Diaphragm-Free Cell for Trace Determination of Water Based on the

to be compatible with all of the commercial reagents. (designed for ... membrane, a glass frit, or a ceramic plug in order to separate the anode and c...
7 downloads 0 Views 185KB Size
Download PDF
Anal. Chem. 1997, 69, 3100-3108

Diaphragm-Free Cell for Trace Determination of Water Based on the Karl Fischer Reaction Using Continuous Coulometric Titration Anders Cedergren* and Svante Jonsson

Department of Analytical Chemistry, Umeå University, S-901 87 Umeå, Sweden

A new type of diaphragm-free coulometric cell for continuous coulometric Karl Fischer titrations of water in the range 0.1-1000 µg is described. The relative standard deviation obtained for titrations of 1 µg amounts of water was typically 1%. The background due to diffusion of water from the air was normally in the range 0.3-0.9 µg of water min-1 depending on environmental humidity. The variation in the background was normally (0.01 µg min-1. The construction makes it possible, at any time in a sequence of titrations, to renew the catholyte by means of a Teflon plunger inside the cathode compartment. In this way, the interference effects caused by oxidizable reduction products of methyl sulfite which are formed at the cathode can be controlled in a very simple way. These products are rapidly eliminated by means of a normal titration before a new titration starts. The need for this draining step differs depending on the type of reagent used. The coulometric titration system makes use of true potentiometric end-point detection, and this principle makes it possible to control the iodine level at the end-point at much lower levels as compared with commercial instrumentation. The analytical advantages gained by this option are demonstrated for the determination of water in ethylenediamine, a task which was found to be impossible when using end-point concentrations in the range (3-7) × 10-5 M, which is typical for the bipotentiometric indicating system used in commercial instruments. Recovery rates in the range 100-102% were obtained and are shown to be dependent on the type of reagent used. The most accurate results were obtained for an imidazole-buffered methanolic reagent in which the concentration of sulfur dioxide was kept relatively low (0.10 M). The diaphragm-free cell described was shown to be compatible with all of the commercial reagents (designed for coulometry) investigated, including the wellknown Hydranal products Coulomat A, AK, AG, AG-H, and AD. Coulometric cells employed for Karl Fischer (KF) titrations normally contain a barrier (diaphragm) such as an ion-exchange membrane, a glass frit, or a ceramic plug in order to separate the anode and cathode compartments. The function of this diaphragm is to restrict the flow between anode and cathode while allowing the passage of current. In this way, the iodine generated at the anode is prevented from reaching the cathode, where it may be reduced. The diaphragm will also prevent iodine from reacting 3100 Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

with reaction products like thiosulfate, hydrogen sulfide, etc.1,2 which are known to be formed in the cathode reaction when the catholyte contains all components of the KF reagent. Commercially available coulometric systems presently make use of a separate electrolyte in the cathode compartment, a solution that does not contain any sulfur dioxide (or methyl sulfite). Scholz recommended3 an electrolyte that contains a suitable ammonium salt so that inert hydrogen is formed with high current efficiency in the cathode reaction. An alternative solution was described by the same investigator,3 who showed that, for catholytes containing a high concentration of carbon tetrachloride, very small amounts of reduction products were formed, even in the presence of 1 M sulfur dioxide, which is a typical concentration of this species in KF reagents. Except for the need to use two different electrolytes in coulometric cells containing a diaphragm, there are additional problems associated with this concept, such as the relatively long conditioning times needed before starting up as well as the lengthy cleaning procedures. The most serious problem, however, is considered to be clogging of the diaphragm, which is especially common when oils are analyzed for water. Clogging may result in a dramatic increase in the electrical resistance of the generating electrode system with subsequent electronic problems. In this context, it should be emphazised that most of the commercially available coulometric analyzers need a relative high electrolyte conductivity in order to function properly. For example, the concentration of sulfur dioxide, which in the reaction with methanol gives the highly conducting species methyl sulfite ion and hydrogen ion (or, more correctly, a protonated base), must be relatively high. In addition, the concentration of low-conducting additives like chloroform, which is frequently used to increase the solubility of long-chain organic substances, has to be restricted to a maximum of 50% for the same reason. Because of these problems, a diaphragm-free coulometric system (using pulse electrolysis) was suggested by Scholz4 and introduced to the market by Metrohm AG a few years ago. This instrument (background drift about 5 µg of water min-1) is capable of determining water in the range from 10 µg to at least 10 mg.4 At the lowest level, the results were about 6% too high, while for larger amounts of water the positive errors were less than 1%. In contrast to diaphragm-equipped cells, which contain a separate electrolyte in the cathodic chamber, a spent KF reagent (containing the easily reducible methyl sulfite) is always in contact with (1) Katoh, H.; Fujimoto, Y.; Kuwata, S. Anal. Sci. 1991, 7, 299-302. (2) Scho ¨ffski, K. Ph.D. Dissertation, Universita¨t Hannover, 1992. (3) Scholz, E. Fresenius’ J. Anal. Chem. 1994, 348, 269-271. (4) Scholz, E. Metrohm Information, 1990/3. S0003-2700(97)00034-6 CCC: $14.00

© 1997 American Chemical Society

Figure 1. Schematic diagram over the diaphragm-free cell.

the cathode in diaphragm-free systems, which means that conditions must prevail so that a minimum fraction of the current flowing through the generating electrode system leads to the formation of reduction products which are oxidizable by iodine. Katoh et al.5 found that, for current densities above 1000 mA/ 2 cm , this requirement is fulfilled2 for a number of reagents containing 1 M sulfur dioxide and bases like diethanolamine, 1,3di(4-pyridyl)propane, pyridine, and imidazole. According to this report,5 the most promising methanolic reagent was shown to be a mixture of 0.75 M diethanolamine and 0.75 M 1,3-di(4-pyridyl)propane. For such a reagent containing 1 M sulfur dioxide, the fraction of hydrogen generated at a current density of 100 mA/ cm2 (cathode surface 0.28 cm2) was as high as 97%. At 30 mA/ cm2, this figure dropped to about 75%. If the results reported by Katoh et al.5 are extrapolated to low current densities, it can be concluded that only a very small fraction of the current leads to the formation of hydrogen. The analytical consequences of these findings were demonstrated by comparison of results obtained by pulse electrolysis (current larger than 1000 mA/cm2) with those obtained with continuous electrolysis (i.e., current flowing in proportion to the polarizing voltage of the bipotentiometric indicating system). The continuous electrolysis technique was found to give about 21% too high values for samples containing 50 µg of water, as compared to +6% obtained with the pulse technique. For water amounts in the milligram range (high current densities prevailed), the positive error dropped to a few percent. The aim of the work presented in this article was to develop a diaphragm-free coulometric cell, based on the continuous coulometric principle, which can yield accurate results even when using low current densities. To achieve this goal, a new type of cell has been constructed and tested for different types of KF (5) Katoh, H.; Fujimoto, Y.; Kakuda, M. Anal. Sci. 1992, 8, 575-577.

reagents. The general applicability of the concept is demonstrated for the most well-known commercially available KF reagents for coulometry. EXPERIMENTAL SECTION Chemicals. Methanol (Merck, p.a.), iodine (resubl.), Hydranal Coulomat A (contains chloroform), Coulomat AK (for ketones; contains 2-methoxyethanol), Coulomat AG, AG-H, and AD (RiedeldeHae¨n, p.a.), imidazole (Fluka, puriss p.a.), sulfur dioxide (>99.97%), ethylenediamine (>99.5%, Fluka), and salicylic acid (BDH, p.a.) were used as received. Safety Considerations. Methanol: highly flammable; toxic by inhalation, in contact with skin, and if swallowed. Chloroform: inhalation and ingestion are harmful and may be fatal; inhalation of vapors may cause headache, nausea, vomiting, and dizziness; prolonged skin contact may result in dermatitis; liquid is readily absorbed through the skin. Imidazole: harmful by inhalation, in contact with skin, and if swallowed. Sulfur dioxide: intensely irritating to eyes and respiratory tract. 2-Methoxyethanol: causes severe irritation; combustible; harmful if inhaled or absorbed through the skin; overexposure may cause male or female reproductive disorders; exceptional health and contact hazards. Ethylenediamine: poison danger; flammable; may be fatal if swallowed; harmful if inhaled or absorbed through skin; keep away from heat, sparks, and flame; avoid breathing vapor; keep in tightly closed container. Iodine: poison danger; may be fatal if swallowed; do not get in eyes, on skin, or on clothing; do not breathe vapor; use with adequate ventilation. Reagents. Except for the Hydranal reagents described under “Chemicals”, a home-made reagent B was prepared by dissolving 0.05 M iodine, 0.15 M sulfur dioxide, 7 M imidazole, and 0.5 M salicylic acid in methanol. The effective concentration of sulfur dioxide (0.10 M) was determined iodometrically (using coulometry) after decoloring of the reagent with water. Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

3101

Figure 2. Percent fraction of the generating current which leads to the formation of reduction products oxidizable with iodine as a function of current density for different types of reagents.

Instrumentation. A schematic description of the diaphragmfree coulometric cell is shown in Figure 1. The titration vessel and the auxiliary electrode compartment were made from polymethylpentene (PTX, Mitsui Petrochemical Industries Ltd.). The auxiliary electrode was made from a platinum wire, 0.5 mm diameter. Draining of the auxiliary electrode compartment was carried out by pushing the Teflon plunger down and up. When the bottom of the Teflon plunger passes the hole on the upward stroke, the liquid level inside the tube falls to the same level as that of the liquid in main compartment, since the normal liquid level is below the hole (see Figure 1). The position of the Teflon plunger was adjusted so that 0.2-0.3 mm of the platinum wire which penetrates the plunger was in contact with the catholyte. The working area of the auxiliary electrode was then about 0.04 cm2. The Teflon plunger was sealed in the upper part of the auxiliary electrode compartment by means of an O-ring made from ethene-propene-rubber/EPDM (STEFA Ltd., England). The diameter of the opening of the auxiliary electrode compartment was 0.2 cm. The indicating electrode system used was the same as that described earlier6 (true potentiometric). The indicating electrode, a 0.5 mm platinum wire, was insulated by a Teflon tube and protruded about 0.5 cm below the seal. The reference electrode consisted of a platinum wire inside a glass tube with a Vycor glass plug fixed at the bottom by means of heat-shrink Teflon tubing. The reference electrode tube was filled with a spent KF solution containing 0.20 M iodide, 5 M imidazole, and 0.5 M (6) Cedergren, A. Anal. Chem. 1996, 68, 3679-3681.

3102

Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

sulfur dioxide, to which 1 µL of water per 5 mL was added. The drift of this low-leakage reference electrode was less than 0.1 mV h-1. The liquid level inside the reference electrode was kept about 1 cm above that of the working compartment. Refilling was normally done once every month. The redox equilibria determining the redox potential of this system probably involve a waterbuffered equilibrium between an iodine complex, hypoiodite, and iodate. This system, as well as its analytical applicability, will be discussed in a future article. The Teflon top of the cell was in contact with a holder having a circular hole and equipped with a screw which could be adjusted in order to make the contact between the Teflon top and the titration vessel as tight as possible. The generating and indicating electrodes were connected to an LKB 16300 coulometric analyzer. This instrument measures the voltage between the indicator and reference electrodes. This voltage is compared with a preset potential, and any deviation is linearly amplified and used to control the current through the generating electrode system. The way in which the end-point is approached can be adjusted with a gain (amplification factor) of the instrument. This means that, for a given gain, the current approaches asymptotically the background value during the titration. The value of the current-time integral can be followed on a display down to 1 × 10-11 equiv, which corresponds to 0.09 ng of water. The current signal from the generating electrode system, as well as the redox potential measured by the indicating electrode system, could be followed (one measurement per second

Figure 3. (a) Six consecutive titrations of a 10 µL sample containing 50.5 µg of water, followed by four draining steps using Hydranal Coulomat A. Numbers 2, 4, 6, 8, 10, and 12 correspond to titrations of the sample. D1, D2, D3, and D4 correspond to the value obtained in the respective draining step. Numbers 1, 3, 5, 7, 9, 11, 13, 14, and 15 denote the different times at which the background was checked. Values obtained (µg): 2, 51.16; 4, 52.61; 6, 52.60; 8, 53.51; 10, 53.38; 12, 55.43; D1, 110; D2, 40; D3, 16; D4, 7. Background values (µg min-1) at point 1, 0.90; 7, 1.02; 13, 1.14. End-point concentration of iodine was 1.0 × 10-5 M; amplification factor was 0.08 mA/mV. (b) Same conditions as in (a) for reagent B. Values obtained (µg): 1, 50.19; 2, 50.30; 3, 50.51; 4, 50.64; 5, 50.59; 6, 50.60; D1, 16.54; D2, 4.52. Background (0-2200 s): 0.49 ( 0.01 µg min-1.

normally) as a function of time by a Fluke 45 dual-display multimeter connected to a computer in which the data collection program Fluke QS 45 was available. Procedure for Preparation of the Coulometric Cell. The excess of iodine in the reagent under investigation was reduced to a transparent solution (containing 10-4-10-3 M iodine) by careful addition of water with a Hamilton syringe. Normally, 13 mL of this solution was then transferred to the cell using a plastic syringe (stainless steel needle). About 0.25 mL of this volume will enter the auxiliary electrode compartment. The cell was then turned upside down and carefully shaken in order to remove all water adsorbed on the walls as well as the water in the gas phase.

After this, the Teflon plunger of the auxiliary compartment was moved up and down, and the stirring motor was switched on. A suitably strong water in methanol solution was then used to remove the remaining iodine. The preset potential, corresponding to a slight excess of iodine (normally 10-5 M), was then selected and a suitable gain (normally 50-100 µA/mV) chosen. The two different types of calibration procedures used, i.e., determination of the relationship between the redox potential and the excess concentration of iodine present, were the same as described recently.6 It should be pointed out that, in situations when reagents are used that tend to be unstable for iodine concentrations above 5 × 10-5 M, the calibration procedure was always Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

3103

Table 1. Summary of Results Obtained for Different Types of Reagentsa reagent

water found, mean value (µg)

recovery (%)

rsd (%)

no. of detns

background shiftb (µg/min)

drainingc (µg of water)

Coulomat A Coulomat AG Coulomat AG-H Coulomat AKd Coulomat AD reagent B

53.11 51.41 50.89 50.75 50.90 50.47

105.1 101.8 100.8 100.4 100.7 99.9

2.55 0.80 0.40 0.17 0.32 0.33

6 6 6 4 6 6

0.24 0.24 0.22 0.01 0.01 0.01

110 37 151 42 42 16

a (50.51 ( 0.12) µg of water was added with a 10 µL Hamilton syringe. b Difference in background before and after a series of determinations (see Figure 3). c The values given correspond to the amount of oxidizable reduction products equivalent to the listed amount of water. d The effective concentration of sulfur dioxide in this reagent was found to be 0.27 M.

Figure 4. Coulometrically generated titration curve obtained for reagent B. Background-corrected generation corresponds to 0.0644 µg of water s-1. The difference between each point in the diagram corresponds to 1 s.

carried out using very low iodine excesses. Such procedures were started by selecting a preset potential corresponding to an iodine concentration in the range 10-6-10-5 M. As discussed before,7,8 one prerequisite for an accurate calibration using such low iodine levels is that the kinetics of the KF reaction are so rapid that the concentration of unreacted water can be neglected. This is the case for the reagents dealt with in this work, except for the relatively slow-reacting Coulomat AK, for which the calibration was carried out using a higher iodine level. After a few minutes, a stable background drift (about (0.01 µg of water min-1) is obtained (normally in the range of 0.3-0.9 µg of water min-1 depending on the environmental humidity) and the actual redox potential registered. The preset potential was then changed by typically 30 mV (which corresponds to about 10 times higher iodine concentration), and the current integral was determined (after correction for background) when the new and stable iodine level was reached. The relationship between the excess iodine concentration and the redox potential can then be calculated by using the value of the integrated current, the theoretical value of the Nernst slope at the actual temperature, the volume of the titration medium, and the exact change in the redox potential. (7) Cedergren, A. Anal. Chem. 1996, 68, 3682-3687. (8) Cedergren, A. Anal. Chem. 1996, 68, 784-791.

3104 Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

Quantitative Determinations of Oxidizable Reduction Products Formed at the Cathode. These experiments were carried out by means of the constant current generator of the coulometric analyzer. An iodine excess concentration in the range 0.1-0.4 mM was selected, followed by draining of the auxiliary compartment. The sample, 60 µg of water for low and 120 µg for high current densities, was then added and current generated until the original iodine level was reached. The potential value before and after a draining step was then measured and used to calculate the amount of oxidizable products formed during the current flow. The results were corrected for a blank value (less than 1 µg of water) determined in the same way without any current flowing. It should be mentioned that the reaction between the reduced species and iodine was rapid, as indicated by the stable potential value which was obtained within a few seconds. RESULTS AND DISCUSSION Determination of Oxidizable Reduction Products Formed at the Cathode. The extent of formation of oxidizable reduction products in the cathode reaction was investigated as a function of current density for a number of different reagent types, and the results are summarized in Figure 2. It should be pointed out that the current densities studied were restricted to about 400

Figure 5. Titration of 0.48 µg of water using reagent B. End-point concentration 10-7 M. The amplification factor used was 2.45 µA/mV.

mA/cm2, which is much lower than that normally used in commercial instruments based on pulse-current coulometry. The shapes of the curves at high current densities indicate that (except for Hydranal Coulomat A) the formation of reduction products tends to decrease further for very large current densities, which is in line with earlier findings reported in the literature.5,9 It can be seen that reagent B, in which the effective concentration of sulfur dioxide (0.10 M) was about one-tenth that of the commercial ones (except for Coulomat AK, 0.27 M), gave much less formation of oxidizable reduction products than others. The same is true for Coulomat AD and AG when compared with the other Hydranal products, and these results are in accord with manufacturers’ recommendations to use these reagents in combination with diaphragm-free coulometric cells. The main differences in composition between these two reagents and the others, i.e., Coulomat A, AK, and AG-H, is that Coulomat A contains chloroform, AK is based on 2-methoxyethanol instead of methanol (developed for determination of water in ketones), and AG-H contains an additive in order to facilitate the solubility of long-chain hydrocarbons like oils. A further study of the reasons for the differences in behavior between various types of KF reagents is in progress based on electrochemical techniques. Comparison of Results Obtained with Different Reagents. Figure 3 illustrates the difference in performance for the diaphragmfree cell for the expected (based on the results given in Figure 2) most troublesome reagent, Hydranal Coulomat A, and the most favorable, reagent B. Considering first the sequence of titrations carried out using Coulomat A, it can be seen that there is an increase in the base line from the initial 0.90 µg/min to 1.14 µg/ min after six injections of a 50 µg sample. In addition, results of the determinations exhibited an increasing trend, which can be explained on the basis of the increase in migration of the oxidizable reduction products forced by the larger flow of current during the titration than under conditions prevailing at the endpoint. This temporary increase in migration will not be compensated for when subtracting for the actual drift value after each (9) Dahms, H. U.S. Patent 5300 207, April 1994.

run. The large accumulation of reduction products in the catholyte is reflected by the high value obtained in the draining step (see D1 in the figure text). The relative decrease in the values for the draining steps D1 to D4 is in line with the results given in Figure 2, taking into consideration that, in order to reflect the true amount of reduced species formed, the D1 to D4 values should be reduced by the blank value caused by minute amounts of water entering the main cell compartment during the draining procedure. In addition, it must also be taken into consideration that larger diffusion/migration losses are expected for D1 because of the longer time for the electrolysis before draining and because of the higher concentration of the reduced species. As compared to Coulomat A, a much more favorable situation is obtained for reagent B, as can be seen in Figure 3b, since there is no trend toward an increase in the base line and the recovery is close to 100%. As expected, the D1 value for this sequence of titrations was very low. A summary of all results is given in Table 1. It can be concluded that, in principle, all reagents are compatible with the presented diaphragm-free cell. However, the need for including a draining step is different. Obviously, very frequent draining is necessary for Coulomat A, in contrast to that needed when using, for example, reagent B. The need for a draining step is clearly indicated when a significant increase in the background drift value takes place. Draining is recommended as soon as the drift increase exceeds 0.1 µg of water min-1 as compared to the value noted before a sequence of determinations is started. Current studies on modified KF reagents indicate that a further reduction in the formation of the oxidizable reduction products is possible, even for reagents in which the concentration of sulfur dioxide is higher than that in reagent B. Choice of Titration Conditions. A titration curve obtained by constant current generation of iodine (background-corrected generation 0.003 58 µmol s-1) using reagent B is shown in Figure 4. The choice of end-point concentration of iodine is primarily governed by the type of sample to be analyzed as well as the amount of water in the sample. In Figure 4, three different endAnalytical Chemistry, Vol. 69, No. 15, August 1, 1997

3105

Table 2. Theoretically Calculated Times (s) for 99.9% Reaction for Different Types of Methanolic KF Reagents and End-Point Concentration of Iodine Based on Published Kinetic Data7,a iodine end-point concentration reagent 1 M pyridine, 0.77 M SO2, 0.26 M 5 M pyridine, 1.37 M SO2, 0.26 M I5 M pyridine, 1.37 M SO2, 0.26 M I-, 30% CHCl3 1 M imidazole, 0.75 M SO2, 0.75 M SO2, 0.20 M I1 M imidazole, 0.75 M SO2, 0.20 M I-, 30% CHCl3 2 M imidazole, 0.85 M SO2, 0.20 M I2 M imidazole, 0.85 M SO2, 0.20 M I-, 30% CHCl3 5 M imidazole, 0.80 M SO2, 0.20 M I5 M imidazole, 0.82 M SO2, 0.20 M I-, 30% CHCl3 I-

a

10-4 M

10-5 M

10-6 M

10-7 M

10-8 M

5 × 10-10 M

74 23 9 35 18 8 3 14 14

7.4 × 2.3 × 102 88 3.5 × 102 1.8 × 102 32 16 17 17

7.4 × 7.3 × 102 3.1 × 102 2.2 × 103 9.4 × 102 1.9 × 102 40 1.4 × 102 69

4.6 × 103 2.3 × 103 3.5 × 104 1.2 × 104 2.6 × 102 1.4 × 102 87 43

6.9 × 102 6.3 × 102 1.4 × 102 69

5.3 × 102 16

102

103

The values for the concentrations of sulfur dioxide were obtained iodometrically after decoloring the reagents with water.

Figure 6. Titration of water in transformer oil using Hydranal Coulomat AG-H. End-point concentration: 7.5 × 10-6 M. The amplification factor used was 0.04 mA/mV. Background before addition of the oil sample was 0.77 µg min-1 and from 9 to 16 min was 0.81 ( 0.01 µg min-1. The latter value was used in evaluation of the result. The value reported by a control laboratory for this oil sample was 35 ( 3 ppm (w/w).

point levels are marked by the letters A, B, and C. For samples containing medium to large amounts of water, level A should be used since the titration can be carried out at a high speed (high gain) without problems in the concentration range 10-5-10-4 M iodine. The slope of the titration curve at this end-point is, according to the Nernst law,

dE/d[I2] ) RT/(2F[I2]) mV M-1

which corresponds to 2 mV/µg of water (cell volume 12.7 mL). A typical variation in the end-point potential at this iodine level is (0.02 mV, which is equal to (2 µA for a gain corresponding to 100 µA/mV. The time required for a titration to be completed at this end-point level is about 30 s, which means that the standard deviation of the results is expected to be about 0.005 µg. For trace determination of water in samples which interfere with the iodine of the KF reagent, a lowering of the end-point concentration (point B) may reduce or even completely eliminate the interference effects, as will be exemplified below under Applications. A lowering of the end-point concentration will lead to less overcorrection when subtracting for the background. The 3106

Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

reason for this is that the rate of such side reactions is slower during the titration compared to the situation in the end-point. It should be pointed out that, at this iodine level, relatively long titration times are required since a low gain has to be selected in order to avoid overshoot. A titration of 0.48 of µg water at 10-7 M using reagent B is shown in Figure 5. The standard deviation of the current values (50 values, one per second, before and 50 values after the peak) was found to be 0.80 µA, which is equivalent to 0.0044 µg of water min-1. Because of the relative long time required for this titration (200 s), the expected smallest amount of water which can be determined with a relative standard deviation of 10-15% is about 0.1 µg (product of time and variation in background, 0.015 µg). However, the long titration time in this case is due to slow kinetics of the KF reaction, since the titration time was found to be about the same as that theoretically calculated (194 s for 99.9% based on a determined rate constant at this iodine level of 3.5 × 106 M-2 s-1). Titrations using the end-point C in Figure 4 are preferred for the same reasons as for the selection of point B but should be used for titration of water amounts larger than 1 µg since the speed of the titration can be relatively high without the risk for overshoot.

Figure 7. Comparison of titration with Hydranal Coulomat AG carried out using different end-point concentrations for 5 µL of ethylenediamine. End-points used: (a) 7.5 × 10-5 and (b) 6.7 × 10-7 M.

The precision for the titrations carried out at this extremely low iodine level will be affected by the slower kinetics of the KF reaction as well as the slower attainment of equilibrium in the indicating electrode system. Conditions for titrations using a number of different types of reagents are given in Table 2. The calculations were based on reported kinetic data7 (t99.9% ) ln 0.001/(k[I2][SO2]), where k is the rate constant determined for the KF reaction at the actual iodine concentration). As can be seen in Table 2, a rapid reaction at the end-point is expected at high concentrations for all reagents except for the “classical” 1 M pyridine medium. However, if we consider the 10-7 M level, it can be seen that only the imidazolecontaining reagents are capable of completing the reaction within a reasonably short time. As was shown recently,7 a remarkable improvement results from the presence of chloroform, which makes it possible to carry out rapid titrations, even at end-points in the range 10-9-10-10 M iodine. Applications. The applicability of the diaphragm-free cell for trace determination of water in a transformer oil using the recommended reagent Coulomat AG-H is shown in Figure 6. As can be seen, 100 mg samples are large enough to permit determinations in the low ppm range. It can be concluded from a recent paper by Margolis12 that the use of small sample volumes should be favorable in view of the problems noted when the oil is not completely dissolved in the titration vessel solution. In such cases, the oil is capable of binding or sequestering a portion of the water so that the moisture is unable to react with the Karl Fischer reagent in an optimum way. The described coulometric system uses a feed-back system based on true potentiometry, which offers the possibility of controlling the end-point concentration at much lower iodine levels compared with what is possible with those pertaining to com-

mercial coulometric instruments, which are almost exclusively based on the bipotentiometric technique. This option is of special importance when the sample contains substances which interact with iodine like, for example, active carbonyl compounds (cyclohexanone)11 and certain pharmaceutical products.13,14 In situations where the KF reagent tends to be unstable at higher iodine concentrations, this possibility can be used to lower the background drift. An illustrative example is given in Figure 7, where a 5 µL sample containing ethylenediamine (EDA) is titrated with Coulomat AG using two different end-point concentrations. The end-point concentration of iodine used in the experiment, the results of which are given in Figure 7a, corresponds to that used in the Metrohm Model 737 (selection, high end-point concentration). As is evident from the figure, the side reaction between iodine and the EDA is too rapid to permit an evaluation of the amount of water in the sample. On the other hand, if the titration is carried out using a much lower end-point concentration of iodine, discrimination of the side reaction is possible. In separate experiments, the kinetics for the reaction between iodine and EDA were investigated in the iodine concentration range 0.1-100 µM. The reaction was found to be first order with respect to iodine (rate constant determined in the concentration range 10-7-10-4 M iodine was found to be 2.7 M-1 s-1, assuming that the reaction is also first order in EDA). The discrimination seen in the figure can then be explained on the basis of earlier findings reported in this journal,7 namely that the rate constant for the KF reaction increases dramatically at lower iodine levels for reagents buffered with imidazole (which is the base used in Coulomat AG12).

(10) Cedergren, A.; Ora¨dd, C. Anal. Chem. 1994, 66, 2010-2016. (11) Ora¨dd, C.; Cedergren, A. Anal. Chem. 1995, 67, 999-1004. (12) Margolis, S. A. Anal. Chem. 1995, 67, 4239-4246.

(13) Lindquist, J. J. Pharm. Biomed. Anal. 1984, 2, 37-44. (14) Kågevall, I.; Åstro ¨m, O.; Cedergren, A. Anal. Chim. Acta 1981, 132, 215218.

CONCLUSIONS The described diaphragm-free cell has been shown to be compatible with all types of reagent mixtures tested. Thanks to

Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

3107

the high sensitivity of the system, sample volumes in the range microliters to milliliters are sufficiently large to cover the concentration range sub-ppm to 100%. It should be emphasized that this concentration range can also be handled by commercial coulometric instruments, but then much larger sample volumes are required. This is a serious drawback not only for trace determinations but also in situations where the samples contain interfering substances. The design of the system restricts the generating current to about 20 mA (100 µg of water min-1) which means that the amount of water which can be determined rapidly covers the range 0.1 µg to a few hundred micrograms. The described cell construction has been shown to be very rigid, and

3108

Analytical Chemistry, Vol. 69, No. 15, August 1, 1997

no major problems have been observed during a test period of one-and-a-half years, including several thousand experiments. ACKNOWLEDGMENT The authors thank Mr. Lars Lundmark for valuable technical support, Dr. Katrin Scho¨ffski, Riedel de-Hae¨n, for reading the manuscript and for kind support of chemicals, and Dr. Michael Sharp for linguistic revision of the manuscript. Received for review January 10, 1997. Accepted April 15, 1997.X AC970034Y X

Abstract published in Advance ACS Abstracts, June 1, 1997.