150
Anal. Chem. 1980, 52, 150-153
Determination of Trace Quantities of Dimethylsulfoxide in Aqueous Solutions Meinrat 0. Andreae Department of Oceanography, Florida State University, Tallahassee, Florida 32306
A method is described for the determination of dimethylsulfoxide at the nanogram level in aqueous solutions. After removal of dissolved volatile sulfur compounds by sparging with helium, dimethylsulfoxide is reduced to dimethylsulfide with sodium borohydride or chromium(11) chloride. The dimethylsulfide is stripped from the solution with a stream of helium, and trapped on a chromatographic packing in a U-tube immersed in liquid nitrogen. Afler gas chromatographic separation from potentlally interfering compounds, the dimethylsulfide Is measured by a flame photometric sulfur detector and/or a flame ionization detector. The detection limit of the method is 1 ng S(DMS0) per sample, corresponding to 0.02 pg S(DMS0) L-'. Precision is 5-lo%, dependlng which varlant of the method is used. Using this procedure, dknethylsutfoxide has been found in a variety of natural waters.
Dimethylsulfoxide is produced industrially in large quantities for applications as industrial solvent ( I ) . It has been used as a solvent for promoting chemical reactions, and as a vehicle for dyes, drugs, steroids and agricultural toxicants, as preservative, radioprotective agent, lubricant, and stabilizer for sulfur trioxide. I t has been suggested for a number of medical uses, including the treatment of acute musculosceletal inflammation ( 2 ) . T h e occurrence of DMSO, together with other sulfoxides and dimethylsulfide, as natural product in plants and animals has been shown by a number of authors ( 3 , 4 ,and references therein). Its distribution in natural waters is correlated with biological activity, in particular primary productivity (Andreae, unpublished data). Its occurrence in rain suggests t h a t it is involved in the atmospheric cycle of sulfur. T h e analytical chemistry of DMSO has been summarized by a number of authors (5,6,and references therein). Various chromatographic methods, e.g., liquid, gas, and thin-layer chromatography, have been employed for the determination of DMSO, mostly a t relatively high concentrations and in nonaqueous solvents. For the quantitative analysis of DMSO in aqueous solution, redox titrations with monoperphthalic acid, titanium trichloride, stannous chloride, or zinc/HCl have been used. These methods are not suitable a t the micro and trace level, however. T h e absence of a method fur the determination of DMSO a t the trace level in aqueous solutions prompted this investigation into a combination of the reduction of DMSO to dimethylsulfide (DMS) and the detection of D M S by gas chromatography/flame photometry. EXPERIMENTAL Standards and Reagents. For the preparation of standards, DMSO (99.970, spectrophotometric grade, Aldrich Chemical Co., Milwaukee, Wis.), was distilled at reduced pressure, dried on activated alumina, and stored over Molecular Sieve 4A. The compound was weighed into a volumetric flask and diluted to make a stock solution containing 1000 mg S(DMS0) L-'. During the dilution, 1 mL of concentrated HC1 was added per 100 mL of solution. This prevents microbial growth and stabilizes the solution. No changes in DMSO concentration have been observed in these standards over several months. This solution was diluted to 0.1 mg S(DMS0) L-' for working standards. Aliquots of 10 0003-2700/80/0352-0150$01 .OO/O
pL to 10 mL of this working standard were further diluted to make 25 mL of solution for the calibration of the procedure. Dimethylsulfide (ICN-K & K Laboratories Inc.) standards were prepared by sequential gas dilution. Sodium borohydride (Alfa-Ventron Corp.) was used either in the form of 0.25-g pellets or a 4% solution. To obtain a more stable solution, 1 mL of 2 N NaOH was added to 50 mL of borohydride solution. In this form it was stable for several days. The chromium(I1) chloride solution was prepared by dissolving reagent grade chromium(II1) chloride (Baker) in water to form a 2 M solution, which was passed through a 25-cm column filled with amalgamated zinc for the reduction to chromium(I1) chloride. This solution, which is readily oxidized by atmospheric oxygen, has to be stored under argon or another inert gas in a stoppered container. The reagent is withdrawn from the container by a syringe inserted through the serum stopper. Apparatus. Two different configurations were investigated in this study. One consisted of a reaction/trapping apparatus connected by a six-way valve to a gas chromatograph equipped with a flame ionization detector (GC/FID system), the other apparatus combined the trapping and separation functions in one column, which was attached to a flame photometric detector (FPD system). The GC/FID system is identical to the apparatus described in (7)for the analysis of methylarsenicals, with the exception that a reaction vessel which allowed the injection of solid sodium borohydride pellets was used. The FPD system (Figure 1) is modified after a design by Braman et al. (8). The first stage in both systems is a reaction vessel, which contains the sample and buffer solutions. Helium is bubbled through the solution by a glass diffuser. A side port allows the injection of the reducing solutions by a hypodermic syringe through a Teflon-coated silicone septum attached by a Teflon Swagelok fitting. In a modified design, a short, bent piece of glass tubing which can hold a borohydride pellet is attached to the upper part of the reaction vessel by a ground glass joint. By turning this tubing, the pellet can be dropped into the solution without opening the system to the atmosphere. From the reaction vessel, the gas stream passes through a 25-cm long, 13-mm 0.d. Pyrex U-tube, which is immersed in an isopropyl alcohol bath at -35 "C. This trap removes most of the water vapor from the gas stream without condensing any dimethylsulfide. In the GC/FID system, the gas stream now passes through a 15-cm long, 6-mm 0.d. Pyrex U-tube which is filled with silanized glass wool and immersed in liquid nitrogen. This trap is interfaced by a six-way valve to a gas chromatograph (Hewlett-Packard Research Chromatograph H P 5750B), equipped with a flame ionization detector. The output of the detector is recorded on a two-pen strip chart recorder with an electronic integrator. The separation is performed on a 4.8-mm o.d., 6-m long stainless steel column, packed with 16.5% silicone oil DC-550 on 80-100 mesh Chromosorb W AW DCMS. The helium carrier gas flow is 80 mL/min air and 30 mL/min auxiliary He. In the FPD system, the sample gas stream is passed through a 6-mm o.d., 30-cm long glass U-tube filled with 15% OV3 on Chromosorb W AW DCMS 60-80 mesh. About 1 m of 0.7 Q/m Chrome1 wire is wound around the outside of this trap, connected to a variable transformer to allow for controlled heating of the trap. This trap is connected to a quartz tube burner, constructed of 11-mm 0.d. quartz glass tubing with an interior tubing drawn out to a flame tip of ca. 1-mm i.d. The flow rates for this burner are 157 mL/min He carrier, 115 mL/min hydrogen, and 200 mL/min air. This burner is mounted in an aluminum housing, which is flanged to a Heathkit photomultiplier system containing a high sensitivity multialkali photomultiplier tube with UV-glass
e 1979 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 52, NO. 1, JANUARY 1980 Reduction vessel with He
sample in
U-tube at
Packed
u
.
~
Burner ~
~
~
E Ammeter
Filter
lame
Recorder Integrator
Figure 1. Apparatus for the reduction of DMSO to DMS and the flame photometric detection of DMS (FPD system)
window (Hamamatsu Model No. R928) operated at 700 V. The PMT current is measured by a Keithley picoammeter in the lo-' to 10 A range, and the peaks are plotted on a strip chart recorder. A Varian colored glass filter (9) is used to monitor the emission from the Szband system centered at 390 nm. This filter results in an adequate signal-to-noise ratio combined with excellent selectivity with respect to the CH emission bands, as discussed in (9). Methods. Reduction ofDMSO to DMS.The sample size to be used depends on the expected concentration of DMSO. As the useful ranges of the FPD and FID detectors are from ca. 1-1000 ng S(DMS) and 2-2000 ng S(DMS), respectively, the sample size has to be chosen to fall within this range. For dilute samples, reaction vessels up to 100 mL can be attached to the bubbler. For concentrated samples, aliquots are pipetted by Eppendorff pipettors into a 25-mL reaction vessel, HC1 is added, and the volume brought to 25 mL with reagent grade water (Culligan Aqua Summa Reverse Osmosis/Deionization system, specific resistance >10 MQ/cm). For the borohydride reduction, 0.1 mL concentrated HCl is added per 25-mL sample; for the reduction with chromium(I1) chloride, 2.5 mL. The reaction vessel is then attached to the system and purged for 5 min with the helium stream. This removes air and volatile sulfur compounds. If large amounts of DMS are present, as in some algal cultures, longer purging times have to be used. In these cases it is advisable to collect the purge gas on the cold trap for 5 min toward the end of the expected purging period, and check for the existence of a DMS blank. Scrubbing is then continued until a zero blank is obtained. After the purging period, the cold trap is immersed in liquid nitrogen, and the reductant [chromium(II) chloride, sodium borohydride pellet or solution] is injected. When sodium borohydride solution is used, it has to be slowly injected into the reactor (ca. 1 min per 2 mL) to avoid an excessively fast reaction. The reductant amounts used are: scdium borohydride, one 0.25 g pellet or 2 mL of 4% solution; chromium(I1) chloride, 5 mL of a 2 M solution. The reaction times are 6 and 30 min for the borohydride and chromium(I1) reductions, respectively. The helium stream is continuously purging the solution during this period in order to remove the volatile DMS formed by reduction of DMSO from the solution. Drying of the Gas Stream. It is necessary to remove most of the moisture from the gas stream to avoid clogging of the liquid nitrogen trap and deterioration of the performance of the electron capture detector, which was operated in parallel with the FID in the instrument used for this study. A number of drying agents [CaC12,CaSO,, Mg(C104)2]were tried, but all of them removed significant amounts of DMS. Therefore a U-tube in a cold bath was used, which at -35 "C removes water vapor adequately without retaining any DMS. The interior of the U-tube was silanized (Silyl-8, Supelco, Inc.) to prevent adsorption of DMS. Trapping and Separation. The DMS and other reaction products in the gas stream are collected in a cold trap immersed in liquid nitrogen. For the GC/FID system, this trap is filled with silanized glass wool. After the collection time, the trap is switched into the carrier gas stream of the gas chromatograph by the six-way valve, and heated rapidly by immersion into hot water. The DMS is then separated on the column from other reaction products and detected by the FID.
151
In the FPD system, the trap serves both to collect the reaction products and to achieve the gas chromatographic separation. For this purpose, a trap filled with a gas chromatographic packing, as described above, is used. After the reaction time, the liquid nitrogen is removed, and the variable transformer is switched on to provide 7 V to the heating coil. The DMS elutes in a sharp peak after 1.1min and enters the flame photometric detector. If interferences are encountered in certain samples, variation of the packing material and heating rate can be used to achieve separation.
RESULTS AND DISCUSSION Reduction of DMSO. A number of reducing agents have been used t o reduce DMSO to DMS: iodide, titanium(", tin(II), zinc/HCl, and lithium aluminum hydride (6). T h e reaction occurs very easily and is usually quantitative. In this study, two new reducing agents for DMSO were investigated: sodium borohydride and chromium(I1) chloride. Sodium borohydride is a very convenient, mild reducing agent, which does not attack carbon-sulfur bonds in most compounds. T h e yield of the reduction was determined by comparison of direct injections of DMS into the system just before the cold trap with the D M S resulting from the reduction of the equivalent amounts of DMSO. The calibration curves obtained from both processes agreed within the precision of t h e determinations (*5-10%), indicating a quantitative yield when the reaction is performed as outlined in the methods section, using 0.1 mL HC1 per m L sample, and one borohydride pellet. In this case the p H of t h e solution increases during the reaction from about 1 t o near 9 owing t o t h e basic character of t h e borohydride. If t h e reaction is performed either by adding enough HC1 t o retain t h e p H in t h e strongly acidic range, or by using Tris-HC1 buffer to maintain a p H of about 7, lower yields are obtained. At pH 1 the yield was found to be 28%, whereas a t p H 7 it was 67% for duplicates of samples containing 100 ng S(DMS0). The pK, of DMSO is -1.80 (IO), therefore it is protonated only to 1. T h e reduced yield therefore a very small extent of p H cannot be due to a slower rate of reaction of the protonated species and has probably t o be attributed to t h e competing acid hydrolysis of the borohydride, which results in t h e formation of hydrogen gas and borate ion. This reaction is very rapid a t p H 1 and leads to the rapid removal of borohydride from solution and to pressure surges in the burner which can extinguish the flame. On t h e other hand, a t the higher pH, very little hydrogen is evolved during t h e reaction. As t h e production of t h e hydrogen bubbles throughout the solution promotes very efficiently the removal of the reaction product DMS into the gas phase, it is desirable to operate the reaction at p H levels conducive t o the controlled release of hydrogen gas in t h e solution. T h e reduced yields at near neutral p H may be attributed either to a p H dependence of the DMSO-NaBH, reaction [which has been observed in a similar fashion for the reaction of NaBH, with other compounds(II)], or to the reduced scrubbing efficiency under these conditions. T h e procedure outlined in the methods section, involving t h e injection of a solid borohydride pellet into a weakly buffered acidic solution permits the relatively slow release of borohydride into solution, controlled by the dissolution rate of the pellet. This results in the continuous supply of reagent to the solution and a controlled rate of hydrogen evolution. As t h e reaction progresses, the p H rises toward t h e neutral range. At the end of the reaction period (6 min), there is still a n excess of borohydride in solution. T h e reduction of DMSO by chromium(I1) ion occurs more slowly than the reaction with borohydride. A reaction time of 30 min gave a yield of 42 % ; attempts to increase the yield by increasing t h e reaction time were not successful. This is in agreement with the findings of Guerin (I2), who investigated
152
ANALYTICAL CHEMISTRY, VOL. 52, NO. 1, JANUARY 1980
Table I. Reaction of Sodium Borohydride and Chromium(I1) Chloride with Various Organic Sulfur Compoundsb sodium
compound dimethylsulfoxide dimethylsulfone dimet hylpropiot het in cystine cysteine homocystine homocysteine S-methylcysteine glutathione methionine methionine sulfoxide met hy lme t hionine
borohydride DMS N.R. DMS N.R. N.R. N.R.
__
N.R. N.R. N.R. N.R.
__
chromium(I1) chloride DMS N.R. N.R. N.R.' N.R.' N.R.' N.R.' N.R.' N.R. N.R.' N.R.'
N.R.'
,*,-a-
?
/
1 ,/ -d
, 3 ~ .5-
. .
.....
I3
.
. .
...
133 ng S DMSO)
.
.
...
,303
3337
: b
Also tested b y Guerin ( 1 2 ) . DMS: dimethylsulfide is produced, N.R.: no detectable reaction product, - : not tested. a
the release of methane thiol from sediments by reduction with chromium(I1) chloride. The lower yield of this procedure and the cumbersome preparation and handling of the reagent, together with the increased reaction time make this procedure advisable only where interfering reactions (see below) make the use of borohydride impossible. Interferences. Two types of the interferences were observed in this investigation: the negative interference due to loss of DMS to the glass container and tubing surfaces, and the positive interference due to reactions of compounds other than DMSO to produce DMS. In earlier studies (12;Andreae, unpublished data) it has been often observed that DMS gave poorly shaped, tailing peaks and that significant losses of DMS to wall surfaces occurred (13). At the nanogram level, losses of over 50% are commonly encountered. These losses can be completely prevented by thorough deactivation of all surfaces that contact the DMS vapor, including the chromatographic support, by application of silylating reagents. In this study, a commercial reagent was used (SILYL-8, Supelco Inc.), but similar success has been obtained with other reagents by Farwell et al. (13). This treatment is essential to obtain reproducible results and adequate sensitivity at the nanogram level. Experiments with different materials for connecting the parts of the glass apparatus showed t h a t losses are also commonly and unpredicably encountered with a variety of materials, including stainless steel and Teflon tubing and fittings. The best results were obtained when the components were connected glassto-glass by short sleeves of Teflon tubing. Recovery of DMS should be occasionally checked by injection of DMS vapor at various locations in the analytical system. A large number of sulfur compounds were tested in order to investigate potential positive interferences due to the formation of DMS from the reaction of sulfur compounds other than DMSO to form DMS. The results are summarized in Table I. No compound other than DMSO showed a volatile reaction product with chromium(I1). This is in agreement with the findings of Guerin (12)who investigated some of the same compounds as tested in this study. The only compound other than DMSO which gave a positive reaction with sodium borohydride is dimethylpropiothetin [(CH,)2S+CH2CH2COO~], an organosulfur compound occurring in some algae. These results show that the method is essentially free from positive interferences, except in the rare cases when dimethylsulfonium compounds are present. In these situations the chromium(I1) reaction has to be used to determine DMSO. It is possible that in some samples volatile compounds may be present which elute a t the same time as DMS. Such
J
IC
I33 ng 5 2MS;
130:
Figure 2. Calibration curves for the determination of DMSO with the FPD (a) and the GC/FID system (b)
substances could give a strong positive interference on the FID and, if present in large enough amounts, even produce a signal on the FPD. No such interferences have been observed in the samples investigated (seawater, freshwater, rain, algal exudates, urine) but may be present in complex samples like sewage and biological fluids. No systematic differences were observed between DMSO concentrations in seawater as measured by the FPD and the GC/FID system. The identity of the DMS peak can be verified by several means, e.g., comparison of the response of the GC/FID and F P D system, use of a gas chromatography/mass spectrometry system, or even by smelling the peak (the human nose can detect and recognize DMS a t the nanogram level). Analytical Parameters. The detection limit of this procedure depends to a certain extent on the specific analytical configuration used. When defined as the amount of analyte that produces a peak twice the height of the instrument noise, a detection limit of 1 ng S ( D M S 0 ) is obtained for the F P D system with chromium(I1) reduction and the GC/FID system. For the largest sample size tested (100 mL), these detection limits correspond to a concentration of 0.01 mg S(DMSO)/L. The precision is dependent on the amount of DMSO in the sample. For the F P D system with borohydride reduction, it is typically better than 5% ; for the other configurations, between 5 and 20%. I t deteriorates as the detection limit is approached. An indication of the precision can be obtained from the error bars in Figure 2b (standard deviations based on 6 to 17 samples per concentration; values obtained throughout several months), and from the results from duplicate runs of standards in Figure 2a. In view of the absence of certified standards for DMSO in water, the accuracy of the method cannot be rigorously determined. To obtain an indication of potential systematic errors with certain sample types, known amounts of DMSO have been added to seawater and freshwater samples. Recovery was quantitative within the experimental precision. The useful range of both the FPD and the GC/FID system extends to several pg S(DMS0). The FID shows a completely
ANALYTICAL CHEMISTRY, VOL. 52, NO. 1 , JANUARY 1980
Table 11. DMSO Concentrations in Some Natural Waters and Culture Media DMSO, !-@ S(DMSO) sample
L"
Seawater, Station 1 0 6 , E. North Pacific 32' 31.8' N , 118" 09.3' W surface
15-m depth 22-m depth 26-in depth 45-111depth 56-m depth 75-m depth 100-m depth 150-m depth Sacramento River, Red Bluff Calif. Colorado River, Parker, Ariz. Tamiami Canal, Fla. rain, La Jolla, Calif. 4 Jan 1977 5 Jan 1977 Growth media at the end of log phase growth Sheletonema costaturn Cricosphaera carterae
0.4Y
0.82 1.05
0.86 1.26 0.39 0.11