Draft version September 29, 2018 Typeset using LATEX twocolumn style in AASTeX61
MICROMETER-SIZED WATER ICE PARTICLES FOR PLANETARY SCIENCE EXPERIMENTS: INFLUENCE OF SURFACE STRUCTURE ON COLLISIONAL PROPERTIES
S. Gartner¨ ,1 B. Gundlach,2 T. F. Headen,3 J. Ratte,2 J. Oesert,4 S. N. Gorb,4 T. G. A. Youngs,3 D. T. Bowron,3 J. Blum,2 and H. J. Fraser1
1School of Physical Sciences, The Open University, Walton Hall, Milton Keynes MK7 6AA, UK 2Institut f¨urGeophysik und extraterrestrische Physik, TU Braunschweig, Mendelssohnstr. 3, 38106 Braunschweig, Germany 3ISIS Facility, STFC Rutherford Appleton Laboratory, Harwell Oxford, Didcot OX11 0QX, UK 4Zoologisches Institut, Christian-Albrechts-Universit¨atzu Kiel, Am Botanischen Garten 1-9, 24118 Kiel, Germany
(Received July 31, 2017; Revised September 11, 2017; Accepted September 11, 2017) Submitted to ApJ
ABSTRACT Models and observations suggest that ice-particle aggregation at and beyond the snowline dominates the earliest stages of planet-formation, which therefore is subject to many laboratory studies. However, the pressure-temperature gradients in proto-planetary disks mean that the ices are constantly processed, undergoing phase changes between different solid phases and the gas phase. Open questions remain as to whether the properties of the icy particles themselves dictate collision outcomes and therefore how effectively collision experiments reproduce conditions in pro- toplanetary environments. Previous experiments often yielded apparently contradictory results on collision outcomes, only agreeing in a temperature dependence setting in above ≈ 210 K. By exploiting the unique capabilities of the NIMROD neutron scattering instrument, we characterized the bulk and surface structure of icy particles used in collision experiments, and studied how these structures alter as a function of temperature at a constant pressure of around 30 mbar. Our icy grains, formed under liquid nitrogen, undergo changes in the crystalline ice-phase, sublimation, sintering and surface pre-melting as they are heated from 103 to 247 K. An increase in the thickness of the diffuse surface layer from ≈ 10 to ≈ 30 A(˚ ≈ 2.5 to 12 bilayers) proves increased molecular mobility at temperatures above ≈ 210 K. As none of the other changes tie-in with the temperature trends in collisional outcomes, we conclude that the surface pre-melting phenomenon plays a key role in collision experiments at these temperatures. Consequently, the pressure-temperature environment, may have a larger influence on collision outcomes than previously thought.
Keywords: accretion, methods: laboratory: solid state, planets and satellites: formation arXiv:1710.02074v2 [astro-ph.EP] 6 Oct 2017
[email protected], [email protected] 2 Gartner¨ et al.
1. INTRODUCTION oratory collision experiments performed under different Dust aggregation is a key step in planet-formation conditions, such that the most appropriate data can be (Testi et al. 2014; Garaud et al. 2013; Blum & Wurm employed in planet-forming models, and where neces- 2008; Wada et al. 2008), enhanced by water ice at and sary such models can be modified to account for the beyond the snowline (Kataoka et al. 2013; Aumatell & influence of ice physics on collision outcomes. Wurm 2011; Gundlach & Blum 2015). But, we can only quantify collisional outcomes empirically, to learn how 2. OUTSTANDING CHALLENGES FROM icy dust sticks under proto-planetary conditions (e.g. EMPIRICAL ICE COLLISION DATA Gundlach & Blum 2015; Hill et al. 2015; Bridges et al. Generally, planet formation requires aggregation of 1996; Higa et al. 1996). small particles to form bigger ones. However, parti- Protoplanetary disk models indicate that icy particles cle sticking (a perquisite of models) is observed only are continually processed as particles traverse tempera- in a small subset of collision experiments and then over ture and density gradients (Woitke et al. 2016; Woitke a range of sticking probabilities (20 − 100 %; Deckers 2015; Visser et al. 2009), resulting in repeated evap- & Teiser 2016; Musiolik et al. 2016; Gundlach & Blum oration and re-formation of the water ice, which may 2015; Shimaki & Arakawa 2012a; Bridges et al. 1996; be amorphous, crystalline or a mixture of both (Sirono Hatzes et al. 1991). Interestingly, all these studies were & Ueno 2017; Sirono 2011a,b; Ros & Johansen 2013). performed at relatively high pressures (1 − 103 mbar), Indeed, both types have been observed in accretion so that we cannot know whether the results would have disks (Boogert et al. 2015; Terada & Tokunaga 2012; been the same at lower pressures as expected beyond Schegerer & Wolf 2010). Previous experiments (Wang the snowline of proto-planetary disks (< 1 mbar; Cieza et al. 2005) showed that crystalline ice films absorb much et al. 2016). less energy from impacts than amorphous ices, implying All laboratory experiments where sticking is observed, that collisional outcomes between proto-planetary disk have in common that they involved micrometer-sized particles could vary as a function of ice phase. structures (small particles or layers of condensed wa- However, all laboratory experiments necessarily have ter often referred to as “frost”). Indeed, models pre- to use analogs rather than interstellar ice, and the ice dict that the particle size strongly influences the stick- particle formation mechanisms in the laboratory diverge ing probability during collisions: the sticking threshold from those in astronomical environments. We therefore velocity vstick decreases with increasing particle radius address the following outstanding questions: −2/3 r: vstick ∝ r for 0.1 to 10 µm-sized particles (see −1 • Are the icy particles we are colliding in laboratory Figure 12 in Gundlach & Blum 2015), and vstick ∝ r experiments good analogs for protoplanetary disk for mm- to m-sized particles (see Figure 7.1 in Heißel- environments? mann 2015). However, micrometer-sized features on the surface of cm-sized particles (as induced by roughening • Does the ice phase of our particles affect the colli- or water condensation) are far less predictable and sev- sional outcome? eral collisional studies on such particles did not observe any sticking (Dilley & Crawford 1996; Higa et al. 1996; • Does the surface structure play a dominant role? McDonald et al. 1989; Hatzes et al. 1988). We have exploited neutron scattering and cryo-SEM In such collision experiments that do not lead to (scanning electron microscopy) to characterize the ice sticking, the coefficient of restitution, , is extracted, particle analogues used in our laboratory collision exper- which describes the loss of translational energy result- iments (Gundlach & Blum 2015) to ascertain whether ing from the collision, and eventually feeds into mod- phase changes in the bulk ice, and/or surface structural els of planet-formation. However, previous experiments changes, tie-in with the temperature-dependencies ob- (Bridges et al. 1984; Hatzes et al. 1988; McDonald et al. served in collisional data, and whether the production 1989; Hatzes et al. 1991; Supulver et al. 1995; Dilley & method of the ice analogs influences the particle struc- Crawford 1996; Higa et al. 1996; Bridges et al. 1996; ture. Higa et al. 1998; Heißelmann et al. 2010; Shimaki & In combination, these data reveal which of the ice Arakawa 2012b,a; Hill et al. 2015; Gundlach & Blum properties can affect collisional outcomes and to what 2015; Deckers & Teiser 2016; Musiolik et al. 2016) dis- extent these properties are altered by the collision envi- agree on whether, and how, varies as a function of ronment. This information is essential to relate labora- temperature, pressure, velocity, size, and shape. The tory data back to planet-formation scenarios and to dis- question is, why is this? We hypothesize that two key entangle the seemingly contradictory results from lab- factors play a role: first the method and prevailing con- AASTEX Collision Experiments: Ice Characterization 3 ditions under which the icy particles are formed, and range of length scales from the mesoscale (≈ 60 nm) second the prevailing conditions under which the parti- down to the intra-molecular level, which means it is cle collisions are investigated. possible to concurrently establish the phase, molecular The two key environmental parameters are pressure P structure and surface properties of icy materials (Mit- and temperature T . However, these two parameters are terdorfer et al. 2014; Hill et al. 2016). not usually varied systematically, resulting in contradic- tory experimental results and making it difficult to as- 3. EXPERIMENTAL METHOD certain exactly which empirical data are most relevant to planet-forming models. From the few cases where The particles for this characterization study were pro- T has been varied at constant P (Bridges et al. 1984; duced as described in detail in Gundlach et al.(2011); McDonald et al. 1989; Higa et al. 1996; Heißelmann Jost et al.(2013); Gundlach & Blum(2015). Briefly, et al. 2010; Gundlach & Blum 2015; Hill et al. 2015), liquid D2O was dispersed by an aspirator and sprayed two clear trends are evident; the collisional outcomes into liquid nitrogen, accumulating sample material for are temperature-independent below T ≈ 210 K (e.g. > 1 hour. The particles were then funneled into the Gundlach & Blum 2015; Heißelmann et al. 2010; Hill pre-cooled (77 K) sample container, which was closed, et al. 2015) and become temperature-dependent above mounted in the neutron beam, and passively cooled at T ≈ 210 K (e.g. Gundlach & Blum 2015), where the a constant P (30 mbar He). coefficient of restitution decreases and the threshold ve- We studied samples with two different mean particle locity to particle sticking increases, as temperature in- radii, ((0.71 ± 0.31) µm and (1.45 ± 0.65) µm), where creases. uncertainties give the particle size distribution’s FWHM The ice projectiles in these collision experiments have (full width at half maximum) (Gundlach et al. 2011). been formed under various conditions, but always from Two independent experiments were conducted at each the liquid phase. While freezing of liquid water in a particle radius. kitchen freezer or under liquid nitrogen is expected to Advantage was taken of the neutron scattering prop- yield some form of crystalline ice, the ice structure on a erties of D2O compared to H2O(Sears 1992), and care molecular scale will depend on the freezing rate as well was taken to minimize sample contamination with H2O as on the conditions (and duration) under which the ice during sample preparation and loading, by maintaining was processed and/or stored between initial freezing and an N2-purged environment, retaining both the container eventual collision. This “thermal history” again has not and sample below 100 K. always been varied systematically and not even always Neutron scattering data was collected over 30 min been fully described. isothermal periods, at 103, 164, 184, 206, 226, and Micrometer-sized particles can be created by shatter- 247 K. The initial data reduction and calibration was ing larger bodies of ice prepared in a freezer (Deckers done using GudrunN software (Soper 2011, 2013), ac- & Teiser 2016), or by rapid freezing of water droplets cording to standard neutron scattering data processing. e.g. on cold surfaces (Musiolik et al. 2016) or by intro- The raw data from our neutron scattering experiments ducing them into a cold gaseous or liquid environments were merged for all detectors, corrected for instrument (Gundlach & Blum 2015; Shimaki & Arakawa 2012a), effects, and normalized on a per atom basis. Examples However, without further characterization, we cannot of the resulting background corrected neutron diffrac- know whether, and how, the production alters the par- tion patterns are shown in Figure1(a) for one of the four experiments. The small angle neutron scattering ticle structure on all length-scales, exactly which form −1 of crystalline ice is produced, and to what extend the P - (Q ≤ 0.1 A˚ ), probing surface structures, was observed −1 T conditions of the collision environment influence the concurrently with the high-Q region (Q ≥ 1 A˚ ), prob- collisional outcomes. ing the bulk ice phase (intra-molecular distances). As no definitive particle characterizations have been To give an impression of the particle structures on made to date, ice phase and micro-scale structure are larger scales (0.1 − 100 µm), we compared the neutron alluded to in icy particle collision experiments and their scattering results to images from complementary cryo- influence on collision outcomes remains a contentious SEM experiments (Hitachi(S-4800); see also Jost et al. issue in the literature, which we address in this work. (2013)), using H2O-particles prepared the same way as By exploiting the unique capabilities of the NIMROD previously described, but necessarily held at lower pres- (Bowron et al. 2010) neutron scattering instrument, we sures (P = 10−3 mbar). characterized the bulk and surface structure of the icy particles. NIMROD can simultaneously observe a wide 4. RESULTS 4 Gartner¨ et al.
) 247 1
- a b c
) 3 2.0 10
m 226 Data o
t b 2 10 Fit (Equation (1)) a 206
r 2 1.5
s 10
( 184
s n
r 164 1 a 101 10 b (
1 1.0 Temperature (K) +
) c 100 Q
( 0 103 I 10 10-1 100 101 0.02 0.04 0.08 2 3 4 5 Q ( -1 ) Q ( -1 ) Q ( -1 ) Figure 1. (a) NIMROD spectra (0.71 µm icy particles), showing the neutron scattering signal I(Q) (barns (sr atom)−1) as a function of the momentum transfer Q, which is inversely proportional to the length scale. Sequential plots show the temperature evolution in six isothermal steps between 103 and 247 K (color bar). Expanded views: (b) low-Q (dashed lines show fit as per Equation1, Section 4.2) and (c) high- Q regions. For clarity, error bars have been omitted (average uncertainty: 6 % of I(Q)). Across the four neutron scattering experiments, no long-range order, which will be discussed in more detail clear differences were seen between scattering from dif- in Section 5.3. ferent icy particle sizes, nor in repeated experiments on Within fitting uncertainties, this deconstruction particles of the same size. However, clear changes with showed that the temperature dependence of the ice increasing sample temperature are evident (Figure1), a (1.45 ± 0.65) m b (A) (B) (C) indicating temperature induced modifications in the icy (0.71 ± 0.31) m 103 164 184 206 226 247 particles, both in the bulk ice phase and the particle sur- 1.0 e.g. Data faces, which will be addressed in detail in the following. 103 K 0.8
Ih y 4.1. Ice Phase I 0.6 N
The high-Q region (Figure1(c)) shows peaks indica- y / Ic x tive of a crystalline-dominated ice structure; the most I 0.4 N I(Q) (a.u.) prominent feature is the triplet of Bragg peaks in exactly Ix the positions expected for hexagonal ice, Ih, (1.59, 1.70, −1 0.2 and 1.80 A˚ (Petrenko & Whitworth 1999)). There are e.g. Residuals 103 K some subtle but significant modifications to this triplet −1 0.0 as a function of temperature; the peak at 1.70 A˚ actu- 2 3 4 5 6 103 164 184 206 226 247 -1 Isothermal Temperature (K) ally corresponds to overlapping features from Ih and Ic Q ( ) (cubic crystalline ice) and loses intensity above 180 K. Figure 2. (a) original high-Q data for the 103 K, 0.71 µm This is also reflected in changes to the smaller diffraction particles (red), together with diffraction patterns of the three peaks at higher Q and indicates that Ic is lost. There are ice phase components Ih (black), Ic (green), and Ix (blue). two possible pathways for this loss; either transforma- Each high-Q pattern was deconstructed into a sum of these tion to the more stable Ih or sublimation. Any increase components as described in the text. Also shown (yellow) are in the amount of Ih would result in an increase in the the residuals when only Ih and Ic are fitted. (b) outcome of this analysis across our entire data set. Cumulative bars intensities of the other Ih Bragg peaks, which appear unaltered. represent the fractions of each ice-phase (colors as in (a)) re- quired to reproduce the high-Q data at each isothermal tem- To quantify these changes, we deconstructed each −1 perature, averaged across two experimental runs, and sepa- high-Q diffraction pattern between Q values of 1−6 A˚ rated by particle radius (see legend). Each cumulative bar into a sum of features representing Ih and Ic, based is normalized to the total amount of ice determined in each on crystallographic calculations of the diffraction pat- sample at the initial temperature point, 103 K. The typical terns for the respective idealized pure ice phases. The uncertainty, propagated from the individual fits, is shown on residuals of this analysis showed a very broad diffraction the final Ih bar; for clarity other error bars have been omit- ted. Dotted lines indicate the critical isothermal steps at peak, whose shape and position closely matched that ex- which drastic changes in the ice phase were observed, break- pected for amorphous ice (Figure2(a) blue and yellow ing the data into three distinct ice-phase regimes; Ih +Ic +Ix; curves), so a third component was added to the decon- Ih + Ix; Ih. struction; Ix, denoting inter-domain bulk material of no AASTEX Collision Experiments: Ice Characterization 5 (A) (B) (C) 3.5 exponent β, which can be related to the ice surface ) 35
( roughness (e.g. Mitterdorfer et al. 2014; Hill et al. 2016).
3.0 r
(1.45 ± 0.65) m e For our icy particles, β values ranged from 4.1 to 6, in- 30 y (0.71 ± 0.31) m a ) 2.5
L creasing non-linearly with increasing temperature. Val-
3 SSA smooth spheres e s m SSA observed ues of β > 4 indicate that no surface roughness on nm c 2.0 25 u / f
Thickness (Diff. Layer) f 2 length scales is introduced by the freezing process, but i m D
( 1.5 f 20 that the surfaces are diffuse, i.e. showing a density gra- o A
S s dient (e.g. Strey et al. 1991; Su et al. 1998). However, s S 1.0 e in the case of diffuse interfaces the particle surface den-
15 n k
0.5 c sity cannot be validly modeled by a step function (as in i h
10 T the Porod analysis) but is best described by convolut- 0.0 100 150 200 250 ing a Gaussian with said step function. The width of Temperature (K) this Gaussian indicates the thickness of the diffuse in- Figure 3. SSA and t, as a function of temperature, terface, t. The resulting fit-function for the background as extracted from the low-Q NIMROD data using Equa- corrected low-Q data is (Strey et al. 1991): tion1. For clarity, the results have been averaged per particle 2 2 size across experimental runs, and consecutive temperature I(Q) = 2π(∆ρ)2 SSA Q−4e−Q t , (1) points joined with a solid line (SSA, left-hand axis) or dashed line (t, right-hand axis). Error bars represent the standard where SSA is the specific surface area, and ∆ρ = −1 deviation of the mean. 5.995 × 10−6A˚ is the scattering length density differ- ence. Under our specific experimental conditions, this structure is independent of particle radius, i.e. a bulk is the scattering length density of D2O, since no other effect. Initially, the icy particles exhibit stacking disor- material is present that has not already been corrected der, dominated by Ih. They comprise areas of both Ih for by the calibration scans. The resulting fits ideally and Ic as well as inter-domain amorphous structures, reproduce the experimental data over the entire low-Q Ix. range, (Figure1(b): dashed lines). As T increases, the normalized fractions of all three A major advantage of this model is that it concur- components change and the normalized fractions no rently gives values of t and SSA, (Figure3). At all tem- longer sum to 1. A measurable fraction of Ic is retained peratures the SSA values are below those calculated for until 184 K (Figure2(b) regime (A)), but has essen- samples of smooth spherical particles with the given size tially disappeared by 206 K. A fraction of Ix persists distributions (Figure3 left-hand axis: light/dark −−×−), until 226 K (Figure2(b) regime (B)); at 247 K the data which will be discussed in detail in Section 5.2. How- is best fitted by Ih only. ever, as expected, the SSA of the 0.71 µm particles is always greater than that of the 1.45 µm particles. Re- 4.2. Surface gardless of particle size, the SSA slightly decreases in Returning to Figure1(b), the low- Q data show much the 103 − 184 K range; the most drastic changes in SSA more drastic changes than the high-Q Bragg peaks. set in beyond 184 K, then this loss rate slows beyond There is little, if any, obvious change in I(Q) between 226 K. 103 and 184 K, but as T increases beyond 184 K, I(Q) It is interesting to note that when the regimes (dotted diminishes rapidly. This indicates a substantial change vertical lines) from Figure2 are transposed to Figure3, in the particles’ surface structure, the onset of which the key temperatures at which ice-phase-compositional coincides with the temperature regime at which Ic is changes occur correspond exactly with the distinctive no-longer measurable in the bulk-ice. changes in SSA. At all temperatures the low-Q slopes approximately Whilst changes in t almost mirror those in SSA, the follow the expected Porod power-law (Feigin et al. 1987), most drastic changes occur above 206 K, thus not match- indicative of compact, granular material. This is ex- ing the temperature regimes observed for ice phase and pected from a sample of non-porous, spherical icy parti- SSA. The absolute values of t are closely comparable be- cles and thus confirms that formation in liquid Nitrogen tween the two particle sizes, starting from around 10 A˚ does not alter the internal particle structures on length at 103 K, which represents roughly 2.5 Ih bilayers, and scales of tens of nanometers. increase (on average) by a factor of 3 with temperature. There are well established methods to extract quanti- From the SSA and t results alone, we cannot distin- tative information on the surface structures from these guish whether the surface changes are caused by par- data. Porod analysis (Sinha et al. 1988) yields a Porod ticle sintering or by sublimation. However, the images 6 Gartner¨ et al. obtained from our complementary cryo-SEM study can solid water), and HDA (high-density amorphous solid answer this question. water), but the resolution of our diffraction data is not well suited to distinguish between them. NIMROD was 4.3. Sintering or Sublimation? not designed as high resolution crystallography instru- Figure4 joins up our findings on all length scales to- ment, but rather to provide atomistically quantitative gether with the SEM images. The vertical dotted lines structural data for highly disordered and complex sys- indicate the same evolutionary stages in the ice-phase tems over a very wide Q-range. Sophisticated models composition and SSA, as identified in Figures2 and3. for the fitting of diffraction patterns from stacking dis- The SEM images reveal that initially the particles are ordered ices have been developed by various groups (e.g. mostly, although not perfectly, spherical (Figure4(a)– Malkin et al. 2015; Kuhs et al. 2012, and references (c)). The icy particles are in contact with each other, therein). However, an in-depth analysis of the obtained but no sintering is evident. With increasing tempera- diffraction patterns is not required for the purpose of ture, sintering is observed (Figure4(b 2)), and as the this work and is not feasible using the moderate resolu- temperature continues to rise, the sintering necks be- tion data obtained. come more pronounced (Figure4(c 2)). Nevertheless, our molecular-scale neutron-scattering Finally, by the highest temperatures, where the par- data are sufficient to characterize three distinct phase ticles only comprise Ih (i.e. beyond the second vertical regimes for our icy particles (Figures2 and4). The dotted line in Figures2,3, and4), material seems to temperature range, across which contributions from Ix be lost from the narrow sinter-neck and the particles are seen, supports the interpretation as inter-domain ice become faceted with straight edges and reduce in size (lacking long-range order) sandwiched between hexago- (Figure4(d 2)), while the smallest particles are lost. nal and cubic domains, as illustrated e.g. in Figures 2 and 3 of Hondoh(2015). These inter-domain amorphous 5. DISCUSSION structures; are not equivalent to, nor to be confused We have characterized micrometer-sized icy particles with, diffuse surface layers or vapor-deposited amor- identical to those used in laboratory collision experi- phous solid water (ASW). We find that none of the ments on planet-formation. Our icy particles were pro- phase-change temperatures matches the collisional tem- duced under liquid Nitrogen and were not dissimilar in perature dependencies, which set in above ≈ 210 K. size to grains of crystallized water ice (≈ 0.8 µm in size) Thus, we conclude that bulk crystalline ice-phase cannot that have been observed in the silhouette disk of a young influence collisional outcomes in our experiments and star (Terada & Tokunaga 2012). further crystallographic studies on a dedicated instru- Our characterization experiments cover almost the ment are not required. whole temperature range exploited in laboratory colli- sional studies over the past three decades, from 80 K 5.2. Does the surface play a dominant role? (e.g. Musiolik et al. 2016; Hatzes et al. 1988) to 269 K Surface features could be connected to collision out- (e.g. Shimaki & Arakawa 2012a; Higa et al. 1996), per- comes and particle aggregation in several ways. Both formed to understand collisions in a variety of environ- surface wetting and surface roughness might be expected ments, like protoplanetary disks, cometary surfaces, and to increase the stickiness of particles via friction effects. planetary rings. Thus, our results provide crucial infor- Molecular scale features (A-scale)˚ such as molecular ori- mation towards the role of ice phase and surface struc- entation, mobility, or density variations on the surface tures in dictating collision outcomes in such environ- might affect particle stickiness. They would affect the ments. small angle scattering slope, but not the SSA. Surface roughness on nm-scales would affect the small angle 5.1. Does the ice phase affect the collisional outcome? scattering slope and increase the observed SSA, particle We find that our particles are initially stacking dis- sintering in the aggregation process would reduce the ordered, as is expected when freezing water droplets in SSA with respect to that of smooth spherical particles. liquid Nitrogen (Malkin et al. 2015, 2012; Kuhs et al. Based on the original Porod analysis of our small angle 2012). The particles comprise both low-pressure crys- scattering data, we can exclude surface roughness, and talline phases of ice, Ic and Ih, but the obtained diffrac- indeed we find that even at the lowest temperatures the tion patterns are best fitted when an amorphous ice observed SSA is below that expected from the given size phase is added. Various amorphous ice candidates could distributions of spherical particles by a factor of ≈ 3 for be attributed to this third phase, e.g. HGW (hyper- both mean particle sizes (Figure3), which could indi- quenched glassy water), LDA (low-density amorphous cate particle sintering. However, the SEM images (Fig- AASTEX Collision Experiments: Ice Characterization 7