Introduction
Measuring the total amount of air trapped by polar ice can provide valuable information on the atmospheric pressure prevailing during the time of ice formation and Consequently on ice-sheet-elevation changes Reference Raynaud, and RobinRaynaud, 1983). The air-content profiles obtained along deep ice cores furthermore reflect a temporal variability of the ice porosity when the air bubbles dose off. They ma) consequently also depend on other past climatic conditions such as wind speed and temperature, which influence ice sintering, and also the initial properties of snow (Reference Martinerie,, Lipenkov, Raynaud, Chapperllaz, Barkov and LoriusMartinerie and others. 1994).
Four different analytical methods have been used until now for measuring the air content in polar ice:
-
1. The first technique is based on gas extraction by ice melting under a liquid and air-volume measurement in a burette(Reference Arnold-Alyabyev,Arnold-Alyabyev, 1930). The error of the method is considered to be as much as 5%, Reference Langway,Langway. 1958) mainly due to uncertainties associated with the air dissolved in the meltwater. The liquid commonly used in this technique is kerosene Reference Langway,Langway. 1958) but air-saturated dimethyl silicone oil Kameda and others, 1990a) and salt-saturated water (Reference Lipenkov,, Barkov, Martinerie and RaynaudLipenkov and others, 1993 have then been used to improve the precision of the measurements.
-
2. The second method consists of chromatographic air-peak measurements after the air has been extracted from the ice by its melting-refreezing under vacuum. This method has been implemented in parallel with methane-concentration measurements (Reference Chappellaz,Chappellaz. 1990). The absolute accuracy is estimated to be about 5% but the relative precision is better.
-
3. A barometric method of air-content measurements has been previously used along with a dry-extraction procedure Reference Kameda,, Nakawo, Nagoshi and MaeKameda and others 1990a) The total amount of air extracted by crushing the ice sample in a vacuum vessel of known volume was estimated with an accuracy better than 1% on the basis of precise air-pressure and temperature measurements. Nevertheless, the experimental error of the overall procedure may be significantly larger due to the limited efficiency of the crushing procedure in extracting all the air occluded in the ice (Reference Raynaud,, Delmas, Ascencio and LegrandRaynaud and others. 1982).
-
4. The most accurate technique for air-content measurement has been developed on the basis of the vacuum-volumetric method (Reference Raynaud, and WhillansRaynaud and others. 1982). A glass vacuum line connected to a glass vessel containing the sample is used to extract air from the ice by melting refreezing and to collect the air. after trapping the water vapor, in the burette of a Toepler pump for precise volume measurement under known room temperature and atmospheric pressure. The calibration tests performed by Reference Martinerie,Martinerie 1990 and recently repeated during the present study have proved the absolute accuracy of the vacuum-volumetric method to be within ±1.5%, Average reproducibility of the method estimated by comparison between results obtained with two different extractiôn lines on ice samples cut from the same horizontal slice of ice core was found to be about ±0.7%. It is interesting to note that the same method has been used coupled with gas extraction by sublimation of ice instead of its melting refreezing (Reference Raynaud,Raynaud. 1977: Reference Raynaud, and WhillansRaynaud and others, 1982). A systematic overestimation of air content measured after the sublimation procedure may result from gas solid interactions occurring during trapping of the water vapor (Ravuaud and others. 1982).
Although the accuracy provided by the vacuum-Volumetric method is good enough for air-content measurements, the experimental procedure has several disadvantages low productivity, the use of health-hazard products: mercury and phosphoric anhydride acid, the relative fragility of the glass extraction line, the need to use vacuum grease) which also cause difficulties in its application under field and laboratory conditions. Therefore, a new experimental technique has been developed.
Here, we describe the new apparatus for air-content measurement based on the barometric method. The sources of experimental uncertainties, absolute accuracy and reproducibility of the measurements are considered on the basis of theoretical calculations and experimental tests.
Natural Variability of Air Content in Polar Ice
The air content V is commonly expressed as a volume (cm3STP) of dry air in 1 g of ice reduced to standard conditions (STP: Ts = 273 K, Ps = 1013mbar). The air content in polar ice ranges from 0.07 to 0.13cm3 (STP) g&minus1 (Reference Martinerie,, Raynaud, Etheridge, Barnola and MazaudierMartinerie and others, 1992) due to the geographical and/or temporal variations of atmospheric pressure Pc (mbar), temperature Tc (K) and pore volume Vc (cm3 g−1 at the close-off depth in the ice sheet where the air becomes isolated from the atmosphere:
Overall variability of the air content, observed along deep ice cores recovered in Greenland and in Antarctica has an amplitude up to 25% (Reference Budd, and MorganBudd and Morgan, 1977; Reference Raynaud,, Lorius, Hiuld and YoungRaynaud and others. 1979; Reference Raynaud,, Delmas, Ascencio and LegrandRaynaud and Whillans, 1982; Reference Herron, and LangwayHerron and Langwav. 1987; Reference Kameda,, Nakawo, Mae, Watanabe and NaruseKameda and others, 1990b: Reference Lipenkov,, Barkov, Martinerie and RaynaudLipenkov and Others, 1993; Reference Martinerie,, Lipenkov, Raynaud, Chapperllaz, Barkov and LoriusMartinerie and others, 1994). On the one hand, this variability may reflect long-term variations in:
Elevation at which the air becomes trapped in the ice because of down-slope ice advection or changes in ice-sheet surface elevation an elevation change of 50 m may generate a change of about 1% in air content).
Atmospheric pressure (about 7 mbar 1%).
Temperature (2°C/1%), including the temperature effect on close-off porosity.
On the other hand, an air-content signal with the same order of magnitude can be induced by variability of close-off porosity, Vc , associated with the stratigraphy of the ice sediments due to different weather conditions causing very short-term variations within a few centimeters of ice core or with wind-speed changes due to climatic variability (Reference Martinerie,, Lipenkov, Raynaud, Chapperllaz, Barkov and LoriusMartinerie and others, 1994). Investigation of the short-term V variations with respect to other structural properties of the ice provides experimental constraints for modeling die ice-sintering and air-trapping processes in ice sheets Reference Martinerie,, Raynaud, Etheridge, Barnola and MazaudierMartinerie and others. 1992).
A common experimental uncertainty for all methods of air-content measurement is due to the so-called “cutbubble effect” because some gas is lost when cutting air bubbles (or cavities formed from air-hydrate decomposition after ice relaxation) at the surface of the ice sample during its preparation. This effect can be estimated quantitatively below a certain depth in the ice sheet 5–10m beneath close-off) where the shape of bubbles is simple enough for determination of their size. For ice samples of about 25 g. the correction of air-content measurements for the effect of cm bubbles varies from 10% near the close-off depth to 1% in the deeper strata (Reference Martinerie,, Lipenkov and RaynaudMartinerie and others. 1990). The corresponding uncertainty for air-content values may exceed 1.5% in the upper part of the experimental profile.
From all these considerations, an absolute precision of about 1% for the analytical technique itself seems to be appropriate with regard to the amplitudes of the V signal to be detected and to the “cut-bubble effect”. In the following sections we focus the discussion on experimental uncertainties of the analytical procedure itself, based on the barometric method.
The Barometric Method
Principle
The barometric method is based on the precise evaluation of the pressure P and temperature T of dry air extracted from an ice specimen by its melting refreezing under a vacuum in a volume-calibrated cell. The real air sample contained in the cell is hereafter considered to be a mixture of two ideal gases: dry air and water vapor. Thus, if-Pmeans is the pressure measured in the cell, the pressure P of dry air will be given by:
where РH2O the partial pressure of saturated water vapor at temperature T. Air content V of the ice is then obtained from the following relationship:
where Mi is the mass of die ice specimen and U is the volume occupied by the air sample in the cell and equal
where Ucal is the volume of the calibrated cell while Mi/ρi represents the volume of the refrozen pure ice whose density. &rhoi , is mostly a function of its temperature (Reference Bader,Bader, 1964).
A diagram of the measuring system is shown in Fig. 1. An ice sample with known mass. Mi , is melted under a vacuum in cell 1 and then gradually refrozen by cooling the bottom of the cell to avoid air-bubble trapping during the freezing process. The cell and part of the line containing the extracted gas is immersed, before pressure measurement, in a liquid cryostat (shown in Fig. 1 by the dotted square to keep the air sample homogeneous at a low temperature T (−30°C). The low temperature ensures a minimum of water-vapor pressure PH2O in the line. The pressure Pmeas of the gas is then measured with a type 590 Barocel (DATAMETRICS) pressure transducer (2 in Fig. Fig. 1).
Fig. 1 Diagram of the measuring system.
The extraction line has three calibrated volumes Uc, UB and UE delimited in Figure 1 by dashed outlines. Volume UC includes the volume of the empty cell plus the internal volumes of the valve v1, and tubes up to the valves v2 and v3. Volume UB is the sum of the internal volumes of the Barocel itself and fitting tube up to the valve v2. The sum UC + UB will allow us to calculate Ucal the calibrated part of the line which is occupied by extracted gas and refrozen ice at the time of pressure measurement. Volume UE is used only for self-calibration of the system. It includes expansion chamber 3 plus internal volumes of the valve v4 and all fitting tubes up to the valves v3. v5 and v6.
Optimum parameters of the measuring system
To estimate the accuracy of the experimental technique, we use the general equation relating the error ΔΥ of the function Y = f(x 1,x 2,...,x n) to the individual errors Δx i of die variables xi obtained from direct measurements
According to Equation (5), the error of the air-content value determined from Equation (3) can be calculated as
where ΔM i and ΔT are the errors of sample mass and cryostat temperature (direct measurements), while ΔP and ΔU are the estimates of pressure P and volume U errors, obtained from Equations (2) and (4) respectively:
and
If all individual errors are determined, the optimum ratio between sample mass. Mi. and cell volume, Ucall , for obtaining the best accuracy for air-content measurement can be evaluated on the basis of Equations (2)–(4) and (6)–(8). Assuming reasonable values and ranges: V =
The New Technique
Description of the device
The vacuum line shown in Figure 1 is built using stainless steel tubing (6.35mm) and CAJON VCR fittings. A diaphragm type of NUPRO valve is chosen for the valves v1− because of its negligible change of internal volume between “close” and “open” positions, while NUPRO bellows valves are installed as V5,6. The line can be evacuated by a turbo molecular pump backed by a rotary oil pump. A Pirani-type gauge is used for vacuum control.
The extraction cell is a glass metal container. The bottom part of the cell is made of glass to allow observation of ice melting and refreezing, while its upper part and the lid are made of stainless steel. A copper “O” ring gasket is used to ensure good air-tightness between the cell and the lid, and to keep the cell volume as constant as possible during successive manipulations. The internal volume UC of the empty cell and related tubing was designed to be about 137 cm3.
The pressure transducer (type 590 Barocel) has a thermal base set at 40°C and measuring a pressure range up to 100 mbar. The pressure applied to its diaphragm is always maintained lower than 100 mbar to avoid hysteresis problems. The output signal is directly proportional to pressure and recorded by a DATA-METRICS-type 1450 pressure read-out. The accuracy of the transducer is about ±0.05% of the reading. The internal volume of the Barocel with no pressure applied is about 5.0cm3, while the diaphragm displacement under a pressure of 100 mbar can be as much as 0.16 cm3, The volume UB of the Barocel and the tubing up to valve v2 is about 15cm3, so that the resultant volume UC + UB ≈ 152 cm3.
The expansion chamber (3 in Fig. 1), which is made of stainless steel, has a lid for introducing a standard volume during the calibration procedure. The lid is coupled to the chamber with a copper “O” ring gasket. The internal volume UE of the chamber, together with the related tubing (up to valves v3, v5 and v6), is about 1363cm3.
The alcohol bath is cooled by a cryostat which maintains the temperature T at −30°C with a long-term temperature stability of ±0.05°C. The bath temperature is controlled by a high-quality stable platinum resistance thermometer. A good homogenization ol 30 1 of ethanol provides a ±01°C; temperature uniformity.
Calibration procedure
The procedure for calibrating the volumes UC,UB and UE is similar to that described by Schwandcr 1984. The calibration is carried out with the aid of two standard stainless steel volumes UI (sphere) and UII (cylinder) which have at 20°C the following characteristics:
The dry air at atmospheric pressure Pa is expanded from a volume Uc into pre-evacuated volumes UE and UB sueecssively. and the resultant pressure in the manifold is measured with Barocel. The operation is repeated under three different conditions: (i) both volumes Uc UE are empty (the resultant pressure after expansion is P1 (ii) volume UC contains the standard solid volume U1(P2); (iii) UC contains the standard volume UII (P3). All parts of the line during the calibration are equilibrated at room temperature and pressure values P1,2,3 are converted to 20°C:. The atmospheric pressure is assumed to remain constant throughout the calibration time. From the above procedure, the volumes UC and (UE+UB) as well as atmospheric pressure Pa , can be obtained:
is a dimensionless parameter of our system which allows calibration of the Barocel by adjusting the pressure P1 read on the transducer in order to have Pa obtained from Equation (11) equal to the atmospheric pressure measured by a mercury barometer with an absolute precision of ±0.01%.
Volume UB can be determined by expanding dry air at a pressure P1 from UB to the pre-evacuated volume UC containing the standard UI . The resultant pressure P4 allows us to calculate UB:
>
The measured pressures during the calibration procedure (P1−4 ) are in the range 13–90 mbar. By applying Equation (5), the accuracy of the volume (UC + UB) is ±0.15% and the absolute calibration of the Barocel has a precision of about ±0.2%.
Analytical procedure
A cubic ice sample of about 20–30g (representing about 3 cm of ice-core thickness) is placed, after weighing, into the pre-cooled extraction cell. The extraction cell is then attached to the vacuum manifold with a CAJON VCR fitting welded to the lid of the cell. The cell is evacuated for 10 min with a turbo molecular pump. During pumping, approximately 1% of the ice is removed by sublimation from the sample surface. After pumping, the cell is isolated by closing stopcock v1 and the bottom part of the flask is immersed for 10min in an ultrasonic bath filled with warm water (50°C) to melt the ice. The experiments have demonstrated an efficiency of the ultrasonic bath in avoiding bubble formation during refreezing of the water. The ultrasonic bath is then replaced with the liquid cryostat and the bottom of the flask containing the melted water is immersed for 40min at −30°C. This results in gradual refreezing from the bottom of the cell. The refrozen ice is considered to be a polycrystalline bubble-free ice with a density of ρi = 0.9207 Mgm−3 at −30°C (Reference Bader,Bader, 1964). The extraction cell and that part of the line representing almost all the volume (UC + UB), except for the internal volume of the Barocel with its short connecting tubing, is then immersed in the alcohol bath as shown in Figure 1 by the dotted square. Valves v1 and v2 are opened to expand the air sample through the volume (UC + UB). The extraction line is kept in the liquid cryostat for 30 min. allowing tin-air sample to equilibrate at −30°C and to reduce the water-vapor pressure to 0.37–0.39mbar. At the end of the procedure, the absolute pressure Pmeas (range within 10–30 mbar) and temperature T (around 243 K) of the extracted air are measured with the Barocel and the platinum resistance, respectively.
The total duration of the analytical procedure is slightly more than 2h. For comparison, the time efficiency of the vacuum-volumetric method described by Reference Raynaud, and WhillansRaynaud and others (1982) is about twice as slow.
Experimental uncertainties and corrections
In this section, we estimate the absolute precision of the air-content data obtained using the new technique, taking account of all possible sources of error and uncertainty associated with the calibration and analytical procedures described above. (The uncertainties due to cut-bubble effect are not considered here.)
An accuracy of about ±0.01 mbar for PH2O evaluation is deduced from a comparison between the theoretical values and the direct measurements of water-vapor pressure using the Barocel, while the line containing an ice sample under vacuum was equilibrated at −30°C. Assuming
The mass loss during the time of sample preevacuation was experimentally defined as 0.2±0.1 g; therefore, the correction δMi = −0.2g must be applied to the measured sample mass and the resultant error
To evaluate Ucal in Equation (4), a number of Corrections have to be made to (UC + UB ) as deduced through the calibration procedure: (i) the thermal shrinkage of the given extraction line immersed in the alcohol bath at −30°C with respect to its volume calibrated at 20°C is estimated to be −0.2 ± 0.01 cm3; (ii) the correction for the effect of the Barocel thermal base (40°C) during the calibration is 0.35 ±0.02 cm3; (iii) the effect of the temperature difference between the part of the extraction line immersed in the alcohol bath (−30°C), the short manifold exposed to the room temperature and the internal volume of Barocel (40°C) during the analytical procedure can be expressed as an apparent change of the volume (UC + UB) which corresponds to a correction of −1.9 ± 0.1 cm3; (iv) the effect of the Barocel diaphragm displacement, which is considered to be a linear function of the applied pressure, can be eliminated by an average volume correction of −0.13 ± 0.02cm3 (an estimate based on the difference between pressure measured in the course of the calibration and analytical procedures). Adding together these different uncertainties, the total volume correction δU is −1.9 ± 0.1 cm3. This leads to Ucal = UC + UB 1.9 cm3 with an absolute accuracy of ±0.25 cm3. As the density of refrozen ice in the extraction cell during our experiments was always the same as the density of pure ice within
Finally, taking ΔT/T≤0.05﹪, after calibration of the platinum resistance in the alcohol bath, we can calculate using Equation (6) an overall error
Experimental Tests with Ice-Core Samples
Ice cores from two boreholes drilled in East Antarctica, 100 and 400 km inland from Mirnyy Station were used for the experimental tests. The two ice-corе increments were collected from depths around 100m (borehole at 100 km) and 124 m (400 km). In total, 22 cubic specimens of about 25–30 g have been prepared and measured for air content using the new apparatus and 11 samples using the old volumetric technique. The experimental results, not corrected for the cut-bubble effect, are given in Table 1. columns 3 and 4.
To estimate the reproducibility of the new analytical procedure, the air-content values obtained using the barometric method for two or three samples cut from the same horizontal slices have been compared (see column 6 in Table 1). The reproducibility, expressed as a relative difference between maximum and mean values for the same slices, was found to be not worse ıhan ±1%. The average reproducibility is ±0.45%.
For inter-calibration between the two different analytical techniques, ten couples of adjacent ice samples of equal sizes cut from ten horizontal 3cm thick slices of the ice ioie were chosen. One sample from each couple was measured for air content using the previous volumetric technique and the other using the new barometric technique. Apart from an abnormal case (slice 124.31m in the ice core from 400km). which gives a difference of 3.7%, the difference between V values measured by the two methods varies from 0.1 to 1.7%. This range can be explained by errors of ±1.5 and ±0.6% quoted for the volumetric and barometric methods, respectively. An average value of die differences is equal In −0.23% when nine slices arc laken into account (without slice 124.31m) and to +0.16% for all studied slices. Therefore, our results indicate no systematic difference between the two analytical techniques.
Conclusion
A new device based on a barometric method of air-content measurements in polar ice has been designed for field and laboratory use. The absolute precision of the analytical technique has been estimated a priori and then checked experimentally to lie within ±0.6%. An average experimenta] reproducibility of the new method is found to be better than ±0.5%. The duration of a single analytical procedure is about 2 h. i.e. half that of the previously used volumetric method. Another advantage of the new device, in addition to the high reliability of the stainless-steel extraction line, is that all health-hazardous products (mercury and phosphoric anhydride acid) used in the previous apparatus, have been eliminated.
Table 1. Results of air-content measurements* in Antarctic ice using volumetric and barometric methods
Acknowledgements
We should like to thank N. Fornier for his contribution in testing the device and P. Martinerie for helpful comments on the paper. This work was supported by Programme National d’Étude de la Dynamique du Climat (PNEDC). One of us (V.L.) wishes to thank the Laboratoire de Glaciologie et Géophysique de I’Environnement (LGGE) du CNRS for support of his work at LGGE.
