Substrates with Discretely Immobilized Silver Nanoparticles for

Creighton, J. A.; Blatchford, C. G.; Albrecht, M. G. J. Chem. Soc., Faraday Trans. 2 1979, 75, 790−798. [Crossref], [CAS]. (51) . Plasma resonance e...
0 downloads 0 Views 186KB Size
Download PDF
Langmuir 2008, 24, 4765-4771

4765

Substrates with Discretely Immobilized Silver Nanoparticles for Ultrasensitive Detection of Anions in Water Using Surface-Enhanced Raman Scattering Siliu Tan,† Melek Erol,‡ Svetlana Sukhishvili,*,‡ and Henry Du*,† Departments of Chemical, Biomedical, and Materials Engineering and of Chemistry and Chemical Biology, SteVens Institute of Technology, Hoboken, New Jersey 07030 ReceiVed December 7, 2007. In Final Form: February 5, 2008 Positively charged silver nanoparticles, Ag [+], obtained by UV-assisted reduction of silver nitrate using branched poly(ethyleneimine) (BPEI) and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) solutions as reducing agents, were immobilized on glass surfaces to produce substrates active in surface-enhanced Raman scattering (SERS). Negatively charged silver nanoparticles, Ag [-], synthesized via a modified citrate reduction method, were also investigated for comparison. At a sparse surface coverage of 30 nanoparticles/µm2, substrates with immobilized Ag [+] showed increasing SERS sensitivity to a variety of anions in water in the order SO42- < CN- < SCN- ≈ ClO4-, with corresponding binding constants of 105, 3.3 × 105, and 107 (for both SCN- and ClO4-) M-1, respectively. This order followed the Hofmeister series of anion binding in water. Significantly, substrates with Ag [+] allowed limit of detection values of 8.0 × 10-8 M (8 ppb) and 2.7 × 10-7 M (7 ppb) for environmentally relevant perchlorate (ClO4-) and cyanide (CN-) anions, respectively. In contrast, substrates with immobilized Ag [-], even upon subsequent modification by a monolayer of BPEI for positive surface charge of the nanoparticles, showed a drastically lower sensitivity to these anions. The high sensitivity of substrates with Ag [+] for anion detection can be attributed to the presence of two types of functional groups, amino and amide, on the nanoparticle surface resulting from UV-assisted fragmentation of BPEI chains. Both amino and amide provide strong binding of anions with Ag [+] nanoparticles due to the synergistic effect through a combination of electrostatic, hydrogen bonding, and dispersive interactions.

1. Introduction Detection of various anions including perchlorate (ClO4-), cyanide (CN-), sulfate (SO42-), and thiocyanate (SCN-) in the parts per billion to parts per million concentration level is of great importance due to their hazardous nature to human health.1 Many techniques have been used for anion detection, including capillary electrophoresis,2-4 ion chromatography,5-8 and electrospray mass spectrometry.9-11 These techniques, while good laboratory analytical tools, are not particularly suitable for in situ on-site detection of anions due to the requirement of additional reagents and lengthy sample preparation or measurements. Raman spectroscopy, a well-established analytical tool for molecular identification, provides an excellent means for measurements of aqueous solutions due to low water background over a vast spectral range. Intrinsically low Raman scattering * To whom correspondence should be addressed. (S.S.) E-mail: ssukhish@ stevens.edu. Phone: (201) 216-5544. (H.D.) E-mail: [email protected]. Phone: (201) 216-5262. † Department of Chemical, Biomedical, and Materials Engineering. ‡ Department of Chemistry and Chemical Biology. (1) U.S. Environmental Protection Agency. National Primary Drinking Water Standards; EPA Documents No. 816-F-03-016; GPO: Washington, DC, 2003. (2) Yang, W.-P.; O’Flaherty, B.; Cholli, A. L. J. EnViron. Sci. Health, Part A 2001, 36, 1271-1285. (3) Liu, Y.; MacDonald, D. A.; Yu, X. Y.; Hering, S. V.; Collett, J. L., Jr.; Henry, C. S. Analyst 2006, 131, 1226-1231. (4) Fung, Y. S.; Lau, K. M. Electrophoresis 2003, 24, 3224-3232. (5) Tian, K.; Dasgupta, P. K.; Anderson, T. A. Anal. Chem. 2003, 75, 701706. (6) Okamoto, H. S.; Rishi, D. K.; Steeber, W. R.; Baumann, F. J.; Kusum Perera, S. J.sAm. Water Works Assoc. 1999, 91, 73-84. (7) Christison, T. T.; Rohrer, J. S. J. Chromatogr., A 2007, 1155, 31-39. (8) Tepus, B.; Simonc, M. Cent. Eur. J. Chem. 2007, 5, 557-569. (9) Magnuson, M. L.; Urbansky, E. T.; Kelty, C. A. Anal. Chem. 2000, 72, 25-29. (10) Winkler, P.; Minteer, M.; Willey, J. Anal. Chem. 2004, 76, 469-473. (11) Koester, C. J.; Beller, H. R.; Halden, R. U. EnViron. Sci. Technol. 2000, 34, 1862.

cross-sections, however, do not render this technique for detection of contaminants at low concentrations. The Raman scattering cross-section can be enhanced by up to 14 orders of magnitude in the presence of a nanostructured noble metal (such as Ag and Au) due to the strong surface plasmon field effect.12-14 Surfaceenhanced Raman spectroscopy (SERS) has thus become a very attractive alternative to fluorescence spectroscopy for ultratrace identification with the potential of single-molecule sensitivity.15,16 Indeed, SERS has been broadly explored for the detection and identification of chemical and biological species in liquid, including water.17-21 A variety of Ag and Au nanostructured substrates have been investigated for SERS measurements, including electrochemically roughened electrodes,22 chemically vapor deposited metal islands,23 highly ordered two-dimensional structures by electron beam lithography,24-26 nanosphere lithography,27,28 microsphere(12) Suh, J. S.; Moskovits, M. J. Am. Chem. Soc. 1986, 108, 4711-4718. (13) Campion, A.; Kambhampati, P. Chem. Soc. ReV. 1998, 27, 241-250. (14) Michaels, A. M.; Nirmal, M.; Brus, L. E. J. Am. Chem. Soc. 1999, 121, 9932-9939. (15) Kneipp, H.; Wang, Y.; Kneipp, K.; Perelman, L. T.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Phys. ReV. Lett. 1997, 78, 1667-1670. (16) Haynes, C. L.; McFarland, A. D.; Van Duyne, R. P. Anal. Chem. 2005, 77, 338 A-346 A. (17) Mosier-Boss, P. A.; Lieberman, S. H. Appl. Spectrosc. 2003, 57, 11291137. (18) Mosier-Boss, P. A.; Lieberman, S. H. Appl. Spectrosc. 2000, 54, 11261135. (19) Wang, W.; Gu, B. Appl. Spectrosc. 2005, 59, 1509-1515. (20) Yea, K. H.; Lee, S.; Kyong, J. B.; Choo, J.; Lee, E. K.; Joo, S. W.; Lee, S. Analyst 2005, 130, 1009-1011. (21) Gu, B.; Tio, J.; Wang, W.; Ku, Y. K.; Dai, S. Appl. Spectrosc. 2004, 58, 741-744. (22) Jeanmaire, D. L.; Van Duyne, R. P. J. Electroanal. Chem. 1977, 84, 1-20. (23) Zeisel, D.; Deckert, V.; Zenobi, R.; Vo-Dinh, T. Chem. Phys. Lett. 1998, 283, 381-385. (24) Gunnarsson, L.; Bjerneld, E. J.; Xu, H.; Petronis, S.; Kasemo, B.; Ka¨ll, M. Appl. Phys. Lett. 2001, 78, 802-804.

10.1021/la703831q CCC: $40.75 © 2008 American Chemical Society Published on Web 04/01/2008

4766 Langmuir, Vol. 24, No. 9, 2008

templated metal nanoparticle self-assembly,29 and colloidal nanoparticles.30,31 In this study, we have explored novel strategies in producing SERS-active substrates via nanoparticle immobilization for in situ detection of various anions in aqueous solutions. One crucial parameter to be considered in designing SERS substrates is the degree of enhancement of local electromagnetic fields in the vicinity of noble metal nanostructure. Such enhancement is determined by the size of individual nanofeatures and their arrangement on a substrate. Another prerequisite for analyte detection at low concentrations is “concentrating” analyte molecules in the vicinity of the nanostructured surface, typically within an interaction range of several nanometers32 for SERS probing. Evidently, such concentrating can be challenging in the case of an ultratrace analyte. One way to overcome this challenge is to dry analyte solutions on an SERS substrate prior to measurements as reported in perchlorate detection for a detection limit of ∼100 ppb.21,33,34 For in situ detection of anions in water, however, conventional Ag or Au nanostructures are not efficient, as they carry prohibitive negative surface charge. One approach to decrease the electrostatic repulsion between the anions and the negatively charged nanoparticles is to premix anions with positively charged molecules, as for the detection of ∼1 ppb CN-.17 Another strategy is to prepare a porous sol-gel matrix embedded with highdensity gold nanoparticles, a method that enabled detection of CN- at ∼9 ppb.35 Surface modification of SERS-active Ag or Au nanostructures offers significant potential for in situ detection of anions in water as no analyte pretreatment is required. Indeed, Mosier-Boss and Lieberman showed that modification of surfaces of roughened Ag/Au electrodes with self-assembled monolayers of cationic thiols was a necessary step for in situ detection of 84 ppm ClO4and 440 ppm SO42-.17,18 Gu et al.19,33,34 produced SERS substrates functionalized with -NH2, -N+(CH3)3, and cysteamine groups for detection of 100 ppb ClO4- in water.19 Low SERS sensitivity has also been attributed to strong electrostatic repulsions between analyte molecules and the nanoparticle surface, according to Aroca et al.36 We report in this paper the use of positively charged Ag nanoparticles immobilized on glass surfaces as SERS substrates for in situ measurements of ClO4-, SCN-, CN-, and SO42- in water with significantly improved sensitivity. A one-step photoreduction process using BPEI/AgNO3/HEPES solutions reported elsewhere37 was employed to produce positively charged Ag nanoparticles (denoted here as Ag [+]) carrying both primary amino and amide groups on their surfaces. The Ag [+] (25) Fe´lidj, N.; Aubard, J.; Le´vi, G.; Krenn, J. R.; Salerno, M.; Schider, G.; Lamprecht, B.; Leitner, A.; Aussenegg, F. R. Phys. ReV. B: Condens. Matter 2002, 65, 0754191-0754199. (26) Kahl, M.; Voges, E.; Kostrewa, S.; Viets, C.; Hill, W. Sens. Actuators, B 1998, 51, 285-291. (27) Hulteen, J. C.; Van Duyne, R. P. J. Vac. Sci. Technol., A 1995, 13, 15531558. (28) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2000, 104, 10549-10556. (29) Tessier, P. M.; Clemente, E. M.; Lenhoff, A. M.; Kaler, E. W.; Velev, O. D.; Christesen, S. D.; Ong, K. K. Appl. Spectrosc. 2002, 56, 1524-1530. (30) Nie, S.; Emory, S. R. Science 1997, 275, 1102-1106. (31) Pristinski, D.; Tan, S.; Erol, M.; Du, H.; Sukhishvili, S. J. Raman Spectrosc. 2006, 37, 762-770. (32) Neacsu, C. C.; Dreyer, J.; Behr, N.; Raschke, M. B. Phys. ReV. B: Condens. Matter 2006, 73, 193406-193404. (33) Wang, W.; Ruan, C.; Gu, B. Anal. Chim. Acta 2006, 567, 121-126. (34) Ruan, C.; Wang, W.; Gu, B. Anal. Chim. Acta 2006, 567, 114-120. (35) Premasiri, W. R.; Clarke, R. H.; Londhe, S.; Womble, M. E. J. Raman Spectrosc. 2001, 32, 919-922. (36) Alvarez-Puebla, R. A.; Goulet, P. J. G.; Aroca, R. F.; Arceo, E.; Garrido, J. J. J. Phys. Chem. B 2005, 109, 3787-3792. (37) Tan, S.; Erol, M.; Attygalle, A.; Du, H.; Sukhishvili, S. Langmuir 2007, 23, 9836-9843.

Tan et al.

nanoparticles exhibited strong affinity to anions due to the synergistic binding affinity of amino and amide groups on the nanoparticle surface toward anions. The substrates with immobilized Ag [+] allowed sensitive and reproducible detection of various anions. The limit of detection values were 1.7 × 10-8 M (1 ppb), 8.0 × 10-8 M (8 ppb), 2.7 × 10-7 M (7 ppb), and 4.2 × 10-5 M (4 ppm) for thiocyanate (SCN-), perchlorate (ClO4-), cyanide (CN-), and sulfate (SO42-) anions, respectively. 2. Experimental Section 2.1. Materials. The following reagents were purchased from the suppliers noted and were used without further purification: N-(2hydroxyethyl)piperazine-N′-2-ethanesulfonic acid (HEPES; reagent grade, Fisher Scientific), silver nitrate (ultrapure grade, Acros), branched poly(ethyleneimine) (BPEI) with a molecular mass of 1200 g‚mol-1 (Polysciences, Inc.), sodium citrate (enzyme grade, Fisher Scientific), sodium perchlorate (99+%, Acros), potassium cyanide (ACS grade, Fisher), sodium sulfate (99.99%, Aldrich), and sodium thiocyanate (99.99+%, Sigma-Aldrich). Water used for the experiments was filtered with Barnstead ion-exchange columns and then further purified by passage through Milli-Q (Millipore) deionizing and filtration columns. B694 n-type (100) silicon wafers with a thickness of 650 µm and resistivity of 7500-9500 Ω cm were purchased from El-Cat Inc., New Jersey. All glassware, glass, and silicon substrates were cleaned in Nochromix (Godax Laboratories, Inc., Maryland) solution in concentrated sulfuric acid overnight, followed by thorough rinsing with Milli-Q water. 2.2. Synthesis and Characterization of Ag Nanoparticles. For the synthesis of positively charged Ag nanoparticles, Ag [+], BPEI and AgNO3 were separately dissolved in 0.1 mM HEPES to yield the desired concentrations. The two solutions were then mixed (50: 50 by volume) to give the final BPEI:AgNO3:HEPES molar ratio of 1:1:0.1. Note that the molar concentration of BPEI repeating units was used throughout this paper. The Ag [+] nanoparticles were produced by exposing 10 mL of a BPEI/AgNO3/HEPES mixture in a glass vial (2 cm in diameter and 5 cm in height) to UV irradiation (BHK mercury grid lamps) for 120 min under continuous stirring, where the illumination occurred from the air side of the air/liquid interface. The standard low-pressure mercury arc lamp used for the production of UV light provides uniform, high-intensity irradiation at various mercury lines including the ozone-producing 185 nm line. Details of the synthesis and mechanisms involved have been described in our previous paper.37 In brief, UV-assisted oxidative chain cleavage of BPEI produced BPEI fragments which contain amide groups,37 as well as retain amino groups, yielding a positive surface charge at neutral pH values. For comparison, negatively charged Ag nanoparticles, denoted as Ag [-], were synthesized according to a modified Lee and Meisel method.38 Briefly, 14.4 mg of AgNO3 was dissolved in 80 mL of H2O, and then a solution of 1% sodium citrate (1.6 mL) was added. The solution was kept under UV light for 1 h with constant stirring. The ζ-potential values of the silver nanoparticles were measured by laser Doppler electrophoresis using Zetasizer Nano-ZS equipment (Malvern Instruments Ltd.). The results were averaged over 30 runs. Each value was obtained by taking the average of the measurements from three samples. Prior to the measurements, Ag [+] colloid was diluted 3 times and citrate-reduced Ag [-] colloid was diluted 10 times with Milli-Q water. Ag [+] and Ag [-] nanoparticles in water had ζ-potentials of +45 and -40 mV, respectively. The silver nanoparticle size and coverage were characterized using scanning electron microscopy (SEM; LEO 982). Ag [+] or Ag [-] nanoparticles were immobilized on oxidized silicon substrates as described in the next section for SEM images. The nanoparticle size and coverage were determined by SEM in conjunction with Digimizer image analysis software developed by MedCalc Software (Mariakerke, Belgium). For the BPEI/AgNO3/HEPES mixture at a molar ratio of 1:1:0.1, the nanoparticle size was 33 ( 8 nm (inset in Figure 1a). Ag [-] nanoparticles had a size of 42 ( 12 nm (inset in Figure (38) Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391-3395.

UltrasensitiVe SERS Detection of Anions in Water

Langmuir, Vol. 24, No. 9, 2008 4767 Scheme 1. Schematic Illustration of (a) Type I, (b) Type II, and (c) Type III Substrates

Figure 1. UV-vis absorption spectra of aqueous dispersions of (a) Ag [+] and (b) Ag [-] nanoparticles with nanoparticle sizes of 33 ( 8 and 42 ( 12 nm, respectively. Nanoparticle concentrations were ∼2 × 1010 particles/mL for Ag [+] and ∼3 × 1010 particles/ mL for Ag [-]. The insets show SEM images of the Ag nanoparticles immobilized on oxidized silicon following the procedures described in the Experimental Section for types I and II substrates. 1b). The solution particle densities for the colloidal solutions before dilution were 5.5 × 1010 and 3 × 1011 particles/mL for Ag [+] and Ag [-] colloidal solutions, respectively.37 The plasmon resonance of the silver nanoparticles was measured using a UV-vis spectrophotometer (Hitachi U3000). Prior to the measurements, Ag [+] colloid was diluted 3 times and citrate-reduced Ag [-] colloid was diluted 10 times with Milli-Q water. Both 33 nm Ag [+] and 42 nm Ag [-] nanoparticles had their UV-vis absorption band centered at ∼420 nm (Figure 1), despite their different sizes. The red shift of the Ag [+] nanoparticle plasmon peak is due to the increase in the refractive index of the surrounding medium39,40 of the nanoparticles caused by the presence of positively charged BPEI fragments adsorbed on the Ag [+] nanoparticle surface. This trend is consistent with our previous report,37 as well as with data by Gittins et al.41 In this study, we have used 33 nm Ag [+] and 42 nm Ag [-] nanoparticles with matching positions of the plasmon bands. 2.3. Immobilization of Ag Nanoparticles on Glass Substrates. Glass tubes (Bellco Glass, 8 mm in outer diameter and 8 mm in height) were adhesively bonded at one end with glass slides (EMS, 22 × 22 × 0.15 mm) to form water-tight sample cells for immobilization of silver nanoparticles on the glass substrates (bottom of the cell) for their later use as SERS substrates. Prior to their use, the glass tubes and slides were cleaned by UV exposure for several hours, followed by soaking in Nochromix solution in concentrated sulfuric acid overnight, and finally by copious rinsing with Milli-Q water. Three types of SERS substrates were produced. Type I substrate (Scheme 1a) was obtained by the adsorption of Ag [+] nanoparticles on the glass substrate by filling the sample cell with 0.2 mL of Ag [+] colloidal solution at pH 5 for 4 h followed by rinsing with Milli-Q water. Type II substrate (Scheme 1b) was prepared first by subjecting the glass surface to BPEI (MW of 1200) surface modification in a solution of 0.2 mg/mL BPEI in 0.1 mM HEPES at pH 8.3 for 20 min with excess BPEI removed from the substrate through Milli-Q water rinsing. The BPEI-modified glass surface (39) Mulvaney, P. Langmuir 1996, 12, 788-800. (40) Underwood, S.; Mulvaney, P. Langmuir 1994, 10, 3427-3430. (41) Gittins, D. I.; Caruso, F. J. Phys. Chem. B 2001, 105, 6846-6852.

was subsequently immobilized with Ag [-] nanoparticles using a 10× diluted Ag [-] colloid solution (particle concentration of 3 × 1010 particles/mL) in 10 mM HEPES buffer at pH 7 for 4 h. Type III substrate (Scheme 1c) was produced by applying an additional BPEI layer onto the type II substrate to mimic the behavior of their positively charged counterparts in Scheme 1a. This BPEI layer was adsorbed from a solution of 0.05 mg/mL BPEI at pH 8.3 (same molecular weight and concentration as used for the reduction of Ag [+] nanoparticles) for 20 min, again followed by removal of excess BPEI with Milli-Q water rinsing. All three substrate types had a comparable nanoparticle coverage density of ∼30 particles/µm2. 2.4. SERS Measurements. The Raman instrumental details are as follows. A 532 nm wavelength light beam from a Laserglow D1-532 laser was spatially filtered and expanded three times, bandpass filtered, and reflected from a Chroma Q540LP dichroic mirror and then illuminated the back aperture of an Olympus 40× objective, NA ) 0.85. The excitation light intensity in front of the objective was ∼10 mW. A sample cell, which was prepared according to the procedures described in section 2.3, was filled with analyte solution as its bottom was situated on a Newport ULTRAlign 561D translation stage equipped with New Focus 8301 computer-controlled piezoactuators. SERS measurements were conducted using the same objective of ∼1.5 µm focusing diameter for excitation and spectral collection. SERS signal from the objective passed through a dichroic mirror, was filtered by a Kaizer SuperNotch filter, and was focused by a collimator into a spectroscopic grade multimode fiber (Newport, 400 µm core). A fiber-coupled Acton SpectraPro 2300 spectrometer with a Roper Scientific liquid nitrogen cooled CCD detector was used in spectrum acquisition. Data were processed using Origin 7.0 software. For all SERS measurements, ∼0.2 mL of aqueous anion solution (100 ppt to 10 ppm; these concentrations were calculated per mass of anion, not for the salt) was added to the sample cell prepared according to the procedures described in section 2.3. Four to seven SERS spectra were collected with an exposure time of 100 s at different illumination spots for each substrate investigated.

3. Results and Discussion 3.1. Comparative SERS Study of Binding of Various Anions on the Type I Substrate. Figure 2 shows SERS spectra of several anions at different concentrations adsorbed on surface-immobilized Ag [+] nanoparticles of the type I substrate. Seen in the spectra are the Raman bands characteristic of symmetric stretching vibrations of various anions: ClO4- band at 921 cm-1, CN- band at ∼2120 cm-1, SCN- band at ∼2110 cm-1, and SO42- band at 1008 cm-1. Background bands at ∼1300-1800 cm-1 from carbon are sufficiently separated from features of interest and thus do not pose a challenge in data analysis. Note that Ag [+] nanoparticles synthesized via the reduction of AgNO3 solution carried nitrate anions as counterions, and anion binding was accompanied by the displacement of NO3-. While in the case of ClO4-, CN-, and SCN- there was no spectral overlap with the NO3- vibrational band at 1036 cm-1, the characteristic

4768 Langmuir, Vol. 24, No. 9, 2008

Tan et al.

Figure 2. SERS spectra of ClO4-, CN-, SCN-, and SO42- in water using the type I substrate with immobilized Ag [+] nanoparticles. The characteristic bands at ∼1008 cm-1 for SO42- anions overlap with that of NO3- at 1036 cm-1, and deconvolution analysis by Origin 7.0 was used to resolve those anion bands. Spectra were acquired using a 10 mW laser power at 532 nm excitation for 100 s.

Figure 3. SERS spectra of SO42- in water using the type I substrate with immobilized Ag [+] nanoparticles, showing the overlap of the ∼1008 cm-1 SO42- peak with the 1036 cm-1 NO3- band, and their deconvolution using Origin 7.0. Spectral acquisition conditions are the same as in Figure 2. The uncertainty originating from different parameters chosen for peak deconvolution was ∼10%. This uncertainty has been added to the experimental error bars in determination of SO42- coverage shown in Figures 4-6.

peak for SO42- at ∼1008 cm-1 overlapped with the 1036 cm-1 NO3- band. Consequently, spectral deconvolution was used for SERS data analysis of sulfate anions as illustrated in Figure 3. The peak area of the deconvoluted peak at ∼1008 cm-1 increased with an increase in SO42- concentration, indicating adsorption of this anion in the Stern layer of Ag [+] nanoparticles. At the same time, the intensity of the NO3- band at 1036 cm-1 decreased, reflecting replacement of NO3- by SO42- ions at the Ag [+] nanoparticle surface. For the case of adsorption of SO42- and SCN- anions, such replacement data are quantified in Figure 4. The data are presented as the fractional coverage of the adsorbed analyte (θ ) I/Imax, where I is the SERS intensity of a characteristic band, proportional to the amount of adsorbed anions, and Imax is the SERS intensity at saturated surface coverage) as a function of the solution concentration. The fractional intensities of the SO42- band at 1008 cm-1 and SCN- band at 2110 cm-1 strongly correlated with the decreasing NO3- peak intensity at 1036 cm-1. Figure 5 shows that, for all anions, the surface coverage increased linearly in the low concentration range of anions and that the slope of the adsorption was dependent on the type of anion. Note that, except SCN-, the binding of all other anions

Figure 4. Adsorption isotherms of SO42- and SCN- anions (shown on the top of panels a and b, respectively) and corresponding displacement of NO3- counterions by SO42- and SCN- shown on the bottom of the panels. Spectral acquisition conditions are the same as in Figure 2.

to the immobilized Ag [+] nanoparticles was reversible; i.e., adsorbed anions desorbed from the surface when rinsed with an excess amount of water. In all cases, the SERS intensity increased linearly with the anion concentration at lower analyte concentrations and saturated at higher concentrations, indicating saturation of the adsorption sites. The limit of detection (LOD) was determined from the linear slope of the anion adsorption isotherms using the following equation:42

LOD )

3σ m

(1)

where σ is the standard deviation in the y-intercept and m is the slope of the line (for the curve of integrated intensity versus analyte concentration). The LODs ranged from 1.7 × 10-8 M (∼1 ppb) for thiocyanate to 4.2 × 10-5 M (∼4 ppm) for sulfate (see Table 1). (42) Skoog, D. A.; Holler, F. J.; Nieman, T. A. Principles of Instrumental Analysis, 5th ed.; Harcourt Brace College Publishers: Philadelphia, 1998.

UltrasensitiVe SERS Detection of Anions in Water

Langmuir, Vol. 24, No. 9, 2008 4769

Frumkin equilibrium isotherm models43,44 were applied to the SERS data:

Langmuir isotherm:

1 1 ) +1 θ KLc

Frumkin isotherm: θ )

cKFe2gθ

or 1 + cKFe2gθ θ ) 2gθ + ln KF ln c(1 - θ)

(

Figure 5. Adsorption of (a) ClO4- and SCN- and (b) CN- and SO42- anions onto the type I substrate shown for low surface coverage (the complete isotherm is not shown). Adsorption was measured from the anion peak area, I. Each data point represents the average of four to seven measurements of integrated peak intensities at ∼921 cm-1 (ClO4-), 2120 cm-1 (CN-), 2110 cm-1 (SCN-), and 1008 cm-1 (SO42-) for each analyte, with error bars showing the standard deviation. Spectral acquisition conditions are the same as in Figure 2. Here and in Figures 6 and 7b, the experimental data are fitted using the Langmuir curve fitting function in the Origin 7.0 software package.

Figure 6. Adsorption isotherms of ClO4-, SCN-, CN-, and SO42on the type I substrate. Error bars show the standard deviation of the measurements. Spectral acquisition conditions are the same as in Figure 2. Table 1. Langmuir and Frumkin Binding Constants and LOD Values of Different Anions on the Type I Substrate type of anion -

ClO4 SCNCNSO42-

Langmuir constant, KL (M-1) 107 107 3.3 × 105 105

Frumkin constant, KF (M-1)/g value

LOD

2.2 × /-1.6 9.2 × 106/-1.6 2.75 × 105/-0.8 5.4 × 104/-0.6

8.0 × 10-8 M (8 ppb) 1.7 × 10-8 M (1 ppb) 2.7 × 10-7 M (7 ppb) 4.2 × 10-5 M (4 ppm)

107

Adsorption isotherms for all anions are plotted in Figure 6 (θ vs c, solution concentration of an anion). The Langmuir and (43) Han, R.; Zhang, J.; Zou, W.; Shi, J.; Liu, H. J. Hazard. Mater. 2005, 125, 266-271. (44) Goswami, S.; Ghosh, U. C. Water SA 2005, 31, 597-602.

(2)

)

(3)

where c is the equilibrium anion concentration in solution (M) and KL and KF (M-1) are the anion binding constants in the Langmuir and Frumkin models, respectively. The Langmuir isotherm is a simple model which assumes no interaction between the analyte molecules. The Frumkin isotherm is an extension of the Langmuir model that counts the interaction between the analyte molecules, reflected by the parameter g. A negative value of g obtained from Frumkin fitting (see Table 1) indicates the presence of a repulsive force between anion adsorbates. Both models gave satisfactory fits to the experimental data, yielding similar binding constants. Table 1 summarizes the data for binding constants of anions on the surface of a substrate of type I. Note that, in this paper, we present data on CN- adsorption, although Lieberman et al.17 showed that CN- anions react with Ag/Au substrates. Our control experiments showed that the SERS signal of CN- anions up to 3.8 × 10-5 M (1 ppm) concentration using substrate I was stable for 24 h. In this investigation, the CNanions were detected in the low concentration range of 3.8 × 10-8 to 3.8 × 10-5 M by SERS, and the surface-enhanced Raman spectrum at each concentration was collected within 10 min, allowing detection of CN-. The first feature obtained from Figures 2, 5, and 6 and Table 1 is that the binding of anions is strongly dependent on the chemistry rather than the anion charge. Specifically, multiply charged sulfate anions had weak binding at the Ag [+] surface despite their higher charge than other anions studied. Rather, the anion binding order follows the Hofmeister series,45,46 which is usually found for specific binding of anionic counterions with synthetic polycations, synthetic proteins, or amide group containing water-soluble polymers.47,48 This order reflects stronger condensation of “chaotropic”, less hydrated anions on the Ag [+] nanoparticle surface, as compared to “kosmotropic”, more hydrated anions. The recent view does not correlate the Hofmeister effect with the ion capacities to affect the interfacial water structure (and therefore argues against the kosmotropic or chaotropic effect of ions on the overall water structure).48,49 Instead, the importance of dispersion forces in the specificity of anion binding in an aqueous environment has been recently demonstrated in direct measurements of the degree of disordering of the hydrophobic regions of a monolayer caused by different anions.49 Typical “chaotropes” such as ClO4- or SCN- ions are therefore more hydrophobic and show stronger binding, while more hydrated “kosmotropes” such as SO42- and CN- have lower binding constants for the polymers in water. In addition, in the case of polymers which contain amide groups, a transition from ion binding of strongly hydrated kosmotropic ions with the mac(45) Ghimici, L.; Dragan, S. Colloid Polym. Sci. 2002, 280, 130-134. (46) Bostro¨m, M.; Williams, D. R. M.; Ninham, B. W. Biophys. J. 2003, 85, 686-694. (47) Hamabata, A.; Von Hippel, P. H. Biochemistry 1973, 12, 1264-1271. (48) Zhang, Y.; Furyk, S.; Bergbreiter, D. E.; Cremer, P. S. J. Am. Chem. Soc. 2005, 127, 14505-14510. (49) Gurau, M. C.; Lim, S. M.; Castellana, E. T.; Albertorio, F.; Kataoka, S.; Cremer, P. S. J. Am. Chem. Soc. 2004, 126, 10522.

4770 Langmuir, Vol. 24, No. 9, 2008

Tan et al. Table 2. Langmuir Binding Constants (M-1) and LOD Values for SCN- Anions on Substrates I-III substrate type I II III

Figure 7. (a) SERS spectra of 8.6 × 10-8 M (5 ppb) SCN- in aqueous solution using types I, II, and III substrates. (b) Adsorption isotherm of SCN- for different substrate types I-III for low anion concentrations (the complete isotherms are not shown). All SERS substrate types have a particle coverage density of ∼30 particles/µm2. Spectral acquisition conditions are the same as in Figure 2.

romolecule hydration shell to direct ion binding of chaotropic anions to the amide groups has been found.48 As we have shown previously, the surface of Ag [+] nanoparticles contains both amino and amide groups,37 and therefore, both dispersive forces and direct anion binding to amide moieties could be responsible for stronger binding of ClO4- and SCN- using the substrate of type I. The second feature is that the LOD values did not follow the binding constants obtained with the Ag [+]-immobilized substrate of type I, as shown in Table 1. While both Langmuir and Frumkin binding constants followed the series ClO4- ≈ SCN- > CN> SO42-, the LOD did not follow the same order and increased in the series SCN- < CN- < ClO4- < SO42-. This is explained by different noise levels, σ (see eq 1), of characteristic vibrational bands of various anions, resulting from the difference of the instrumental diffraction grating efficiency in different spectral regions. Specifically, significantly lower spectral noise at wavenumbers higher than 2000 cm-1 assured a better signalto-noise ratio in the detection of SCN- and CN- as compared to other anions and provided LODs for these anions significantly lower than those expected from binding constants. Note that using cationic coatings on the Ag/Au electrode and aqueous anion solutions, Lieberman et al.17 obtained binding constants for various anions from 220 to 7380 M-1, which are 1000 times lower than our values. Zhang et al.50 were able to obtain a much higher binding constant of 1013 M-1 for the case of subtilis spores, which, unlike monovalent anions, are capable of forming multiple adsorption points with an SERS substrate. The LOD values obtained with the Ag [+]-immobilized substrate of type I (8.0 × 10-8 M (8 ppb) for ClO4- and 4.2 × 10-5 M (4 ppm) for SO42-) are the lowest LOD values reported by in situ SERS detection of these anions in the literature so far. To delineate the superior performance of the Ag [+]-immobilized SERS substrates for in situ monitoring of anions, we have compared the SERS activity of several substrates with variable surface charge and chemical functionality. Similar trends were obtained with all the anions investigated. We present in the following section our results for one selected anion type: a representative chaotrope, the SCN- anion. 3.2. Comparison of Substrates of Types I-III for in Situ SERS Analysis of SCN-. Figure 7 illustrates adsorption of SCNon substrates of types I-III. Figure 7a shows raw SERS data for the three substrates collected with 8.6 × 10-8 M (5 ppb) SCNsolutions. Figure 7b depicts SCN- adsorption isotherms at SCN(50) Zhang, X.; Shah, N. C.; Van Duyne, R. P. Vib. Spectrosc. 2006, 42, 2-8.

Langmuir constant, KL (M-1) 107

5 × 105 1.7 × 106

LOD 1.7 × 10 M (1 ppb) 8.8 × 10-6 M (51 ppb) 6.9 × 10-8 M (4 ppb) -8

concentrations lower than 2 × 10-7 M, where adsorption was measured as the peak area of the SCN- vibrational band at 2110 cm-1. Table 2 summarizes the values of the Langmuir constants, KL, and LOD values for substrates of types I-III toward SCN- anions. A striking observation from Figure 7 and Table 2 is that adsorption isotherms were drastically dependent on the type of Ag nanoparticles attached to the substrate. Evident is a unique nature of the type I substrate (with Ag [+]), which showed the strongest anion binding with a high association constant, KL, of 107 M-1, and a reliable quantitative detection of SCN- with an LOD value of 1.7 × 10-8 M (1 ppb). In contrast, the type II substrate (with Ag [-]) did not yield detectable vibrational features for SCNat the low parts per billion concentration level due to a 20-fold lower binding constant, KL, of 5 × 105 M-1. The low SERS sensitivity for the type II substrate is a result of the electrostatic repulsion between Ag [-] nanoparticles and anions, which inhibits anion adsorption on the nanoparticle surface. Indeed, bare silver or gold nanoparticles prepared via the standard salt reduction using sodium citrate38 or sodium borohydride51 are negatively charged at neutral pH values, and usually their modification with cationic surface functionalities is required for SERS detection of anions in aqueous solutions.17-19,33,34 In our case, a control substrate (type III) was produced by coating a BPEI monolayer on immobilized Ag [-] nanoparticles to check the efficiency of surface functionalization with only amine groups toward improved anion binding. The presence of an adsorbed layer of BPEI at the surface of Ag [-] nanoparticles resulted in an improved binding of SCN- ions with a KL of 1.7 × 106 M-1 due to electrostatic attraction between the anions and the amino groups of BPEI on the nanoparticle surface. However, the SCN- binding constant was 6-fold lower than with the Ag [+] particles immobilized on the type I substrate, and an LOD value of 6.9 × 10-8 M (4 ppb) could be achieved through such modification. The presence of UV-induced amide groups, in addition to amino groups, in type I substrates contributed to their higher SERS sensitivity compared to that of type III substrates. Amide groups have been used as the main element in the synthesis of anion-binding receptors (macrocyclic or acyclic ones), which capture anions through hydrogen bonding in a nonaqueous environment (organic solvents).52,53 A combination of hydrogen bonding and electrostatic attractions has also been used to increase synthetic receptors to increase anion capturing in organic solvents.53 Water, as a highly competitive environment for hydrogen bonding, is expected to decrease the overall strength of hydrogen bonding between anions and the amide groups, but hydrogen bonding of amides with anions has been reported in an aqueous medium.47,52,54 Binding of anions to amide groups of proteins is also well-known.47,55 In addition to hydrogen (51) Creighton, J. A.; Blatchford, C. G.; Albrecht, M. G. J. Chem. Soc., Faraday Trans. 2 1979, 75, 790-798. (52) Kubik, S.; Goddard, R. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 51275132. (53) Bondy, C. R.; Loeb, S. J. Coord. Chem. ReV. 2003, 240, 77-99. (54) Spencer, J. N.; Berger, S. K.; Powell, C. R.; Henning, B. D.; Furman, G. S.; Loffredo, W. M.; Rydberg, E. M.; Neubert, R. A.; Shoop, C. E.; Blauch, D. N. J. Phys. Chem. 1981, 85, 1236-1241. (55) Voet, D.; Voet, J. G.; Pratt, C. W. Fundamentals of Biochemistry; Wiley: New York, 1999.

UltrasensitiVe SERS Detection of Anions in Water

bonding, dispersive interactions between anions and amide functional groups in water have recently been found to play a significant role in anion binding with the strongest contribution for the less hydrated chaotropic anions, such as ClO4- or SCN-.48 Finally, we state that large anion binding constants observed for Ag [+] particles (type I substrate) are a result of cooperative binding to amino and amide groups, which is facilitated through molecular mobility of such groups included in the adsorbed BPEI fragments. A similar situation occurs in the case of nonaqueous synthetic anion-binding amide-based receptors, where the anionreceptor binding constants can be enhanced through the introduction of a flexible backbone in amide-based receptor molecules, affording rearrangement and optimization of receptoranion hydrogen bonding.53 We suggest that a similar enhancement of cooperativity of anion binding occurs through molecular mobility of amino and amide groups included in loops and tails of BPEI fragments adsorbed on Ag [+] nanoparticles.

4. Conclusions Positively charged silver nanoparticles synthesized by photoreduction of silver nitrate using a mixture of BPEI and HEPES solutions as reducing agents have been used to produce robust SERS platforms for ultrasensitive measurements of a variety of anions, including environmentally relevant ClO4- and CN- anions in water. Anion detection was achieved without further elaborate

Langmuir, Vol. 24, No. 9, 2008 4771

surface functionalization treatment of the Ag [+] nanoparticles with LOD values of 8.0 × 10-8 M (8 ppb) and 2.7 × 10-7 M (7 ppb) for ClO4- and CN-, respectively. In contrast, SERS substrates with immobilized Ag [-] nanoparticles without or with a BPEI overlayer are less sensitive for SERS-based anion measurements, indicating critical importance of the control of the surface charge state as well as the surface chemical functionality of the nanoparticles. The key to ultrahigh SERS sensitivity of the Ag [+]-immobilized substrate is the synergistic effect of primary amino groups, which facilitate electrostatic attraction of anions, and the amide groups, which promote anion attraction by hydrogen bonding and dispersion interactions, of BPEI fragments on the Ag [+] nanoparticle surface. The high density and likely conformational mobility of amino and amide groups included in the loops and tails of adsorbed BPEI fragments ensure a significant synergistic effect and provide superior anion binding constants and SERS performance of Ag [+]-immobilized substrates. Acknowledgment. We thank Mr. Maung Khaing Oo for his help with SERS measurements of Ag [+] nanoparticle stability toward cyanide anions. This work was supported by the NSF under Grant No. ECS-0404002. LA703831Q