Anal. Chem. 2007, 79, 5078-5081
Technical Notes
Vapor Deposition Method for Sensitivity Studies on Engineered Surface-Enhanced Raman Scattering-Active Substrates Thomas H. Reilly, III,† Jordan D. Corbman,† and Kathy L. Rowlen*,‡
Department of Chemistry and Biochemistry, University of Colorado, Boulder, Colorado 80309, and InDevR, LLC, 2100 Central Avenue, Suite 106, Boulder, Colorado 80301
The design and optimization of a vapor-phase analyte deposition method for limit of detection (LOD) studies on engineered surface-enhanced Raman scattering (SERS)active substrates is presented. The vapor deposition method was designed to overcome current challenges in quantitative analysis of lithographically produced SERS substrates that are relatively small (hundreds of square micrometers). A custom-built flow cell was used to deposit benzenethiol from the vapor phase onto SERS-active Ag thin films, as the control substrates, and nanoaperture arrays that were generated by electron-beam lithography. The surface coverage of benzenethiol as a function of time was monitored using the ring stretching mode 1070-cm-1 band and the trend was fit to Langmuir adsorption kinetics. The method was deemed reliable based on agreement between the LOD determined on the control substrates and previously reported values for those substrates. Application of the new method to a 20 × 20 µm2 nanoaperture array yielded a LOD of 4.2 ( 0.3 amol. The recent emergence of rationally designed micrometer-sized surface-enhanced Raman scattering (SERS) substrates necessitates the development of new analytical methods to quantitatively evaluate substrate performance. While substrates such as nanoaperture and nanoparticle arrays are highly interesting to the SERS community, and the broader field of plasmonics, they pose a unique challenge when the active area is only a few hundred square micrometers in area (see Figure 1).1,2 Such a small active area requires homogeneous deposition of the SERS analytes over the same length scale. In order to determine the limit of detection (LOD) in a SERS experiment, signal must be monitored as a function of known surface coverage.3 Surface coverage is generally varied by drop * To whom correspondence should be addressed. E-mail:
[email protected]. † University of Colorado. ‡ InDevR, LLC. (1) Brolo, A. G.; Arctander, E.; Gordon, R.; Leathem, B.; Kavanagh, K. L. Nano Lett. 2004, 4, 2015-2018. (2) Reilly, T. H., III; Chang, S.-H.; Corbman, J. D.; Schatz, G. C.; Rowlen, K. L. J. Phys. Chem. C. In press. (3) McCreery, R. L. Raman Spectroscopy for Chemical Analysis; John Wiley & Sons: New York, 2000.
5078 Analytical Chemistry, Vol. 79, No. 13, July 1, 2007
coating analyte molecules across the surface.4 For large (mm2) substrates, this technique works well since the excitation spot size is typically much smaller than the area covered, which allows for multiple measurements on the substrate to account for any inhomogeneities in surface coverage. When the substrate is smaller, on the order of the excitation beam diameter, as is the case for nanoaperture arrays, drop coating is a poor choice. Drop coating deposits the analyte in an ill-defined gradient that is dependent on environmental (e.g., humidity) conditions and the nature of the substrate. Alternate methods exist for analyte deposition. For example, coverage across the surface of a SERS substrate can be controlled using adsorption/desorption equilibria in solution.5 While the solution equilibrium method avoids the analyte gradient that occurs from the drop coating technique, it exposes the SERS substrate to solvent. Solvent effects have been shown to alter the optical response and morphology of SERS substrates.6 The changes that occur when the SERS substrate is immersed in the solvent are not readily predictable. Furthermore, quantitative deposition requires direct measurement either as a function of time or, as is typically done, assumptions regarding adsorption/ desorption kinetics. To evaluate small SERS sensors, a promising means to overcome the drawbacks of drop coating and solution equilibriumbased analyte deposition techniques is vapor-phase analyte deposition. SERS has been demonstrated to be sensitive to vapordeposited organic species.7-13 In particular, Hill et al. demonstrated (4) Norrod, K. L.; Sudnik, L. M.; Rousell, D.; Rowlen, K. L. Appl. Spectrosc. 1997, 51, 994-1001. (5) Zhang, X. Y.; Young, M. A.; Lyandres, O.; Van Duyne, R. P. J. Am. Chem. Soc. 2005, 127, 4484-4489. (6) Roark, S. E.; Semin, D. J.; Lo, A.; Skodje, R. T.; Rowlen, K. L. Anal. Chim. Acta 1995, 307, 341-353. (7) Vo-Dinh, T.; Stokes, D. L. Field Anal. Chem. Technol. 1999, 3, 346-356. (8) Stokes, D. L.; Pal, A.; Narayanan, V. A.; Vo-Dinh, T. Anal. Chim. Acta 1999, 399, 265-274. (9) Carron, K. T.; Kennedy, B. J. Anal. Chem. 1995, 67, 3353-3356. (10) Wehling, B.; Hill, W.; Klockow, D. Int. J. Environ. Anal. Chem. 1999, 73, 223-236. (11) Hill, W.; Fallourd, V.; Klockow, D. J. Phys. Chem. B 1999, 103, 47074713. (12) Hill, W.; Wehling, B.; Klockow, D. Appl. Spectrosc. 1999, 53, 547-550. (13) Taranenko, N.; Alarie, J. P.; Stokes, D. L.; VoDinh, T. J. Raman Spectrosc. 1996, 27, 379-384. 10.1021/ac070121i CCC: $37.00
© 2007 American Chemical Society Published on Web 06/06/2007
Figure 2. Schematic of the vapor-phase thiol deposition apparatus. Ultrahigh-purity N2 was passed over a 10-4 M BT solution and into the sample cell. The quartz SERS substrate was used as a window on the bottom of the sample cell. A 40× (0.75 NA) objective was focused through the quartz to the SERS-active film on the inside of the sample cell. An argon ion laser (514.5-nm plasma line, 0.5 mW) was used as the excitation source with 1-s collection times on the CCD.
Ag thin metal films and then extended to a 20 × 20 µm2 nanoaperture array.
Figure 1. Representative images of nanoaperture array SERS substrates. Image a is a white light transmission image of a nanoaperture array with 800-nm lattice spacing. The image contrast has been enhanced for clarity. Image b is an AFM micrograph of a nanoaperture array with 600-nm lattice spacing. The dimensions of such small SERS substrates make them exciting for their potential use in miniaturized sensors but challenging to quantitatively evaluate.
that vapor-phase adsorption of methyl mercaptan on Ag films is consistent with a Langmuir mechanism of adsorption. The observation of a Langmuir mechanism of adsorption indicates that it is possible to quantify the surface coverage of analyte molecules as a function of time. Under homogeneous mixing conditions of vapor-phase analyte and carrier gas, it is expected that the surface coverage across a substrate would be uniform to a very small length scale. Uniform analyte deposition and a surface coverage that can be varied with exposure time would enable a SERS LOD study to be performed on SERS substrates of any size with minimal perturbation of the substrate. In this work, a vapor-phase deposition method for LOD determination was evaluated. The method was first validated using
EXPERIMENTAL SECTION The flow cell was constructed to deposit volatile organics onto a SERS substrate. A schematic of the vapor deposition procedure is featured in Figure 2. The details of the flow cell have been previously described.2 Benzenethiol (BT) was used as the model nonresonant Raman analyte in this study. Exposure to BT can damage the eyes, liver, and nerves. Extreme caution should be used when handling BT. Details of the home-built Raman spectroscopy and imaging experiment have been described in detail elsewhere.14 SERS detection of BT on thin metal films (TMFs) and on the nanopatterns used an excitation wavelength of 514.5 nm from an Ar+ laser with an excitation power of 0.5 mW and 1-s integration on the CCD. The 40×, 0.75 numerical aperture (NA) objective was used as both the excitation and collection objective. All Raman spectra were collected using WinSpec 32 software. Data sets were exported to ascii format from WinSpec. MatLab was used to fit a linear baseline to the Raman spectra and extract temporal changes in the intensity of the Raman bands as well as calculate signal-to-noise factors. Regression to data was performed in SigmaPlot 8.0. Quartz (Technical Glass Products, Inc., 1-in. rounds, 0.5-mm thickness, optically polished on both sides) was used as the substrate for spin coating the electron beam resist. The quartz was first triple washed using a soap solution (Alconox) and plastic sponge. The sample was thoroughly rinsed under flowing water (Elga, 18-MΩ resistance). Using plastic hemostat tweezers, the sample was immersed in a 75 °C solution of 4:1:1 H2O/HCl/H2O2 (30%), commonly known as SC-2 solution, for 15 min. The sample was removed from this solution, rinsed thoroughly under flowing water, immersed in a 4:1:1 H2O/NH4OH/H2O2 solution, commonly known as SC-1, at a temperature of 75 °C for 15 min, and subsequently rinsed under flowing water. Prior to deposition, the quartz rounds were heated on a hot plate at ∼150 °C for ∼5 min to drive off residual water from the cleaning process. Five nanometers of Ag (99.9%, Aldrich) was evaporated out of a tantalum boat at 0.2 Å/s at a pressure of 10-7 Torr. After deposition, the samples were left under high vacuum for 15 min. (14) Jacobson, M. L.; Rowlen, K. L. Chem. Phys. Lett. 2005, 401, 52-57.
Analytical Chemistry, Vol. 79, No. 13, July 1, 2007
5079
Figure 3. Adsorption isotherms for BT on 5-nm Ag thin films. The change in surface coverage through time was monitored by following the change in the 1070-cm-1 peak of BT. The data were normalized using the maximum intensity value at long times according to the Langmuir fit of the data.
The nanoaperture array was fabricated according to a previously published procedure.2 All samples were used for thiol deposition within 2 h of Ag evaporation. RESULTS AND DISCUSSION Characterization of the Vapor-Phase Thiol Deposition Apparatus. Thin metal films were chosen to evaluate the performance of the vapor-phase deposition flow cell and method because of the abundance of quantitative information that exists in the literature.8,15-20 With careful control of the flow rate, the increase in benzenethiol SERS intensity was monitored as a function of time during vapor deposition. The 1070-cm-1 band [ν(C-C), δip(C-H), ν(C-S)]21 was employed as a quantitative measure of adsorption. During vapor deposition, the signal reached a constant value within a few minutes. The adsorption data as a function of time were interpreted using simple Langmuir adsorption kinetics. Langmuir adsorption kinetics predict that the rate of adsorption is given by22
dΘ ) ka(1 - Θ)C - kdΘ dt
(1)
where Θ is the fraction of monolayer coverage, C is the molecular flux on the surface, ka is the adsorption rate constant, and kd is the desorption rate constant. Jakubowicz et al. simplified eq 1 and integrated to yield the coverage as a function of an observed rate constant: (15) (16) (17) (18) (19) (20) (21)
Semin, D. J.; Rowlen, K. L. Anal. Chem. 1994, 66, 4324-4331. Roark, S. E.; Rowlen, K. L. Anal. Chem. 1994, 66, 261-270. Roark, S. E.; Semin, D. J.; Rowlen, K. L. Anal. Chem. 1996, 68, 473-480. Mosier-Boss, P. A.; Lieberman, S. H. Appl. Spectrosc. 1999, 53, 862-873. Aroca, R.; Corio, P.; Rubim, J. C. Ann. Chim. (Rome) 1997, 87, 1-7. Vo-Dinh, T. TrAC, Trends Anal. Chem. 1998, 17, 557-582. Taylor, C. E.; Pemberton, J. E.; Goodman, G. G.; Schoenfisch, M. H. Appl. Spectrosc. 1999, 53, 1212-1221. (22) Jakubowicz, A.; Jia, H.; Wallace, R. M.; Gnade, B. E. Langmuir 2005, 21, 950-955.
5080
Analytical Chemistry, Vol. 79, No. 13, July 1, 2007
Figure 4. (top) Signal-to-noise plots of the TMFs used to characterize the vapor-phase thiol deposition apparatus. Each plot corresponds to a different TMF evaluated in the flow cell. (bottom) S/N data at early times for the TMFs. To ensure a conservative estimate of the LOD, the point at which the S/N achieved a value of 3 and never fluctuated below 3 again was used as the time to determine the surface coverage.
Θ(t) ) K[1 - exp( - kobst)]
(2)
where Θ(t) is the coverage as a fuction of time, kobs) kaC + kd, and K is the equilibrium surface coverage; in this work, K was assumed to be a monolayer (K ) 1) because there should have been minimal desorption back into the vapor phase once the BT reacted with the Ag surface. That assumption was based on a ∆Gads in the range of -60 to -140 kJ/mol, similar to values for thiols bound to gold.22-25 The time-dependent trends were normalized using the plateau value for the intensity in order to yield surface coverage as a function of time, Θ(t). Normalized adsorption isotherms for BT on 5-nm Ag TMFs are shown in Figure 3. The isotherms were fit to eq 2 via nonlinear regression. Data analysis indicated that the adsorption of benzenethiol on Ag from the vapor phase is well described by a Langmuir mechanism. Three of the four Langmuir fit curves had R2 values of 0.92 or greater, and the fourth fit curve had a R2 value of 0.83. The kobs values obtained from nonlinear regression to the data ranged from 0.6 × 10-2 to 2.6 × 10-2 s-1. (23) Shadnam, M. R.; Amirfazli, A. Chem. Commun. 2005, 4869-4871. (24) Nara, J.; Higai, S.; Morikawa, Y.; Ohno, T. J. Chem. Phys. 2004, 120, 67056711. (25) Kolega, R. R.; Schlenoff, J. B. Langmuir 1998, 14, 5469-5478.
Figure 5. Signal-to-noise ratio of BT 1070-cm-1 band through time. A horizontal line has been added to indicate when the BT signal reached the minimum signal to noise for a positive identification. The vertical line indicates the data point used to derive the surface coverage of BT at the point of the LOD.
The limit of detection for these substrates was determined by evaluating the signal-to-noise ratio on each adsorption curve, as shown in Figure 4. The signal-to-noise ratio in this case was determined using the baseline-corrected signal of the 1070-cm-1 BT peak. Noise was determined by taking the standard deviation of a 45 cm-1 (50 pixels in the spectral dimension of the CCD) wide flat portion of the BT spectra. To ensure a conservative estimate of S/N, the maximum spectral noise for the first 100 spectra was used. The fluctuation of the BT peak at monolayer coverage could have been used, but this would have provided an unrealistically low spectral noise value because at long times it was consistently observed that the background SERS diminished. Using the plots of surface coverage versus time, it was possible to calculate the time, and therefore coverage, at which the BT signal was 3× the noise. The average LOD of a range of Ag TMF substrates was determined to be (1.1 ( 0.7) × 10-2 fmol within the excitation spot. The value for the TMF LOD was in reasonable agreement, considering the differences in optical collection efficiency, with the work of Norrod et al.,4 who determined a LOD for trans-1,2-bis-4-pyridylethene on 5.0-nm Ag TMFs of (4.2 ( 0.3) × 10-1 fmole. The method was therefore deemed reliable for LOD determination.
Nanoaperture Array LOD. The same procedure for a vaporphase SERS LOD was applied to a 350-nm lattice spacing nanoaperture array with a Ag film thickness of 50 nm. The raw adsorption data were fit to a Langmuir adsorption mechanism to determine the signal at monolayer coverage. This raw signal was normalized to the plateau value to give a surface coverage as a function of time. An R2 value of 0.926 indicated this adsorption profile was well fit by a Langmuir adsorption isotherm. The Langmuir fit yielded a kobs of (2.4 ( 0.2) × 10-2. This value is within the range of values observed for the kobs from the TMF study. Figure 5 shows the signal-to noise values as a function of time for BT on the nanoaperture array. The coverage of BT at the LOD was determined to be 0.12 of a monolayer. The LOD at this coverage was calculated to be 4.2 ( 0.3 amol within the excitation spot. Here the precision was determined from the standard deviation of the adsorption profile. The slightly lower LOD is consistent with a SERS enhancement from the nanoaperture array.2 CONCLUSIONS A new method of performing a SERS LOD study was presented. The design and implementation of a vapor-phase thiol deposition apparatus allowed for in situ monitoring of selfassembled monolayer formation on discontinuous Ag island films and on enhanced transmission nanoaperture arrays. The LOD for the Ag island films was determined to be in adequate agreement with previous work. The LOD determined for a nanoaperture array with a lattice spacing of 350 nm was 4.2 ( 0.3 amol. The results indicate that the method of evaluating a SERS sensor based on the adsorption isotherm of a volatile analyte offers an alternate means of evaluating SERS substrates that have areas of only a few hundred square micrometers. ACKNOWLEDGMENT The authors gratefully acknowledge funding from the AFOSR MURI grant program and the NSF.
Received for review January 21, 2007. Accepted April 22, 2007. AC070121I
Analytical Chemistry, Vol. 79, No. 13, July 1, 2007
5081
