Origin of the Filaments

Water veins exist along the triple junctions in high-purity ice close to its melting point (Reference SteinemannSteinemann, 1957; Reference Ketcham and HobbsKetcham and Hobbs, 1969; Reference Nye and FrankNye and Frank, 1973). Dissolved impurities increase the diameters of the water veins and decrease their freezing temperatures (Reference MaderMader, 1992). Thus, an obvious question regarding our observations is: are the filaments frozen impurity-containing water veins that remain after the ice has sublimated away in the SEM? This seems unlikely since we know the positions of both the triple junctions and the grain boundaries from optical microscope observation of each specimen before SEM examination, and filaments have been found along grain boundaries in which there were no nearby triple junctions from which they could have originated. Further, it is very unlikely that we could section horizontally through several triple junctions simultaneously to produce images such as that shown in Figure 1 (Reference Cullen and BakerCullen and Baker, 2000).

In order to investigate this further, an ice-core specimen from 150 m depth at GISP2 was examined after it was shaved to a thickness of 0.2 mm. The average grain-size in the ice, determined using the linear intercept method, is 2.6 mm. Thus, the specimen is significantly less than one grain diameter thick, and no obvious (non-vertical) triple junctions, from which frozen water veins could be produced, were observed. Figure 3a shows a SE image of a typical1–2 mm grain (labeled ‘À”) from this specimen. A representative X-ray spectrum from the grain boundary (Fig. 3b) shows largely Na and Cl, with small amounts of K and Ca. Figure 3c shows a ribbon-like NaCl filament which “grew” out of the grain boundary during sublimation of the specimen in the SEM for 40 min at 183 K. An X-ray spectrum from the filament again showed that it contained Na and Cl. It was quite clear that this filament came from the grain boundary and that there was no triple junction from which it could have emerged.

Fig. 3 Continued overleaf.

Fig. 3 (a) SE image after sublimation for 20 min at 183 K in the SEM (ice sublimated is approximately 0.1 mm); (b) typical X-ray spectrum from the grain boundary; (c) SE image of ribbon-like NaCl filament (arrowed) from grain boundary (GB) indicated in (a) after sublimation for ∼40 min at 183 K (ice sublimated is approximately 0.15 mm).

Thus, it appears that formation of the grain-boundary filaments occurs as a result of the preferential sublimation of the ice that surrounds the impurities, which had segregated to the grain boundary prior to examination. The examination temperature in the SEM was well below the eutectic temperature of the H₂O–impurity systems for the observed impurities, and it is therefore likely that the filaments are hydratedFootnote ¹ salts, which coalesced (to reduce surface energy) after the surrounding ice sublimated away.

Thus, it appears that the filaments are artifacts. However, they indicate the presence of water–impurity eutectics at the grain boundary.

Do Impurities Exist within Grains?

If, as indicated above, a water–NaCl eutectic exists at the grain boundaries of GISP2 ice, then one concern is that specimen preparation at 253 K could have spread the impurities over the grains since the eutectic temperature is ∼250 K, i.e. there would have been a liquid film at the grain boundaries (personal communication from A. Rempel, 2001). To examine whether specimen preparation introduced impurities into the grains, several experiments, as outlined below, were performed.

1. 1. A specimen was shaved in one direction only, with the blade being cleaned after each pass. If the impurities were spread into the grain from the grain boundary, then a higher concentration of impurities would be observed in the grain shaved after the grain boundary. Experimentally, it was found that there was no evidence of more impurities in the grain on one side of the grain boundary than on the other.

2. 2. All the grain boundary regions were cut off a single large grain before the surface was shaved and examined in the SEM. Impurity spots were still observed in the grain interiors (see X-ray spectrum in Fig. 4), even though no grain boundary films were present.

3. 3. A specimen was shaved under liquid nitrogen, when all water–impurity eutectics would be frozen and immediately placed in the SEM. Upon SEM examination at 173 K, impurity spots were found straightaway on the ice. Although the spots could immediately be seen, they had to “grow” before the EDS system could detect the impurities present, hence the 105 min sublimation indicated in Figure 5. It is worth noting that specimen preparation was not routinely performed under liquid nitrogen since the specimen is much harder at this temperature and hence is much more difficult to shave.

4. 4. A specimen was cleaved under liquid nitrogen and, after sublimation in the SEM, impurities could still clearly be observed on the cleaved surface within the grains (see Fig. 6).

5. 5. Ion chromatography was performed on melt from pairs of large-grained specimens from Byrd Station and GISP2. In each pair, one specimen had the grain boundary region removed and the other was analyzed with the grain boundaries intact. The results (Table 1) showed that while the specimens which contained grain boundaries generally had slightly higher impurity concentrations than the grain interiors, indicating that there was some concentration of impurities in the grain boundary region, the grain interiors contained comparable impurity levels.

It is also worth noting that for Byrd Station ice, where EDS identified mostly Mg and S at the grain boundaries, the specimen preparation temperature was below the MgSO₄–water eutectic temperature of ∼270 K. Hence, no liquid film would be present to spread Mg and S over the grain interiors.

Fig. 4 X-ray spectra from impurities in an ice specimen from 1504 m Byrd Station ice after 15 min at 183 K. The grain boundaries were removed prior to specimen preparation to preclude any possible boundaries.

Fig. 5 EDS spectrum from an impurity spot in a 1102 m Byrd Station ice specimen shaved under liquid nitrogen. Sublimation time is 105 min at 173 K (ice sublimated is approximately 0.4 mm).

Fig. 6 (a) SE image of 1102 m Byrd Station ice cleaved under liquid nitrogen, after 190 min at 173 K in the SEM; (b) X-ray spectrum from the point indicated within the grain (ice sublimated is approximately 0.7 mm).

We also addressed the concern that the impurities arose in the grain interior by diffusion from the grain boundary during examination (Reference Rempel, Waddington, Wettlaufer and WorsterRempel and others, 2001) through the following experiments and observations:

1. (i) In (3) above, we noted that impurity spots could immediately be observed on the specimen surface after shaving at liquid-nitrogen temperatures and examination in the SEM at 158 K (again, the impurity spots had to “grow” before the impurities could be detected), suggesting that no diffusion of impurities was necessary to form them.

2. (ii) We examined the possibility that vapor transport could produce the impurities in the grain interior, by placing a specimen containing ∼1 ppm H₂SO₄ next to a high-purity ice specimen in the SEM, and examined whether impurity spots formed on the surface of the undoped ice. They did not.

3. (iii) In specimens cleaved under liquid nitrogen and examined immediately in the SEM, impurity spots were immediately present, although, as noted above, they had to “grow” before the EDS system could detect them, indicating that the impurities were present in the lattice.

4. (iv) Many elements were detected in impurity spots in grain interiors, while the grain boundaries in the GISP2 ice contained largely Na and Cl and those in Byrd Station ice contained largely Mg and S. Thus, the impurities within the grains could not all come from the grain boundaries.

Therefore, our observations indicate that the impurity spots in the grain interiors arise from impurities within the grains and are not artifacts. It is not wholly clear whether these impurities are present as precipitates or solutes (or a combination of both). However, it is worth noting that the longer a specimen was allowed to sublimate, the larger the white spots became, and the more likely the spots were to contain detectable impurities (since larger spots provide a larger interaction volume for X-ray production, which occurs from a region of the order of 1 μm in diameter). Thus, one possibility is that some impurities were present as precipitates and other impurities coalesced on them, moving there by localized surface diffusion.