Zachary D. Schultz

Surface enhanced Raman scattering (SERS), and the high-resolution imaging analog tip-enhanced Raman scattering (TERS), have emerged as ultrasensitive methods of chemical analysis (1). SERS derives from Raman scattering, and is a nondestructive, label-free, chemicalspecific method. In Raman spectroscopy, laser radiation incident on a molecule is inelastically scattered at frequencies that can be correlated to the vibrational modes of the molecule (2). These vibrational modes are similar to the energy levels probed in infrared spectroscopy and arise from how elements are connected within a molecule. The energy of each vibrational mode is sensitive to interactions with other molecules, temperature, and other conditions, providing insight into physical properties. Additionally, the unique organization and combination of elements that compose a molecule give rise to unique combinations of bands in the Raman spectrum, which is often considered a molecular fingerprint for identification. The challenge with Raman spectroscopy is that the probability of a photon inelastically scattering is very low (about a billion times less than that of a photon fluorescing). Fortunately, in the presence of silver and gold nanostructures, SERS can enhance Raman scattering signals by as much as 1011, such that single molecules can be investigated (3).

The general phenomenon responsible for SERS is shown in Figure 1 and arises from an intense localized electromagnetic field on the nanostructures when the laser excites a localized surface plasmon resonance (LSPR) (4). The LSPR is the collective oscillation of electrons in the conduction band of the metal nanoparticle. Silver and gold nanoparticles are most commonly used because their LSPR is found in the visible region of the electromagnetic spectrum, where most Raman spectroscopy experiments are performed. Other materials can also generate SERS; for example, aluminum nanoparticles exhibit a LSPR at ultraviolet wavelengths and can generate SERS (5). When excited, the oscillation of the electrons concentrates the laser radiation into intense local electromagenetic fields at the nanostructure surface. Molecules in close proximity (generally <5nm) to the nanostructure surface experience this field and generate increased Raman scattering. The Raman scattering occurs at a wavelength similar to the LSPR, causing the nanoparticle to further re-radiate the increased Raman scattering from nearby molecules. In essence, the nanoparticle acts as an antenna to amplify the Raman signal from molecules (6–8).

The electric field in the junction between two plasmonic nanoparticles is significantly larger than on the surface of a single particle and gives rise to very large Raman enhancements. Similarly, asymmetric nanoparticles, such as nanorods and nanostars (9,10), have higher enhancements at points and vertices. Highly enhancing locations are often called “hotspots” and have been shown to dominate the SERS signal (11). Plasmon-enhanced scattering has enabled Raman detection of single molecules and, in some applications, is more sensitive than fluorescence (12,13). At this point, it is well understood the physical interactions that generate a large SERS enhancement; however, controlling the interactions for quantitative analysis has been more challenging. The observed Raman intensity arises from the organization of the nanoparticles and the location of the analyte molecule interacting with the nanoparticles. Although nanoparticles are often modeled as spheres, small features on the surface can modulate the enhanced fields. Additionally, the spacing between particles affects the resulting fields, and forming reproducible aggregates can be challenging and affect the observed signals. Because only molecules near the surface can be detected, the signal saturates at higher concentrations.

The plasmonic effects that give rise to SERS can also be localized at the apex of a scanning probe microscope tip to perform TERS (14,15). This approach has been demonstrated using both atomic force microscopy (AFM) and scanning tunneling microscopy (STM) to control the position of the tip. The advantage of AFM-based TERS is it can be done on almost any sample, while STM–TERS requires a conductive sample. In either approach, a plasmonic metal nanoparticle (or sharp metal point) is located at the very apex of the tip to generate the Raman signal enhancement from molecules beneath the tip. By controlling the location and configuration of the enhancing nanostructure, high sensitivity spectroscopy can be obtained from precise locations on a sample (16). Figure 2 illustrates how, by scanning the sample under the TERS tip in the laser focus, high spatial-resolution images of samples can be obtained. In ultrahigh-vacuum (UHV) experiments, where molecular motion can be minimized, TERS imaging within a molecule has been demonstrated. Under ambient conditions, TERS experiments are capable of imaging with a spatial resolution of 1 to 2 nm (17–19).

A number of applications of SERS and TERS have been reported in the literature. Many SERS applications use the enhanced signal for trace detection. A large number of papers have been published describing many ways to generate a SERS substrate, and commercial solutions are available.

One common use is to create “SERS tags,” where a reporter molecule with a strong Raman response is attached to the surface of the nanoparticle along with some kind of recognition agent (such as an antibody or nucleic acid sequence) (20,21). The unique SERS signal from the reporter molecule is used to quantify, and image, the detected analyte. Although this is similar to fluorescence, SERS tags do not photobleach, and the narrow Raman bands allow multiple different tags to be detected and unmixed simultaneously (22). 

Other recent work has investigated novel ways to utilize the molecular enhancements for chemical sensing, such as SERS pipettes for optophysiology (23) and online detection with chromatographic separations (24). In addition to analytical experiments, new physical understanding is being uncovered at length scales suggesting picoscience (25). As SERS and TERS continue to mature, one can expect to see these techniques enabling new physical insights and methods of chemical analysis.

1. J. Langer, D. Jimenez de Aberasturi, J. Aizpurua, R.A. Alvarez-Puebla, B. Auguié, J.J. Baumberg, G.C. Bazan, S.E.J. Bell, A. Boisen, A.G. Brolo, J. Choo, D. Cialla-May, V. Deckert, L. Fabris, K. Faulds, F.J. García de Abajo, R. Goodacre, D. Graham, A.J. Haes, C.L. Haynes, C. Huck, T. Itoh, M. Käll, J. Kneipp, N.A. Kotov, H. Kuang, E.C. Le Ru, H.K. Lee, J.-F. Li, X.Y. Ling, S. Maier, T. Mayerhoefer, M. Moskovits, K. Murakoshi, J.- M. Nam, S. Nie, Y. Ozaki, I. Pastoriza-Santos, J. PerezJuste, J. Popp, A. Pucci, S. Reich, B. Ren, G.C. Schatz, T. Shegai, S. Schlücker, T. Li-Lin, K.G. Thomas, Z.-Q. Tian, R.P. Van Duyne, T. Vo-Dinh, Y. Wang, K.A. Willets, C. Xu, H. Xu, Y. Xu, Y.S. Yamamoto, B. Zhao, and L.M. Liz-Marzán, 

Modern Raman Spectroscopy: A Practical Approach

 (John Wiley & Sons, Hoboken, New Jersey, 2005).

4. P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.R. Van Duyne, 

5. S. Tian, O. Neumann, M.J. McClain, X. Yang, L. Zhou, C. Zhang, P. Nordlander, and N.J. Halas, 

6. K. Yoshida, T. Itoh, H. Tamaru, V. Biju, M. Ishikawa and Y. Ozaki, 

7. P. Bharadwaj, R. Beams, and L. Novotny, 

9. B. Nikoobakht, Z.L. Wang, and M.A. El-Sayed, 

11. Y. Fang, N.H. Seong, and D.D. Dlott, 

12. D. Graham, R. Stevenson, D.G. Thompson, L. Barrett, C. Dalton, and K. Faulds, 

13. I. A. Larmour, K. Faulds, and D. Graham, 

14. R.M. Stockle, Y.D. Suh, V. Deckert, and R. Zenobi, 

15. N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, 

16. R. Zhang, Y. Zhang, Z.C. Dong, S. Jiang, C. Zhang, L.G. Chen, L. Zhang, Y. Liao, J. Aizpurua, Y. Luo, J.L. Yang, and J.G. Hou, 

18. Z. He, Z. Han, M. Kizer, R.J. Linhardt, X. Wang, AM. Sinyukov, J. Wang, V. Deckert, A.V. Sokolov, J. Hu, and M.O. Scully, 

19. T. Deckert-Gaudig, E. Kaemmer, and V. Deckert, 

20. D. Graham, B.J. Mallinder and W.E. Smith, 

21. T. Vo Dinh, K. Houck, and D. L. Stokes, 

23. F. Lussier, D. Missirlis, J. P. Spatz, and J. F. Masson, 

24. L. Xiao, C. Wang, C. Dai, L.E. Littlepage, J. Li, and Z.D. Schultz, 

25. J. Lee, K.T. Crampton, N. Tallarida, and V.A. Apkarian, 

Articles

SERS and TERS

It is fairly easy to generate a surface enhanced Raman scattering (SERS) signal. However, there are severalissues to be aware of when trying to use these signals in analytical applications. The SERS signal arises from the combination of the number of molecules (

), the polarizability or cross section of the molecule (σ), and the electric field (

) experienced by the molecules. Understanding how these different variables interact to generate the enhanced SERS response is the key to applying SERS accurately. Here, we try to share experiences and some helpful references to assist people interested in applications using SERS.

It has been more than40 years since the realization that Raman signals are enhanced on nanostructured metal surfaces (1). Figure 1 illustrates the advantages of surface enhanced Raman scattering (SERS), which provides a larger signal from fewer molecules. Over those past 40 years, significant progress has been made toward understanding the origin of the signal enhancement, enabling SERS to become a promising technique for chemical-specific analytical analysis. It is now understood how the electric fields originating from localized surface plasmon resonances on nanoparticles can concentrate the incident laser and generate significant enhancements in Raman scattering from molecules at the surface (2). In particular, nanoparticle dimers and larger aggregates with nanometer-scale gaps and crevices produce some of the largest electromagnetic field enhancements (3). In fact, it is fairly easy to generate a SERS signal; however, as many SERS practitioners will testify, there are a few things to be aware of when trying to use these signals in analytical applications (Table I). The SERS signal arises from the combination of the number of molecules (

) experienced by the molecules. Understanding how these different variables combine to generate the SERS response is the key to applying SERS. Here, I share my experiences and try to provide some helpful references to help people interested in applications using SERS. The following sections highlight some key concepts that people should be aware of when using SERS.

Much of the pioneering work in SERS has been performed using a molecule such as rhodamine or other molecules that have electronic resonance in the visible region (2). These experiments are often called 

surface enhanced resonance Raman scattering

 (SERRS). Just as resonance Raman increases the cross-section (σ) of spontaneous Raman scattering, it also enables the lowest limits of detection in SERS experiments. Most people studying single molecules use this additional enhancement (4). It turns out that many of these molecules are fluorescent dyes, and even with excitation away from the peak absorption, some level of pre-resonance is accessed boosting the SERS response.

The other class of molecules that is commonly detected in SERS is aromatic thiols and pyridines. Although these molecules do not have an intrinsic resonance, they are reported to form charge transfer complexes on the plasmonic surfaces that also increase the relevant cross-section of the analyte and boost the enhancement. These charge-transfer mechanisms are commonly associated with “chemical enhancement” (which has become a default category for enhancement that is not from the enhanced electric field).

Some molecules are just difficult. Glucose is one example. Although glucose has been detected by SERS, this often involves some form of surface functionalization or capture, such as reacting with boronic acid on the surface (5,6). It is often worth measuring the spontaneous Raman limit of detection for a molecule; with SERS one should be able to detect the molecule at a lower concentration. This is one way to distinguish SERS from weak spontaneous Raman scattering. However, if molar concentrations are needed to see the spontaneous Raman signal, the SERS signal may be difficult to measure.

SERS Enhances the Signal from the Molecules at the Surface

The SERS effect is a very short-range enhancement. Studies done on the distance dependence of SERS show the signal enhancement goes away within a few nanometers (7,8). Other studies indicate the molecule is physisorbed to the surface (9). There was even discussion of a “first layer effect,” where the molecules adsorbed to the surface showed a larger enhancement (10).The number (

) of molecules at the surface can often be modeled with a Langmuir isotherm, which correlates the surface coverage to the affinity of the molecule to the surface. At concentrations where your molecule is unlikely to have significant coverage, it is unlikely that one will detect SERS. Both electromagnetic and chemical enhancement mechanisms increase the signal of molecules at the surface. For nanoparticles dropped onto a substrate and dried, the molecules that stick to the nanoparticles are easier to detect.

The idea of a SERS tag is to put a molecule on the surface that gives a strong signal (see above) and then incorporate some form of chemical recognition agent (such as DNA, an antibody, or an aptamer) that binds the target molecule (11,12). Although these approaches have been very successful, it is worth remembering that your limit of detection and sensitivity will now be controlled by the chemical recognition agent on your tag.

Intensities Depend on the Nanostructures, Too

It has been shown that the large majority of the SERS signal originates from “hotspots,” (13) generally regarded as the high electric-field (

) regions associated with gaps and crevices. This means that one molecule in an electric field enhanced region of 106 will be as bright as either 106 molecules in solution, or 105 molecules in a region with an electric field enhancement of 10. This tells us that small changes in the number of molecules in “hotspots” can create large intensity variations. This is particularly problematic when using colloidal nanoparticles, where it can be very challenging to aggregate nanoparticles in a reproducible manner.

A lot of effort has gone into making patterned nanostructures as SERS substrates (14). Many of these substrate approaches still have some variation in the electric field enhancements, and it is common to observe intensity variation on the scale of 10% in these systems. One approach to averaging out this heterogeneity is to measure multiple spots. One study suggested that more than 100 spots are needed to properly capture this variance (15).

This is not to say that SERS cannot be quantitative (16). The use of internal standards, assumed to have the same intensity distribution as an analyte, can correct for this variance. This correction can be a co-adsorbed molecule (17), or, perhaps preferably, a stable isotope variant of the target molecule (18).

The metric used to compare the enhancement observed from a SERS active material is the enhancement factor (

) (19). In theory, this number represents the increase in the Raman signal expected from a molecule. In practice, this is really difficult to determine. It requires that you measure a SERS signal (

SERS), and you have the spontaneous Raman signal (

Raman) from a known number of molecules (

Raman) under similar experimental conditions. With these four numbers, the EF can be calculated as follows:

In perhaps the simplest case, a monolayer of thiol on a patterned nanostructured surface, approximations are still required. Because of the influence of hotspots, the number of molecules detected by SERS (

SERS) is difficult to know. Further, the free molecule may have a different cross-section than the adsorbed molecule, making 

Raman also uncertain. Given these uncertainties, the 

 is at best a relative number for the conditions under which it was determined. When reporting 

s, it is important to provide as much information as possible about how the 

 was determined. As noted above, just because one molecule enhances well, another type of molecule may have a very different 

The Molecule You Put in Your Sample May Not Be What You Detect

One frustrating aspect of SERS is the vibrational frequencies observed may vary from those of the molecules in solution or powder forms, such as is observed in Figure 1. In SERS, there is a strong polarization dependence that selectively enhances those modes aligned with the enhanced electric field. In the past 10 years, experiments show that the electrons in the metals, which generate the plasmon resonance and enhance the Raman signal, can also do chemistry on the analytes. One classic example is para-aminothiophenol, where theories on changes in selection rules to activate Raman-inactive B2 modes were developed, only to find out the new frequencies arose from the formation of dimercaptoazobenzene on the surface (20). Understanding these effects is an active area of SERS research today. To avoid these effects, it is important to use low laser powers, generally less than 1 mW of power, in a diffraction-limited focus of a microscope objective focus. These low powers can minimize these photoreactions as well as heating that can cause problems in your measurement. For those seeking to use SERS, the best practice is to generate a calibration curve with known concentrations of your analyte at low laser powers, and use a strong band (or multivariate model) that provides the most accurate concentration determination.

P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.R. Van Duyne, 

K.L. Wustholz, A.-I. Henry, J.M. McMahon, R.G. Freeman, N. Valley, M.E. Piotti, M.J. Natan, G.C. Schatz,and R.P. Van Duyne, 

B. Sharma, P. Bugga, L.R. Madison, A.I. Henry, M.G. Blaber, N.G. Greeneltch, N.H. Chiang, M. Mrksich, G.C. Schatz, and R.P. Van Duyne,

X. Gu, H. Wang, Z.D. Schultz, and J.P. Camden, 

J.A. Dieringer, A.D.McFarland, N.C. Shah, D.A. Stuart, A.V. Whitney, C.R. Yonzon, M.A. Young, X. Zhang, and R.P. Van Duyne, 

S. Lal, N.K. Grady, G.P. Goodrich, and N.J. Halas, 

D. Graham, B.J. Mallinder, and W.E. Smith, 

N.R. Isola, D.L. Stokes, and T. Vo-Dinh, 

B. Sharma, M.F. Cardinal, S.L. Kleinman, N.G. Greeneltch, R.R. Frontiera, M.G. Blaber, G.C. Schatz, and R.P. Van Duyne, 

A. Subaihi, Y. Xu, H. Muhamadali, S.T. Mutter, E.W. Blanch, D.I. Ellis, and R. Goodacre, 

P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.P. Van Duyne, 

Y.-F. Huang, H.-P. Zhu, G.-K.Liu, D.-Y. Wu, B. Ren, and Z.-Q. Tian, 

is with the Department of Chemistry and Biochemistry at The Ohio State University, in Columbus, Ohio. Direct correspondence to:


SERS offers many advantages. However, there are several issues to be aware of when trying to use SERS signals in analytical applications.

Demystifying SERS: A Newcomer’s Guide to Using Surface Enhanced Raman Scattering

	Advances in instrumentation and data analysis suggest that surface enhanced Raman scattering (SERS) is a powerful tool to characterize complex samples. Since the initial experiments and subsequent realization that Raman signals are significantly enhanced on the surface of plasmonic, structured, metal surfaces, surface enhanced Raman scattering (SERS) has evolved into an ultrasensitive method for chemical analysis (1–3). Indeed, understanding of how light is concentrated and how the local fields are enhanced and re-radiated by plasmonic structures has driven SERS research forward (4). Advances in SERS substrates, now commercially available from a number of sources, and improved sampling strategies, suggest that SERS is poised to be potent analytical technique. Indeed, it is now relatively straightforward to produce a SERS signal, and with this greater understanding of signal enhancements, increased attention is being given to applications using SERS.

	However, since the early days of SERS, one of the more befuddling aspects of SERS analysis has been that the SERS spectrum observed from a molecule is not always the same as the spontaneous Raman spectrum. The origin of these spectral differences has been attributed to surface selection rules, charge-transfer effects, photo-reactions, and other phenomena (5–8). Indeed, these signal variations have raised suggestions that libraries of SERS spectra are needed to properly identify analytes in mixtures. This need for libraries is complicated by observations that the SERS signal from a sample can vary depending on common variables like the substrate or laser excitation wavelength.

	Although the identity of an unknown molecule, or molecules, in a complex system (such as human biofluids) that give rise to a SERS signal can be elusive, new advances in computing are enabling SERS analysis of complex systems through machine learning and related algorithms (9). The detected bands in a SERS spectrum provide an information-rich signal that can be analyzed with computer algorithms, even if an analyst does not know the molecular origin of the signal. Algorithms include neural networks, random forest, and artificial intelligence algorithms, as well as multivariate regression and other chemometric routines. With a rise in companies producing commercial software and the availability of web-based shareware routines, the activation barrier to these analyses methods continues to decrease.

	The underlying principle of machine learning is that given enough data, a computer can use small fluctuations and patterns to classify the outcome. The most common approach is to use a training set‚ samples of known classes to generate a neural network or other machine learning model that will properly capture the variance in the sample data and predict to what class an unknown sample belongs. As the cost of computing power decreases, more-sophisticated methods are accessible to analyze data. The clear advantage of machine learning is that the chemically rich SERS signal, where each vibrational mode detected acts as a variable, can classify the complex sample signals as arising from different specific classes, such as healthy or diseased cells (Figure 1). As the instrumentation and methods using SERS increase the ability to generate large amounts of SERS data quickly, these analyses provide an incredibly powerful way to utilize the chemical information in the SERS spectra for diagnostic assays and to monitor chemical changes. With increased training sets, there is potential to refine such classes further, such as to determine which treatment will likely be most effective for a disease.

	The downside to machine learning algorithms is that the models rely on appropriate training data and, for the most part, can only classify a sample into classes that have been predetermined. Both supervised and unsupervised methods exist, but all models require validation against data sets with known classifications. A further downside to these methods is the loss of information. Although the variance in the SERS data arises from molecular differences, the identity of these molecules can be difficult to recover. Returning to the biofluid example, the machine learning algorithm may classify a sample correctly; however, information, such as a specific upregulated biomolecule to be targeted for therapy, is lost. Complementary methods are often necessary to further assess the disease.

	In a recent study we were able to perform SERS detection following a high-performance liquid chromatography (HPLC) separation of a mouse tumor cell lysate and show that the molecular composition significantly changed between healthy and various types of cancerous cells (10). The molecules with consistent increases in concentration generated patterns that appear to be characteristic of the sampled cells. These patterns can be readily differentiated by machine learning algorithms. Again, although the identity of the molecules detected by HPLC are unknown, computer models can use the patterns in the observed spectra to track and diagnose disease. Additionally, analysis of the underlying data may indicate key biomolecules important in the disease, without knowing the identity of the actual molecules. In such cases where knowing the identity of a specific molecule is not necessary, the ability to rapidly generate SERS spectra and analyze them with machine learning makes it possible to capture the chemical information in SERS data to enable informative characterization.

M. Fleischmann, P.J. Hendra, A.J. McQuillan, Chem. 

J. Electroanal. Chem. Interfacial Electrochem. 

J. Langer, D. Jimenez de AberÂ­asturi, J. Aizpurua, R.A. Alvarez- Puebla, B. Auguié, J.J. BaumÂ­berg, G.C. Bazan, S.E.J. Bell, A. Boisen, A.G. Brolo, J. Choo, D. Cialla-May, V. Deckert, L. Fabris, K. Faulds, F.J. García de Abajo, R. Goodacre, D Graham, A.J. Haes, C.L. Haynes, C Huck, T. Itoh, M. Käll, J. Kneipp, N.A. Kotov, H. Kuang, E.C. Le Ru, H.K. Lee, J.-F. Li, X.Y. Ling, S.A. Maier, T. Mayerhöfer, M. Moskovits, K. Murakoshi, J.-M. Nam, S. Nie, Y. Ozaki, I. Pastoriza-Santos, J. Perez-Juste, J. Popp, A Pucci, S Reich, B Ren, G.C. Schatz, T. Shegai, S. Schlücker, L.-L. Tay, K.G. Thomas, Z.-Q. Tian, R.P. Van Duyne, T. Vo-Dinh, Y. Wang, K.A. Willets, C Xu, H. Xu, Y. Xu, Y.S YaÂ­mamoto, B. Zhao, and L.M. Liz- Marzán, 

J. SzczerbiÅski, L. Gyr, J. Kaeslin, and R. Zenobi, 

Y. Zhang, S. He, W. Guo, Y. Hu, J. Huang, J.R. Mulcahy, and W.D Wei, 

Y.-F. Huang, H.-P. Zhu, G.K. Liu, D.Y. Wu, B. Ren, and Z.Q. Tian, 

F. Lussier, V. Thibault, B. Charron, G.Q. Wallace, and J.-F. Masson, 

L. Xiao, C. Wang, C. Dai, L.E. Littlepage, J. Li, and Z.D. Schultz, 

is an associate professor at The Ohio State University. Direct correspondence to 

In celebration of Spectroscopy’s 35th Anniversary, leading spectroscopists discuss important issues and challenges in analytical spectroscopy.

Rise of the Machines: SERS Instrumentation and Machine Learning Enabling Complex Bioanalysis

It is fairly easy to generate a surface enhanced Raman scattering (SERS) signal. However, there are several  issues to be aware of when trying to use these signals in analytical applications. The SERS signal arises from the combination of the number of molecules (

		
		It has been more than 40 years since the realization that Raman signals are enhanced on nanostructured metal surfaces (1). Figure 1 illustrates the advantages of surface enhanced Raman scattering (SERS), which provides a larger signal from fewer molecules. Over those past 40 years, significant progress has been made toward understanding the origin of the signal enhancement, enabling SERS to become a promising technique for chemical-specific analytical analysis. It is now understood how the electric fields originating from localized surface plasmon resonances on nanoparticles can concentrate the incident laser and generate significant enhancements in Raman scattering from molecules at the surface (2). In particular, nanoparticle dimers and larger aggregates with nanometer-scale gaps and crevices produce some of the largest electromagnetic field enhancements (3). In fact, it is fairly easy to generate a SERS signal; however, as many SERS practitioners will testify, there are a few things to be aware of when trying to use these signals in analytical applications (Table I). The SERS signal arises from the combination of the number of molecules (

) experienced by the molecules. Understanding how these different variables combine to generate the SERS response is the key to applying SERS. Here, I share my experiences and try to provide some helpful references to help people interested in applications using SERS. The following sections highlight some key concepts that people should be aware of when using SERS. 

	Much of the pioneering work in SERS has been performed using a molecule such as rhodamine or other molecules that have electronic resonance in the visible region (2). These experiments are often called 

 (SERRS). Just as resonance Raman increases the cross-section (σ) of spontaneous Raman scattering, it also enables the lowest limits of detection in SERS experiments. Most people studying single molecules use this additional enhancement (4). It turns out that many of these molecules are fluorescent dyes, and even with excitation away from the peak absorption, some level of pre-resonance is accessed boosting the SERS response.       

	The other class of molecules that is commonly detected in SERS is aromatic thiols and pyridines. Although these molecules do not have an intrinsic resonance, they are reported to form charge transfer complexes on the plasmonic surfaces that also increase the relevant cross-section of the analyte and boost the enhancement. These charge-transfer mechanisms are commonly associated with “chemical enhancement” (which has become a default category for enhancement that is not from the enhanced electric field).

	Some molecules are just difficult. Glucose is one example. Although glucose has been detected by SERS, this often involves some form of surface functionalization or capture, such as reacting with boronic acid on the surface (5,6). It is often worth measuring the spontaneous Raman limit of detection for a molecule; with SERS one should be able to detect the molecule at a lower concentration. This is one way to distinguish SERS from weak spontaneous Raman scattering. However, if molar concentrations are needed to see the spontaneous Raman signal, the SERS signal may be difficult to measure.

	The SERS effect is a very short-range enhancement. Studies done on the distance dependence of SERS show the signal enhancement goes away within a few nanometers (7,8). Other studies indicate the molecule is physisorbed to the surface (9). There was even discussion of a “first layer effect,” where the molecules adsorbed to the surface showed a larger enhancement (10).The number (

	The idea of a SERS tag is to put a molecule on the surface that gives a strong signal (see above) and then incorporate some form of chemical recognition agent (such as DNA, an antibody, or an aptamer) that binds the target molecule (11,12). Although these approaches have been very successful, it is worth remembering that your limit of detection and sensitivity will now be controlled by the chemical recognition agent on your tag.

	It has been shown that the large majority of the SERS signal originates from “hotspots,” (13) generally regarded as the high electric-field (

) regions associated with gaps and crevices. This means that one molecule in an electric field enhanced region of 10

 molecules in a region with an electric field enhancement of 10. This tells us that small changes in the number of molecules in “hotspots” can create large intensity variations. This is particularly problematic when using colloidal nanoparticles, where it can be very challenging to aggregate nanoparticles in a reproducible manner.

	A lot of effort has gone into making patterned nanostructures as SERS substrates (14). Many of these substrate approaches still have some variation in the electric field enhancements, and it is common to observe intensity variation on the scale of 10% in these systems. One approach to averaging out this heterogeneity is to measure multiple spots. One study suggested that more than 100 spots are needed to properly capture this variance (15).

	This is not to say that SERS cannot be quantitative (16). The use of internal standards, assumed to have the same intensity distribution as an analyte, can correct for this variance. This correction can be a co-adsorbed molecule (17), or, perhaps preferably, a stable isotope variant of the target molecule (18).

	The metric used to compare the enhancement observed from a SERS active material is the enhancement factor (

), and you have the spontaneous Raman signal (

) under similar experimental conditions. With these four numbers, the EF can be calculated as follows:

	In perhaps the simplest case, a monolayer of thiol on a patterned nanostructured surface, approximations are still required. Because of the influence of hotspots, the number of molecules detected by SERS (

) is difficult to know. Further, the free molecule may have a different cross-section than the adsorbed molecule, making 

 also uncertain. Given these uncertainties, the 

	One frustrating aspect of SERS is the vibrational frequencies observed may vary from those of the molecules in solution or powder forms, such as is observed in Figure 1. In SERS, there is a strong polarization dependence that selectively enhances those modes aligned with the enhanced electric field. In the past 10 years, experiments show that the electrons in the metals, which generate the plasmon resonance and enhance the Raman signal, can also do chemistry on the analytes. One classic example is para-aminothiophenol, where theories on changes in selection rules to activate Raman-inactive B

 modes were developed, only to find out the new frequencies arose from the formation of dimercaptoazobenzene on the surface (20). Understanding these effects is an active area of SERS research today. To avoid these effects, it is important to use low laser powers, generally less than 1 mW of power, in a diffraction-limited focus of a microscope objective focus. These low powers can minimize these photoreactions as well as heating that can cause problems in your measurement. For those seeking to use SERS, the best practice is to generate a calibration curve with known concentrations of your analyte at low laser powers, and use a strong band (or multivariate model) that provides the most accurate concentration determination.

		P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.R. Van Duyne, 

		K.L. Wustholz, A.-I. Henry, J.M. McMahon, R.G. Freeman, N. Valley, M.E. Piotti, M.J. Natan, G.C. Schatz,and R.P. Van Duyne, 

		B. Sharma, P. Bugga, L.R. Madison, A.I. Henry, M.G. Blaber, N.G. Greeneltch, N.H. Chiang, M. Mrksich, G.C. Schatz, and R.P. Van Duyne,  

		X. Gu, H. Wang, Z.D. Schultz, and J.P. Camden, 

		J.A. Dieringer, A.D.McFarland, N.C. Shah, D.A. Stuart, A.V. Whitney, C.R. Yonzon, M.A. Young, X. Zhang, and R.P. Van Duyne, 

		S. Lal, N.K. Grady, G.P. Goodrich, and N.J. Halas, 

		D. Graham, B.J. Mallinder, and W.E. Smith, 

		N.R. Isola, D.L. Stokes, and T. Vo-Dinh, 

		B. Sharma, M.F. Cardinal, S.L. Kleinman, N.G. Greeneltch, R.R. Frontiera, M.G. Blaber, G.C. Schatz, and R.P. Van Duyne, 

		A. Subaihi, Y. Xu, H. Muhamadali, S.T. Mutter, E.W. Blanch, D.I. Ellis, and R. Goodacre, 

		P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.P. Van Duyne, 

		Y.-F. Huang, H.-P. Zhu, G.-K.Liu, D.-Y. Wu, B. Ren, and Z.-Q. Tian, 

is with the Department of Chemistry and Biochemistry at The Ohio State University, in Columbus, Ohio. Direct correspondence to: 

The SERS signal arises from the combination of the number of molecules, the polarizability or cross-section of the molecule, and the electric field experienced by the molecules. Understanding how these variables interact to generate the SERS response is the key to applying SERS accurately.

A Newcomer’s Guide to Using Surface Enhanced Raman Scattering

Thin films of noble metals deposited by physical vapor deposition (PVD) are a simple substrate commonly used in surface-enhanced spectroscopies. The morphology and gap spacing of these plasmonic substrates are key factors in the Raman enhancements observed from molecules in the optical near field. The nanometer gaps between nanostructured features of the substrate are associated with the strongest fields and the greatest enhancement of light-driven processes (scattering, absorption, fluorescence) that occur near the surface. Results shown here illustrate that chemical adsorbates and solvents interacting with the substrate can induce changes in the atomic, nanoscale, and mesoscale morphology of the plasmonic substrate, possibly leading to in situ species that play a role in the enhancement mechanisms associated with plasmonic nanostructures. In this work, we report on the morphology changes induced by functionalizing nano-island films (NIF) of silver and gold (film thickness = 4–30 nm) in thiol solutions (thiophenol and 1-hexanethiol). The adsorbate substrate interaction is hypothesized to cause aggregation or annealing-like surface rearrangements. NIF substrates of thickness <10 nm exhibit more significant changes than thicker films, as evidenced by SEM and UV-vis data. Insights into substrate design to account for analyte substrate interactions for surface enhanced Raman spectroscopy (SERS) are discussed in the context of strongly and weakly adsorbing analytes.

	The enhancement of Raman scattering from molecules on the surface of plasmonic nanostructures (so-called 

, or SERS) has been studied extensively for trace detection of analytes in a wide variety of applications (1–5). The localized surface plasmon resonance (LSPR) supported on nanostructured metal surfaces (the most commonly studied are silver, gold, and copper) under irradiation at optical frequencies is a result of the strong interaction between the conduction electrons confined in a sub-wavelength geometry and the incident electromagnetic wave. The oscillating electromagnetic field polarizes the electron cloud in the metal, and induces a resonant oscillation of the conduction electrons, giving rise to a new field concentrated around the nanostructure. Molecules within these enhanced fields exhibit increased Raman scattering. The strongest fields are found at the tip of nanoscale features with high aspect ratio or atomically sharp edges due to the "lightning rod effect," or in the nanometer gaps of neighboring nanoparticles, termed 

, due to field enhancements from interparticle coupling (6–9).

	To capture these enhancements for trace chemical detection, considerable effort has been put into fabricating noble metal nanoparticle systems and studying their plasmonic properties. The structure, composition, and local dielectric environment of a plasmonic nanostructure influence the LSPR frequency and spatial intensity distribution of the excited field around the nanostructure. The choice of these parameters will vary depending on the application, and many fabrication methods have been developed to explore a vast design space for plasmonic substrates (10–12). These fabrication methods, ranging in complexity and performance, include nanolithography, physical vapor deposition (PVD), electrochemical roughening, colloidal nanoparticle immobilization, wet chemical reduction, and others.

	One of the first-studied and simplest methods of preparing a SERS substrate is the physical vapor deposition of silver or gold onto a glassy substrate to produce nano-island films (NIFs). Within the category of PVD, films can be produced from sputtering, thermal evaporation, electron beam evaporation, and laser ablation, which largely exhibit similar properties. Thin films of these metals do not deposit conformally on smooth, glassy surfaces, due to low surface energy and poor wetting of the metal on the surface, but instead form isolated nanostructures with gaps between particles of the order of the particle diameter or smaller, thus facilitating plasmonic activity (13–19). The experimental parameters governing the resultant morphology of as-deposited films (and typical values found in the literature) are nominal film thickness (1-15 nm), deposition rate (0.001-0.1 nm/s), substrate temperature (100-500 °K), and vacuum pressure (<10

 torr) (20-24). The morphology and optoelectronic properties of as–deposited films and those subject to various post-deposition treatments (thermal annealing [25-39], solvent annealing [31–35], chemical functionalization [21,31,36–46]) have been intensely investigated. The identity of the underlying substrate (glass [20–23, 25–27,31–40], quartz [42,45,46], calcium fluoride [16], silicon and geranium [28,41], formvar [47], and carbon [48,49]) and cleaning procedures also factor into the NIF morphology. For as–deposited films, the morphology is tunable from isolated ellipsoidal particles to randomly shaped particles coined 

 to semi-continuous film. Still, due to the random nature of the deposition process, nanoparticle size and shape distributions are rather broad, leading to some variation in the optical properties across the substrate. Above a certain thickness threshold (~20–30 nm), the film will become fully continuous and acquire a mirror-like appearance, and the plasmonic activity is significantly diminished as the average particle size is larger than the incident wavelength. Thick films (thickness >50 nm) can be prepared that display plasmonic activity given careful considerations of the deposition parameters, but these films will not be discussed here (50).

	Although the physics giving rise to the enhanced electric fields is now well understood, SERS has been plagued by circumstances and effects that created a perception of irreproducibility. Some of this irreproducibility arises from large variations in the optical properties associated with differences in the nanostructure preparation. Other variations have been attributed to chemical effects, some of which date back to the initial discovery of SERS (51,52). It is also understood that changes in the nanostructures can alter the enhancements and, in some cases, quench the electric field (53).

	The topic of this report is the much-studied thermally evaporated thin films of silver and gold (nominal film thickness <30 nm). The results of this work describe significant changes to the film morphology, and thus the corresponding optical properties, upon functionalization with self-assembled monolayers (SAMs) of thiols. Characterization by SEM and ultraviolet (UV)-vis spectroscopy indicates the formation of the thiol SAM perturbs the interparticle van der Waals forces, leading to a structural rearrangement of the film that mimics coalescence or aggregation of the nanoparticles. The observed structural rearrangement significantly alters the observed SERS enhancements.

	Silver and gold NIF substrates were prepared by PVD. White glass microscope slides (Globe) were soaked in a Nochromix acid bath (Godax Laboratories) for >12 h. The slides were removed from the cleaning bath, rinsed thoroughly with ultrapure water (Thermo Fisher Scientific GenPure, 18.2 Mâ¦ cm), and subsequently dried under nitrogen. The cleaned slides were loaded into the commercial PVD vacuum chamber (PVD Nano 36, Kurt J. Lesker Co.), the metal shot (silver, 99.999%, Sigma-Aldrich, and gold, 99.999%, Kurt J. Lesker Co.) to be resistively heated was placed in a tungsten boat, and the chamber was evacuated to a pressure of <10

 torr. A quartz crystal microbalance (QCM) located near to the deposition surface monitored the deposition rate of the metal. During the deposition, the substrates were rotated at 10 rev/s. The pre-deposition procedure for heating the metal shot to the point of evaporation involved ramping the output power set point in two stages to ~80% and ~90% of the working deposition set point (ramp rate = 0.5 units/s, soak time = 45 s), after which a proportional-integral-derivatice (PID) rate control function was enabled to stabilize the rate at the working set point (rate = 0.003 nm/s). Then, the substrate shutter was opened to begin the deposition of metal onto the substrates, which proceeded until the nominal film thickness set point was reached as determined by the QCM (thickness = 4–30 nm). The temperature of the substrate was not monitored during the deposition.

	Self-assembled monolayers of thiol (benzenethiol, benzene thiol [BT], Sigma-Aldrich, 99.9%; 1-hexanethiol, hydroxytryptophan [HT], Sigma-Aldrich, 95%; chemicals used as received) were deposited onto the as-prepared nano-island films by two methods: (1) soaking in a 10 mM ethanolic solution of the respective thiol for >12 h, after which the films were rinsed with ethanol for 10 s and subsequently dried under nitrogen. To study the impact of thiol concentration, some films were soaked in lower concentration BT solutions: 5 nM, 50 nM, 1 µM, and 50 µM; (2) positioning the substrate above a 10 µL droplet of the respective thiol inside a closed container to expose the surface to thiol vapor for 1 h; after removal no further treatment was performed.

	UV-vis transmission measurements of the thin film substrates were carried out by placing the substrate into the optical path of a VWR UV-1600PC spectrophotometer just in front of the sample cuvette holder that could not be modified to fit the microscope slides or cover slips. A blank slide or cover slip was used to obtain baseline reference spectrum. The extinction spectra were recorded using scan rate of 1 nm/s. The data were analyzed in MATLAB to determine LSPR wavelength and extinction maxima.

	Raman measurements were collected using a Renishaw inVia Raman microscope. The Renishaw mapping software was used to raster over approximately 20 µm × 20 µm areas, collecting a spectrum every 3 µm; the mapped areas varied from sample to sample but at least 50 spectra were recorded per map. The acquisition time was 1 s, and the microscope objective used had a 50x magnification and a numerical aperture (NA) of 0.75 (Olympus). The excitation power and laser was ~1 mW from a 632.8 nm HeNe laser (Thorlabs). The spectral windows collected for BT and HT were 700-1275 and 800-1350 cm

 Raman shift, respectively. The data was analyzed using MATLAB and an open-source peak-fitting routine (54) to calculate peak areas based on Gaussian fits. The number of Gaussian line shapes included in the fit was 3 for BT and 1 for HT for the selected spectral windows 900-1200, and 1080-1140 cm

 Raman shift, respectively. A linear baseline was subtracted from the spectrum before fitting to the Gaussian line shape.

	The nanoscale morphology of thin film substrates was characterized by imaging in the field emission mode of a FEI Apreo LoVac SEM. Imaging conditions were 1-2 kV, 6.3 pA, working distance 3-4 mm, stage bias -4000 V. The in-lens T1 component of the Trinity suite backscatter detectors of the SEM was used to acquire the signal.

	Figure 1 shows scanning electron micrographs of the silver NIFs obtained from different deposition thicknesses before and after functionalization with thiol. Isolated nanoparticles were observed in the 6 and 8 nm films (nominal film thickness measured during the deposition by QCM) that were roughly ellipsoidal, while a more heterogeneous mixture of shapes was found for the 14 nm film. After functionalization with thiols by soaking in ethanolic solutions, the observed particles appear closer together, suggesting smaller gap sizes between the islands formed.


	Figure 1: Scanning electron micrographs of silver nano-island films, demonstrating morphological changes upon functionalization. Images (a) and (b) correspond to two different 6 nm samples, and (b) was functionalized with benzene thiols (BT). Images (c) and (d) correspond to a 8 nm sample that was imaged before and after functionalization with BT. The scale bar for the inset image in (d) is 1 µm. Images (e) and (f) correspond to different 14 nm samples, (f) was functionalized with HT.

	This change in spacing is apparent upon visual inspection. Figure 2 shows the color of the films are also observed to change in different chemical environments. In Figure 2, the pristine NIF is observed to have a yellow hue. The extinction max of the pristine film is observed at 470 nm. When immersing the film in ethanol, a slight discoloration is observed and a shift in the extinction maxima to 490 nm is measured. When the films are allowed to dry over the course of a day, the change observed from exposure to ethanol is observed to be reversible, with the LSPR returning to near the same value as observed from the pristine film. However, upon functionalization with BT, a dramatic color change is observed visually that corresponds to change in the extinction maxima to 665 nm. The measured extinction spectrum is consistent with coupled or aggregated nanoparticles reported in the literature (55–57).


	Figure 2: Color changes in the NIF are observed when exposed to different chemicals. (a) Photograph of colored silver 8 nm films diced from the same microscope slide at different stages of treatment (wavelength of extinction maximum indicated). (Left) The as-prepared film appears yellow colored to the eye. (Center) After soaking for 12 h in ethanol and drying under nitrogen, the film exhibits a slight discoloration. The appearance nearly returns to the as-prepared state over the course of a day. (Right) After soaking for 12 h in BT, the film becomes blue-colored, and the change is permanent. All samples exhibit small shifts in color over the course of a week when exposed to ambient atmosphere. (b) UV-vis spectra of the silver 8 nm films from (a).

	The data in Figures 1 and 2 suggest a significant change in the NIFs upon adsorption of BT. To understand this effect, we investigated how different deposition conditions and functionalization concentrations affect the change in optical properties. Figure 3 shows that the single peak observed in the extinction spectrum broadens and red-shifts as the thickness of the metal deposition increases. Wavelength (Figure 3b) and peak absorbance value (Figure 3c) are shown to increase with film thickness. Similar trends were observed for both silver and gold films. The observed trends agree with previous reports of changes in the LSPR of NIFs with changes in deposition thickness (15,17,22,23).


	Figure 3: UV-vis characterization of NIF. (a) UV-vis extinction spectra acquired in transmission mode of silver (Ag) films of thickness = 2, 6, 8, 10, 12, 14, and 20 nm, indicating general trends of increased extinction, peak broadening, and a red shift of the LSPR wavelength for thicker films. (b) Plot of LSPR peak wavelength, and (c) extinction for silver (Ag) (black circles) and gold (Au) (red squares) films of various nominal thickness.

	The broadening of the extinction spectrum with increasing nominal film thickness shown in Figure 3a is commonly observed as films transition from nanomaterials to mesoscale materials (15,17,19,26). At a deposition thickness of 20 nm, the width of the feature in the extinction spectrum is consistent with a heterogeneous collection of particles with varying shapes and sizes. Analysis of extinction spectra obtained from multiple films indicates that above a deposition of 15 nm, the nano-islands percolate into larger structures; this increased interconnectedness has been observed in this regime of film thickness (15,17,26). The trends in the extinction spectra agree with the observations from the electron micrographs (Figure 1). Increased film thickness gives rise to a larger size nanoparticle consistent with increased and red-shifted extinction.

	Figure 4 illustrates the changes in extinction spectrum observed from different deposition thicknesses after functionalization with either BT or HT. In all cases, a marked change in the extinction spectrum is observed as the thiols adsorb onto the NIFs. It is well understood that the change in the dielectric environment will shift the frequency of the LSPR (the basis of surface plasmon resonance sensing [10,42,44]). The change in line shape observed from smaller thickness films (Figures 4a, b, and d, thickness < 10 nm) show splitting more consistent with plasmon hybridization from the aggregation of discrete particles. At thicker depositions (Figure 4c, 14 nm), the change in the extinction spectrum is best described by a shift in frequency but not line shape. The changes in line shape for the different thickness film is observed from both solution and gas phase adsorption (data not shown) of the thiols onto the NIFs. This indicates that the deposition thickness influences the morphological and optical changes resulting from molecular adsorption, but that adsorption is a key event.


	Figure 4: Characterization of NIF LSPR by UV-vis before and after functionalization. Spectra of silver (Ag) and gold (Au) films of thickness (a) Ag 6 nm, (b) Ag 8 nm, (c) Ag 14 nm, and (d) Au 4 nm. Solid curves correspond to measurements of as-prepared films, dash-dotted curves correspond to the same film after soaking in a 10 mM ethanolic thiol solution overnight (black = BT, red = HT).

	Given the known relationship between nanostructure and the plasmonic fields responsible for surface-enhancement effects, we sought to understand the impact of this adsorption-driven change in film structure on the observed SERS signals.

	Figure 5 plots the results obtained from the Raman mapping experiments of BT and HT-functionalized films. SERS spectra of the two molecules are shown in Figure 5b, demonstrating that BT has a much larger Raman cross-section compared to HT. A representative map analysis of a silver 8 nm NIF functionalized with BT is displayed in Figure 5a and 5c. In general, the SERS response from similarly prepared films is quite reproducible. In the map shown, the step size of approximately 3 µm, which is larger than the laser spot size (diffraction limited at approximately 500 nm), ensures that each spectrum is obtained without possible photodamage from the previous acquisition. With a 1 s integration, a large signal of approximately 30,000 counts is observed. The spectra show excellent signal-to-noise ratios. When examining the variation in intensity across the silver NIF substrate, the peak area of the 1000 cm

 band in BT changes by about 15% from the brightest to dimmest spectra.


	Figure 5: SERS mapping of silver 8 nm NIF functionalized with BT. (a) Optical image of the mapped area at 50x magnification. (b) Representative SERS spectrum of BT (black) and HT (red). The bands marked with an asterisk are used for calculating the peak area from Gaussian fits in (d). (c) Colormap of the peak area of the 1000 cm

 band of BT. The map was obtained by rastering in 3 µm steps, acquiring a spectrum at each pixel for 1 s (laser power = 0.7 mW). (d) Map-averaged peak area of the 1000 cm

 band of HT (red) vs. nominal film thickness. Error bars indicate the standard deviation for a mapped area of 50-100 pixels.

	Comparing across different films, the average SERS signal obtained from the NIF substrates shows little dependence on the deposition thickness below 15 nm (Figure 5d). Once the thickness of the deposited metal exceeds 15 nm, the SERS intensity is observed to decrease for both BT and HT. As discussed above, NIFs obtained from thicker depositions exhibit larger interconnected particles. These interconnected features decrease the density of hot spots (nanometer gaps between particles) and therefore these films exhibit less SERS. Beyond the change in observed SERS intensity with thicker deposits (thickness > 15 nm), the trends associated with the as-deposited film properties are less discernible. The correlation between the topography and SERS intensity suggests the reorganization is observed for smaller features and that once the as-deposited features coalesce into larger islands, the ability to reorganize and boost the SERS signal is diminished.

	As noted above, the identity of the adsorbate influenced the reorganization of the nano-islands in the film, where ethanol did not produce the same effect as BT or HT. In Figure 6, we examined how the concentration of BT in solution alters the extinction (indicative of the island reorganization) and the observed SERS signal. In Figure 6a, the extinction is observed to remain largely unchanged at adsorbate concentrations below 50 µM (spectra for films soaked in 5 nM and 1 µM not shown). When a film was initially soaked at a 50 nM concentration of BT, the extinction was largely unchanged. Upon resoaking the same film in 10 mM of BT, the extinction was observed to change consistent with the reorganization of the NIFs described above. The observed change with re-soaking was the same as soaking a second pristine film in a 50 µM solution. This indicates that a critical coverage of BT (or other adsorbate) is necessary to induce the reorganization of the metal islands.


	Figure 6: Characterization of BT-functionalized silver NIF at low soaking concentrations. (a) UV-vis transmission spectra series of two silver 8 nm samples after different BT treatment steps. (b) SERS spectra for four silver samples soaked in various concentrations of BT. SERS spectra were recorded with 633 nm laser excitation for 10 s at laser power 3, 0.6, 0.6, and 0.05 mW for the respective films. The inset magnifies the spectral region 950–1150 cm

 where the prominent BT modes are present in the 5 nM, 50 nM, and 50 µM samples, but not the 1 µM sample.

	In Figure 6b, the SERS spectrum observed from the NIFs is shown to change as function of adsorbate concentration. Solutions of BT below 1 µM did not evince a significant SERS signal, suggesting this concentration is below the limit of detection. Peaks are sporadically observed at frequencies consistent with BT, but do not increase appreciably in intensity until higher concentrations. At a 50 µM concentration of BT, a markedly different spectrum is observed. The peaks are observed with significantly greater intensity. The change in intensity at 50 µM is dramatically larger than the signals observed from 50 nM and 1 µM adsorbate concentrations. Additionally, the change in intensity correlates with a dramatic change in the extinction spectrum (Figure 6a) suggesting that this change in the spectrum is related to the reorganization of the NIFs.

	The characterization of silver and gold NIFs was accomplished with SEM, UV-vis, and SERS measurements. Our analysis of as-deposited films from our work agrees well with the literature with respect to the investigated deposition parameters (17,21–23,26,36). The plasmonic properties of these thin films are apparent upon visual inspection, since the films are colored and semi-transparent, as opposed to bulk metal. Similar to the Lycurgus cup, the films show differences in reflection and transmission (15,20). What has not been directly correlated in the literature is the increased SERS response that arises from the structural reorganization of the plasmonic metal islands and the analyte dependence of this effect.

	Previous studies of the morphological and optical properties of PVD thin films suggest the NIFs can be characterized as oblate spheroids with an aspect ratio (R) between 0.1 and 0.5, reflecting the island diameters are ~2-10x their height (17). Thus, with sufficient interparticle spacing (greater than a few nanometers, as is the case for thicknesses <15 nm), the LSPR of NIFs is observed to be a single peak reflecting the in-plane mode along the major axis of the isolated ellipsoids. Changes in the LSPR have been described when the films are annealed, which causes a reorganization of the islands into larger, more spheroidal particles (17,25–30), and when molecular adsorption changes the dielectric environment of the film. A number of studies have noted the change in the local dielectric environment from adsorbates such as solvents (31–35), dye molecules (36–39), silanes (58), thiols (33,42–44,46), and others (21,31,45); however, a structural rearrangement has not been widely implicated (21,45–46). Indeed, to facilitate LSPR sensing, the substrates for the deposition of NIFs were pre-functionalized with mercaptopropyltrimethoxysilane (MPTS) to improve film stabilization (44). In that case, when the NIFs deposited on MPTS-treated glass were functionalized with a dithiol, only a small red-shift of the LSPR (< 50 nm) was observed, typical of a change in the local dielectric environment with increasing coverage (thickness) of the adsorbed layer. This change indicates stabilized films provide morphological stability which may impact utility as SERS sensors. On the other hand, structural rearrangements of NIF on unmodified quartz substrates were described upon functionalization with two different molecules by measurements of the LSPR and atomic force microscopy (45,46).

	The change in morphology we claim suggests that intermolecular forces, such as increased van der Waals forces, between the nano-islands can drive aggregation. The change in the intermolecular forces derives from the coverage of the adsorbate and produces a marked change in the observed SERS signal. While not a classical "chemical effect," here the electromagnetic enhancement clearly derives from chemical functionalization, in this case changing the spacing and the local electric fields experienced by the adsorbate. Indeed, there are now SERS substrates commercially available that take advantage of this effect, using molecular adsorption to pull nano-pillars together (Silmeco, reference [59]). Our results illustrate how these aggregation affects alter the observed signals and likely affect selectivity for different analytes. Ethanol did not drive rearrangement in the same manner as the alkane thiols. There is clearly a balance between coverage and chemical properties of the adsorbate that drive the rearrangement. The changes we report in the UV-vis spectrum provide an easy experiment to assess if these effects are involved in experiments with other adsorbates and applications. Understanding these effects may enable selective sensors for future applications.

	We have shown that molecular adsorption can drive structural rearrangements in silver and gold NIFs commonly used as SERS substrates. Adsorption of a significant portion of a monolayer changes the structure of small NIFs, typically derived from deposition thicknesses less than 15 nm, that are evident in SEM, UV-vis, and SERS measurements. At low analyte coverages, the SERS spectrum is significantly weaker and does not change with concentration, suggesting stochastic effects. Above a critical coverage, changes are observed that correlated with SEM and UV-vis measurements to suggest the nano-islands are closer and coupling more strongly. These results have implications for understanding signal variation in SERS and in designing future SERS assays.

	This work was supported by National Science Foundation awards CHE-1830994 and DBI-1830153 and funding from the Ohio State University. Electron microscopy was performed at the Center for Electron Microscopy and Analysis (CEMAS) at The Ohio State University.

	(1) P. L. Stiles, J. A. Dieringer, N. C. Shah, R. P. Van Duyne, 

	(2) N. J. Halas, S. Lal, W. S. Chang, S. Link, P. Nordlander, 

	(3) A. H. Nguyen, E. A. Shangle, and Z. D. Schultz, 

	(4) F. Sun, D. D. Galvan, P. Jain, Q. Yu, 

	(5) F. Wang, S. Cao, R. Yan, Z. Wang, D. Wang, H. Yang, 

	(7) E. C. Le Ru, M. Meyer, E. Blackie, P. G. Etchegoin, 

	(8) K. L. Wustholz, A. I. Henry, J. M. McMahon, R. G. Freeman, N. Valley, M. E. Piotti, M. J. Natan, G. C. Schatz, and R. P. Van Duyne, 

	(9) K. D. Alexander, K. Skinner, S. Zhang, H. Wei, and R. Lopez, 

	(10) Y.-S. Shon, H. Y. Choi, M. S. Guerrero, and C. Kwon, 

	(11) M. Fan, G. F. S. Andrade, and A. G. Brolo, 

	(14) G. S. A. Shawki, M. G. El-Sherbiny, and F. B. Salem, 

	(15) C. A. Davis, D. R. McKenzie, and R. C. McPhedran, 

	(16) P. Dawson, K. B. Alexander, J. R. Thompson, J. W. Haas III, and T. L. Ferrell, 

	(17) R. P. Van Duyne, J. C. Hulteen, and D. A. Treichel, 

	(18) U. K. Chettiar, P. Nyga, M. D. Thoreson, A. V. Kildishev, V. P. Drachev, and V. M. Shalaev, 

	(19) G. Santoro, S. Yu, M. Schwartzkopf, P. Zhang, S. Koyiloth Vayalil, J. F. H. Risch, M. A. Rübhausen, M. Hernández, C. Domingo, and S. V. Roth, 

	(20) J. G. Bergman, D. S. Chemla, P. F. Liao, A. M. Glass, A. Pinczuk, R. M. Hart, and D. H. Olson, 

	(23) R. Gupta, M. J. Dyer, and W. A. Weimer, 

	(24) K. M. McPeak, S. V. Jayanti, S. J. P. Kress, S. Meyer, S. Iotti, A. Rossinelli, and D. J. Norris, 

	(25) G. Kalyuzhny, A. Vaskevich, G. Ashkenasy, A. Shanzer, and I. Rubinstein, 

	(26) I. Doron-Mor, Z. Barkay, N. Filip-Granit, A. Vaskevich, and I. Rubinstein, 

Optoelectron. Adv. Mater., Rapid Commun. 

	(28) E. Thouti, N. Chander, V. Dutta, and V. K. Komarala, 

	(29) M. G. Sreenivasan, S. Malik, S. Thigulla, and B. R. Mehta, 

, 247045 (2013). doi:10.1155/2013/247045.

	(32) S. E. Roark, D. J. Semin, A. Lo, R. T. Skodje, and K. L. Rowlen, 