Anal. Chem. 1998, 70, 3579-3584
Chemiluminescence Flow-Through Sensor for Copper Based on an Anodic Stripping Voltammetric Flow Cell and an Ion-Exchange Column with Immobilized Reagents Wei Qin
Department of Chemistry, Nanjing University, Nanjing 210008, China Zhujun Zhang* and Huajun Liu
Department of Chemistry, Shaanxi Normal University, Xi’an 710062, China
A novel chemiluminescence (CL) flow-through sensor based on immobilizing all the ingredients involved in the analytical reaction for the determination of copper is proposed. The analytical reagents including luminol and cyanide were coimmobilized permanently on an anionexchange column, while the analyte copper was retained temporarily by electrochemical preconcentration on a gold electrode placed in an anodic stripping voltammetric flow cell. By injection of a volume of sodium hydroxide through the column with immobilized reagents, luminol and cyanide were eluted from the resins in alkaline aqueous solution and then reacted with copper stripped from the gold electrode to produce a CL signal, by means of which copper could be sensed. The sensor was not susceptible to interference by other metal ions associated with the CL reaction. The response to the concentration of copper was linear in the range of 0.01-10 µg/L and an extremely low detection limit of 8.0 × 10-3 µg/L was achieved. A complete analysis could be performed in 4 min with a relative standard deviation of less than 8%. The column with immobilized reagents was readily prepared and could be reused over 200 times. The sensor was applied successfully to the determination of copper in human serum and natural water samples. Flow-through (bio)chemical sensors have appeared particularly promising in the past few years. These sensors offer advantages relative to their conventional counterparts (probes) in that they facilitate sample transport and conditioning, as well as calibration and sensor preparation, maintenance, and regeneration, all of which result in enhanced analytical features and a wider scope of application. Several reviews summarize progress in this field.1-3 In general, integrated reaction (immobilization) flow-through sensors consist of a flow cell located in a nondestructive detector that is packed with a suitable support for immobilizing one or (1) Valcarcel, M.; Luque de Castro, M. D. Analyst 1990, 115, 699-703. (2) Luque de Castro, M. D.; Valcarcel, M. Lab. Robot. Autom. 1991, 3, 199205. (3) Valcarcel, M.; Luque de Castro, M. D. Analyst 1993, 118, 593-600. S0003-2700(97)00917-7 CCC: $15.00 Published on Web 07/22/1998
© 1998 American Chemical Society
more of the ingredients of the analytical reaction either permanently for the catalyst (enzyme) and analytical reagent or temporarily for the analyte and reaction product.4 In recent years, chemiluminescence (CL) flow-through sensors based on immobilized enzymes and analytical reagents have been extensively investigated.5-14 In these sensors, analytes are detected by the CL reactions with the immobilized enzymes or reagents directly and/or with the dissolved reagents which are released from the immobilized substrates or the solid forms by appropriate eluents. As compared to the use of continuously delivered reagents in the conventional CL flow systems, these CL flow-through sensors are advantageous not only for operational convenience and instrumental simplification but also for cost, environment, and resource considerations. To date, no CL flow-through sensors based on immobilized analytes have been reported. Although CL flow analysis has been one of the most powerful techniques for the determination of metal ions due to high sensitivity, wide linear dynamic range, reproducibility, simplicity, and rapidity,15-17 a limitation to widespread application of this technique is often poor selectivity of the CL detection itself, which is commonly based on the use of established CL reagents such as luminol, lucigenin, and lophine.18 So it is essential to separate (4) Fernandez-Romero, J. M.; Luque de Castro, M. D. Anal. Chem. 1993, 65, 3048-52. (5) Blum, L. J. Enzyme Microb. Technol. 1993, 15, 407-411. (6) Spohn, U.; Preuschoff, F.; Blankenstein, G.; Janasek, D.; Kula, M. R.; Hacker, A. Anal. Chim. Acta 1995, 303, 109-120. (7) Gubitz, G.; van Zoonen, P.; Gooijer, C.; Velthorst, N. H.; Frei, R. W. Anal. Chem. 1985, 57, 2071-74. (8) van Zoonen, P.; Kamminga, D. A.; Gooijer, C.; Velthorst, N. H.; Frei, R. W.; Gubitz, G. Anal. Chim. Acta 1985, 174, 151-61. (9) Rao, N. M.; Hool, K.; Nieman, T. A. Anal. Chim. Acta 1992, 266, 279-86. (10) Nieman, T. A. Mikrochim. Acta 1988, 3, 239-47. (11) Nakagama, T.; Yamada, M.; Hobo, T. Anal. Chim. Acta 1990, 231, 7-12. (12) Lu, J. Z.; Qin, W.; Zhang, Z. J.; Feng, M. L.; Wang, Y. J. Anal. Chim. Acta 1995, 304, 369-73. (13) Zhang, Z. J.; Qin, W. Talanta 1996, 43, 119-24. (14) Qin, W.; Zhang, Z. J.; Zhang, C. J. Analyst 1997, 122, 685-688. (15) Kricka, L. J.; Thorpe, G. H. G. Analyst 1983, 108, 1274-96. (16) Townshend, A. Anal. Proc. 1985, 22, 370-71. (17) Lewis, S. W.; Price, D.; Worsfold, P. J. J. Biolumin. Chemilumin. 1993, 8, 183-99. (18) Robards, K.; Worsfold, P. J. Anal. Chim. Acta 1992, 266, 147-73.
Analytical Chemistry, Vol. 70, No. 17, September 1, 1998 3579
the analyte metal ions from other metal ions or sample matrix for the improvement of selectivity in CL flow analysis. This can be achieved by on-line ion chromatography, although incompatibility of separation conditions and postcolumn reaction conditions becomes a serious problem in many applications.19-23 An effective and promising alternative way for such separation is selective preconcentration or immobilization of analytes on solid-state substrates. With the use of this preconcentration/separation step, not only selectivity but also sensitivity can be dramatically improved. In recent years, CL flow systems based on on-line preconcentration of certain metal ions on chelating resin columns have received much attention and been applied successfully to the determination of metal ions in seawater.24-27 However, in these systems, it is required to elute the preconcentrated metal ions from the substrates usually with strong acids and deliver them to the flow cells where the CL reactions take place, thus inevitably resulting in problems of analyte dilution and incompatibility of the acid eluents and the CL reactions which commonly proceed at basic conditions. Anodic stripping voltammetry (ASV) has gained wide acceptance for electrochemically well-behaved analytes. This technique offers two advantages compared to the chelating resin method for preconcentration of metal ions in CL flow systems: first, since the deposition of a certain detected metal ion at a suitable working electrode and its later stripping out can be readily and selectively carried out just via control of electrode potential, the analytical operation with use of ASV is very convenient and there is no need to use an acidic eluent; second, if the working electrode is placed in the flow cell, the stripping process and the CL reaction would take place simultaneously at the electrode surface and therefore the problem of analyte dilution can also be eliminated. These unique characteristics of ASV technique prompt us to design an ASV flow cell for analyte preconcentration for use in CL flow systems. In this paper, a novel CL flow-through sensor for copper has been developed with excellent selectivity and high sensitivity. It was based on the Cu(II)/CN-/luminol CL system.12 The analyte copper was retained temporarily in an ASV flow cell and the analytical reagents luminol and cyanide were both immobilized permanently on an anion-exchange resin column. To the best of our knowledge, this paper reports the first flow-through sensor based on immobilizing all the elements involved in the analytical reaction including the analyte and the analytical reagents. EXPERIMENTAL SECTION Reagents. All the reagents were of analytical grade; the water used for the preparation of solutions was deionized and doubly distilled. Nitric acid and hydrochloric acid were purified by subboiling distillation. Sodium chloride solution of 5.0 M was electrolytically purified by controlled potential electrolysis at -1.3 V with a mercury pool electrode. A stock solution of copper(II) (19) Yan, B.; Worsfold, P. J. Anal. Chim. Acta 1990, 236, 287-92. (20) Gammelgaard, B.; Jons, O.; Nielson, B. Analyst 1992, 117, 637-40. (21) Sakai, H.; Fujiwara, T.; Yamamoto, M.; Kumamura, T. Anal. Chim. Acta 1989, 221, 249-58. (22) Jones, P.; Williams, T.; Ebdon, L. Anal. Chim. Acta 1989, 217, 157-63. (23) Neary, M. P.; Seitz, W. R.; Hercules, D. M. Anal. Lett. 1974, 7, 583-90. (24) Elrod, V. A.; Johnson, K. S.; Coale, K. H. Anal. Chem. 1991, 63, 893-98. (25) Coale, K. H.; Johnson, K. S.; Stout, P. M.; Sakamoto, C. M. Anal. Chim. Acta 1992, 266, 345-51. (26) Sakamoto-Arnold, C. M.; Johnson, K. S. Anal. Chem. 1987, 59, 1789-94. (27) Obata, H.; Karatani, H.; Nakayama, E. Anal. Chem. 1993, 65, 1524-24.
3580 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
Figure 1. Schematic diagram of the CL flow-through sensor for copper: a, sample; b, HCl; c, NaCl; d, water carrier; e, NaOH; S, switching valve; P, peristaltic pump; I, injection valve; A, anionexchange column; E, electrochemical analyzer; F, flow cell; D, detector; R, recorder; W, waste.
Figure 2. Flow cell and electrodes.
was made by dissolving 1.000 g of pure of copper metal in the minimum volume of (1 + 1) nitric acid and diluting to exactly 1 L with 1% (v/v) nitric acid. Testing standard solutions were prepared by appropriate dilution of the stock solution with water. A 0.25 M luminol solution was prepared by dissolving 4.43 g of luminol in 100 mL of 0.5 M NaOH solution. The electrolyte employed for copper deposition and stripping out was 0.2 M sodium chloride. All glassware and polyethylene bottles were cleaned by soaking in 4 M hydrochloric acid for at least one week and then carefully rinsing several times with water. D201 anionexchange resin purchased from Nankai University was used for reagent immobilization. Apparatus. The flow system employed in this work (Figure 1) consisted of a peristaltic pump, an ASV flow cell, a column with immobilized analytical reagents, a switching valve, and an injection valve. PTFE tubing (0.8-mm i.d.) was used to connect all components in the flow system. The ASV flow cell (Figure 2) was made of a glass minicolumn (20 × 3 mm i.d.), which contained a gold working electrode with an area exposed to the detector of 2.5 × 8 mm and a screw-in Ag/AgCl reference electrode placed 3 mm away from the working electrode. The surface of the gold electrode was polished and carefully rinsed ultrasonically as described before.28 To complete the electrochemical cell, a stainless steel tube (10 × 1 mm i.d.) placed at the exit of the flow cell with a distance of 10 mm from the reference electrode was used as the counter electrode. For all experiments, the cell potential was controlled vs Ag/AgCl by an SDP-1 electrochemical analyzer (Jiangsu Jintan Analytical Instrument Plant). The CL emission with the maximum emission wavelength of 425 nm and the half-width of 83 nm was transduced to an electric (28) Jagner, D.; Sahlin, E.; Renman, L. Talanta 1995, 1447-1455.
signal without wavelength discrimination by a Hamamatsu R456 photomultiplier tube placed close to the flow cell and recorded with an XWT-204 recorder (Shanghai Dahua Instrument and Meter Plant). Absorbance monitoring was done using a UV spectrophotometer model-752 (Shanghai Third Analytical Instrument Plant). Preparation of Immobilized Reagent Column. D201 anionexchange resin (0.5 g) was stirred with 25 mL of 0.25 M luminol or potassium cyanide for 12 h, and then the resin was filtered, washed with water, and kept dry for storing. The most convenient method to determine the amounts of immobilized luminol and cyanide is to measure the change of the concentration in their immobilization solutions. This was done by UV-visible absorbance at 360 nm for luminol and titrimetry for cyanide. The amounts of luminol and cyanide immobilized were 2.05 and 2.36/g of resin, respectively. To prepare a column with immobilized reagents, 0.06 g of the resin with immobilized luminol and 0.12 g of that with immobilized cyanide were mixed together and packed into a glass column with an internal diameter of 3 mm and a total volume of about 0.5 mL and furnished with glass wool at both ends to prevent loss of the resins. Procedures. As shown in Figure 1, flow lines were inserted into sample solution (containing 0.2 M NaCl), 0.2 M hydrochloric acid washing solution, 0.2 M sodium chloride electrolyte, water carrier, and sodium hydroxide solution, respectively. With the switching valve in the washing position and the injection valve in the load position, the peristaltic pump was started at a constant speed of 3.5 mL/min on each flow line to clean the whole flow system, and the potential of the gold electrode was scanned from +0.3 to +1.2 V at 200 mV/s for several times to verify complete removal of the residue metals from the gold electrode. Then the switching valve was rotated to the preconcentration position so that the sample solution was introduced into the flow cell, and deposition of copper was simultaneously performed at -0.6 V for 2.5 min. During the next stage of 20 s, the switching valve was returned to the washing position and hydrochloric acid was drawn in to wash the flow system of any remaining sample solution. The direction of the switching valve was then changed to the detection position, and sodium chloride electrolyte was passed into the flow system. After the hydrochloric acid cleaning solution flushed from the flow cell, 150 µL of sodium hydroxide was injected via the injection valve through the column with immobilized reagents, and the eluted luminol and cyanide under basic conditions was delivered into the flow cell to react with copper that had just been stripped out at the gold electrode surface by potential scanning from -0.6 to +0.25 V at 200 mV/s, thus producing a CL light signal. The concentration of copper was quantified by the CL intensity. Safety Considerations. Care has to be taken when handling potassium cyanide, as it is toxic. General guidelines for work with acids and bases have to be respected. RESULTS AND DISCUSSION Optimization of the Chemiluminescence Flow-Through Sensor. The various experimental parameters were examined for optimum analytical performance. These variables either affected the ASV process in the flow cell, including sample pH, deposition potential, deposition time, final potential of anodic
Figure 3. Effect of sample pH on CL intensity: 1 µg/L copper(II) standard; deposition potential, -0.6 V; deposition time, 2.5 min; final potential of anodic stripping scan, +0.25 V; scan rate, 200 mV/s; eluent, 0.1 M NaOH; mixing ratio between resins with immobilized luminol and cyanide, 1:2. Error bars represent one standard deviation for four measurements.
Figure 4. Effect of deposition potential on CL intensity: sample pH, 2.0; other conditions as in Figure 3.
stripping scan and scan rate, or affected the reagent-eluting process in the ion-exchange column with immobilized reagents, including eluent, eluent concentration, and mixing ratio between resins with immobilized luminol and cyanide. The influence of sample pH on CL response is illustrated in Figure 3. It can be seen that the CL intensity increased upon increasing the sample acidity until pH 3.0 and then remained almost constant. The low CL intensities at high pH values might be ascribed to the hydrolysis of Cu(II), which affects the deposition of Cu(II) onto the gold electrode. The sample pH of 2.0 with higher CL intensity and better reproducibility was selected for the present sensor. Figure 4 shows that maximum CL intensities could be obtained when applying the deposition potentials held on the gold electrode from -0.6 to -1.0 V; however, at potentials more positive or negative than this range, the decrease of CL intensity would occur, probably because of deficient reduction of Cu(II) onto the gold Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
3581
Figure 5. Effect of final potential of anodic stripping scan on CL intensities of (b) 2 µg/L copper(II) standard and (O) the blank: sample pH 2.0; other conditions as in Figure 3.
Figure 6. Effect of scan rate on CL intensity: sample pH, 2.0; other conditions as in Figure 3. Table 1. Characteristics of Eluents for Copper Determinationa
electrode or the evolution of hydrogen at the electrode surface, which could prevent the deposition of Cu(II). To achieve high CL intensity and also to avoid interference species from accumulating on the electrode, a relative positive potential of -0.6 V was employed for subsequent work. In this case, serious interference by other metal ions associated with the CL reaction could be effectively eliminated (discussion in detail below). The effect of deposition time on CL intensity was checked by testing 0.1 µg/L Cu(II) standard in the time range from 0 to 6 min. The results showed a good linearity over the whole range. Considering a compromise between analysis time and sensitivity, a deposition time of 2.5 min was chosen for the present sensor. In the case of the final potential of the anodic stripping scan, the CL intensity increased with increase in final potential rapidly until +0.25 V and thereafter increased gradually; on the other hand, an appreciable and inescapable background CL which results from electrochemical oxidation of luminol10 would occur at potentials more than +0.3 V (Figure 5). To maximize the signalto-noise ratio and also to prevent interfering metals from being stripped out, a final potential of +0.25 V was selected as optimum. Since faster scan rates applied for the anodic stripping process can result in higher concentrations of copper(II) at the moment of being stripped from the gold electrode, it can be naturally predicted that the CL response would be improved with increasing scan rate. This situation is illustrated in Figure 6. The scan rate of 200 mV/s with the highest CL intensity was used for the present flow-through sensor. The immobilized luminol and cyanide can be eluted from the resins by various anions injected through the anion-exchange column. The elution characteristics of several different eluents including NaOH, NaAc, NaCl, NaNO3, and Na2SO4 have been examined in the flow system without the ASV flow cell as described before.12 The results are shown in Table 1. It was found that sodium sulfate was the best eluant with the highest CL intensity, which corresponds with the observation that sulfate has the highest affinity and can be most strongly retained by the anion-exchange resin, thus releasing the largest amounts of luminol and cyanide. However, for operational convenience and 3582 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
eluent relative CL intensityb
NaOH 35
NaAc 40
NaCl 55
NaNO3 74
Na2SO4 100
a CL conditions: 2 µg/L copper(II) standard; 0.1 M NaOH medium; 0.05 M eluent. b Corresponding to the normalized maximum light intensity.
instrumental simplicity, sodium hydroxide was chosen to be injected through the resin column in this sensor, which was used not only as the eluent but also as the medium to create the alkaline environment necessary for the luminol CL reaction. The amounts of released luminol and cyanide could be controlled by the concentration of sodium hydroxide injected. Various concentrations of sodium hydroxide were injected through the anionexchange resin column with immobilized luminol and cyanide and the downstream solutions were collected. The amounts of luminol and cyanide released were measured by UV-visible absorbance and the CL method,12 respectively. The results are shown in Figure 7. It can be seen that the amounts of luminol and cyanide released both increased linearly with the concentration of sodium hydroxide injected. Figure 7 also illustrates that increasing the eluent concentration gives increasing CL intensity. Considering a compromise between higher CL intensity and longer lifetime of the column, 0.1 M sodium hydroxide was used for the present sensor. In this case, the column with immobilized CL reagents could be used 200 times. Finally, the influence of the mixing ratio between resins with immobilized luminol and cyanide was examined. As shown in Table 2, the optimum mixing ratio between the amount of luminol resin and that of cyanide resin was 1:2. Performance of the Flow-Through Sensor for Copper(II) Measurements. Under the selected conditions, response to copper(II) was linear in the concentration range of 0.01-10 µg/L and the detection limit was 8 × 10-3 µg/L (3σ). The regression equation was I ) 3.74 + 42[Cu(II)] (µg/L) with a correlation coefficient of 0.9991 (n ) 11). A typical peak of CL intensity is shown in Figure 8. The whole process for the determination of
Table 3. Results of the Determination of Copper in Natural Waters sample
present sensora (µg/L)
AASa (µg/L)
diff (%)
rainwater 1 rainwater 2 tap water 1 tap water 2 underground water 1 underground water 2
3.65((4.6) 2.82((5.6) 4.70((4.6) 3.52((5.4) 0.98((6.4) 1.58((5.8)
3.77((6.2) 2.72((8.3) 4.85((6.0) 3.34((7.4) 1.06((9.4) 1.66((8.8)
-3.2 +3.7 -3.1 +5.4 -7.5 -4.8
a
Average value of four determinations ((RSD, %).
Table 4. Results of the Determination of Copper in Human Serum
Figure 7. Effect of eluent concentration on (O) amount of luminol released, (4) amount of cyanide released and (0) CL intensity: conditions as in Figure 3 except sample pH of 2.0. Table 2. Effect of Mixing Ratio between Resins with Immobilized Luminol and Cyanidea mixing mass ratio relative CL intensityb
4:1 36
2:1 44
1:1 66
1:2 100
sample
present sensora (mg/L)
AASa (mg/L)
diff (%)
serum 1 serum 2 serum 3 serum 4
1.35((3.2) 1.05((3.6) 0.87((4.6) 0.98((4.6)
1.29((5.2) 1.09((6.4) 0.82((7.0) 1.03((5.8)
+4.7 -3.7 +6.1 -4.9
a
Average value of four determinations ((RSD, %).
1:4 82
a Sample pH, 2.0; other conditions as in Figure 3. b Corresponding to the normalized maximum light intensity.
Figure 8. Observed CL signals for duplicate measurements of (a) 0 (the blank), (b) 0.2, and (c) 0.4 µg/L copper(II).
Cu(II) could be performed in 4 min, giving a throughput of about 15 h-1. Multiple determinations (n ) 7) gave relative standard deviations of 3.5, 6.0, and 7.4% at Cu(II) concentrations of 4, 0.4, and 0.04 µg/L, respectively. Interference Study. The influences of foreign species were investigated by analyzing a standard solution of 1 µg/L Cu(II) to which increasing amounts of interfering species were added. The tolerable limit of a foreign species was taken as a relative error not greater than 5%. The interference study of the Cu(II)/CN-/ luminol CL system was carried out without the use of an ASV flow cell. It was found that the selectivity of the CL system was rather poor and serious interference would result from 0.5-fold Fe2+, Fe3+, and Mn2+ and 2-fold Co2+ and Ni2+. However, with the use of an ASV flow cell, none of these metal ions at a concentration of 10 mg/L showed any effect on the CL signal
compared to the 1 µg/L Cu(II) standard. In addition, studies also showed that more than a 1000-fold excess of Ca2+, Mg2+, Ba2+, Zn2+, Sn2+, Al3+, Ti(IV), Mo(VI), Cr(VI), V(V), and Cr3+, less than a 200-fold excess of Sn(IV), Pb2+, and Cd2+, and less than a 50fold excess of Hg2+, Ag+, and Bi(III) did not interfere. Under the applied specific conditions, all these metal ions either could not be accumulated effectively by the gold electrode or could not be stripped from the electrode to interfere with the CL detection of copper, thus confirming the efficiency of the ASV flow cell in removing the metal ion interferences. Besides metal ions, some nonmetal ions were also tested for their potential interferences in the present sensor. The tolerable concentration ratios with respect to 1 µg/L Cu(II) were more than 1000 for NH4+, NO3-, HCO3-, HPO42-, F-, Ac-, SO42-, Br-, SO32-, and NO2-, 50 for Te(IV), and 20 for Se(IV) and As(III). A satisfactory selectivity of the proposed sensor with an ASV flow cell is therefore evident. APPLICATION Analysis of Copper in Natural Waters. Rainwater, tap water, and groundwater were collected in 125-mL polyethylene bottles and filtered through a standard 0.45-µm filter. The filtrates (20 mL) were treated with 5 mL of 2 mol/L NaCl and 5 mL of 0.1 mol/L HCl and then diluted to 50 mL with water for the determination of copper. The results are given in Table 3. It is shown that the values obtained by the present sensor agree well with those obtained by graphite furnace atomic absorption spectrometry (AAS). Analysis of Copper in Human Serum. A 0.2-mL aliquot of human serum sample was added into a 25-mL beaker, followed by 0.2 mL of 70% distilled nitric acid, 0.2 mL of 60% distilled perchloric acid, and 0.5 mL of 95% distilled sulfuric acid. The solution was heated to near dryness and the residue dissolved in 5 mL of water. The clear solution was quantitatively transferred into a 50-mL volumetric flask, and then 5 mL of 2.0 mol/L sodium chloride and 5 mL of 0.1 mol/L hydrochloric acid were added. Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
3583
The mixture was completed to the mark with water. The concentration of copper was determined by the proposed sensor and the results are shown in Table 4, which again correspond with those obtained by AAS. CONCLUSIONS This paper reports a novel CL flow-through sensor for copper based on immobilizing all the ingredients of the CL analytical reaction, including the analyte and the analytical reagents, either temporarily in an ASV flow cell or permanently on an anionexchange resin column. With the combination of the ASV technique and flow injection CL detection, the sensor shows excellent selectivity and extremely high sensitivity for the determination of copper and the operation procedure is convenient and rapid. In addition, the use of immobilized analytical reagents
3584 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
including luminol and cyanide offers great cost and safety advantages in the application of this flow-through sensor. This type of CL flow-though sensor with the use of an ASV flow cell and immobilized reagents is attractive for extension to determinations of other metal ions, and this aspect is now under study. ACKNOWLEDGMENT This study was supported by the National Natural Science Foundation of China (Grant 39730160).
Received for review August 21, 1997. Accepted May 22, 1998. AC970917P
