Fiber-Optic Microsensor Array Based on Fluorescent Bulk Optode

A corresponding Ag+-selective fluorescent optical microsensor array for the rapid sensing of trace level Ag+ was then developed. It was fabricated usi...
0 downloads 0 Views 178KB Size
Download PDF
Anal. Chem. 2005, 77, 4706-4712

Fiber-Optic Microsensor Array Based on Fluorescent Bulk Optode Microspheres for the Trace Analysis of Silver Ions Katarzyna Wygladacz,† Aleksandar Radu,§ Chao Xu,† Yu Qin,‡ and Eric Bakker*,†

Department of Chemistry and Biochemistry, Auburn University, Auburn, Alabama 36849

An optical microsensor array is described for the rapid analysis of silver ions at low parts per trillion levels. Because the ionophore o-xylylenebis(N,N-diisobutyldithiocarbamate) (Cu-I) was reevaluated and shown to exhibit excellent selectivity for silver ions, ion-selective electrode (ISE) membranes were optimized and found to exhibit the lowest reported detection limit so far (3 × 10-10 M). A corresponding Ag+-selective fluorescent optical microsensor array for the rapid sensing of trace level Ag+ was then developed. It was fabricated using plasticized PVC-based micrometer-scale fluorescent microspheres that were produced via a sonic particle casting device. They contained 156 mmol/kg Cu-I, 10 mmol/kg 9-(diethylamino)5-[4-(15-butyl-1,13-dioxo-2,14-dioxanodecyl)phenylimino]benzo[a]phenoxazine (chromoionophore VII, ETH 5418), 2.3 mmol/kg 1,1′′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (internal reference dye), and 14 mmol/kg sodium tetrakis[3,5-bis(trifluoromethyl)phenyl]borate and were deposited onto the etched distal end of a 3200-µm-diameter optical fiber bundle. The microarray was characterized by fluorescence spectroscopy in samples containing 10-12-10-8 M AgNO3 at pH 7.4, with selectivity characteristics comparable to the corresponding ISEs. The response time of the microsensor array was found to be less than 15 min for 10-9 M AgNO3, which is drastically shorter than earlier data on optode films (8 h) and corresponding ISEs (30 min). A detection limit of 4 × 10-11 M for Ag+ was observed, lower than any previously reported optode or silver-selective ISE. The microsensor array was applied for measurement of free silver levels in buffered pond water samples. Silver is considered a metal of commercial importance as well as a good indicator of urban pollution. It is used in photography, batteries, and the semiconductor industry. Because of its antibacterial properties, silver compounds have been used to disinfect water used for drinking and recreational purposes, in dental and pharmaceutical antibacterial and anti-HIV preparations, and in * To whom correspondence should be addressed. E-mail: [email protected]. † Present addresss: Department of Chemistry, Purdue University, West Lafayette, IN 47907. § Present address: Adaptive Information Cluster, Dublin City University, Glasnevin, Dublin 9, Ireland. ‡ Present addresss: Department of Chemistry, Renmin University of China, Beijing, P. R. China, 100872.

4706 Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

implanted prostheses.1 It has been estimated that every year ∼450 000 kg of silver enters the aquatic system and the atmosphere. Although only a small portion of total silver is biologically available in the water, it has been found that an excess of silver is toxic to fish and microorganisms at a concentration as low as 0.17 µg/L.1 It is therefore crucial to closely monitor the activity of silver in the environment. Inductively coupled plasma atomic emission spectrometry,2,3 inductively coupled plasma mass spectrometry,4,5 and flame atomic absorption spectrometry6,7 have been commonly used for trace determination of silver in various samples. These techniques normally require sample preconcentration and are considered labor and consumables intensive. Ion-selective electrodes (ISEs) have been established for many years as powerful analytical tools. In contrast to other analytical techniques, where normally total concentrations are measured, they are known to respond to the analyte ion activity. More recently, ISEs have been successfully applied for the trace analysis of ions of environmental and physiological importance.8-11 They can now be considered competitive methods for the determination of free ion levels because their detection limits have been successfully reduced. This was accomplished by effectively reducing zero current ion fluxes from the membrane interior, inner electrolyte solution, or both, which otherwise give rise to elevated local concentration levels at the membrane surface and therefore higher detection limits.12-15 Note that ion-selective electrodes with an inner aqueous electrolyte contact will invariably show a kinetic (1) EPA Drinking Water Criteria Document for Silver; Environmental Protection Agency: Washington, DC,1989; p∧pp EPA CASRN 7440-7422-7444. (2) Wang, T.; Jia, X.; Wu, J. J. Pharm. Biomed. Anal. 2003, 33, 639-646. (3) Yang, X. J.; Foley, R.; Low, G. K. C. Analyst 2002, 127, 315-318. (4) Jitaru, P.; Tirez, K.; De Brucker, N. At. Spectrosc. 2003, 24, 1-10. (5) Krachler, M.; Mohl, C.; Emons, H.; Shotyk, W. Spectrochim. Acta 2002, 57B, 1277-1289. (6) Bermejo-Barrera, P.; Moreda-Pineiro, J.; Moreda-Pineiro, A.; BermejoBarrera, A. Talanta 1996, 43, 35-44. (7) Dadfarnia, S.; Haji Shabani, A. M.; Gohari, M. Talanta 2004, 64, 682687. (8) Ceresa, A.; Radu, A.; Peper, S.; Bakker, E.; Pretsch, E. Anal. Chem. 2002, 74, 4027-4036. (9) Ceresa, A.; Bakker, E.; Hattendorf, B.; Guenther, D.; Pretsch, E. Anal. Chem. 2001, 73, 343-351. (10) Radu, A.; Telting-Diaz, M.; Bakker, E. Anal. Chem. 2003, 75, 6922-6931. (11) Malon, A.; Radu, A.; Qin, W.; Qin, Y.; Ceresa, A.; Maj-Zurawska, M.; Bakker, E.; Pretsch, E. Anal. Chem. 2003, 75, 3865-3871. (12) Sokalski, T.; Ceresa, A.; Zwickl, T.; Pretsch, E. J. Am. Chem. Soc. 1997, 119, 11347-11348. (13) Sokalski, T.; Zwickl, T.; Bakker, E.; Pretsch, E. Anal. Chem. 1999, 71, 1204-1209. 10.1021/ac050856s CCC: $30.25

© 2005 American Chemical Society Published on Web 07/07/2005

bias, even with very high membrane selectivities, and observed detection limits are typically still much higher than their thermodynamically predicted ones.8 ISEs with a solid inner contact are promising alternatives in this regard, but their detection limits have so far remained in the nanomolar range as well.16 Corresponding ion-selective bulk optodes contain membrane components similar to those of their ISE counterparts and normally function on the basis of a competitive ion-exchange or coextraction mechanism.17,18 These true two-phase sensing principles are conveniently miniaturizable and have the promise of reaching detection limits that reflect the thermodynamically predicted values. Indeed, extremely selective and sensitive optode films have been prepared for determining subnanomolar levels of heavy-metal ions.19-21 Still, thin optode films have suffered from some limitations, such as very long response times in highly diluted sample solutions.20 This may be overcome by decreasing the volume of the sensing phase.22 Today, bulk optodes are attractive because they can be easily miniaturized and mass fabricated.22-25 Here we introduce microarray-based silver-selective microspheres for the trace analysis of silver ions based on the ionophore o-xylylenebis(N,N-diisobutyldithiocarbamate)) (Cu-I),26 originally introduced as an ionophore for divalent copper. Fluorescent and potentiometric sensors for the detection of silver with extremely high selectivity and detection limits in the 10-10 M range at physiological pH are reported here for the first time. EXPERIMENTAL SECTION Reagents. Poly(vinyl chloride) (PVC), bis(2-ethylhexyl) sebacate (DOS), Cu-I, and 9-(diethylamino)-5-[4-(15-butyl-1,13dioxo-2,14-dioxanodecyl)phenylimino]benzo[a]phenoxazine (ETH 5418, chromoionophore VII), were purchased from Fluka (Milwaukee, WI). The internal reference dye 1,1′′-dioctadecyl-3,3,3′,3′tetramethylindocarbocyanine perchlorate (DiIC18) was from Molecular Probes (Eugene, OR); sodium tetrakis[3,5-bis(trifluoromethyl)phenyl]borate (NaTFPB) was purchased from Dojindo Laboratories (Gaithersburg, MD). Tetrahydrofuran (THF) (Fluka), ethyl acetate (Fisher), xylenes (EM Sciences), cyclohexanone (99.8%) (Aldrich Baker), poly(ethylene glycol) (PEG) (Polysciences, Inc.), tris(hydroxymethyl) aminomethane (Tris) (Sigma), 3-morpholinopropanesulfonic acid (MOPS), and magnesium acetate (Fluka) were reagents purchased from the indicated suppliers. The cation-exchanger resin (Dowex C-350 H+ form, 30(14) Sokalski, T.; Ceresa, A.; Fibbioli, M.; Zwickl, T.; Bakker, E.; Pretsch, E. Anal. Chem. 1999, 71, 1210-1214. (15) Ion, A. C.; Bakker, E.; Pretsch, E. Anal. Chim. Acta 2001, 440, 71-79. (16) Sutter, J.; Radu, A.; Peper, S.; Bakker, E.; Pretsch, E. Anal. Chim. Acta 2004, 523, 53-59. (17) Seiler, K.; Simon, W. Anal. Chim. Acta 1992, 266, 73-87. (18) Bakker, E.; Buehlmann, P.; Pretsch, E. Chem. Rev. 1997, 97, 3083-3132. (19) Lerchi, M.; Bakker, E.; Rusterholz, B.; Simon, W. Anal. Chem. 1992, 64, 1534-1540. (20) Lerchi, M.; Reitter, E.; Simon, W.; Pretsch, E.; Chowdhury, D. A.; Kamata, S. Anal. Chem. 1994, 66, 1713-1717. (21) Lerchi, M.; Orsini, F.; Cimerman, Z.; Pretsch, E.; Chowdhury, D. A.; Kamata, S. Anal. Chem. 1996, 68, 3210-3214. (22) Telting-Diaz, M.; Bakker, E. Anal. Chem. 2002, 74, 5251-5256. (23) Tsagkatakis, I.; Peper, S.; Bakker, E. Anal. Chem. 2001, 73, 315-320. (24) Tsagkatakis, I.; Peper, S.; Retter, R.; Bell, M.; Bakker, E. Anal. Chem. 2001, 73, 6083-6087. (25) Wygladacz, K.; Bakker, E. Anal. Chim. Acta 2005, 532, 61-69. (26) Kamata, S.; Murata, H.; Kubo, Y.; Bhale, A. Analyst 1989, 114, 10291031.

80 mesh) was purchased from Fluka. All solutions were prepared with freshly deionized water (18 MΩ cm) using a Nanopure Millipore water purification system. Salts, acids, and bases of the highest available quality were used. Potentiometric Sensors. ISE Membranes. The membrane used for the fabrication of the silver-selective electrodes contained 12.38 mmol/kg (0.64 wt %) Cu-I, 5.53 mmol/kg (0.51 wt %) NaTFPB, 32.60 wt % PVC, and 66.25 wt % DOS. The components totaling 417 mg were dissolved in 5.0 mL of THF and poured into a glass ring (i.d. 70 mm), giving ∼200-µm-thick membranes. The membranes used for complex formation constants determination contained 8.94 mmol/kg (0.46 wt %) Cu-I, 4.62 mmol/kg (0.43 wt %) NaTFPB, 31.32 wt % PVC, and 67.80 wt % DOS and a reference membrane of the same composition but without ionophore. Electrodes. A 6-mm-diameter disk was punched from the membrane and glued to plasticized PVC tubing with a THF/PVC slurry, and the electrode assembled as described.10 The membrane was contacted with 10-3 M AgNO3 as inner and outer conditioning solutions overnight. Subsequently, the inner solution was changed to 4 × 10-4 M AgNO3 + 10-4 M NaNO3 equilibrated with 0.5 g of resin in 5 mL of H2O providing 3 × 10-7 M free Ag+, and the electrode was further conditioned in a 10-5 M NaNO3 + 10-9 M AgNO3 solution for ∼24 h. The free Ag+ concentration in the inner solution was calculated with the selectivity coefficient of the resin (KAg,Na ) 1.86) and a dry resin capacity of 5.14 mequiv/g, determined as explained elsewhere.27 The electrode was then fully assembled by inserting a plastic pipet tip filled with a cotton plug into the back of the electrode tube,8 and the top compartment was filled with 10-3 M NaCl as a bridge electrolyte. The electrodes for the determination of the complex formation constant were prepared by conditioning both parent membranes in 10-3 M AgNO3 overnight and then punching 6-mm-diameter disks that were mounted into an IS-561 Philips electrode body (Moeller, Zurich, Switzerland) as described.28 EMF Measurements. Potentials were monitored through a PCI MIO16XE data acquisition board (National Instruments, Austin, TX) utilizing a four-channel high Z interface (WPI, Sarasota, FL) at room temperature (22 °C), in the galvanic cell: Ag/AgCl/3 M KCl/1 M LiOAc/sample solution/ISE membrane/IFS/10-3 M NaCl/AgCl/Ag with a double-junction reference electrode (type 6.0729.100, Methrom AG, CH-9101 Herisau, Switzerland). The experiments were performed in a 100-mL polyethylene beaker pretreated overnight in 0.1 M HNO3. All EMF values were corrected for liquid-junction potentials according to the Henderson equation. Activity coefficients were calculated by the DebyeHu¨ckel approximation. Rotating electrode experiments were performed with a Pine Instrument analytical rotator (model ASR, Grove City, PA) and an ASR motor control box (1000 rpm/V, 20010000 rpm range) for all experiments, with some adjustments described elsewhere.29 Optical Sensors. Particle Preparation. Fluorescent microspheres were prepared using a previously described sonic particlecasting apparatus.22,24,25 The casting procedure is based on the coexistence of two streams, a diluted membrane cocktail (core) and purified water (sheath). The membrane cocktail containing (27) Qin, W.; Zwickl, T.; Pretsch, E. Anal. Chem. 2000, 72, 3236-3240. (28) Qin, Y.; Mi, Y.; Bakker, E. Anal. Chim. Acta 2000, 421, 207-220. (29) Ye, Q. S.; Meyerhoff, M. E. Anal. Chem. 2001, 73, 332-336.

Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

4707

the sensing ingredients 156 mmol/kg Cu-I, 10 mmol/kg ETH 5418, 2.3 mmol/kg DiIC18, and 14 mmol/kg NaTFPB was dissolved in 2.5 mL of cyclohexanone, diluted with 50 mL of ethyl acetate, and filtered with a 0.45-µm syringe filter to remove any solid impurities. Both streams were directed to the mixing chamber of the particle caster. The organic core stream was broken into droplets by oscillating a piezoelectric crystal and formed polymeric particles after curing. The following setup was applied: a ceramic tip with a 0.0018-in. diameter orifice, a 0.75 mL/min water stream flow rate, 21-kHz frequency, and 0.58 mL/ min polymer flow rate. Microspheres suspended in the receiving water phase were collected in 10-mL glass vials. To prevent their agglomeration, a 3% v/v PEG surfactant solution was continuously added. After casting, the microspheres were stored in the dark for at least 7 days in order to remove residual solvent from the particles and to allow the microspheres to settle to the bottom of the glass vials. Microsensor Platform Preparation and Microsphere Deposition. Commercially available 3200-µm-diameter imaging fibers (Edmund Industrial Optics) comprising ∼7000 individual fibers with 12-µm core diameters were used as a sensing platform for the microsensor array preparation. The microwells were formed by wet chemical etching in a buffered HF solution.25 After rinsing with tap and deionized water, the fiber was left to dry. Subsequently, a 2-µL aliquot of the particle suspension was deposited onto the distal face of the fiber bundle. After 3-min settlement time, the fiber was washed in distilled water for 1 min to remove unbound microspheres. The fiber was confirmed to contain at most 5-10 particles (this was desired in order to minimize sample depletion), immersed in distilled water for another 24 h for further curing, and characterized. For the imaging experiments on glass slides, a 30-µL aliquot of the microsphere suspension was deposited onto Fisher brand microscope cover glass slides. After 10 min, the slides with adhered particles were cured in distilled water for ∼24 h and characterized. Sample Solutions. Calibrations curves and selectivities were recorded in 1 mM buffer solutions (Tris-HCl or MOPS-NaOH for pH 7.4 and magnesium acetate buffer for pH 4.7) containing the appropriate electrolytes. All calibrating solutions were performed in polyethylene beakers that had been pretreated with 0.01 M HNO3. The calibration curve for silver was recorded in a 1.5 mM sodium background. Instruments. A PARISS imaging spectrometer (Light Form, Belle Mead, NJ) in combination with a Nikon Eclipse E400 microscope was used to characterize the microspheres and imaging fibers, as described.23 Scanning electron microscopy (Zeiss DSM 940 scanning electron microscope) and atomic force microscopy (Autoprobe CP atomic force microscope, Park Scientific Instruments) were used to characterize the etched optical fiber bundle. Procedures. Microspheres immobilized on the covered glass slides were normally equilibrated in 100-mL buffered sample solutions for ∼60 min before measurement. Calibration curves were recorded in 10-11-10-7 M buffered AgNO3 solutions at both pH 4.7 and 7.4. Selectivity coefficients were evaluated using the separate solutions method as the horizontal distance between logarithmic activities of primary and interfering ions at R ) 0.5 4708

Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

and at pH 7.4, as established.30 The exposure time for the fluorescence data acquisition was 200 ms. To minimize photobleaching, neutral density filters ND 4 or 8 were used. Full protonation and deprotonation spectra of the chromoionophore ETH 5418 were recorded at 10 mM HCl and 10 mM NaOH, respectively, and ratiometric measurements were performed by comparing the fluorescence emission peaks of ETH 5418 and the reference dye DiIC18 at 708 and 608 nm, respectively. Microspheres incorporated onto the etched wells of the fiber were characterized at pH 7.4 in the range of 10-12-10-8 M silver nitrate. During all measurements, the distal face of the fiber covered with the sensing microspheres was immersed in a 1.5-mL sample container. Excitation and fluorescence emission collection were performed as above, but from the backside of the fiber. Pond Water. A 500-mL water sample was collected from a nearby pond, filtered to eliminate visible impurities, and buffered with MOPS/NaOH to pH 7.4. Silver was measured by direct determination of the buffered pond water sample and by spiking with solutions containing known amounts of silver nitrate: 10 µL of 10-6 M AgNO3 (aliquot 1) followed by 25 µL of 10-5 M AgNO3 (aliquot 2) added to a 20-mL buffered sample. RESULTS AND DISCUSSION The ionophore o-xylylenebis(N,N-diisobutyldithiocarbamate) was originally reported by Kamata as a copper(II) selective ionophore.26 The resulting membrane selectivity was, at the time, characterized by the fixed interference method under routine conditions, and a reasonable preference for copper ions over other common interferences was reported. Since this ionophore has a structure related to a commercially available silver ionophore (S,S′methylenebis(diisobutyldithiocarbamate)),31,32 its characteristics as a silver ionophore were explored here for the first time. Unbiased selectivity determinations were performed by conditioning the membrane in a sodium nitrate solution before recording calibration curves of separate solutions of interfering ions in the order shown in Table 1, followed by a silver nitrate calibration curve.32 As shown in Figure 1, near-Nernstian response slopes were observed, suggesting unbiased selectivity coefficients (Table 1). Surprisingly, this ionophore does not appear to exhibit a preference for copper. The most preferred ion is silver, with a selectivity that surpasses that of recent work reporting silver ISEs with low detection limits.8 Before attempting to explore this ionophore in fluorescent microsphere arrays, ISEs with low detection limits were prepared for comparative purposes. In analogy to earlier work,8,11,27 the inner solution contained an ion-exchange resin that was calculated to give a robust logarithmic molar detection limit of -10.3 in a 10 µM sodium nitrate background. Calculations were done with eq 15 from ref 8 with diffusion coefficients in the aqueous and organic phases of 1.65 × 10-5 and 1 × 10-8 cm2 s-1, respectively, a 200µm membrane thickness, a 33-µm Nernst diffusion layer thickness at 2000 rpm,10 a selectivity coefficient of silver over sodium of log pot KAg,Na ) - 9.4 (Table 1), and a lipophilic ion-exchanger concentration of 5.3 mmol/kg. Figure 2 shows the corresponding experimental response curve. The observed detection limit of log c(DL) ) -9.5 was somewhat higher than predicted (-10.3). (30) Bakker, E.; Simon, W. Anal. Chem. 1992, 64, 1805-1812. (31) Mi, Y.; Bakker, E. Anal. Chem. 1999, 71, 5279-5287. (32) Bakker, E. Anal. Chem. 1997, 69, 1061-1069.

Table 1. Experimental Selectivity Coefficients for Microspheres based Silver-selective Optodes (log kIOsel ,J ) and Corresponding Ion-selective Electrode (ISE, log a KIpot ,J ) selectivity coefficients log ion J Ag+ Mg2+ Cd2+ Ca2+ Na+ K+ Li+ Cu2+ Hg2+ H+

pot KAg,J

Osel log kAg,J ; pH 7.4, R ) 0.5

ISE

fiber wells

glass slides

0.0 -11.6 -11.3 -11.2 -9.4 -8.9 -9.7 -11.1 n.d. -9.4

0.0 -11.5 -10.6 -10.3 -10.0 -9.9 -9.6 -7.3 -4.0 n.d.

0.0 -10.2 -10.4 -8.5 -9.3 -9.4 -8.5 -8.2 -3.9 n.d.

a Note that the selectivity coefficients for divalent ions are not directly comparable between ISEs and optodes.37

Figure 2. EMF response of the optimized silver-selective electrode based on Cu-I in a 10-5 M NaNO3 background, exhibiting a logarithmic detection limit of -9.5.

µm wells. Fluorescent plasticized PVC microspheres were doped with the silver ionophore (L), a H+-selective fluoroionophore (Ind), and a lipophilic cation-exchanger (R-) to function according to the following established ion-exchange sensing mechanism:18

Ag+(aq) + L(org) + IndH+(org) ) AgL+(org) + Ind(org) + H+(aq) (1)

Figure 1. Determination of unbiased selectivity coefficients of a silver-selective electrode containing Cu-I in separate nitrate solutions of the indicated cations (see also Table 1). Conditioning and inner solutions were 10-3 M NaNO3.

Recent work has shown that altering the aqueous diffusion layer thickness by rotating the electrode in the absence of primary ions in the sample may give potential changes that are indicative of the level of optimization of the inner solution.10 Large potential changes indicate strong electrolyte coextraction from the inner membrane side, for instance. In keeping with this work,10 the electrode was rotated repeatedly between 100 and 2000 rpm in a 10 µM sodium nitrate background in the absence of silver ions.10 As established,10 a well-optimized membrane is expected to show a 19.3-mV potential difference between these two rotational speeds. Here, however, the stir effect was found to be just 2.3 mV. Since no super-Nernstian response slope is observed in Figure 2 that could otherwise explain this small stir effect, it is possible that low levels of silver impurities in the sample were present. The observed potential difference may be explained with a silver impurity level in the sample of ∼3 × 10-10 M. The excellent detection limit and selectivity motivated us to explore this ionophore in fluorescent microspheres deposited onto an imaging fiber microarray comprising ∼7000 individual optical fibers, the distal end of which was chemically etched to form 12-

The concentration and binding characteristics of all sensing components determine the analytical response range. A fluoroionophore with a low pKa value is often desirable because it allows one to shift the response range to lower silver concentrations by shifting the equilibrium described by eq 1 to the right.21 The previously reported azo dye ETH 5315 with a pKa of 5.533 was not suitable, however, because it exhibits poor fluorescence emission properties.22 The more basic fluoroionophore ETH 5418 with a pKa of 8.8 33 was chosen instead. The silver response range was shifted to low concentrations by using a 16-fold excess of ionophore relative to indicator. An internal reference dye (DiIC18), emitting at 608 nm, was included in the composition because ETH 5418 shows only a single emission peak for its protonated form around 708 nm. Figure 3a shows fluorescence images for the 10µm beads after appropriate curing. The particles were found to be spherical and uniform in size, and their surface appeared to be smooth. The 3D renderings of the fluorescence emission spectra collected from the sensing microspheres exposed to 10-2 M HCl and 10-2 M NaOH are illustrated in Figure 3b and c, respectively. The microspheres appeared to be sufficiently bright, and a proper dye distribution within each polymeric particle was confirmed in fluorescence mode. Initial response studies were performed on microspheres deposited on cover glass slides at pH 4.7 and pH 7.4, respectively, as shown in Figure 4. The theoretical optode response function for this case is described by18,30

(1 -R Ra )L

-1 aAg+ ) (KAg exch)

H+

T

RT- - (1 - R)IndT - (RT- - (1 - R)IndT)

(2)

where KAg exch is the ion-exchange constant and subscripts T Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

4709

Figure 4. Response function of silver-sensing particles on glass slides to AgNO3 solutions in Mg(OAc)2 buffer at pH 4.7 and in MOPS buffer at pH 7.4, and theoretical response (solid lines) calculated according to eq 2. Both curves are described with log KAg exch ) 0.0 (eq 2).

Figure 3. Fluorescence image of 10-µm-diameter silver-selective particles deposited on a cover glass slide (A) and 3-D rendering of the fluorescence spectra of single microspheres exposed to 10-2 M HCl (B) and 10-2 NaOH (C).

denote total concentrations. The fluorescence signal is expressed as the mole fraction of unprotonated fluoroionophore, R, which is related to the fluorescence signal as24

R) 4710

(

)

Rmax - R [Ind] ) 1+ IndT R - Rmin

-1

Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

(3)

where R, Rmin, and Rmax are the fluorescence intensity ratios (at two wavelengths) for a given equilibrium and at minimum and maximum protonation of the fluoroionophore, respectively. The experimental data points were compared to the theoretical curve by assuming a 1:1 stoichiometry (eq 2). The error bars (calculated as standard deviations from the obtained data expressed as 1 - R) were quite small, demonstrating good sensing reproducibility of the Ag+-selective fluorescent microspheres (see Figure 4). Note that the same exchange constant log KAg exch ) 0.0 (eq 2) was used to describe the experimental data at two pH values, confirming that the designed system obeys classical optode theory with attractive tunability of the measuring range for silver ions.19 The detection limit of silver-selective optodes may be described on the basis of a number of definitions reported earlier.18-20,34 Briefly, they are calculated (1) by considering the standard deviation of the background noise in analogy to other analytical methods,19 (2) by assuming a limiting response slope,18 or (3) by considering the level of background interference in analogy to established ISE practice.18,34 This last method was chosen here since the comparison of ISE and optode detection limits was a primary focus of this work. For the microspheres characterized at pH 7.4, a logarithmic detection limit of log aAg(DL) ) -10.4(4 × 10-11 M) was determined (Figure 4). This is clearly superior to the characteristics of the corresponding ISE (see Figure 2, giving log aAg(DL) ) -9.5; 3 × 10-10 M). Note that the electrolyte background chosen for the optode experiment was nearly 100fold more concentrated than for the ISE, meaning that a 10-fold lower detection limit was achieved under more challenging selectivity conditions. The system was further characterized by evaluating the stability constant of the silver-ionophore complex in the membrane phase. For this purpose, the so-called sandwich membrane technique was used, which records the potential of a fused sandwich membrane, with only one side containing the ionophore.28,31 The membrane potential is related to the complex formation constant of the ionophore complex because the activity (33) Bakker, E.; Lerchi, M.; Rosatzin, T.; Rusterholz, B.; Simon, W. Anal. Chim. Acta 1993, 278, 211-225. (34) Bakker, E.; Willer, M.; Pretsch, E. Anal. Chim. Acta 1993, 282, 265-271.

Figure 5. Response time behavior of a silver-selective microsphere on an optical fiber bundle as a function of the indicated logarithmic molar silver concentrations at pH 7.4.

ratio in both membrane halves reflects the effective binding strength between the primary ion and the ionophore. Here, a potential difference of ∆E ) 401.05 ( 0.6 mV was observed, which translates into a complex formation constant of log β ) 9.22 ( 0.04. The stability constant was also estimated from the exchange constants of the optode response curve, calculated according to eq 15 in ref 35 with - log KIexch ) 8.833 and log KAg exch ) 0.1 (experimental data), and a similar complex formation constant of log β ) 8.9 was calculated. These data point to a relatively strong stabilization of silver compared to other common ionophores,28,31,36 although the silver ionophore S,S′-methylenebis(diisobutyldithiocarbamate was reported to exhibit even stronger binding characteristics (500 mV, log β ) 12.4).31 Response Time. In the past, classical thin optode films exhibited extremely long response times in highly diluted samples. Indeed, optode membranes proposed earlier for the analysis of nanomolar levels of Ag+ required ∼8 h to reach a stable response.20 The key limitation was recognized as the slow mass transport from the dilute aqueous sample to reach a required equilibrium preconcentration at millimolar levels in the sensing film. Furthermore, equilibration of the films required very large sample volumes. As demonstrated in Figure 5, typical response times of a single sensing particle incorporated into the well of an etched optical fiber are much shorter. The microspheres studied here were exposed to the test samples at pH 7.4 with silver concentrations ranging from 10-12 to 10-8 M, which are 3 orders of magnitude lower than that obtained earlier for silver-selective thin optode films. The 10-µm particle exposed to Ag+ ion concentrations of 10-12 and 10-11 M was found to exhibit a response time of ∼2 h. The required equilibration time was significantly shortened with higher concentrations. For instance, at 10-10 M concentrations, the time needed for complete response was ∼35 min, less than one-tenth the response time of 500 min reported by Lerchi et al. at a silver concentration of 3 × 10-9 M, 30 times higher than the concentrations used here. The response times for 10-8 and 10-9 M silver ions were recorded as 5 and 15 min, respectively, which are attractive values for routine analysis. Micrometer-sized sensing particles may clearly exhibit drastically shortened response times relative to their thin-film counter(35) Bakker, E.; Willer, M.; Lerchi, M.; Seiler, K.; Pretsch, E. Anal. Chem. 1994, 66, 516-521. (36) Ceresa, A.; Pretsch, E. Anal. Chim. Acta 1999, 395, 41-52.

Figure 6. Silver calibration curve and ion selectivity observed for microspheres incorporated into the etched wells of an optical fiber bundle (compare to Figure 4). Silver calibration curve with log KAg exch ) 0.1 (eq 2).

parts. Indeed, a 10-µm-diameter sensing particle (5 × 10-13 L) requires just 1 × 10-15 mol of silver ions (at 2 mM) for full equilibration. If it is tolerable to accept a 10% sample depletion of a 1 nM sample for accurate measurement, a mere 10-µL sample volume may be sufficient in an analysis. In contrast, required volumes for thin optode films for the measurements of such concentrations were in the multiliter range.20 The detection limit found here is perhaps the best so far reported in the literature for any ionophore-based optode. It is ∼1 order of magnitude lower than that of thin optode films reported earlier20 and surpasses the detection limit of the corresponding ISEs by at least 0.7 orders of magnitude (see Figure 2), which was measured with a 100-fold lower level of interfering ion. At a nanomolar silver sample concentration, the ISE required ∼30 min to lower the observed potential drift to 0.2 mV/min. Additionally, ISE experiments were carried out in 100-mL volumes, while the microspheres were measured in 1.5-mL samples. Hopefully, efforts in ISE miniaturization and elimination of the inner solution by a solid inner contact16 will further improve the sensing characteristics of potentiometric sensors. Currently, however, the properties listed above make fluorescent bulk optode microspheres a more powerful tool for trace analysis applications than ISEs. Additionally, they can be mass produced, easily miniaturized, integrated into optical fiber arrays as well as fluidic devices, and appear to be highly adaptable building blocks for more complex, integrated analytical instrumentation. Selectivity. Figure 6 illustrates the recorded calibration curves and observed selectivity for 10-µm-sized silver-selective microspheres in 1 mM Tris buffer at pH 7.4. The selectivity coefficients were determined as the horizontal distance between corresponding curves at R ) 0.5.30 Clearly, all interfering ions were extremely well discriminated, including mercury, which was expected to be a potential competitor for the silver ions. No additional experiments to evaluate the potential interference by mercury were performed, however. The calculated logarithmic selectivity coefOsel ficients,37 log kAg,I , are summarized in Table 1. Note that the Osel values of log kAg,I for the microspheres incorporated into fiber wells are in accordance with those observed for the microspheres from the same batch but characterized on microscope glass slides. (37) Bakker, E. Anal. Chim. Acta 1997, 350, 329-340.

Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

4711

The values were also found to be in good agreement with data obtained for ion-selective electrodes (Table 1). As an exploration to apply the fabricated silver-selective microspheres to measure free silver ion content in real aqueous samples, water samples were taken from a nearby pond and buffered with MOPS and NaOH to pH 7.4. The calculated free silver ion activity for the pond water sample was 2.00 × 10-10 M (RSD ) 12%, n ) 15 particles), according to the theoretical curve fit with log KAg exch ) 0, the value used for Figures 4 and 5. Two additions of known amounts of silver ions (final concentrations 5.0 × 10-10 and 1.34 × 10-8 M) to the buffered pond water gave increased silver responses, and the observed free silver ion -10 (RSD ) activities (with log KAg exch ) 0) were found as 8.91 × 10 -8 17%, n ) 15) and 1.41 × 10 M (RSD ) 16%, n ) 15), respectively. This suggests that silver complexation at subnanomolar levels in the studied pond water sample is insignificant. Note that the particle-to-particle reproducibility of the activity measurement at 0.2 nM (22 ppt) was just 12%, suggesting high analytical promise for this class of chemical sensors. CONCLUSIONS The experimental bias often observed in the past in the determination of selectivity coefficients of ISEs makes it necessary to reevaluate many existing ionophores in terms of their selectivity. We showed that a commercially available copper(II)-selective ionophore is, in fact, strongly silver selective. Its unbiased selectivity coefficients are among the best reported thus far. The lowest reported detection limit for any silver-selective ISE was observed here with this ionophore (log DL ) -9.5). Such potentiometric sensors, however, still comprise a three-phase system (sample, membrane, inner solution), and the detection limit is normally not dictated by simple ion-exchange processes

4712

Analytical Chemistry, Vol. 77, No. 15, August 1, 2005

alone but strongly influenced by ion fluxes across the ion-selective membrane. Further reduction of detection limits with such devices is only possible if these fluxes are drastically reduced or even eliminated. Thin-film optodes are an attractive approach, since they use the same sensing components but comprise a simple twophase equilibration system. In the past, however, the large surface area and sensing volume made true trace level measurements impractical. Silver-selective fluorescent microsphere sensing arrays deposited onto the distal end of an optical fiber bundle solve this problem. The characteristics of such a miniaturized sensing platform were compared to the corresponding ISEs and an even lower detection limit of less than 10-10 M, or low parts per trillion levels, was found. The observed selectivity was comparable to data obtained for ISEs. The much faster response times observed at nanomolar concentration levels (∼15 min) is credited to the drastically reduced sensing volume, which now requires only ∼1 fmol to be extracted into the microsphere. This translates into required sample volumes that are only in the microliter range. The attractive sensing characteristics, ability of mass production, tunable miniaturization, and integration into more complex instrumentation, as well as excellent sensor reproducibility (RSD ) 12% around the detection limit with n ) 15), makes this technology very promising for practical use in trace level analysis. ACKNOWLEDGMENT The authors thank the National Institutes of Health (DE14950 and EB002189) and Beckman-Coulter, Inc. for financial support of this research. Received for review May 17, 2005. Accepted June 10, 2005. AC050856S