Introduction

Sea-water contains some 34%₀, by weight of dissolved inorganic substances: this quantity (essentially) is known as its salinity (Reference Sverdrup, Sverdrup, Johnson and FlemingSverdrup and others, 1942). While this parameter may vary from place to place, the ratios of the major ion concentrations stay reasonably constant at the values shown in Table I. Salinity is commonly established through the determination of one of these ions (often chloride). The same concepts are frequently extended to sea ice. It is less clear that this is justified : that is, we should be confident that the major ion ratios remain reasonably close to their sea-water values. The chemical analyses of Reference Bennington and Kingery Bennington (1963), on young Arctic sea and pack ice, do suggest that this is the case and that there is relatively little selective concentration of any particular ions with depth. The present paper reports the results of additional chemical analyses. It includes an investigation of natural sea ice by means of thinner sections than those employed by Bennington giving finer resolution with depth. Analyses are also reported upon artificial ice prepared from sea-water in the laboratory and upon the clear glassy ice which is sometimes obtained when brine is frozen slowly (Reference Weeks, Lofgren and Öura Weeks and Lofgren, 1967), Little evidence of selective concentration was found in any region, and the results are, at least generally, in accord with those of Bennington.

Sea ice is produced when sea-water freezes downward from the surface under unidirectional cooling. It is an inhomogenous material which undergoes a complicated series of phase changes with decreasing temperature as various solid salts precipitate out from the entrapped liquid brine. In a well-known paper, Reference AssurAssur (1958) has analyzed the data of other workers to arrive at a postulated scheme of phase relationships. At about — 1°C (relatively) pure ice crystals appear, thereby concentrating the brine in their immediate vicinity. At about —8°C the precipitation of solid Na₂SO₄-IOH₂O commences and, at successively lower temperatures, other solid salts appear: NaCI-2H₂O near — 23°C, KC1 near — 37°C, and MgCI₁₂H₂O near — 43°C. The latter four transitions have been indirectly observed by-means of electrical measurements (Reference AddisonAddison, 1969; Reference AddisonAddison, 1975). Additional transitions were seen: an unidentified one at — 47°C, that of (probably) CaCl₂.6H₂O near —53°C, and the probable ultimate solidus point of the system between —70 and — 75°C. Assur, in addition, suggests the deposition of minor amounts of CaCO₃.6H₂O near —4°C and of MgCl₂.8H₂O near — 20°C, although he mentions thai the behaviour of both these salts may be somewhat erratic.

The structure and formation of sea ice have been frequently described (for example, Reference Harrison, Tiller and KingeryHarrison and Tiller, 1963; Reference PounderPounder, 1965; Reference Weeks and AssurWeeks and Assur, 1968). The region just below the surface often arises from the solidification of frazil ice and consists of small ice crystals having random orientation of the (optic) c-axes, along with entrapped brine and solid salts. The body of the ice shows the well-known columnar structure where the crystal zones have predominantly horizontal c-axis orientations. This arises from the preferred growth of ice crystals, under a temperature gradient, in directions perpendicular to the optic axes (Reference Perey and PounderPercy and Pounder, 1958). A “transition layer”, a few millimeters in thickness, lies between these two regions. It is created after the solidification of the frazil layer, when, of all those crystals nucleated, the ones most favourably oriented with respect to the temperature gradient grow preferentially. These overwhelm the others and establish the columnar structure. The transition layer is often found 5 to 25 cm below the surface. It is a region especially high in salt content. This arises partly because of its chaotic structure, and partly because of brine which drains from the more porous upper material and is retained by the less permeable material below.

New sea-ice salinities range from some 4%₀ in regions well below the surface, to io%₀, or more, in the upper regions. In general, the faster the rate of freezing, the higher will be the salinity (Reference Fertuck, Fertuck , Spyker and Husband.Fertuck and others, 1972; Reference Weeks, Lofgren and Öura Weeks and Lofgren, 1967). The salts are associated with liquid brine entrapped within the ice: either in rows of thread-like brine cells lying between the platelets of pure ice which make up a crystal zone, or at the inter-zone boundaries. The formation of these platelets has been described by Reference Harrison, Tiller and KingeryHarrison and Tiller (1963) and the phenomenon has been considered, with some success, in terms of constitutional supercooling (for example, Reference Milośević-Kvajić, Miloševic–Kvajić, Kvajić and BrajovićMilošcvić-Kvajić and others, 1973; Reference Kvajć, Kvajić, Brajović, Milošević–Kvajić, Pounder, Whalley, Whalley, Jones and GoldKvajić and others, 1973)

Once the ice has been formed however, the brine is not static. Various methods for the transport of brine have been suggested (for example by Reference Kingery, Goodnow and KingeryKingery and Goodnow, 1963; Reference Untersteiner and ÖuraUntersteiner, 1967; Reference Lake and LewisLake and Lewis, 1970). These include drainage mechanisms, flushing by melt water, and expulsion through temperature cycling. Such processes are thought to account for a large part of the brine transport in natural sea ice.

A second, more microscopic, process also exists: the migration of closed liquid-brine inclusions (towards warmer regions) along a temperature gradient (see, for example : Reference PounderPounder, 1965; Reference HarrisonHarrison, 1965; or Reference Hoekstra, Hoekstra, Osterkamp and WeeksHoekstra and others, 1965). While accounting for a relatively small portion of the total brine movement, it seems to be the mechanism most likely to alter the ion ratios after the ice has formed. Specifically, the first solid salt to be deposited, Na₂SO₊- ioH₂O, could be left behind in such a migration and thereby appear somewhat displaced from the other impurities. Brine-cell migration is slow below 15°C

The motion of a liquid inclusion under a temperature gradient has been the object of considerable study recently. Of interest is the observation of Reference JonesJones (1973) that, depending on the shape of the inclusion, its path need not necessarily be parallel to the temperature, gradient, but may in fact lie at a considerable angle to it.

Artificial sea ice, frozen in the laboratory under conditions approximating unidirectional cooling, shows a structure which may be described as a “scale model" of the natural ice structure. There is no frazil-ice region. The transition layer, a millimeter or so in thickness, is found in the first centimeter beneath the surface. Below it, we again find the tendency towards vertical columnar crystal zones. These, however, are rather smaller than in natural sea ice: about 1 cm in diameter and, at most, 7 or 8 cm in length. By keeping the surface temperature approximately constant during freezing, we eliminate brine-transport mechanisms dependent upon temperature cycling and, in general, produce a material which has a simpler thermal history. Any selective concentration might be expected to show up more clearly in artificial ice.

The Present Study

Two samples of artificial ice and one of young sea ice (in the form of a 7.6 cm (3 inch) core, transported in dry ice from Fort Churchill, Manitoba, Canada) were sectioned along horizontal planes by means of a lathe with a milling device. The resulting “snow" was melted and the melts analyzed quantitatively for Cl−, SO₄ ²⁻, Na+, Mg²⁺, Ca²⁺, and K+.

Artificial samples were prepared in a cold room by freezing 30 1 of natural sea-water (salinity, 34.4%₀; measured ion ratios similar to Table I) in a container about 40 cm in diameter by 65 cm deep. With the exception of the top, which was protected but not sealed, this vessel was completely insulated. Small weighted air bladders were placed on the bottom to relieve pressure during freezing. Examination of thin sections of the ice under crossed polaroids, showed that there was always a small central cylindrical volume, some 12 cm in diameter, at the top of the mass, where the temperature gradient had been essentially vertical during freezing. As this freezing system was closed, rejected ions accumulated in the underlying brine. Experience showed that some 15 cm of ice could form before effects of this enrichment were apparent. Consequently, the study was confined to the top 10 cm of ice. Freezing of the mass took approximately 6 weeks.

Two lots of artificial ice were produced : one at a constant ambient temperature of — 30°C and the other at — I5°C, a temperature chosen to effect the slow growth rates necessary for the production of glassy ice.

The natural sea ice was studied to a depth of 30 cm by means of sections approximately 5 mm thick. The — 30°C artificial ice was shaved in 2 mm sections. Because of its extremely low salinity, the — 15°C artificial ice was cut into 5 mm sections in an attempt to satisfy the requirements of chemical analysis.

In spite of the air bladders, a small quantity of brine was forced up along the sides of the freezing vessel on to the surface of the — 30°C ice. This took place when the freezing was virtually complete, long after the upper ice had formed. The liquid was removed quickly and the ice was sectioned soon afterwards. The effects of this flooding were, therefore, confined to the surface itself and may be readily identified.

The thickness of the sections was usually chosen to yield about 10 ml of melt. Each sample was eventually diluted quantitatively by a factor of 20 to provide a suitable working volume. Analytical methods were chosen for simplicity and suitability, rather than for accuracy.

In any mixture of ions, some usually interfere in the determination of others. The sea-water group is especially troublesome in this regard. As a consequence, the anions and the cations were studied separately. Before the anion determinations, half of the sample was passed through an acidic cation exchange resin (to replace the cations by hydrogen) and the pH adjusted to neutrality. Similarly, before the cation determinations, the remaining half was subjected to an anion exchange resin to replace these ions by Cl−.

The chloride ion was determined by the well-known silver nitrate (Mohr) titration (Reference KolthoflT and SandellKolthoff and Sandell, 1952). At lower concentrations, the end-point was somewhat subjective, but little real difficulty was encountered. Sulphate was measured colorimetrically with barium chloranilate at 330 nm (Bertolacini and Barney, Reference Bertolacini and Barney1957, Reference Bertolacini and Barney1958).

Na+, K+, and Ca⁺⁺ were determined by flame photometry at 589 nm, 768 nm, and 554 nm, respectively (Reference Burriel-Marti and Ramirez-MunozBurriel-Mani and Ramirez-Munoz, 1957). These ions interfere mutually in the flame in a rather complex manner (Reference Kuemmel and KarlKuemmel and Karl, 1954 ; Reference Broderick and ZackBroderick and Zack, 1951). A method was devised to obviate this difficulty (at least partially). When potassium was being determined, an amount of sodium equal to the average amount of this ion present in the particular group of samples under study was added to the potassium standards. Similarly, average amounts of both sodium and potassium were added to the calcium standards during the determination of this ion. Magnesium was measured colorimetrically at 435 nm by means of the dye titan yellow (Reference SandcllSandell, 1959).

It is felt that this scheme generally served to determine these ions to within ±5%, or better, with the exception of calcium, where the error might be as high as ±8%. Analyses of typical synthetic mixtures of known concentrations confirmed these estimates. (An additional check (of a sort) was also carried out. For each sample, the ion concentrations were converted to normalities and the positive and negative sums compared. Usually, these sums “balanced" to within ±4%) or better.) For the extremely dilute solutions from the — 15°C artificial ice the errors were sometimes greater. A number of these samples were not measurable and, for some of the others, the methods were pushed to the limits of usefulness.

Discussion

Fig. 1. Ion concentration profile.·, for saline ice grown at -30C. (Zero depth indicates the true surface oj the ice.)

The concentration profiles for the — 30°C and — 15°C artificial ices and the natural sea ice are shown, respectively, in Figures 1, 3, and 5; and the resulting ion concentration ratios in Figures 2, 4, and 6. In each of the latter, vertical lines indicate the values of these ratios in sea-water.

With some exceptions, the ion ratios show a tendency to remain fairly near the sea-water values. This suggests that the application of the concept of salinity to sea ice, at least to bulk samples, is reasonable. There is little evidence of selective ion concentration.

Fig. 2. Ion ratios for saline ice grown at — 30°C.

Fig. 3. Ion concentration profiles for saline ice grown at —15°C. (The top 4 cm showed the clear, glassy structure of fresh ice. )

The ion profiles of the — 30°C ice (Fig. 1) contain few surprising features. As a consequence of some visible condensation, and the previously mentioned flooding during freezing, the natural surface of this ice (as revealed by sections) lay somewhat lower than the apparent surface. The zero level in Figures 1 and 2 indicates this natural surface. (The erratic behaviour of some of the ions in the first one or two sections is a consequence of the flooding.) The apparent transition layer, with enriched salt content, is clearly indicated about 5 mm below the true surface. The section corresponding to the peak in concentration, and the three directly below it, form a pattern which is reproduced in all of the profiles except that of sulphate. This ion does not, on first glance, appear to show a transition-layer peak. However, it seems probable that this feature is present in the sulphate profile at a level one or two sections higher and that the peak has merged with the high spurious concentration of this ion at the surface. If so, it suggests that the transition-layer process took place about 2 mm closer to the surface than the other profiles indicate, and that the liquid brine migrated downwards leaving the sulphate behind. The behaviour of the Na+/Cl− ratio at this level is consistent with this speculation. The ion ratios suggest a slight enrichment of magnesium in all but the top 2 cm.

Fig. 4. Ion ratios for saline ice grown at — 15°C.

The — 15°C ice contained a layer which was very similar in appearance to fresh ice. Except at the very surface, the lop 3.5 cm were absolutely clear and glass-like. The usual columnar structure became established only below 6 cm. Analysis confirmed that the clear ice had rejected virtually all of the impurities. Several of the ions could not be determined here. (The vertical dotted bars in Figure 3 provide visual continuity and are not meant as detection limits.) Below this layer, the salinity increases just about monotonically with depth. There was no visual sign of a transition layer, nor are there concentration peaks to indicate the presence of one. (The peaks on the sulphate profile are attributed to experimental error at these low concentrations.) It is interesting to note that the ion ratios of Figure 4, where experimental error permits them to have meaning, are, even here, reasonably close to the sea-water values.

Figure 5 shows the results obtained for the natural sea-ice sample. The top 7 cm have the expected high salt content. At the bottom of this region, small peaks coinciding with the transition region are seen in the Na⁺, Cl⁻, and SO₄ ²− curves. Below this, the salinity falls off as the structure became more regular. It then rises again, possibly because of faster freezing due to increasingly cold winter weather at the surface. The sulphate profile is again somewhat different from the others. The zone of decreased salinity between 7 and 15 cm, and the differing behaviour of the sulphate curve, undoubtedly arise, at least in part, from downward brine transport. For example, the rather high values of the SO₄ ^2minus;/Na⁺ ratio (Fig. 6), at (and above) the transition layer, suggest that solid sodium sulphate was left behind in a region which consequently suffered a depletion of brine through drainage and migration.

Fig. 5. Ion concentration profiles natural sea ice.

Fig. 6. Ion ratios for natural sea ice.

The K⁺ /Cl⁻ and (especially) the Ca²⁺/Cl⁻ ratios are smaller than the expected values. For our other samples, more reasonable concentrations of these ions were found, with one ion or the other showing a slight preponderance. This suggests the possibility of some systematic error in the Ca²⁺ determination. However, for just this reason, known control samples were included with each batch of samples during analysis. It is hoped that the probability of such an error having escaped detection is not high. In any case, the concentrations of these ions are small and they make only a small contribution to the salinity calculation. (Unfortunately, a sample of sea-water from the site was not available.)

Our profiles for natural sea ice may be compared with those of Bennington (in particular, with his figures 4 and 7). The comparison must be rather indirect since the 30 cm depth of the present study corresponds to just two of his thicker sections. Our curves show larger variations than do his: this is reasonable since his measurements were averaged over much larger samples. Within these limitations, there seem to be few significant differences between the two studies, especially when we take into account the wide variability of this natural material. Bennington’s K⁴ /Cl⁻ and Ca²⁺/Cl²⁺ ratios were similar both to each other and to the sea-water values. Depending on his particular sample, either ion might show a slight excess in concentration over the other. In general, all of his ion ratios exhibited smaller departures from the sea-water values than were observed in the present case.

On the basis of these two investigations, we may conclude, with a fair degree of generality, that the impurity ratios in sea ice remain encouragingly similar to those in sea-water. The results of the present work are largely corroborative. In view of the large expenditures of time and effort demanded by the chemical analyses, the project was not pressed further.