Determination of phenols by solid-phase microextraction and gas

Chanbasha Basheer , Hian Kee Lee , Jeffrey Philip Obbard. Journal of ..... Albert C. Censullo , Dane R. Jones , Max T. Wills. Journal of Coatings Tech...
0 downloads 0 Views 1MB Size
Environ. Sci. Technol. 1993, 27, 2844-2848

Determination of Phenols by Solid-Phase Microextraction and Gas Chromatographic Analysis Karen D. Buchholz and Janusz Pawliszyn' The Guelph-Waterloo Centre for Graduate Work in Chemistry and the Waterloo Centre for Groundwater Research, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1

Solid-phase microextraction (SPME) is a fast and simple analytical technique which uses coated fused silica fibers to extract analytes from aqueous samples. The analytes are desorbed in the injector of a gas chromatograph and subsequently analyzed. A SPME method based on poly(acrylate) coated fibers has been developed €or phenols as an alternative to the U.S. Environmental Protection Agency (EPA) wastewater methods 604 and 625-acid extractables. Limits of detection are typically at nanograms per liter, and the precision is at 5 % RSD, including nitrophenols. The sensitivity of the method is enhanced at low pH levels and with the addition of salt. The method was applied to the analysis of groundwater, and wastewater samples and SPME data are compared with more traditional techniques. The results indicate the suitability of the SPME approach for rapid screening for phenols at high and low concentration levels.

Introduction Phenol and its derivatives are used widely in industry and can be serious health hazards if released into the environment through accidental spills or poor disposal practices. They are highly toxic and cause taste and odor contamination of drinking water supplies at trace levels (1).

Current analytical methods, such as EPA methods 604 and 625 (acid extractables section), are based on liquidliquid extraction (LLE). LLE of phenols is often difficult because of their high solubility in water. These methods call for the acidification of the sample, followed by the extraction into methylene chloride, concentration, and analysis by gas chromatography with flame ionization (GC/ FID) or mass spectrometry (GUMS)detection (2,3).There is the potential for analyte loss at each step in the cleanup and extraction procedure. Typical recoveries range from 40 to 89 % , with the precision varying widely from 38 to 64% RSD ( 4 ) . Beside being time consuming, LLE techniques require the use of toxic and expensive solvents, which are undesirable for health, economic, and disposal reasons. There is a definite need for a phenol analytical method that is fast and simple, yet still capable of meeting the detection limits of the current standard methods. Solid phase microextraction (SPME) meets these requirements and essentially eliminates solvents from the extraction process. SPME has been used for environmental applications such as substituted benzenes in water and headspace (5),the chlorinated compounds of U S . EPA method 624 (6),and selected PCBs (7)in water. The method has also been fully automated (8). An outstanding feature of the SPME sample preparation method is its simplicity. Organic pollutants are absorbed from aqueous or gaseous samples by the solid phase coating of a silica fiber support. The analytes are then directly transferred to the injector 2844

Envlron. Sci. Technol., Vol. 27, No. 13, 1993

Stainless Steel Rod

I

EPOXY

1I Syxinge Needle

Fibre

' !

Plunger

cap

Figure 1. Syringe assembly used in SPME. When the plunger is pulled back, the fiber is drawn inside the needle to protect it from breakage when the septum of a sample vial or GC injector is pierced.

of a gas chromatograph (GC) using a modified syringe assembly where they are thermally desorbed and analyzed. This process is much more simple than conventional techniques and significantly reduces the potential for analyte loss during the extraction process. There is also a tremendous savings in terms of analytical time. Extraction times are in the order of min compared to hours for LLE techniques, while still maintaining the high sensitivity required for trace organic analysis. In this paper, we are describing the first application of SPME to the analysis of polar compounds. Most of the previous methods have been based on poly(dimethy1si1oxane)-coatedfibers, which are relatively nonpolar. The selectivity of the fiber coating toward polar compounds has been achieved by changing its chemical nature. For this work, poly(acry1ate)-coatedfibers have been used to demonstrate the applicability of this rapid sample preparation technique to phenols.

Experimental Section The modified syringe assembly used in SPME is shown in Figure 1. A 1-cm length of a poly(acry1ate) fiber (PolyMicro Technologies, Phoenix, AZ) was glued into a stainless steel rod with high-temperature epoxy resin. The fibers had been previously conditioned at 300 OC for 3 h under helium. The thickness of the coating used was 95 Fm. The rod was inserted into a Hamilton 7005 syringe and glued to the top of the plunger. When the plunger is retracted, the fiber is drawn into the syringe needle. This protects the fiber when the syringe needle is used to pierce the septum of a sample vial or GC injector. The plunger can then be pushed down to expose the fiber to the sample solution during the extraction or to the GC carrier gas during thermal desorption. The SPME extractions were done with magnetic stirring to ensure the proper mixing of the sample solution. The fibers were conditioned prior to use at 300 "C under helium for several hours to reduce bleed. The fibers were 0013-936X/93/0927-2844$04.00/0

0 1993 Amerlcan Chemlcal Society

placed inside a stainless steel tube (U4-h. diameter), which was then placed inside a muffle furnace. A compressed gas cylinder provided the helium flow for the conditioning. Copper tubing leading from the regulator of the helium tank was connected to a piece of fused silica tubing. The silica tubing was threaded through a gap in the furnace door and connected to the tube containing the fibers using a 1/16-1/4-in. stainless steel reducing union(Swagelock, Niagara Valve, Hamilton, Ontario). Graphite ferrules were used to seal the connection. At the other end of the tube containing the fibers, a similar connection was made to a second piece of fused-silica tubing. This piece was threaded through the gap in the muffle furnace door, providing an outlet for the helium flow. The furnace door was then closed carefully so as not to crush the silica tubing, and the furnace was then turned on to the required temperature. A small vial of water was placed at the end of the silica tubing leading out of the furnace. The appearance of bubbles in the water served as an indicator that the gas was flowing through the system during the whole process. Stock standard mixtures of the 11 phenolics in EPA 604/625 dissolved in methanol were purchased from Supelco Canada in 1-mL aliquots. The stock contained phenol; 2-chlorophenol; 2-nitrophenol, 2,4-dimethylphenol; 2,4-dichlorophenol; 2,4,6-trichlorophenol; 2,4-dinitrophenol; 4-chloro-3-methylphenol;4-nitrophenok 2-methyl-4,6-dinitrophenol; and pentachlorophenol. The stock mixture was diluted by a factor of 10 with 2-propanol to obtain a spiking standard. This standard was used to spike 30-mL water samples that were prepared fresh daily. The GC/FID work was done using a Varian Vista 6500 GC that was interfaced to the data system of a Model 6000 GC. The column used for all experiments was a 30 m PTE-5 (Supelco, Bellefonte, PA) capillary column with a 0.25-mm-i.d. and 0.25-pm film thickness. Helium was used as the carrier gas with a flow rate of 1.5 mL/min. A split/ splitless injector was used in the splitless mode and maintained at 200 "C. A 7.0-min desorption time was used for all fiber injections. The column program was as follows: initial temperature, 35 "C for 7.0 min; ramp to 190°C at a rate of 10°C/min; final hold time of 7.7 min. The detector was kept at 275°C with a nitrogen makeup gas flow at 23.0 mL/min, hydrogen at 30.0 mL/min, and air at 280 mL/min; range 12 and attenuation 32. A Varian 3400 GC/Saturn ion-trap mass spectrometer was used for the GC/MS experiments. The same GC column and operating conditions were used; however, this GC had an SPI (septum programmable) injector, which was maintained at 200°C. The temperatures of the transfer line and ion-trap manifold were maintained at 280 and 25OoC,respectively. The mass spectrometer was tuned to FC-43 (perfluorotributylamine). The filament emission current was set so that the ionization time was 20 ms. The segment breaks were left at the default values of 10-99/ 10O-249/250-399/400-650 amu, and the segment tune factors were adjusted to 80/80/120/85. The electron multiplier voltage and automatic gain control target were set automatically. The mass range scanned was 45-325 amu, and the detector was turned off for the first 300 s of the run to prevent overloading from the 2-propanol used in the spiking standards. The first step in the SPME phenol method development procedure was to determine when equilibrium had been reached between the analytes in the coating and in the sample. This corresponds to the concentration optimum

sampling time. A standard water solution was prepared and extracted for varying lengths of time to obtain an absorption time profile. All results were done in duplicate or triplicate to ensure reproducibility. A fresh solution was prepared for each time interval. A plot was made of GC/FID response (in area counts) vs time to determine graphically when equilibrium had been reached. To determine if any of the analytes remained on the fiber after desorption, carryover profiles were performed. Three consecutive desorptions of the same fiber were run after the initial desorption, following exposure to a standard solution. The results were reported as a percent of area of the initial desorption peak. For the fiber linearity studies, arange of standard water samples were prepared, with each successive sample being 10 times more dilute than the previous one. A plot was made of GC area counts vs concentration for each analyte to graphically determine the linear range. All procedures were carried out in duplicate. Detection limits were determined for both FID and MS detectors. For FID, the calculated limits were determined by comparing the GC area counts of the lowest detectable standard concentration to a peak threshold level of 10 000. For example, if a 3 ppb standard gave an area count of 30 000 for a particular analyte, then the detection limit was estimated to be 1.0 ppb. A peak threshold of 10 000 was arbitrarily chosen as a conservative measure of the instrument's noise. For the GC/MS results, detection limits were calculated by comparing the signal to noise ratio (S/N) of the lowest detectable concentration to an S/N of 3. For example, if a 10 ppb solution gave a S/N of 6, the detection limit was calculated to be 5 ppb. All procedures were carried out in duplicate. The precision of the method was determined by running 10 consecutive samples. The average GC area counts and corresponding relative standard deviations ( 75 RSD) were determined for each analyte in the mixture. The effect of low pH and salt were examined as a means of enhancing the extraction. The spiking standard was added to a 30-mL sample of pH 2 buffer solution, which had been saturated with sodium chloride. The buffer solution was prepared with 25 mL of 0.2 M KC1 and 6.5 mL of 0.2 M HCI in 100 mL of water (9). The amount extracted for each analyte was compared to that of a control sample at neutral pH and without any salt added. Groundwater and paper mill effluent samples were analyzed by adding 7.5 g of NaCl and 2 drops of sulfuric acid (pH < 1) to 30 mL of the sample prior to SPME extraction. The quantitation was performed using external calibration in distilled water under the same salt and pH conditions. Results and Discussion SPME, unlike most conventional techniques, is not based on an exhaustive extraction of the sample, but on an equilibrium between the analyte concentration in the sample and that in the solid phase fiber coating. The distribution constant for the equilibrium, K , is defined as (7):

K = CJC,, (1) where C, is the analyte concentration in the solid phase, and C,, is the concentration in the aqueous phase. The moles extracted at equilibrium, n,, can be defined by the Environ. Sci. Technol., Vol. 27, No. 13, 1993 2845

2'-7

cDI=---

0

10

20

40

30 TIHk

eo

50

( m s )

4.0Oer7

2W+7

o.ooe+o

20

10

30 TPQ

40

50

60

IYnmrrS)

Figure 2. Absorption time profiles. All analytes equilibrate within 40 mln. Abbreviations: phenol (PHE); 2-chlorophenol (2C); 2-nitrophenol (2N), 2,4-dimethylphenol (24DM); 2,4dlchlorophenol (24DC); 2,4dinitrophenol (24DN); 4-chloro-3-methylphenoi(4C3M); 2,4,6-trichlorophenol (246TC); 4-nitrophenol (4N); 2-methyl-4,6-dinitrophenol (2M46DN); and pentachlorophenol(PCP).

following equation: (2) na = COaqVaq - C'aqVaq where Vaqis the volume of the aqueous sample, and Coaq and Cuaqare the initial and equilibrium concentrations of the analyte in the sample, respectively. This is essentially the difference between the analyte concentration in the sample prior to the extraction and the concentration remaining at equilibrium. The direct relationship between the amount extracted and K can been seen by rearranging eq 1 to obtain

ns Vaq (3) VB(~,,Coaq - n,) A coating with a high affinity for an analyte will have a high K value associated with it, and a large amount of the analyte will be extracted at equilibrium. The amount extracted can also be increased by using a thicker coating. Figure 2 shows the absorption time profiles for the analytes in the mixture. All of them reach equilibrium within 40 min. The total sample preparation time is considerably lower than the hours required for conventional techniques and can be reduced even further by using alternative agitation techniques. Previous SPME work has shown that sonication is much more efficient than magnetic stirring (10); however, care must be taken to avoid analyte loss. It should also be noted that it is not essential for equilibrium to be reached in the SPME process. Shorter times can be used as long as the extractions are timed carefully and the mixing conditions remain constant. The area counts at equilibrium were used to calculate the moles of analyte extracted:

K=

n, = (AC X R)IMW (4) where AC is the GC area counts from a fiber injection, R 2848

Envlron. Scl. Technol., Vol. 27, No. 13, 1993

is a response factor (of pg/area count) determined from a syringe injection of a liquid phenol standard, and MW is the molecular weight of the analyte. This value was then used to calculate the K value according to eq 3. V,, Vas, and Caq in eq 3 were 0.00196 cm3, 30 mL, and 1.7-8.3 pgI mL, respectively. Caq depended on the analyte's initial concentration in the stock standard. The analytes with the greatest number of area counts indicate the highest affinity for the fiber and, therefore, had the highest K values. The K values for all of the compounds in the mixture are listed in Table I. Pentachlorophenol has by far the highest affinity for the fiber, followed by the other chlorophenols, nitrophenols, and phenol itself. Figure 3 shows a GC/FID chromatogram of the water sample spiked with the EPA 604/625compounds. Phenol and 2-chlorophenol coeluted as one peak on the column used at the time of this sample run. The pentachlorophenol peak is by far the largest, as expected from its high affinity for the fiber coating. Beside the peaks for the 11 target compounds, there are two others of interest. The large peak at the beginning of the chromatogram is from 2-propanol used to spike the phenols into the water samples. There is also a peak occurring just after the phenol and 2-chlorophenol peak. This peak results from compounds desorbing from the coating itself and decreases after prolonged use of the fiber and eventually disappears altogether. It can also be eliminated by conditioning the fibers prior to use at 3OOOC under helium for several hours (see Experimental Section for details). The results from the carryover profiles showed that most of the target analytes were efficiently desorbed off the fiber coating during the initial desorption. Carryover was only observed for 4-chloro-3-methylphenol, 2,4,6-trichlorophenol, and pentachlorophenol. The carryover was measured with three blank injections following the initial desorption. In the blank injections, the fiber was placed in the injector without prior exposure to the sample solution. The carryoverresults were calculated by dividing the GC area counts from the blank injections by the analyte peak area from the initial desorption just after the fiber's exposure to the sample solution, all multiplied by 100%. 4-Chloro-3-methylphenoltypically gave values of 0.7,0.3, and 0.1% on three consecutive injections following an initial desorption. The corresponding values for 2,4,6trichlorophenol were 0.9,0.3, and 0.2%. Pentachlorophenol had the highest carryover at 1.1,0.5, and 0.3 %. This can be improved by using higher desorption temperatures or longer desorption times. Table I1 shows the linear ranges for the GC/FID and GC/MS results. Most analytes were linear over a 100 fold concentration range. For the FID, the bottom of the linear range was limited by the sensitivity of the detector. At the top end, the range was limited because of the solubility in water for some of the analytes, particularly pentachlorophenol. For the MS, the higher end of the linear range was also limited by the capacity of the ion trap, which was easily overloaded at higher concentrations. This indicates that the linearity of the method is limited by the detector and not the extraction technique itself. Since the fiber coating concentrates the analytes, the amount deposited on the column is higher than if the same concentration of the sample was introduced as a liquid syringe injection. For example, with a 0.2 pL splitless injection of a sample containing 0.5 ppm of the 2,4dichlorophenol sample, 0.1 ng is injected onto the column. If a 40-mL sample of the same solution is extracted by

Table I. Poly(acrylate)/Water Distribution Constants ( K ) , Detection Limits, and Precision

K values

compound phenol 2-chlorophenol 2-nitrophenol 2,4-dimethylphenol 2,4-dichlorophenol 4-chloro-3-methylphenol 2,4,6-trichlorophenol 2,4-dinitrophenol 4-nitrophenol 2-methyl-4,6-dinitrophenol pentachlorophenol a

SPME GUMS (pg/L)

1.3 9.3 3.7 9.1 47 16 60 1.7 2.4 7.3 170

detection limits enhanced SPME EPA 625 GUMS (fig/LP GUMS (rg/L)

0.80 0.24 0.38 0.02 0.02 0.01 0.08

1.6 0.75 0.44 0.11

0.13 0.06 0.004 0.01

0.01 0.004 0.04 0.09

0.24 0.07 0.08

1.5 3.3 3.6 2.7 2.7 3.0 2.7 42.0 2.4 24.0 3.6

EPA 604 GC/FID (rg/L)

precision ( % RSD)

0.14 0.31 0.45 0.32 0.39 0.36 0.64 13.0 2.8 16.0 7.4

4.2 4.2 5.2 4.8 4.9 4.0 4.5 8.9 9.3 5.6 12

Estimated limits considering factor increases achieved with pH 2 and saturated salt conditions.

I

6

7

Time (minutes)

Flgure 3. GCIFID chromatogram of a water sample spiked with the 11 phenolics in EPA Method 604. Phenol (1); 2-chlorophenol (2); 2-nitrophenol (3); 2,rldimethylphenol (4); 2,4dichlorophenol (5); 4-chloro-3-methylphenol(6);2,4,6-trichlorophenoI(7); 2,4dinitrophenol (8); 4-nitrophenol (9); 2-methyl-4,6dinitrophenol (10); and pentachlorophenol (1 1).

Table 11. Tested Linear Range of EPA 604/625 Compounds linear range (pg/mL) compound phenol 2-chlorophenol 2-nitrophenol 2,4-dimethylphenol 2,4-dichlorophenol 4-chloro-3-methylphenol 2,4,6-trichlorophenol 2,4-dinitrophenol 4-nitrophenol 2-methyl-4,6-dinitrophenol pentachlorophenol

GC/FID

GUMS

0.2-2.0 0.02-2.0 0.02-2.0 0.02-2.0 0.002-2.0 0.008-8.0 0.05-5.0 0.5-5.0 0.08-8.0 0.08-8.0 0.008-8.0

0.007-0.7 0.007-0.7 0.007-0.7 0.007-0.7 0.007-0.7 0.007-0.7 0.007-0.7 0.07-0.7 0.007-0.7 0.007-0.7 0.007-0.7

SPME, 46 ng is deposited onto the column. Table I lists GC/MS detection limits, along with the regulatory limits. All of these results easily surpass the EPA 625 guidelines and also meet the more stringent 604 limits. Similar results can be obtained using FID. With this detector most of the results meet the EPA 625 limits and are within an order of magnitude of the 604 specifications, with the exception of phenol and 2-nitrophenol which have FID limits of 30 and 11 pg/L, respectively. In general, the compounds with the highest Kvalues had the lowest detection limits. There were some deviations to this pattern due to poor instrument sensitivity. For example, pentachlorophenol should theoretically have the

lowest detection limit since it has the highest affinity for the fiber coating. However, this compound has one of the poorest instrument response factors of the mixture, likely due to its poor peak shape and loss in the injector(l1). The precision of the SPME method is significantly better than typical LLE results. Table I shows the relative standard deviations (RSD) of 10 consecutive extractions. These results indicate that SPME is a reliable analytical tool for phenol analysis. Most of the analytes have a RSD of 5% or better with the exception of the nitrophenols and pentachlorophenol. The poorer precision for these compounds is partly due to chromatograhic reasons. 2,4Dinitrophenol and pentachlorophenol can be difficult to chromatograph reproducibly at low levels with a splitless injector. The chromatographicprocessis highly dependent on the cleanliness of the injector liner and loss of these analytes can occur due to discrimination in the flash vaporization process ( I 1). The amount of analytes extracted can be enhanced in a low pH and saturated salt environment. The bar graphs in Figure 4 compare amounts extracted from the pH 2 buffer-saturated salt solution and the control sample at neutral pH and with no salt added. The low pH ensured that all of the phenols were in their neutral form, and the addition of sodium chloride essentially salted the analytes out of solution and into the fiber coating (8). The enhancement factors are very significant. They are typically 2-5 for most analytes in the mixture. 2,4Dinitrophenol shows the largest increase with the amount extracted being 17 times greater than with pH 7 and pure water conditions. These enhancement techniques potentially can be used to further lower limits of detection (Table I). Additional improvement in sensitivity for polar compounds can be obtained by using in situ derivatization (12). In the case of SPME, this can be performed directly in the coating of the fiber by absorbing the reagent into the polymer prior to extraction. The disadvantage to using the acidic and salt conditions was an increase in the extraction time. Under these conditions, 60 min was required for all of the analytes to equilibrate. This is due to the fact of a greater mass being absorbed by the coating. Also, the diffusion of the analytes through the aqueous phase would have been slower than in pure water. In applications where shorter extraction times are necessary, some of the increased sensitivity can be compromised for shorter extraction times by using a thinner coating. Salt and pH conditions are also currently being investigated as ways of compensating for sample matrix variations. Envlron. Sci. Technol., Vol. 27,

No. 13, 1993 2847

IW*

mr-

Flgure 6. Extraction of 2,4dichlorophenol(l) from pulp and paper mill extraction stage effluent.

of pulp and paper extraction stage effluent. The SPME technique gave a result of 2.5 ppb compared to 3.2 ppb determined by in situ acetylation and liquid-liquid extraction with petroleum ether (13). Acknowledgments

4t.

2 w - y

P

Flgure 4. Extraction enhancement under salt and pH 2 conditions.The abbreviations are the same as in Figure 2.

Thanks to Professor Jim Barker of the Waterloo Centre for Groundwater Research (WCGR) for his encouragement and advice as well as Kim Hamilton (WCGR) and Brian MacGillivray of the Waste Water Technology Centre, Burlington, Ontario for providing the samples and analytical data. The help of Lin Pan and Zhouyao Zhang with the analysis of the samples was greatly appreciated. Financial support was provided by the Natural Sciences and Engineering Research Council of Canada, Varian, and Supelco. Literature Cited

188.

1

(1) Fresenius, W.; Quentin, K.; Schneider, W., Eds. Water

(2) (3) (4)

(5) (6)

SPME extraction of 2,4dimethyIphenol(l) from a contam inated groundwater sample.

Flgure 5.

Although an intensive investigation of the SPME technique involving various matrices and analytes is required prior to validation for quantitative analysis, preliminary results show that SPME can be used as an effective screening tool. Figure 5 is a chromatogram of a groundwater sample from a creosote contaminated site found to contain 1.0 ppm of 2,4-dimethylphenol by the SPME technique. When the sample was analyzed by liquid-liquid extraction with methylene chloride, it was found to contain 0.2 ppm of this analyte, which indicates poor solvent extraction efficiency. The chromatogram in Figure 6 shows 2,4-dichlorophenol extracted from a sample

2848

Envlron. Sci. Technol.. VoI. 27. No. 13. 1993

(7) (8) (9) (10) (11) (12) (13)

Analysis: A Practical Guide to Physico-Chemical, Chemical and Microbiological Water Eramination and Quality Assurance; Springer-Verlag: Germany, 1988. US.EPA Method 604. Fed. Regist. 1984,49,43290. US. EPA Method 625. Fed. Regist. 1984, 49, 153. Hall, J. R.; Florance, J. R.; Strother, D. L.; Wass, M. N. EPA Method Study 14: Method 604-Phenols; Contract No. 68-03-2625; U. S. Environmental Protection Agency: Washington, DC, 1984. Zhang, Z.; Pawliszyn, J. Anal Chem. 1993, 65, 1843. Arthur, C.; Pratt, K.; Motlagh, S.; Pawliszyn, J. J. High Res. Chromatogr. 1992, 15, 141. Arthur, C.; Pawliszyn, J. Anal. Chem. 1990, 62, 2145. Arthur, C.; Killam, K.; Buchholz, K.; Berg, J.; Pawliszyn, J. Anal. Chem. 1992,64,1968. Weast,R.C.,Ed. CRCHandbook OfChemistryandPhysics, 59th ed.; CRC Press: West Palm Beach, FL, 1979. Motlagh, S.; Pawliszyn, J. Anal. Chimica Acta, in press. Keith, L. H., Ed. Advances in the Identification and Analysis of OrganicPollutants in Water:Ann Arhor Science Publishers: Ann Arbor, MI, 1981. Ohkura, Y.; Nohta, H. Adu. Chromatogr. 1989, 29. 221. Lee, H.-B.; Hong-You, R.; Fowlie, P. J. Assoc. O f f .Anal. Chem. 1989, 72,979.

Received for review March 19, 1993. Revised manuscript received September 2, 1993. Accepted September 13, 1993.0

@Abstractpublished in Advance ACSAbstracts, October 15,1993.