Introduction

Ice cores provide a valuable and productive means of studying past changes in climatic conditions. Concentrations of gases such as CO₂, CH₄ and CO and isotopes of oxygen have been measured in ice cores from Greenland and Antarctica in order to deduce information about past atmospheric chemistry and its connection with major climatic shifts such as ice ages (Reference Barnola, Raynaud, Korotkevich and LoriusBarnola and others, 1987, Reference Barnola, Anklin, Porcheron, Raynaud, Schwander and Stauffer1995; Reference Chappellaz, Barnola, Raynaud, Korotkevich and LoriusChap-pellaz and others, 1990; Reference Etheridge, Steele, Langenfelds, Francey, Barnola and MorganEtheridge and others, 1996). Concentrations of dissolved species also provide valuable indicators for past atmospheric chemistry. Among the numerous species detected in polar ice, H₂O₂ and HCHO, which are oxidation products of CO and CH₄, are important constituents (Reference Sigg and NeftelSigg and Neftel, 1991; Reference Staffelbach, Neftel, Stauffer and JacobStaffelbach and others, 1991; Reference Fuhrer, Neftel, Anklin and MaggiFuhrer and others, 1993; Reference Jacob and KlockowJacob and Klockow, 1993; Reference Neftel, Bales, Jacob and DelmasNeftel and others, 1995; Reference Van Ommen and MorganVan Ommen and Morgan, 1996). Since CH₄ and CO in the gas phase are oxidised almost exclusively by the OH radical, knowledge of the levels of their oxidation products in ice cores may permit assessment of past changes in OH concentration and consequently in the oxidising ability of the atmosphere (Reference Thompson and DelmasThompson, 1995). However, deconvolution of ice-core records of species such as hydrogen peroxide and formaldehdye in order to study past changes in atmospheric chemistry is complicated by other factors, such as post-depositional processes, that may modify the atmospheric signal. Thus, in addition to high-quality ice-core data, a fundamental understanding of air–ice transfer processes and post-depositional change is required in order to infer past atmospheric chemical conditions.

The aim of this paper is to address the first of these requirements, which is the establishment of a high-quality ice-core data record, in our case from Law Dome, East Antarctica. The second requirement, interpretation of the data record in terms of past atmospheric processes, will be addressed in due course after additional work on air–ice exchange at Law Dome, and validation of an atmospheric photochemical model for this site.

This work reports measurements of H₂O₂, methyl hydroperoxide (CH₃OOH) and HCHO made in ice-core samples from Law Dome, Antarctica. Although CH₃OOH has not been detected in studies elsewhere, measurements were made on Law Dome samples because the detection technique used was more sensitive than before, and the organic peroxide is expected to be considerably more important in Antarctic than in Arctic air (due to a much higher atmospheric CH₄/CO ratio in the Southern than in the Northern Hemisphere).

Measurements were made in samples from the modern and pre-industrial periods and the Last Glacial Maximum (LGM). The HCHO measurements reported here are the first such analyses for any Law Dome core, and the CH₃OOH analysis constitutes the most sensitive search for this species in polar ice/snow yet reported in the literature. The new ice-core data presented here were obtained from the Dome Summit South (DSS) ice core, drilled on Law Dome. This core provides a very high-resolution record which extends back beyond the last glacial and has clearly preserved seasonal cycles in hydrogen peroxide (Reference Van Ommen and MorganVan Ommen and Morgan, 1996) that have been detected at ages of several thousand years. The site has a high accumulation rate (0.7 m ice a⁻¹) and a relatively low mean wind speed (Morgan and others, 1997) for East Antarctica. These properties lead to rapid burial and minimisation of wind-induced ventilation of the near-surface snow. Such conditions are thought favourable (Reference Waddington, Cunningham, Harder, Wolff and BalesWaddington and others, 1996) for preservation in the ice of such reversibly deposited species as H₂O₂ and HCHO at, or close to, the concentrations at the time of deposition. The DSS material used for the pre-industrial and LGM periods was augmented, for the modern period, with material from a fresh shallow core (denoted as DSS96P) retrieved from the site in February 1996. To minimise the risk of drilling in a previously disturbed site, DSS96P was drilled 840 m from the DSS borehole at a bearing of 208°, which is approximately on the same accumulation isopleth.

Experimental Methods

The ice was sampled at four deposition periods corresponding to ages (before present: 1996.0) of 0–15, 330–450, 6900 and 13 700 years. The core dating was achieved by direct counting of annual layers down to AD 1300, below which an ice-flow model constrained by spot layer-thickness measurements was used (Reference Van Ommen and MorganVan Ommen and Morgan, 1996; Morgan and others, 1997). We believe that this model (which assumes a simple flow law) generates ages which are systematically too young beyond about several thousand years, probably as a result of unmodelled stagnant ice near the base of the sheet. The LGM at this site is defined as the coldest stage of the last glaciation as estimated from isotope measurements. The samples used in this study are dated at around 13 700 BP, when the temperature was estimated to be about 15.9°C colder than now (Reference Van Ommen and MorganVan Ommen and Morgan, 1997).

For all epochs, ice was cut in discrete samples that were melted immediately prior to analysis. The sampling in the modern and pre-industrial sections was arranged to provide a resolution of 0.1 year, while the samples from older sections represented averages over many years to decades. As noted in the introduction, the core material used for the modern epoch was retrieved in February 1996, just 1 month before analysis. This rapid analysis was undertaken to minimise concerns about possible post-drilling changes in peroxide levels. Two parallel sets of samples were prepared from identical depths to provide independent samples for the subsequent analyses. One of the sets was analyzed for total peroxides. The other was divided, upon melting, into two aliquots. One aliquot was analyzed for HCHO and the other was used for the measurement of individual peroxides.

Total peroxide measurements were made with a commercial diffraction-grating fluorimeter (GBC 1250 Fluoro detector) using the enzymatic fluorimetric analysis method (Reference Lazrus, Kok, Gitlin and LindLazrus and others, 1985; Reference Sigg and NeftelSigg and Neftel, 1988) as discussed in an earlier paper (Reference Van Ommen and MorganVan Ommen and Morgan, 1996). The detection limit was approximately 0.015 μmol L⁻¹, with a 1σ deviation for single measurements in the range 0.030–0.06 μmol L⁻¹, based on response of calibration standards.

Formaldehyde concentrations were measured using the fast-flow injection analysis of Reference Dong and DasguptaDong and Dasgupta (1987). This technique relies upon the Hantzsch reaction which produces a derivative, 3, 5-diacetyl-1, 4-dihydrotolutidine, from the cyclisation reaction of formaldehyde with NH₃ (ammonium acetate) and 2, 4-pentanedione. Immediately after the ice samples were melted, an aliquot was injected into the system via a six-port Rheodyne 7125 injector with a loop having a volume of approximately 800 μL. High-purity water, used as the mobile phase, was pumped through the system with a peristaltic pump so that the ice-core sample was injected into the water stream; the 2, 4-pentanedione/ammonium acetate reagent was added to the liquid line downstream of the injector. The liquid was heated to 95°C in an in-line heating block to ensure the reaction was complete, and then passed through a small length of Gore-Tex TA-001 hydrophobic membrane tubing to remove air bubbles introduced by the heating. The concentration of the derivative in the liquid was measured with a fluorescence detector (GBC LC 1250 Fluoro Detector) at an excitation wavelength of 412 nm and an emission wavelength of 510 nm. Concentrations of HCHO were calculated using standard formaldehyde solutions which were serially diluted daily from a standardised formaldehyde solution. Peak heights of the standards were found to be linear over a concentration range of < 0.03 to >6 μmol L⁻¹.

Samples were analyzed for individual peroxides using high-pressure liquid chromatography (HPLC), allowing the separation and quantification of H₂O₂ and individual organic hydroperoxides. Detection was based on the horseradish peroxidase (Sigma-Aldrich, type II) enzyme-catalysed di-merisation of p-hydroxyphenylacetic acid (Reference Lazrus, Kok, Lind, Gitlin, Heikes and ShetterLazrus and others, 1986).

Samples were injected via a Rheodyne 9125 metal free injector valve onto a 100 μL sample loop from a sample volume of 250 μL. To reduce particulate contamination, samples were passed through 0.2 μm ion-chromatography syringe filters (Anachem) prior to injection. Separation was achieved using a 5 μm Alltech Adsorbosphere (4.6 × 250 mm) MF-Plus C18 column, with an eluent of 1 mmol L⁻¹ H₂SO₄ and100 μmol L⁻¹ ethylenediaminetetra-acetic acid in Milli-Q water. Eluent was delivered by a Merck–Hitachi L-6200 Intelligent pump at a flow rate of 0.6 mL min⁻¹.

Separated compounds were then derivatised by the addition of 26 mmol L⁻¹ p-hydroxyphenylacetic acid with 10⁵ units L⁻¹ of horseradish peroxidase, in potassium hydrogen phthalate buffer, at pH 5.8 delivered by a Gilson peristaltic pump (0.25 mm inside-diameter tubing). After reaction in a Teflon mixing coil, derivatised peroxides were passed through a Nafion membrane reactor (Omnifit, England), immersed in 150 mL of 30% NH₄OH which raised the pH above 10, to convert the dimer to an anionic form, maximising fluorescence for detection. A Merck–Hitachi detector was used, at excitation and emission wavelengths of 310 and 405 nm, respectively.

Peroxide peak identities were confirmed by comparison with retention times of authentic standards. Commercially available 30% H₂O₂ and 36–40 wt% peroxyacetic acid (Sigma-Aldrich) were used, whilst methyl hydroperoxide ethyl hydroperoxide, hydroxymethyl hydroperoxide, 2-hydroxyethyl hydroperoxide and 2-hydroxypropyl hydroperoxide were synthesised using modified methods of Reference Reiche and MeisterReiche and Meister (1933, Reference Reiche and Meister1935) and Reference Davies and DearyDavies and Deary (1992).

All hydroperoxides react to form the same fluorescent dimer, producing an equivalent molar response. Therefore, HPLC calibration was performed using H₂O₂ only, with calibration standards made daily by the serial dilution of commercially available 30% H₂O₂ (Sigma-Aldrich) with Milli-Q water. Calibration was found to be linear over the range 1 nmol L⁻¹ to 10 μmol L⁻¹.

The limit of detection and limit of quantification for the HPLC have been taken as three times and ten times the standard deviation at zero analyte, respectively, at 95% confidence, as defined by Reference Heavner, Ogden and NelsonHeavner and others (1992). This leads to calculated values of 3 and 11 nmol L⁻¹, respectively. Precision was found to be 1% for repeat injections of a 1.7 μmol L⁻¹ H₂O₂ standard solution.

Results and Discussion

The individual peroxides analysis failed to show any CH₃OOH or other organic peroxides in the ice from any of the epochs, even in the relatively fresh (1 month old) firn samples. This confirms at Law Dome the work of others (Reference Sigg and NeftelSigg and Neftel, 1988, Reference Sigg and Neftel1991; Reference Jacob and KlockowJacob and Klockow, 1993) who have reported an absence of organic peroxides in ice cores from Greenland and Antarctica. The notable difference is that our HPLC method places a new upper limit of approximately 0.003 μmol L⁻¹ on the ice-core CH₃OOH level, compared with prior results which reported lower sensitivities of ∼0.015 μmol L⁻¹.

The parallel measurements of H₂O₂ with the HPLC system and of “total peroxides” with the standard flow-injection system showed extraordinary agreement (Fig. 1), confirming that “total peroxides” equals H₂O₂ and that the ice is essentially devoid of organic peroxides. The remarkable agreement between the independent analytical methods is evident from the mean difference from 120 measurements of only 15 ± 9 nmol L⁻¹ on an average absolute level of > 1.5 μmol L⁻¹ _. The corresponding 1σ variation in measurement difference between single measurements using the two techniques is 110 nmol L⁻¹. Although this is larger than the error quoted in the previous section for calibration standard measurements, it includes a contribution due to the variability between separate samples prepared in parallel. Minor differences in sample handling, cut geometry and local porosity variations in the firn can all be expected to lead to small discrepancies in parallel samples from the same piece of core.

Fig. 1. Comparison of total peroxides and hydrogen peroxide concentrations in ice cores, 1980–85.

The results of the analysis are summarised in Tables 1 and 2. The modern epoch consists of 13 separate years in three separate periods, “modern-a” (1993.0–1996.0), “modern-b” (1986.0–1992.0) and “modern-c” (1981.5–1985.5). The modern-b data come from earlier analysis for H₂O₂ only (Reference Van Ommen and MorganVan Ommen and Morgan, 1996) and are presented here to provide improved statistics for this period. In similar fashion, the pre-industrial HCHO and individual ROOH measurements (pre-industrial-b) have been augmented with a mean (pre-industrial-a) from a longer series of earlier H₂O₂ measurements (Reference Van Ommen and MorganVan Ommen and Morgan, 1996). The data are plotted in Figure 2.

Fig. 2. Mean H₂O₂ and HCHO concentrations with age. Error bars are one standard deviation amplitude, computed from estimated interannual variability as described in the text.

The data were reduced to averages over a whole number of years to avoid biases (which would arise for peroxide, at least, as a result of the strong seasonality). Both H₂O₂ and HCHO series display large interannual variability, no doubt due to the stochastic nature of precipitation events and the effect of this on snow retention, as well as fluctuations in tropospheric gas-phase concentrations. In the case of H₂O₂ these interannual differences are known to show a 1σ variability (Reference Van Ommen and MorganVan Ommen and Morgan, 1996) on the order of 30% for annual averages (computed from peroxide analysis results from the DSS core for a 262 year period, AD 1580–1842). This figure of 30% variability (scaled by the square root of the number of years averaged) was used to estimate the standard error in the modern and pre-industrial mean values plotted in Figure 2. Standard errors quoted for the early Holocene and LGM measurements are calculated directly as for the N measurements in each period. For the HCHO means, these are the first results available on the DSS core, and the best estimate we have for the variability comes from the annual values measured here. We have taken 20% as a representative figure for the 1σ annual variation in HCHO, and computed standard errors for the HCHO means as for the H₂O₂ means. This lower interannual variability for HCHO than for H₂O₂ is consistent with the much higher diffusion smoothing exhibited by HCHO.

The peroxide measurements agree well with earlier measurements detailed in Reference Van Ommen and MorganVan Ommen and Morgan (1996), although the modern series presented here was analyzed under more controlled conditions than the previously published modern data and is of higher quality. Of particular note is the fact that the summer peroxide peak of 1982, which was absent in the earlier DSS87P core (see discussion in Reference Van Ommen and MorganVan Ommen and Morgan, 1996), is clearly present in these data from the new DSS96P core.

The data from this DSS96P core show increases in H₂O₂ and HCHO over the modern period. This increase is not statistically significant in the case of H₂O₂, but the increase in HCHO is unambiguously significant (>5σ). Some of the change during the modern period is likely to be related to the effects of progressive burial in the upper layers. Processes of ventilation and diffusion plus changes in temperature, irradiance and porosity affect the final concentrations. We assume here that these effects are largely confined to the upper firn and do not make a significant contribution to the changes seen between the modern-c period (at depths of 14–19 m) and the pre-industrial. Clearly, this assumption requires confirmation, which may come from future study of other shallow cores from different accumulation regions on Law Dome.

Examination of the H₂O₂ and HCHO concentrations over the modern period (Fig. 3) reveals some interesting features. Seasonality is clearly shown in the H₂O₂ concentrations. The HCHO concentrations are more erratic, but do show some suggestion of systematic behaviour in addition to the change in mean levels noted earlier. From the surface, concentrations rise markedly to a large peak in the previous autumn/winter layer. The correspondence may be fortuitous, but there is a suggestion of weaker peaks in the previous two autumn layers, and even possibly in the 1980s series. The noisy nature of the HCHO concentrations is interesting in light of the very uniform peroxide cycles, which would suggest minimal snow redistribution and reasonably regular precipitation. It would seem that the processes responsible for largely removing seasonality in HCHO concentrations cannot simply be diffusional, or else these erratic fluctuations would also be smoothed away (occasional spikes up to several times mean concentrations are also seen in the sub-annual resolution measurements for the pre-industrial period). Reference Neftel, Bales, Jacob and DelmasNeftel and others (1995) discuss HCHO levels seen in pit studies in Greenland. They report a pattern where HCHO concentration shows a minimum at about 10 cm depth which they attribute to photolysis. A similar argument may explain the results seen here. The lower HCHO levels at the surface, it could be reasoned, are caused by photolytic destruction, at least at the snow-grain boundaries. The rise in concentration seen at around 0.6–1.0 m may then be attributed to uptake of HCHO from the atmosphere by forced ventilation, photolysis being no longer important at these depths. If so, the enhanced HCHO levels seen in autumn are the result of the competing effects of photolysis and ventilation and the depth scales over which they operate. If this argument is correct, the consequent pattern should be different at a site with different accumulation.

Fig. 3. Concentration of hydrogen peroxide (upper trace) and formaldehyde (lower trace) 1980–85 and 1993–96.

The changes seen in both species from modern to pre-industrial, early Holocene and LGM periods are all statistically significant if the additional H₂O₂ (“pre-industrial-a”) measurement is used for the pre-industrial period. Most of the decrease in peroxide levels seen in the early Holocene and LGM periods is probably the result of destruction in ice and not of changes in atmospheric levels. Reference Van Ommen and MorganVan Ommen and Morgan (1996) showed an overall loss of approximately 0.05 μmol L⁻¹ per 1000 years in the DSS peroxide record.

The elevated HCHO value in the early Holocene is not understood. Data from Summit (Greenland) ice cores (Reference Fuhrer, Neftel, Anklin and MaggiFuhrer and others, 1993) show elevated HCHO in the last glacial (as is also seen here) but not during the early Holocene. HCHO measurements for Dye 3 and Byrd cores (Reference Staffelbach, Neftel, Stauffer and JacobStaffelbach and others, 1991) also show low levels in the early Holocene but which extend into the last glacial period without the higher levels seen in Summit or DSS data. Since, in this study, early Holocene results showed elevated HCHO concentrations, it was judged prudent to check the concentration in those cores. HCHO concentrations were measured again in samples freshly cut from the same DSS core and an adjacent DSS core. The results, given in Table 2, show concentrations very similar to the original ones even though more than 2 years elapsed between the two sets of measurements.

Previous studies in Antarctica have found H₂O₂ concentrations in ice cores of about 0.3–2.1 μmol L⁻¹ during the pre-industrial period (Reference Neftel, Bales, Jacob and DelmasNeftel and others, 1995). The pre-industrial H₂O₂ concentration of 1–1.2 μmol L⁻¹ measured in the present study places our results in the middle of the reported range. In contrast, H₂O₂ concentrations in Greenland ice cores are 1.2–5.9 μmol L⁻¹ (Reference Neftel, Bales, Jacob and DelmasNeftel and others, 1995) which appears to be significantly higher than in Antarctica. This may be due to the differences in the modern CH₄ : CO ratio between the Northern and Southern Hemispheres. Although CH₄ concentration globally shows only a small interhemispheric difference, this is not the case for CO. Reference Martinerie, Brasseur and GranierMartinerie and others (1995) provide a convenient picture of a strong gradient in CO concentrations decreasing from the Northern to the Southern Hemisphere, with the modern CH₄ : CO ratio at about 25 in Antarctica and 15 in Greenland. Since H₂O₂ is a product of CO oxidation by hydroxyl radical and CH₃OOH from the analogous oxidation of CH₄, it appears that the increased relative importance of CO compared with CH₄ in the Northern Hemisphere may explain the higher H₂O₂ concentrations in Greenland ice cores, compared with those measured in Antarctica.

Conclusions

Concentrations of total peroxides, speciated peroxides and formaldehyde were measured in a high-accumulation ice core from Law Dome, covering the modern, pre-industrial, early Holocene and LGM epochs. Two techniques were used to measure peroxides. In the first, total peroxide concentration was measured, and in the second, peroxides were measured individually by speciation on a HPLC column. No organic peroxides were detected using the HPLC technique, and total peroxide and hydrogen peroxide concentrations were in close agreement in all samples measured. From this we conclude that virtually all the peroxide in the ice was found to exist in the form of H₂O₂ at a level of ∼1.5 μmol L⁻¹ in modern ice. Based on analytical detection limits, the concentration of CH₃OOH can be set at < 0.003 μmol L⁻¹.

H₂O₂ concentrations were 1.6–2.1 μmol L⁻¹ in modern ice with mean ages ranging from 1.5 to 12.5 years before 1996.0. In the pre-industrial period, H₂O₂ concentrations were significantly lower, with values of 1.0–1.2 μmol L⁻¹ in ice having mean ages of 348 and 455 years before 1996.0. During the early Holocene, H₂O₂ concentrations were 0.6 μmol L⁻¹ in ice with an average age of 6900 years before 1996.0; in the LGM, concentrations decreased to 0.2 μmol L⁻¹.

HCHO concentrations were measured in ice in the modern, pre-industrial, early Holocene and LGM epochs. During the modern period, concentrations were 0.22–0.10 μmol L⁻¹ for ice with mean ages of 1.5 and 12.5 years before 1996.0. Formaldehyde concentrations dropped to a mean of 0.07 μmol L⁻¹ during the pre-industrial period, when ice had a mean age of 455 years before 1996.0. In the early Holocene, formaldehyde concentrations increased to 0. 13 μmol L⁻¹. Since previous studies have not measured such high concentrations in the early Holocene, measurements were repeated. More than 1 year after the original measurement, formaldehyde concentrations were determined in fresh cores cut from the same and an adjacent DSS core. The results of these analyses were very similar to the original measurements. The mean HCHO concentration measured in the LGM was 0.14 μmol L⁻¹ and was more consistent with previous measurements.