Ice Shell Structure and Composition of Ocean Worlds: Insights from Accreted Ice on Earth

Ice that crystallizes within a supercooled water column, as opposed to at a solid interface, is referred to as frazil ice. Frazil ice is formed in the presence of turbulent water that has been supercooled by tenths to hundredths of a degree (Weeks and Ackley, 1986; Mager et al., 2013; Robinson et al., 2019), where increased supercooling generally promotes increased frazil production (Ettema et al., 1984). There are a number of mechanisms in nature that can promote supercooling and thus the production of frazil ice. Examples of such mechanisms include the adiabatic rise of water masses to a lower-pressure environment and double diffusion occurring between two adjacent water bodies at different temperatures and salinities (Mager et al., 2013).

Ice crystals formed from collisions of larger ice crystals, the refreezing of spray, or snow can serve as nucleation sites for frazil ice crystals (Osterkamp, 1977). It was long believed that foreign particles, such as organic matter, could serve as nucleation sites for frazil ice, but no experimental or field observations have demonstrated that this is possible at the degrees of supercooling observed in nature (<1°C) (Daly, 1984; Robinson et al., 2019).

Turbulence is also necessary to promote secondary nucleation, responsible for generating meaningful quantities of frazil crystals (Ettema et al., 1984). Because frazil ice forms from individual crystals that can nucleate independent of each other, it has a granular texture and no preferred orientation (Fig. 2). Once a stable frazil ice layer has formed, congelation ice growth can occur.

Congelation ice refers to ice produced by the direct freezing of water at an existing ice interface, driven by conductive heat losses (Weeks and Ackley, 1986). In congelation ice, the microstructural morphology of the ice-water interface (e.g., planar, cellular, dendritic) is highly dependent on the purity of the source water and the growth velocity (Harrison and Tiller, 1963; Lofgren and Weeks, 1969; Wettlaufer, 1992, 1998). Ultimately, the microstructural morphology of the ice-ocean interface is related to the phenomenon of constitutional supercooling (Harrison and Tiller, 1963; Eicken, 2003), originally proposed and studied in the field of metallurgy (Rutter and Chalmers, 1953; Jackson, 2004).

Constitutional supercooling refers to supercooling that occurs in advance of the freezing front. The role of constitutional supercooling in congelation ice growth is critical to governing its substructure and, in turn, its properties (Eicken, 2003; Weeks, 2010; Petrich and Eicken, 2017). The rejection of impurities locally enhances the concentration and depresses the freezing point at the interface, promoting supercooling ahead of the interface. If perturbations occur in the presence of constitutional supercooling, the supercooled fluid serves as a heat sink that promotes further growth, forming cells or dendrites. In the absence of this supercooled layer, small perturbations in the interface morphology are not energetically favorable and a planar interface remains stable.

Characteristics of the interface are significant to the efficiency of impurity incorporation in ice (Nagashima and Furukawa, 1997). A planar interface is more efficient at rejecting impurities, whereas a cellular interface retains impurities through the entrapment of brine between cells (Osterkamp and Weber, 1970; Eicken, 2003; Weeks, 2010; Petrich and Eicken, 2017). For the growth rates typical of sea ice on Earth (Table 2), it has been demonstrated that congelation ice forming from seawater will always result in the development of a cellular interface (Wettlaufer, 1992). Congelation ice is typically characterized by a columnar texture, where crystals preferentially elongate parallel to the direction of the temperature gradient (Harrison and Tiller, 1963; Tison et al., 1998).

In low salinity environments, such as freshwater lakes, constitutional supercooling during freezing is minimal and the morphology of the microstructural interface can remain planar for higher growth velocities than it would for seawater (Leppäranta, 2015).

3.2. Marine ice and sub-ice-shelf congelation ice

Marine ice is specific to frazil ice that collects and consolidates beneath ice shelves or within ice shelf rifts that are characterized by a low temperature gradient environment. The formation of marine ice is generally believed to occur in two phases, defined by Tison et al. (2001) as (i) the frazil ice phase and (ii) the consolidation phase.

The frazil phase encompasses the formation and accumulation of frazil ice crystals beneath the ice shelf. These crystals preferentially form and collect where the ice draft thins rapidly—features such as inverted channels, rifts, or crevasses beneath the ice shelf (Tison et al., 1993; Khazendar et al., 2001; Khazendar and Jenkins, 2003). The consolidation phase involves the buoyancy-driven compaction of accumulated frazil crystals. In this phase, crystals agglomerate and collect, forming a permeable layer. As more frazil accumulates, buoyant pressure builds up at the ice-water interface, compressing the layer and forcing out interstitial water, reducing the brine volume fraction. The bulk density of the ice-brine system is, thus, counter-intuitively reduced by compaction. At a certain stage in the consolidation phase, the ice becomes impermeable and any remaining brine is trapped in the ice as inclusions at triple-junctions and along grain boundaries (Moore et al., 1994).

The final stage of consolidation involves the freezing of remaining interstitial water through congelation growth, analogous to the incorporation of frazil ice layers beneath growing sea ice known as platelet ice. Unlike platelet ice, this interstitial congelation growth occurs at a much slower rate due to the insulation from atmospheric thermal forcing by overlying glacial ice. The lower unconsolidated portion of the marine ice layer is a hydraulically connected region that can extend from tens of meters up to ∼100 m from the base of the ice shelf (Craven et al., 2009). The formation of marine ice beneath ice shelves is part of a process that has been referred to as an “ice pump,” where the pressure dependence of the freezing point supports the operation of a continuous cycle involving the melting of ice at depth and the accretion of ice at a shallower location (Lewis and Perkin, 1986).

The term “marine ice” is sometimes broadly applied to ice that forms beneath ice shelves. Here, however, we distinguish between marine ice and sub-ice-shelf congelation ice to emphasize the distinct formation mechanisms between these forms of accreted ice. Because the ice-ocean interface beneath ice shelves is fairly insulated from atmospheric forcing (i.e., the ocean is shielded from frigid air temperatures by hundreds of meters of ice), the formation of congelation ice at the base of an ice shelf is rare (Fig. 4); however, it has been observed beneath certain ice shelves in Antarctica (Gow and Epstein, 1972; Zotikov et al., 1980; Souchez et al., 1991).

OPEN IN VIEWER

FIG. 4.

A summary of properties and characteristics of terrestrial accreted ice from published ice core data. The first two columns specify the location where the ice core was collected and a description of the sample site. The sites are color and texture coded by the ice shelf and presented in the map of Antarctica in (a). Where multiple cores were collected from a single location, the second core is represented as dashed. The third column provides the name of the ice core as referenced in the sources in the rightmost column. The type of accreted ice is specified in the fourth column according to the following codes: CS = congelation sea ice; M = marine ice; SISC = sub-ice-shelf congelation ice. The plots representing the (b) absolute and (c) depth-normalized salinity profiles follow the same color and texture coding represented in the table and map.

A simple model to predict the formation of congelation ice beneath an ice shelf was proposed by the Ross Ice Shelf Project (RISP) and summarized by Neal (1979). When water at the pressure-melting temperature flows in the direction of increasing ice shelf thickness, it must dissipate heat to remain at the pressure-melting temperature. Under conditions where the thickness gradient and flow speed are such that the sensible heat conduction to the overlying ice layer exceeds that which must be dissipated at the boundary layer to maintain the pressure-melting temperature, bottom freezing will occur (Neal, 1979).

The J-9 Ross Ice Shelf core represents a unique and valuable sample of congelation ice acquired at a depth of ∼400 m within a zone of bottom freezing (Zotikov et al., 1980). The published sample is uniquely well characterized for sub-ice-shelf congelation ice and includes measurements of salinity, grain size, texture, and freezing rate. The freezing rate estimate was obtained from an observed transition in growth conditions at the bottom 2 cm, which was attributed to localized melting caused by a drilling expedition the previous year (Zotikov et al., 1980). The estimate was validated by a simple heat transfer calculation (Zotikov et al., 1980) and represents the only estimate of sub-ice-shelf congelation ice growth rate obtained through direct inspection of a sample of the basal accreted ice.

Congelation ice can also form beneath ice shelves experiencing high rates of surface ablation (e.g., locations with strong katabatic winds) (Souchez et al., 1991).

4.1. Congelation ice across growth regimes

The partitioning of salt into ice, S i c e , from ocean water of salinity, S ₀, can be represented by the effective solute distribution coefficient, k v = S i c e ∕ S 0 , which is a function of ice growth velocity (Burton et al., 1953; Weeks and Lofgren, 1967). Although models for the effective solute distribution coefficient do not directly represent the physics of sea ice desalination as it is understood today (Notz and Worster, 2009), existing models fit the data well for both natural and artificial ice over a range of freezing rates relevant to sea ice (Fig. 3).

OPEN IN VIEWER

FIG. 3.

Summary of relationships representing the effective solute distribution coefficient, k = S_ice/S ₀, as a function of ice growth velocity. The markers represent data points from experimental or field data. Solid lines through data points represent least squares fits of the data to published models for the solute distribution coefficient, where dashed lines represent extensions of the model beyond the available data range. The green curve is a smoothed representation of multiple runs of the mushy-layer model of Buffo et al. (2018), assuming a critical porosity inferred from the salinity of the J-9 core from Ross Ice Shelf, Antarctica. The light blue curve represents the model presented in Equation 6.

Parameterizations of salt partitioning based on growth velocity represent a computationally inexpensive approach to augment simple freezing models that do not directly model ice desalination processes. Furthermore, representing the salinity of ice as a fraction of ocean salinity allows salt entrainment in ice to be parameterized independent of the source water salinity. Even though more complex numerical models of ice desalination processes exist (Griewank and Notz, 2013; Buffo et al., 2018; Wells et al., 2019), effective solute distribution coefficients are invaluable for certain planetary applications where high-resolution salinity profiles are not needed and properties of the ocean are poorly constrained.

At growth velocities above those naturally occurring on Earth (Fig. 3), ice experiences minimal fractionation ( k ≈ 1 ) on freezing, implying that it serves as a relatively unaltered chemical fingerprint of the source water. This rapid freezing regime would include processes occurring at or near the surface at Europa and Enceladus, such as the flash freezing of brine infiltrating porous ice at the surface or plume material that is frozen as it ascends through the fractured ice shell from a subsurface reservoir below (McCord et al., 2002; Schmidt et al., 2011; Fox-Powell and Cousins, 2021).

Published measurements of sea ice growth rates span from ∼ 1 0 − 6 to 1 0 − 4 cm/s (Table 2). Salt partitioning in this regime has been characterized by using both natural (Nakawo and Sinha, 1981; Granskog et al., 2006) and artificial (Weeks and Lofgren, 1967; Cox and Weeks, 1975) samples of congelation ice. Studies of natural sea ice samples are more challenging due to the difficulties in obtaining samples and the uncertainties in natural growth rates.

The dataset of Nakawo and Sinha (1981) is particularly valuable because of the high sampling frequency of ice salinity and temperature they obtained over the growth season that produced nearly continuous profiles of ice salinity and growth rate. Although growth velocities for the sea ice regime and above are not directly applicable to accretion occurring at the ice-ocean interface of ocean worlds (Fig. 3), this represents the regime where the effective solute distribution coefficient is the most sensitive to growth velocity and where more significant variations in bulk ice shell salinity might occur.

At a certain stage in growth, the salinity profile of the ice no longer evolves in time due to progressive brine drainage. This salinity has been referred to as the stable salinity (Nakawo and Sinha, 1981; Petrich et al., 2006) or steady-state salinity (Petrich et al., 2011). The natural congelation ice samples of Nakawo and Sinha (1981) in Fig. 3 are believed to be representative of this stable salinity and, as such, fall below the experimental data, which were not given sufficient time to reach this steady-state condition.

The Baltic sea ice samples of Granskog et al. (2006) in Fig. 3 represent the stable salinity of ice formed from a lower salinity source water. These data suggest that a lower salinity source water may enhance the efficiency of salt rejection, possibly due to a change in interface morphology (Granskog et al., 2006). Their data are consistent with those of Weeks and Lofgren (1967), which included samples formed from low salinity source waters.

Because salt in ice is predominantly trapped interstitially as brine, the steady-state salinity is believed to be coupled to a critical porosity (∼5%) below which ice is believed to be impermeable to brine transport (Golden et al., 1998, 2007). The critical porosity is typically a prescribed parameter in numerical models of sea ice desalination (Petrich et al., 2011; Griewank and Notz, 2013; Buffo et al., 2018, 2020; Wells et al., 2019) and governs the finite ice salinity that the model asymptotically approaches as the growth velocity approaches zero (i.e., the system reaches equilibrium) (Fig. 3).

The distribution coefficient associated with this limit has been referred to as the effective equilibrium distribution coefficient, k e q (Burton et al., 1953; Weeks and Lofgren, 1967) and represents the bulk salinity of congelation ice as the growth velocity approaches zero. Other sea ice desalination models do not explicitly impose a critical porosity, but they instead use permeability-porosity relationships that represent a percolation threshold as a significant reduction in permeability that occurs as the critical porosity is approached (Petrich et al., 2006; Buffo et al., 2021a).

Figure 3 demonstrates that at the growth velocities predicted for the ice shells of ocean worlds, the effective equilibrium solute distribution coefficient should govern the bulk salinity of the ice shell.

For ocean worlds, we are interested in the accretion of ice in low temperature gradient environments characterized by growth velocities within the ice shelf regime ( < 1 0 − 6 c m ∕ s ), where k ≈ k e q (Fig. 3). Because experimental studies cannot sample this growth velocity regime, we must leverage the Earth's natural laboratory to estimate the salinity of ice formed in these environments. We present a survey of the available published ice core data from Antarctica and the Arctic, including samples of marine ice and sub-ice-shelf congelation ice (Fig. 4).

We provide characteristics of the environment in which the ice formed, including depth from the surface as a proxy for temperature gradient (i.e., deeper ice implying a lower temperature gradient) and estimates of growth velocity where available. We also include properties of the ice such as salinity and δ¹⁸O where known, which can serve as a proxy for modification of the seawater by glacial meltwater (i.e., values close to 2‰ implying minimal modification). δ¹⁸O is often used to determine the origin of the ice (i.e., marine or meteoric) when the salinity signal is ambiguous (Gow and Epstein, 1972; Morgan, 1972; Oerter et al., 1992). Estimates for the effective solute distribution coefficient were obtained by dividing the ice salinity by the salinity of seawater, assumed to be 35 ppt.

Although it represents a relevant analogue, we exclude the ice core from Lake Vostok because the mechanism of accretion remains debated, and the properties of the lake water are not well constrained (Souchez et al., 2000, 2004; Lipenkov et al., 2015). Table 3 presents the values of k e q estimated from selected ice cores in Fig. 4. We discuss how these values are obtained in the following sections.

5.1. Bulk salinity of a congelation ice shell

We estimate the bulk salinity of a congelation ice shell using a 1D solidification model known as the Stefan problem, where heat is conducted from the interface through the overlying ice. This model has been previously applied to Europa's ice shell (Quick and Marsh, 2015; Buffo et al., 2021b); however, we specifically adopt the form published in Lior (1996). The analytical solution to this problem represents the temperature in the ice, T, as a function of position, x, and time, t, and is given byT x , t = T s + T f − T s e r f x 2 α t e r f λ ′(1)

where T_s is the surface temperature, T_f is the freezing temperature, α is the thermal diffusivity of the ice, and λ ′ is the solution to the equationλ ′ e λ ′ 2 e r f λ ′ = N S t e π(2)

where N S t e is the Stefan Number, defined asN S t e = c p T s − T f L(3)

where c_p is the specific heat capacity of ice, and L is the latent heat of fusion. The position of the ice-water interface as a function of time can be expressed in terms of these variables asX t = 2 λ ′ α t(4)

and the velocity of the ice-water interface isv = Ẋ t = λ ′ α t(5)

which corresponds to the time derivative of Equation 4. We use Equation 5 to estimate ice shell growth rate as a function of the ice-water interface position, represented by ice thickness (Fig. 5). We assume the ocean is at the melting temperature of 270 K and that the thermophysical properties of the ice shell are represented by pure ice at this same temperature and a pressure of 1 atm (Feistel and Wagner, 2006).

OPEN IN VIEWER

FIG. 5.

(a) Ice shell thickness versus growth velocity and (b) effective solute distribution coefficient versus growth velocity. The effective solute distribution coefficient curves are taken from Fig. 3. The dashed line represents an illustration of the transition from a cellular interface to a planar interface at a critical growth velocity, v_c , represented by two possible values. The solid line depicts the equilibrium solute distribution coefficient from Gross et al. (1977). The diamond markers represent the bounds of growth velocity estimated by Zotikov et al. (1980) for the J-9 Ross Ice Shelf core.

We evaluate four cases, assuming upper boundary conditions of 50 K, 100 K, 200 K, and 250 K to approximate surface temperatures expected at icy ocean worlds. 50 K represents a lower bound surface temperature for both Europa and Enceladus, 100 K represents the mean annual surface temperature of Europa's ice shell (Ojakangas and Stevenson, 1989; Ashkenazy, 2019), 200 K represents the maximum temperature near the tiger stripes of Enceladus (Spencer et al., 2018), and 250 K is intended to represent a terrestrial boundary condition. Higher surface temperatures result in lower growth rates for a given ice shell thickness. Using this model, we can estimate an upper bound on ice shell growth rate and thus estimate the maximum salinity of the bulk ice shell.

Instead of explicitly modeling salt rejection, as in the works of Buffo et al. (2020) and Buffo et al. (2021a), we represent the incorporation of salt as a function of growth velocity by using a model for k v , adapted from Petrich et al. (2011). We prescribe a critical porosity equal to the effective equilibrium distribution coefficient for congelation ice ϕ c = k e q , as opposed to ϕ c = 0 . 0 5 , which is used in the model of Petrich et al. (2011), and we force the model to approach this value at low growth velocities. This yields an expression for the effective solute distribution coefficient given byk v = k e q 1 + k e q 2 v γ s w 0 − 1 + 1 + 4 1 − k e q k e q 2 γ s w 0 v(6)

where k e q represents the effective equilibrium solute distribution coefficient for congelation ice, v is the ice growth velocity, and γ s w 0 represents a scaling parameter related to the interstitial brine velocity (Petrich and Eicken, 2017). Note that our version includes an additional factor of 2 that was excluded from a term in the radicand in the published versions of the original model (Petrich et al., 2011; Petrich and Eicken, 2017).

We fit Equation 6 to the data of Nakawo and Sinha (1981) to obtain γ s w 0 = 3 × 1 0 − 8 m ∕ s , which is similar to the value of γ s w 0 = 4 . 5 × 1 0 − 8 m ∕ s obtained by Petrich et al. (2011). Both our model (cyan) and the original model (yellow) are shown alongside the data of Nakawo and Sinha (1981) in Fig. 3. We estimate the bulk salinity of the ice shell using Equation 6 and the growth velocities obtained from the 1D freezing model (Eq. 5).

We find that the growth velocity transitions to the ice shelf regime (Fig. 3) below ∼100 m depth for all surface temperatures considered. This is similar to the results of Buffo et al. (2020), which found that the salinity profile approaches an asymptotic value below ∼300 m. This supports the conclusion that the bulk salinity for a large fraction of the ice shell will correspond to a value approaching the effective equilibrium solute distribution coefficient.

Note that the lower limit bulk ice shell salinity predicted by Buffo et al. (2020) corresponds to an effective equilibrium solute distribution coefficient governed by the apparent critical porosity in congelation ice (∼0.05). We instead adopt an effective equilibrium solute distribution coefficient of k e q = 6 . 7 × 1 0 − 2 , derived from the Ross Ice Shelf core in Section 4.2.1, to represent the bulk salinity of a congelation ice shell.

Although the critical porosity appears to be a significant factor governing the effective equilibrium solute distribution coefficient in natural congelation ice, as the growth velocity approaches zero, the ice-water interface geometry should become planar and as a result will be incapable of entrapping brine (Eicken, 1998). A planar interface is generally stable for lake ice on Earth because of the relative purity of the water (≲1 ppt); however, the same phenomenon can occur if the growth velocity falls below a critical growth velocity for a higher salinity of water (Wettlaufer, 1992; Maus, 2007a).

The development of a stable planar interface under the appropriate growth conditions is a phenomenon that has been studied in both nature and laboratory experiments for decades (Weeks and Lofgren, 1967; Grothe et al., 2014). In experiments, the transition from a cellular to planar interface coincides with a drastic change in appearance (cloudy to clear) and a reduction in effective solute distribution coefficient that can exceed an order of magnitude (Weeks and Lofgren, 1967; Osterkamp and Weber, 1970; Kvajić and Brajović, 1971; Maus, 2007b). This suggests the potential existence of a congelation ice shell where the bulk salinity is not governed by the critical porosity.

Although the existence of a critical growth velocity is not controversial, the magnitude of the critical growth velocity is challenging to constrain. Morphological stability theory, originally proposed by Mullins and Sekerka (1964), has been leveraged by a number of authors to investigate the development of a cellular interface in the freezing of saltwater systems (Wettlaufer, 1992; Maus, 2007a). The theory has been augmented through the years (Coriell et al., 1985; Sekerka et al., 2015) and is still an active area of research (Maus, 2020).

The theory predicts the existence of a critical growth velocity below which a planar ice-water interface should be stable for any wavelength perturbation. The magnitude of this critical growth velocity is poorly constrained by the theory and is highly sensitive to parameters, including the solution concentration (i.e., salinity), the interfacial solute distribution coefficient ( S i c e ∕ S i n t ), and the temperature gradient in the liquid (Terwilliger and Dizio, 1970; Wettlaufer, 1992; Maus, 2007b, 2020).

To illustrate the onset of this transition during the thickening of an ice shell (Fig. 5), we adopt the values obtained by Wettlaufer (1992) from a linear stability analysis applied to the interface morphology of a sodium chloride system for a solution concentration approximately equal to the Earth's ocean (∼35 ppt). The critical growth velocity of v c ≈ 1 0 − 8 c m ∕ s assumes an interfacial solute distribution coefficient of 0.3, whereas v c ≈ 1 0 − 1 0 c m ∕ s assumes an interfacial solute distribution coefficient of 0.003.

Their results demonstrate that the more efficient the ice is at rejecting the solute, the lower the critical velocity for the onset of interface instability for a given solution concentration. The upper estimate for critical growth velocity is reached for an ice shell thickness less than ∼10 km for all surface temperatures considered (Fig. 5). Again, note that we adopt the values in Fig. 5 for illustration purposes only and the true value could be orders of magnitude lower.

Below the critical velocity, we assume that a planar-ice water interface remains stable and that the bulk salinity of the ice shell will be governed by the equilibrium distribution coefficient (note the absence of “effective”) for congelation ice, where impurities are retained predominantly within the ice lattice (i.e., not incorporated interstitially as brine). However, soluble salts can be accommodated in the ice lattice only up to a certain concentration referred to as the solubility limit. From both natural and artificial samples, the solubility limit for chloride in ice has been inferred to be ∼300 μM (Seidensticker, 1972; Gross et al., 1977; Moore et al., 1994), although in the presence of ammonium the solubility limit increases (Gross et al., 1977).

There is some evidence that the solubility limit may be higher in ice that has undergone recrystallization (Moore et al., 1994), suggesting that marine ice may be able to accommodate more chloride than sub-ice shelf congelation ice. The chloride distribution coefficients obtained by Gross et al. (1977) represent salt entrainment through incorporation of impurities in the ice lattice and serve as the lower bound of equilibrium distribution coefficients for congelation ice.

Their values are similar to earlier works that estimated equilibrium distribution coefficients on the order of 1 0 − 3 for dilute ( ∼ 2 × 1 0 − 4 M ) chloride solutions (Osterkamp and Weber, 1970). For solution concentrations where chloride could be entirely accommodated within the ice lattice and did not occupy interstitial sites, the average equilibrium distribution coefficient was determined to be k e q = 2 . 7 × 1 0 − 3 (Gross et al., 1977).

Note that this distribution coefficient applies to chloride and not the associated cation pair, which was found to be significantly less soluble (Gross et al., 1977). In the presence of ammonium, the equilibrium distribution coefficient increased to k e q = 1 . 4 × 1 0 − 2 (Gross et al., 1977). For more concentrated solutions, the solubility limit was exceeded on crystallization, forcing residual impurities to be accommodated interstitially along grain boundaries. In the case of hydrogen chloride, the distribution coefficient increased from k e q = 2 . 9 × 1 0 − 3 to k e q = 4 . 7 × 1 0 − 3 (Gross et al., 1977; Tison et al., 2001).

Although the distribution coefficient almost doubled at this transition, it was independent of the solution concentration both below and above this transition. It is unclear whether a solution composed entirely of insoluble salts, such as magnesium sulfate, would be accommodated as efficiently because it would be limited to interstitial sites. It is also possible that because of its inability to be accommodated in the lattice, a solution dominant in lattice insoluble salts may promote interface breakdown and enhance interstitial entrapment.

These models imply that the native bulk salinity of a congelation ice shell should be <10% of the ocean salinity, where sub-ice-shelf congelation ice cores imply a bulk salinity between 6% and 7%. There are two cases where we might expect a higher salinity layer to be present near the ice shelf surface: (i) catastrophic melting and subsequent refreezing of an ice shell, although this would likely only extend to ∼100 m depth, and (ii) rapid refreezing of intrusive features, if they extend far enough into the cold ice shell interior (Buffo et al., 2020).

If the ice shell growth velocity is sufficiently slow, such that a planar interface remains stable as the ice shell thickens, the ice shell salinity reduces to <1% of the ocean salinity. For a planar interface at near-equilibrium conditions, the salts entrained are dominantly lattice soluble salts, such as chloride. The experiments of Gross et al. (1977) suggest that the ice chlorinity will be 0.27% of the ocean chlorinity. In the case that the chlorides cannot be entirely accommodated within the lattice, the ice shell chlorinity will be 0.47% of the ocean chlorinity and permit some interstitial incorporation of impurities.

Diagenetic processes can operate to alter the bulk ice shell salinity post-accretion. The flushing of interstitial impurities by meltwater could locally reduce the ice shell salinity, whereas refreezing of meltwater could locally enhance the ice shell salinity. In the ice shells of ocean worlds, meltwater may be generated through tidal heating (Sotin et al., 2002), through frictional heating caused by tectonic activity (Gaidos and Nimmo, 2000; Nimmo and Gaidos, 2002), or by convective currents (Kalousová et al., 2014).

Whether this melt can drain through the ice shell is critically dependent on the ice shell permeability (Kalousová et al., 2014; Hesse et al., 2022). If interstitial impurities are removed due to flushing or drainage, the bulk salinity would be governed by the concentration of impurities accommodated in the ice lattice. For ice saturated with chloride, this would imply an ice shell chlorinity of ∼10 mg/kg, which is on the order of ice shell salinity predicted by Steinbrügge et al. (2020).

Fluctuations in ice shell growth rates have the potential to generate vertical and regional heterogeneities in ice shell salinity. However, predicted growth rates suggest that fluctuations are likely to fall within the low growth velocity regime (Table 2 and Fig. 3), where the effective solute distribution coefficient is relatively insensitive to changes in growth velocity. Peddinti and McNamara (2019) predict an increase in growth rate from 5.67 to 8.22 km/Myr associated with the merging of convective cells within Europa's ice shell, which translates to growth velocities of 1 . 8 × 1 0 − 8 to 2 . 6 × 1 0 − 8 cm/s.

Growth rate estimates obtained by other authors are typically on the order of 1 0 − 9 or 1 0 − 8 cm/s (Table 2). Comparing the lowest estimate of freezing rate for Europa's ice shell in Table 2 (∼1.5 km/Myr) with the age of the surface (∼100 Ma) would imply an ice shell thickness of ∼150 km (Bierhaus et al., 2009). Because Europa's ice shell thickness is believed be an order of magnitude thinner, this suggests that (i) the ice shell has reached a near-equilibrium thickness, (ii) the ice shell is in a state of thermodynamic disequilibrium where melting and re-freezing are occurring continuously, as suggested by Green et al. (2021), or (iii) the estimated freezing rates are potentially an order of magnitude higher than reality.

At Enceladus, observed topographic anomalies are believed to be maintained by melting/freezing less than a few millimeters per year (Čadek et al., 2019; Kang et al., 2021), which translates to growth velocities on the order of 1 0 − 9 c m ∕ s . These growth velocities are comparable to the upper estimate for critical growth velocity at which an ice-water interface becomes planar for a terrestrial ocean (Wettlaufer, 1992). If transitions in growth velocity are such that the ice-water interface stability is affected, this could result in a salinity contrast of up to an order of magnitude associated with this event (Fig. 5). A similar magnitude salinity contrast could be generated by the local and regional accretion of marine ice beneath congelation ice.

The ice-ocean interfaces of icy ocean worlds represent dynamic environments characterized by gradients in ice thickness on both regional and local scales (Nimmo et al., 2007; Nimmo and Bills, 2010; Čadek et al., 2019; Hemingway and Mittal, 2019; Soderlund et al., 2020).

Rifts and basal features, such as crevasses and troughs, represent favorable locations for the formation and accretion of frazil ice in an ice shell. A number of processes have been demonstrated to generate stresses that are sufficient to cause fracturing in the ice shell, including impacts (Roberts and Stickle, 2021; Turtle and Pierazzo, 2001), pressurization due to cooling and thickening (Nimmo, 2004b; Manga and Wang, 2007; Johnston and Montési, 2017; Hemingway et al., 2020), tidal forcing/nonsynchronous rotation (Helfenstein and Parmentier, 1985; Geissler et al., 1998; Greenberg et al., 1998; Hoppa, 1999; Lee et al., 2005; Hurford et al., 2007; Rhoden et al., 2012; Patthoff et al., 2019), and true polar wander (Schenk et al., 2008; Rhoden et al., 2011; Tajeddine et al., 2017).

The fracturing of an ice shell has important implications for surface-ice-ocean exchange and, as such, has been studied extensively. Early work by Crawford and Stevenson (1988) examined both surface and basal fractures as resurfacing mechanisms for Europa's ice shell. They found that direct conduits extending from the surface through an ice shell were unlikely due to the need for high stresses applied rapidly that cannot be supplied by any process believed to be operating at Europa. Basal fractures were also shown to be incapable of extending to the surface; however, they extended over an order of magnitude farther than surface fractures.

Although basal ice is ductile, Crawford and Stevenson (1988) argue that crack initiation and propagation is possible if the ice is strained sufficiently rapidly compared with the Maxwell time. This condition is possibly satisfied by the eccentricity tides that are ∼10⁵ s and comparable to the Maxwell time of ∼10⁴ s (Crawford and Stevenson, 1988).

The model of Lee et al. (2005) showed that surface fractures could penetrate the entire brittle part of the ice shell, in the case where a brittle and ductile layer are mechanically decoupled. They did not study basal fractures, citing that they were less likely to occur than surface fractures based on the increase in ice strength with depth, due to pore closure, and their interpretation of the results of Crawford and Stevenson (1988).

Rudolph and Manga (2009) show that in the presence of a relaxed basal layer, fractures on Europa cannot penetrate the ice shell for thicknesses greater than a few kilometers. Because the gravitational acceleration at Enceladus is a fraction of that at Europa, fractures could penetrate the ice shell for thicknesses up to tens of kilometers (Rudolph and Manga, 2009).

The ice shell thickness where the tiger stripes are located is believed to be <10 km (Hemingway et al., 2020), supporting the interpretation that these features are fractures connecting the ice shell surface to a subsurface ocean (Postberg et al., 2011; Spencer et al., 2018). The ice collapse model of Walker and Schmidt (2015) suggests that basal fractures could form above a subsurface water pocket; however, this mechanism would not necessarily translate to the formation of basal fractures at an ice-ocean interface.

Hemingway et al. (2020) argue that a surface fracture could penetrate a ductile ice layer in an ice shell, so long as it is not too thick, because the layer will exhibit elasticity on timescales that are relevant to fracture propagation. Walker et al. (2021) show that tensile fractures initiating from the base of an ice shell can propagate further into the interior than surface fractures. Furthermore, they showed that a connection between the surface and ice-ocean interface can be achieved if basal tensile fractures connect to the surface through shear failure.

Broadly, these works suggest that basal fractures extending into the ice shell interior are possible—if the basal ice is subject to a sufficiently high strain rate—and that rifts extending through the entirety of an ice shell are unlikely for Europa but possible under specific conditions. Still, many authors attribute surface features at Europa such as domes, pits, and lenticulae to the presence of sills within the ice shell and implicate vertical fractures extending from the ice-ocean interface in their formation (Michaut and Manga, 2014; Craft et al., 2016).

Furthermore, observations and interpretations of putative plume activity at Europa (Sparks et al., 2017; Jia et al., 2018) and Enceladus (Postberg et al., 2011) provide strong evidence that fractures in the ice shell serve as a connection between the surface and some subsurface water reservoir. Where cracks may penetrate the entirety of an ice shell, such as the tiger stripes at Enceladus, the resulting plumes would likely include samples of relatively unfractionated ice formed from agglomerated frazil crystals that nucleated within the turbulent, supercooled water column as the ocean water was brought to the surface.

Given the fast rate of ice formation, the salinity and compositional signal likely experiences minimal fractionation, k ≈ 1 . Conversely, if the plume material were sourced from a reservoir generated from the melt of native ice shell material and not the ocean, our estimate of the effective equilibrium solute distribution coefficient for a congelation ice shell ( k e q = 6 . 7 × 1 0 − 2 ) would predict a saturated ocean at Enceladus (20 ppt/0.067 ∼ 300 ppt). This estimate neglects the effect of brine concentration that may occur during freezing of a reservoir. At Enceladus, plume material is believed to be sourced directly from a sub-ice ocean (Spencer et al., 2018); however, the origin of plumes at Europa is more ambiguous (Sparks et al., 2017).

Ice shell thickness variations on regional scales have been inferred from models and observations of ocean worlds. Models of the ice shell thickness of Enceladus based on observations of the shape (Tajeddine et al., 2017) and gravity (Iess et al., 2014) by Cassini suggest the presence of lateral variations in the ice shell thickness (Čadek et al., 2019). Limb profiles of Europa suggest either a thin ice shell (<35 km) with lateral thickness variations below the detection threshold or a thicker shell in which lateral flow or convection promote a uniform ice shell thickness (Nimmo et al., 2007).

Although the ice shell thickness of Europa is more poorly constrained than Enceladus (Billings and Kattenhorn, 2005; Howell, 2021), multiple models have demonstrated that variations in surface temperature and basal heat flux could promote lateral thickness gradients (Soderlund et al., 2013; Ashkenazy et al., 2018; Čadek et al., 2019; Soderlund, 2019). These lateral thickness gradients could plausibly occur in any icy ocean world with large surface temperature gradients in latitude and/or heterogeneous tidal heating. Because these lateral thickness gradients are unstable (from both a mechanical and a thermodynamic perspective), mechanisms will operate to homogenize the ice shell thickness.

Two mechanisms have been proposed for the homogenization of ice shell thickness: (i) the pressure gradient induced by the variable ice thickness will drive basal ice flow from thicker to thinner regions of the ice shell (Ojakangas and Stevenson 1989; Nimmo 2004a; Nimmo et al., 2007; Ashkenazy et al., 2018) and (ii) an “ice pump,” described by Lewis and Perkin (1986), will operate to melt ice where the ice shell is thick and accrete ice where the ice shell is thin (Vance and Goodman, 2009; Soderlund et al., 2013).

Both properties likely play a role in homogenizing ice shell thickness gradients, although environmental factors such as ocean circulation and tidal velocity will determine which process dominates (Goodman 2018). The ice flux resulting from viscous flow at the base of the ice shell has been estimated to range from fractions of a millimeter to centimeters per year at Europa (Ashkenazy et al., 2018) and less than a few millimeters per year at Enceladus (i.e., on the same timescales as melting) (Kamata and Nimmo 2017), whereas marine ice accretion rates on Earth, driven by the “ice pump,” are on the order of meters per year (Craven et al., 2009).

We, thus focus our discussion on the “ice pump,” which could infill these features on shorter timescales than viscous flow. As the buoyant meltwater is transported along the ice-ocean interface in the direction of decreasing ice thickness, it will become supercooled due to the reduction in pressure and prime the generation of frazil ice.

For terrestrial ice shelves, the ice pump process is approximately adiabatic (Foldvik and Kvinge, 1974; Tison et al., 1998; Koch et al., 2015; Hoppmann et al., 2020). Neglecting heat transfer between water masses is likely only a valid assumption over certain spatiotemporal scales, which may be exceeded when applied to regional scale thickness gradients in the ice shells of ocean worlds. Crevasses, troughs, and rifts, on the other hand, represent high gradient features that can promote substantial supercooling through the operation of a highly localized ice pump.

The magnitude of potential supercooling will be governed by the feature's vertical extent in the ice shell, equivalent to the difference in the pressure melting temperature expected by a reduction in overburden pressure (Fig. 1). These high gradient features also provide a means to shelter the frazil ice from potentially strong sub-ice currents (Soderlund et al., 2020), allowing crystals to accumulate and consolidate, forming marine ice. This process is analogous to the infilling of rifts at the Nansen and Roi Baudouin Ice Shelves by marine ice (Fig. 4).

The texture of the Nansen Ice Shelf core was not columnar, suggesting that no congelation growth had occurred within the rift (Khazendar et al., 2001). This suggests that the infilling of high gradient features in the ice shells of ocean worlds would likely be dominated by frazil ice, as opposed to congelation ice, by nature of both the localized ice pump and the relatively low temperature gradients expected near the base of the ice shell. In this case, the salinity profile will likely decrease with depth within the consolidated layer.

At the permeable-impermeable boundary, the salinity may appear to level off before increasing again as the brine volume fraction increases with depth (Fig. 4). It is possible that if the fracture penetrated far enough into the ice shell such that the surrounding ice was substantially colder, congelation ice could play more of a role as modeled in Buffo et al. (2020).

6.1. Geophysical implications of heterogeneous accretion

The accretion of frazil ice within basal features in a congelation ice shell has significant implications for processes governing surface-ice-ocean exchange. Frazil ice accretion serves as a vehicle to deliver both sensible heat and latent heat into the ice shell interior. Sensible heat is delivered through the introduction of warm ice (relative to the ice shell interior), as frazil ice infills and consolidates within basal features. The relative warmth of marine ice within an ice shelf is supported by borehole measurements from Amery Ice Shelf, which show that the temperature profile within the marine ice layer is nearly isothermal at a temperature close to the freezing point of the underlying seawater (Craven et al., 2009).

The gradual consolidation and interstitial freezing of brine pockets further releases latent heat into the ice shell, serving as an additional mechanism to thermally perturb the ice shell. Because of the timescales of tidal cycles on Enceladus, it is unlikely that a highly consolidated marine ice would be able to form within the tiger stripes; however, the formation and accumulation of frazil in the fissures could be capable of modulating eruptions, a role previously attributed to turbulent dissipation alone (Kite and Rubin, 2016).

Marine ice is more ductile than meteoric ice (Holland et al., 2009; Jansen et al., 2013; Kulessa et al., 2014; McGrath et al., 2014); however, it is still an open area of research as to whether this could be an intrinsic material property or can be attributed to elevated temperatures alone (Dierckx and Tison, 2013; Craw, 2020). The infilling of basal features by more ductile ice could affect the mechanical properties of the ice shell. On Earth, marine ice accretion is believed to play an important role in stabilizing ice shelves against collapse through the infilling of regions of weakness (Holland et al., 2009; Khazendar et al., 2009; Kulessa et al., 2014) and could play a similar role in ice shells.

The observation that fractures propagating in ice shelves arrest when encountering features infilled with marine ice (McGrath et al., 2014) could guide inferences of subsurface properties of an ice shell by using observations of the fractured surface terrain. The accretion of marine ice within suture zones has been shown to channel shear deformation, enabling the decoupling of adjacent units of ice flowing at different velocities (Jansen et al., 2013). As such, accretion in pre-existing fractures could facilitate strike-slip and lateral displacement, believed to be responsible for the linea observed on Europa's surface (Hoppa, 1999; Hoppa et al., 2000; Prockter et al., 2000; Hammond, 2020).

Enhanced ductility within these features might also favor heating over fracturing when subject to tidal deformation, potentially resulting in positive feedback. The enhanced ductility would also increase the Rayleigh number (ratio of buoyancy to diffusion), influencing convective vigor and modulating its responses to tidal forcing. This suggests that marine ice accretion could also play a role in transitioning between convective and conductive regimes in an ice shell.

The marine ice infilling these features is not only warmer but could also be significantly purer than the native ice shell material (Table 3). As such, marine ice is both thermally and compositionally buoyant, which could further promote the formation of narrow diapirs that are believed to be responsible for forming Europa's domes (Pappalardo and Barr, 2004). Soderlund et al. (2013) proposed that marine ice accretion on regional scales, modulated by thickness gradients established by heterogeneous ocean-driven heating, could play a role in the formation of chaos terrain through a similar mechanism (Schmidt et al., 2011).

6.2. Fractionation

To constrain the habitability of an ocean world, it is important to determine whether the composition of the ice shell is representative of the underlying ocean. Knowledge of the composition of the sub-ice ocean would help decipher whether the ocean is in an oxidizing or reducing state (Zolotov, 2008), which, in turn, would help guide future life-detection missions. In a reducing ocean, organisms might concentrate at the ice-ocean interface where surface oxidants delivered to the ocean could drive metabolism-supporting redox gradients, whereas in an oxidizing ocean organisms might concentrate at the sea floor where water-rock interactions and/or hydrothermal activity could supply reductants (Vance et al., 2016; Russell et al., 2017).

Directly sampling the sub-ice ocean to constrain its composition is a significant engineering challenge (Bryson et al., 2020). Studying the surface or near-surface composition, though still an engineering challenge, represents a more feasible approach in pursuit of constraining the composition and thus redox state of the sub-ice ocean (Hand, 2017). The mode of salt entrainment, whether salt is accommodated within the ice lattice or interstitially as brine pockets, can influence the ice shell composition. A cellular interface would be more favorable for the entrapment of brine pockets than a planar interface, resulting in a bulk ice composition that is more representative of the underlying ocean in terms of the relative concentrations of major ionic species.

Because there are very few studies of the chemistry of low-temperature gradient ice, we include studies of sea ice, in addition to marine ice, to identify processes that can result in fractionation of an ice shell.

The composition of sea ice is generally assumed to be representative of seawater (Petrich and Eicken, 2017), although published studies of accreted ice chemistry suggest that some chemical fractionation occurs in sea and marine ice (Table 4). There does not appear to be any evidence that sulfate or calcium is consistently either enriched or depleted in sea ice, although potassium appears to be depleted across all sea ice samples presented in Table 4.

This is consistent with the idea that the degree of fractionation should scale with ion diffusivity, because potassium represents the fastest diffusing ion and thus is more efficiently removed from the ice through networks of brine channels (Maus et al., 2011). The consistent enrichment of magnesium observed in sea ice (Table 4) cannot be attributed to known cryohydrate precipitation and is likely related to its slow diffusivity relative to chloride (Granskog et al., 2004; Maus et al., 2011). Although calcium and sulfate are also slow diffusing relative to chloride, these ions participate in cryohydrate formation early on in sea ice growth (T > −8°C), which could further influence the fractionation signal.

Studies of fractionation in multi-year sea ice cores (Anderson and Jones, 1985; Gjessing et al., 1993) and changes in fractionation with depth observed in young sea ice cores (Maus et al., 2011) suggest that the fractionation signal may evolve as the ice thickens and ages. The mixing model of Reeburgh and Springer-Young (1983) suggests that melt produced from warming as the ice ages removes ionic species conservatively; however, the sea ice samples of Gjessing et al. (1993) show strong sulfate depletion due to washout from melting snow.

The enrichment observed in certain low salinity samples was interpreted to be the result of refreezing of meltwater (Gjessing et al., 1993). Although mirabilite precipitation is often implicated in observed sulfate enrichment (Granskog et al., 2004), the results of Gjessing et al. (1993) and Maus et al. (2011) suggest that sulfate enrichment could be due to the relatively low diffusivity of sulfate. Because chloride can be accommodated in the lattice, it can be preserved in the ice as other insoluble ions retained in interstitial brine are rejected (Moore et al., 1994).

This phenomenon can be observed in samples of marine ice, where the degree of fractionation appears to increase and chloride becomes more enriched as brine volume fraction and salinity decreases (Moore et al., 1994). Snow ice similarly appears to retain chloride relative to other ions when flushed by meltwater, through a process termed preferential elution (Brimblecombe et al., 1987; Davies et al., 1987). Some studies have shown that sodium is removed at a similar rate to chloride and is the least mobile cation (Brimblecombe et al., 1985, 1987; Tsiouris et al., 1985; Davies et al., 1987), which has been attributed to the role of sea salt in atmospheric condensation by Tsiouris et al. (1985) but could also be related to adsorption effects (Davies et al., 1987). These early works were validated by a recent study that was able to quantify the ion exclusion rates governing the process of preferential elution (Costa et al., 2020).

These studies of terrestrial ice fractionation allow us to identify the processes that may alter the chemical fingerprint of the sub-ice oceans of Europa and Enceladus in their ice shells: differential diffusion and flushing by meltwater. An ice shell that entrains salt through the entrapment of brine pockets should initially be representative of the underlying ocean. This is also true for locations where frazil ice accretion occurs, although there will be some chloride enrichment that will decrease with depth, inversely correlated to salinity and brine volume fraction.

If permeable brine networks remain stable over geologic time, differential diffusion may result in a relative enrichment in magnesium and depletion in potassium. This diffusion can still occur through the ice crystals in the absence of brine networks, although far less efficiently (Price, 2000). The presence of magnesium in the ice shell supports the hypothesis put forth by Brown and Hand (2013) that magnesium salts from the ocean contribute to the radiolytic formation of magnesium sulfate salts at the surface of Europa.

The presence of sulfate salts at surface of the ice shell is not necessarily incompatible with their early precipitation. If the ice becomes impermeable at a temperature above which any cryohydrates precipitate, then the composition of the ice should not differ significantly from that of the sub-ice ocean. If cryohydrates were to precipitate in a permeable medium, there is the potential that flushing from melt could remove these impurities from the ice, assuming brine veins were large enough to transport the minerals.

In the case where a planar microstructural ice-water interface remains stable at very low growth velocities, only impurities that are soluble in the ice lattice, such as chloride, would be incorporated in the ice shell. A similar mechanism to generate an ice shell dominated by chloride is by continuous flushing of interstitial impurities by meltwater. In the case where all brine is drained from the ice shell, chloride could still be preserved within the ice lattice.

This indicates that although chloride salts have been observed on the surface and are correlated with endogenous features (Trumbo et al., 2019), this does not necessarily imply that the ocean is dominantly composed of chloride salts.

Furthermore, the association of chloride with resurfacing features is compatible with the near-surface injection of a chloride-rich brine, where sulfate minerals remain in the subsurface, consistent with the hypothesis of Schmidt et al. (2011) for chaos terrain formation and evolution. Vance et al. (2019) also suggest that an ocean rich in sulfates may not be reflected in Europa's surface composition and attribute this to fractional crystallization (i.e., sulfate minerals precipitate out of solution earlier than chloride minerals). The drainage and subsequent refreezing of melt will likely play an important role in redistributing sulfate in the ice shell, generating regions of local sulfate depletion and enrichment, respectively (Gjessing et al., 1993; Maus et al., 2011).

Constraining the detailed physical structure and chemical characteristics of planetary ices has important implications for potential ice-ocean habitats and their ability to retain biosignatures. In icy world systems (e.g., Europa, Enceladus), the stratigraphic and structural evolution of the ice shell, including porosity, temperature, and chemistry, will determine the spatial habitability of the respective cryosphere and determine the preservation/degradation of biosignatures as they are transported through the ice shell (Schmidt, 2020).

Water activity, the availability of water in an environment (Grant, 2004), is an important metric that governs the ability for organisms to grow and reproduce (Stevenson et al., 2015). Water activity strongly influences the ability of organisms to persist in extreme environments such as hypersaline brines on Earth and the surface of Mars (Oren, 2008; Tosca et al., 2008).

Liquid vein networks and brine pockets are important habitats in both sea ice and glacial ice on Earth (Price, 2000, 2007, 2009). Although the brine channels that form in sea ice are recognized as a significant cryosphere habitat (Loose et al., 2011; Arrigo, 2014), the ice must maintain sufficient permeability to enable nutrient exchange in support of maintaining these habitats. For this reason, the cool, impermeable sea ice interior is considered to be a less favorable environment for organisms relative to the ice-water interface, even though the interior represents an environment where sunlight is more accessible (Arrigo, 2014).

Beneath glacier ice, nutrient exchange within liquid vein networks is considered important to maintaining in-ice habitats in the absence of sunlight (Price, 2000, 2007, 2009). Even so, the discovery of sub-ice-shelf anemones that burrow into relatively impermeable glacial ice suggests that organisms may not be inhibited by the lack of available pathways through the ice (Daly et al., 2013). The complex relationship between ice permeability and habitability is highlighted by studies of sea ice microorganisms that generate extracellular polysaccharide substances (EPS) (Krembs et al., 2011; Raymond, 2011). Although the brine volume fraction in ice is increased in the presence of EPS, the similarly enhanced tortuosity results in a net decrease in permeability, which allows brine to be retained in the ice (Ewert and Deming, 2013).

The ability for ice to entrain biosignatures can also be examined independently from its suitability as a habitat. Studies of sea ice have shown that frazil ice can concentrate biological material through mechanical incorporation resulting from the buoyant consolidation of frazil ice crystals (Garrison et al., 1983, 1989; Clarke and Ackley, 1984). Frazil ice also possesses the unique ability to scavenge material, as it is transported through a water column (Garrison et al., 1989; Reimnitz et al., 1993; Arrigo et al., 2010).

A notable example of these scavenging capabilities can be observed in McMurdo Sound, where benthic fauna, mobilized by anchor ice (i.e., frazil ice that accretes at the seabed), have been found at the surface of the ice shelf (Mager et al., 2013). There have not been many dedicated studies examining the incorporation of biosignatures in marine ice; however, one study of protists in the marine ice of Amery Ice Shelf revealed that these organisms were likely sourced from melting sea ice in the neighboring bay and were entrained in the ice as the meltwater was transported beneath the ice shelf (Roberts et al., 2006). This is significant because although marine ice did not serve as the original habitat to these organisms, it could incorporate and preserve these life forms even in the uppermost portion of the ice.

On ocean worlds, radiolytically generated oxidants transported from the surface may represent a viable alternative to sunlight for sub-ice organisms (Chyba, 2000). If oxidant-limited, organisms within the sub-ice ocean will preferentially inhabit the ice-ocean interface where the ice shell serves as source of oxidants. Vertical motion of the ice-water interface driven by tidal deflection of the ice shell could promote nutrient exchange by permitting an influx of ocean water into the ice interior that might replenish habitats at the ice-ocean interface, similar to the tidally driven recharge of nutrients in sea ice (Arrigo et al., 1995; Arrigo and Thomas, 2004).

The enhanced permeability of frazil ice relative to congelation ice may translate to more efficient tidally driven nutrient exchange at interfaces dominated by such an ice texture and thus perhaps a more favorable habitat. An impermeable ice shell interior may imply a lower concentration of preserved biosignatures if organisms migrate with brine toward the ice-ocean interface. Alternatively, the presence of EPS might prevent the drainage of brine habitats, through altering the structure of the ice, and preserve biosignatures even as the habitat becomes progressively more depleted in nutrients over time.

Even if the ice-ocean interface is not inhabited, if life is present in the source water where frazil ice forms, biosignatures will likely be entrained as the frazil ice rises buoyantly to accumulate and consolidate at the ice-ocean interface. Because the relatively pure frazil ice is also buoyant relative to the surrounding ice shell, it can serve as a vehicle to deliver samples toward the surface where they might be sampled by a lander. Features associated with conditions favorable to the accretion of frazil ice can, thus, serve as promising sites for in situ investigations searching for signs of life.

An additional constraint on biological viability as well as biosignature preservation is the chaotropicity and kosmostropicity of fluids within the shell. A measure of the tendency for solutes to stabilize (kosmotropes) or destabilize (chaotropes) proteins and membranes, chao-/kosmo-tropicity impacts the habitability of brines and could limit the survivability of detectable biosignatures as they are transported through the ice shell and subjected to thermal cycling (Hallsworth et al., 2007; Oren, 2013; Pontefract et al., 2017). In many naturally occurring, charge balanced systems, the presence of kosmostropes offsets the destabilizing nature of chaotropes (e.g., seawater); however, if ions are preferentially fractionated through freezing or precipitation reactions, this balance can be upset and lead to toxic chaotropic solutions (Pontefract et al., 2017; Buffo et al., 2022).

One notable chaotrope is chloride, suggesting that an amplified presence in an ice shell due to fractionation could challenge resident biology if concentrations are high enough (Fox-Powell et al., 2016). The ice salinity and fractionation, thus, play an important role in determining the contemporary habitability of the ice shell as well as in controlling the preservation of relict biosignatures. As such, constraining the ice-ocean interface dynamics—which govern the solute entrainment within and biogeochemical evolution of the shell—is an imperative part of assessing the habitability of ice-ocean worlds and designing life detection missions (Des Marais et al., 2008; National Research Council, 2011; Hendrix et al., 2019).