ANALYST, NOVEMBER 1991, VOL. 116 1131 Potentiometric Determination of Proton Activities in Solutions Containing Hydrofluoric Acid Using Thermally Oxidized Iridium Electrodes Michael L. Hitchman and Subramaniam Ramanathan" Department of Pure and Applied Chemistry, University of Strathclyde, 295 Cathedral Street, Glasgo w GI IXL, UK A robust sensor based on iridium oxide is shown to be suitable for use in determining proton activities in solutions containing hydrofluoric acid. Fabricated by a thermal growth process, the low-impedance sensor possesses analytical utility for regulating acidity levels in etching media and pickling-baths. The thermally oxidized electrodes function satisfactorily in hydrofluoric acid solutions with concentrations up to 28 mol dm-3 and show responses comparable to electrodes prepared by the more involved method of reactive sputtering .Keywords: Iridium oxide sensor; h ydrofluoric acid monitoring; thermally oxidized electrode; direct current reactive sputtered electrode Common formulations used for etching in the microelectron- ics industry are based on hydrofluoric acid.132 For example, silicon dioxide and silicate glasses are readily attacked by HF at room temperature, whereas both single crystal and poly- crystalline silicon are wet-etched in mixtures of nitric and hydrofluoric acid. In both instances in order to control the rate of etching it is necessary to regulate the acidity of the media by the addition of buffering agents; NH4F is often used for SiOZ etching and CH3COOH for Si etching. Hydrofluoric acid based pickling- and etching-baths used in the metallurgical industries are similarly dependent on pH control for their optimum functioning3 Thus, there is a need to monitor the proton activity in both etching media and pickling-baths containing HF in order to ensure good process control.The glass electrode is, of course, the most commonly used pH probe, but it is not suitable for use in solutions containing HF because of the etching of the silica-based glass by HF. It has been found, for example, that a glass electrode does not give reliable results when the concentration of HF >5 mmol dm-3.4 This has been commonly attributed to the formation of hydrofluorosilicic acid on the glass membrane, which leads to a masking of exchange sites. The hydrogen electrode can be used reliably to monitor proton activity in media containing HF,5-7 but there is, of course, the need to equilibrate the test solution with a defined partial pressure of hydrogen and this is not very practicable for industrial processes. The quinhydrone electrode has also been used in strong HF solutions4.8~9 where no other redox couple has been present.The necessity to spike the test media with quinhy- drone in this instance is also a serious disadvantage for process control, particularly in flowing solutions. The drawbacks of the glass, hydrogen and quinhydrone electrodes have led to significant interest in the use of alternative probes based on various solid-state configurations. Among the advantages attributed to such sensors are rugged- ness, low-impedance characteristics and ease of miniaturiza- tion.l0 One example of a solid-state sensor is a semiconductor electrode, which makes use of the dependence of etch-rate kinetics on proton activity in HF solutions.11-14 A potential is applied between a semiconductor and stainless-steel electrode and the resulting current flow can be correlated with the acidity, up to about 50 mmol dm-3.Beyond this level the rapid kinetics of the etching process requires sample dilution before a measurement can be made. Another solid-state sensor that * Present address: Singapore Science Centre, Science Centre Road, Singapore 2260, Singapore. has been investigated is a palladium electrode cathodically charged to form palladium hydride.15 This electrode has been used to monitor the acidity of NH4F-HF baths.The electrodes remain viable for 1-2 weeks in this environment. For low concentration fluoride baths, a tungsten oxide electrode has been shown to operate satisfactorily as a pH probe.16 For more concentrated HF solutions, an iridium oxide electrode has been found to be particularly useful.17 However, it has to be prepared by d.c. reactive sputtering, which is not a particularly straightforward or cheap fabrication technique. Therefore, the feasibility of using a simple, readily implemen- ted, thermal method of growth for preparing iridium oxide sensors for monitoring the proton activity in solutions containing HF has been investigated. Experimental The iridium oxide electrodes were made from iridium wires of length 1 cm, diameter 0.5 mm and purity 99.9% (Goodfellow Metals).An electroactive coating was formed by heating the wire, which had previously been soaked in 2 mol dm-3 NaOH, at 800 "C in a furnace for 30 min.18 This process was repeated three times in order to obtain a uniform blue-black coating. The electrode was then cooled in air and immersed in doubly distilled water for 2 d . This step ensured that the iridate coating formed by the thermal treatment was converted into iridium oxide. A small area of the oxide coating was then scraped off at one end of the wire in order to expose the base metal and onto this a length of platinum wire (Johnson Matthey) was spot welded to form an ohmic contact. This junction and most of the platinum wire was encapsulated within heat-shrink poly(tetrafluoroethy1ene) (PTFE) tubing (Farnel Electronics) such that only the iridium oxide surface was exposed.When not in use the electrode was stored in distilled water. For the measurements involving HF a simple PTFE cell was used. The PTFE lid on the cell had two holes drilled through it to accommodate the iridium oxide electrode and a salt bridge consisting of a porous PTFE diaphragm. A saturated calomel electrode (SCE) dipping into saturated KCI contained in the salt bridge provided a long term, stable reference electrode; earlier attempts to use a porous ceramic frit for the salt bridge only gave stable reference potentials over a matter of days. The volume of the sample solution in the PTFE cell was 300 ml. All of the reagents used were of AnalaR grade and the solutions were made up in water, which had been twice1132 - - - ANALYST, NOVEMBER 1991, VOL.116 710 W 690 2 670 2 v) 2 210 I+\ \h 730 W $J 170 $ 150 v) 5 190 1 - . > E 130 \ al f110 Lu" 90 c . ' ' 650 Po q-J- -1.5 -1.3 -1.1 -0.9 -0.7 - Log[ H F] Fig. 1 Variation in potential of iridium oxide electrodes with concentration of HF. A, Thermally oxidized iridium electrode (this work); and B, d.c. reactive sputtered iridium oxide electrode, ref. 17 distilled in glass. Buffers were prepared according to standard recipes. 19 A Corning 150pH-ion meter was used for the potentio- metric measurements. It had an impedance of 1012 Q and the current drawn was typically 1 PA. A voltage output from the pH meter was fed to a Keithley Model 175 digital voltmeter, which was used for logging the potentiometric data.All experiments were performed at 25 k 0.1 "C. Results and Discussion The variation in electrode potential, of a thermally oxidized iridium electrode, with HF concentration in the range 5.6 < [HF] < 28mol dm-3 is shown in Fig. 1. The upper limit represents the concentration of HF as usually supplied, i.e., about 48% m/m. The lower limit represents a typical concen- tration of HF that might be used for etching silicon dioxide and binary and ternary silicate glasses. * Initially, on the introduc- tion of the electrode into the solution the potential drift was typically 0.5-1.0 mV min-1, but after 10 min the drift had decreased by more than one order of magnitude (see below) and this allowed an effective steady-state reading to be obtained.Thus, all of the potential values plotted in Fig. 1 were obtained after a 10min stabilization period. Clearly, a thermally oxidized electrode responds to variations in the activity of the protons. The operative equilibrium, in general, for such an electrode, has been suggested to be as follows:20 211-0~ + 2H+ + 2e- Ir203 + H20 E(mV versus SCE) = 681.0 - 59.1 pH where the pH response is conferred by the acid-base properties of a mixed-valence oxide. However, as can be seen from Fig. 1, the response in HF does not show either a Nernstian slope or, indeed, linearity. The super-Nernstian behaviour ( i e . , a slope >59mV per pH unit) cannot be attributed to variations in the concentrations of H+ with HF as expected from equilibria involving protons and HF;17 e .g . , H+ + F- e HF. Neither would contributions from liquid junction potentials be expected to have such a significant influence on electrode potential. The most likely reason for the deviations from Nernstian behaviour is that in concentrated HF, the proton activity coefficient is a complex function of ionic strength. However, although the dependence of potential on the concentration of HF is non-linear, the reproducibility for separate runs made with the same electrode over several days was good. At each concentration of HF the variation in the steady-state potential measured was never greater than k6 mV and at best it was reproducible to within about +1 mV. As far as the stability of the electrode is concerned, after the initial drift, already mentioned, of =1 mV min-1 the reading Table 1 Fundamental pH electrode parameters measured for ther- mally oxidized iridium electrodes at 25 "C in standard buffer solutions (2.0 d pH S 12.0) before and after treatment in: ( a ) 5% HF + 5% HCI; and ( b ) concentrated HF Before treatment After treatment Electrode E"'ImV (aEl3pH)lmV E"'1mV (3ElapH)lmV identifier versus SCE per pH unit versus SCE per pH unit ( a ) HFl 586.8 58.3 596.3 58.1 HF2 559.8 58.1 551.2 56.8 HF3 620.5 59 .O 547.8 56.7 HF5 583.6 59.2 552.4 54.2 ( 6 ) HF4 680.8 59.1 621.4 54.9 was stable to better than 0.1 mV min-1, and so it would be possible to monitor solutions of HF continuously over several hours without the need for re-calibration.Drift in the electrode potential was also assessed in more detail for mixed electrolytes, and is discussed below.The variation in potential measurements made for the same concentration of HF but with different electrodes could be as great as +lo-15 mV, but this variation would be consistent over the HF concentration range shown in Fig. 1; i e . , the whole curve in the plot would be shifted by the same amount. This variation between different electrodes is not unexpected and is, in fact, commonly observed for pH glass electrodes. It arises from variations in the apparent standard potentials of electrodes.21 The advantage of solid-state electrodes, such as the iridium oxide system, over glass electrodes on this matter is that close agreement between different electrodes can be achieved by a field-induced poising technique. This technique is based on the concept of the pH response of thermally oxidized iridium electrodes being dependent on the acid-base properties of a mixed-valence oxide.** By using this concept, the intervalency transitions induced by an electric field can be used to shift the Fermi level of electrons in the d-n* band to either higher or lower electronic energy levels and so allow the enhancement of the agreement between apparent standard electrode potentials (E"') in a batch of sensors.This can be important not only when sensors are freshly made, and invariably values of E"' differ, but also when they are used in particularly aggressive environments and shifts of E" ' occur as a result of the differential attack of the mixed oxide; for example, the use of electrodes in solutions containing HF.After the experiments in HF, no sign of damage to the thermal oxide coating could be detected on microscopic examination, but changes in the fundamental electrode constants, and the Nernstian slope (3EI3pH) could be seen. Table 1 summarizes the data for three electrodes for which measurements were made in a series of standard buffer solutions over the pH range 2.0-12.0 after being exposed to a mixture of 5% HF and 5% HCI, a typical metal etching solution,3 or to concentrated HF. For each, a diminution in the Nernstian slope is seen to occur in addition to a shift in the values of E"', with a greater effect being observed after exposure of the electrodes to concen- trated HF. The shift in E"' values can, as indicated above, be understood in terms of the variations in the acid-base characteristics of a mixed-valence oxide, and this can be overcome by the application of an appropriate electric field to the electrode in order to restore the original, or for that matter, any desired value of E" '.The variations in the Nernst slope might be due to the incorporation of F- ions onto the oxide sites, possibly in the form of oxy-fluoro species. This would mean that the open circuit potential would no longer be controlled solely by a proton equilibrium but might also include a component from a parallel equilibrium involving coordinated F- ions. A mixed equilibrium involving, on average, more than one electron would then give a sub- Nernstian pH dependence. Some support for this line ofANALYST, NOVEMBER 1991, VOL.116 1133 Table 2 Differences in potential for a thermally oxidized electrode (El) and a sputtered electrode (E2) as a function of the concentration of HF -Loglo[HF] -0.80 -0.90 -1.00 -1.10 -1.20 -1.30 -1.35 -1.40 -1.45 (El -E2)/mV 534 536 542 544 545 550 545 547 551 61 0 590 W 0 2 570 E 2 , 550 E G 530 I I HF1 H F2 L L 0 20 40 60 80 100 120 140 160 180 Time/h Fig. 2 Time dependence of potential of iridium oxide electrodes in 5% HF + 5% HCl. HF1 and HF2 were previously conditioned in 5% HF + 5% HCl for 1 week whereas HF3 was a newly prepared electrode reasoning is given by the observation that electrodes eventu- ally recover a near-Nernstian slope after storage in distilled water when, presumably, the F- ions adsorbed are leached out of the oxide layer.Also plotted in Fig. 1 are the results of Lauks et aZ.17 obtained with electrodes prepared by d.c. reactive sputtering. The general shapes of the two curves are very similar. As has already been indicated the absolute potential value, at any concentration of HF for a given electrode, will largely depend on the E"' for the electrode, so the significantly different electrode potentials shown in Fig. 1 for the two electrodes are probably simply a reflection of the different E"' values. If that is so then taking the difference of the two potentials at the same concentration of HF should give a constant value. Table 2 shows that this is almost true. The mean value is 543.7 k 4.6mV at the 95% confidence level and so the two electrodes show the same dependence of potential on the concentration of HF to within a few per cent.Thus, thermally oxidized electrodes clearly show comparable behaviour to d.c. reactive sputtered electrodes and the much simpler, less involved and cheaper method of preparation of the former type of electrode is obviously a distinct advantage. In addition to concentrated HF being used in the semicon- ductor industry, mixtures of HF and HCI are, as mentioned above, used for metal etching.3 For example, a typical mixture used for aluminium etching would be 5% HF and 5% HCI. The long-term stability of electrodes in this mixture was evaluated and the results are shown in Fig. 2. The electrodes are the same as those used in the investigation of the fundamental parameters reported in Table 1.The greater drift of potential with time of electrode HF3 is consistent with the significantly greater variation observed for this electrode in its electrode parameters, particularly in the E"' arising from the shift in the acid-base properties of the mixed-oxide sensor, as discussed above. Conclusions Iridium oxide electrodes can be readily made and re-gener- ated by thermal oxidation. The reproducibility of the response and stability in HF used in typical industrial environments are sufficiently good to warrant further investigation for process analysis applications. The support of an ORS award for S. R. is gratefully acknowledged. We also acknowledge financial support for this work from Ingold Messtechnik AG, and thank Drs. R. Bucher and H. Buehler of that company for the many useful discussions.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 References Kern, W., and Deckert, C. A., in Thin Film Processes, eds. Vossen, J. L., and Kern, W., Academic Press, New York, 1978, Wolf, S., and Tauber, R. N., Silicon Processing for the VLSZ Era, Volume I-Process Technology, Lattice Press, California, 1988, ch. 15. Smithells, C. J., Metals Reference Book, Butterworth, London, 1976, pp. 310 and 1490. Warren, L. J., Anal. Chim. Acta, 1971, 53, 199. Wynne-Jones, W. F. K., and Huddleston, L. J., J. Chem. SOC., 1924, 125, 1031. Jahn-Held, W., and Jellinek, K., 2. Elektrochem., 1936, 42, 401. Broene, H. H., and de Vries, T., J. Am. Chem. SOC., 1947,69, 1644. Entwistle, J., Weedon, C., and Hayes, T., Chem. Ind. (London), 1973,9,433. Farrer. H., and Rossotti, F., J . Znorg. Nucl. Chem., 1964, 26, 1959. Ives, D. J., and Janz, G. J., Reference Electrodes, Academic Press, New York, 1969, ch. 7. Turner, D. R., Anal. Chem., 1961, 33, 959. McKaveney, J. P., Anal. Lett., 1970, 3, 17. McKaveney, J. P., and Byrnes, C. J., Anal. Chem., 1970, 42, 1023. McKaveney, J. P., and Byrnes, C. J., Anal. Chem., 1972, 44, 290. Jasinki, R., J. Electrochem. SOC., 1974, 121, 1579. Veselkov, E. A., Elektrokhimiya, 1970, 6, 1701. Lauks, I., Yuen, M. F., and Dietz, T., Sens. Actuators, 1983,4, 375. Macur, R. A., US Patent., 1 348 912, 1974. CRC Handbook of Chemistry and Physics, ed. Weast, R. C., CRC Press, Boca Raton, F1, 20th edn., 1985, p. D-149. Pourbaix, M., Atlas of Electrochemical Equilibria in Aqueous Solutions, National Association of Corrosion Engineers, Hous- ton, TX, 1974, p. 397. Hitchman, M. L., and Ramanathan, S., Electroanalysis, in the press. Hitchman, M. L., and Ramanathan, S., Talanta, in the press. ch. V-I. Paper 1 I022236 Received May 13th, 1991 Accepted June 25th, 1991