General considerations

There are two separate problems connected with ice orientation: (1) the start of freezing, and (2) orientation selection or change after freezing has commenced. The best way of studying the start of freezing is to watch freeze-up in the field, but this has not been done; therefore, the discussion must be in terms of later observations. Suffice it to say that the first growth of ice in lakes may be random, or consistently orientated with either horizontal or vertical c-axes. In agreement with the experiments of Reference Lyons and StoiberLyons and Stoiber (1959), an initial c-axis-horizontal orientation is always very fine-grained (order of 1 mm.) at the top and randomly orientated in the horizontal. Ice which starts with vertical c-axes is generally very coarse-grained (order of centimeters), as would be expected for freezing in quiet water. Ice of initial random orientation starts with intermediate grain size. It is certainly reasonable to postulate that wind action is responsible for these varying orientations, as indicated by Reference Lyons and StoiberLyons and Stoiber (1959).

Observations of the ice on Peters and Shrader Lakes in the spring of 1961 support the idea that wind is the main factor. When all the snow had ablated and the ice had candled, different surface orientations over the whole lake could be determined by inspection (Figs. 1–3) The c-axis-vertical ice was white and the c-axis-horizontal ice was darker, because of the detailed mechanism of candling. (The same phenomenon is reported by Reference BarnesBarnes (1961), also at Peters Lake.) Candling in the fine-grained, c-axis-horizontal ice is simply grain-boundary melting, while the whiteness of the c-axis-vertical ice is attributable to tubules (Ragle, unpublished) and air bubbles formed when solar radiation causes internal melting in large crystals of this orientation.Footnote *

Fig. 1.

Candled ice surface of Peters Lake, Alaska (Shrader Lake in distance). The light-colored ice has vertical c-axes, whereas the dark-colored ice has horizontal c-axes. Note frozen lead traversing the lake

Fig. 2.

Close-up of the surface of Peters Lake showing frequent orientation changes

Fig. 3.

The surface of Peters Lake showing patches of c-axis-vertical ice in c-axis-horizontal ice. Note the track of a taxiing DC–3 airplane; the wheels dug grooves about 20 cm. deep in candled, c-axis-horizontal ice but they made no impression on the c-axis-vertical ice

Patterns of orientations suggested wind action: breaking of an initial ice skim and refreezing between the “pancakes” or in tears in an initial skim.

Although about half the surface ice on both Peters and Shrader Lakes was orientated with the c-axis vertical and half horizontal, all of the ice more than 60 cm. beneath the surface in Peters Lake and below 15cm. in Shrader Lake was c-axis horizontal. This was determined from 10 cores in Peters and 5 in Shrader Lake, taken through surface ice of the various orientations and surroundings. Evidently, phenomena occurring during growth of ice at the bottom of an ice sheet can be more important to the orientation of the bulk of the ice than the initial ice formation.

In the consideration of ice growth at the bottom of a sheet of ice, there is again a convenient two-fold sub-division of the subject. First, as grain size increases with depth in the ice, some grains are wedged out, and this process is an orientation mechanism in so far as it is selective; and secondly, there may be the introduction of new orientations as conditions of freezing change. The second consideration may be ruled out of the discussion, because the ease of supercooling of water makes discrete re-nucleation in the process of freezing extremely unlikely, except in conditions of most rapid freezing, which cannot occur in nature at the bottom of an appreciable sheet of ice. No unequivocal cases of nucleation of new crystals during freezing of the ice were found, except in the upper 6 in.(15.2 cm.).

As a third possibility, it is not inconceivable that a crystal may change its orientation as it grows, with the introduction of certain types of dislocations, but this effect was specifically looked for and not found. Therefore, the process of wedging out and its results are the subject of the remaining part of this paper dealing with orientations.

Historical

The literature on ice orientations is quite extensive, surprisingly so, in view of its usual inconclusiveness. The work up to 1938 is completely reviewed by Dorsey (Reference Dorsey1940, p. 407–12) and is not further referred to here.

The best attempt at a comprehensive theory of ice orientation to date is that of Reference ShumskiyShumskiy (1955). (Reference Perey and PounderPerey and Pounder (1958) discuss ice orientations in a somewhat similar fashion and list several additional references.) Shumskiy’s theory is also restricted to the wedging out process and considers the fast growth direction of ice in relation to temperature gradients in the water. If supercooling decreases away from the freezing surface, the ice crystals whose fast growth directions (a-axes) are parallel to the freezing surface are thought to wedge out ice crystals of other orientations, and a preferred orientation develops with c parallel to the heat flow direction (perpendicular to the ice/water interface). Similarly, as freezing progresses, impurities concentrate in the water near the interface and lower its freezing point, causing supercooling to increase away from the interface, and crystals with c parallel to (a perpendicular to) the interface are favored. (Shumskiy does not explicitly mention impurities but he deals directly with temperature or supercooling gradients.) No measurements of the concentration of impurities at the ice/water interface have been reported, and consequently the theory has no direct confirmation, but the general orientation trends are in agreement with its predictions. The crystals in sea ice are generally orientated with their c-axes horizontal (Reference Weeks and LeeWeeks and Lee, 1958; Reference SchwarzacherSchwarzacher, 1959). Lake and artificial fresh-water ice may have either orientation, but where there is a transition, the vertically orientated ice is normally above, as predicted (Reference Perey and PounderPerey and Pounder, 1958; Reference Lyons and StoiberLyons and Stoiber, 1959; the author’s observations).

The argument for constitutional supercooling causing c-axis-horizontal grains to wedge out others seems quite clear and reasonable. The first grain to reach the supercooled layer will grow sidewise at the expense of the others, and this grain will have an a-axis most nearly vertical. The argument for c-axis-vertical grains being favored when supercooling decreases away from the interface is not so clear at all, when one thinks of the details involved. Certainly, if it occurs, it must be very slow.

One theory of ice orientation has been based on measured or assumed anisotropy of thermal conductivity (Reference WeeksWeeks, 1958). Anisotropy of fresh-water ice is very small (Reference Landauer and PlumbLandauer and Plumb, 1956), and the fact that fresh-water ice (ice with no visible liquid inclusions) can orientate either way, seems to rule out this theory. It is intuitively much less appealing to the present author than Shumskiy’s theory, especially for thick sheets of ice in which lateral temperature gradients at the bottom must be very slight.

Melt-pond data and discussion

Frozen melt ponds on Arctic pack ice provide some striking visual confirmation of Shumskiy’s theory. The water in these ponds comes from rain, melted snow and melted old sea ice, and is consequently quite fresh, usually with a salinity less than 1‰. These ponds are only one or two feet (0.3−0.6 m.) deep and as freezing progresses the concentration of salt in the remaining water increases rapidly. A study of the ice reveals the response of the crystals to the continuously increasing impurity level of the water from which they freeze. As expected, as the water becomes more salty, the c-axis-horizontal crystals commence growing dendritically (that is, they start growing downward by means of discrete blades or platelets, extending into the liquid) and very rapidly wedge out the vertical crystals (Fig. 4b). An unexpected effect was that, as dendritic growth started, the grain boundaries became irregular in a rather periodic fashion—doubtless reflecting the platelet structure, and further verifying Shumskiy’s theory (compare Figs. 4a and b). Up to the point where the grain boundaries of the horizontal crystals became irregular, there was little consistent selection between vertical and horizontal crystals. Both, however, would tend to wedge out tilted crystals.

Fig. 4.

Horizontal sections of melt-pond ice under crossed polaroids. Both contain a single c-axis-vertical crystal surrounded by c-axis-horizontal crystals. (b) is in a zone of dendritic growth hut (a) is not. Note the crystal with striations being wedged out in (a). The c-axis-vertical crystal in (b)—the black one—is being wedged out by all its neighbors

Melt-pond ice also shows clear evidence that appreciable freezing takes place upward from the bottom of ponds. The ice which forms the bottom in most ponds, being originally sea ice, is polygonized (Reference KnightKnight, 1962). Overgrowths on these crystals are detected by their lack of polygonization. This proves the presence of constitutional supercooling during the last stage of freezing and further confirms Shumskiy’s theory. (“Constitutional supercooling” is the term used in metallurgy for the condition in which supercooling increases away from the freezing interface.) A good review of constitutional supercooling and dendritic growth in metals, and other related effects, is given by Reference RutterRutter (1957).

It is concluded that Shumskiy’s theory is correct in principle, but it is questionable whether impurities build up enough at the ice/water interface in deep, “fresh” lakes for it to apply in these more important situations.

In two ponds a-axis orientations were measured. In both cases the initial crystals were very small and “perfectly” c-axis-horizontal. In both ponds a-axis orientation was also “perfectly” horizontal, as determined by inspection of orientated etch pits formed by the method of Reference HiguchiHiguchi (1958). This parallels the findings of Reference Lyons and StoiberLyons and Stoiber (1959) on artificial ice frozen on a water surface agitated by wind. The same etching technique was tried with sea ice, to compare its a-axis orientation with that of the pond ice, but difficulties were encountered. The etchant did not work properly, presumably because of a salt-water film on the ice surface. However, overgrowths of fresh ice could be produced on sea ice, with fair crystal continuity, and orientations were determined on these overgrowths. Seventy-one crystals were “measured” by inspection of etch pits on horizontal sections. Sixty-three were judged to be most nearly a-axis vertical and eight were most nearly a-axis horizontal. In many cases considerable judgement was exercised but the trend seemed definite. In addition, the trend was checked qualitatively by measuring the orientations of crystals in several vertical thin sections, using a straining technique described elsewhere (Reference KnightKnight, 1962). There is no question of these orientations being the result of initial growth; they must be the result of wedging out, since the sea ice used in these determinations was taken near the bottom of one- or two-year-old pack ice (about 2 m. down).

It is interesting that a perfect horizontal a-axis orientation was found in one case, and an imperfect vertical a-axis orientation in the other, clearly of two different origins. Since the a-axis direction is the direction of fast growth, an imperfect vertical a-axis orientation resulting from a wedging out process is to be expected according to Shumskiy’s theory. The horizontal a-axis orientation in initially orientated ice is explained by Reference Lyons and StoiberLyons and Stoiber (1959) in terms of mechanical considerations in ice formation on a wind-blown water surface.

To the author, it seems odd that, if the mechanical effect of the wind is important, no orientation of c with respect to wind direction is found. In his experience c is entirely random in the horizontal plane. However, no better explanation can be offered.

Lake ice data and discussion

In late spring of 1960 a series of orientation profiles was measured in Ikroavik and Imikpuk (also called Fresh) Lakes, and single orientation profiles were measured in Malikpik, Peters and Shrader Lakes. The first three lakes are near Barrow, Alaska and the last two are in the Brooks Range just south of Barter Island, Alaska. In the spring of 1961 orientation profiles were repeated in Ikroavik and Fresh Lakes, and in Peters and Shrader Lakes as mentioned above. The data are presented in Figures 5–9. Conventional thin-section techniques were used throughout, the sections being spaced originally 15.0 cm. apart along the 7.6 cm. diameter cores, and closer together when needed to follow an orientation transition.

Fig. 5.

Profiles of crystal orientations; spring, 1960. H, V and T stand for horizontal, vertical and tilled c-axes. Malikpik Lake was frozen to the bottom

Fig. 6.

Orientation profiles

Fig. 7.

Orientation profiles

Fig. 8.

Orientation profiles

Fig. 9.

Orientation profiles

In 1960 the cores in Ikroavik and Fresh Lakes were taken in a rather restricted area in the center of each lake—within a radius of approximately 50 m. In 1961 the cores were purposely taken farther apart, except for two (Fresh Lake, 1961, e and f), which were taken within 1 m. of each other to check a specific feature. The sequence in the figures is not significant.

In 1960 the initial orientations in both lakes were more or less random in all the cores, as indicated by the dotted fines. In 1961 a sharply defined, bubbly layer of ice was usually found at the top, possibly representing a flooding of the original ice surface from beneath because of a weight of snow (Ragle (unpublished) calls this “snow ice”). Beneath this layer there was ordinarily a sharp transition to some preferred orientation.

The vertical, horizontal and tilted (V, H and T) orientations indicated on the figures were completely distinct in almost all cases. On the few occasions when there was doubt the arbitrary definition for T was taken as more than 15° deviation from H or V.

Grain size increased downward, and below about 60 cm. in the ice it was rare to find one grain completely enclosed in a 7.6 cm. diameter thin section, though it was also unusual to find a section composed of one single crystal. Grain-size measurements were not made. Thin sections of a larger area would be necessary for studies of this kind.

A look at the graphs of orientations (Figs. 5–9) shows that when vertical and horizontal ice occur in the same core, the vertical ice tends to be above, with some exceptions. This is compatible with Shumskiy’s theory, but some other features are not. In the first place, in only one case grain boundaries became noticeably irregular in the manner which reflected dendritic growth and accompanied rapid wedging out in melt-pond ice; this was 84 in. (2.13 m.) down in Fresh Lake (1961, e). It is true that the bottom interface, when the c-axis of the ice at the interface was horizontal, always exhibited a system of approximately 1 mm. high ridges 2 or 3 mm. apart, orientated with the basal plane; but it seems doubtful if this growth is sufficiently dendritic for Shumskiy’s mechanism to apply. Indeed, one observation of the bottom surface contradicts it. The bottom surface from one core in Fresh Lake (1961, c) intersected several horizontal crystals which for the most part surrounded one tilted crystal. These horizontal crystals had ridges which extended about 1 mm. below the relatively flat surface of the tilted crystal, yet the tilted crystal was evidently in the process of wedging out the others with exceptional rapidity. This was the only time a bottom surface containing crystals of differing orientations was encountered, with the exception of the one core from Shrader Lake taken while the ice was melting.

The tilted orientations also appear to contradict Shumskiy’s theory. There is a point in the continuum of possible supercooling situations at the ice/water interface at which there is no preference for vertical or horizontal orientations, but it is difficult to imagine any positive principle of selection for tilted orientations. However, just this selection seemed to occur in 1960. It may or may not be significant that all of these cases had initial random crystal orientation.

Whether the tilted orientations represent a real selection, or just an imperfection in the scheme of horizontal and vertical selection, is a question which cannot be answered with the data at hand. The diameter of the ice cores is too small to obtain grain-size data at a sufficient depth in the ice, or to permit one to observe inter-grain relationships with any completeness. While studying the 1960 cores, it was felt that the tilted orientations were distinct enough to imply a selection principle as a cause. Unfortunately, simpler ice was encountered in the more detailed study in 1961 and few tilted crystals were observed.

Another puzzling aspect is presented in the rough layering of orientations observed in 1960. More data are required, again only easily obtainable from larger diameter cores. This layering must be fortuitous if it is to be assumed that nucleation does not occur during the freezing process.

In 1960 several sequences of thin sections spaced 2 cm. apart and closer were made in the lake ice, through the zone where the initial random orientations gave place to a specific orientation. In the winter of 1960−61 similar series of sections were made in melt-pond ice in the zone above dendritic growth, when interesting orientation situations arose. Visible clues of factors significant toward orientation selection were looked for and various photographs and petrofabric diagrams were made. Apart from the obvious fact of an orientation selection occurring, no consistent correlations involving wedging out were observed. As far as could be determined, any grain was capable of wedging out any other, depending on local circumstances, for which no general ordering principles were discovered.

However, in a number of cases specific features of local significance were observed, and apparently several of them have not been previously described. All of these features involved grain boundaries and sub-boundaries (as they must in a monomineralic substance), and many of them are of more interest in themselves than they arc in relation to orientations. They are described separately below and their relation to orientation will be only incidentally mentioned here. The wedging out process is, however, fundamentally a grain boundary phenomenon, and any new understanding of grain boundaries is likely to elucidate it. Large-angle grain boundaries will be discussed first, followed by a discussion of sub-boundary (striation) features.