Phase equilibrium in the system CO2-H2O: Application to Mars



[1] Thermodynamic analysis of the system CO2-H2O shows that although CO2 clathrate and liquid CO2 are stable over most of the pressure and temperature range expected for the Martian regolith, hydraulic equilibrium of the regolith pore gas with the overlying atmosphere, a condition most likely in equatorial regions, will destabilize both phases. Condensation on Mars comes in four predictable assemblages: pure H2O ice, CO2 clathrate (hydrate), a eutectic mixture of solid CO2 and either clathrate or H2O ice (if the clathrate fails to nucleate), and under limited conditions, pure solid CO2. Seasonal frosts are likely to consist of thin layers of water ice and/or clathrate, overlain by the eutectic assemblage. Growing ice caps are also likely to be layered, with sets of ice (lower) and eutectic (upper) layers interspersed with pure solid CO2 layers. The saturated water content of a 1-bar CO2 atmosphere with surface temperatures >273 K would have reached 20 precipitable millimeters. Basal melting, triggered by trapped heat flow from the interior, limits the thickness and composition of the polar ice caps and is to be expected during periods of low obliquity, when there is extensive cooling and condensation at the poles. Introduction of liquid CO2 into the Martian regolith by basal melting at the poles may be a major mechanism for loss of the ancient greenhouse atmosphere and its long-term sequestration. Explosive release of CO2 fluid coexisting with groundwater may trigger outflow channel floods and reset atmospheric isotope ratios.

1. Introduction

[2] Many observations show that the dominant condensed phases on the Martian surface are dry (solid CO2) ice and water ice despite a large disparity in atmospheric abundance — 95.3 and 0.03 volume%, respectively [Kieffer et al., 1992]. CO2 and H2O are obviously the dominant reactive species in the Martian volatile system. The reason for this disparity, as I will show below, is that these proportions of CO2 and H2O are close to the solubility limit of ice in gas at average Martian surface temperature (210 K) and pressure (5.6 mbar). Other minor atmospheric constituents [Kieffer et al., 1992] are either inert (N2, 2.7%; Ar, 1.6%) or too dilute (O2, 0.13%; CO, 0.07%) to have an effect. CO2 clathrate and liquid CO2 should also be stable in the crust, although neither phase has been detected on the surface [Carr, 1996]. Also, there are numerous geomorphic features (valley networks, outflow channels) that suggest liquid water near or on the surface in the past [Carr, 1996] and possibly even the present [Heldmann et al., 2005], so any treatment of phase stabilities must account for six phases — three solids, two liquids and gas.

[3] Recent observations indicate that the Martian polar ice caps are composed mostly of water ice [Bibring et al., 2004]. Why this is so is not entirely clear because extensive winter frosts of CO2 cover the polar regions each year and a 1-m-thick residual cap of solid CO2 covers the south polar cap [Kieffer, 1979]. The precipitated equivalent of the present atmosphere is about 10 cm of CO2; if concentrated in the area of the present polar ice caps the thickness of solid CO2 multiplies to about 15 m. Nearly all atmospheric models predict that the atmosphere would have been much thicker earlier in Martian history the atmosphere [e.g., Jakosky and Phillips, 2001]. Estimates range from ∼10 [Jakosky and Jones, 1997] to several hundred times the present thickness [Kasting, 1991]. The models with thinner atmospheres are derived from the ability to model the exchange of CO2 between the atmosphere and observable reservoirs (ice caps, polar layered deposits, porous regolith) and an inability to explain the fate of the CO2 that would have comprised thicker atmospheres. For example, sequestration of CO2 from a 1 bar greenhouse atmosphere in carbonate would require a global layer ∼23 m thick. So far no observation has given any indication of carbonate deposits on this scale. Although the basic justification for positing a thicker ancient atmosphere is cosmochemical — by analogy with models of Venus and the Earth — the major incentive for these models has been to find a set a of conditions under which liquid water would have been stable on the surface [Kasting, 1991].

[4] Figure 1 illustrates a number of salient features of Mars' near-surface environment superimposed on the phase diagrams for pure CO2 and H2O. The range of surface temperatures averaged according to latitude [Kieffer et al., 1976] lies within the fields of CO2 gas and H2O ice, well removed from the stability range of water. A regolith thermal profile (0.015 K/m), typical of those calculated by Clifford [1993], passes through the melting curve of CO2 at a pressure of 30 bars, which corresponds to a depth of ∼0.5 km. Thick arrows illustrate atmospheric thermal profiles of 2.5 K/km [Kieffer et al., 1992] leading to possible formation of H2O clouds at equatorial and midlatitudes and to formation of CO2 clouds in polar regions. Although phase equilibria in the two pure systems account for most of the general features of the Martian volatile systems, what has been missing from previous inquiries into Martian climate is a clear understanding of the phase equilibria that govern gas-ice interactions on the surface and in the porous regolith of Mars and an accurate knowledge of the solubility of H2O in CO2 gas at low temperatures. The first part of this paper provided the necessary framework of phase equilibria. This paper will apply some of these phase equilibria to Mars as well as provide evidence that previous models of ice saturation, based on ideal mixing of CO2 and H2O, may have underestimated H2O solubility at polar conditions by a factor of 10 to 100 or more.

Figure 1.

Pressure-temperature regimes of the Martian atmosphere, surface, and regolith compared to phase relations of the pure H2O (dash-dotted curves) and CO2 (dashed curves) systems. Most Martian data are from Kieffer et al. [1992]. CO2 system data are from Angus et al. [1976], and H2O data are from Weast et al. [1964]. See text for details.

2. Clathrate Stability

[5] Figure 2a illustrates binary equilibria in the system CO2-H2O [Longhi, 2005] along with the regolith thermal profile from Figure 1. Starting with experimental work of Miller and Smythe [1970] and continuing with Kargel and Lunine [1998], a number of workers have pointed out a wide range of environments in which CO2 hydrate or clathrate, (CO2)8-y·46H2O (where y is the number of cage site vacancies per unit cell), should be stable — from the polar surfaces to most of the regolith. However, most spectroscopic techniques cannot distinguish clathrate from water ice and, consequently, few discussions of the regolith have included clathrate. Furthermore, observations of the south polar, residual CO2 ice cap have revealed scant evidence of the H2O [Kieffer et al., 2000] that is integral to clathrate, despite the fact that the clathrate stability field lies between the average P-T conditions at the south pole (the star in Figure 2a) and the stability field of solid CO2.

Figure 2.

(a) Binary system CO2-H2O at low temperatures after Longhi [2005] (reprinted with permission from Elsevier). Here t.p. is the pure system triple point, where solid, liquid, and gas coexist, and c.e.p. is the critical end point. Stable phases in the pure systems are indicated by lowercase italics (ice, water, gas, liq, and solid). Binary phases are given by capital letters as shown in the legend. Binary invariant points (four phases) are labeled according to the absent phases. There are six phases of interest, so the point involving G, H, W, and I becomes {SL}, and the degenerate point involving W plus a GL azeotrope becomes {SHI}. Regolith thermal profile (thick gray curve) is from Clifford and Hillel [1983]. The star represents average south polar temperatures based on the climate modeling and TES temperatures reported by Haberle et al. [2001]. (b) Phase equilibria effects of hydraulic equilibrium between Martian atmosphere and pore gas in Martian regolith. Thick curves are binary equilibria from Figure 2a. Regolith thermal profile is from Clifford and Hillel [1983], and atmospheric profile is from Zurek et al. [1992].

[6] Experimental studies of clathrate nucleation on the surface of 40 to 60 μm ice spheres indicates that at 193 K and 500 mbar it takes on the order of 20 hours for 10 percent of the available CO2 to react to form clathrate [Genov et al., 2004]. Predictably (from their Figure 8), these rates will be a factor of 2 lower at present Martian polar temperatures, and are likely to be lower still at Martian surface pressures (6 mbar). Indeed, Schmitt et al. [2003] report in abstract very slow nucleation rates for clathrate at actual Martian surface conditions, albeit without much detail. Sluggish nucleation therefore may contribute to the apparent absence of clathrate on the surface at present in polar regions, where temperatures are the lowest and kinetics the slowest. Abetting slow nucleation is a very limited temperature stability range at Martian surface pressures, evident in Figure 2a. Furthermore, as I will discuss next, clathrate is not always stable where predicted at first glance by the phase diagram.

[7] The equilibrium mapped in P-T space by Miller and Smythe [1970] is I + G = H where I is water ice, G is CO2-rich gas, and H is hydrate (clathrate). Given a typical estimate of the thermal gradient in the regolith of 15 K/km [Clifford, 1993] and a lithostatic pressure gradient of 140 bars/km [Kieffer et al., 1992], modified for porosity, we might expect clathrate to become stable at depths ≥10 m (Figure 2b). However, one of the conditions of the equilibrium is that all 3 phases be at the same pressure. In a porous regolith exposed to the atmosphere, this is not likely to be so. In the solid framework of the regolith where the mass of the overlying regolith is carried by the framework, the pressure is essentially that of the lithostatic gradient. However, where the regolith pores are connected by open paths to the surface, the pressure of gas in the pores will be given by the atmospheric gradient. Because most of the mass of the regolith + gas system is in the solid, the temperature will be that of the regolith framework. The regolith gas thermal gradient is thus given by the intersection of isobars from the atmospheric gradient (Figure 1) projected into the regolith with isotherms from the lithostatic gradient as shown in Figure 2b. Pressures of the regolith gas gradient are well below the lower stability limit of clathrate for depths up to at least 5 km. So clathrate is not likely to be found in abundance anywhere near the surface in porous regolith. Where are regolith pores open to the atmosphere? Not at high latitudes where abundant ground ice (e.g., Feldman et al. [2004] may seal pores and eliminate communication with the atmosphere. Here, clathrate will form at depth by reaction of ice with either solid CO2 or a CO2-rich fluid. However, quantitative models of ice stability [e.g., Clifford, 1993] indicate that in equatorial regions ice may be absent in the regolith from the surface down to depths of tens to hundreds of meters.

[8] The one situation in which clathrate stability on Mars is likely to be enhanced relative to the binary system shown in Figure 2a is at very low temperatures. The phase diagram shows a lower temperature limit of ∼130 K for clathrate. At this temperature, however, condensation of solid CO2 would have dropped CO2 in the atmosphere to only about 5% of its present level, and at lower temperatures nitrogen would become the major constituent of thinning atmosphere. Predictably, the increasing N2/CO2 ratio would drive the {LW} quadruple point and its associated three-phase equilibria to lower temperatures, and any clathrate that formed would of course contain N2, but with N2/CO2 lower than the N2/CO2 ratio in the atmosphere.

[9] A somewhat different analysis applies to liquid CO2 (L), which the phase diagram (Figure 2a), predicts to be stable at pressures higher than 30 bars (depths > 300 m). If the regolith were to become saturated with liquid CO2 at some depth (a “carbonifer”), the solid regolith matrix would transmit some of its load to the interstitial CO2 liquid and thus stabilize most of the liquid CO2 layer. However, at the interface between the pore gas and the carbonifer liquid CO2 would be subject to the projected atmospheric pressure in the regolith and thus would boil. Gradually the gas-liquid interface would move downward and consume all the available liquid CO2. Thus liquid CO2 will be stable only in crustal environments that are sealed off from the atmosphere.

3. Ancient Atmospheres

[10] Higher erosion rates and dendritic drainages in the Noachian period suggest a warmer climate with more fluid (probably water) activity near the surface [Baker, 2001]. Numerous models of dense ancient atmospheres have been advanced to provide higher surface temperatures [Kasting, 1991; Forget and Pierrehumbert, 1997]. Most of these models have invoked CO2-rich atmospheres for two obvious reasons: one is that CO2 constitutes more than 95% of the present atmosphere; the other is the example of the other major terrestrial planets, Venus and the Earth. Venus has a dense CO2-rich atmosphere and the Earth would have one also if it were not for the burial of carbonates and hydrocarbons in the crust [Holland et al., 1986]. A minimum estimate of Venus' carbon, 2.7 × 10−5 kg/kg [Hoffman et al., 1977], is derived simply from what is observed in the atmosphere. Holland et al. [1986] estimate 1.7 × 10−5 kg/kg of carbon to be in the Earth's crust, oceans, and atmosphere. A conservative estimate of the mantle abundance brings the total terrestrial carbon abundance to Venus' level. Ratios of volatile to involatile elements in the Martian meteorites suggest that volatile elements on Mars should have abundances comparable to, if not in excess of, those of Venus and the Earth [Dreibus and Wänke, 1985]. If so, then Mars' total C abundance is at least 19 × 1018 kg or 7.1 × 1019 kg of CO2, which is more than 3000 times the mass of the present atmosphere [Kieffer et al., 1992] and more than 450 times the amount of CO2 thought to be adsorbed on regolith particles [Fanale and Jakosky, 1982]. Even allowing for incomplete degassing of the interior and extensive loss of the atmosphere during the ancient epoch of heavy meteorite bombardment [Melosh and Vickery, 1989], the former existence of a dense (≥1 bar) CO2 atmosphere is certainly worth considering. Consequently, it is necessary to examine the applicable phase equilibria and solubility limits in such atmospheres. Unfortunately, there is very little data on the CO2-H2O system in the range of pressures (0.1 to 4 bars) at which ancient atmospheres have been modeled [Spycher et al., 2003], so it is necessary to extrapolate available data from higher pressures.

[11] Figure 3 illustrates temperature-composition (T-X) sections on the basis of my earlier work [Longhi, 2005] and on published solubility data [Song and Kobayashi, 1987; Wiebe, 1941]. The sections are drawn with partial log scales on the composition axis to emphasize CO2-rich compositions. The compilation of Spycher et al. [2003] shows that in the range of pressure and temperature of Figure 3 the concentration of CO2 in liquid water coexisting with a CO2-rich phase varies only from ∼1.5 to 3 mole%. The higher-pressure sections show that at regolith depths >220 m liquid is the stable form of pure CO2 (if the pores are sealed) and that the CO2-rich limbs of the various two-phase fields are relatively steep with respect to the log-scaled composition axis. Because of the extremely limited solubility of H2O in CO2 fluids at low temperatures there is less than 1 K depression of the CO2 melting curve near the triple point or anywhere along the sublimation curve.

Figure 3.

Temperature-composition (T-X) sections at constant pressure modified from Longhi [2005] (reprinted with permission from Elsevier): (a) 34.5 bars, (b) 20.7 bars, (c) 6.9 bars, and (d) 1.0 bar. Published solubility data (open circles) are from Song and Kobayashi [1987]; open circles indicate data from the CO2 system [Angus et al., 1976]. Note logarithmic scale on the composition axis. A mole fraction of 0.00001 H2O is assumed to be “pure” CO2. Dashed curve illustrates the ideal mixing approximation for the composition of gas coexisting with ice. Depths in Figures 3a and 3b are approximate regolith depths in meters based upon a surface density of 2600 kg m−3 that increases linearly to 3000 kg m−3 at 10 km depth. Shaded areas labeled Treg in Figures 3a and 3b are temperatures given by thermal profile in Figure 2a. Cloudless and cloudy greenhouse conditions in Figure 3d are from Pollack et al. [1987] and Mischna et al. [2000], respectively.

[12] The topologies of the sections change as various invariant points are crossed. At pressures <74 bars the gas-liquid azeotrope (the binary loop with a minimum) develops in CO2-rich fluids. At pressure ∼6.9 bars the azeotrope disappears and a gas + ice field, relevant to the surface of Mars, becomes extensive; but unfortunately there are no more data at lower pressure for this assemblage. At 6.9 bars (Figure 3c) an estimate of the composition of CO2-rich gas in equilibrium with ice is shown as a broken curve. This estimate is based on the Lewis and Randall Rule for ideal gases

equation image

where pi is the partial pressure of component i, x is mole fraction, and P is total pressure. In the present case pi is calculated from an empirical fit to the ice sublimation data tabulated by Weast et al. [1964],

equation image

where T is temperature in Kelvins. Xi is then calculated from equation (1). At temperatures where ice is stable, the agreement between the experimental and calculated solubility (mole fraction of H2O in CO2 gas in equilibrium with ice) is excellent. At lower temperatures (i.e., below ∼260 K), the metastable extension of the gas + ice field is constrained to lie within the gas + hydrate field. Although the exact slope of the limb of the gas + ice two-phase field is not known, it is clear that the limit of the gas + ice two-phase field must be steeper than the ideal gas approximation and thus the difference between the ideal gas approximation and the real composition of gas in equilibrium with ice widens with decreasing temperature and may be as much as an order of magnitude at temperatures typical of the Martian surface (∼220 K). Similar disparities also apply to the composition of gas coexisting with clathrate — the ideal gas approximation of this solubility limit must lie to the low-H2O side of the ideal gas/ice curve. Nonideal mixing of CO2 and H2O in the gas phase is, of course, to be expected because of the presence of the CO2-rich critical point at 304 K [Wendland et al., 1999] and the coexistence of two CO2-rich fluids at lower pressures.

[13] One bar is the lowest pressure at which there are solubility data for H2O in CO2-rich fluids (Figure 3d). Unfortunately, the extant data are limited to fairly high temperatures, so the two-phase field boundaries at low temperatures are drawn with slopes similar to those in the sections at higher pressure. At the upper limit of ice + gas stability (∼270 K) the graphical estimate of the H2O concentration in the gas closely agrees with the ideal gas approximation, but as in Figure 3c, the disparity between the ideal gas approximation and the graphical estimate of H2O solubility in gas widens at lower temperature. A pressure of 1 bar CO2 is well within the range of dense atmospheres modeled for early Mars. An average Martian surface temperature of 223 K, which is the result calculated by Pollack et al. [1987] for a 1-bar CO2 atmosphere at 70% solar insolation with no clouds, is shown in Figure 3d. At 223 K the gas coexisting with ice has a mole fraction of 0.00035 H2O, which is similar to the present atmospheric concentration (0.0003) but is nearly a factor of 10 greater than the ideal gas approximation. Because the total pressure in Figure 3d is higher than that of the present atmosphere, a similar concentration implies a higher H2O partial pressure: in this case 0.00035 × 1 = 0.35 mbar, which is ∼200 times the present value of 0.0017 mbar. Nonetheless, the precipitable water content of the 1-bar atmosphere remains a paltry 2 mm distributed globally, as compared to the 45 m currently in the perennial polar ice caps. As an illustration of how arid the climate might be, imagine that seasonal temperatures under optically thick CO2 clouds [Mischna et al., 2000] were locally high enough to stabilize a body of water. Saturated air above the water would have a H2O content of 0.0045. If a parcel of air with average water content (0.00035) were then heated and blown over the body of water, its relative humidity would be <10%.

[14] These considerations suggest that Mars with a 1-bar atmosphere and an average surface temperature of 223 K would be much like the present: poleward of 40°–50° latitude ground ice would probably be stable, while equatorial regions would be arid. For the most part the surface of Mars under such an atmosphere would have looked very similar to the present surface. The major climatic difference would have been the vastly increased water transport capacity of the atmosphere from the equator to the poles. On the other hand, if continuous CO2 cloud coverage increased surface temperature sufficiently to stabilize liquid water (from 223 K to 273 K), then the water content of the atmosphere could increase by approximately a factor of 10 at 1 bar, reaching ∼20 precipitable millimeters (pr mm). Such amounts of H2O not only imply further increases in the transport capacity of the atmosphere, but also suggest the possibility of significant precipitation as rain or snow consistent with interpretations of higher erosion rates on early Mars [e.g., Carr, 1996]. Note also, however, that the H2O concentration in ice-saturated gas decreases with increasing pressure at constant temperature, so that the total water content (partial pressure) does not increase linearly as the atmosphere pressure increases. For example, at 273 K the H2O concentration in saturated gas at 273 K drops from 0.0045 at 1 bar (Figure 3d) to 0.001 at 6.9 bars (Figure 3c). Thus the partial pressure of H2O at 273 K in this pressure range increases from 0.0045 × 1 = 0.0045 bar to 0.0010 × 6.9 = 0.0069 bar, which is only a factor of ∼1.5 despite a sevenfold pressure increase. Therefore 30 pr mm of H2O is a probable upper limit for ancient, CO2-rich atmospheres in which temperatures remained close to ice stability.

[15] Another potentially significant feature of a denser atmosphere would be a much wider range of stability for clathrate. Because of the higher atmospheric density, Mars' average surface conditions would lie very close to or within the stability field of clathrate. Whether clathrate would precipitate directly from the atmosphere at 1 to 2 bars remains problematical because clathrate nucleation remains relatively sluggish even at 3 to 5 bars and T ≤ 230 K [Genov et al., 2004]. However, it is possible that the increased atmospheric pressure would have speeded up reaction kinetics sufficiently to transform exposed ground ice to clathrate in polar regions.

4. Present Conditions

[16] Even though there are no experimental solubility data available at present for Martian surface conditions, it is possible to approximate the solubility limits and draw a phase diagram for CO2 and H2O mixtures near the Martian surface. The phase diagram in Figure 2a provides the temperatures of the unary two-phase and binary three-phase equilibria. There are two binary equilibria of interest involving CO2-rich gas at Martian surface pressure, a gas eutectic e (S + H = G) and a peritectic p (H = G + I). It is possible to draw a qualitatively correct diagram for temperature and composition with only this information. However, to make the diagram quantitative some estimate of the H2O content of CO2-rich gas in equilibrium with ice is needed. Figure 4a illustrates such an estimate based upon orbiter measurements of H2O atmospheric abundances [Jakosky and Haberle, 1992] and Viking Lander measurements of surface pressure and frost point temperatures [Hess et al., 1977]. These observations fix the gas-ice solubility limit at a mol fraction of approximately 0.00045 to 0.0006 at 192–194 K and 7 mbar. At higher temperatures there is little difference between the solubility curve and the ideal gas approximation for H2O-rich compositions. However, the frost point estimate of concentration is a factor of 10 higher than the ideal gas approximation at the same temperature. Thus the slope of the gas + ice field boundary must steepen with respect to ideal gas approximation at low H2O concentrations in order to pass through the frost point estimate and, consequently, the disparity between the ideal gas approximation and the estimated H2O content of Martian air increases with decreasing temperature, reaching is a factor of 100 at polar temperatures. Qualitatively, the departure from ideal gas behavior is similar to that inferred for 6.9 and 1.0 bars, but the departure takes place at a lower temperature, which is to be expected for a lower total pressure. Because of elevation and hence pressure differences, the mixing properties of CO2 and H2O in the gas phase are likely to be less nonideal on top of the southern ice cap (∼2 mbar in southern winter) and more nonideal in the north polar region at the same temperature.

Figure 4.

Temperature-composition relations for the present Martian atmosphere. (a) Surface pressure and frost point temperature at the Viking Lander 1 (VL1) at LS (areocentric longitude of the Sun) site from Hess et al. [1977]; H2O concentrations are based upon modeling of Jakosky and Haberle [1992]. Here p is a peritectic point: G + I = H, and e is a eutectic: S + H = G. (b) T-X relations at the limits of Martian surface pressure.

[17] The diagram in Figure 4a not only allows the observer to predict precipitation in the Martian atmosphere (see below), but also allows an estimate of the average temperature below which surface ice is stable (and hence a latitude above which surface ice is stable). Clifford and Hillel [1983] made such an estimate on the basis of the intersection of the average atmospheric composition and the ideal gas-ice saturation curve at ∼200 K. They concluded that ground ice should be stable poleward of ±40° latitude. Schorghofer and Aharonson [2005] constructed a more sophisticated model based on diffusive exchange between the atmosphere and ground ice buried beneath a layer of dry, porous regolith. They place the limit of ground ice at 50°. In Figure 4a the average atmospheric composition crosses the empirical gas-ice saturation curve at ∼163 K. The zonally averaged Thermal Emission Spectrometer (TES) data presented by Haberle et al. [2001] gives a southern hemisphere latitude of ∼52° for 163 K. Other factors, such as dust and elevation content will affect the stability of surface ice, but it is interesting to note that Feldman et al. [2004] report ground ice deposits poleward of ±50° latitude. The approximate agreement between prediction and observation of ground ice has an important implication: the water content of the Martian atmosphere is largely buffered by water ice.

[18] Figure 4b shows estimated constant pressure sections for the approximate limits of surface pressure limits of Mars at present. Kieffer and Zent [1992] calculated an average pressure of 4.1 mbar for the top of the south polar ice cap. However, reported winter temperatures <134 K [Kieffer et al., 1976] are consistent with seasonal pressures of 1 mbar and less for pure CO2. To avoid complications caused by the increasing N2/CO2 ratio, I am assigning a low-pressure limit of 2 mbar for the purpose of illustration. The high-pressure limit of 10 mbar, which is appropriate for the northernmost lowlands is taken from the figures of Pollack et al. [1990]. These sections show that with decreasing pressure the temperature of condensation of a given composition decreases, the temperature interval over which clathrate may condense decreases, and the composition of the eutectic assemblage becomes more H2O rich. These sections thus make it possible to predict a range of Martian condensation sequences. For example, for an atmosphere with average composition ice condenses first over most of Mars, followed by clathrate (at peritectic p any ice crystals in the atmosphere will react with the gas at constant temperature and gas composition to form clathrate until the ice is consumed), then by the eutectic assemblage, solid CO2 + clathrate. Because of slow nucleation, clathrate may not condense from the atmosphere [Schmitt et al., 2003] in which case the condensation of ice is followed by a solid CO2 + ice eutectic assemblage at a temperature slightly below the true eutectic at the intersection of the gas + ice and the solid CO2+ gas field boundaries. However, if the eutectic layers are buried and become nonporous, clathrate will very likely form at solid CO2/ice contacts. In general, the amount of ice that precipitates before the onset of the eutectic assemblage decreases with decreasing pressure or increasing surface elevation. The eutectic assemblage is predominantly solid CO2: the approximate compositional range shown in Figure 4b as mole fractions of 0.00009 at 10 mbar to 0.0003 at 2 mbar implies volume fractions from 60 (71) to 200 (220) ppm water ice.

[19] Because of the low H2O content of the atmosphere, condensation of ice and/or clathrate will have negligible effect on the pressure. However, once eutectic condensation begins, continued cooling will cause the atmospheric temperature and pressure to drop and the H2O concentration of the saturated atmosphere to increase. Local drops in pressure will induce winds that advect air from areas of higher pressure. The most common case will involve air with higher water contents blowing toward the winter pole from equatorial regions [Leighton and Murray, 1966]. Repetitions of this scenario will produce a repeated sequence of layers: ice, and solid CO2 + ice. Thus seasonal CO2 frosts in the northern hemisphere, at least, should typically have a thin layer of ice beneath them. However, advection is also possible within polar regions if cooling is localized and more CO2 condenses in one area than another. The area that condenses more of the eutectic assemblage will have higher H2O concentration, lower pressure, and lower temperature than adjoining areas. The consequent pressure gradient could drive in polar air having a lower water concentration than the ambient eutectic. In such cases, which are more likely in the southern hemisphere, solid CO2 will condense first, followed by the eutectic assemblage, giving rise to a different sequence of layers.

[20] In many situations sublimation will necessarily be the reverse of condensation: the eutectic assemblage will be the first to sublime, producing a local pressure increase and lower H2O concentration in the overlying atmosphere. What happens next depends on what phase(s) is exposed. In the case of northern seasonal frosts, once the eutectic assemblage is consumed, ice will sublime next (or clathrate, if present, and then ice) with increasing temperature and atmospheric H2O concentration at constant pressure. Sublimation of southern seasonal frost may be similar or may involve a pure CO2 layer. In the latter case pressure would increase and H2O concentration decrease at nearly constant temperature until the CO2 layer was consumed.

[21] Figure 5 shows the phase diagrams for the pressure extremes from Figure 4 along with measured seasonal water concentrations from Jakosky and Haberle [1992]. A condensation column is drawn on the basis of the composition extremes for each hemisphere. The relative heights of the water ice columns have been increased by a factor of 1.3, which is the ratio of the molar heats of sublimation of dry ice (25.7 kj) and water (12.5 kj) at 175 K divided by their respective molar volumes (28.8, 19.3 cm3 [Angus et al., 1976]). Consequently, for the same initial pressure the columns represent the same amount of cooling. Figure 5 demonstrates that although both types of condensation sequences are likely, the ice-first (I) sequence is likely to be prevalent in the northern polar region, whereas the CO2-first (II) sequence, if present at all, will likely be restricted to the southern polar region. The actual situation will be more complex because surface pressure changes with the season [Hess et al., 1977]. Nonetheless, it should be clear that different condensation sequences are possible on Mars. Furthermore, temperature alone is not the sole controlling factor of stability. If it were, a residual CO2 layer on the south pole cap would be very unlikely because the average temperature at the south pole (∼155 K [Kieffer et al., 1976]) is higher than the sublimation temperature of pure CO2. Undoubtedly, the CO2 residual layer owes its persistence to sluggish sublimation kinetics. Burial, compaction, and recrystallization of the perennial CO2 layer would certainly make it less susceptible to sublimation than a CO2 seasonal frost that remained on the surface. However, sublimation kinetics are also likely to be slower in pure solid CO2 than in the eutectic assemblage because the presence of even a miniscule amount of ice (or clathrate) in the eutectic assemblages creates loci of high surface energy that facilitate the jumping of atoms from the solid to the gaseous state. So high purity may contribute to the persistence of the residual CO2 layer at the south pole, as well.

Figure 5.

Temperature-composition relations for the present Martian atmosphere at the limits of surface pressure: (a) north polar region and (b) south pole. Clathrate is assumed not to condense, so its phase boundaries have been left out. Seasonal water concentrations (shaded regions) are from Jakosky and Haberle [1992]. Columns show the normalized thicknesses of the various assemblages derived by complete condensation of the indicated compositions. Condensation fills columns from the bottom up.

[22] Recent measurements by Doute et al. [2005] suggest additional complications. They reported 300 to 800 ppm H2O (by volume) in south polar ice cap measured with the Mars Express imaging spectrometer. Converting volume proportions of the solid ices to molar proportions yields mole fractions of 0.00044 to 0.0011 H2O. These concentrations are higher than and do not overlap the range of ice proportions in the eutectic assemblage estimated from the phase diagrams in Figure 4b. The extra ice may reflect scattered layers of water ice and/or clathrate lying on the surface or beneath holes in the CO2 layer, but in either case the amount of detected water obscures the distinction between a pure CO2 layer and a eutectic assemblage.

5. Climate Change

[23] A final application of CO2-H2O phase equilibria comes in the area of climate change, specifically basal melting of polar ice caps. Clifford [1993] has suggested that addition of layers of ice and dust would induce basal melting of ice deposits already close to their melting points, and that this melting would lead to recharge of the global groundwater system. At present the atmospheric H2O abundance is the equivalent of 10 μm. If all of this water were to be deposited on the polar ice caps, the thickness would increase by 150 times to ∼1.5 mm. In a cloudless 1- to 2-bar CO2 atmosphere the potential thickness of atmospheric water is ∼200 times the present value or 0.3 to 0.6 m (based on Figure 3d). In a cloudy atmosphere with 30 pr mm (see above) focusing at the poles could produce ice thickness of 4.5 to 9 m. This is hardly enough ice to cause any significant melting or recharge. The recharge of the groundwater at the poles would require transfer of many times the ambient atmospheric H2O content via equator-to-pole transport. Such transport would imply the evaporation of a large near-surface reservoir — one produced by giant floods, episodes of volcanism, or sublimation of ground ice deposited in equatorial regions during periods of high obliquity. On the other hand, the mass of the present Martian atmosphere (150 kg m−2) is equivalent to a 0.10 m thick global layer of CO2 or 15 m concentrated at the poles [Kieffer et al., 1992]. In a 1- to 2-bar atmosphere the potential polar CO2 thickness is 3 to 6 km. The potential of such thicknesses is very significant.

[24] Both Mellon [1996] and Kargel and Lunine [1998] and have called attention to the fact that water ice is a much better heat conductor than solid CO2 or CO2 clathrate (hydrate). Consequently, much greater thicknesses of ice caps composed of water ice are possible before basal melting ensues. Moreover, mixtures of solid CO2 and hydrate melt more than 50 K below ice (Figure 2a). These two factors greatly limit the thickness of solid CO2 and/or clathrate-rich polar caps. Figure 6a illustrates two thermal profiles calculated by Mellon [1996] for polar ice caps according to the equation

equation image

where Tz is the temperature at some depth z in the ice cap, Tms is the mean surface temperature, H is the heat flow, and ke is the effective conductivity of the ice cap. The higher pressure curve is appropriate for the conductivity of ice (2.8 Wm−1 K−1); the other reflects the lower conductivity of solid CO2 or hydrate (0.5 Wm−1 K−1). Where the thermal profiles cross the melting curves, the depth equivalents of pressure are indicated in kilometers. The calculations show that the maximum height of a pure CO2 cap is 1 km. This result is independent of a similar result reported by Nye et al. [2000] based on mechanical strength. On the other hand, the maximum height of a water ice cap is 11 km. Secondary results are that even traces of solid CO2 would melt out of a thick water ice cap at depths ≥6 km. Somewhat more difficult to appreciate is the fact that a pure hydrate ice cap could attain a maximum thickness of ∼2 km (the intersection of the thermal profile with the H = L + W curve). Such an ice cap would be unlikely in a thin atmosphere because of slow nucleation rates and a small temperature stability range. However, extensive cooling of atmospheres with several bars of CO2 (cf. Figure 3c) might easily produce thicker layers of clathrate than ice. Obviously, mixtures of water ice, dry ice, and clathrate will have intermediate thermal profiles, although Mellon [1996] showed that the effective conductivity is skewed toward pure CO2 or clathrate conductivities. That is, 20% dry ice may reduce the effective thermal conductivity by as much 50% if the mixture consists of alternating layers rather than random inclusions.

Figure 6.

CO2-H2O phase relations of Martian polar ice caps. (a) Calculated one-dimensional conductive thermal profiles of H2O ice (2.8 Wm−1 K−1) and CO2/clathrate (0.5 Wm−1 K−1) for a heat flow of 0.03 Wm−2 and an average surface temperature of 155 K. (b) Polythermal/polybaric section along the CO2/clathrate conductive profile in Figure 6a.

[25] Results from the Mars Orbiter Laser Altimeter (MOLA) reported by Smith et al. [2001] show both polar ice caps to be approximately 3 km above datum. This means that the upper limit for solid CO2 and/or clathrate is about 20% at present. As shown in Figure 6b, which is a polybaric/polythermal section along a conductive heat flow profile, solid CO2 and H2O ice are incompatible in the ice caps and so metastable eutectic assemblages of solid CO2 and H2O ice will convert to solid CO2 + clathrate and selvages of clathrate will develop between the contacts between H2O ice and any layer containing solid CO2. Perennial ice caps are most likely to have reached maximum thickness during periods of low obliquity when insolation at the poles is weakest [Kieffer and Zent, 1992; Jakosky et al., 1995], although cooling of the winter poles during the periods of highest obliquity may have been sufficient to promote growth of seasonal cap at the winter pole [Mischna et al., 2003]. Calculations reported by Toon et al. [1980], Haberle et al. [1994], and Nakamura and Tajika [2001] have predicted three general climatic states: a high-pressure (0.5–5.0 bars) CO2 atmosphere with perennial CO2 ice caps, an intermediate pressure condition in which enhanced atmospheric heat transport is likely to have inhibited perennial CO2 ice caps even during periods of low obliquity, and a third low-pressure state similar to the present one in which CO2 layers are found only as seasonal frosts and as perennial layers with limited extent. In the latter 2 regimes growth of an ice cap is likely to have involved burial of CO2 frosts by H2O ice in successive years of decreasing obliquity. Interbedded layers of different ices are consistent with observations of folds and boudinage structures in the south polar ice, as pointed out by Kargel and Tanaka [2002]. As shown above, however, even greenhouse atmospheres never contained sufficient H2O to contribute to growth of the caps: the growth of the caps required extensive near surface H2O, either as equatorial ground ice [Jakosky et al., 1995] or as large bodies of water [e.g., Clifford and Parker, 2001]. The structure of a thick cap is predictable: CO2 layers will be present in the top and melted out in the bottom as illustrated in Figure 7. Precisely how thick depends on mean annual temperature, heat flow, dust, and the proportions of CO2 and H2O ices. The most important feature of basal melting, however, is that melting of solid CO2 (+ clathrate) begins at least 50 K below the melting of H2O ice.

Figure 7.

Cartoon of the effects of obliquity variation on basal temperatures and melting of polar ice caps.

6. CO2 Crustal Reservoir

[26] Various models predict condensation of the equivalent of several times the present Martian atmosphere at the poles during periods of low obliquity [Ward et al., 1974; Bills, 1990]. In view of the potential for condensing several kilometers of solid CO2 ± hydrate at the poles from a greenhouse atmosphere during such an episode and the ∼1-km limitation placed by thermal models, the prospects for generating large amounts of liquid CO2 (Figure 7) seem quite reasonable. At pressures more than a few tens of bars liquid CO2 is not only denser than ice, but at pressures > 170 bars liquid CO2 is denser than water at the melting point of ice (Figure 8). Also, the thermal gradients are shallower than the melting curve. This means that masses of liquid CO2 will tend to sink through or flow through surrounding ice into the underlying regolith without freezing. Depending upon the depth at which the liquid CO2 encounters groundwater, the CO2 may either flow above or below the groundwater. In either case at depths corresponding to the 285 K isotherm the two liquids will each dissolve some of the other and thereafter come to equilibrium.

Figure 8.

Densities of liquid CO2, water, and ice as functions of temperature and pressure. Data are from Angus et al. [1976] and Grigull [1989]. Numbers give depth equivalents in km of ice/regolith.

[27] The coexistence of two liquids, as shown in Figure 2a (the H = L + W curve) and Figure 6b (the liquid CO2 + water two-phase field), at depth in a porous regolith is a topic that has received relatively little attention, but is one that merits attention from several points of view. Perhaps the most significant feature is the prospect for long-term sequestration of CO2 and concomitant thinning and cooling of the atmosphere. Another application may lie in understanding the mechanism of giant floods where generating sufficient flow through an aquifer is a problem [e.g., Wilson et al., 2004]. Subjacent layers of liquid CO2 that are much more compressible (hence expandable) than liquid water, may play an important role in augmenting flow. Not only may CO2 force the water above it to percolate through an aquifer more rapidly, but the expanding CO2-rich fluid may also induce hydrofracturing in the aquifer, thereby greatly increasing the flow. On the other hand, if liquid CO2 lies above the water table and the frozen crust above it cracks, an explosive eruption is possible that may disrupt many square kilometers of crust. Widespread explosive evacuation of a confined fluid is likely to leave jumbles of crust on the kilometer scale that are highly porous on a fine scale. Explosive disruption of the crust is a plausible explanation for chaotic terrains, which are the apparent sources of many outflow channels [Carr, 1979]. Most investigators have concluded that water is the main fluid agent in outflow channels [Carr, 1996], whereas Hoffman [2000] has proposed density flows of rock, soil, and ice riding of a bed of CO2 gas. Given the possibility of both liquids in the crust, we must consider the possibility of a two-stage flow — the first being a CO2 debris flow, the second being a carbonated water flow. Finally, dissolution of CO2 in water will greatly increase its corrosiveness — raising the possibilities of groundwater increasing the porosity and permeability of the Martian crust in some places and precipitating carbonates in others.

[28] Although the thermal profiles in ice and solid CO2 derived from present heat flow and atmospheric conditions (Figure 6a) suggest that much thicker water ice caps would be stable with respect to basal melting, it is a relatively simple matter to show that this excess capacity would not have existed early in Mars' history. Even though Mars' current heat flow can only be estimated at present, it is certain that the heat flow would have been significantly higher in the past. Precisely how much more depends upon time and distribution of heat-producing elements (K, Th, U). Accordingly, I calculated conduction profiles in ice and solid CO2 with equation (3) for various heat flows that were converted to age according to the model of Stevenson et al. [1983]. Two mean surface temperatures were employed: one being 155 K, which is 5 K higher than the CO2 sublimation curve at 6 mbar; the other being 197 K, which is 5 K higher than the CO2 sublimation curve at 1 bar. The various thermal profiles were plotted on the phase diagram and the pressures at the points of intersection of the thermal profiles with the melting curves were then converted to thickness of ice or solid CO2. Figure 9 illustrates the results of these calculations. At heat flows expected for 4.0 Ga ago [Stevenson et al., 1983] the thickness of a CO2-rich ice cap in a 6-mbar atmosphere is only a few tens of meters, and the thickness of a CO2-rich ice cap in a 1-bar atmosphere is essentially zero. Equally important, however, is the result that the maximum thickness of a water ice cap drops below the current thickness of the northern ice cap (3 km) at ∼4.0 Ga for a 6-mbar atmosphere and at ∼1.8 Ga for a 1-bar atmosphere. This means that early in Mars' history much of the water presently tied up in the upper levels of the ice caps would have resided somewhere else. One possibility is that the ice caps would have occupied a greater area. A larger area of surface ice would have decreased surface heating due to insolation and perhaps caused an ice age. Alternatively, basal melting may have kept the ice caps at their maximum thickness while charging the global aquifer system envisioned by Clifford [1993] and Clifford and Parker [2001]. In this case the regolith beneath the ice caps would have remained relatively free of ice and hence porous.

Figure 9.

Maximum thicknesses of solid CO2 and ice in polar caps derived from intersection of thermal profiles and melting curves as in Figure 6. Time–heat flow model is from Stevenson et al. [1983].

[29] These conditions are conducive not only for the flow of water into the deep regolith, but also for the flow of liquid CO2. However, because CO2 is highly compressible in the fluid state, its disposition would have varied considerably throughout Mars' history. The higher heat flow early in Mars' history would have effectively limited the amount of solid CO2 that could reside in the polar ice caps, but would not have limited the amount of CO2 that condensed at the poles (potentially several kilometers) in low-obliquity excursions. As discussed above, the ice caps would have initially consisted of three or four assemblages: pure water ice, possibly primary clathrate layers, eutectic mixtures of solid CO2 and water ice or clathrate, and pure solid CO2. Within the ice caps the interfaces between pure CO2 and H2O ices would have developed reaction layers of clathrate, as required by the phase diagram. Similarly, any metastable eutectic mixtures would have transformed to mixtures of solid CO2 and clathrate. As the temperature increased within the ice sheet, melting would have begun at the solid CO2 + clathrate eutectic (S + H = L) in Figure 6. From Figure 8 a minimum of 1.1 km of overlying ice is required for the liquid CO2 to be denser than the surrounding ice and hence to sink. For the case of a H2O >> CO2 ice cap the CO2-rich liquid would be constrained to follow the CO2-rich limb of the CO2 liquid + clathrate (L + H) two-phase field as it descended along cracks in ice. In doing so, the CO2 liquid would dissolve H2O and react with ice to form clathrate selvages. Reaction of liquid CO2 (37.7 cm3/mol [Angus et al., 1976]) and water ice (5.75 × 19.49 cm3/mol) to form clathrate (118 cm3/mol [Circone et al., 2003]) involves a 20% volume reduction, so formation of clathrate would actually enlarge cracks and channels. If the ice cap were at its maximum thickness as suggested above, then the CO2-rich liquid would reach the bottom of the ice sheet approximately at the temperature of the clathrate + water eutectic (∼271 K: H + I = W). However, the density crossover between liquid CO2 and water at this temperature is approximately 170 bars, as shown in Figure 8, which corresponds to 5 km of overlying ice. Figure 9 shows that this thickness of ice is not possible in a thin atmosphere until ∼1.8 Ga ago, according to the Stevenson et al. [1983] thermal model. A greenhouse atmosphere would have inhibited such thicknesses until ∼0.6 Ga. Consequently, beneath thinner ice caps the liquid CO2 would have floated above the water.

[30] If the regolith were saturated with water, then the liquid CO2 would react with the water to form a clathrate septum separating the two as-yet incompatible liquids. This reaction involves a 12% volume decrease, so the clathrate would not form an impervious barrier. Rather, as more liquid CO2 accumulated above, the clathrate boundary layer would migrate downward increasing in both temperature and pressure along the lithostatic gradient until the temperature reached approximately 285 K — the temperature of the H = L + W equilibrium. At greater depths and higher temperatures clathrate is no longer stable, and the two liquids would now coexist stably. Alternatively, if the regolith were not saturated with water, the CO2 would initially expand to fill the regolith pores, possibly turning to gas. However, as long as the ice cap and surrounding ground ice sealed the regolith off from the atmosphere, the pores would eventually become filled, the pore pressure would eventually approach the lithostatic pressure, and the trapped CO2 would condense to form a liquid again.

[31] The eventual disposition of the liquid CO2 would depend on the temperature and depth (pressure) at which it encountered groundwater. At 285 K the density crossover between pure liquid CO2 and water occurs at approximately 280 bars, as shown in Figure 8. Dissolved salts in the groundwater would make the water denser and easily double the pressure of the density crossover. Lowering the temperature at which clathrate breaks down to liquid CO2 plus water is a competing but small effect — 6 wt% NaCl in water lowers the temperature of the H = L + W equilibrium only 2 K [Larson, 1955]. Within the Martian regolith, the CO2 liquid might lie entirely above or below the water layer, or the liquid CO2 might wind up sandwiching the layer of water. As described by Clifford [1993], the weight of the ice cap should drive both liquids initially down, then away from the poles, and eventually up toward the surface at low latitudes, following the base of the cryosphere. As long as an impermeable layer of ice persists in the regolith and the temperature of the two-liquid contact remains above 285 K, both liquids should coexist stably. However, there are a variety of possible disturbances. If the two-liquid contact rises above the level of the 285 K isotherm, clathrate would precipitate at the interface. Decompression might lower the density of the liquid CO2 below that of the water, producing an overturn. Likewise, increased dissolution of salts into the groundwater might have increased its density to the point of overturn. Geological evidence suggests that subsurface ice affects topography poleward of 40° latitude [Mustard et al., 2001] and observations by the orbiting Mars Odyssey gamma ray spectrometer [Feldman et al., 2004] are consistent with surface ice poleward of 50°. Conversely, ice is not generally present in sufficient quantities near the surface in equatorial regions to affect topography and models of ice stability in the Martian regolith [e.g., Clifford, 1993] indicate that ice may be absent to depths of a few hundred meters near the equator. As illustrated in Table 4 of Clifford [1993] the absence of ice in the regolith in equatorial regions at present, although not likely, is within the uncertainty of the models. The possibility of ice-absent regions increases significantly with increased heat flow (age), greenhouse heating, and insolation (low obliquity). If ice were absent, then regolith pores would be in hydraulic equilibrium with the atmosphere, thereby destabilizing liquid CO2 and clathrate. Liquid water would also not stable in the regolith in these regions — not because of low pore pressure — but because the atmosphere is not saturated with H2O in the equatorial region. Leakage of CO2 out of the regolith at low latitudes would have undermined sequestration over geologic time, unless the pores were somehow sealed. Migration of groundwater may provide just such a mechanism.

[32] Beneath the poles groundwater is made acidic in response to its saturation with CO2. However, groundwater that has migrated away from the poles through the regolith is very likely carrying a complement of ions dissolved along the way. If the groundwater were to encounter a region of ice-free regolith with pore gas in hydraulic contact with the atmosphere, the groundwater would begin to evaporate back into the atmosphere, causing the dissolved ions to precipitate as salts and carbonates within the regolith. The edge of the ice-free ground ice is likely to have moved poleward at low obliquity and toward the equator during high-obliquity excursions [Kieffer and Zent, 1992]. Thus there may be a geographically extensive band of carbonate veins and cemented regolith at low latitudes to midlatitudes, developed over geological time, that not only is now a sink for CO2, but also seals off from the atmosphere the putative global aquifer and a large reservoir of liquid CO2 sequestered deep within the regolith. The carbonate vein fillings in Martian meteorite ALH84001 that have C and O isotope ratios consistent with precipitation from low-temperature crustal fluids [Valley et al., 1997] may be products of this process.

[33] Finally, it is probable that other sources contributed to the crustal fluid reservoirs. General mantle degassing and localized igneous activity would have produced juvenile fluids that migrated through the crust. Although the temperature-depth gradients along which these fluids traveled changed with time and location, predicting a typical sequence of phase stabilities is nonetheless possible on the basis of the gradient shown in Figure 1 and the phase diagram in Figure 6b. Given the much greater solubility of H2O in magmas relative to CO2 [Dixon and Stolper, 1995], initial degassing produces fluids with CO2 >> H2O. H2O-rich fluids, evolved later from the magma, are likely to be highly reactive and not as mobile. Ascending CO2-rich fluids will eventually unmix to form a CO2-rich liquid and carbonated water. If the 2 liquids separate, the water will eventually solidify to a mixture of ice and clathrate, while the CO2-rich liquid precipitates clathrate and becomes richer in CO2. In this scenario, which is appropriate for low to intermediate heat flow and low surface temperature, the juvenile fluids are likely either to become fixed on their own in the crust or to become incorporated in a fluid reservoir within the crust. During greenhouse episodes thermal gradients may miss the liquid CO2 stability field entirely, ultimately leading to the addition of CO2 gas to the atmosphere and water to the crust. Stable isotope studies of Martian meteorites summarized by Jakosky and Jones [1997] reveal that H, O, and C isotope ratios of trapped gases and alteration phases are all well fractionated with respect to expected magmatic values, implying that addition of juvenile fluids to the Martian crust-atmosphere system has not been sufficient to eliminate the fractionations developed over geological time.

7. Timing

[34] Although introduction of liquid CO2 into the Martian regolith by basal melting of the polar ice caps became possible relatively early in Mars' history as the polar ice caps began to exceed thicknesses of ∼1 km and liquid CO2 became denser than the basal ice, the subsequent fate of the CO2 would have depended on density contrasts with groundwater. Figure 10 illustrates some of the interplay between heat flow (time), latitude, and the density of volatile phases at the approximate temperature limit (285 K) of two-liquid stability. The plus and minus symbols indicate the presence or absence of a 3 km ice (H2O) cap. Obviously, the depth to the 285 K isotherm is greater at high latitudes than at the equator and it increases with time as the heat flow decreases. There is a modest decrease in the depth of the isotherm beneath the ice cap itself. Also, liquid CO2 remains less dense than water until the depth equivalent of ∼180 bars. Figure 11 depicts some of these interplays schematically. As discussed above, liquid CO2 would have been less dense than any groundwater it encountered until the ice cap thickness reached 5 km. Up until this time liquid CO2 would have floated on top of any aquifer and during the earliest period of Martian history most CO2 melted out of the ice caps would have quickly returned to the atmosphere. As heat flow diminished and the ice caps thickened, liquid CO2 became denser than the relatively pure water beneath the polar ice caps and thus basally melted CO2 would initially have sunk through the aquifer, leaving selvages of clathrate at the interfaces between the two liquids because liquid CO2 and water do not coexist until ∼285 K. However, because the thermal expansion of CO2 is much greater than that of water, there would have been conditions where the density of liquid CO2 would have fallen below that of water before the sinking CO2 blobs reached the 285 K isotherm, whereupon the two liquids would come to equilibrium with each other and coexist without the intervening clathrate.

Figure 10.

Pressure and depth in regolith to the 285 K isotherm as functions of time and latitude. Pressure in the regolith is calculated by assuming a surface density of 2600 kg m−3 that increases 0.04 units per meter, such that at 10 km the density is 3000 kg m−3. Temperature is calculated from equation (3), where thermal conductivity is roughly approximated by a model based upon the discussion of Clifford [1993] that assumes ke = 1.0 W m−1 s−1 at the surface and increases with depth, such that ke = 3 W m−1 s−1 at 10 km. Heat flow versus time is from Stevenson et al. [1983]. Surface temperature versus latitude is from Haberle et al. [2001]. Shape of symbols changes with latitude; color of symbols changes with heat flow/time. Note that the x axis is offset for different heat flow values.

Figure 11.

Sketch of liquid H2O and liquid CO2 relationships on Mars as functions of time (heat flow). Here 285 K marks the low-temperature stability limit of coexisting water and liquid CO2. Where the liquids come into contact at T < 285 K, a thin layer (selvage) of clathrate develops (thick dashes). The 4.0 Ga diagram shows a thin but laterally extensive solid CO2 layer to accommodate the atmospheric model of Haberle et al. [1994] and heat flow constraints discussed in the text. There are two outcomes for fluids migrating away from poles: Water deposits carbonate veins as it rises toward the surface equatorward of the ground ice limit; local decompression (fault, meteorite impact) leads to CO2-driven blowout.

[35] The entry of liquid CO2 into the regolith offers the possibility of sequestration of atmospheric CO2 as liquid, as carbonate, and as clathrate. Confined CO2 might also have acted as an accelerant or propellant for the outflow channels. However, long-term storage of CO2 in the regolith would have competed with sputtering processes at high altitude [Jakosky et al., 1994] and chemical weathering in the crust [Carr, 1999] to reduce the mass of the atmosphere. Evaluating the contribution of basal melting through time requires knowledge of the distribution of heat-producing elements. If most of Mars' incompatible elements were fractionated into the crust during primordial differentiation, as argued by Zuber et al. [2000], then heat flow would have diminished much earlier than predicted by the Stevenson et al. [1983] model (heat producing elements remain in the mantle), so ice caps of a given thickness and basal melting would have been possible earlier than predicted by the models presented above. Conversely, a more gradual drop off in heat flow may have allowed sputtering and weathering to reduce the atmosphere significantly before basal melting became effective.

[36] The possibility of all three processes operating simultaneously is likely to complicate interpretation of stable isotopes. Sputtering losses tend to leave behind a remnant atmosphere with higher values of stable isotope ratios, such as 13C/12C; whereas sequestration of CO2 in condensed phases, like carbonate and liquid CO2, ought to produce a remnant atmosphere with lower values. Briefly, as discussed by Jakosky et al. [1994], the 13C/12C isotope ratio in the Martian atmosphere is higher than expected if a dense CO2 atmosphere had collapsed by formation of thick carbonate deposits now buried in the crust. The observed 13C/12C isotope ratio is apparently more consistent with sputtering processes that leave an isotopically heavy residue. Consequently, Jakosky et al. [1994] favor a relatively thin postbombardment atmosphere with no more than a few tens of millibars of CO2. On the other hand, they also point out that in the absence of a dense CO2 atmosphere loss of a thin atmosphere by sputtering and photochemical escape should have also produced a much larger 15N/14N fractionation than observed. However the occasional release of isotopically heavy water and CO2 stored in the crust into a thinning atmosphere by catastrophic floods may resolve some of these contradictory observations. Carr [1996] estimates the volume of individual outflow channel floods to be on the order of 105 km3; a detailed analysis of flow laws by Wilson et al. [2004] indicates that typical flood volumes have been overestimated by a factor of 25. These volumes translate to a range of approximately 0.5 to10 × 1016 kg of water, which is comparable to the mass (2 × 1016 kg) of the current atmosphere [Kieffer et al., 1992]. If, as suggested above, CO2, sequestered as a liquid in the crust, was a propellant or accelerant for the extruding water and comprised even a modest fraction of the extrusive mass, then it is easy to envision these extrusions repeatedly resetting atmospheric 13C/12C isotopic values to an extent not accommodated by existing atmospheric models. Furthermore, sequestration of CO2 in the crust as liquid and as carbonate veins permits a denser early CO2 atmosphere that would reduce sputtering effects on nitrogen thus accounting for the limited fractionation of nitrogen isotopes noted by Jakosky et al. [1994].


[37] I thank M. Max and an anonymous reviewer for thoughtful comments that helped to improve the clarity and substance of this paper. This research was supported by NASA grants NAG5-12074 and NNG05GL91G. Lamont-Doherty Earth Observatory contribution 6896.