Sir,

The spatial variation of impurities found in natural polycrystalline ice has been of interest for over a century. Based on thermodynamic arguments, Reference Harrison and RaymondHarrison and Raymond (1976) suggested that impurities in temperate glacier ice reside to varying degrees in three-grain intersections or triple junctions (TJs). Thus far, however, there have been only two studies of the microchemistry of natural ice. In the first study, Wolff and co-workers (Reference Mulvaney, Wolff and OatesMulvaney and others, 1988; Reference Wolff, Mulvaney and OatesWolff and others, 1988) showed, using X-ray microanalysis in a scanning electron microscope (SEM), that the TJs contained substantial concentrations of sulfate ions. The ice, obtained from Dolleman Island, Antarctica, was coated with aluminum and held at a temperature of −160°C during analysis. Neither S nor Cl were detected elsewhere in the ice. Later, Reference Fukazawa, Sugiyama, Mae, Narita and HondohFukazawa and others (1998) used micro-Raman spectroscopy to study ice from two Antarctic sites. At temperatures between −8° and −35°C in Nansen ice, NO₃ ⁻ and HSO₄ ⁻ were found as liquids at the TJs, while at temperatures between −8° and −20°C in South Yamato ice, SO₄ ²⁻ was found as a liquid at the TJs.

The ice studied here was from an ice core (specimen depth 214 m) at the Greenland Ice Sheet Project II (GISP2) site. It was stored at −20°C at the National Ice Core Laboratory, Boulder, Colorado, U.S.A., before examination at Dartmouth. Specimens with dimensions of ∼25 mm × 25 mm by 10 mm thick were cut from the ice, and the surface carefully shaved with a razor blade in a high-efficiency particle-air-filtered, laminar-flow hood following standard clean-room practices. The specimens were examined using a JEOL 5310LV low-vacuum SEM, equipped with a PGT IMIX energy-dispersive X-ray microanalysis system (EDS) utilizing a pure germanium, thin-window detector. The SEM was operated at 10 kV and utilized a cold stage cooled to −115 ± 5°C. Four SEM images, and associated EDS data (normalized to 100 000 total counts), are shown here from uncoated GISP2 ice. The elements indicated on the X-ray spectra are the impurities typically found in glacier ice (Reference MayewskiMayewski and others, 1993). Nitrogen, which is present in both NO₃ ⁻ and NH₄ ⁺, is not indicated since it was not found in any specimen. EDS data were collected from a number of different regions in the ice: TJs, grain boundaries (GBs), grain interiors, “filaments” and inclusions were all analyzed.

Figure 1 shows the ice after it was allowed to sublimate in a small sealed container for 8 weeks at −20°C. The large GB channels (see inset) are typical of the long sublimation time. Sublimation preferentially occurs at the GBs because these are regions of high energy. Sublimation also led to grain faceting, a feature previously noted by Reference CrossCross (1969). At the upper left is a ∼1.0 μm diameter filament in the center of a GB channel (labeled 1). EDS data from the filament (Fig. 2a) showed that it is composed chiefly of Cl, Mg and Na, although small S (from SO₄ ²⁻) and Ca peaks are also present. Filaments were observed along many GBs, and all contained large amounts of Cl and Na, while the occurrence of the other common glacier impurities appears to be happenstance. Note that in Figure 1 filaments were absent along the other two GBs, although a smaller filament fragment is present in the horizontal GB (labeled A). Filaments appear to be attached relatively loosely to the ice and move around readily, particularly when the electron beam is focused on them or when the specimen is undergoing rapid cool-down. EDS of the small white spot located within the TJ center (labeled 2) showed predominantly Cl and Na (Cl is less than in the filament), with a minute S peak, while EDS of the white spot in the grain interior (labeled 3) showed low impurity levels, with minute S, Cl and Na peaks, typical of spectra from other white points within grains.

Fig. 1. Three GBs meet at a TJ (2) in ice which has been allowed to sublimate for 8 weeks. Note the filament lying along the upper left GB (1) and the smaller filament at the right of the horizontal GB (A). 3 is a white spot in the grain interior. The inset is a reduced version of the figure in which the lines indicate the GB channels.

To study the effect of a short sublimation time, the specimen in Figure 1 was reshaved with a razor and then allowed to sublimate in a sealed container at −20°C for 2 hours. The GB channels (Fig. 3) are not present, indicating minimal sublimation, but the GBs are easily distinguished by the small white impurity spots and the different sublimation patterns of the grains (which depend on their crystallographic orientation). The inset in Figure 3 is a higher magnification image of an inclusion and a filament from the GB adjacent to the lower TJ. EDS data (Fig. 2b) showed that the inclusion (labeled 4) consisted of S, Cl, Na, Ca and a small amount of Mg, while the filament (labeled 5) and the white spot on the GB labeled 6 contained only Cl and Na.

Fig. 2. (a) EDS spectra from the points indicated in Figure 1: (1) GB filament; (2) TJ white spot; (3) white spot in grain interior. (b) EDS spectra from the points indicated in Figure 3: (4) spherical inclusion near TJ; (5) filament adjacent to inclusion; (6) GB white spot. The ordinate axis indicates X-ray counts, and the abscissa indicates the energy; the relative concentration of each element is given by the area under the peak. The peak on the far left is oxygen from the ice and/or hydrated impurities; the small unlabeled peaks some times present to the right of the S or C lpeaks are the K_β peaks for these elements.

Fig. 3. Two TJs and connecting a GB in ice. Inset: Spherical inclusion and filament from GB area near the lower TJ. The specimen had been allowed to sublimate for 2 hours.

The formation of the GB filaments is thought to occur as a result of preferential sublimation of pure ice, which surrounds the impurities that have segregated to the GB. The examination temperature was well below the eutectic temperature of any of the H₂O-impurity systems for the observed impurities, so it is possible that the filaments are hydrated salts which coalesced (to reduce surface energy) after the surrounding ice sublimated. An estimate can be made of the expected filament diameter by assuming that a sheet of impurities 1 nm thick resided at the GB, and that the depth of the GB channel in Figure 1 is half its width of 250 μm. The resulting cylindrical filament would have a diameter of 0.4 μm that is comparable to the observed ∼1 μm filament width in Figure 1. Similarly, coalescence of impurities, via surface diffusion, could explain the formation of the white impurity spots in the grain interiors. It is worth noting that the longer a specimen sublimates, the more numerous and larger become the white spots on both the grain interiors and the GBs, and the more likely the spots are to contain detectable levels of impurities. No detectable impurities were found in the dark areas of any specimen.

Any doubt that the filaments observed in the previous images originated from the GB should be alleviated by Figure 4 in which the association between the filaments and the GB region is clearly illustrated along with the transient nature of the filaments. The thinner filament sections are peeling away from the GB channels, the impetus for this movement probably being the rapid cool-down of the specimen and the large difference in the thermal expansion coefficient of NaCl (∼28°C⁻¹ at −120°C) and ice (∼90°C⁻¹ at −120°C).

Fig. 4. Image of ice showing GB filaments. Note that some of the filaments have peeled out of the GB.

Wolff and co-workers stated that between 40% and quite possibly all of the H₂SO₄ in the Dolleman Island ice they studied was at the TJs. They suggested that if this result could be generalized to all Antarctic and polar ice it could explain the d.c. conductivity of the ice. Our results, showing high concentrations of S in inclusions (Fig. 2b), trace amounts in the grain interiors, but only small amounts at the TJs, suggest that all the H₂SO₄ is not located in TJs in polar ice. This observation was indirectly confirmed by Reference Fukazawa, Sugiyama, Mae, Narita and HondohFukazawa and others (1998), who estimated that <3% of the H₂SO₄ in their South Yamato ice was at the TJs. On the other hand, Wolff and co-workers proposed (but did not observe, since they did not find any Cl) that it is thermodynamically preferable for NaCl in polar ice to form “pure volumes” of NaCl which would eventually partition to the GB. Our data (small Cl concentrations in the grain interiors, but large concentrations in the form of filaments in the GB) would seem to confirm this suggestion. It should be noted that the GISP2 site is located in the opposite hemisphere, is colder and is farther from the ocean, and that the specimens studied here were obtained from different depths than the sites from which Wolff and co-workers or Reference Fukazawa, Sugiyama, Mae, Narita and HondohFukazawa and others (1998) obtained their ice. Therefore, a direct comparison between their data and ours is not possible.

In summary, filaments composed mostly of NaCl were observed in GBs of ice from GISP2. The filaments appear to be the result of preferential sublimation of ice at the GBs leaving behind the impurities. In addition, small amounts of impurities (mainly S and Cl⁻) were found in the grain interiors, while large concentrations of S were observed in inclusions.