Continuous voltammetric monitoring of hydrogen and oxygen in water

pursued a similar route in the development of a system for the continuous monitoring of dissolved hjrdrogen as well as oxygen in water containing both...
0 downloads 0 Views 528KB Size
Anal. Chem. 1082, 54, 1651-1654

1851

Continuous Voltammetric Monitoring of Hydrogen and Oxygen in Water L. W. Niedrach" and UY. H. Stoddard General Electric Company, Corporate Research1 and Development, P.0. Box 8, Schenectady, New York 12301

Srinivasan and Tarcy described a sensor for hydrogen which they prepared by modifyiry: a Beckman oxygen sensor (1). We pursued a similar route in the developmeint of a system for the continuous monitoring of dissolved hydrogen as well as oxygen in water containing both gases. For the hydrogen we made relatively minor changes to the sensor and electronics of a Beckman Type 7002 oxygen monitor. Unlike the previous authors we did not find it necessary to coat the Teflon diffusion barrier with palladium; rather we found that we could prevent interference frorn dissolved oxygen by appropriate biasing of the sensing electrode potential. For dissolved oxygen we use unmodified versions of the Beckman Type 7001 and 7002 monitors. In addlition to the sensing units our system also provides for in situ calibration of both sensors. In the discussion that follows emphasis is placed upon the hydrogen sensor and the calibration system, but performance data are included for the) complete system. It is also to be noted that while the focw has been on instrumentation from a specific manufacturer, the modifications are general and should be applicable to other analogous systems.

EXPERIMENTAL SECTION Equipment. The Hydrogen Sensor. A Beckman oxygen sensor (Part No. 190403) has been used as the basic element. It is modified, however, by coating the gold sensing electrode with platinum black and by forming a coating of silver chloride on the counterelectrode. To maintain a relatively constant pH throughout the life of the sensor, we emplloyed a substitute electrolyte, 0.1 M KCl-O.1 M HCl, in place of the 5% (-0.7 M) KC1 solution used with the oxygen sensor. Apart from these changes the sensor is identical with that for oxygen, and Beckman's recommendations for membrane installation and electrolyte replacement apply (2). The first step in modifying the basic unit iis to form the silver chloride on the countereleclmde. In preparation for this step the cap and the Teflon membrane assembly are removed and the original filling solution is diiscarded. A pin connector or suitable piece of wire is then used to make a connection to position B on the socket. Fresh fiiing solution (the 5% KC1 solution provided by Beckman for filling the oxygen sensor) is then added to refill the electrolyte chamber almost completely. With the unit mounted vertically in a clamp, a platinum wire (10-20 mils is a convenient weight) with a loop of sufficient size to encircle the post containing the sensing electrode is inserted down into the KCl solution;cf. Figure 1. Connections are then made to a small d.c. power supply with the wire from position B going to the positive terminal. A current of about 100 mA is maintained for about 45 min; then the electrolyteis replaced with a fresh portion and the electrolysis is continued for another 45 min. After completion of the electrolysis the elecfxolyte is discarded and the chamber is rinsed with distilled water. The sensor is then ready for platinization of the sensing electrode. For this purpose we have used a solution containing 3 g of chloroplatinicacid and 0.03 g of lead acetate dissolved in 100 mL of water as recommended by Reilly and Rae (3). Sufficient solution is placed in a glass container such that the gold sensing electrode can be immersed to a depth of about 1mm. A small magnetic stirring bar and a platinum wire counterelectrode for use in the electrolysis are also placed in the container. Our arrangement is shown in Figure 2. A pin or wire is then used to make connection between position A of the socket and the power supply. The platinum counterelectrode goes to the other terminal of the power supply. While the solution is gently stirred, electrolysis is performed at a current of 100 mA for 30 s with the gold sensing electrode negative. The current it3 then reversed for about 30 s, and, finally, the current is reversed again to repeat 0003-2700/82/0354-1651$01.25/0

the first step. At this point an even coating of platinum black is evident on the electrode. After gentle rinsing with water, the sensor is reassembled with a fresh Teflon membrane and the 0.1 M KC14.1 M HCl electrolyte using the Beckman procedure for rebuilding an oxygen sensor. Instrumentation for the Hydrogen Sensor. A Beckman Model 7002 oxygen monitor can be used with minor modifications. Connection of the sensor cable to the monitor is identical with that for the oxygen sensor with a single exception-THE CONNECTIONS FOR THE SENSING AND COUNTERELECTRODES ARE INVERTED FROM NORMAL. This enables existing circuitry in the monitor to bias the sensing electrode positive instead of negative with respect to the counterelectrode, a necessity to accomplish hydrogen oxidation. It also results in the meter needle deflecting in the proper direction even though the output current from the sensor is the reverse of that from the oxygen sensor. The circuitry in the Beckman unit requires a minor change to permit achieving the desired bias of +550 mV for the sensing electrode w. the counterelectrode. We have done this by shunting resistance R19 (cf. circuit diagram in operating manual for the 7002 monitor (2))with a 10 kQ resistor. The bias can then be adjusted with potentiometer R18 until a value of 550 mV is measured between the anode and cathode at the terminal board (with a sensor in place). Although we used a digital voltmeter connected in parallel with the analog meter on the 7002 monitor for our measurements, a minor change in the circuitry permits one to obtain a useful scale reading on the analog meter itself. This consists of shunting resistance R14 with a 1kQ resistance and setting the range selector switch to position 2. One can then use the calibration knob to adjust so that a full scale deflection is equivalent to 200 ppb of dissolved hydrogen. We have made no effort to modify the temperature compensating thermistors. Calibration Unit and System Layout. For the calibration of oxygen sensors, Beckman suggests a procedure involving exposure of the sensing electrode to ambient air (2). Alternatively calibration with a grab sample analyzed for oxygen by the Winkler method is suggested (2). Neither of these procedures is suitable for calibration of a hydrogen sensor. As an alternative, we equilibrate gases of known composition with water which is then pumped over the sensor. (This approach would seem a useful alternative for the calibration of oxygen sensors as well. It has the virtue over the Beckman method of exposure to air of not requiring removal of the sensor from the system. By proper choice of the calibrating gas it also avoids exposure of the sensor to grossly atypical concentration levels and hence large extrapolations from the calibration point to the operating point. For example, when monitoring oxygen at the 200 ppb level, calibration by exposure to air uses a calibration point equivalent to approximately 8000 ppb. Finally, our approach provides for a simple zero check by equilibrating the water with nitrogen. This is a useful feature because we have found that zero drift can be a cause of malperformance of the oxygen sensor, often because of the deposition of silver dendrites on the sensing electrode.) Operation of the system is best considered in connection with Figure 3, which shows a schematic diagram. All elements of the system are mounted on a single rack assembly. The hydrogen sensor and an unmodified oxygen sensor, both mounted in standard flow chambers, are connected in series. In order to assure adequate mass transport in the water at the sensing electrode interface, we maintained a metered flow rate of -500 mL/min with a gear pump. This circulation rate is independent of the sampling rate from the system being monitored. For calibration the sensors can be connected to the equilibration column rather than the system being monitored. In our unit the column consists of a 4 ft length of 1.0 in. 0.d. glass tubing. A coarse @ 1982 American Chemical Society

1652

ANALYTICAL CHEMISTRY, VOL. 54, NO. 9, AUGUST 1982

800

I-

600

c

> E

KCL SOLUTION

4000

Aa COUNTER

I-

::

20oc

W E

I l

ot 0

m

Figure 1. Arrangement for depositing AgCl on the counterelectrode.

-400 -600

I

LEAD FROM TERMINAL TO POWER

A

SUPPLY

I

I

-3

-2

I

1

1

-I 0 I 2 C U R R E N T - ARBITRARY UNITS

1 3

I 4

Figure 4. Current-voltage curves for a modified sensor exposed to hydrogen, oxygen, and nitrogen: (0)hydrogen, (0) oxygen, (A)nitrogen.

SENSING ELECTRODE

Pt SOLUTION MAGNETIC

STIRRER Figure 2. Arrangement for platlnlzlng the sensing electrode. RETURN T O LOOP

FLOII'METER

VCLVE CALI BR ATIhG

GCSES

pling line terminating in Swagelok fittings. Flow in this sampling line is not believed to be critical and can probably vary from rates as low as 5C-75 mL/min up to several hundred mL/min or higher. The pump in the analytical arm maintains an adequate flow rate over the sensors regardless of the flow in the sampling arm. As noted above, we have been using a circulation rate of about 500 mL/min over the sensors. When calibration is desired, the two three-way valves in Figure 3 are adjusted to circulate water from the equilibration column over the sensors. We use 10% hydrogen in nitrogen and 0.5% oxygen in argon as our calibration gases and nitrogen to check the background current. It is also feasible-though slightly less reliable-to check the background current of the oxygen sensor while equilibrating with the 10% hydrogen, and that of the hydrogen sensor while equilibrating with the 0.5% oxygen. It is to be noted that calibration is performed with the range switch on the monitor set to read in ppb (range 2 for the system modified to monitor hydrogen) not in position 4 as when calibrating an oxygen sensor by exposure to ambient air. To calculate the concentration of hydrogen (oxygen) dissolved in the water from the equilibration column, we use solubility data from Himmelblau ( 4 ) . Knowing the barometric pressure and the temperature of the water in the equilibration column, one can derive .the partial pressure of the gas of interest. P = (atm pres - vap pres H 2 0 + P,) X percent in equ Gas where P, is the head of water in the equilibration column. Then

Figure 3. Schematic diagram of the system.

glass frit is employed to disperse the gas about 8 in. above the water outlet in the bottom of the column. Water is drawn from the bottom of the column and passes through the pump, the sensors, and a flowmeter before it returns to the top of the column. Provision has been made for adding makeup water through a small funnel which also serves as a level control. Because of the moderately high temperature coefficient of solubility of hydrogen and oxygen in water we have jacketed our equilibration column with a 1.5 in. 0.d. glass tube through which water thermostated at 25 "C is flowed. While the thermostating was desirable in developing and studying the system, we do not consider it a necessary refinement for field use. When doing a calibration, however, it is important that the temperature of the water be known so that the proper solubility value can be used in calculating the gas concentration in the water with the aid of solubility data in the literature. A different correction for the vapor pressure of water is also required at each temperature. Operation of the System. For background data on the performance of the system, the equilibration column was connected to the sensor at all times while the equilibrating gases and the flow rates were changed. During normal use in monitoring dissolved gas concentration in process streams, the measuring system is connected to the system being monitored with a sam-

where K is the appropriate solubility coefficient from ref 4.

RESULTS AND DISCUSSION Effect of Bias Potential on the Response of the Modified Sensor. Experiments were first performed to determine the effect of the bias potential on the response of the sensor to hydrogen in an effort to find a range in which interference, particularly from dissolved oxygen, would be minimized or eliminated. These were performed by using the recommended water circulation rate for the oxygen sensor of 500 mL/min. A gas flow rate of 300 mL/min was used in the equilibration tower. The operating temperature was held at 25 "C and the head of water in the equilibrator was maintained at 24 in. (=45 torr). Early experiments had indicated that the gold electrode and a bright platinum electrode are unsatisfactory for sensing hydrogen. The platinum black electrode, however, showed an excellent response to hydrogen over a wide range of bias potentials; cf. Figure 4. Also shown in Figure 4 are response curves obtained with air and with nitrogen flowing through the equilibration tower.

ANALYTICAL CHEMISTRY, VOL. 54, NO. 9, AUGUST 1982

1653

z Y 0

(L

a W

I I I

I J 600 800 1000 GAS FLOW RATE TO EQUILIBRATOR m l / m i n

0

i L 2 0 400 600

Flgure 5. Effect of water ciirculation rate on sensor response: (0) oxygen sensor, (A)hydrogein sensor.

It is evident that over much of the potential range for hydrogen oxidation, oxygen was reduced arid would therefore interfere. At a potential of about 500 mV vs. the counterelectrode, the oxygen reduction became negligible, but at potentials above 600 mV, a significant oxidation current was obtained. A similar oxidation current was obtained when the circulating water was equilibrated with nitrolgen. We interpret this as indicating that the platinum black was being oxidized according to the reaction

+ 4C1- = PtCld- + 2e-

400

Flgure 6. Effect of gas flow rate In equilibrator on sensor response: (0)oxygen sensor with air, (A)hydrogen sensor with forming gas.

2OC 800 1000 WATER C I R C J L A T I O N R A T E m l l m i n

Pt

200

(1)

which has a standard potential of 730 m'V (European sign convention) (5). On the basis of these results we chose 550 mV vs. the counterelectrode as the preferred bias potential for the hydrogen sensor. Effect of the Water Circulation Rate on the Response of the Sensor. The effect of the water circulation rate on both the hydrogen and the unmodified oxygen sensors is illustrated by the data in Figure 5. Clearly the hydrogen sensor is slightly more sensitive to the flow than the oxygen sensor. This undoubted1.y stems from the fact that the permeation coefficient for hydrogen through 'Teflon (the membrane material) is hgher for hydrogen than oxygen (6),thereby resulting in a larger concentration drop in the boundary layer a t equivalent flow rates, i.e., a larger "stirring error". On the basis of these data and the fact that our system operated more stably at 5100 mL/min than higher circulation rates, we chose this as ouI standard rate. Since the flow rate is monitored and the same value is used during normal operation and calibration, it is not necessary that the rate be such as to be on the response plateau. In future modifications, however, it would seem desirable to operate with somewhat higher circulation rates. Effect of Gas Flow R&e in the Equilibrator. We have also examined the effect (of the gas flow on the steady-state performance. This is illustrated by the datri in Figure 6. On the basis of these data we adopted a gas flow rate of 300 mL/min as standard. Validity of the Calibiration Procedure. To check the validity of the proposed calibration procedure an oxygen sensor was calibrated by 13eckman's suggested procedure of exposing the sensor to air. The conditions summarized in Table I prevailed. With the monitor set in the calibration mode (selector switch in position 4) the barometric pressure reading was set on the output meter after correcting for the relative humidity of the air. The latter was measured with

Table I. Evaluation of Calibration Procedure ambient temp (dry bulb temp) wet bulb temp re1 humidity partial pressure of water in air barometric pressure corrected pressure equilibrium 0, concn in water under above conditions (4) concn from monitor calibrated on air difference equilibrium 0, concn in water at 752 torr total pressure, T = 25 "C, and a 24 in. head of water (=45 torr) concn from monitor calibrated on air difference

21.5 " C 15.0 "C 58% 11 torr 752 torr 741 torr 8.79 ppm 8.47 pprn 0.32 ppm (3.6%) 8.65 ppm 8.62 ppm 0.03 ppm (0.3%)

a Psychro-dyne Psychrometer (Environmental Tectonics Corp., Southampton, PA, Model 22012). The monitor was then switched to the 0-10 ppm range position with the sensor still exposed to the air. Under this condition the meter should then have given a reading corresponding to the dissolved oxygen concentration in water and in equilibrium with air under the existing conditions of pressure and temperature. The equilibrium value calculated after Himmelblau (4)is indicated in Table I. The measured value although 3.6% low, is in satisfactory agreement. The sensor was then reinserted into the test unit and exposed to the circulating water which was being equilibrated with air while the temperature was being maintained at 25 "C and the excess pressure wm held at 24 in. of water (45 torr). Under these conditions one calculates a dissolved oxygen concentration of 8.65 ppm after Himmelblau. This, in turn, compares with the measured value of 8.62 ppm based upon the original calibration on ambient air. In examining the data in Table I it is evident that although all of the values are not in perfect agreement, they do correspond within about 2-3 %. This is considered extremely satisfactory and strongly supportive of the proposed approach to calibration. Range of Linear Response to Hydrogen. Having established the operating conditions at a bias of 550 mV, a water circulation rate of 500 mL/min and a gas flow rate in the equilibrator of 300 mL/min, we investigated the response of the hydrogen sensor over the range of concentration from 1.5 to 1500 ppb dissolved hydrogen. For this purpose pure hydrogen, standard mixtures of 1.0 and 10.0% hydrogen in nitrogen, and further dilutions of the latter two with nitrogen were employed. The 10% mixture was used to calibrate the scale of the sensor readout in terms of parts per billion of

1654

Anal. Chem. 1982, 54, 1654-1659 22Or

Flgure 7. Responses of hydrogen and oxygen sensors upon exposure to water equilibrated with various admixtures of formlng gas and 0.5% 02-Ar: (0)oxygen sensor, (0) hydrogen sensor, (A)hydrogen sensor with solutions contalning -2000 ppb oxygen; open points obtained with lncreaslng formlng gas content, closed polnts with decreaslng formlng gas content; crosses Indicate callbration points. i l

- 4 J O%Fi

-

s;

kE l,FFPiMING 5

-I p;t;yloN 2

L

75IFG

666hFG 100%FG

L

7

i

L

GPS

1

i_

J

SENSOR1

12 u

i _

Flgure 8. Traclngs showing response times in obtaining data of Figure 7.

dissolved hydrogen. On the basis of 14 data points, the intercept and slope of the linear regression of sensor readout against actual concentration were, with 95% confidence, 1.009 f 0.807 ppb and 0.999 f 0.002, respectively. The correlation coefficient was 0.999, Additional Performance Data. Additional data have been obtained to illustrate the response of the system over a range of hydrogen and oxygen concentrations. Figure 7

shows data for both a hydrogen sensor and an unmodified oxygen sensor when different mixtures of 10% hydrogen in nitrogen and 0.5% oxygen in argon were passed through the equilibration column. At the start of the experiment the output meters were calibrated for each sensor by equilibrating the water with the individual gas mixtures. It is of interest to note that the background readings on the opposing sensors were very low-1-2 ppb for the hydrogen sensor on the 0,-Ar mix and 4-5 ppb for the oxygen sensor on the Hz-Nz mix. Upon equilibration with intermediate concentrations of the two active gases the anticipated linear variation associated with simple dilution was observed. Clearly there is no sign of any interaction of the hydrogen and oxygen under any circumstances. As a further illustration, two additional points are included in Figure 7. They were obtained by using mixtures of forming gas, air, and nitrogen such that the oxygen content of the water was maintained a t 2000 ppb. Even a t this elevated concentration of oxygen no effect is evident on the hydrogen response. Data in Figure 8 provide an indication of the response rate of these sensors. The data were obtained in connection with those appearing in Figure 7. We have also cycled the system between the H2-N2 and 02-Ar gases employing 6- and 12-h periods over several days with excellent stability and reproducibility and the unit has functioned well for over a month of continuous operation in a test loop for monitoring the effects of water chemistry on corrosion films on steel. In this case solenoid valves and a timer were used to alternate measurements between streams entering and leaving the autoclave.

ACKNOWLEDGMENT We are grateful to John Steadwell and Craig Palmer for the use of an oxygen monitoring system which they had designed and assembled for another application. This unit served as the foundation for our hydrogen/oxygen monitoring system and its associated calibration unit. LITERATURE CITED (1) Srinivasan, V. S.;Tarcy, G. P. Anal. Chem. 1881, 53, 928-929. (2) “Beckman Instructions 015-555321 for Model 7002 Oxygen Monitor”; Beckman Instruments, Inc.: Fullerton CA, 1976. (3) Rellly, R.; Rae, W. N. “Physlco-chemlcal Methods. Vol. 11”, 5th ed.; Van Nostrand: New York, 1954; p 608. (4) Hlmmelblau, D. M. J. Chem. Eng. Data 1860, 5, 10-15. (5) Latimer, W. M. “Oxldatlon Potentials”, 2nd ed.; Prentlce-Hall: New York, 1952; p 206. (6) “Modern Plastlcs Encyclopedia”; McGraw-HIII: New York. 1981; Vol. 58, p. 541.

RECEIVFD for review March 10,1982. Accepted April 29,1982.

Long Term Stabllity of Spectral Interference Calibrations for Inductively Coupled Plasma Atomic Emission Spectrometry Robert I. Botto Exxon Research and Englneerlng Company, Baytown Research and Development Divlsion, Analytlcal Research Laboratory, Baytown, Texas 77520

Spectral interferences may be seriously detrimental to the accuracy and reliability of trace element determinations by inductively coupled plasma atomic emission spectrometry (ICPAES), particularly when they are performed in the presence of high concentrations of matrix constituents: Si, Al, Fe, Mg, Ca, Ni, etc. Trace characterizations of geological, metallurgical, and marine samples by ICPAES are often hampered by severe spectral interference problems (1-4). Spectral interferences arise from direct spectral line overlaps (5-7) and background shifts due to recombination continuum 0003-2700/82/0354- 1654$01.25/0

emission (8))molecular bands (9))stray light effects (IO), and overlap with nearby broadened line wings (8,I.l). Corrections for background shifts are often accomplished by making off-line background measurements (12). Corrections for spectral line overlaps are usually made by using experimentally determined correction factors according to the method of Boumans (13). Multielement ICP emission spectrometers using a fixed array of detectors (polychromators) lack the flexibility of scanning instruments for avoiding spectral interferences 0 1982 American Chemical Soclety