Anal. Chem. 1997, 69, 4051-4055
Electrochemical Determination of Water in Environmental Hydraulic Fluids Using the Karl Fischer Reaction Anders Cedergren* and Marcus Lundstro 1m
Department of Analytical Chemistry, Umeå University, S-901 87 Umeå, Sweden
Different procedures based on the Karl Fischer reaction were investigated with respect to their applicability for water determinations in environmental hydraulic fluids: (i) continuous coulometry using a recently described diaphragm-free cell; (ii) on-line stripping of water at elevated temperature using either continuous coulometry or direct potentiometry for detection of the liberated water. Except for one of the oils, Statoil PA, which is a poly(rolefin) with certain polymers added, no significant difference was found among coulometry using an optimized imidazole-buffered methanolic reagent containing 75% (v/v) chloroform, the two different stripping techniques (working in the temperature interval 100-110 °C), and the commercially available Hydranal Coulomat AG-H. The high stability and sensitivity of the coulometric technique described made it possible to work with sample amounts in the low milligram-range, and this is shown to increase the reliability of the coulometric method as compared to normally used procedures. Oils based on natural or synthetic esters have been of great importance for lubricating and hydraulic techniques for more than 60 years. Products based on the use of esters, for example, were early recognized for their good lubricating properties, heat stability, low volatility, and low viscosity as well as high viscosity index. During recent years, several new types of biodegradable ester-based oils have been developed by the big oil companies using renewable raw materials like rape or different kinds of this cereal. Unlike petroleum-based products, these oils are capable of binding water in the range 300-700 ppm at room temperature and 40-70% relative humidity. The problems associated with the determination of water in oil products with special reference to transformer oils were recently discussed by Margolis1 in this journal. This author reported on a collaboratory study among 14 analytical laboratories in the United States, which were coordinated by NIST during 1991 with the assistance of the electric power industry and the ASTM D-27 committee. Results reported were obtained by standard coulometric KF methodology and showed good precision. However, Margolis in his own work obtained about 50% higher values when using a volumetric method based on the use of 65-70% chloroform. The discrepancy was attributed to effects arising when the oil is not completely dissolved in the titration medium. When this happens, Margolis proposed that the undissolved oil is capable of binding or sequestering a portion of the water so (1) Margolis, S. A. Anal. Chem. 1995, 67, 4239-4246. S0003-2700(97)00302-8 CCC: $14.00
© 1997 American Chemical Society
that moisture is unavailable for reaction with the KF reagent. It should be mentioned that all of the coulometric methods investigated, including the newly introduced Hydranal Coulomat AGH, gave low results (50-90% recoveries) and this was explained by the low concentration of organic modifier (chloroform) present in the titration medium. Attempts to use higher concentrations were not successful since this resulted in a breakdown of the voltage across the indicating electrodes of the coulometric instrument. It should be mentioned that the methods recommended by ASTM use methanol/chloroform solutions (1:2 and 1:3) for the determination of water in insulating liquids2 and petroleum products.3 Surprisingly, there appears to be no papers in the literature dealing with the determination of water in environmental hydraulic fluids. Commercial laboratories that carry out oil analyses use the same procedure as for petroleum-based oils, which is based on the KF reaction. The aim of the work presented in this paper was therefore to investigate the applicability of different techniques based on the KF reaction for various types of environmental hydraulic fluids. The following techniques were evaluated: (i) continuous coulometry using a recently developed diaphragmfree cell; (ii) on-line stripping of water at elevated temperature using either continuous coulometry or direct potentiometry (zero current) for detection of the liberated water. In continuous coulometry, a current, proportional to the difference in redox potential in the cell and a preselected value, is used to generate iodine until the background level is reached. Five different commercially available hydraulic oils were selected including the following: (i) OK Biohydraul ES (OKES), which is a synthetic polyol ester, (ii) Raisio Biosave 32L, a refined rape oil, i.e., a natural ester; (iii) Statoil VE, a vegetable refined oil with 25-30% synthetic di-, tri-, or polyol ester; (iv) Statoil ES, a synthetic ester of unknown origin based on a renewable raw material like rape; and (v) Statoil PA, a poly(R-olefin), which is a refined petroleum product to which certain polymers are added. EXPERIMENTAL SECTION Chemicals. Methanol (pa) and sodium iodide (pa) were from Merck. Sulfur dioxide (>99%) and imidazole (puriss pa) were from Fluka. Iodine (resublimed), Hydranal Coulomat A (contains imidazole and chloroform), Hydranal Coulomat AG-H (imidazolebuffered mixture of pentanol/methanol), and Hydranal standard 5.00 (50 µg/10 µL) were from Riedel-deHae¨n. Chlorofrom (pa) was from Prolabo. Dehydrite (granular magnesium perchlorate) was from BDH, and high-purity helium was from AGA, Lidingo¨, (2) Gedemer, T.; Frey, T. Am. Lab. 1975, 8 (March), 47-53. (3) MacLeod, S. Anal. Chem. 1991, 63, 559A-566A.
Analytical Chemistry, Vol. 69, No. 19, October 1, 1997 4051
Figure 1. Schematic diagram over the strip cell: (1) strip cell made in TPX (polymethylpentene), (2) silicone membrane, (3) isolated Teflon tubing, PTFE, 0.8 mm i.d., (4) heating device, (5) gas inlet via drying tube filled with granular magnesium perchlorate, (6) copper tubing, (7) thermocouple, and (8) aluminium block.
Sweden. Three of the oils under investigation, Statoil VE, Statoil ES, and Statoil PA, were fresh oils while OKES was a used oil and Raisio Biosave had been exposed to water for an extended period in hot air. Safety Considerations. Methanol is highly flammable and is toxic by inhalation, in contact with skin, and if swallowed. Chloroform is harmful by inhalation and ingestion and may be fatal. Inhalation of vapor may cause headache, nausea, vomiting, and dizziness. Prolonged skin contact may result in dermatitis. The liquid is readily absorbed through the skin. Imidazole is harmful by inhalation, in contact with skin, and if swallowed. Sulfur dioxide is intensely irritating to eyes and the respiratory tract. KF Reagents Used. For the direct coulometric determinations, reagents A, B, C, and Hydranal Coulomat AG-H were used. The compositions of these methanolic reagents were as follows: (A) 75% chloroform/5 M imidazole/0.5 M sulfur dioxide/0.1 M iodine; (B) 75% chloroform/2 M imidazole/0.5 M sulfur dioxide/ 0.1 M iodine; (C) 75% chloroform/2 M imidazole/0.15 M sulfur dioxide/0.1 M iodine. For the experiments using the combination of stripping at elevated temperature and continuous coulometry, the following reagent was used: 5 M imidazole/0.5 M sulfur dioxide/0.05 M iodine in methanol. For the method based on stripping/direct potentiometry, a methanolic reagent containing 20% (v/v) chloroform/1 M sulfur dioxide/2 M imidazole/0.6 M sodium iodide was used. Instrumentation. The diaphragm-free coulometric cell, recently described in this journal4 was used in all experiments including those for which the combination strip cell/direct potentiometric determination was tested. The strip cell was made from TPX (Mitsui Petrochemical Industries Ltd.) and is shown in Figure 1. The temperature of the aluminium block could be controlled to within 1 °C. The inner diameter of the isolated Teflon tube connecting the strip cell with the coulometric cell was 0.8 mm. Inside the coulometric cell, the Teflon tubing was arranged so that it dipped 2-3 cm below the surface of the electrolyte. The outlet of the titration cell consisted of a 0.4 mm (4) Cedergren, A.; Jonsson, S. Anal. Chem., in press.
4052 Analytical Chemistry, Vol. 69, No. 19, October 1, 1997
Figure 2. Typical potentiometric response curve for the procedure based on stripping/direct potentiometry. At point A, a water standard, normally 50 µg, was added, and at point B, a 20-100 mg oil sample was introduced into the strip cell. Calculations of the background was normally made using potential values obtained in the time interval 5-10 min after the introduction of the sample.
i.d. Teflon tubing, positioned in the cover of the titration vessel. The generating and indicating electrodes were connected to an LKB coulometric analyzer, the function of which was described earlier.5 The current signal from the generating electrode system, as well as the redox potential of the indicating electrode system, was followed as a function of time by a Fluke dual-display multimeter connected to a computer in which the data collection program Fluke QS 45 was available. Procedure. Direct Coulometric Titration. The procedure for preparation of the diaphragm-free coulometric cell including calibration was the same as that described recently.4 Draining of the auxiliary electrode compartment was normally carried out once an hour after which a few minutes was required in order to obtain a stable end point. Samples, 20-500 mg, were delivered with a 1 mL plastic syringe equipped with a 5 cm needle (0.9 mm). Before injection of the oil under investigation, the syringe was washed twice with the sample to be delivered. Stripping/Continuous Coulometric Titration. The strip cell was initally washed with 0.5 cm3 chloroform and rinsed with a few milliliters of methanol. The strip cell was then put in place and heated to 150 °C under passage of about 50 cm3 min-1 helium for about 15 min without connection to the coulometric cell. No significant difference in the results was observed when air was used instead of helium. These gases passed through a tube containing granular magnesium perchlorate before entering the stripping cell. The working temperature was then preset, and the strip cell was connected to the coulometric cell by means of a Teflon tubing which could be pressed through a fine hole in the silicone membrane of the top of the cell. After about 5 min, a stable background level, normally about 1 µg min-1 (calculated as water) was obtained. The variation in the background was typically in the range 0.01-0.02 µg min-1. The background value with air as the carrier was typically 1.5 µg min-1. Stripping/True Potentiometry. A suitable starting concentration of iodine (see level at A in Figure 2) in this method was found to be about 1 mM. At this level, the change in redox potential due to the change in iodide concentration (2 equiv of iodide is formed for each mole of water reacted) will be less than 0.1 mV for an addition of 100 µg of water to 12 mL of reagent containing 0.6 M (5) Cedergren, A. Anal. Chem. 1996, 68, 3679-3681.
Table 1. Comparison of Results Obtained for the Water Concentration in Raisio Biohydraul 32 L Using Hydranal Coulomat AG-H and Different Types of Methanolic Reagents Containing 75% (v/v) Chloroforma type of reagent
results, mean value ppm (w/w)
std dev, ppm (w/w)
no. of detns
A B C Hydranal Coulomat AG-H
388 418 406 420
19 2.5 2.5 4.2
5 5 5 6
a
Sample amounts in the range 20-40 mg were used. End point concentration of iodine was 5 × 10-6 M.
Table 2. Coulometric Determination of Water in the Different Oils Using Reagent Ba and Hydranal Coulomat AG-Hb type of oil reagent
Raisio OKES Biosave Biohydraul
B 418 ( 2 Coulomat AG-H 417 ( 3
235 ( 3 233 ( 6
Statoil ES
Statoil VE
Statoil PA
329 ( 7 360 ( 20 93 ( 7 337 ( 6 357 ( 12 65 ( 9c
a For composition, see text and Table 1. b Sample amounts in the range 20-40 mg were used. The results given are mean values and standard deviations of three to five determinations expressed in ppm (w/w). End point concentration of iodine was 5 × 10 -6 M. c Always a trend toward lower values for increasing amounts of oil in the titration vessel.
iodide, assuming that the following equation holds (Nernst law) - 3
E ) constant + RT/2F ln([“I2”]/[I ] )
(1)
where “I2” denotes the uncertainty concerning which iodine species takes part in the redox equilibrium. The water capacity of a reagent with this starting concentration is 216 µg. After addition of the water standard, normally 50 µg of water (Hydranal standard), the remaining working range corresponds to about 150 µg. As can be seen in the figure, only a few seconds is necessary to obtain a stable potential after the water standard had been added. Most of this time is required for mixing since it can be estimated6 that the time for 99.9% reaction under these conditions is less than 1 s. Knowing the temperature, the volume of the KF medium, the change in potential value (from A to B in Figure 2) for a given amount of water and, the value of RT/2F in eq 1, the relation between the redox potential and the concentration of iodine excess can be established. It should be mentioned that in separate experiments based on coulometric generation of iodine, RT/2F was found not to deviate by more than 0.1% from the theoretical value at a given temperature. In Figure 2, the oil sample is added at point C where the time and redox potential are known. By noting the time/potential values in the interval 5-10 min, a background value (micrograms of water per minute) corresponding to a constant addition of water can be determined and this value can then be used to compensate for the total background from point C in the figure. RESULTS AND DISCUSSION Direct Coulometric Determinations. Results obtained with different types of reagents containing 75% (v/v) chloroform were studied and compared with those obtained with the well-known commercial reagent Hydranal Coulomat AG-H (see Table 1). In contrast to the former types of reagents, the latter was not capable of completely dissolving the oil samples under investigation. As was mentioned earlier, results reported by Margolis1 showed that low values will be obtained if the oil is not completely dissolved in the titration medium. Nevertheless, as can be seen in Table 1, the agreement with the results obtained with reagent B and Coulomat AG-H are very good although the oil is not completely dissolved in the latter reagent. The table also shows that the recovery rates for reagents A and C are significantly below 100%. These reagents have a higher pH than reagent B, which might explain the difference in terms of a change in stoichiometry. As (6) Cedergren, A. Anal. Chem. 1996, 68, 3682-3687.
was discussed earlier,7,8 this may occur if a reaction between a hydrolysis product of iodine (hypoiodite) reacts with the hydrolysis product of sulfur dioxide, i.e., bisulfite or sulfite, since for this scenario 1 mol of iodine will correspond to 2 mol of water. The probability for both of these hydrolysis reactions increases at higher pH. If the time required for the release of water from the oil phase is long, the probability that the free sulfur dioxide will react with water will of course increase. A comparison between reagent B and Hydranal Coulomat AG-H is given in Table 2 for five different types of oils. The agreement between the reagents seems to be very good for the Raisio and OKES and reasonably good for Statoil ES and Statoil VE while the results for Statoil PA differ significantly. As will be shown below, the value obtained with Coulomat AG-H for the latter oil is likely to be too low. It should be mentioned that a clear trend toward lower values was always found for Coulomat AG-H and this oil. In contrast to results reported by Margolis,1 the recovery of water from a water standard (water in xylene/butanol) was in the range 99-100% even when a relatively large amount (0.5 g) of oil was present in the titration vessel containing Hydranal Coulomat AG-H. Signal shapes obtained for reagents B and Hydranal Coulomat AG-H under identical conditions, i.e., the same end point concentration of iodine and amplification factor, are shown in Figure 3. It is evident from the recorded titration curves that it is favorable to use small sample volumes, since the time for a complete titration will then be relatively short. For the largest amounts shown in the figure (0.4 g in 12.5 mL of titration medium), the titration times are as long as 15-25 min and will, of course, affect the accuracy and precision. It should be pointed out that normally about 2 g of oil is added to 40-100 mL of titration medium.1 Recently, it was reported4 that the relative standard deviation obtained for titration of 1 µg of water, using the same coulometric instrumentation, was typically 1%. Since the types of oils discussed in this paper normally have a water content in the range 300700 ppm, sample weights in the low-milligram range are sufficiently large to give precision values of 1-2%. The different types of oils investigated in this paper are very complex, and their compositions are not known in detail. In order to investigate possible interference effects, the water reaction was frozen by excluding sulfur dioxide from the reagent. It is of course difficult to design a reagent with the same redox properties as that of reagent B. Nevertheless, a methanolic reagent buffered (7) Cedergren, A. Anal. Chem. 1996, 68, 784-791. (8) Ora¨dd, C.; Cedergren, A. Anal. Chem. 1995, 34, 999-1004.
Analytical Chemistry, Vol. 69, No. 19, October 1, 1997
4053
Table 3. Potentiometric Determination of the Iodine Production Caused by the Different Oils Using a Methanolic Reagent Containing Sulfur Dioxidea type of oil OKES Biohydraul Statoil PA Statoil ES Statoil VE Raisio Biosave
sample wt (mg)
iodine prodctn (ppm water)
35.68 32.52 44.20 49.16 67.63 56.10 57.51
-48 -49 -50 -14 -86 -190 -51
a Composition: 75% chloroform/0.4 M imidazole/0.2 M imidazolium iodide. The results are expressed in ppm (w/w) water by assuming a 1:1 stoichiometric ratio between iodine and water.
Figure 3. (a) Coulometric response curves for consecutive additions of different amounts of OKES Biohydraul to the coulometric cell containing 12.5 mL of reagent B. (b) Same as (a) except that the reagent used was Hydranal Coulomat AG-H. Figure 5. Influence of temperature on the results obtained for the different oils using the stripping/continuous coulometric method. The results are related to those obtained by direct titration using reagent B.
Figure 4. Potentiometric response curves obtained for 35.68, 32.52, and 44.2 mg of OKES Biohydraul added to a methanolic reagent (0.4 M imidazole/0.2 M imidazolium iodide/75% chloroform (v/v)) that did not contain any sulfur dioxide. The amount of iodine produced is expressed in ppm water (i.e., a potential negative interference effect).
to the same pH as reagent B was chosen for these studies. It became clear that all of the oils reacted stoichiometrically with this reagent forming iodine. Signal shapes obtained for OKES are given in Figure 4. As can be seen, the kinetics for the formation of iodine are slow but sufficiently fast to cause a negative interference effect provided that the reaction proceeds with at least the same rate in a KF reagent. A summary of the results obtained 4054 Analytical Chemistry, Vol. 69, No. 19, October 1, 1997
with this reagent is given in Table 3, where it can be seen that a large negative interference effect is to be expected for most of the oils; especially Statoil VE (-190 ppm). As will be shown below, the good agreement between the methods including the alternative technique comparing stripping of the water and electrochemical determination indicates that the potential interference effect shown in Table 3 is eliminated by the presence of methyl sulfite (high concentration) in the KF reagent. This means that this compound will rapidly reduce the potential interferent so that it will be unable to cause any oxidation of iodide to iodine. Stripping at Elevated Temperature. Results obtained for the various oils using different stripping temperatures are shown in Figure 5. The results are related to those obtained with reagent B (see Table 2). It can be seen that a prerequisite for accurate results is that the stripping temperature is maintained within the interval 100-110 °C. At higher temperatures, a complex situation arises probably because hydrolysis of the ester as well as esterification reactions takes place. The situation is further complicated since volatile components involved in the esterification equilibria are likely to be removed from the system during the stripping process. The time required to recover the water in the
Table 4. Comparison of Results Obtained for Different Oils with Continuous Coulometry Using Reagents B and Hydranal Coulomat AG-H, Stripping at 105 °C/ Continuous Coulometry and Stripping at 105 °C/Direct Potentiometry method direct coulometry type of oil
reagent B
OKES Biohydraul Raisio Biosave Statoil ES Statoil VE Statoil PA
235 ( 3 418 ( 2.5 329 ( 7 360 ( 20 93 ( 7
stripping at 105 °C
Coulomat direct AG-H continuous potentiometry 233 ( 6 417 ( 3 337 ( 6 357 ( 12 65 ( 9
236 ( 2.5 417 ( 2.5 330 ( 10 375 ( 15 92 ( 4
236 ( 6 417 ( 4 332 ( 9 370 ( 15 90 ( 5
a Sample amounts in the range 20-100 mg were used. The results given are mean values of three to five determinations expressed in ppm (w/w).
Figure 6. (a) Signal shapes for Statoil PA using stripping at different temperatures and continuous coulometric titration. (b) Signal shapes for OKES and Statoil ES using 105 °C as stripping temperature and continuous coulometric titration.
range 100-110 °C was typically 6-10 min. For Statoil VE, however, 10-15 min of stripping was needed. In separate stripping experiments using the sulfur dioxide-free reagent discussed above instead of the normal KF reagent, no signal was obtained for any of the oils added to the strip cell in the temperature range 105-150 °C. This indicates that the high results obtained at elevated temperatures are caused by water formed in esterification reactions. Normally, stripping temperatures in the range 120-150 °C are selected9 for mineral oils. As is evident from the results given in Figure 5, the use of such temperatures for the types of oils discussed in this paper will result in poor accuracy was well as poor precision. Signal shapes obtained at 105, 135, and 150 °C for Statoil PA are illustrated in Figure 6a. Response curves for
OKES and Statoil ES, using the optimal 105 °C stripping temperature, are given in Figure 6b. In Table 4 results from the stripping methods are compared with those obtained with direct coulometry. It can be seen that the agreement between the stripping technique and direct coulometry using reagent B is relatively good. The precisions obtained for OKES and Raisio are much better when compared with the other oils. It should also be observed that the precision of the results obtained with the very simple technique that combines stripping and direct potentiometry is only marginally worse than for the results obtained by using continuous coulometry for detection although the latter technique requires a much more complicated instrumentation. Typical current/time curves for two of the oils under investigation are shown in Figure 6b. For these oils 5-10 min is required in order to attain a baseline drift of e0.01 µg of water min-1. This means that the expected error in subtracting the background will be about 0.1 µg of water. For a 50 mg sample, this corresponds to 2 ppm. CONCLUSIONS Access to a sensitive coulometric system seems to offer great advantages for water determinations in various types of oil products since such instrumentation makes it possible to work with very small sample volumes. Many of the problems recently discussed in the literature can be eliminated in this way. Received for review March 19, 1997. Accepted July 1, 1997.X AC970302S
(9) Bestimmung des Wassergehaltes nach Karl Fischer. Deutsche Normen, DIN 51 777, Blatt 2, Sept 1974.
X
Abstract published in Advance ACS Abstracts, August 15, 1997.
Analytical Chemistry, Vol. 69, No. 19, October 1, 1997
4055
