ABUNDANT CIRCUMSTELLAR SILICA DUST AND SiO GAS CREATED BY A GIANT HYPERVELOCITY COLLISION IN THE ∼12 MYR HD172555 SYSTEM

The recent Deep Impact hypervelocity experiment, one of the few direct man-made astrophysical experiments on record, produced a mix of materials—volatilized cometary material, silicaceous droplets, and excavated, largely unprocessed bulk material—after a small bolide of ∼370 kg mass impacted the nucleus of comet 9P/Tempel 1 (hereafter Tempel 1) at 10.2 km s⁻¹ on 2005 July 4, when the comet was 1.51 AU from the Sun (A'Hearn et al. 2005; Melosh et al. 2006; Richardson et al. 2007). Due to the very low bulk modulus (<10 kPa) and escape velocity (∼1 m s⁻¹) of the nucleus, the ejected material was heavily dominated, by 2–3 orders of magnitude in mass, by the excavation of bulk unprocessed material, as were the Spitzer emission spectra. This bulk material was excavated from the nucleus from depths as large as 30 m and was largely unaltered, except for disruption of loosely held macroscopic fractal particles into their individual subfractal, micron-sized components, as demonstrated by the presence of significant amounts of nonrefractory water ice in the ejecta (Lisse et al. 2006; Sunshine et al. 2006). Spitzer IRS 5–35 μm spectra were taken within minutes of the impact, both before and after. The ejected material cooled from effects due to the impact within seconds to minutes, and separation of the ejecta emission spectrum from blackbody emission at local thermal equilibrium (LTE) due to large, optically thick dust particles in the ambient coma was easily made. The resulting highly structured spectrum of the ejecta showed over 16 distinct spectral features at flux levels of a few Janskys that persisted for more than 20 hr after the impact (Lisse et al. 2006).

The emission flux from a collection of dust is given by

$$\begin{equation*} F_{\lambda,\bmod } = \frac{1}{{\Delta ^2 }}\sum\limits_i {\int_0^\infty {B_\lambda (T_i (a,r_*))Q_{{\rm abs},i} (a,\lambda)} \pi a^2 \frac{{dn_i (r_*)}}{{da}}da}, \end{equation*}$$

where T is the particle temperature for a particle of radius a and composition i at distance r_* from the central star, Δ is the distance from Spitzer to the dust, B_λ is the blackbody radiance at wavelength λ, Q_abs is the emission efficiency of the particle of composition i at wavelength λ, dn/da is the differential particle size distribution (PSD) of the emitted dust, and the sum is over all species of material and all sizes of particles for the dust. Our spectral analysis consists of calculating the emission flux for a model collection of dust, and comparing the calculated flux to the observed flux. The emitted flux depends on the composition (location of spectral features), particle size (feature to continuum contrast), and the particle temperature (relative strength of short-wavelength versus long-wavelength features). We discuss each of these effects below.

To determine the mineral composition, the observed IR emission is compared with the linear sum of laboratory thermal IR emission spectra. As-measured emission spectra of randomly oriented, μm-sized powders were utilized to directly determine Q_abs. The material spectra were selected by their reported presence in interplanetary dust particles, meteorites, in situ comet measurements, young stellar objects (YSOs), and debris disks (Lisse et al. 2006). By building up a spectral library, we tested for the presence of over 80 different species in the T1 ejecta, in order to approach the problem in an unbiased fashion. We also expected, given the large number of important atomic species in astrophysical dust systems (H, C, O, Si, Mg, Fe, S, Ca, Al, Na, ...) that the number of important species in the dust would be on the order of ∼10.

The list of materials tested included multiple silicates in the olivine and pyroxene class (forsterite, fayalite, clino- and ortho-enstatite, augite, anorthite, bronzite, diopside, and ferrosilite); phyllosilicates (such as saponite, serpentine, smectite, montmorillonite, and chlorite); sulfates (such as gypsum, ferrosulfate, and magnesium sulfate); oxides (including various aluminas, spinels, hibonite, magnetite, and hematite); Mg/Fe sulfides (including pyrrohtite, troilite, pyrite, and niningerite); carbonate minerals (including calcite, aragonite, dolomite, magnesite, and siderite); water-ice, clean and with carbon dioxide, carbon monoxide, methane, and ammonia clathrates; carbon dioxide ice; graphitic and amorphous carbon; and neutral and ionized polycyclic aromatic hydrocarbons (PAHs).

A model phase space search easily ruled out the presence of a vast majority of our library mineral species from the T1 ejecta. Only convincing evidence for the following as the majority species in the Tempel 1 ejecta was found (Lisse et al. 2006): crystalline silicates such as forsterite, fayalite, ortho-enstatite, diopside, ferrosilite, and amorphous silicates with olivine- and pyroxene-like composition; phyllosilicates similar to nonerite; sulfides similar to ningerite and pyrrohtite; carbonates similar to magnesite and siderite; water gas and ice; amorphous carbon (and potentially native Fe:Ni); and ionized PAHs. This list of materials compares well by direct comparison to numerous in situ and sample return measurements (e.g., the Halley flybys and the STARDUST sample return; see Lisse et al. (2007a) for a detailed list). Before HD172555, these seven classes of minerals (Ca/Fe/Mg-rich silicates, carbonates, phyllosilicates, water-ice, amorphous carbon, ionized PAHs, and Fe/Mg sulfides; 15 species in all) were successfully used to model thermal emission from dust emitted by six solar system comets, three extra-solar YSOs, and two mature exo-debris disks (Lisse et al. 2006, 2007a, 2007b, 2008).

Particles of 0.1–1000 μm are used in fitting the 5–35 μm data (although results to date have shown the strongest IRS spectral sensitivity to the 0.1–20 μm particle size range), with particle size effects on the emissivity assumed to vary as

$$\begin{equation*} 1 - {\rm Emissivity}(a,\lambda) = [1 - {\rm Emissivity}(1\,\,\mu {\rm m},\lambda)]^{(a/1\,\,\mu {\rm m})}. \end{equation*}$$

The PSD is fit at log steps in radius, i.e., at [0.1, 0.2, 0.5, 1, 2, 5, ..., 100, 200, 500] μm. Particles of the smallest sizes have emission spectra with very sharp features, and little continuum emission; particles of the largest sizes are optically thick, and emit only continuum emission. Both the T1 ejecta spectrum and the HD172555 spectrum show strong, sharp, high-contrast features, indicative of the presence of small (∼1 μm) grains. For the Tempel 1 ejecta, the PSD was found to be unusually narrow, consisting predominantly of 0.5–2.0 μm particles (Lisse et al. 2006). For the HD172555 excess, a steep power-law spectrum of particle sizes, dominated by small grains, was found to be necessary to fit the Spitzer data. New for this work was the additional requirement of including a separate population of large cool grains with a > 100 μm.

Dust particle temperature is determined at the same log steps in radius used to determine the PSD. Particle temperature is a function of particle composition and size at a given astrocentric distance. The highest temperature for the smallest particle of each species is free to vary, and is determined by the best fit to the data; the largest, optically thick particles (1000 μm) are set to the LTE temperature, and the temperature of particles of in between sizes is interpolated between these extremes by radiative energy balance. We model the unresolved dust excess around HD172555 as a relatively localized dust torus, and a disk as a linear sum of individual torii.

The T1 ejecta were, by experimental design, all highly localized at 1.51 AU from the Sun. We use this fact to empirically determine the effective distance of the emitting material from HD172555. To do this, the best-fit model temperature for the smallest and hottest (0.1–1 μm) particles for each material from our analysis is compared to the temperatures found for the (0.5–2.0 μm) particles of the Tempel 1 ejecta (Lisse et al. 2006), using the relation

$$\begin{equation*} T_{{\rm dust}} = T_{{\rm T}1{\rm ejecta}} ({L_* }/{L_{\rm solar}})^{1/4} (1.51\,\,{\rm AU}/{r_* })^{1/2}, \end{equation*}$$

where T_dust is the temperature of the dust around HD172555, T_T1ejecta are as given in Lisse et al. (2006) (∼340 K), L_* = bolometric luminosity of HD172555, and r_* is the distance of the dust particle from HD172555.

Our method has limited input assumptions, uses physically plausible emission measures from randomly oriented powders, rather than theoretically derived Mie values, and simultaneously minimizes the number of adjustable parameters. The free parameters of the model are the relative abundance of each detected mineral species, the temperature of the smallest particle of each mineral species, and the value of the PSD at each particle size (Table 1). Best fits are found by a direct search through (composition, temperature, size distribution) phase space. The total number of free parameters in the model used to fit the IRS HD172555 spectrum = 11 relative abundances + 11 hottest particle temperatures + 1 power-law size index = 23. (Since the results of our modeling show we could describe the dust temperature using two parameters, in application we actually used only 14 adjustable values.)

The model has allowed us to get beyond the classical, well known olivine–pyroxene–amorphous carbon composition to the second-order, less-emissive species such as water, sulfides, PAHs, phyllosilicates, and carbonates. We are able to determine the overall amounts of the different major classes of dust-forming materials (olivines, pyroxenes, sulfides, water, etc.) and the bulk elemental abundances for the most abundant atoms in these materials (H, C, O, Si, Mg, Fe, S, Ca, Al). Applying our analysis, with a series of strong ''ground truth'' checks of its validity, is highly diagnostic for interpreting mid-IR spectra of distant dusty systems such as YSOs, debris disks, and PNs.

The main difference between the approach used to model the mid-IR spectrum of HD172555 and that used to decompose previously studied cometary- and dusty exo-systems is that the usual linear sum of laboratory thermal IR emission spectra do not fit the HD172555 spectrum. That is, no combination of emission from the Fe/Mg olivines, Ca/Fe/Mg pyroxenes, Fe/Mg sulfides, phyllosilicates, amorphous inorganic carbon, water ice/gas, PAHs, and carbonates commonly found in circumstellar dust, or from the other 80 species in our spectral library was able to reproduce the observed spectrum. A host of new mineral species were studied and compared to the HD172555 spectrum. The majority of these species focused on matching the pronounced emission feature from 7.5 to 10 μm, in the usual range for an Si–O vibration mode. Lunar species such as anorthite and basalt, and all the allotropes of silica available in the literature were studied: quartz, cristoblite, tridymite, amorphous silica, protosilicate, obsidian, and various tektite compositions. Also examined were mineral sulfates (such as gypsum, ferrosulfate, and magnesium sulfate) and oxides (including various aluminas, spinels, hibonite, magnetite, and hematite).

We found the high signal-to-noise ratio of the Spitzer data in the 7–13 μm region to be highly constraining on the possible species present. A clear match to the Spitzer spectrum at the 95% confidence level (C.L.) was found only in the case of a spectrum dominated by emission from the glassy silicas: obsidian and tektites. The presence of obsidian and tektite-like material, produced on Earth as the kinetic product of very high temperature processing of silicaceous materials in lavas and impact craters (i.e., by quick quenching of rapidly melted rocks), was immediately suggestive of a violent origin for the circumstellar material found around HD172555.

Just as interesting was the discovery of two sharp residuals in the emission after all the solid-state emission features had been removed. These features were found with good fidelity at 7.5–10 μm and in a much noisier feature rising toward short wavelengths at 5–6 μm (Figure 4, right panel). A similar interesting excess was found at 8–9 μm in the 8–13 μm ground-based observations of HD172555 by Schütz et al. (2005). We could not find any solid-state material in the literature that matched these residuals. On the other hand, a literature search of possible gas species produced an excellent match with the fundamental rovibrational linear stretch complex of the SiO molecule—it has P, Q, and R branches at 7.5–10 μm, centered at ∼8 μm. The first overtone of the SiO stretch lies at ∼4 μm, and the rovibrational manifolds have transitions extending from 3 to 6 μm. Examination of the transition moment calculations of Drira et al. (1997) and the IRAS/LRS and ISO-SWS measurements of α Tau, an oxygen-rich K-star on the asymptotic giant branch (AGB) with a known SiO absorption feature (Cohen & Davies 1995), demonstrated a good fit to the residuals. Once we adopted an opacity for SiO gas based on ISO-SWS observations of α Tau, we found that we could easily account for the sharp fall in emissivity at 5–6.5 μm, the extremely flat emissivity at 6.5–7.5 μm, and the sharp rise just longward of 7.5 μm.

SiO gas is the species created when olivines and pyroxenes are vaporized in high-temperature processes—the minerals decompose into SiO and metal (FeO, MgO, CaO, AlO, etc.) oxides. The same processes can form silica from the olivine and pyroxene silicates.

A third component we have not had to use in our previous analyses, but find necessary to obtain a good fit to the Spitzer data (especially at the longer wavelengths), is material contained in a very large particle dust population (a > 100 μm), radiating as blackbodies at ∼200 K. As described in Section 2.2.4, for all of our previous analyses of Spitzer bright dusty disk spectra we have been able to fit both the sharp features and the continuum of the spectrum with a power-law size distribution of dust, 0.1 < a < 1000 μm. At 80% relative surface area of all the other solid dust components, the large dust particle population is clearly important in the overall mix of materials. It is, however, difficult to characterize due to its relative lack of spectral structure. No mineralogical information can be derived for the large cold particle population, and only a lower limit for the mass.

The best-fit blackbody function to the total 5–35 μm HD172555 dust excess spectrum has T_bb = 335 K. While extremely useful for quickly removing the zeroth-order effects of dust temperature on the as-observed spectrum, a blackbody fit is a gross simplification, lumping in the emission behavior of multiple chemical species and particle sizes. Our more detailed compositional modeling finds a maximal temperature for the smallest silicaceous dust grains of 305 K (compare to 260 K estimate from IRAS photometry, and a 520 K + 170 K estimate from Chen et al. 2006), for a primary with L_* = 9.5 L_Sun (Wyatt et al. 2007b; Table 1). To determine where the dust lies with respect to the primary, we scale the results of the Deep Impact experiment, where fine micron-sized dust was released at 1.51 AU into the Sun's radiation field and promptly observed by the Spitzer IRS (Section 2.2.5). This technique has worked well in comparison to imaging measures of the dust location. For example, we placed the location of the dominant emitting dust from the inner cavity wall of the HD100546 disk at 13 ± 1.0 AU (Lisse et al. 2007a), and STIS one-dimensional spectroscopic measurements find a cavity for this disk of radius 13.2 AU across (Grady et al. 2001, 2005). Our best-fit model for the HD69830 circumstellar dust ring puts the dust at 1 ± 0.1 AU from the K0V primary (Lisse et al. 2007b), while recent Gemini AO measurements have shown that any warm dust in the system resides at <2 AU from the primary (C. A. Beichman et al. 2007, private communication). VISIR imaging constraints on the location of the warm dust in HD69830 are r < 2.4 AU (Smith et al. 2009).

For the HD172555 circumstellar material, scaling from T_T1ejecta = 340 K for the hottest Deep Impact dust at 1.51 AU and L_* = 1.0, we find for the smallest grains with T_max = 335 K that r_dust = 5.8 ± 0.6 AU. LTE for grains larger than 100 μm in radius at this distance is ∼206 K. This agrees well with the ∼200 K temperature of the large, cool grain dust grains from this work, and with Wyatt et al.'s (2007b) Spitzer 24/70 μm based estimate of 4–5 AU for the location of the dust with respect to the HD172555 primary. The equivalent location in our solar system is at ∼1.9 AU from the Sun, near the inner edge of the main asteroid belt.

A location of 5.8 ± 0.6 AU for the circumstellar material is also consistent with the location of the innermost enhancement of dust surrounding β Pic, 6.4 AU from the star, as determined by Okamoto et al. (2004). From our best-fit model, the total surface area of dust that we find creating the observed mid-IR radiation is 4 × 10¹⁵–2 × 10¹⁶ km². At 5.8 AU, this is equivalent to a relatively narrow and localized torus or dust belt, of width 0.01 AU/sin(i) (where i is the inclination of the torus to Spitzer's line of sight).

There is 10¹⁹–10²⁰ kg of fine dust in the system, corresponding to the mass of an asteroid of radius 150–200 km. The fine dust, dominated by μm-sized particles, is a minority component by mass but dominates much of the observed IR emission due to its large surface area and strong IR emissivity. The location of the dust, at 5.8 AU from the primary, is analogous to a locus in the inner main belt of our solar system. Because of its small size and relative transparency, we are able to determine the mineralogical composition of the fine material in some detail. High-temperature, nonequilibrium amorphous silica dominates the Si- and O-rich fine silica dust. There is a deficit of the relatively unrefractory amorphous silicate species, and a deficit of the less refractory pyroxene versus the more refractory olivine (as compared to the primitive material found in other young systems, such as the TTS and Herbig objects, Figure 7). There is a noted lack of any low-temperature organic or hydrous material. It appears probable that the fine dust material was produced from a differentiated, SiO-rich body in an impulsive, high-temperature event.

Zoom In Zoom Out Reset image size

A component we have not had to use in our previous analyses, but find necessary to obtain a good fit to the Spitzer data (especially at the longer wavelengths) is material contained in a very large particle dust population (a > 100 μm), radiating as blackbodies at ∼200 K, close to the local equilibrium temperature for 5.8 AU. Because it is optically thick, we cannot determine much about its mineralogical nature. While hard to characterize, most of the solid circumstellar mass, 10²²–10²³ kg, is in this large cold dust. It is likely that this population is macroscopic rubble from the event that created the fine dust material.

At ∼10²² kg, SiO gas is one of the two major reservoirs of circumstellar mass in the HD172555 system. As SiO gas is typically formed when rocky material, such as olivines and pyroxenes, are vaporized in high-temperature processes, it is another piece of evidence for high-temperature processing. On the other hand, while the condensation temperatures for most SiO species are high, greater than 900 K, the temperatures we find for the solid dust range from 200 to 450 K (Table 3). If the system were in true thermodynamic equilibrium, then there would not be any SiO gas present. We conclude that the SiO gas reservoir and the dust reservoirs are not in equilibrium, but are instead still relaxing from the recent violent event that created them—much of the material appears to be still condensing out of the vapor phase.

Whatever processes are operative in the HD172555 system, they are acting on an important mass scale. 10²² kg of dust is about the total mass estimated to be in the solar system's main asteroid belt today. The amount of mass involved in the silica features of HD172555 is similar to the estimated amount of chondritic asteroidal material, the major reservoir of primitive silica in the present-day solar system. What we are observing is not merely due to the transformation of one averaged sized asteroid, producing a dense zodiacal cloud (e.g., the HD69830 system; Lisse et al. 2007b) or due to the destruction of a large comet. The scale of the causal event is like that of an ''oligarch''-sized planetesimal (on the order of the size of Ceres or larger). On the other hand, it is much too small to be an entire young asteroid's belt of circumstellar material—our own asteroid belt is purported to be diminished by a factor of 100–1000 from its original mass due to perturbations by Jupiter during the era of terrestrial planet growth (Chambers 2004).

The process is also operating on an interesting timescale. The presence of glassy amorphous silica-like material, produced on Earth as the kinetic product of very high temperature processing of silicaceous materials in lavas and impact craters (i.e., by quick quenching of rapidly melted rocks), and abundant SiO gas, the product of silicate vaporization, is suggestive of a quick, violent origin for the circumstellar material found around HD172555. The dominance of the SiO gas mass implies that it has not had time to recondense and settle out of the HD172555 circumstellar medium.

We estimate the timescale for SiO recondensation to occur via two-body collisions in a circumstellar gas torus in the molecular flow regime at ∼0.1 Myr, from estimates of the collisional lifetime of zodiacal dust particles in the solar system (Burns et al. 1979), assuming that when two molecules collide they stick with unit efficiency and begin to condense. A similar timescale applies for sweeping the fine dust particles out of the system via radiation pressure and P–R drag (Chen et al. 2006). If the SiO gas is localized around a planetary body, it should recondense even more quickly. A lower limit to the timescale, 10²–10³ years, can be estimated from the time for lunar formation from a dense circumterrestrial torus of dust and gas created by a hypervelocity collision between the proto-Earth and the Mars-sized impacting body Theia (Pahlevan & Stevenson 2007). An absolute observational lower limit to the timescale for blowout and recondensation of 22 years can derived from the apparent lack of change between the IRAS photometry from 1983 and Spitzer fluxes from 2005 (Figure 1; Chen et al. 2006).

All of these processes provide strong, short upper limits (in terms of solar system evolution timescales) to the time elapsed since the event creating the observed material, explaining why such systems like HD172555 are rarely seen.

Given our current understanding of young stellar system evolution, the possible causes of silica and SiO gas formation in HD172555 are fourfold: nebular shocks created by giant planet formation in a protoplanetary disk dense with gas (Desch et al. 2005); β-meteoroid formation and outflow, as in the Vega system (Su et al. 2005); massive stellar flares common to young pre-main-sequence (PMS) stars (Ribas et al. 2005); or hypervelocity collisions between planetesimal sized bodies (this work).

We can quickly rule out nebular shocks as the possible source of the HD172555 silica and SiO gas, as the system is gas-poor and unable to support strong energetic shocks, having already formed gas giants (if any) at 12 Myr of age and swept out its supply of primordial gas (Chen et al. 2006).

We can also rule out formation by infall of ''normal'' dust onto the stellar photosphere in a process similar to the creation of the β Meteoroids in the solar system (i.e., P–R drag induced infall of circumstellar dust, evaporation of the dust as it nears the stellar surface, and radiation pressure blowout of the micron-sized remnants; McDonnell & Dohnanyi 1978). The photo-evaporation mechanism would create copious amounts of SiO gas and smelt large amounts of silicaceous material, separating SiO from MgO, FeO, CaO, etc., into ∼1 μm particles at temperatures ∼1000–1500 K, and spread them over 10–100 AU as they are ejected from the system (e.g., the Vega system morphology; Su et al. 2005). By contrast, the highest temperatures measured in our study for the (smallest) HD172555 dust particles is ∼300 K, the dust is localized at ∼5.8 AU, and it is commingled with a population of large dust in radiative thermal equilibrium with temperatures ∼200 K.

Finally, we can rule out a stellar flare driven silica formation scenario. As a late-type A-star, HD172555 should be relatively X-ray flare quiet, and indeed has a low measured fractional X-ray luminosity L_x/L_bol of 2 × 10⁻⁶ (Chen et al. 2006). Also, the size of a flare required to melt and vaporize 10²² kg of SiO gas (roughly equivalent to a 1000 km radius asteroid) is enormous (at ΔH_vap ∼ 50 MJ kg⁻¹; Gail 2002), 10²² kg of rock requires 10³⁰ J total vaporization energy, and a 1000 km radius asteroid at 5.8 AU would require a total flare energy of 4 × 10⁴³ J emitted into 4π sr. Even for flares lasting years (10⁷–10⁸ s), we require average luminosities of L_x ≥ 10³⁵ J s⁻¹, prohibitive by many orders of magnitude versus the maximal observed L_x values of 10²³ J s⁻¹ for Herbig Ae stars and main-sequence A-stars (Skinner et al. 2004).

We are thus left with a planetesimal scale hypervelocity collision as the likely cause of the observed circumstellar material. The fine silica dust, SiO gas, and a large amount of macroscopic rubble were created in recent large hypervelocity (v_impact > 10 km s⁻¹) impact(s), via high-temperature processes operating at or near the surface of contact of the impacting bodies. In these impacts, multiple processes occur: the kinetic energy of the incoming bolide is transformed directly into heating and vaporization of rock into SiO gas at the surface of contact, a large amount of material is spalled off in the liquid phase, to solidify in-flight as tektite-like material, while at the edges of the affected region, shock waves propagate and excavate solid material in rubble form, producing a population of macroscopic, large dust particles (Melosh 1989). The impact disrupts the colliding body(s), while releasing into circumstellar orbit a massive amount of material. The results of this collision are being seen soon after the impact, so there has been little time for the silica dust and SiO gas to relax to the more thermodynamically stable olivine and pyroxene materials, or to be superceded by the products of subsequent grinding of the macroscopic fragments.

Hypervelocity impacts with enough mass and specific kinetic energy (v_impact > 10 km s⁻¹; Grieve & Cintala 1992; Cintala & Grieve 1998) could occur in events similar to the lunar formation event between large differentiated planetesimals. They could also occur due to the stirring up or by destabilization of an asteroid belt during oligarch formation, or by planetary migration through a local dynamical resonance, although these are somewhat more complicated scenarios, requiring a large number of similar impacts. Another possibility for achieving a relative impact velocity of >10 km s⁻¹ is gravitational focusing of a bolide onto a large body. The relative impact velocity achieved will then be roughly the escape velocity of the large body. A simple calculation for a large body made of terrestrial planet material shows that it would have to mass about as much or more than the Earth (v_{escape, Earth} = 11 km s⁻¹).

Observationally, the hypervelocity collision mechanism appears highly plausible for the HD172555 system, as (1) the closely analogous β Pic system shows evidence for large clumped regions of enhanced dust due to collisions (Telesco et al. 2005), three belts of enhanced IR emission from dust created by planetary formation (Okamoto et al. 2004), and evidence of the existence of a planet at ∼8 AU (Lagrange et al. 2009); (2) the amount of dusty material and gas we detect in a narrow torus at 5.8 AU is equivalent to at least a 1000 km radius asteroid of 2.5 g cm⁻³ density, the size of oligarchs and protoplanets; (3) the 10–100 Myr time of formation of the Moon in a massive impact on the proto-Earth and the stripping of Mercury's surface layers, is consistent with the estimated ∼12 Myr age of HD172555; and (4) after a small bolide of ∼370 kg mass impacted onto a comet nucleus at 10.2 km s⁻¹, the recent Deep Impact hypervelocity experiment produced analogous materials to what we have found for HD17255 : refrozen silicaceous droplets, i.e., fine dust, volatilized cometary material, i.e., SiO gas, and excavated, unprocessed bulk material, i.e., large cool dust grains (A'Hearn et al. 2005; Melosh et al. 2006; Richardson et al. 2007).

One consequence of the hypervelocity impact model is that the silica dust and SiO gas in the HD172555 system should be relatively transient in nature, clearing on timescales of <1 Myr (Section 4.1), somewhat in contrast to the predictions of Wyatt et al. (2007a). Another consequence is that we can use the presence of strong mid-IR 8–9 μm silica and SiO gas features to locate massive hypervelocity events in nonprimordial exo-solar systems. There is some indication, for example, that the ∼100 Myr old star HD23514, reported to have massive amounts of dust created by binary collisions, also demonstrates strong ∼9 μm emission spectral features (Rhee et al. 2008).

Comparison to other astrophysically similar spectra of comparable quality (Figure 2) verifies the unusual nature of HD172555. Abundant silica has not been reported in the ISM nor is it found in the majority of dusty exo-systems. Some of these other systems, such as BD+20 307 (Song et al. 2005), HD69830 (fragmentation of the equivalent of a 30 km radius P or D asteroid at 2 Gyr; Beichman et al. 2005; Lisse et al. 2007b) and HD113766 (fragmentation of the equivalent of >300 km s-type asteroid; Lisse et al. 2008) have been implicated as having large amounts of silicate-rich dust due to low velocity collisional processes that rubbleizes bodies, without creating much alteration of their mineralogy or chemistry. Evidence for silica-based emission has also not been reported in cometary dust. This is not surprising, as silica is not the low energy thermodynamic equilibrium state of Si, O, Mg, and Fe (Lewis 1989). Si–O gas masers do exist around late-type oxygen-rich AGB stars, but these require a large amount of free Si–O gas created through stellar mass shedding, driven by the local high stellar photon flux, to produce a population inversion and 8–9 μm emission. A handful of dense cloud cores have been reported to have Si–O masers under similar conditions.

Out of hundreds of mid-IR spectra compiled by E. Furlan (2008, private communication), we have found five other spectra which exhibit strong silica features at ∼9 μm: four TTS (FN Tau, CY Tau, V410 Anon 13, Hen3-600, published by Sargent et al. 2006) plus the 100 Myr old HD23514 (Rhee et al. 2008). The IR emission spectrum of HD172555 is in fact very similar to the FN Tau TTS spectrum of Sargent et al. (2006; Figure 8). The close similarity of the FN Tau and HD172555 spectra have in fact led us to consider the misidentification of HD172555 as an early MS star. We can rule out the misidentification of HD172555 as a late-type HAe or early-type TTS, however, due to its advanced age by association with β Pic, its low X-ray luminosity (L_x/L_bol < 2 × 10⁻⁵; de la Reza & Pinzon 2004; Schröder & Schmitt 2007), and its low disk gas content (Chen et al. 2006 and this work).

Zoom In Zoom Out Reset image size

While HD172555 is clearly not a TTS, could the same mechanisms be forming silica in the two cases? The physical situation around a TTS is much different than for circumstellar material in a debris disk around a young MS star like HD172555. Massive amounts (∼0.01 M⊙) of finely divided primordial material are located around a PMS star in an optically thick circumstellar disk many AUs in radius. This material is continually acted upon and altered over the 10⁶–10⁷ years of existence of the TTS phase. The most likely causes of silica formation in T Tauri systems are nebular shocks, massive stellar flares, and hypervelocity collisions (Section 4.3). While we expect shocks and stellar flares in the gas-rich, X-ray active T Tauri systems, we do note that collisions due to non-Keplerian rotation and accretion should be very frequent in a thick massive T Tauri disk, as the rate of collision goes at least as ∼disk density squared, and can even be enhanced due to gravitational focusing as giant planet formation proceeds though collisional accretion in the first few Myr.

There is abundant inferential evidence for massive collisions in the early solar system (Hartmann & Vail 1986): Mercury's high density; Venus' retrograde spin; Earth's Moon; Mars' North/South hemispherical cratering anisotropy; Vesta's igneous origin (Drake 2001); brecciation in meteorites (Bischoff et al. 2006); Uranus' spin axis located near the plane of the ecliptic. Local geological evidence for widespread impact melting includes tektites found on Earth and glass beads found in lunar soils (chemically distinguished from volcanic glasses also found in the soil; Warren 2008).

Solar system bodies without significant tectonic reworking, weathering or erosion of their surfaces, such as the Moon, Mercury, and Vesta, have the best evidence for giant impacts during their early growth stages. In the lunar formation event, the incoming bolide Theia had to have impacted at a low relative speed about equal to the escape velocity for the Earth–Moon system (v_escape = 12 km s⁻¹), so that the bulk of the excavated material either fell back to the Earth or remained bound in a circumterrestrial torus or disk rather than going into solar orbit (Canup 2004, 2008). Massive amounts of iron- and volatile-poor rock were vaporized, melted, and extracted from the upper layers of the proto-Earth, and a magma ocean was induced on the Earth's surface. However, while physically analogous to the event we expect formed the HD172555 materials, an event channeling most of the excavated material into a small, dense disk with dimensions on the order of the Earth–Moon distance would be hard to detect by Spitzer. In fact, the huge surface area of the dust creating the strong HD172555 mid-IR signature, ∼10¹⁶ km², implies an equivalent disk radius of ∼0.4 AU (6 × 10⁷ km), more than 2 orders of magnitude larger than the Earth–Moon distance of ∼4 × 10⁵ km, and far outside the Earth's Hill sphere.

Little is currently known about the origins of Vesta, other than it was formed in the first few million years of the solar system, its surface is highly variegated and heavily scarred by massive cratering, and that it consists of a highly basaltic igneous mineralogy that has been thoroughly heated and differentiated. Spectra of Vesta's surface (Vernazza et al. 2005), and Vesta-like meteorites (the so-called HED meteorites; Usui & McSween 2007) indicate the presence of silica, but as a minority component at the ∼10% level. Thus while we note its possible connections with the case of HD172555, as local evidence for an asteroidal source of silica dust and SiO gas, we do not discuss it further, other than to note the following plausibility argument for its early participation in a hypervelocity (>10 km s⁻¹) impact event. Keplerian orbital velocities of main belt asteroids (MBAs) in the solar system are on the order of v_Kepler ∼$\sqrt(G{M_\odot}$/2.6 AU) = 18.5 km s⁻¹. Along with v_Kepler, the dispersion in orbital inclinations of the MBA population is a critical defining characteristic for the velocity of interaction v_inter between two objects, as most collisions and velocity mismatches occur mainly in the direction perpendicular to the general Keplerian flow (Bottke et al. 1994). For example, in today's solar system, the ±15 deg range of inclinations versus the ecliptic in today's main belt leads to an average v_inter of ∼5 km s⁻¹. The young main asteroid belt at a few Mya age is predicted to have been frequently dynamically excited as planetary accretion progressed and oligarch formation occurred, leading to much larger inclination dispersions (up to ±45 deg), and much more energetic collisions than in the present day (Chambers 2004; Kenyon & Bromley 2004, 2006). Collision rates were also orders of magnitude higher in the much denser primordial belt. The dust and particles in the HD172555 system we have found at 5.8 AU should be moving at Keplerian orbital velocities with respect to the HD172555 primary of roughly v_Kepler = $\sqrt(GM_{*}$/r_*) = $\sqrt(G2M_ \odot$/5.8 AU) = 17.5 km s⁻¹, similar to the velocities of MBAs in our solar system. Also like the young main belt, inclination dispersions expected of young systems such as HD172555 should be dynamically excited and much more ''puffed up,'' with a large inclination dispersion and a correspondingly much greater chance of undergoing a hypervelocity interaction with v_inter > 10 km s⁻¹.

The details of a planetary crustal stripping event are fully consistent with our mineralogical findings, and can easily create the amount of dust we observe in HD172555. A giant impact event akin to the one proposed to have stripped Mercury of its surface layers of light crustal rock, throwing ejected material into a circumstellar torus centered on the impacted body's orbit, is expected to have occurred at high relative velocities of v_inter ≥ 20 km s⁻¹ (D. Stevenson 2008, private communication) and to have lasted long enough (∼10⁵ yr, Section 4.1) to be detected by Spitzer. Modeling the Mercury-stripping event, Benz et al. (1988) found that proto-Mercury was ∼2.25 times more massive than at present, before suffering a hypervelocity impact at 20–35 km s⁻¹ by a body 0.16 as massive as today's Mercury. In their models, the impactor became totally vaporized, indicating that temperature and pressure regimes leading to silica and SiO gas formation were reached. The 1.4 M_Mercury of ejected material is dominated by material from the differentiated surface layers of Mercury, potentially creating the depletion of Mg, Fe, S, etc. versus solar that we see in the HD172555 material. Modern predictions of the short-term recondensation products from the Mercury stripping event, following models similar to the equilibrium study of Fegley & Cameron (1987) would be very interesting to compare to what we find in the Spitzer spectra of HD172555.