Extraction of CO2 from air samples for isotopic analysis and limits to ultra high precision delta18O determination in CO2 gas

The determination of delta18O values in CO2 at a precision level of +/-0.02 per thousand (delta-notation) has always been a challenging, if not impossible, analytical task. Here, we demonstrate that beyond the usually assumed major cause of uncertainty - water contamination - there are other, hitherto underestimated sources of contamination and processes which can alter the oxygen isotope composition of CO2. Active surfaces in the preparation line with which CO2 comes into contact, as well as traces of air in the sample, can alter the apparent delta18O value both temporarily and permanently. We investigated the effects of different surface materials including electropolished stainless steel, Duran glass, gold and quartz, the latter both untreated and silanized. CO2 frozen with liquid nitrogen showed a transient alteration of the 18O/16O ratio on all surfaces tested. The time to recover from the alteration as well as the size of the alteration varied with surface type. Quartz that had been ultrasonically cleaned for several hours with high purity water (0.05 microS) exhibited the smallest effect on the measured oxygen isotopic composition of CO2 before and after freezing. However, quartz proved to be mechanically unstable with time when subjected to repeated large temperature changes during operation. After several days of operation the gas released from the freezing step contained progressively increasing trace amounts of O2 probably originating from inclusions within the quartz, which precludes the use of quartz for cryogenically trapping CO2. Stainless steel or gold proved to be suitable materials after proper pre-treatment. To ensure a high trapping efficiency of CO2 from a flow of gas, a cold trap design was chosen comprising a thin wall 1/4" outer tube and a 1/8" inner tube, made respectively from electropolished stainless steel and gold. Due to a considerable 18O specific isotope effect during the release of CO2 from the cold surface, the thawing time had to be as long as 20 min for high precision delta18O measurements. The presence of traces of air in almost all CO2 gases that we analyzed was another major source of error. Nitrogen and oxygen in the ion source of our mass spectrometer (MAT 252, Finnigan MAT, Bremen, Germany) give rise to the production of NO2 at the hot tungsten filament. NO2+ is isobaric with C16O18O+ (m/z 46) and interferes with the delta18O measurement. Trace amounts of air are present in CO2 extracted cryogenically from air at -196 degrees C. This air, trapped at the cold surface, cannot be pumped away quantitatively. The amount of air present depends on the surface structure and, hence, the alteration of the measured delta18O value varies with the surface conditions. For automated high precision measurement of the isotopic composition of CO2 of air samples stored in glass flasks an extraction interface ('BGC-AirTrap') was developed which allows 18 analyses (including standards) per day to be made. For our reference CO2-in-air, stored in high pressure cylinders, the long term (>9 months) single sample precision was 0.012 per thousand for delta13C and 0.019 per thousand for delta18O.