Investigating adsorption/desorption of carbon dioxide in aluminum compressed gas cylinders

Between June 2010 and June 2011, the National Institute of Standards and Technology (NIST) gravimetrically prepared a suite of 20 carbon dioxide (CO2) in air primary standard mixtures (PSMs). Ambient mole fraction levels were obtained through six levels of dilution beginning with pure (99.999%) CO2. The sixth level covered the ambient range from 355 to 404 μmol/mol. This level will be used to certify cylinder mixtures of compressed dry whole air from both the northern and southern hemispheres as NIST standard reference materials (SRMs). The first five levels of PSMs were verified against existing PSMs in a balance of air or nitrogen with excellent agreement observed (the average percent difference between the calculated and analyzed values was 0.002%). After the preparation of a new suite of PSMs at ambient level, they were compared to an existing suite of PSMs. It was observed that the analyzed concentration of the new PSMs was less than the calculated gravimetric concentration by as much as 0.3% relative. The existing PSMs had been used in a Consultative Committee for Amount of Substance-Metrology in Chemistry Key Comparison (K-52) in which there was excellent agreement (the NIST-analyzed value was -0.09% different from the calculated value, while the average of the difference for all 18 participants was -0.10%) with those of other National Metrology Institutes and World Meteorological Organization designated laboratories. In order to determine the magnitude of these losses at the ambient level, a series of "daughter/mother" tests were initiated and conducted in which the gas mixture containing CO2 from a "mother" cylinder was transferred into an evacuated "daughter" cylinder. These cylinder pairs were then compared using cavity ring-down spectroscopy under high reproducibility conditions (the average percent relative standard deviation of sample response was 0.02). A ratio of the daughter instrument response to the mother response was calculated, with the resultant deviation from unity being a measure of the CO2 loss or gain. Cylinders from three specialty gas vendors were tested to find the appropriate cylinder in which to prepare the new PSMs. All cylinders tested showed a loss of CO2, presumably to the walls of the cylinder. The vendor cylinders exhibiting the least loss of CO2 were then purchased to be used to gravimetrically prepare the PSMs, adjusting the calculated mole fraction for the loss bias and an uncertainty calculated from this work.