Determination of the analytical performance of a headspace capillary gas chromatographic technique and karl Fischer coulometric titration by system calibration using oil samples containing known amounts of moisture

Over the past few years, concerns have been raised in the literature about the accuracy of the Karl Fischer (KF) method for assessing moisture in transformer mineral oils. To better understand this issue, the performance of a static headspace capillary gas chromatographic (HS-CGC) technique was compared to that of KF coulometric titration by analyzing moisture in samples containing known amounts of water and various samples obtained from the National Institute of Standards and Technology (NIST). Two modes of adding samples into the KF vessel were used: direct injection and indirect injection via an azeotropic distillation of the moisture with toluene. Under the conditions used for direct injection, the oil matrix was totally dissolved in the anolyte, which allowed the moisture to be titrated in a single-phase solution rather than in a suspension. The results have shown that when HS-CGC and combined azeotropic distillation/KF titration are calibrated with moisture-in-oil standards, a linear relation is observed over 0-60 ppm H(2)O with a correlation coefficient better than 0.9994 (95% confidence), with the regression line crossing through zero. A similar relation can also be observed when calibration is achieved by direct KF addition of standards prepared with octanol-1, but in this case an intercept of 4-5 ppm is noted. The amount of moisture determined by curve interpolation in NIST reference materials by the three calibrated systems ranges from 13.0 to 14.8 ppm for RM 8506 and 42.5 to 46.4 ppm for RM 8507, and in any case, the results were as high as those reported in the literature with volumetric KF titration. However, titration of various dehydrated oil and solvent samples showed that direct KF titration is affected by a small bias when samples contain very little moisture. The source of error after correction for the large sample volume used for the determination (8 mL) is about 6 ppm for Voltesso naphthenic oil and 4 ppm for toluene, revealing a matrix effect on the measurement. Finally, the results revealed that HS-CGC is a good technique for measuring moisture in oil samples and that the use of azeotropic vapors for introducing moisture into the titrator almost completely eliminates the matrix effect observed with the oil components. Direct KF injection could also be used provided the system is calibrated with moisture-in-oil standards prepared in the same matrix which is to be used for the determination.