Quantitative Surface Acoustic Wave Detection Based on Colloidal

Mar 26, 2008 - The immobilization scheme of monodispersed gold nanoparticles (10-nm diameter) on piezoelectric substrate surfaces using organosilane ...
0 downloads 0 Views 774KB Size
Anal. Chem. 2008, 80, 3318-3326

Quantitative Surface Acoustic Wave Detection Based on Colloidal Gold Nanoparticles and Their Bioconjugates Chi-Shun Chiu and Shangjr Gwo*

Institute of Nanoengineering and Microsystems and Department of Physics, National Tsing-Hua University, Hsinchu 30013, Taiwan, Republic of China

The immobilization scheme of monodispersed gold nanoparticles (10-nm diameter) on piezoelectric substrate surfaces using organosilane molecules as cross-linkers has been developed for lithium niobate (LiNbO3) and silicon oxide (SiO2)/gold-covered lithium tantalate (LiTaO3) of Rayleigh and guided shear horizontal- (guided SH) surface acoustic wave (SAW) sensors. In this study, comparative measurements of gold nanoparticle adsorption kinetics using high-resolution field-emission scanning electron microscopy and SAW sensors allow the frequency responses of SAW sensors to be quantitatively correlated with surface densities of adsorbed nanoparticles. Using this approach, gold nanoparticles are used as the “nanosized mass standards” to scale the mass loading in a wide dynamical range. Rayleigh-SAW and guided SH-SAW sensors are employed here to monitor the surface mass changes on the device surfaces in gas and liquid phases, respectively. The mass sensitivity (∼20 Hz‚cm2/ng) of Rayleigh-SAW device (fundamental oscillation frequency of 113.3 MHz in air) is more than 2 orders of magnitude higher than that of conventional 9-MHz quartz crystal microbalance sensors. Furthermore, in situ (aqueous solutions), real-time measurements of adsorption kinetics for both citrate-stabilized gold nanoparticles and DNAgold nanoparticle conjugates are also demonstrated by guided SH-SAW (fundamental oscillation frequency of 121.3 MHz). By comparing frequency shifts between the adsorption cases of gold nanoparticles and DNA-gold nanoparticle conjugates, the average number of bound oligonucleotides per gold nanoparticle can also be determined. The high mass sensitivity (∼6 Hz‚cm2/ng) of guided SH-SAW sensors and successful detection of DNA-gold nanoparticle conjugates paves the way for realtime biosensing in liquids using nanoparticle-enhanced SAW devices. Gold nanoparticles are the most stable and developed metal nanoparticles. Several methods have been developed to synthesize monodispersed (size and shape) gold nanoparticles in water or organic solutions. At present, they have been successfully used as bottom-up building blocks in various fields of nanoscience and * To whom correspondence should be addressed. E-mail: [email protected]. edu.tw.

3318 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

nanoengineering due to their fascinating size-dependent optical, electronic, and catalytic properties.1,2 One of the important aspects about gold nanoparticles is the capability to functionalize biocompatible gold nanoparticles with a large number of surface ligands such as organic molecules and polymers. Especially, gold nanoparticle bioconjugation schemes with oligonucleotides, proteins, peptides, lipids, enzymes, drugs, and viruses have been developed, which greatly enhance their perspective for their biological sensing applications.1-4 Recently, various DNA detection methods based on functionalized gold nanoparticles have been proposed and demonstrated, which include light absorbance and scattering, localized surface plasmon resonance imaging, electrical detection, electrochemical detection, and gravimetric method using quartz crystal microbalances (QCMs).5,6 The principle of QCM detection is to measure the resonant frequency shift that can be correlated with very small mass changes at the oscillating quartz surface. Since the frequency readout from the quartz crystal can be acquired very rapidly, it is also possible to observe the real-time events of surface mass loading using QCMs. Very recently, it has been demonstrated that the use of gold nanoparticle mass tags can significantly enhance the sensitivity of QCM-based sensors due to the high specific mass of gold nanoparticles and the mass amplified effect.7-9 In comparison with QCM sensors, surface acoustic wave (SAW) sensors utilize a pair of interdigital transducers (IDTs) to generate and detect an acoustic wave propagating on the same surface of a piezoelectric substrate. Because of high operating frequencies of SAW sensors (typically >100 vs 5-10 MHz for (1) Shipway, A. N.; Katz, E.; Willner, I. ChemPhysChem. 2000, 1, 18-52. (2) Daniel, M. C.; Astruc, D. Chem. Rev. 2004, 104, 293-346. (3) Penn, S. G.; He, L.; Natan, M. J. Curr. Opin. Chem. Biol. 2003, 7, 609615. (4) Katz, E.; Willner, I. Angew. Chem., Int. Ed. 2004, 43, 6042-6108. (5) Fritzsche, W.; Taton, T. A. Nanotechnology 2003, 14, R63-R73. (6) Foultier, B.; Moreno-Hagelsieb, L.; Flandre, D.; Remacle, J. IEE Proc.Nanobiotechnol. 2005, 152, 3-12. (7) Zhou, X. C.; O’Shea, S. J.; Li, S. F. Y. Chem. Commun. 2000, 953-954. (8) (a) Patolsky, F.; Ranjit, K. T.; Lichtenstein, A.; Willner, I. Chem. Commun. 2000, 1025-1026. (b) Weizmann, Y.; Patolsky, F.; Willner, I. Analyst 2001, 126, 1502-1504. (c) Willner, I.; Patolsky, F.; Weizmann, Y.; Willner, B. Talanta 2002, 56, 847-856. (9) (a) Lin, L.; Zhao, H. Q.; Li, J. R.; Tang, J. A.; Duan, M. X.; Jiang, L. Biochem. Biophys. Res. Commun. 2000, 274, 817-820. (b) Zhao, H. Q.; Lin, L.; Li, J. R.; Tang, J. A.; Duan, M. X.; Jiang, L. J. Nanopart. Res. 2001, 3, 321-323. (c) Liu, T.; Tang, J.; Zhao, H.; Deng, Y.; Jiang, L. Langmuir 2002, 18, 56245626. (d) Liu, T.; Tang, J.; Jiang, L. Biochem. Biophys. Res. Commun. 2004, 313, 3-7. 10.1021/ac702495g CCC: $40.75

© 2008 American Chemical Society Published on Web 03/26/2008

QCM sensors), the sensitivities of SAW sensors are extremely high. In addition to their high sensitivity, SAW sensors also possess the advantages of small size, robustness, multiplexed detection, and low fabrication cost. Therefore, they are promising candidates to be used as electronic noses and tongues. Until now, most of the QCM and SAW sensitivities were theoretically estimated based on known material parameters (via the Sauerbrey equation for the QCM case) and indirect measurements of analyte coverages (e.g., assumption of monolayer coverage or layer thickness measurement by optical means10,11). In this work, we report on the quantitative SAW sensing of adsorption of citratestabilized and oligonucleotide-functionalized gold nanoparticles onto piezoelectric crystal surfaces such as lithium niobate (LiNbO3, LN) and lithium tantalate (LiTaO3, LT), which are widely used as the transducer materials for Rayleigh-SAW and shear horizontal (SH)-SAW devices. When the surface acoustic wave propagates on a piezoelectric crystal surface, its propagation characteristics (such as phase velocity and amplitude) vary with the changes of physical parameters at the surface or the adjacent medium. The first report of gas-phase chemical sensing with SAW devices was given by Wohltjen and Dessy in 1979.12 Rayleigh-SAW was used in their study. Detailed discussion about the theory and operation of SAW devices has since been published.13,14 For the gas-phase environment, Rayleigh-SAW is usually utilized due to the fact that the acoustic wave propagates with most of its energy on the surface confined to around one acoustic wavelength in depth. Therefore, it is sensitive to the surface perturbations. However, the Rayleigh wave involves both longitudinal component (out-of-plane) and transverse component (in-plane) motions, where a longitudinal motion has a radiation loss into liquid because the longitudinal wave velocity in liquid is less than the phase velocity of RayleighSAW. To avoid the radiation loss into the surrounding liquid, modes with pure or dominant shear polarization should be employed since liquids do not support shear waves. For SH-SAW devices, the elastic displacement is parallel to the surface and perpendicular to the wave propagation direction, so that the losses associated with this type of device in liquids are not significant. It has been previously demonstrated that the SH-SAW device can be used as a sensor platform for chemical detection in a liquidphase environment.15-17 However, SH-SAW propagates slightly deeper into the substrate; hence, the detection sensitivity to surface perturbations is reduced. Nevertheless, the mass sensitivity can be enhanced by using a guiding layer on the device surface. In this way, the acoustic energy is concentrated near the sensing surface of the guiding layer, thus increasing the sensitivity to surface perturbations. The guided SH-SAW devices typically (10) Ebara, Y.; Itakura, K.; Okahata, Y. Langmuir 1996, 12, 5165-5170. (11) Branch, D. W.; Brozik, S. M. Biosens. Bioelectron. 2004, 19, 849-859. (12) Wohltjen, H.; Dessy, R. Anal. Chem. 1979, 51, 1458-1464. (13) Wohltjen, H. Sens. Actuators 1984, 5, 307-325. (14) Ballantine, D. S.; White, R. M.; Martin, S. J.; Ricco, A. J.; Zellers, E. T.; Frye, G. C.; Wohltjen, H. Acoustic Wave Sensors: Theory, Design and PhysicoChemical Applications; Academic Press Inc.: San Diego, CA, 1997. (15) (a) Moriizumi, T.; Unno, Y.; Shiokawa, S. Proc. 1987 Ultrason. Symp. 1987, 579-582. (b) Campitelli, A. P.; Wlodarski, W.; Hoummady, M. Sens. Actuators, B 1998, 49, 195-201. (16) Yamazaki, T.; Kondoh, J.; Matsui, Y.; Shiokawa, S. Sens. Actuators, A 2000, 83, 34-39. (17) Jacesko, S.; Abraham, J. K.; Ji, T.; Varadan, V. K.; Cole, M.; Gardner, J. W. Smart Mater. Struct. 2005, 14, 1010-1016.

consist of an overlayer with a lower shear wave velocity. Both SiO218,19 and poly(methyl methacrylate)20,21 have been reported as the overlayer materials. SiO2 might be more advantageous because it features a low damping coefficient and a sufficiently low bulk shear velocity. Moreover, it is compatible with the existing integrated circuit fabrication process. Guided SH-SAW has already been applied for biosensing such as specific protein binding based on ligand-receptor pairing, which is very useful for immunosensors.22-24 However, nanoparticle-based SH-SAW sensing has not yet been demonstrated. In this paper, the following main results are presented: (1) The immobilization of recognition species on the sensing surfaces is the key step to construct a SAW sensor. Positively charged organosilane self-assembled monolayers (SAMs) have been demonstrated before as the immobilization films for selective adsorption of negatively charged gold nanoparticles onto substrates ranging from glass, quartz, alumina, metals, mica, and indium tin oxide to silicon with a native oxide layer.25-28 Here, we demonstrate the immobilization scheme of gold nanoparticles and their derivatives based on the formation of organosilane SAMs on the plasma-activated LN and SiO2/gold/LT substrates (Scheme 1). The specific type of SAMs used here is the amino-functionalized monolayer of 3-aminopropyltrimethoxysilane (APTMS) due to their stable charged property in water (amino terminal groups are protonated under suitable pH conditions). However, it is not the only possible choice, and other possible choices have been described in detail.26 (2) Because of the monodispersed nature of citrate-stabilized gold nanoparticles (10-nm diameter, ∼7% in size distribution width), quantitative analysis of adsorbed mass on SAW substrates is feasible here by correlating the surface densities of adsorbed gold nanoparticles (“nanosized mass standards”) observed by fieldemission scanning electron microscopy (FE-SEM) and the measured SAW frequency shifts. (3) Using APTMS-functionalized, SiO2-layer-guided SH-SAW devices, we demonstrate in situ (liquid phase), real-time detection of adsorption kinetics for both citrate-stabilized gold nanoparticles and DNA-gold nanoparticle conjugates. It is shown that the SiO2 guiding layer can greatly enhance the sensitivity of SH-SAW (a (18) Du, J.; Harding, G. L.; Collings, A. F.; Dencher, P. R. Sens. Actuators, A 1997, 60, 54-61. (19) Freudenberg, J.; von Schickfus, M.; Hunklinger, S. Sens. Actuators, B 2001, 76, 147-151. (20) Gizeli, E. Smart Mater. Struct. 1997, 6, 700-706. (21) Josse, F.; Bender, F.; Cernosek, R. W. Anal. Chem. 2001, 73, 5937-5944. (22) Harding, G. L.; Du, J.; Dencher, P. R.; Barnett, D.; Howe, E. Sens. Actuators, A 1997, 61, 279-286. (23) Gizeli, E.; Liley, M.; Lowe, C. R.; Vogel, H. Anal. Chem. 1997, 69, 48084813. (24) Bender, F.; Cernosek, R. W.; Josse, F. Electron. Lett. 2000, 36, 1672-1673. (25) (a) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J. Science 1995, 267, 1629-1632. (b) Grabar, K. C.; Freeman, R. G.; Hommer, M. B.; Natan, M. J. Anal. Chem. 1995, 67, 735-743. (c) Doron, A.; Katz, E.; Willner, I. Langmuir, 1995, 11, 1313-1317. (d) Lahav, M.; Gabai, R.; Shipway, A. N.; Willner, I. Chem. Commun. 1999, 19371938. (26) Grabar, K. C.; Allison, K. J.; Baker, B. E.; Bright, R. M.; Brown, K. R.; Freeman, R. G.; Fox, A. P.; Keating, C. D.; Musick, M. D.; Natan, M. J. Langmuir 1996, 12, 2353-2361. (27) Schmitt, J.; Ma¨chtle, P.; Eck, D.; Mo ¨hwald, H.; Helm, C. A. Langmuir 1999, 15, 3256-3266. (28) Chen, C. F.; Tzeng, S. D.; Lin, M. H.; Gwo, S. Langmuir 2006, 22, 78197824, and references therein.

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3319

Scheme 1. Schematic Representation of DNA-Gold Nanoparticle Conjugate Adsorption onto the APTMS-Coated Silicon Oxide Sensing Surface of a Guided SH-Mode Lithium Tantalate SAW Sensor Covered with a Gold Layer (for Prevention of Acoustoelectric Effect) and a SiO2 Wave Guiding Layer

factor of 10). Besides measuring the mass sensitivity, the average number of bound oligonucleotides per gold nanoparticle can also be determined by using this approach. EXPERIMENTAL SECTION Materials. Citrate-stabilized gold nanoparticles with a mean diameter of 10 nm and a particle concentration of ∼8 nM were purchased from Sigma (product no. G1527). APTMS was purchased from Aldrich (product no. 281778) and stored in desiccators. Sodium citrate was purchased from Sigma. Lithium niobate and lithium tantalite wafers were all SAW grade substrates, polished on single sides. Gold-coated substrates were composed of 150-nm Au and 30-nm Ti. Oigonucleotides were prepared by standard phosphoramidite chemistry. The thiolated probe DNA corresponding to the structurally important L2 zinc binding domain of the TP53 gene is a 28-base oligonucleotide with the following sequence: 5′-HS-(CH2)6-CATGGTGGGGGCAGTGCCTCACAACCTC-3′. The molecular weight of this 28-base oligonucleotide is 8786.79. All water used has a resistivity above 18 MΩ‚ cm, distilled through a Barnstead Nanopure water purification system. An Eppendorf 5415D centrifuge was used for centrifugation of DNA-gold nanoparticle conjugate solutions. Preparation of SAW Substrates. All SAW substrates were cut into 1 cm × 0.5 cm pieces for the fabrication of SAW devices, sequentially cleaned with acetone and methanol in an ultrasonic bath for 10 min, respectively. They were then rinsed with deionized water and blown dry with nitrogen. After wafer cleaning, the surface modification of sample was performed by air plasma. The samples treated in this way are rich in hydroxyls at the surface and suitable for the silanization process.28 Functionalization of the Sensing Surfaces. Cleaned samples were functionalized by an APTMS solution of original concentration as received. The silanization proceeded in APTMS solution for 12 h at room temperature. The samples were removed from the solution, rinsed rigorously with deionized water, sonicated for 20 min, and then rinsed again several times. Finally, the samples were blown dry with nitrogen. The samples were stored in desiccators if not used immediately. These APTMS monolayers on hydroxyl-containing substrates were used as immobilization layers for electrostatic adsorption of charged gold nanoparticles. 3320 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

Figure 1. Statistical analysis of gold nanoparticle size distribution as evaluated from FE-SEM imaging of 1252 gold nanoparticles. (a) Representative FE-SEM image of adsorbed gold nanoparticles on an APTMS/gold/Si substrate and (b) size histogram of adsorbed gold nanoparticles.

Adsorption of Gold Nanoparticles. The silanized samples were immersed in a solution of gold nanoparticles for different immersion times, then rinsed with water, and blown dry with nitrogen. Samples were coated with a ∼5-nm-thick gold thin layer before observation using a field emission scanning electron microscope (Zeiss Ultra 55 FE-SEM). The quantification of the nanoparticle density absorbed on the surface was performed by FE-SEM image analysis. In order to obtain reliable statistics of the nanoparticle surface density, at least five FE-SEM images were analyzed for each sample. For quantitative information, an image analysis software (Scion Image) was utilized to obtain the surface densities of adsorbed gold nanoparticles. Size Distribution and Average Mass of Gold Nanoparticles. In order to determine the average diameter of gold nanoparticles, it is necessary to analyze the distribution of gold nanoparticles. For this purpose, the substrates were gold-coated silicon wafers. Gold nanoparticles of ∼10-nm diameter were deposited on silanized samples by immersion into the as-received gold nanoparticle aqueous solution for 1 h at room temperature. Immersion time was adjusted to ensure a high density of homogeneously distributed gold nanoparticles over the substrate surfaces, and then the nanoparticle size distribution was obtained by analyzing high-resolution FE-SEM images. The resulting size distribution of immobilized nanoparticle is shown in Figure 1. The size statistics was obtained with more than 1000 nanoparticles (the size histogram is shown in Figure 1b). The result is indicative of a monodispersed distribution of nanoparticle diameter, with the

Figure 2. (a) Photograph showing the configuration of dual delay lines for the Rayleigh-SAW devices. (b) Photograph showing the configuration of dual delay lines for the guided SH-mode SAW devices.

average diameter of 10.0 nm and a standard deviation 0.7 nm, corresponding to a distribution width of ∼7%. Thus, it is reasonable to use 10 nm as the nanoparticle diameter. In this study, we assume that gold nanoparticles possess a spherical shape. Since the specific mass of gold is 19.30 g/cm3, the average mass (m0) for each gold nanoparticle can be estimated to be ∼1.0 × 10-17 g (∼10 ag). Moreover, the surface mass density of gold nanoparticles on the surface can be defined as n×m0/A, where n is the number of gold nanoparticles adsorbed on a measured area A. Rayleigh-SAW Device Configuration. The Rayleigh-SAW devices were fabricated on lithium niobate (128° rotated Y-cut, X-propagating LiNbO3, 128YX.LN) with 150-nm Au/30-nm Ti metallization using a standard photolithographic process. The structure of IDTs consisted of 50 finger pairs with an acoustic aperture w of 2.38 mm, a center to center separation L of 5.1 mm, and a periodicity p of 34 µm, which corresponds to a fundamental oscillation frequency of ∼113.3 MHz. A dual-delay-line configuration was used in these devices. Both delay-line paths were free surfaces, and a sensitive length l of 3.4 mm was constrained by silicone placed on gold markers (Figure 2a). Each Rayleigh-SAW device was mounted on a printed circuit board (PCB). The pads of the SAW sensor were welded to the PCB with aluminum wires by using a wire bonder, and then silicone was used around the sensing area to constrain the nanoparticle solution. These delay lines were connected in a feedback circuit of a high-frequency amplifier, resulting in SAW oscillators. In this configuration, one delay line was used as a reference channel with an APTMS-coated

surface; the other delay line was used as a sensing channel. For the sensing delay line, the whole propagation path was coated with the SAM formed from the APTMS solution, and then a gold nanoparticle solution was dropped on the delay-line path between input and output IDTs of the device surface for different immersion times, blown with nitrogen, and rinsed with water several times. Adsorption of gold nanoparticles causes an increase in the surface mass density of the SAM film, resulting in a decrease of the SAW resonant frequency. In the present sensor design, the dual-delayline configuration eliminates environment changes which are common to the two delay lines, such as the effects of the changes in temperature and pressure. However, the nonsystematic effects, such as the changes to the SAM film, cannot be cancelled by the present device configuration. Therefore, mainly the change in the surface mass loading due to the adsorption of gold nanoparticles is detected as a sensor response. The frequency output from the readout electronics package was measured with an Agilent 53132A frequency counter or an Advantest R3131 spectrum analyzer. The difference between the resonant frequencies of the reference and sensing channels was measured. Guided SH-SAW Device Configuration. The guided SHSAW devices were fabricated on lithium tantalate (36° rotated Y-cut, X propagating LiTaO3, 36YX.LT) piezoelectric substrates with 150 nm/30 nm thick Au/Ti metallization using standard photolithographic process. IDTs consisted of 50 finger pairs with an acoustic aperture W of 2.38 mm, a center to center separation L of 5.1 mm, and a periodicity p of 34 µm, which corresponds to Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3321

a SAW frequency of ∼121.3 MHz in aqueous solution. Dual-delay lines were employed with a gold metallized delay path between input and output IDTs to eliminate the acoustoelectric interaction with analytes in phosphate buffer solution (PBS) on the sensing surface (Figure 2b). A SiO2 layer, 3.2 µm thick, was then deposited by plasma-enhanced chemical vapor deposition (PECVD) on both the IDTs and the sensing area. The effect of this layer is to convert the SH-SAW into a guided SH-SAW, to increase the coupling coefficient, and to provide electrical passivation of IDTs. Moreover, the SiO2 waveguide layer also decreases the temperature dependence of resonant frequency. For the real-time measurement of nanoparticle adsorption based on mass loading, another dual-delayline configuration is employed (Figure 2b). Both delay-line paths are electrically screened by an Au/Ti film to eliminate the acoustoelectric effect. The SiO2 waveguide used in these devices can be easily functionalized with hydroxyl surface group for SAM coupling. A silicone liquid cell (volume ∼30 µL) was fabricated and placed on top of the propagating path, to localize sample liquid to the sensing area of the guided SH-SAW device. The guided SH-SAW sensing process used in this study is shown in Scheme 1. Preparation of Probe DNA-Modified Gold Nanoparticles. Gold nanoparticles were modified with thiolated oligonucleotides according to the salt-aging protocol described by Mirkin and coworkers.29,30 Briefly, the oligonucleotides were prepared by adding 100 µL of oligonucleotides solution (3.76 µM) to 300 µL of aqueous gold nanoparticle solution. After standing for 24 h, the solution was buffered at pH 7.5 with 0.1 M NaCl, 10 mM PBS for 48 h at room temperature. Excess reagents were then removed by centrifugation for 30 min at 13 200 rpm. The DNA-gold nanoparticle conjugates were repeatedly washed and then redispersed in 0.3 M NaCl, 10 mM PBS (pH 7.5). It was estimated that an amount of ∼10-20% nanoparticles was discarded along with the supernatant during the preparation process.30 In this work, we estimated 20% of gold nanoparticles were consumed during this preparation step. Estimation of Thiolated Oligonucleotides Loaded on Gold Nanoparticles. The adsorption measurements of gold nanoparticles and DNA-gold nanoparticle conjugates onto the sensing surfaces of guided SH-SAW sensors were employed to quantify the amount of oligonucleotides bound on individual gold nanoparticles. After the guided SH-SAW measurements, the final surface densities of gold nanoparticles were derived from FE-SEM micrographs, corresponding to the maximum SAW frequency shifts measured in situ using guided SH-SAW sensors. The frequency shifts per nanoparticle on unit sensing area were obtained by dividing the measured frequency shift by the surface densities of nanoparticles. The additional frequency shift of oligonucleotides loaded on gold nanoparticles was derived from the difference of frequency shifts per adsorbed particle between bare gold nanoparticle and DNA-gold nanoparticle conjugate. The average mass of oligonucleotides loaded on gold nanoparticles was then estimated using the known mass of 10-nm-diameter gold nanoparticle (∼1.0 × 10-17 g). Finally, the average number of bound oligonucleotides per particle was obtained by dividing the (29) Mirkin, C. A.; Letsinger, R. L.; Mucic, R. C.; Storhoff, J. J. Nature 1996, 382, 607-609. (30) Demer, L. M.; Mirkin, C. A.; Mucic, R. C.; Reynolds, R. A., III; Letsinger, R. L.; Elghanian, R.; Viswanadham, G. Anal. Chem. 2000, 72, 5535-5541.

3322

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

average mass of oligonucleotides bound on a single gold nanoparticle by the mass of single oligonucleotide. RESULTS AND DISCUSSION Gold Nanoparticle Immobilization. FE-SEM images are employed to visualize surface densities of gold nanoparticles adsorbed on the APTMS-coated piezoelectric substrates. For the cases of insulator substrates, surface charging is likely to occur during high-resolution FE-SEM imaging. Hence, a ∼5-nm-thick gold thin film was sputtered on the insulating substrate surfaces before FE-SEM observation. FE-SEM imaging for APTMS-coated lithium niobate samples shows that gold nanoparticles sparsely adsorbed (without aggregation) on the APTMS-covered surfaces under different immersion times (Figure 3). Initially, the surface density of gold nanoparticles increases sharply with increasing adsorption time. Gold nanoparticles are mostly well separated from each other. This can be explained by electrostatic repulsion between adsorbed nanoparticles due to the charged nature of citrate-stabilized gold nanoparticles. Eventually, after a long adsorption time, a saturation surface coverage of gold nanoparticle monolayer is reached. Various factors are responsible for the saturation surface density of adsorbed gold nanoparticles during the immobilization process. For instance, the choice of the organosilane monolayer, nanoparticle concentration, nanoparticle size, solution temperature, and pH value and salt concentration (ionic strength) in nanoparticle solution can all play an important role.28 We examine here the effects of nanoparticle concentration on adsorption kinetics and saturation nanoparticle surface coverage with all other parameters fixed. As shown in Figure 3e, with increasing adsorption time, surface density of gold nanoparticles increases steeply within the first 15 min and then approaches a saturated coverage. In contrast with the adsorption case of nondiluted (as-received) nanoparticle solution, the case of diluted (10% of the original concentration) solution of gold nanoparticles exhibits much slower adsorption kinetics. Beside the difference in adsorption kinetics, there is a big difference between the saturated adsorption coverage and the nanoparticle concentration. After nanoparticle adsorption in aqueous solution, Rayleigh-SAW measurements were employed by a dip and dry method to detect surface densities of gold nanoparticles in gas phase. The adsorption kinetics of gold nanoparticles is dominated by electrostatic attraction between the nanoparticles in solution and the oppositely charged surface. In order to describe the electrostatic adsorption kinetic processes of nanoparticles immobilized onto solid surfaces, diffusion-limited mechanisms have been introduced for the adsorption behavior of citrate-stabilized gold nanoparticles at the initial adsorption stage.31 However, this model could not adequately describe the behavior of submonolayer-type surface coverages, which is very different from the initial adsorption stage. Instead, a site-filling process seems to be more appropriate for describing the immobilization process of charged nanoparticles onto oppositely charged solid surfaces. The basic idea behind the site-filling process is that the surface coverage of nanoparticles reaches a maximum value (ns, saturated number density of adsorbed nanoparticles, which depends on the detailed (31) Grabar, K. C.; Smith, P. C.; Musick, M. D.; Davis, J. A.; Walter, D. G.; Jackson, M. A.; Guthrie, A. P.; Natan, M. J. J. Am. Chem. Soc. 1996, 118, 1148-1153.

Figure 3. Surface micrographs obtained by FE-SEM for the cases of citrate-stabilized gold nanoparticle adsorption onto the APTMS-coated lithium niobate substrates under different immersion times: (a) 20, (b) 100, (c) 180, and (d) 3600 s. (e) Kinetic curves of gold nanoparticle adsorption (number of nanoparticles vs time) were obtained by analyzing FE-SEM images for different nanoparticle concentrations, including as-received and diluted (10% of the original concentration) solutions. The sample surfaces were coated with a thin gold layer prior to FE-SEM imaging to prevent surface charging.

conditions of nanoparticle solutions) after sufficiently long adsorption time. This saturation behavior corresponds to the formation of a self-assembled nanoparticle monolayer. Therefore, the adsorption rate at a given time linearly depends on the remaining adsorption sites ns-n(t), where n(t) is the surface density of adsorbed nanoparticles at time t. Therefore, the adsorption rate can be described as

dn(t) ) kads[ns - n(t)] dt

(1)

where kads represents the adsorption rate constant. Under the present circumstance, kads proportional to c (the solution concentration of nanoparticles), the average impinging velocity of nanoparticles onto the surface, and their sticking coefficient. Under the present circumstance, kads is close to a linear function of c. The solution of eq 1 has the following form:

n(t) ) ns[1 - exp(-t/τ0)],

where

τ0 )

1 kads

(2)

The experimental data of gold nanoparticle adsorption onto the APTMS-coated lithium niobate substrates under different immersion times are plotted in Figure 3e. The dashed and solid curves are least-squares fits of eq 2 to the experimental data shown in Figure 2e. The fitted value of τ0 is 156 s for the nondiluted gold nanoparticle solution, and it is 1842 s for the diluted (10%) gold nanoparticle solution. The value of τ0 of the original nanoparticle concentration is roughly 11.8 times smaller than that of diluted (10%) solution, in good agreement with the assumption of sitefilling model. Rayleigh-SAW Gas-Phase Measurement. Surface acoustic waves are very sensitive to the surface changes, which can lead

Figure 4. Adsorption kinetics of gold nanoparticles measured by Rayleigh-SAW sensor for different nanoparticle concentrations, including as-received and diluted (10% of the original concentration). The SAW measurements were performed by a dip and dry method.

to the shift of resonant frequency in SAW sensors. The response of 113.3-MHz dual-delay-line Rayleigh-SAW sensor to gold nanoparticles adsorbed on the propagation path is relatively rapid within the first 15 min (Figure 4). At a higher nanoparticle concentration in solution, the change of SAW resonant frequency is very rapid and the frequency shift reaches a steady value after 20 min due to the saturated nanoparticle coverage. And, the SAW frequency can maintain at a constant value even after many hours of measurement. At a lower nanoparticle concentration in solution, the resonant frequency shifts much slower than that for the case of original concentration and the saturated state can only be reached after several hours of nanoparticle adsorption. Quantitative comparison with the FE-SEM micrographs (Figure 3e and Figure 4) indeed shows that the total frequency shift is dependent on the surface density of adsorbed nanoparticles. It Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3323

Figure 5. Linear response curves between the measured RayleighSAW frequency shift and the adsorption mass (estimated from the FE-SEM images) for (a) as-received gold nanoparticle aqueous solution and (b) diluted (10% of the original concentration) gold nanoparticle aqueous solution. The inset in (b) is the FE-SEM micrograph of the measured surface with the smallest frequency shift. It shows that a frequency shift of 203 ( 10 Hz corresponds to an adsorbed nanoparticle surface density of 10 ( 4 nanoparticles per µm2.

can be concluded that the 113.3-MHz dual-delay-line RayleighSAW sensor fabricated by us has a wide, linear dynamic range for detecting adsorbed nanoparticle surface density (results for both cases of nanoparticle concentration are shown in Figure 5). The mass sensitivity of our Rayleigh-SAW device can be determined to be ∼20 Hz‚cm2/ng. These results demonstrate that selfassembled nanoparticle monolayers not only can be used as sensing films (e.g., for detection of functionalized nanoparticles) but also can be applied to calibrate and quantify the performance of SAW sensors. Gas-phase measurement of citrate adsorption on a gold film was also performed in this study to estimate the amount of adsorbed citrate ions. For this purpose, both delay-line paths were coated with 50-nm Au/10-nm Ti for the measurement of citrate adsorption on the gold surface. A 0.04% sodium citrate solution was contacted with the sensing surface for 30 min. After dip and dry, the frequency shift between before and after citrate adsorption is ∼1.21 kHz (∼60 ng/cm2 surface mass loading density according 3324

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

to our Rayleigh-SAW sensitivity). Therefore, the surface citrate mass bound to 10-nm gold nanoparticle can be derived to be ∼1.90 × 10-19 g (only ∼2% of the mass of 10-nm gold nanoparticle) under the assumption of the same surface coverage between the gold film and the gold nanoparticle surface. Hence, the estimated amount of citrate capped on the gold nanoparticle surface is ∼600 ions/particle (equivalent to 1.9 × 1014 ions/ cm2). Including the mass of citrate, the mass of single citrate-stabilized gold nanoparticle is ∼1.02 × 10-17 g. SH-SAW Liquid-Phase Measurement and Acoustoelectric Interaction. Liquid-phase measurement is often necessary for real-time monitoring of the interaction between the functionalized nanoparticles and the piezoelectric sensing surface. As a test, the dual-delay-line configuration with a bare sensing area (without gold coating) is employed. However, in situ measurement shows that no detectable frequency shift is observed during the contact time of nanoparticles and the sensing surface. After the SH-SAW measurement, FE-SEM imaging was performed to confirm the surface state of the sensing area. The resulting image showed that no gold nanoparticles were adsorbed on the lithium tantalate surface. This can be explained by the acoustoelectric effect generated from the piezoelectric material. In other words, in this testing case, the moving evanescent electric field prevents the adsorption of nanoparticles at the surface. Thus, a metallized delay path between the input and output IDTs was used in this work to eliminate the acoustoelectric interaction (as shown in Figure 2b). In Situ Measurements Based on Guided SH-SAW Device. For these measurements, guided SH-SAW devices were employed to monitor the adsorption kinetics of gold nanoparticles and DNA-gold nanoparticle conjugates. The reference channel and the sensing channel were filled with DI water or PBS, and the liquid in the sensing channel contained gold nanoparticles (in DI water) or DNA-gold nanoparticle conjugates (in PBS), respectively. A PECVD-deposited 3.2-µm-thick SiO2 layer was used as the waveguide layer to trap acoustic wave energy near the surface. For comparison, guided and unguided SH-SAW sensors were employed first for in situ, real-time monitoring of gold nanoparticle adsorption (Figure 6). At the nanoparticle concentration of ∼4 nM, the resonant frequency of guided SH-SAW device decreased very quickly within the first 10 min and then changed very slowly. The saturated frequency shift of gold nanoparticle solution on APTMS-coated, guided SH-SAW surface was ∼4.12 kHz (Figure 6a). FE-SEM image of the resulting surface after guided SH-SAW measurement was used to estimate the surface density of adsorbed nanoparticles after 1-h exposure, which is 683 ( 67 particles/ µm2 (Figure 6a). By comparing results of guided SH-SAW and FE-SEM, the average frequency shift per gold nanoparticle adsorbed on unit surface area (µm2) can be obtained to be ∼6 Hz (∼5.81 Hz after liquid viscosity calibration; see detailed description below). The enhancement of sensitivity by using a SiO2 guiding layer can also be confirmed in this figure. Using SH-SAW sensor without the SiO2 waveguide to measure the response of gold nanoparticle adsorption onto the APTMS-coated gold surface, at the gold nanoparticle concentration of ∼5.6 nM, a saturated frequency shift of ∼0.57 kHz (∼0.40 kHz after liquid viscosity calibration) was observed after 1-h exposure (Figure 6b), and the FE-SEM image of gold nanoparticles adsorbed on the SH-SAW surface after 1-h exposure was used to estimate the nanoparticle

Figure 6. Real-time, continuous monitoring of SH-SAW frequency shift during gold nanoparticle adsorption onto the sensing surfaces in liquid at room temperature for two different types of SH-SAW devices. (a) Adsorption of citrate-stabilized gold nanoparticles onto the APTMS-coated SiO2 surface of guided SH-SAW device using a ∼4.0 nM gold nanoparticle aqueous solution. The corresponding SEM image was taken after 1-h exposure time. (b) Adsorption of citratestabilized gold nanoparticles onto the APTMS-coated gold surface of unguided SH-SAW device using a ∼5.6 nM gold nanoparticle aqueous solution. The corresponding SEM image was also taken after 1-h exposure time. The SAW response is greatly enhanced (∼10 times larger) by using a SiO2 guiding layer.

surface density, which is 768 ( 20 particles/µm2 (Figure 6b). Comparing these two cases, the factor of sensitivity enhancement using a SiO2 guiding layer is ∼10. Next, DNA-gold nanoparticle conjugates of different concentrations were immobilized onto the surfaces of APTMS-coated guided SH-SAW sensors to form a self-assembled nanoparticle monolayer. The measured SAW frequency decreased with time, resulting from the binding of DNA-gold conjugates. Meanwhile, the surface coverage of DNA-gold conjugates also increased with time. At the concentration of DNA-gold nanoparticles of ∼3.2 nM, a saturated frequency shift of ∼3.96 kHz was observed after ∼40 min exposure (Figure 7a). There is a major difference in solvents for gold nanoparticles and for DNA-gold nanoparticle conjugates. While low-conductivity water was used for gold nanoparticle solution, high-conductivity phosphate-buffered solution was adopted for DNA-gold nanoparticle conjugate solution. The screening effect of PBS remarkably decreases the electrostatic interaction between negative charged DNA-gold nanoparticle conjugates and the positive charged APTMS layer, resulting in slower electrostatic adsorption kinetics. At the higher concentration of DNA-gold nanoparticle conjugates of ∼6.4 nM, a saturated

Figure 7. Real-time, continuous monitoring of DNA-gold nanoparticle conjugate adsorption onto the SiO2 surfaces of guided SHSAW devices in liquid at room temperature. Two kinds of DNA-gold conjugate solutions were used, corresponding to nanoparticle concentrations of (a) ∼3.2 and (b) ∼6.4 nM. Below the SH-SAW responses curves are surface micrographs of DNA-gold nanoparticle conjugates adsorbed onto the APTMS-coated SiO2/gold/LiTaO3 surfaces imaged by FE-SEM after 1 h exposure time.

frequency shift of ∼7.13 kHz was observed after ∼30-min exposure (Figure 7b). These results are consistent with the surface densities of DNA-gold nanoparticle conjugates obtained from SEM images. The final surface densities of adsorbed nanoparticles under different nanoparticle concentrations were determined to be 565 ( 43 and 1009 ( 75 particles/µm2 (Figure 7a and b), respectively. In order to measure the viscosity effect, we used the following configuration for testing: The reference channel had no liquid loading, and the sensing channel was filled with different solutions. Moreover, the surface of the sensing channel was not functionalized in order to prevent the complication of mass loading by nanoparticle adsorption. The measurements of different viscous solutions (DI water, 4 nM gold nanoparticle solution, 5.6 nM gold nanoparticle solution, PBS, 3.2 nM DNA-gold nanoparticle conjugate solution, and 6.4 nM DNA-gold nanoparticle conjugate solution) were performed on the same guided SH-SAW device to compare the viscosity effect with different liquid loading. The measured frequency shifts due to the viscosity effect for different solutions are summarized in Table 1. According to this table, the differences of frequency shift due to the viscosity effect for the cases of DI water/4 nM gold nanoparticle solution and PBS/6.4 nM DNA-gold nanoparticle conjugate solution are 0.15 and 0.74 kHz, respectively. After deduction of the frequency shifts due to the viscosity effect, the average frequency shift per adsorbed Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3325

Table 1. Frequency Shifts Due to Viscosity Effect for Different Solutions

Table 2. Mass Sensitivities for Different Acoustic Devices

DNA-Au nanoparticle conjugates

frequency shift (kHz) b

DI watera

4.0 nM Au NPsb

5.6 nM Au NPs

0

0.15

0.17

PBS

3.2 nM

6.4 nM

0.43

0.83

1.17

a The frequency shift of DI water is used as the reference base value. Au NPs is abbreviation of gold nanoparticles.

DNA-gold nanoparticle conjugate on unit area (µm2) can be estimated, which is ∼6.33 Hz. The guided SH-SAW measurement results of adsorption of DNA-gold nanoparticle conjugates can then be used to estimate the oligonucleotides loaded on gold nanoparticles. As a result, the frequency shift due to the mass of oligonucleotides bound on gold nanoparticle can be determined to be ∼0.52 Hz per DNA-gold nanoparticle conjugate on unit area (µm2). Because the mass of 10-nm gold nanoparticle is known to be ∼1.0 × 10-17 g (we can neglect the effect of surface citrates because of their small molecular weight), the total mass of oligonucleotides conjugated on nanoparticle is ∼9.0 × 10-19 g/gold nanoparticle. This value is equivalent to ∼60 oligonucleotides/ nanoparticle (∼0.20 oligonucleotides/nm2) based on the molecular weight of DNA used in this study. The estimated surface coverage of oligonucleotides bound on the nanoparticle is in good agreement with the result obtained by the fluorescence-based method (∼0.21 oligonucleotides/nm2, 159 oligonucleotides bound on 15.7nm-diameter gold nanoparticle).30 Comparison of Mass Sensitivities for Various Devices. Mass loading effect is the main sensing mechanism of surface acoustic wave devices employed in this study. The sensitivity of mass loading effect is often defined as the frequency change per unit loaded mass density. For comparison, the mass sensitivities of QCM,32,33 Rayleigh-SAW, and guided SH-SAW used in this study are listed in Table 2. Sensitivities of mass loading effect for Rayleigh-SAW and guided SH-SAW are adopted from our measurements, and it should be noted that the design of the present guided SH-SAW has not been fully optimized. The most different part in our sensitivity measurements is that we obtained the mass sensitivities by direct comparison between adsorbed surface mass (32) Caruso, F.; Furlong, D. N.; Niikura, K.; Okahata, Y. Colloids Surf., B : Biointerface 1998, 10, 199-204. (33) Bizet, K.; Gabrielli, C.; Perrot, H. Analusis 1999, 27, 609-616.

3326 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

device

center frequency, MHz

sensitivity, Hz·cm2/ng

QCM32,33 Rayleigh-SAWa guided SH-SAWb

9 113.3 121.3

0.183 20 6

a Rayleigh-SAW device cannot be operated in the liquid environment. The design of guided SH-SAW device used in this study has not been fully optimized.

b

density of nanoparticles and the resulting frequency shift, and not through a theoretical equation. The results reported here indicate that our approach can provide a direct method to precisely calibrate the performance of SAW devices. Comparing with the mass sensitivity of a 9-MHz QCM device, the mass sensitivity of our Rayleigh-SAW device is ∼100-fold higher. Another important consideration for sensing applications is the limit of detection (LOD), which depends on the background noise level. For our Rayleigh-SAW sensors, the fluctuation (standard deviation) of frequency readout is ∼10 Hz (Figure 5b), and it is ∼15 Hz for our guided SH-SAW sensors in liquid phase. Both numbers are comparable with that reported for a 9-MHz QCM sensor (∼10 Hz).32 Therefore, the LOD is ∼2 orders of magnitude smaller than that of a 9-MHz QCM sensor (ref 32, ∼100 ng/cm2). From the result shown in Figure 5b, the LOD for Rayleigh-SAW sensor is ∼1 × 108 nanoparticles/cm2 (∼1 ng/cm2). CONCLUSIONS In this study, we show that monodispersed gold nanoparticles can be used as the “atto gram mass standards” to quantitatively determine the mass sensitivities of SAW sensors. Both RayleighSAW and guided SH-SAW sensors are employed here to monitor the surface mass changes on the device surfaces in gas and liquid phases, respectively. The achieved mass sensitivity of 113.3-MHz Rayleigh-SAW is ∼2 orders of magnitude larger than the conventional 9-MHz QCM sensors. Because of the excellent properties of gold nanoparticles in terms of monodispersity of size and shape, high specific mass, and availability of a wide range of surface functionalization schemes, gold nanoparticle-based quantitative SAW gravimetric detection methods can be very useful for future chemical and biological sensing applications. Received for review December 8, 2007. Accepted February 21, 2008. AC702495G