Anal. Chem. 2009, 81, 3669–3675
Measurement of Parts per Million Level Gaseous Concentration of Hydrogen Sulfide by Ultraviolet Spectroscopy using 1,1,1,5,5,5-Hexafluoropentan-2,4-dione as a Derivative by Reaction of Cu(hfac)(1,5-Cyclooctadiene) J. Michael Davidson,*,† Zoe Pikramenou,‡ Adrian Ponce,§ and Richard E. P. Winpenny⊥ School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh EH9 3JJ, and School of Engineering, The University of Edinburgh, Mayfield Road, Edinburgh, EH9 3JL, United Kingdom Hydrogen sulfide reacts rapidly and quantitatively at ppm levels with cycloocta-1,5-diene-1,1,1,5,5,5-hexafluoropentan-2,4-dionatocopper(I) (Cu(hfac)(COD)) to yield 1,1,1,5,5,5-hexafluoropentan-2,4-dione (Hhfac) having a strong ultraviolet absorption at 268 nm, and which can be used without interference as a derivative in the continuous, fast online spectroscopic determination of the H2S concentration. The analytical method has excellent selectivity and can be used in the presence of N2, H2, CO, COS, SO2, moist air, CH3OH, C2H4, C6H6, and light alkanes including fuel gases. Using a standard spectrometer, the level of detection is about 10 ppb. Concentrations down to 1 ppm H2S can be analyzed readily, usually with an error in the range 0 to -7% dependent largely on the sorption characteristics of the apparatus. These are systematic errors due to adsorption; however, the analytical apparatus was very stable, and relative standard deviations as low as 0.1% of the mean can be obtained. The method can also be applied to the analysis of methanethiol. The analysis of trace levels of hydrogen sulfide in gases, of great importance in process engineering, quality control of foods, environmental chemistry, and geochemistry, is difficult because of problems connected with selectivity, adsorption of the gas within the analytical equipment, the corrosive nature of the gas, and also the sensitivity and stability of suitable analytical equipment. There is an extensive literature of the analysis of sulfides in solution, especially using colorimetric methods,1 but these are generally not applicable to gases without prior absorption of the * To whom correspondence should be addressed. E-mail: mike.davidson@ ed.ac.uk. Phone: (44)(0) 131 650 4853. † School of Engineering, The University of Edinburgh. ‡ Current address: School of Chemistry, University of Birmingham, Edgebaston, Birmingham B15 2TT, U.K. § Current Address: Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109, U.S.A. ⊥ Current address: Department of Chemistry, University of Manchester, Oxford Rd., Manchester M13 9PL, U.K. (1) Lawrence, N. S.; Davis, J.; Compton, R. G. Talanta 2000, 52, 771–784. 10.1021/ac9001035 CCC: $40.75 2009 American Chemical Society Published on Web 04/10/2009
H2S, which excludes on line analysis of gas flow systems. One variation is to immobilize the analytical reagent in a thin polymer film; the color change can then be monitored on contacting with the flowing gas.2 Generally online spectroscopic methods, solid state sensors, or electrochemical sensors are highly desirable in flow systems as opposed to the use of batch procedures. Williams has given a general review of the performance of gas sensitive semiconductor metal oxide (MOS) resistors.3 There are several approaches to the problem of H2S gas analysis: (i) Gas chromatography using a thermal conductivity detector is often the method of choice above 50 ppm; column separation gives high selectivity in multicomponent analyses, and the detector is very stable with a linear response and very little drift. The batch sampling from a flow is instantaneous, but the delay between sample injections in reporting the peak area may be considerable. If time allows replication, an accuracy of 0.1% of the concentration value can be obtained. A chemiluminescence4,5 or a flame photometric detector, especially in the pulsed form,6 is more sensitive, but the responses are highly nonlinear, calibration is difficult, and must be repeated frequently. The flame photometric detector in combination with a suitable column is particularly suitable for rough quantitative analysis of streams such as fermentation CO2 that contain H2S together with several organosulfur compounds. (ii) Advantage can be taken of equilibrium physical adsorption or chemisorption of H2S on a variety of metal oxide semiconductors, especially SnO2, copper doped SnO2,7 ZnO,8 and also WO3,9 which then undergo changes in physical and electrical (2) Wallace, K. J.; Cordero, S. R.; Tan, C. P.; Lynch, V. M.; Anslyn, E. V. Sens. Actuators, B 2007, 120, 362–367. (3) Williams, D. E. Sens. Actuators, B 1999, 57, 1–16. (4) Shearer, R. L. Anal. Chem. 1992, 64, 2192–2196. (5) Shearer, R. L.; O’Neal, R.; Dee, L.; Rios, R.; Baker, M. D. J. Chromatogr. Sci. 1990, 28, 24–28. (6) Forsyth, D. S. J. Chromatogr., A 2004, 1050, 63–68. (7) Patil, L. A.; Patil, D. R. Sens. Actuators, B 2006, 120, 316–323. (8) Kaur, M.; Bhattacharya, S.; Roy, M.; Deshpande, S. K.; Sharma, P.; Gupta, S. K.; Yakhmi, J. V. Appl. Phys. A: Mater. Sci. Process. 2007, 87, 91–96. (9) Royster, T. L.; Chatterjee, G.; Paz-Pujalt, G. R.; Marrese, C. A. Sens. Actuators, B 2001, 53, 155–162.
Analytical Chemistry, Vol. 81, No. 9, May 1, 2009
3669
properties. MOS sensors can be based on electrical conductivity,8 chemiluminescence,10 a photoacoustic effect11 or capacitance.12 The last of these (in the form of a metal-dielectricsemiconductor sensor) offers a sensitivity of 5 ppb. Such MOS sensors are now frequently used for online environmental monitoring in the field, for quality control in the food industry and for process controls and alarms, but there are significant problems in calibration, slow response, and selectivity. For multicomponent gas analysis, array sensors can be used with multivariate analysis,3,13-16 or a single sensor can be used with temperature modulation.17 (iii) Electrochemical sensors are usually three electrode cells having a liquid electrolyte contacted with the gaseous analyte through a suitably specific diffusion membrane or other restrictor.18,19 Hydrogen sulfide is oxidized at the controlled anode potential, and the cell current is limited by the rate of diffusion of H2S through the membrane into the cell. These sensors are both accurate and sensitive at the ppm level. Drawbacks are the relatively short lifetime of the cell and the need for frequent recalibration. (iv) Direct methods of spectroscopic analysis can be made continuous but present problems of selectivity and sensitivity. Hydrogen sulfide itself absorbs in the near-ultraviolet spectral region. Although the absorption maximum at 202 nm is just below the range of inexpensive laboratory instruments, in the absence of interfering substances, simple, commercially available nondispersive photometers can be used satisfactorily down to 100 ppm at a slightly longer wavelength. If there are interfering chromophores then a multivariate analysis can be carried out using an array sensor of tuned photodiodes.20 In this case, at the cost of additional complexity in computation, a multicomponent analysis is permitted. Similar principles would apply to the use of FTIR but unfortunately the vibration spectrum is very weak. (v) Spectroscopic analysis of H2S by derivatization within a film or tape has been reviewed by Wallace et al.2 Monitoring the rate of deposition of lead sulfide by reaction with H2S in a lead acetate tape has been automated in commercial instruments that remain in the market and allow analysis at the ppm level.21 A similar principle can be applied to the color change in a tape impregnated with silver.22 The procedure described below makes use of the intense ultraviolet absorption of a gaseous derivative formed continuously (10) Zhu, Y.; Shi, J.; Zhang, Z.; Zhang, C.; Zhang, X. Anal. Chem. 2002, 74, 120–124. (11) Varga, A.; Bozoki, Z.; Szakall, M.; Szabo, G. Appl. Phys. B: Laser Opt. 2006, 85, 315–321. (12) Nikolaev, I. N.; Galiev, R. R.; Litvinov, A. V.; Utochkin, Yu. A. Meas. Tech. 2004, 47, 633–636. (13) Saltzman, R. S. In Encylopedia of Analytical Chemistry; Meyers, R. A., Ed.; Wiley-Interscience: published online, 2006. (14) Vogt, F.; Dable, B.; Cramer, J.; Booksh, K. Analyst 2004, 129, 492–502. (15) Alfassi, Z. B.; Boger, Z.; Ronen, Y. Statistical Treatment of Analytical Data; Blackwell: Oxford; 2005. (16) Martens, H.; Naes, T. Multivariate Calibration; Wiley: New York; 1989. (17) Ngo, K. A.; Lauque, P.; Aguir, K. Sens. Actuators, B 2007, 124, 209–216. (18) Schiavon, G.; Zotti, G.; Toniolo, R.; Bontempelli, G. Anal. Chem. 1995, 67, 318–323. (19) www.alphasense.com. (20) Getino, J.; Ares, L.; Robla, J. I.; Horrillo, M. C.; Sayago, I.; Fernandez, M. J.; Rodrigo, J.; Gutierrez, J. Sens. Actuators, B 1999, 4, 249–254. (21) ASTM 4084-07, e.g., Analytical Systems International: Tomball, TX. (22) Hawkins, S. J.; Ratcliffe, N. M.; Sagastizabal, A. Anal. Chim. Acta 1998, 359, 125–132.
3670
Analytical Chemistry, Vol. 81, No. 9, May 1, 2009
and quantitatively from H2S by conversion of a flowing stream; it avoids complicated sample conditioning and is suitable for accurate online analysis with high sensitivity and good selectivity using only a general purpose ultraviolet spectrometer. Previously we established by gravimetry23 that Cu(hfac)(COD) (hfac ) 1,1,1,5,5,5-hexafluoropentan-2,4-dionate; COD ) 1,5cyclooctadiene) undergoes quantitative conversion into copper(I) sulfide by sulfiding in flowing gaseous H2S (0.2-500 ppm in N2) (eq 1). 2Cu(hfac)(COD) + H2S ) Cu2S + 2Hhfac + 2COD
(1)
The purpose of this paper is to establish the quantitative uptake of gaseous H2S and its conversion to 1,1,1,5,5,5-hexafluoropentan-2,4-dione (Hhfac) according to the stoichiometry of eq 1 which then allows the use of the reaction to analyze H2S by derivatization. The fast rate of this reaction, and the characteristics of the substances involved render this suitable as a method of trace analysis of H2S using Hhfac as a derivative on account of the latter’s strong ultraviolet absorption, a 1 ppm mixture giving an absorbance of about 0.01 at 268 nm in a 10 cm gas cell. A plot of its symmetrical Gaussian absorption peak is shown in Supporting Information, Figure S1. The light absorption of COD is negligible by comparison. In addition to N2, a number of other common carrier gases, (CO2, C2H6, n-C4H8, air, and fuel grade C3H8) have been used without their reaction on the packing and without spectroscopic interference; likewise important components of process streams (H2, CO, COS, SO2, CH3OH, C2H4, and C6H6) have been screened for interference variously at the 100-4000 ppm level. A few experiments were carried out using methanethiol in place of H2S and, allowing for a factor of 0.5 in the calibration constant, the reaction appears also to be quantitative. There was no marked color change in the absorbent. Thiols are used in stenching, and therefore the derivative analytical method described here should yield the concentration of -SH groups in fuel gas streams. In packed bed flow experiments, Cu(hfac)(COD) was used in the form of single crystal particles (106-300 µm), the presumption being that the large negative change of molar volume in the solids leads to a highly porous solid product (sulfide) phase. The kinetic model is therefore that of fast reaction at the surface of a shrinking non-porous core with easy diffusion through a porous external sulfide layer. The progress of the reaction front was readily monitored by the yellow to black color change and was consistent with high conversion of the solid within a narrow advancing zone. Within the gas phase a sharp front (plug flow) can be maintained by means of a sufficiently fast mass flow rate in relation to the particle size and a suitably high ratio of particle size to tube diameter. Carberry24 gives a discussion of the particle Reynolds number criterion for good plug flow. Safety Note. Because of its toxicity, the inventory of H2S in gas cylinders was held below the amount that could be tolerated in an accidental total discharge into the laboratory, being at most 70 standard mL H2S into 450 m3 of air or 0.15 (23) Davidson, J. M.; Grant, C. M.; Winpenny, R. E. P. Ind. Eng. Chem. Res. 2001, 40, 2982–2986. (24) Carberry, J. J. Chemical and Catalytic Reaction Engineering; McGraw-Hill: New York; 1976.
Figure 1. Apparatus Diagram a, H2S/carrier gas supply cylinder; b, N2 gas supply cylinder; c, diaphragm control valves; d, two way solenoid valve; e, capillary restrictor; f, on/off solenoid valve; g, derivatization flow reactor; h, two way solenoid valve; i, UV spectrometer with measurement of absorbance and cell compartment temperature; j, baratron pressure measurement; k, soap film meter; m, zinc oxide dump bed. Analysis mode: valves 1,2 open, valve 4 closed; instrument baseline mode: valves 3,4,5 open.
ppm. As a further precaution, a zinc oxide dump bed (ICI absorbent 75-1 in the form of 3 mm spheres) was placed in the outflow of the apparatus, before discharge to the fume cupboard. EXPERIMENTAL SECTION Two alternative ultraviolet spectrometers were used at 268 nm wavelength in single beam mode: a Pye-Unicam standard bench instrument (absorbance to 3 decimal places) with a 1 cm flow cell mainly for gases of >100 ppm H2S and a Perkin-Elmer Lambda 2000 instrument (absorbance in 4 decimal places) with a 10 cm flow cell for dilute gases in the lower ppm range. For calibration purposes, mixtures of Hhfac with air were first used to test the Beer’s Law behavior in static mode, that is, without flow. A 5 cm silica cell in the Pye-Unicam instrument was connected directly to a vacuum line, allowing mixtures of known partial pressures of Hhfac to be admitted. Typically a 341 ppm mixture of Hhfac in air at 265 mm Hg total pressure in a 5 cm cell (CHhfac ) 0.00487 mol/m3) gave a steady value absorbance of 0.226; nine different samples (CHhfac ) 0.01209 - 0.003747 mol/m3) gave a mean value of ε ) 9.173 with an adjusted root-mean-square deviation25 of s ) 1.3 × 10-3 (relative standard deviation, rsd ) 0.15%) (from eq 2). From the stoichiometry of eq 1, this gives a calibration factor for CH2S of f ) 0.0545/l in molH2S/m3 per unit absorbance (AHhfac) where l is the path length in cm. All calculations were carried out in terms of concentration values which were converted into parts per million fractions only for convenience (see below). AHhfac ) εCHhfacl
(2)
A linear least-squares fit25 of the same data gave ε ) 9.058 (correlation coefficient ) r ) 0.9998 with an absorbance intercept of 8 × 10-5 which is close to the level of detection of the method, confirming eq 2 with zero intercept. The measurements using the Pye-Unicam spectrometer were perfectly steady, but significant noise became evident at higher resolution in the Perkin-Elmer instrument using mixtures having less than about 90 ppm H2S. Furthermore, the absorbance values were subject to slow downward drift presumably because of adsorption onto the silica, and so initial values were used. A group of 24 calibration (25) Miller, J. C., Miller, J. N. Statistics for Analytical Chemistry, 2nd ed.; Ellis Horwood: Chichester; 1988.
points from both instruments were combined, and a leastsquares fit to all of the data gives a higher value of ε ) 9.340 (rsd ) 0.4%; r ) 0.9997) but within 2% of the Pye-Unicam value (data displayed graphically in Supporting Information, Figure S2). Given the use of two different instruments under diverse conditions, this a measure of the good reproducibility of the calibration. Cu(hfac)(COD) was prepared by heating together Cu2O and the ligands.26 After removal of the excess of the latter in vacuo crude Cu(hfac)(COD) was recovered by sublimation at 70 °C/ 0.05 mmHg pressure. Small samples were then held for long periods (several weeks) in sealed evacuated tubes at 50 °C to grow large single crystals that could be crushed and sized in the range 106-300 µm using standard sieves. Whereas crude Cu(hfac)(COD) showed signs of sensitivity to air, the purified material could be handled in air. The reactor was of glass (5 mm i.d. packed bed in downflow) with 1/8′′ Teflon tube cylinder and spectrometer connections and low volume solenoid valves to avoid dead spaces (Figure 1); these were used to switch between pure N2 and N2/H2S and between reactor and bypass. The latter was used to pass pure carrier gas through the spectrometer cell for setup purposes. Pressures were monitored using an MKS type 627 baratron (0-999 mmHg in increments of 0.1 mmHg), and flow rates were measured using a soap film meter, both situated at the outlet of the measuring cell. The temperature of the instrument compartment for the open flow cells was measured at the time of measurement while the results were found to be unaffected by the reactor temperature in the range 16-33 °C. The response time of the derivatization reactor is the residence time of the gas which was typically about 50 ms. The response times of analyses depend mainly on the ratio of the sample flow rate to the volume of the spectrometer cell together with that of the transfer line from the reactor; these were less than a minute for the 1 cm cell and less than 10 min for the 10 cm cell. Thus, the former was rapid enough for the measurement of kinetic transients in the reaction of H2S with nickel carbonates.27 Results of analyses are given in Table 1 and Table 2. Benesch et al.28 have discussed the handling of gas mixtures containing (26) Chi, K. M.; Shin, H.-K.; Hampden-Smith, M. J.; Duesler, E. N. Polyhedron 1991, 10, 2293–2299. (27) Davidson, J. M.; Glass, D. H. Ind. Eng. Chem. Res. 2007, 46, 4772–4777. (28) Benesch, R.; Haouchine, M.; Jacksier, T. Anal. Chem. 2004, 76, 7396– 7399.
Analytical Chemistry, Vol. 81, No. 9, May 1, 2009
3671
Table 1. Analysis of Hydrogen Sulfide in Various Carrier Gases and with Other Added Gases analyte composition source vessela
pressure (mmHg)
temperature (Kelvin)
carrier gas
HP HP HP BOC BOC BOC BOC BOC LP LP LP LP LP BOC LP LP LP LP LP cylinder
761 748 732 732 730 730 733 731 740 758 755 751 753 718 744 753 754 750 757 752
300 300 299 305 308 302 305 307 295 297 297 295 296 293 295 295 299 299 295 293
N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 CO2 Dry air C2H6 n-C4H10 C3H8d
LP LP
753 752
293 293
N2 C3H8d
H2S (ppm)
H2S (ppm)b from Hhfac
352 169 8.68 257 199 173 149 93.6 334 258 271 294 257 131 354 331 317 216 360 none CH3SH (ppm) 470 399
354 159 8.07c 237 194 163 141 85.5 320 246 252 294 240 131 335 316 290 193 345 0.0 CH3SH from Hhfac 440 382
interference test (ppm)
CO 4800 COS 462 SO2 326 C2H4 302 C6H6 450 CH3OH 495 H2 4600
a Source vessels: (HP) aluminum high pressure supply cylinder (∼80 atm); (LP) polypropylene low pressure supply cylinder (
