 cm thickness. However, these “true fallback” ejecta have not been identified in the Ries crater so far; this fact may be interpreted as an indirect evidence for intense postimpact processes. Indeed, there is strong observational evidence for intense reworking of the upper section of CS by fast and heavy floods (Jankowski 1977b; Füchtbauer et al. 1977).
Summary of Numerical Results
Standard models of impact cratering discussed above allow us to reproduce some typical features of the Ries crater, namely:
Crater shape, size, and morphology, including uplifted rocks in the crater center, but a minimal topographic uplift.
Composition and extension of the continuous ejecta blanket.
Proportions of shocked basement clasts and unshocked sediments within Bunte Breccia.
Maximum radial extension of specific strata of the pre-impact stratigraphy as a function of their depth in the target including the observed inverted stratigraphy.
Some of our results detailed above are in contradiction with observations and/or with previously proposed qualitative hypotheses of the Ries suevite formation:
The impact plume above the crater consists mainly of a sedimentary vapor/melt mixture, containing very little material in comparison with the ejecta curtain.
Thickness of true fallback inside the crater does not exceed a couple of meters.
At the end of the modification stage, the crater floor is covered by a thick layer of impact melt with a total volume of 6–11 km3.
Ejecta from all stratigraphic units are transported ballistically; no separation between sedimentary and crystalline rocks occurs. However, the mass fraction of crystalline rocks strongly decreases with radial distance as observed in the Ries.
In summary, our current understanding of the formation of a vapor plume and ballistic ejection of material based on a more or less standard model of crater formation cannot explain many of the observed features at Ries crater. The model predicts a pool of coherent melt inside the inner crater that does not seem to exist at Ries; neither drilling nor geophysical exploration revealed any evidence for the existence of a coherent melt sheet inside the crater. The model also fails to simulate the formation of the suevite and the upper part of the ejecta layer as the result of ejecta plume collapse. These observational facts which contrast the model are clearly not due to any errors in our models, but can be only explained by the lack of some physical processes that play an important role during crater formation or shortly thereafter, and that have not been considered in our current model and in any of the previous models.
Quantitative Assessment of Known Suevite Formation Hypotheses
To test the previously proposed hypotheses for the genesis of suevite, we first carried out a series of simplified model calculations, which are not directly connected to Ries formation, but describe similar physical processes and thus allow for a quantitative feasibility assessment. The different “geological” hypotheses for the formation of suevite have been briefly described in the introduction, namely (1) OS and upper part of CS represent fallback ejecta from a plume, whereas the lower part of CS was deposited as a ground-hugging flow (Stöffler 1977; Engelhardt and Graup 1984); (2) impact melts and suevites were deposited as nonviscous melt flows during crater collapse (Osinski 2004; Osinski et al. 2004); (3) OS is a product of plume collapse and was deposited as a pyroclastic flow, i.e., OS is similar to ignimbrite deposits known from volcanology (Newsom et al. 1986, 1990; Branney and Brown 2011).
Plume Concept: The Role of Gas and Particle Size in Nonballistic Ejecta Transport
It has been widely accepted for decades that the bulk of the suevite represents a product of deposition from a collapsing ejecta plume (Stöffler 1977; Engelhardt and Graup 1984; Melosh 1989). Accordingly, the bulk of the suevite deposits (inside the inner basin, in the megablock zone, and beyond) were assumed to reside initially in the ejecta plume, suspended in the atmosphere for some time. If this was the case, then the “plume material” would have had to stay long enough (approximately 2 min) in the atmosphere to allow deposition of ballistically ejected material (Bunte Breccia) before the vapor plume particles rained down at relatively low velocity on top of the continuous blanket of ballistic ejecta. To confirm this assumption, at least four conditions would have to be fulfilled: (1) a substantial mass of ejecta has to be involved in the plume; (2) melt and lithic clasts from the crystalline basement have to dominate these ejecta; (3) the retention time of particles in the plume has to be sufficiently long to enable the early deposition of ballistic ejecta; and (4) the deposition velocity has to be low enough so that the observed sharp contact between Bunte Breccia and suevite was not eroded/disturbed, but at the same time fast enough to avoid sorting of the descending material, as there is no obvious evidence for gradation of particle sizes observed in both crater suevite and outer suevite.
Under vacuum conditions (or at low atmospheric pressure), ejected solid and molten fragments move along ballistic trajectories and form the ejecta curtain as observed in laboratory experiments, numerical models, and in nature (e.g., Gault 1973; Stöffler et al. 1975; Wünnemann et al. 2005). Only a minor amount of highly shocked target material may be involved sometimes in an early plume (probably related to the presence of atmosphere) and later covers the crater floor (Stöffler et al. 1975). In contrast, hot gases/vapors create a hemisphere above a target plane, i.e., tend to expand isotropically. If ejected material is a mixture of solid and molten particles with vapor (or if the atmosphere is present), the trajectories of at least some part of the material may be affected by gas–particle interaction in the expanding gas plume, and the final deposition may significantly deviate from pure ballistic estimates. This process depends at least on two factors: the gas/solid mass ratio and particle size distribution (ejection velocity may be another one). To quantify gas–particle interaction, we have modeled a simplified ejecta curtain: at a distance of 5 km from a fictitious crater center, a mixture of gas and particles is ejected with a constant velocity of V =0.5 km s−1 at an angle of α = 45° to the horizon. The process lasts 10 s, and the total ejecta volume is 7.5 km3. We carried out 20 runs in total and varied the particle size from 0.1 mm to 1 m, and the gas/solid mass ratio from 10−4 to 0.1. In each test run, all particles have the same size and the gas/solid ratio is fixed. Assuming pure ballistic motion, where the single force affecting the ejection trajectory is terrestrial gravity with g =9.8 m s−2, the travel distance is given by V2/g ≈25 km corresponding to a distance from the crater center of 30 km (25 + 5 km). At this distance, the total ejecta thickness averaged over a 1 km wide ring will be approximately 40 m. Indeed, the largest particles mixed with a negligible amount of gas (10−4) move ballistically (Fig. 13, black curve). Under the same conditions, small particles are substantially redistributed by the atmosphere (Fig. 13, gray curve), although they are deposited rather close to the “ballistic” distance. If the gas/solid ratio is high, the trajectories of the largest particles are substantially disturbed, whereas the smallest particles are redirected to the crater center where they are deposited, eventually (Fig. 13, right panel). Figure 14 shows the deviation of ejecta from the ballistic distance (25 km) for all modeled combinations of particle sizes and gas/solid ratios. Obviously, ejecta move nonballistically if ejected fragments are small and the gas/solid ratio is high. These conditions are typical for early ejecta with a high proportion of partially vaporized and molten materials. This means that it is exactly these early ejecta that create an impact plume and move within it. The opposite situation (large fragments, low gas content) is typical for late ejecta, which move within the ejecta curtain and are deposited ballistically.
If ejecta consist of particles with two significantly different sizes (1 mm and 1 m) and the amount of gas is high (0.1), separation of ejecta into ballistic ejecta and plume ejecta takes place (Fig. 15). Small fragments are deposited mainly inside the fictitious crater, whereas large fragments create the ejecta blanket (Fig. 16). However, this scenario is not typical for a real impact event, in which a size-frequency distribution of particles is a continuum with maximum particle size gradually increasing with decreasing ejection velocity.
This simplified approach confirms the results of our Ries models, described in the previous section: plume ejecta exist at the earliest stage of crater formation and consist mainly of extremely small particles of partially vaporized/molten projectile and sedimentary rocks. Late coarse ejecta are incorporated into the ejecta curtain and are deposited ballistically.
Plume Concept: Collapse of a Fictitious Plume
Let us assume that somehow, in contrast to the results presented above, a substantial amount of ejected material is not deposited ballistically, but instead is “trapped” within a plume above a growing crater. Although this contradicts our statements of the previous paragraph, we want to test whether the plume concept can serve as an explanation for the formation of suevite, if we significantly underestimate the amount of particles laden in the plume. We model the collapse of this plume with two goals: (1) to check the deposition velocity of the particles and their sorting by the atmosphere, and (2) to model the formation of density currents, which would explain a patchy distribution of outer suevite. We ran two variants with the total mass of particles equivalent to a 10 m thick deposit (“dense plume”) and to a 1 m thick deposit (“rarified plume”). Even the dense plume is not massive enough to reproduce the upper suevite in the Ries, which could be a result of an impact plume collapse as suggested in the 1970s. In both variants, we started with a simple size distribution of particles: the largest particles are 1 m in diameter, particles in the next bin are 10 times smaller (down to 0.1 mm), and the total mass of particles is the same in each of 5 bins. All particles have zero initial velocity and are distributed evenly in the atmosphere from an altitude range from 50 to 100 km. Any hot gas/vapor initially present in the plume has to expand, cool down, and condense at the time of a fully developed plume. Hence, standard “cold” terrestrial atmosphere gives us the upper limit of the available gas (and the maximum possible gas/solid mass ratio). The results are shown in Fig. 17.
“Dense plume” ejecta (bottom panel in the figure) are deposited within 150 s (99% of all particles) with a high final velocity of 300–1000 m s−1 and without grading. Only the upper few tens of centimeters of the deposit are normally graded with particle sizes from 1 cm (at the bottom) to 100 μm (at the top). In the second case (1 m thick ejecta, “rarified plume”), particles reach the surface with a lower velocity of 100–200 m s−1 and are graded from top to bottom (top panel in the figure). While 90% of the particles are deposited within the same time as from the dense plume (approximately 150 s), another 10% (the finest particles, <1 mm) remain suspended in the atmosphere for much longer and, in addition, may be redistributed by atmospheric winds beyond approximately 100 km. Thus, material form a high-density plume is deposited unsorted, but with high velocity. This implies substantial secondary mass wasting and is inconsistent with a knife-sharp boundary between Bunte Breccia and OS. In contrast, material from a rarified plume is deposited with terminal velocity of a few m s−1, and it is inevitably graded. Consequently, true plume ejecta have to form a thin layer and have to be graded. Such fallback ejecta, indeed, have been observed in the Bosumtwi crater (Koeberl et al. 2007): a 30 cm thick normally graded layer that contains accretionary lapilli, microtektite-like glass spherules, and shocked quartz grains, and that is directly overlain by postimpact sediments; the composition of the fallback spherules is very similar to the composition of Ivory Coast microtektites (i.e., both types originate from the uppermost part of the target). The first results of the El'gygytgyn drilling project by ICDP (Raschke et al. 2013; Wittmann et al. 2013) have also shown a small component of impact spherules as a component of the uppermost impact breccia fill of the crater, directly underneath earliest deposited crater sediment. Such a thin, normally graded layer has not yet been observed at the Ries crater.
We do not observe formation of density currents in our simulations, which would support the observed patchy distribution of OS. If particles are distributed evenly (as in our model), they fall to the surface as a more or less homogeneous flow. Instabilities (density currents) could be expected if there was a non-even initial distribution of particles. The final ejecta thickness varies slightly from site to site, but its distribution is certainly not patchy—see Fig. 18. Both impactite layers, i.e., the upper part of CS and all OS, cannot be the result of the collapse of the primary ejecta plume, whether dense or rarified.
Impact Melt Flow From the Central Uplift
Standard models of complex crater formation show consistently the formation of a large melt pool inside the inner crater (see the 'Transient Cavity, Final Crater, and Internal Melt Pool' section). All hypotheses proposed to explain the origin of suevite inevitably have to incorporate the fact that any evidence for the existence of such a melt pool at Ries is lacking (except for the 20 m thick drilled section of coherent melt rock at the inner slope of the inner ring at Enkingen [Reimold et al. 2011], which may be considered either as a lump of melt rock in suevite or as a remnant from a melt pool). To test the proposed hypothesis that OS represents the final deposition of ground-hugging impact melt flow (Osinski 2004; Osinski et al. 2004), we consider a melt layer initially overlying a 1 km high central uplift (two-fold exaggeration of the approximately 0.5 km maximum height proposed by Collins et al. 2008a; their fig. 8).
Note, the Ries morphology and geophysical expression (Pohl et al. 1977; Stöffler et al. 2013) do not show any surface or other expression of a central uplift, although a minor uplift could exist below the lake deposits. However, it has been proposed that a raised peak may have collapsed during crater formation giving rise to the presumed relatively flat morphology of the crater floor in the inner basin (Pohl et al. 1977; Stöffler et al. 2013). Standard numerical models of impact cratering (e.g., presented in the previous section) cannot reproduce a melt flow accurately, as the resolution is too low in comparison with the melt sheet thickness d (> 20 m versus approximately 200 m in the case of the Ries crater). A higher resolution of 20–40 cells per thickness of the melt layer allows for modeling this flow more realistically, although in a simplified manner. In particular, motion of underlying rocks has been neglected in the following calculation, i.e., the central uplift was frozen at its maximum height. Furthermore, the melt has zero velocity in the cells adjacent to rocks (nonslip boundary conditions). As a radial spread of melt results in quick thinning (and, hence, cooling) of the melt, a planar approximation is used to mimic a flow within a channel. Melt cooling due to heat exchange with colder rocks, mixing with lithic clasts, and heat radiation to the atmosphere have been neglected as well. All these simplifications tend to increase the range of the modeled flow in comparison with a more realistic flow. So our simplified modeling may be understood as an upper estimate of a potential melt-flow distance. We varied the viscosity μ from zero to a low value of 100 Pa × s (typical for high-temperature basaltic magma) and up to 107 Pa × s (typical for highly viscous dacitic magma). The Reynolds number, the ratio of inertial and viscous forces in the fluid (Re = ρdU/μ, where U is the flow velocity and ρ is the flow density), is high for a low-viscous flow; thus, we may expect a highly turbulent behavior. Our results are presented in Fig. 19. Indeed, the low-viscosity flow is turbulent (although we cannot resolve this turbulence in detail, many splashes are visible in the figure).
Although the maximum mass-averaged velocity of the low-viscous flow may be up to a hundred of m s−1, the flow cannot reach the crater rim, possibly with exception of a few small splashes. After a few cycles of melt moving outward and inward, the non- to low-viscous melts are deposited on the crater floor. Highly viscous melts are deposited on the crater floor without any outward–inward cycles. A much higher central uplift is required to propel the melt outside the crater. Still, it seems highly improbable that impact melt, even if it is very hot and almost nonviscous, can flow en masse a long distance against the crater rim topography. Thus, the hypothesis proposed by Osinski et al. (2004) is not in agreement with our quantitative models of the dynamic processes presented here. We think we provide convincing data to revise or dismiss the melt-flow model, which is also severely questioned by observational evidence discussed in the companion paper by Stöffler et al. (2013).
Pyroclastic Flow Hypothesis
A pyroclastic flow, resulting from an impact plume collapse, was suggested by Newsom et al. (1986, 1990) to explain the deposition of the outer suevite. However, the numerical results presented in the previous section (Ries Crater Formation) do not support suitable initial conditions for this scenario. In particular, we do not have a massive plume above the crater, which would consist mainly of hot basement material. We do observe an inward atmospheric flow at the end of crater formation (see Fig. 9 at 120 s), but this flow does not sweep up a substantial amount of material from the surface into the atmosphere. Moreover, any massive plume above a crater falls back with high velocity and without substantial lateral spreading (see Plume Concept: Collapse of a Fictitious Plume). Seemingly, substantially more vapor (or other volatiles, or denser atmosphere) would be required to create a long-lasting pyroclastic flow resulting in the deposition of outer suevite.
Natural and Man-Made Analogs of Ejecta Plumes
In the preceding section, we have demonstrated by simplified quantitative numerical models that none of the previously proposed explanations (collapsing ejecta plume, impact melt flow, and pyroclastic flow) for the formation of suevite seems to be adequate. Apparently, some important processes have not been taken into account so far. Before we propose our alternative concept, it is worth comparing the Ries impact plume with those plumes generated by man-made nuclear explosions (mushroom clouds), volcanic plumes, and recently observed plumes after high-velocity impacts.
Mushroom Clouds Formed in the Atmosphere after a Nuclear Explosion
After a shallow-depth nuclear explosion, ejected material forms an ejecta curtain; this curtain collapses and creates a base surge propagating outward while part of the curtain forms a rising plume (Fig. 20). One of the largest man-made cratering events was the Sedan explosion in Nevada with a total yield of 104 kt (United States Nuclear Tests 2000). This explosion produced a 390 m diameter and 100 m deep crater; 1.2 × 1010 kg of soil and rock were lifted into the air, 0.8 × 1010 kg of it falling outside the crater. The Sedan shot resulted in two plumes, rising to 3.0 and 4.9 km, respectively.
The particle sizes in the ejecta ranged from sub-micrometer and micrometer (created by condensation of plasma in the fireball) to 10–500 μm (surface material agitated by the blast wave and raised by the afterwinds), and to millimeter and larger sizes (ejecta from the crater). Large particles precipitate quickly: there were no particles larger than 177 μm in the clouds (the base surge and the mushroom cloud) after the first couple of minutes (Izrael 2002). The amount of smaller particles reached 3 × 107 kg for the Sedan explosion, or approximately 0.38% of the total ejecta. Thus, any long-lasting mushroom cloud consists exclusively of extremely small particles. These particles cannot create any “visible” layer on the surface.
Although extrapolation to the Ries event (1020 to 1021 J, or 105 Mton, or 1 million “Sedans”) is not straightforward, we can use standard scaling laws to estimate the Ries ejecta volume. Very large explosions are expected to scale roughly as the power ¼ if energy and gravity are the only important scaling variables (Melosh 1989, p. 113). This means that the Ries crater should be 30 times larger than the Sedan crater, i.e., approximately 12 km in diameter. This value is in good agreement with the diameter of the transient cavity (see the Results—Transient Cavity section; the size of the transient crater is thought to be the most representative measure for the released impact energy). The total ejecta volume in gravity-dominated craters is roughly proportional to the cubed crater diameter, resulting in 2.2 × 1014 kg of ejecta for the Ries event. Again, this value is in good agreement with the estimated volume of Bunte Breccia (100–140 km3, approximately 3 × 1014 kg). The extrapolation of dust load in the plume (0.38% of the total ejecta) leads to a value of 0.4–0.5 km3, which is at least three times lower than the minimum estimated volume of outer suevite (1.2 km3, see Stöffler et al. 2013, table 3) and fifty times lower than the maximum estimated volume of outer suevite and upper crater suevite (25.6 km3). Moreover, this extrapolation is not well justified, as the entrainment of gas into the cloud assumes substantial interaction of particles with the atmosphere, i.e., the process may be much less efficient in the Ries case.
Although nuclear “cratering” tests have certain similarities to natural impact events and may be used to derive scaling laws for impact cratering, the amount of suevite in the Ries crater by far exceeds the amount of dust that may be involved in a man-made nuclear plume (with similar cratering efficiency). Moreover, suevite particles are much larger than their “nuclear test” counterparts.
Umbrella Region of a Volcanic Plume and Volcanic Fallout Deposits
Huge plumes (umbrellas) rising far above the troposphere are routinely observed during Plinian volcanic eruptions. Here, we shortly discuss the conditions of their origin, particle concentration, and their residence time in the atmosphere.
Long-lasting volcanic clouds exist due to the constant addition of hot material from below, i.e., from the vent. Numerous observations allow for making predictions of two favorable conditions for plume formation: (1) eruption mass rate in the range 106–109 kg s−1; (2) magma water content of 1–3 wt% (Parfitt and Wilson 2008). If the mass rate is higher (or water content lower), a buoyant plume cannot be formed and the eruption column collapses creating intense pyroclastic flows and ignimbrite deposits. Numerical models (Valentine and Wohletz 1989) confirm these parameters.
Pyroclastic particles precipitate with a velocity depending on their size (mm-sized particles with a velocity of a few m s−1, 10 μm particles—with a velocity of cm s−1, at sea level), and their final position on the surface is defined by the initial altitude in the umbrella (which, in turn, is defined by the mass eruption rate and by the gas content in the magma) and by local winds. Cross-wind deposition distances of these particles rarely exceed a few tens of km (see fig. 8.7 in Parfitt and Wilson 2008), whereas the smallest aerosol particles may travel around the planet for months and years.
Volcanic fallout deposits have a few common features: (1) the size of the largest clasts found at any location decreases with increasing distance from the vent; (2) at any distance, a fallout deposit exhibits a range of clast sizes due to turbulence in the umbrella cloud; (3) the thickness of the deposit exponentially decreases with distance from the vent and is usually <10 cm at a distance of a few km from the vent; (4) a deposit may exhibit vertical grading (usually normal but sometimes inverse, due to changes in the eruption mass rate or in local weather conditions).
Are impact conditions during Ries crater growth favorable for “umbrella” formation? The total mass of impact ejecta (3 × 1014 kg) divided by the transient cavity formation time of 10 s (see Fig. 8) gives an average mass rate of 2 × 1013 kg s−1, which is by far higher than the maximum mass rate of plume-forming volcanic eruptions. Moreover, late impact ejecta are mainly cold, contain very little volatiles, and cannot entrain and heat the atmosphere. Plume-forming conditions may take place at the very beginning of impact crater formation, when the mass rate of ejecta is low and the ejecta are hot. Our modeling of the Ries event does show, however, that a plume is formed and that this plume consists of hot and high-velocity, mainly sedimentary rock-derived material, see the Results section). Thus, a pure volcanic plume concept is not applicable for impact crater formation—including that of the Ries crater.
Plumes Produced by Experimental High-Velocity Impacts
Laboratory experiments are restricted to impact velocities of 5–8 km s−1 (bulk shock pressures of 20–40 GPa), and, hence, impact plumes are observed mainly after impacts into highly porous or volatile-rich targets (Sugita et al. 1998; Schultz et al. 2005). These plumes predate ballistic ejecta and consist of molten and/or vaporized materials from the upper layers of the target.
In the Deep Impact experiment, a copper projectile collided with the nucleus of comet 9P/Tempel 1 at 10 km s−1 and an impact angle of 20–35° to the horizon, delivering 19 GJ of kinetic energy (see A'Hearn et al. 2005 for details). The set of observations included (1) a very early (<200 ms) flash associated with vaporization of the impactor and part of the comet; (2) a much brighter, downrange displaced second flash (for at least 120 ms) caused by the fastest ejecta of surface material; (3) the observed plume of hot self-luminous material moving outward at a projected velocity of 5 km s−1 (true velocity of 7–10 km s−1); and (4) a much slower mechanically excavated cone of ballistic ejecta. Thus, this experiment is in good agreement with laboratory experiments and with our knowledge of impact cratering: a plume relates to the very early, hot and fast ejecta, whereas late ejecta obey ballistics principles. Recent analysis of the Deep Impact experiment by Groussin et al. (2010) showed that most (85%) of the impact energy goes into the hot plume, which only represents a very small fraction (<0.01%) of the total ejected mass. The dust/H2O-vapor ratio in the hot plume is approximately 1.3 (Groussin et al. 2010) confirming the simple idea that a plume—whatever its origin—cannot be dominated by solid materials. Again, these recent results confirm that the Ries suevite cannot be a product of impact plume collapse.
New Hypothesis for the Genesis of Suevite
The sharp boundary between the Bunte Breccia and suevite and an extremely low content of sedimentary target rock particles (<1 vol%) in suevite demand (1) a substantial hiatus between ballistic ejecta emplacement and subsequent suevite deposition, and (2) low-velocity, high-turbulent flow of suevite above the ballistic ejecta or after ejecta deposition (Hörz et al. 1983; Osinski et al. 2004). Besides the observed distribution of suevite with the largest volume fraction emplaced inside the inner crater basin and a substantial amount of suevite deposited outside the inner ring and beyond the outer tectonic rim on top of the Bunte Breccia, one of the most striking problems is where the melt predicted to have been generated by all models has gone—or in other words, how the melt pool inside the transient crater evolved into the observed suevite deposits with particles (mainly in the size range of μm to decimeters) of melt inside and outside of the inner basin? Here, we suggest that an important postimpact process—not previously included in crater modeling—could substantially modify ejecta blankets and intracrater melt deposits.
Fuel-Coolant Interaction in Impact Craters
Explosive-like interaction of hot melt and water is observed in volcanic steam explosions (phreato-magmatic eruptions, PMEs) where magma interacts with near-surface water or ice (Colgate and Sigurgeirsson 1973; Wohletz and Sheridan 1983). A similar (but not identical) effect, known as “melt-coolant” (or fuel-coolant) interaction, occurs in foundries and nuclear power plants. In both cases, explosion efficiency depends on the water-melt mass ratio and has a maximum at values of 0.1–0.4, depending on the efficiency of water/melt mixing. “Ideal” numerical modeling, assuming a perfect melt-water mixture, gives a value of about 0.2, and the maximum expansion velocity for a vapor plume loaded with molten particles is approximately 700 m s−1 (Shuvalov and Artemieva 2004). It seems possible that similar effects may occur in terrestrial impact craters and that suevite may be produced by postimpact interaction of water with a melt pool (Artemieva et al. 2009; Grieve et al. 2010; Branney and Brown 2011).
To model this scenario, we start with a 250 m thick and 6 km radius flat layer, in which melt is thoroughly mixed with water and some solid fragments (currently we do not discuss the mechanism of this mixing). The water content varies between 2 and 10 wt%, and the initial temperature of the mixture is varied between 800 K (low-temperature mixture) and 1500 K (high-temperature mixture). We also assume that water is instantaneously vaporized by heat exchange with melt, reaches an equilibrium temperature depending mainly on the melt temperature and amount of involved nonmolten fragments, but has no time and/or space to expand. It means that the initial vapor pressure is about 0.4–0.8 GPa, and an explosive fuel-coolant interaction (FCI) occurs. First, we allow the mixture to expand as a heavy gas (the molecular weight corresponds to a silica-vapor mixture) without phase separation. An initial expansion velocity of this FCI-cloud exceeds 1 km s−1, and the cloud creates weak shock waves in the atmosphere. When the layer expands up to a few km, pressure drops to 3–10 bar, and the vapor volume exceeds the particle volume by an order of magnitude—we switch to a two-phase model.
As described above (Numerical Models and Initial Conditions: SOVA Code), the particle phase is simulated by representative particles and the vapor phase is treated as an expanding continuum. At this moment, the cloud temperature is only a few percent lower than the initial temperature. Now particles move within the gas and exchange momentum and heat with the vapor. The latter is an important mechanism to prevent quick condensation of the expanding vapor cloud and its collapse. The dynamics of the cloud interacting with the atmosphere is shown in Fig. 21 for the low T mixture with a water content of 2 wt%. The first important result is that temperature of the cloud remains practically constant during the first minute and may even increase during the final collapse (the next few minutes). However, bulk deposition of modeled “suevite” occurs during this first minute, whereas the smallest particles are deposited later and, hence, in hotter conditions. As geophysical studies (Pohl et al. 1977; Stöffler et al. 2013) revealed that the suevite temperature after deposition was above 575 °C (and certainly below melting of silicate rocks), our model with low-temperature mixture suits the observations better than the model with high-temperature mixture. Figure 22 shows the density distribution at the lower part of the cloud. The cloud base is 10–100 times denser than the atmosphere, i.e., it is a gravitationally driven flow. At the same time, the volume fraction of particles is 100–10 times lower than the gas volume, i.e., the “dusty flow” approximation still works. After the first minute, the density of the basal layer decreases below atmospheric density. This means that the uppermost layer of suevite (a few cm thick) is deposited similarly to deposits from volcanic clouds, i.e., this layer should be graded.
The thickness of deposits from the FCI cloud as a function of distance is shown in Fig. 23 for three initial values of water content (2, 5, and 10 wt%) and two initial values of temperature. In the high-temperature case and for all modeled water contents, suevite layers are too thin (<80 m) within the inner ring and spread too far from the crater center (up to 26 km for the lowest water content of 2 wt%, and even further for higher water contents). In the low-temperature case and for all modeled water contents, we see reasonable distribution of materials deposited from the FCI cloud (compare with fig. 19 in Stöffler et al. 2013): the thickest layer of particles (>100 m for the lowest water content, approximately 50 m for the highest) is inside the inner ring (at distances <6 km) and corresponds to the upper CS; then, the thickness decreases gradually to a distance of 20 km (25 km for the highest water content) and correlates with OS deposits (90–5 m); beyond this limit, only a few particles, not a layer, may be found. A water content of 5 or 10 wt% results in a rampart shape of ejecta deposits. The rampart moves farther from the crater at higher amounts of volatiles (17 km versus 22 km for 5 and 10 wt%, respectively). For an even better correlation with observations, we would need, first, a thicker initial melt pool or more lithic clasts in the melt, and, second, the temperature should be slightly higher than 800 K.
As the flow is gravity-driven, any nonflat features on the surface may lead to nonhomogeneous deposits (thicker in valleys, thinner on hills). Although particle sizes vary from 10 μm to 10 cm (melt bombs), final deposits are nongraded except for the uppermost 50 cm where the smallest particles prevail.
Source of Volatiles
Graup (1977) presented measurements of water content in various basement rocks found in the drill core Nördlingen 1973. This varies over a wide range from 0.56 wt% in granite to 5.65 wt% in a mixed-gneiss sample. If we calculate an average water content of the crystalline basement rocks on the basis of their frequency distribution in the outer suevite (Engelhardt 1997), we obtain approximately 1.9 wt% (see table 9 in Stöffler et al. 2013). This largely crystal-bound water in the target rocks is released from rocks compressed to shock pressures >45 GPa (Stöffler and Grieve 2007).
On first glance, our results (for the lowest water content of 2 wt%) of Fig. 23 are consistent with water content in the Ries basement rocks. However, two important remarks have to be made. First, it is important to inject water into the melt pool after crater formation, otherwise an explosion does not happen. Instead, higher ejection velocities (Artemieva 2007) of the excavation flow and, finally, a larger crater (for a given projectile size) may occur akin to the experimental cratering results for dry and wet sandstone targets discussed by Kenkmann et al. (2011). Second, we assume instantaneous and homogeneous mixing of two materials, melt and water, which is difficult to explain for the kilometer-scale melt pool envisaged for the Ries crater. Most probably, there were a few explosions separated in time and space, which created a single cloud above the melt pool. Consequently, additional sources of water or other volatiles mixed into the melt pool after its formation have to be identified.
There are three sources for water or water-rich components available after crater formation: (1) rocks slumping into the melt pool from the central uplift which have a relatively low water content; (2) rocks from the inner ring and megablock zone, which are mainly represented by water-rich Triassic sedimentary rocks dominated by (OH)-bearing sheet silicates; (3) surface water from a quickly re-established drainage system inside the structural rim of the crater.
In our models, the inner ring is “non-permeable” and we do not obtain any slumping of sediments into the inner part of the crater. However, nature is more complicated than our model—the inner ring has lots of breaches and is not even well pronounced in the NW-N-NE parts of the crater (see fig. 8 of Stöffler et al. 2013). Thus, water and water-CO2-rich sediments may slump into the hot inner part of the crater releasing their volatiles (water vapor at T > 373 K, or carbon dioxide at T > 1150 K as the result of thermal decomposition of limestone).
We also have some time constraints for the described process. If water, after coming into contact with the hot melt, can escape to the atmosphere, no phreato-magmatic eruption can be triggered. Thus, at least a thin crust of solidified melt should be present. This crust is formed shortly after crater formation due to thermal radiation from the hot melt surface and due to true fallback ejecta, consisting of small (and cold) particles (Onorato et al. 1978). Later, water could have penetrated deep into the layer of still molten material along thermal expansion cracks, giving rise to sudden vaporization similar to phreato-magmatic explosions. The maximum duration of the FCI process is defined by two factors: (1) cooling of melt below the glass transition temperature of approximately 1000 K occurs after 300 yr and does not allow formation of glass bombs observed in the OS, i.e., the OS has to be deposited within this time interval; (2) FCI may continue within the crater during the next 3000 yr—as long as the temperature of the melt pool is above the boiling temperature of water.
Geological Evidence for the Presence of Water in the Crater Shortly After the Impact
In general, suevite is a nonsorted mixture of lithic and mineral clasts and impact melt particles with some secondary matrix mineralization. However, there is a 17 m thick sequence (314–331 m) of normally graded suevite in borehole Nördlingen 1973 (Jankowski 1977a; Stöffler 1977; Füchtbauer et al. 1977) on top of regular suevite. In this graded section, fragment size increases with depth from silt to pebble. Füchtbauer et al. (1977) also described the overlying unit (reworked suevite between 256 and 314.3 m) as flysch-type graded beds deposited under water by strong rainfalls, possibly even debris flows. Considering two possible mechanisms to deposit graded suevite, an airfall or a turbidity current-type flow in water, the authors chose the former.
We are able to demonstrate that “graded suevite” was in fact deposited in water (playa-type lakes) present at the surface in the last phase of the FCI-induced suevite formation. We modeled settling of different particles (grain sizes between 10 μm and 1 cm, with each bin for 10 times larger diameters) through atmosphere and through a water layer to find out how much air/water is needed to grade this mixture via the sedimentation process. The model assumes that the mass of each particle fraction equals ¼ of the total mass, and particles are distributed evenly within the water layer (or in the atmosphere up to an altitude of 10 km). The results are shown in Fig. 24. The atmospheric case (left) shows well-mixed deposits—all particles are present in equal proportions throughout the entire layer, but with a deficiency of the largest particles at the very top end. In contrast, both water cases (50 and 100 m water depth) give well-graded deposits: the largest particles are deposited in the lowest 4–5 m (together with 1 mm particles that were initially close to the bottom); then 1 mm particles form a 2–3 m thick layer with minor presence of fines; and the smallest particles are deposited in the topmost section. Substantial amounts of fines are still suspended in water half an hour after the beginning of the sedimentation process. Thus, the thick graded unit in the Nördlingen 1973 core cannot have been created without the presence of water (or any other dense and viscous environment).
Accretionary lapilli are numerous in crater suevite of the Deiningen, 1001, and Nördlingen 1973 drill holes (Graup 1981; Meyer 2011, 2012). They occur on the top of graded suevite over a core length of approximately 35 m. Sometimes, they are concentrated in distinct layers where they form up to 23% of the suevite. These lapilli are identical to volcanic lapilli and consist of a very fine-grained outer shell and a coarser grained core. Their diameters range from 1.1 to 12 mm (Meyer 2012; Stöffler et al. 2013).
The possible mechanisms of impact lapilli formation have not been discussed in detail in the literature in contrast to their volcanic counterparts that have been studied in great detail. The crucial factor for their formation is the presence of moisture in the eruptive column allowing rapid ash agglomeration during descent of an embryonic lapillus through the eruptive cloud (Moore and Peck 1962). Brown et al. (2009) found that lapilli occur exclusively in the uppermost part of ignimbrites, not in co-ignimbrite ash-fall deposits where, instead, ash pellets (ash aggregates without layering) are abundant. Hence, these authors argued that formation of accretionary lapilli requires substantially higher concentration of ash, i.e., they were formed within volcanic density currents. Still, moisture is a key factor for their formation. Thus, the presence of accretionary lapilli is another indirect confirmation of water presence in (and above) the crater when (or shortly after) CS was deposited by the FCI process
Possible Scenario of the Ries Suevite Origin
Based on our new numerical modeling results, the following scenario can be summarized for the formation of the crater and emplacement of suevite as a function of time:
Projectile/target contact, propagation of shock waves, formation of a primary impact plume, early ejecta, i.e., tektites and Reuter blocks (Fig. 2);
Excavation of a transient cavity (TC) and ejection of Bunte Breccia and megablocks; formation of a melt layer and of an allochthonous layer of fragmented, shocked and unshocked crystalline basement rocks (“melt-poor crater suevite”) overlying the TC floor (Fig. 8, plates for 10 s and 18 s after impact);
Gravitational collapse of the TC (Fig. 8, plates for 30–160 s after impact) and formation of a melt pool;
Partial cooling of the melt pool due to radiation and deposition of true fallback ejecta. Inflow of external water and slumping of water-rich and limestone-rich sediments (from the megablock zone) into the melt pool. Irregular mixing of water and melt, fuel-coolant interaction, thermal degassing of limestone (at temperature >1150 K). Ejection of the products of FCI (lithic clasts, melt particles, and water vapor) and formation of secondary plume(s) leading to the deposition of “melt-rich crater suevite” (Figs. 21 and 22);
Propagation of the FCI-cloud beyond the inner ring and deposition of ignimbrite-like flows mainly within valleys but also on local elevations, including megablocks (Figs. 21 and 22);
Cooling of suevite, formation of degassing pipes, hydrothermal activity (not included in our models).
The suggested scenario and the models presented above allow us to explain the following five main characteristics of suevites presented at the beginning of this article:
The precursor rocks of suevite are mainly crystalline basement rocks, whereas Bunte Breccia is mainly derived from sedimentary rocks;
The components of suevite represent a wide range of shock metamorphism including whole rock melting, whereas Bunte Breccia contains only weakly shocked or unshocked clasts;
Outer suevite is separated from the underlying Bunte Breccia by a knife-sharp boundary and does not show any substantial mixing with the substrate Bunte Breccia;
Outer suevite extends to only 1.8 crater radii, whereas the radial extension of Bunte Breccia is up to 3.3 crater radii;
Both OS and CS are unsorted on a meter-sized scale.
These characteristics can be explained by the modeling results in the following way: Bunte Breccia are ballistic ejecta derived mainly from weakly shocked sedimentary target rocks with a minor addition of weakly shocked crystalline rocks. On the contrary, suevite is formed from the melt pool, which consists mainly of molten crystalline rocks, obviously mixed with more highly shocked lithic clasts.
Furthermore, the FCI model allows a substantial hiatus between BB deposition and suevite emplacement, and shows a highly turbulent, but low-velocity suevitic flow on the top of Bunte Breccia. Continuous ballistic ejecta deposits (Fig. 7) thicker than at least 10 cm reach a distance of 30 km (up to 40 km in downrange direction) from the crater center corresponding to 2.3 and 3.1 crater radii, respectively. FCI deposits (Fig. 23) may be found up to a distance of 20–25 km (1.5–1.9 crater radii).
In addition to the five main characteristics, the modeling results can explain a number of other observations (see Stöffler et al.  for detailed geological data):
Limestone blocks slump into the high-temperature melt pool and initiate FCI. If they are in contact with melt for a long interval, i.e., reside within the melt pool, they are thermally decomposed. For this reason, we do not observe limestone clasts in CS. Clasts of limestone ejected during the FCI are subjected to high temperature for a limited time interval and, hence, may survive within OS deposits.
Large melt bombs observed only in the OS are shaped within an ignimbrite-like flow initiated by FCI. Melt particles in crater suevite are usually much smaller and do not have flow-like features. These differences are directly related to a multiple-explosion-character of FCI within the inner crater ring and to a more quiet flow outside the ring.
The patchy distribution of OS is the result of heterogeneously distributed FCI explosions within the crater and of channelized flows outside.
The presence of a graded layer in CS may be explained by sorting of particles in a water layer. Accretionary lapilli could be formed from the smallest particles in the presence of water vapor and deposited with a time delay on top of graded CS. Double layers of accretionary lapilli may be due to additional slumping within the crater.
Perspectives for a Re-Evaluation of Suevite Formation in Other Terrestrial Impact Structures
A puzzling and so far unexplained observation of impact crater geologists has been that in many midsized terrestrial craters, suevite deposits are found not only on top of an extended melt sheet but also below the melt sheet. The former has been interpreted as a “plume deposit,” i.e., “fallback suevite.” Our numerical models clearly show that the term “plume” has been somewhat misused in the literature and true fallback deposits should be much thinner, in the order of a few meters, and may range in thickness from a few tens of cm in the case of medium-sized craters to about one hundred meters for the largest terrestrial craters such as Sudbury (Grieve et al. 2010). True fallback may be found only in craters with a well-preserved upper impactite layer, i.e., in craters rapidly buried by postimpact sediments under quiet environmental conditions, as observed, for example, in the form of a few cm thick layer in drill core LB-5A from Bosumtwi (Koeberl et al. 2007). Suevite underlying a melt pool seems to be an obvious product of shocked and unshocked materials being mixed during transient cavity growth-collapse and deposited on the top of the crater floor. Although this process cannot be resolved by numerical models over the entire duration of crater formation, this type of suevite should occur in any terrestrial or extraterrestrial impact structure.
In this article, we describe an additional mechanism for suevite formation for those cases in which impact occurs into mixed and/or wet target. This “third” type of suevite may be the result of the interaction of shock-generated melt with water or volatile-rich sediments. The process has so far not been included in numerical models, which have been tested and worked fairly well for lunar cratering and for terrestrial cratering in crystalline rocks, but failed if cratering occurs in an area covered by sedimentary or water layers.
A few recent drilling projects in terrestrial craters (Bosumtwi, Chesapeake Bay, and Chicxulub) have particularly emphasized the problems with impact melt volume and thickness of impactite layers in craters formed in mixed or water-covered targets. In the following section, we present re-interpretation of suevite observations from several terrestrial impact structures in the context of the previously described FCI mechanism. This compilation is certainly not comprehensive, but supports the importance of FCI in terrestrial “wet” target conditions.
Sudbury (D = 200 km)
The 1.4–1.6 km thick Onaping Formation consists of a complex series of breccias and “melt bodies” stratigraphically above the Sudbury Igneous Complex (SIC). Based on the presence of shocked lithic clasts and various “glassy” phases, the Onaping Formation has been described as a “suevitic” breccia, with an origin, at least in part, as fallback material (French 1970) or a combination of fallback with other postimpact processes (Peredery and Morrison 1984; Mungall et al. 2004). The apparent excessive volume of the Onaping Formation and other geological evidence allowed for suggesting a new mechanism of its formation (Grieve et al. 2010): seawater could have entered the inner crater and reacted violently with the underlying impact melt, leading to the initiation of the FCI process. It may be a coincidence, but the total thickness of the final Onaping sequence (1.4–1.6 km) is such that it would create a lithostatic load on the underlying impact melt sheet of approximately 30 MPa that could suppress further water vaporization and, hence, would have terminated the FCI activity. Also, the total thickness of the SIC (2.5–3 km) plus the melt component in the Onaping Formation is in good agreement with the total estimated melt thickness of the postulated primary Sudbury melt pool (Ugalde et al. 2005).
Chicxulub (D = 180 km)
Borehole Yaxcopoil-1 (Yax-1) is located some 60 km south of the center of the approximately 180 km diameter Chicxulub impact structure (between the inner “peak ring” and the crater rim) and extends to a depth of 1511 m (Dressler et al. 2003). Analyses of the drill core revealed approximately 100 m of suevite-type allochthonous breccias that were subdivided into six subunits. The uppermost four units (middle suevite, upper suevite, lower sorted suevite, and upper sorted suevite) with a total thickness of 66 m were interpreted as “plume deposits” (Stöffler et al. 2004). The suevite thickness of Yax-1 is only slightly less than the thickness of suevite within intracrater boreholes, whereas the coherent melt sheet is totally absent in Yax-1 (except for approximately 24 m of brecciated impact melt rock) in contrast to intracrater boreholes that contain hundreds of meters of melt (Stöffler et al. 2004; Fig. 10).
The previously suggested model of suevite deposited through a re-established atmosphere (Stöffler et al. 2004) and the estimated intervals for particle precipitation are inconsistent with a model of plume collapse, as discussed above (see the 'Plume Concept: Collapse of a Fictitious Plume' section): a tens of meter thick sequence would collapse with free-fall velocity preventing any sorting and grading of particles by size, but that was not observed at Yax-1. Additionally, the total thickness of suevites in Yax-1 and the fact that their precursor rocks are mainly crystalline are in contradiction with numerical models of plume deposits (Artemieva and Morgan 2009; Grieve et al. 2010).
Recent marine seismic data acquired across the Chicxulub crater revealed that at the time of impact, both the water depth and sediment thickness varied with azimuth around the impact site (Gulick et al. 2008). If the water depth reached 2 km in the northeast direction from the crater center, water could rush into the northern part of the crater in a few minutes after its collapse (Collins et al. 2008b). This scenario resembles the situation proposed for the formation of the Onaping unit at Sudbury, although the total thickness of suevitic units in Yax-1 is substantially less than the thickness of the Onaping. One argument may be that the uplifted crater rim and ejected material prevented an immediate resurge of seawater into the crater from the south. According to Bahlburg et al. (2010), refilling took place by permeation through and localized erosion of the ring wall and lasted probably much longer.
Popigai (D = 100 km)
At the impact site Archean gneisses were covered by inhomogeneous sedimentary and metasedimentary rocks with a thickness increasing from zero up to approximately 1 km in the SE-NE direction. This crater represents a unique example of an intensively studied crater, where the total length of sampled drill cores exceeds 100 km (Masaitis et al. 1999). These drilling data have enabled the creation of a 3-D picture of the crater subsurface that is strongly asymmetric. Moreover, Masaitis et al. (1999) claim that impactite deposits “were formed by the ejection of separate jets and currents that differ in composition, impulses, and trajectories,” i.e., by mechanisms substantially different from the well-established models of ejecta deposition at lunar craters. Although there is no obvious evidence for the presence of water during Popigai formation so far, sedimentary rocks could have provided volatile-rich fragments to initiate a series of localized FCI explosions. Another important lesson has to be learned from the Popigai crater—one (or even two) drill core(s) cannot be representative for the whole impact structure.
Chesapeake Bay (D = 80 km)
The 85 km diameter Chesapeake Bay impact structure is among the largest and best-preserved impact structures on Earth. It was formed in the late Eocene, at approximately 35.2 ± 0.3 Ma, in a shelf environment (Horton et al. 2005). The Eyreville boreholes (Gohn et al. 2008) were drilled by USGS and ICDP into the moat of the inner crater basin at approximately 9 km distance from the crater center and revealed the following sequence (from top to bottom): 444 m of postimpact sediments, 652 m of resurge sediments, 275 m of basement rocks (also part of resurge deposits), 175 m of impactites (suevite, polymict lithic impact breccia, minor impact melt rock), and 215 m of basement rocks (which may alternatively represent the true crater floor or a giant megablock—Kenkmann et al. 2009). The lower 80 m-thick part of the impactites has been interpreted as ground-surge material, whereas the upper section has been described as a product of the ejecta plume that was deposited simultaneously to ground surging (Kenkmann et al. 2009; Wittmann et al. 2009).
In principle, this interpretation is consistent with the proposed scenario in the present paper; however, one important proviso is required—the above-mentioned plume is not a primary impact plume (which has low density and consists mainly of sediments), but a secondary plume resulting from the interaction of an early water-rich resurge with a primary melt pool. This secondary plume is substantially denser than the primary plume, contains mainly materials from deeper lithologies, collapses within a few minutes, and disperses some impactites from the crater. A late massive resurge (Exmore breccia) suppresses the FCI process. This re-interpretation is more consistent with the modeling results presented in this article and helps to explain the discrepancy between the only observed 175 m thickness of impactites and the modeled 500 m thick impactites expected to form in the course of an impact event of this magnitude (Kenkmann et al. 2009).
Bosumtwi (D = 10 km)
The crater was drilled in 2004 and two cores were retrieved from impactites (Koeberl et al. 2005). The most surprising results of this drilling project are (1) a total absence of a coherent melt sheet that had been predicted by geophysical studies (Plado et al. 2000) and numerical models (Artemieva et al. 2004), as well as a very low melt content in suevite (Coney et al. 2007; Deutsch et al. 2007; Ferrière et al. 2007); (2) a well-mixed sequence of impactites, i.e., the absence of any trends in shock metamorphism of clasts with depth (Koeberl et al. 2005; Reimold et al. 2006; Coney et al. 2007; Deutsch et al. 2007); (3) Koeberl et al. (2007) described an only 30 cm thick, fine-grained layer from drill core LB-5A, which contains accretionary lapilli, microtektite-like spherules, splash-form impact glass, and shocked quartz and K-feldspar, at the contact between postimpact sediments and the underlying impact breccias. The description of this thin layer suggests that it represents true fallback material.
Additional modeling of the Bosumtwi crater (Artemieva 2007) clearly showed that an absence of the melt pool cannot be explained by a very shallow impact angle or by a low impact velocity. On the contrary (see also Table 1 of this article), the melt pool volume increases with decreasing impact angle. If drill cores LB-7A and LB-8A provide representative samples of the crater fill and crater floor, the melt deficiency and intensive mixing of the impactites are most probably due to intensive dispersion of impactites driven by the vaporization of water in the target, although other volatiles may have played a similar role (see also Kieffer and Simonds 1980). This scenario requires that the target rocks at Bosumtwi contained pore water (10 wt% in the model or may be less) to a depth of several kilometers. Observations of decorated PDFs in Bosumtwi samples (Ferrière et al. 2007) have been interpreted to favor an impact into a water-bearing target. Molten ejecta could be intensively fragmented and dispersed and, hence, cannot be easily identified near the crater.
Craters on Mars
For many years, the Ries crater ejecta have been considered as an analog of crater ejecta on Mars. Indeed, 89% of cataloged Martian craters are surrounded by a layered (fluidized) ejecta blanket with approximately 14% having double or multiple ejecta layers (Barlow 2005). The origin of fluidized ejecta and their multiple characters are still debated, with the most popular hypothesis concerning gas/vapor influence on ejecta emplacement. Barnouin-Jha and Schultz (1998) advocated the idea that the Martian atmosphere itself may be a possible source of this gas. A volatile-rich (H2O ice, in particular) target was suggested by others (e.g., Wohletz and Sheridan 1983; Mouginis-Mark 1987). The latter hypothesis is further supported by recent discoveries of water-rich Martian substrates (Boynton et al. 2002), decrease in minimal crater diameter with layered ejecta at high latitudes (Barlow et al. 2001), and an obvious correlation between DLE/MLE craters with topographic lows (Barlow and Perez 2003). The study presented in our article clearly shows that neither the terrestrial atmosphere nor subsurface volatiles lead to double/multiple ejecta layers. Essentially, FCI has been suggested by Wohletz and Sheridan (1983) to explain lobate ejecta on Mars. They wrote: “We conclude that water either on or beneath the surface of Mars interacted with impact melt during crater excavation to produce vapor explosions. These explosions altered initial parabolic trajectories of ejecta to form partly fluidized flows that deposited the rampart and terraced ejecta.” Although it seems questionable to produce “vapor explosions” during the crater excavation stage (instead, higher ejection velocities are observed—see Artemieva 2007; Kenkmann et al. 2011), postimpact FCIs may be responsible for multiple ejecta layers.
In conclusion, interaction of impact-derived melt and water may occur in many terrestrial craters. However, the consequences of this interaction (and, hence, final impactite distribution) depend on many factors, for example, water distribution in target rocks (intrinsic, or surficial), the melt/water ratio, the time delay between crater formation and water influx, or the total amount of water. Currently, we cannot provide a complete quantitative model of the suggested FCI process, but it seems evident that this process has to be taken into account in future studies of terrestrial impact craters. To prove/disprove our ideas, we may suggest the following: (1) intensive collaboration between volcanologists and impact geologists in their field study of suevite and ignimbrite deposits; (2) experiments to model melt/volatiles interaction (specific materials and conditions are needed; however, similar experiments were conducted more than 30 yr ago); (3) numerical modeling of melt/water interaction on a microscale taking into account heat transfer between melt and water and fragmentation of melt, which leads to an increase in the interface area and to an explosive character of the interaction; (4) numerical modeling on a macroscale with sophisticated (mixed) equations of state for the materials involved.
We thank the Deutsche Forschungsgemeinschaft (German Science Foundation) for the generous financial support through projects RE 528/5-1 and -2 during the period 2007–2011. We thank Boris Ivanov for his careful review, which allowed us to improve the manuscript.