Anal. Chem. 2001, 73, 520-526
Bias Assessment of Current Technologies Used for the Determination of Low Levels of Moisture in Mineral Oil Samples Roland Gilbert,* Jocelyn Jalbert, and Pierre Te´treault
Institut de recherche d’Hydro-Que´ bec (IREQ), 1800, boulevard Lionel-Boulet, Varennes, Que´ bec, Canada J3X 1S1
For the past few years, the accuracy of the Karl Fischer (KF) coulometric titration in mineral oil using method B of ASTM standard D 1533-881 has been questioned. In a paper published in 1994, Jones and Mayne2 showed that determinations are
affected by low positive biases (3-9 ppm). (Unless specified, all concentrations are expressed in ppm (w/w).) They observed similar biases for coulometric and volumetric titrations and deduced that the deviations were caused by a chemical phenomenon. This paper was followed by the publication between 1995 and 1999 of the work of Margolis, from the National Institute of Standards & Technology (NIST).3-6 In contradiction with the previous finding, this author claimed that coulometry is affected by a high negative systematic error. The demonstration is not direct since it is obtained using comparisons with volumetric titration under conditions specific to NIST, which show much higher water values on duplicate samples. The explanation provided by the author is that when coulometry is applied, there is sequestration of part of the water in the oil collected above the reagent in the cell, which is no longer accessible to the KF reagent. This work led NIST to modify the investigation reports of the reference oil materials.7 A recommended value obtained from a NIST modified ASTM D 1533-88 volumetric method (39.7 ( 2.8 ppm for RM 8506 and 76.8 ( 2.3 ppm for RM 8507) is reported together with a biased low value from a consensus among different laboratories which had applied the ASTM D1533-88 coulometric method (21 ( 3 ppm for RM 8506 and 47 ( 4 ppm for RM 8507). In a paper published in 1999, static headspace capillary gas chromatography, direct oil sample injection into a KF coulometric titrator, and indirect oil sample injection via azeotropic distillation combined with continuous KF coulometric titration were used to investigate the NIST reference materials.8 For the first time, a coulometric titrator was calibrated using oil samples containing known amounts of water. The study revealed intercepts of about 4-5 ppm on the calibration curves when the standards were introduced directly into the KF cell, which was not the case for the curves obtained using the other two approaches. The intercept value was of the same order and sign as the deviations reported by Jones and Mayne on similar oil types.2 The interpolated moisture values obtained with the three calibrated techniques ranged from 13.0 ( 0.4 to 14.8 ( 0.6 ppm for RM 8506 and from
* Corresponding author: (e-mail)
[email protected]. (1) ASTM D 1533-96. Standard Test Method for Water in Insulating Liquids (Karl Fischer Reaction Method: Method B). Annual Book of ASTM Standards; American Society for Testing and Materials: Philadelphia, PA, 1998. (2) Jones, C. F.; Mayne, A. Application and Limitations of the Karl Fischer Technique of Moisture Analysis in Electrical Plant Insulation. Proc. 4th Int. Conf. Properties Appl. Dielectric Mater.; 1994; pp 895-898.
(3) Margolis, S. A. Anal. Chem. 1995, 67, 4239-4246. (4) Margolis, S. A. Anal. Chem. 1997, 69, 4864-4871. (5) Margolis, S. A. Anal. Chem. 1998, 70, 4264-4270. (6) Margolis, S. A. Anal. Chem. 1999, 71, 1728-1732. (7) National Institute of Standards and Technology, Report of Investigation, Moisture in Transformer Oil, Reference Materials 8506 and 8507. Gaithersburg, MD, March 27, 1992; May 7, 1996; June 13, 1997. (8) Jalbert, J.; Gilbert, R.; Te´treault, P. Anal. Chem. 1999, 71, 3283-3291.
The problem in the current debate on the accuracy of Karl Fischer (KF) titrations lies in the fact that coulometry is being compared to volumetry on mineral oil samples for which the true moisture content is unknown. To clarify this point, dehydrated oil samples equilibrated under known temperature and relative humidity conditions and equilibrated oil samples containing known amounts of added moisture were used to assess the accuracy of the determinations. In addition, the measurements were extended to other techniques given that it is unlikely that they would be affected by the same phenomenon causing the KF systematic errors. The samples sent to different laboratories were analyzed by headspace/capillary gas chromatography, gas-phase H2 sensor, oil-phase or gasphase RH sensors, KF coulometric titration with direct or indirect injection, and KF volumetric titration using a standard or NIST modified procedure. The laboratory comparison showed that with the exception of 4 techniques out of 10 that were tested, the measurements gave results in the expected concentration range. Considering the exceptions, two techniques based on volumetric titration yielded results tainted with an important positive bias for both sample types. This bias, tentatively associated with the high iodine end point concentration used by these systems, was estimated at ∼22 ppm under the conditions applied by NIST. On the other hand, the two RH sensors showed a marked tendency to underestimate the value of the samples containing high moisture content. In this case, a loss of analyte through wall adsorption during the time required to achieve steady-state conditions in the measuring chamber seems to be at the origin of the negative biases.
520 Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
10.1021/ac0003495 CCC: $20.00
© 2001 American Chemical Society Published on Web 01/03/2001
42.5 ( 1.9 to 46.4 ( 0.8 ppm for RM 8507. In a more recent paper, using two techniques, one based on direct coulometry with modified reagents designed for diaphragm-free coulometry and the other on stripping at elevated temperatures with continuous coulometry, Cedergren and Nordmark9 reported respectively 23.1 ( 0.6 and 14.1 ( 0.1 ppm for RM 8506 and 47.2 ( 1.2 and 45.2 ( 1.9 ppm for RM 8507. The higher direct coulometric value measured for RM 8506 was attributed to interference caused by the direct injection of the oil matrix into the coulometric vessel. By using an SO2-free reagent in which no water reaction can take place, these authors estimated the bias at 6-8 ppm on the RM 8506 sample under the conditions of their coulometric system. As for the Jones and Mayne data, these results confirmed the value obtained with a coulometric titrator calibrated with waterin-oil standards. The debate around these results reported by different groups is sustained by the fact that the data were obtained with mineral oil samples for which the exact moisture content is unknown (e.g., RM 8506 and RM 8507). The determination of a higher moisture value for such samples by volumetry may indicate a high positive bias on these measurements rather than water sequestration under the coulometric KF conditions, as claimed by one author. To clarify this point, dehydrated oil samples equilibrated under known conditions of temperature and relative humidity (RH) and equilibrated oil samples containing known amounts of added moisture were used to assess the accuracy of the determinations. In addition, the measurements were extended to other techniques based on different physicochemical principles knowing that it is highly unlikely that they would be all affected by the same phenomenon causing the KF systematic errors. The techniques selected were headspace/capillary gas chromatography, gas-phase H2 sensor, oil-phase and gas-phase RH sensors, KF coulometric titration with direct and indirect injection, and KF volumetric titration using a standard or NIST modified procedure. The measurements were carried out at the laboratories where most of the techniques were developed. Before proceeding with the laboratory comparison, a study was done to verify the integrity of the water content of our samples over time. The measurement techniques used for this verification were first validated. These results are presented and discussed in this paper. EXPERIMENTAL SECTION Sample Preparation. The study was performed with samples prepared in the Voltesso 35 gas-absorbing naphthenic oil used by Hydro-Que´bec in high-voltage equipment. A patent pending dehydrating process10 was applied to the oil to remove the moisture. As explained in the application, the process consists of introducing 10 g of CaC2 into 50-mL polyethylene bottles (Nalgene Centrifuge Ware), which are then filled to the top with an oil in equilibrium with ambient air (30-35% RH and 21-22 °C). The bottles are then agitated at 70 °C for 1 h and centrifuged at 14 000 rpm for 10 min to separate the liquid phase consisting of dried oil and the solid phase consisting of a mixture of Ca(OH)2 and unreacted CaC2. A first step of dried oil recovery is then carried (9) Cedergren, A.; Nordmark, U. Anal. Chem. 2000, 72, 3392-3395. (10) Jalbert, J.; Gilbert, R. Process for Dehydrating a Mineral Oil or Other Solvents for the Preparation of Moisture-In-Oil or Moisture-In-Solvent Standards. Hydro-Que´bec, Que´bec, Canada, Patent Pending, s/n 08/644,432, 1996.
out by introducing the bottles into the glovebox so that their contents can be decanted into a second series of polyethylene bottles. The humidity of the glovebox was maintained at 0.71.7% RH during sample handling. The new series of polyethylene bottles is then again centrifuged under the same conditions as above. This centrifugation-decantation cycle is applied four times before pouring the oil of the bottles into a 1-10-mL dispenser located in the glovebox. The dispenser oil was then equilibrated under the atmospheric conditions of the glovebox. On the basis of the Voltesso 35 water saturation data (moisture content at 100% RH ) 16, 27, 35, 45, and 56 ppm at 0, 10, 15, 20, and 25 °C, respectively),11 the amount of moisture picked up by the dispenser oil was estimated to be 0.4-0.8 ppm, depending on the prevailing RH and temperature conditions (assuming a linear relation between 0 and 100% RH as indicated in ref 2). The dispenser was used to discharge 8-mL aliquots of oil with an accuracy of (50 µL into opened 10-mL gastight SGE glass syringes (Supelco). Moisture-in-oil samples were prepared by adding known volumes (1, 2, 4, 5, and 8 µL) of a water-saturated octanol standard into the opened syringe containing 8 mL of oil by using a Digital 0-50µL syringe. The syringe plunger was then attached to the barrel and the content was homogenized during 1 h at 70 °C by shaking. Prior to this step, some stainless steel bearings were added in the syringes to assist homogenization. Some samples were collected directly from the polyethylene bottles using the 10-mL SGE glass syringes. In this case, the dehydrated oil was in contact with the atmosphere of the glovebox for a very short period of time (∼2 min). All the samples were kept in the dark at room temperature in the laboratory or in the glovebox. To minimize moisture desorption or adsorption by the glass and metal wall material, the SGE syringes were preconditioned with the same type of samples that were intended to be used for the study. Techniques for Qualifying the Samples. The KF coulometric titrations were performed with a Mitsubishi Moisturemeter, model CA-06 (Mitsubishi Chemical Corp.), equipped with a membrane titration cell, model CAMCL2, for coulometric measurements. All the trials took place in a closed titration vessel using two sample injection methods. For direct injection, the samples contained in the gastight SGE syringes were introduced into the vessel through a septum. For each determination, the total 8-mL volume of the syringe was injected into the titration cell. For indirect injection, an azeotropic distillation was applied using an Aquastar oil evaporator, model EV-6L (EM Science). A transfer line was used to connect the evaporator chamber with the KF vessel. After the 10 mL of toluene used as the azeotropic former had been dried by partial distillation, the 8 mL of oil contained in the SGE syringes was introduced into the chamber through a septum. Ultrahighpurity nitrogen was used to carry off the azeotropic vapor into the titrator cell, where titration was performed with the distillate. Each titration using the direct injection mode was performed in strict accordance with the NIST requirement of having the oil sample completely dissolved in the vessel solvent.7 This necessitated refreshing the anolyte every two titrations for a maximum addition of 16 mL of oil into the compartment. The instrumental conditions of both injection modes are summarized in Table 1. (11) Technical information on Voltesso 35 electrical insulating oil, Imperial Oil Ltd., 90 Wynford Drive, North York, ON, Canada.
Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
521
Table 1. Instrumental Conditions for KF Coulometric Titration Titrator Parameters ac-bipotentiometric indicating electrode system polarization current 19-20 µA polarization voltage 102 mV end point iodine excess concn 1.5 × 10-4 Ma anodic solution 70 mL of Hydranal-Coulomat AG-H + 30 mL of CHCl3 cathodic solution 2.5 mL of HydranalCoulomat CG sensitivity 0.01 µg of H2O/s titration fixed time 5 min (
