INTRODUCTION

More than 100 years ago, Reference SchôneSchöne (1874) found that H₂O₂ is present in hoar frost. Nevertheless, only during the past few years has it been appreciated that H₂O₂ is one of the most important species in atmospheric chemistry. Because of its intermediate position between extreme reactive species such as OH or Ho₂radicals and the long-living species, it is a useful tool for testing atmospheric–chemistry model calculations. Reliable measurements of the atmospheric concentrations have existed only for a few years. Because of the reactivity of H₂O₂ in the liquid phase and its high solubility, it is used up quickly in clouds over mid-latitudes (Reference Kelly, Daum and SchwartzKelly and others 1985) and the interpretation of the measured data in precipitation samples is extremely difficult. In 1984 it was realized that H₂O₂ is present in relatively large amounts in polar snow and ice samples (Reference Neftel, Jacob and KlockowNeftel and others 1984). It is quite astonishing that hydrogen peroxide, a rather unstable and reactive species, survives the firnification process and is still detectable in ice several thousand years old. The observed long-term trend in Greenland deep ice cores points to a very slow disintegration of H₂O₂ with a half–life of the order of several thousand years (Reference Neftel, Jacob and KlockowNeftel and others 1986). The suppressed chemical reactivity in snow which is well below the freezing point and the ease with which H₂O₂ is built into the ice matrix favour the preservation of records of H₂O₂ in the polar ice sheets. It seems, therefore, that the polar precipitation archives could be used to trace back in time the atmospheric H₂O₂ for high latitudes, if the transfer function from the atmospheric concentration to the ice concentration is fully understood.

In this paper we present two continuous H₂O₂measuring series: from Siple Station, West Antarctica, covering the last 83 years, and from Dye 3, south Greenland, covering the last 10 years. The different steps required to evaluate the atmospheric concentration from the measured ice data are discussed.

MEASURING TECHNIQUE

The data presented in this paper are based on discrete samples. The samples are cut with a band saw to give a cross-section of about 2 cm². The sample length is chosen according to the desired vertical resolution. Saw-dust is removed as far as possible, with a Teflon brush. The samples are put into polyacrylic vessels and melted in a moderately warm-water bath. Immediately after melting, aliquots of these samples are inserted into the measuring system. It is important to keep the time between melting and measurement as short as possible, because H₂O₂ tends to decay in aqueous solutions. Typical values are 1—5% in 15 min.

We have two independent systems to measure H₂O₂ concentrations in aqueous solutions at the ppb (= μg/kg²) level. Both methods are fast and need only about 200 μ1 of sample per measurement.

The first method is based on the chemiluminescent reaction of hydrogen peroxide with TCPO, i.e. bis(2,4,6-trichlorophenyl) oxalate (Reference Klockow and JacobKlockow and Jacob 1986). The emitted light during the reaction is proportional to the H₂O₂ concentration. The solubility of TCPO in water is insufficient; therefore this reagent has to be dissolved in acetone. Perylene is used as a fluorescer. It can be added to the acetone solution. Because the reaction is pH–dependent, the samples have to be adjusted to pH = 7.8 with a borate buffer solution.

In the second method, called the enzymatic method, hydrogen peroxide reacts with 4–ethylphenol and builds dimers of this compound quantitatively. This reaction needs the presence of the enzyme peroxidase. The concentration of this dimer is proportional to the H₂O₂ concentration and can be determined fluorometrically. Addition of borate buffer is necessary also in this method. A similar method is described in Reference Lazrus, Kok, Gitlin and LindLazrus and others (1985).

The two systems are presented schematically in Figure 1. Sample input is exactly the same for both devices. The input tube is moved from sample to sample while the peristaltic pump is running. The sampling time is about half a minute. While the tube is moved, an air segment is inserted automatically between two sample segments. This air bubble is necessary to separate the samples. Without these air segments, mixing in the thin tubes would be stronger. Ten samples are measured in one series between two blanks. The blank water is bi-distilled and a platinum–coated catalyst in the water container removes any remaining H₂O₂. From time to time, the systems have to be calibrated with standard solutions, diluted from a ‰ H₂O₂ stock solution. This stock solution is more stable than the diluted standards, but at least every month it has to be calibrated by titration with KMnO₄.

Fig. 1. Measuring systems: (a) Chemiluminscent method; (b) Enzymatic method.

Both methods need only small samples, and are fast and very sensitive; both methods, however, exhibit interferences with other trace species:

The enzymatic method is sensitive to organic peroxides, such as methylhydroperoxide and peroxyacetic acid, too (Reference Lazrus, Kok, Gitlin and LindLazrus and others 1985). This interference can be eliminated: the enzyme catalase destroys H₂O₂ much faster than other peroxides. The presence of organic peroxides can therefore be tested by adding catalase to the sample before measurement. Until now, we have never found organic peroxides in polar ice.

On the other hand, the chemiluminscent method is much less sensitive to organic peroxides, but another interference leads to problems with this method: the sensitivity to Fe²⁺ is comparable to that for H₂O₂ (Reference JacobJacob unpublished). This would be a minor problem if we had clean samples, because the Fe²⁺ content of polar ice is certainly well below the H₂O₂ content. But when the samples are cut with a band saw, traces of iron may contaminate the sample.

That this problem really exists, is demonstrated in Figure 2. The chemiluminescent method leads sometimes to much higher concentrations than the enzymatic method. If catalase is added, a high residual signal is still found. The difference between the original chemiluminescent signal and this residual signal corresponds to H₂O₂ only. The H₂O₂ concentration determined in this way is in good agreement with the enzymatic measurement. Differences between the two methods were always observed after replacing the saw blade. Therefore the connection between this phenomenon and contamination by the new saw blade is obvious. This problem could be solved by measuring routinely both the total chemiluminescent signal and the residual signal after destroying H₂O₂ with catalase. But the enzymatic method offers some other practical advantages, especially for in–situ measurements: first, the calibration of the system is very stable, because all reagents are present in abundance and an equilibrium concentration after the reaction is measured. The calibration therefore does not depend on the exact reagent concentrations or the pump velocity. Secondly, no acetone is needed and all reagents are soluble in water.

Fig. 2. Interference of the chemiluminescent method: contamination with Fe(II). Squares: chemiluminescent method (total signal); circles: fluorometric method; stars: chemiluminescent method (H₂O₂ signal).

RESULTS FROM SIPLE AND DYE 3

During summer 1983–84, a 200 m core was drilled at Siple Station (Antarctica) by the Polar Ice Coring Office. Siple Station combines a high accumulation rate of 50 cm ice equivalent with an annual mean temperature of –24°C. This is very advantageous for investigations of seasonally varying parameters. The almost complete absence of melt layers is a great advantage of this site for many investigations.

Figure 3a shows the H₂O₂ record of the top 64 m of the Siple core. The sample resolution is 5 cm, corresponding to about 10—20 samples per year. It is remarkable how little noise is present in the H₂Oj signal. The seasonal signal is very strong and is still present at the lower end of the record, which corresponds approximately to the firn—ice transition. That these wiggles really represent seasonal cycles is confirmed by the fact that independent dating of this core by H₂O₂ and electrical conductivity measurements (Reference SchwanderSchwander unpublished) lead to exactly the same result. The electrical conductivity method (ECM) dating by counting annual layers is supported by absolute time marks, provided by historically known volcanic events.

Fig. 3. (a) The Siple Station H_sO₂ record.

An interesting point is the relation between the seasonal parameters H₂O₂ and δ¹⁸O. Unfortunately, no δ¹⁸O measurements on this core segment have been made yet. However, parallel measurements of these parameters in deeper pieces of the same core show that H₂O₂ maxima coincide more or less with δ¹⁸O maxima. Nevertheless, more measurements are necessary to determine the exact phase relationship between H₂O₂ and δ¹⁸O.

A strong seasonality of H₂O₂ does not only exist in Antarctic ice. Measurements on the top 10 m of a 300 m core from Dye 3, drilled in summer 1986, show a similar picture. Dye 3, located in southern Greenland, has about the same accumulation rate as Siple Station, but a higher annual mean temperature of –19.6°C. The occurrence of melting in summer is normal. In Figure 4 the H₂O₂ profile of the top 10 m is plotted, together with the δ18^O profile. The resolution is 5 cm or about 20 samples per year. The variability of the H₂O₂ concentration is even greater than in the Siple Station core and the concentration level is higher. Comparison with the δ¹⁸O record shows that there is a small phase lag between the two parameters. H₂O₂ seems to be shifted to the left, i.e. H₂O₂ maxima are later than δ¹⁸O maxima.

Fig. 4. Dye 3: comparison of H₂O₂ with δ¹⁸O.

DISCUSSION

Hydrogen peroxide is of interest in atmospheric chemistry. First, H₂O₂ is believed to be the most efficient oxidizer for dissolved SO₂ in cloud droplets with pH lower than 5 (Reference Penkett, Jones, Brice and EggletonPenkett and others 1979). Secondly, H₂O₂ is involved in tropospheric photochemistry. It acts as a major sink for HO₂ radicals which are members of the odd hydrogen cycle (Reference Logan, Prather, Wofsy and McElroyLogan and others 1981).

If we want to translate our ice-core data into atmospheric H₂O₂ concentrations, we have to investigate mainly two processes: the scavenging mechanism of H₂O₂ from the atmosphere to the snow, and the influence of firnification and of summer melting.

Condensation of H₂O₂ during precipitation

The initial stage from the atmosphere to the ice core is the formation of snow in the atmosphere. Growth of snow crystals out of the gas phase is the most probable process in polar regions. We therefore neglect the presence of supercooled water droplets, particularly the aggregation of these droplets on to snow crystals (riming). According to Reference JacobJacob (unpublished), our conception about the condensation process is very simple: H₂O₂ is supposed to behave like water vapour, i.e.

(1)

We call this equation the co-condensation model. The idea is the following: H₂O₂ has a freezing point very similar to that of water (-0.41 °C at 1 atmosphere) and is easily built into the ice matrix (see below). If the sticking probabilities of gaseous H₂O₂ and H₂O on a growing ice surface are similar, we do not expect a fractionation between H₂O₂ and H₂O.

We have evidence from experiments that co–condensation is a reasonable model for the incorporation of H,0, into snow. It is actually possible to measure H₂O₂ gas–phase concentrations by a cryogenic technique (Reference Jacob, Tavares and KlockowJacob and others 1986). In this method, the air is pulled through a cooled glass cylinder; the water vapour freezes on the cold surface and with it the H₂O₂ vapour. Jacob checked a possible fractionation, using a cold trap (after the glass cylinder) which removed all remaining water and H₂O₂; there was no difference between the H₂O₂ concentrations in the cylinder and in the cold trap. This means that one can determine the gas–phase H₂O₂ concentration by this method, if the absolute air humidity is known. In another experiment, test gas with known H₂O₂ concentration and humidity was generated, and H₂O₂ was measured using the cryogenic method. The measured concentration was close to the expected value.

We successfully applied the cryogenic method at Dye 3 during June 1986. We have performed 40 gas–phase measurements, about ten of them during or immediately after precipitation events. We also monitored the surface-snow H₂O₂ concentration and found that, within 20% error, the frozen air humidity exhibited the same concentration as the freshly fallen snow. The concentrations of 170—300 ppb fit well with the summer H₂O₂ level of the core (Fig. 4). During sunny periods, however, the cryogenic technique generally led to higher concentrations in the condensed water vapour, between 400 and 800 ppb. The gas–phase H₂O₂ concentrations were calculated using the measured water-vapour pressure. They ranged from 0.4 to 1.5 ppb volume. These values agree well with model calculations. Reference Hov and IsaksenHov and Isaksen (in press) calculate a H₂O₂ concentration of about 0.3 ppb volume for Dye 3 at the end of May. Reference Logan, Prather, Wofsy and McElroyLogan and others (1981) predict about 0.8 ppb volume for this location in summer conditions.

An important feature of co-condensation is the influence of water-vapour concentration. Consequently, the air temperature of snow formation has a strong influence on the relation between atmospheric and snow H₂O₂ concentration. The saturation water–vapour pressure of ice decreases by at least a factor of 10 when temperature decreases 25 °C. This temperature range is a lower limit for the variations between summer and winter at Siple Station and Dye 3. Thus a constant gas-phase H₂O₂ concentration would produce snow concentrations ten times higher in winter than in summer. However, the measured summer–snow H₂O₂ concentration is about five to ten times higher than that of winter snow. Taking into account this temperature effect of co-condensation, we end up with an atmospheric H₂O₂ summer-to-winter ratio of at least 50. This does not contradict model calculations. The two global tropospheric chemistry models mentioned (Reference Hov and IsaksenHov and Isaksen, in press, and Reference Logan, Prather, Wofsy and McElroyLogan and others 1981) predict strong meridional H₂O₂ gradients, from high values in low latitudes to low values in polar regions. The drop to low concentrations is very pronounced in the winter hemisphere. This is a consequence of the missing sunlight, which is necessary for the production of the precursors of atmospheric H₂O₂. Quantitative statements are difficult, because polar locations represent marginal regions in the models, but Logan’s model suggests that summer-to-winter ratios of 50 are not unreasonable.

Firnification

The annual mean concentrations of the Siple Station record exhibit some variations, but no decreasing trend with depth (Fig. 3b). Therefore we assume that the firnification process smooths the initial H₂O₂ distribution, but does not affect the mean concentrations.

Fig. 3b. The Siple Station H₂O₂ record. Annual mean values.

Smoothing of a signal in the firn due to evaporation, gas diffusion and recondensation is typical for “water–related” components like hydrogen and oxygen isotopes. H₂O₂ is the first chemical impurity in which a similar smoothing can be observed. We have tried to reconstruct the original H₂O₂ signal in a similar way to Reference JohnsenJohnsen (1977). Diffusion is most effective in the upper firn, due to high porosity and temperature variations (Reference Whillans and GrootesWhillans and Grootes 1985). For the following calculation we therefore assume that diffusion is limited to the upper firn. Then we can reconstruct the original signal for the lower firn. The effect of diffusion on a given isotope or H₂O₂ signal x(z), z being the depth, can be described in the frequency domain by a Gaussian low-pass filter F(k), k being the wave number, if x(z) is the original signal, y(z) the diffused signal and X(k) and Y(k) their Fourier transforms:

(2)

λ is the wavelength and L the (empirical) diffusion length. We adopt from Reference JohnsenJohnsen (1977) a value L = 8 cm of ice. The reconstruction of x(z) requires the inversion of the filter F(k). To avoid the blow-up of high-frequency noise, we take the filter G(k) instead of F⁻¹(k):

(3)

ε determines the degree of low-pass filtering. The result of this reconstitution is shown in Figure 5, calculated with ε = 0.15. Further calculations showed that a smaller ε increases the high-frequency noise, but does not affect the annual signal substantially. It is clear that this calculation should not be over–interpreted, but the reconstituted signal at least shows a variability similar to the uppermost, only weakly diffused, annual cycles, and the minima are still always positive. That means that the diffusion length of S cm is a reasonable value for diffusion in the firn not only for δ¹⁸O, but also for H₂O₂.

Fig. 5. Deconvolution of the H₂O₂ signal in the lower firn (correction for diffusion). The diffusion length is estimated to 8 cm of ice.

CONCLUSIONS

Hydrogen peroxide is found to be one of the clearest seasonal tracers in both Arctic and Antarctic high-accumulation ice cores. A typical summer-to-winter ratio of 5 in the snow concentration corresponds to an atmospheric ratio of about 50. Model calculations of the atmospheric H₂O₂ distribution confirm this result, and they also predict a concentration level which is in good agreement with the values derived from our in-situ measurements.

Firnification has a smoothing effect on the H₅Oj signal. For the Siple Station record, this smoothing can be described by a diffusion length of about 8 cm of ice at the firn—ice transition, which is similar to that of δ¹⁸O. The annual mean H₂O₂ concentrations of the Siple Station core exhibit large variations, but not a significant temporal trend.

The perfect seasonality, the possibility of in-situ measurements, and the small sample volume make hydrogen peroxide a very interesting parameter for ice-core dating.