Anisotropic metal nanoparticles for surface enhanced Raman scattering

Chem. Soc. Rev., 2017, 46,3866 Anisotropic metal nanoparticles for surface enhanced Raman scattering Javier Reguera, abc Judith Langer, ab Dorleta Jime ´nez de Aberasturi ab and Luis M. Liz-Marza ´n* abc The optimization of the enhancement of Raman scattering by plasmonic effects is largely determined by the properties of the enhancing substrates. The main parameters behind this effect are related to the morphology of plasmonic nanoparticles and their relative distribution within the substrate. We focus this tutorial review on the effects of nanoparticle morphology, for the particular case of anisotropic metal nanoparticles. Anisotropy in silver and gold nanoparticles offers the possibility to tailor their plasmonic properties and intrinsic electromagnetic ‘‘hotspots’’. We describe the effect of varying particle size and shape on the SERS signal, focusing on the most common anisotropic morphologies used for SERS. Especial emphasis is made on existing comparative studies that shed light on the effect of nanoparticle anisotropy on their enhancement capabilities. We aim at providing a general perspective toward understanding the general key factors and highlighting the difficulty in quantitatively determining SERS performance. Key learning points Some fundamentals to understand SERS Importance of anisotropy toward SERS detection The most relevant anisotropic gold and silver nanoparticles in SERS Effects of shape and morphology on SERS (comparative studies) Key factors for SERS enhancement 1. Introduction Noble metal nanoparticles (NPs) display a rich plasmonic behavior, which can be tuned via variations in the size and shape of the particles (ref. 1, 2 and references therein). Engineering of plasmon resonances in metal NPs is of great interest toward applications in very diverse fields. Sensing, light- and energy-harvesting, optoelectronics, biomedicine, and catalysis, are some of the most prominent examples. Among all metals, NPs made of silver (Ag) and gold (Au) play a central role in plasmonics as their resonances occur in the visible and infrared (IR) range of the electromagnetic spectrum, rendering them suitable for many applications. A large number of simple chemical methods have been developed for large scale production of NPs made of these two metals, featuring low polydispersity, chemical inertness, bio-compatibility, and easy functionalization, so that they offer great potential in biological, biochemical and biomedical diagnostics, both in vitro and in vivo. A classical example is the use of spherical Au NPs in flow immunoassays, such as home pregnancy tests. Engineering anisotropy in metal NPs has proven to be an extremely powerful tool for plasmon resonance tuning. Plasmon resonances in spheres can span a relatively small wavelength range (few tens of nm) by varying NP diameter, whereas shape anisotropy provides an additional degree of freedom that allows controlling plasmon resonance wavelengths, ranging from the visible through the mid IR, by varying the aspect ratio (AR) of the NPs. This is of high relevance in various sensing techniques, among which surface enhanced Raman scattering (SERS) is one of the most prominent examples. Raman scattering, which is based on inelastic light scattering upon vibrational excitation of molecules and materials is an extremely inefficient process with ca. one photon out of 10 7 being inelastically scattered. This drawback can however be overcome when the molecules are a CIC biomaGUNE, Paseo de Miramo ´n 182, 20014 Donostia-San Sebastia ´n, Spain. E-mail: llizmarz[email protected] b Biomedical Research Networking Center in Bioengineering Biomaterials and Nanomedicine, Ciber-BBN, 20014 Donostia-San Sebastia ´n, Spain c Ikerbasque, Basque Foundation for Science, 48013 Bilbao, Spain Author version TUTORIAL REVIEW located near a rough metal surface or metal NPs, which results in a boost of Raman scattering, in a process commonly known as SERS. Since its discovery in 1973, 3 the interest on SERS spectroscopy has increased tremendously, in turn pushing a fast technical development of high-end setups (with highly focused laser beams, efficient filters, highly sensitive detectors, etc.) as well as robust, potent and portable instruments. SERS spectroscopy has become an attractive technique for ultrasensitive detection, identification and, sometimes, quantification of a broad range of molecules. A wide variety of anisotropic NPs have been developed with high potential for use in SERS sensing and imaging. Opposite to spherical NPs, aggregation is not required to generate hotspots and plasmonic tunability can be readily obtained. In this tutorial, we review recent research on the use of the most common anisotropic Ag and Au NP morphologies toward SERS. We focus on the intrinsic shape-dependent SERS performance of anisotropic NPs, highlighting comparative and systematic SERS studies on NPs, mainly in solution. The article is organized as follows: in Section 2, we give a basic theoretical introduction to the optical properties of NPs and the SERS effect, as well as the relevant experimental techniques included in this report. In Sections 3–6 we present the most common anisotropic NP shapes, namely nanorods, nanotriangles, nanocubes and nanostars. Section 7 is dedicated to more complex structures that can offer additional advantages. After a short introduction of Javier Reguera Javier Reguera is an Ikerbasque research fellow at CIC Biomagune in San Sebastia ´n (Spain). He received a bachelor degree in Physics and a PhD in Materials Science (2008) at the University of Valladolid (Spain). He was a Fulbright fellow at Massachusetts Institute of Technology (USA) 2009–2011, and a postdoctoral scientist at E ´cole Polytechnique Fe ´de ´rale de Lausanne (Switzerland) 2011–2014. He joined CIC BiomaGUNE (Spain) in 2014 where his research interests include the synthesis of patchy and multicomponent nanomaterials, self-assembly at interfaces and applications in nanobiomedicine, imaging and sensing. Judith Langer Judith Langer obtained a PhD from Freie Universita ¨t Berlin in 2003 for work on low-energy electron attachment on clusters in supersonic beams, under the supervision of Prof. Eugen Illenberger. After postdoctoral research at the Institute of Materials Science of Barcelona (ICMAB), the Institute of Physical Chemistry at Freie Universita ¨t Berlin and the Institute of Optic and Atomic Physics at Technische Universita ¨t Berlin, working on different materials and infrared (multiphoton) dissociation, she joined the Bionanoplasmonics Lab at CIC biomaGUNE headed by Prof. Luis Liz-Marza ´n in 2012. Her current research focuses on the application of plasmonic nanoparticles for bioimaging and sensing by SERS. Dorleta Jime ´nez de Aberasturi Dorleta Jimenez de Aberasturi graduated from the University of the Basque Country (UPV/EHU) in 2006. In 2013, she received her PhD entitled ‘‘Design of multifunctional colloidal particles with ion selective ligands’’ in cotutelle between the UPV/EHU and Philipps University of Marburg under the joint supervision of Prof. Dr Teo ´filo Rojo, Dr Idoia Ruiz de Larramendi and Prof. Dr Wolfgang Parak, obtaining the special doctoral award. Since 2014 she has worked in the group of Prof. Liz-Marza´nat CICbiomaGUNE. Her current research involves plasmonic nanoparticle synthesis, assembly and their functionalization for biosensing and imaging. Luis M. Liz-Marza ´n Luis Liz-Marza ´nisIkerbasque Professor and Scientific Director of CIC biomaGUNE, in San Sebastia ´n (Spain), since September 2012. He graduated in chemistry from the University of Santiago de Compostela, was postdoc at Utrecht University and Professor at the University of Vigo (1995– 2012). He has been Invited Professor at several universities and research centers worldwide and received numerous research awards. His major research activity is devoted to understand the growth mechanisms of metal nanocrystals, to tailor their surface chemistry and direct self-assembly. He also works on the design of biomedical applications based on the plasmonic properties of well-defined metal nanoparticles and nanostructures. Tutorial Review and a dielectric and are laterally confined due to the nanoscale dimensions of the NPs. They are thus termed localized surface plasmon resonances (LSPRs). The LSPR wavelength depends on the dielectric function of both the metal eand the surrounding medium e m , the NP morphology and the number of delocalized electrons. As the excitation wavelength is typically much larger than the NP dimension (lcr), LSPR modes possess dominantly a dipolar character. In the case of a sphere (isotropic), a dipole is induced along the electric field vector of the light wave. The corresponding moment m ind is determined by the incoming EM field E 0 (o 0 ) and the polarizability a(the degree on how easily the electron cloud can be displaced) of the NP (m ind =aE 0 (o 0 )). The situation in anisotropic NPs becomes more complex. As ais a tensor, its components can differ along the three main axes and several dipole modes can exist depending on the NP symmetry and orientation. This is exemplarily shown for nanorods (Fig. 1A), which can accommodate two different dipolar LSPRs, namely the transverse mode (electron oscillation perpendicular to the NR main axis) and the longitudinal mode (electron oscillation Fig. 1 (A) Orientation-dependence of dipolar LSPRs in NRs, longitudinal (top) and transverse (bottom) mode. 4 (B) UV-Vis spectrum of AuNRs showing the position and intensity of transverse and longitudinal LSPR modes. 4 (C) Illustration of the quadrupolar (center) and hexapolar (right) character of LSPRs in long NRs compared to the dipole (left), and corresponding EM field distributions. With increasing pole order high-field areas along the rod axis appear. 5 (D) EM field distributions and intensities for nanorods and nanostars, compared to spherical NPs, upon excitation at two different wavelengths indicating that the highest near-field is generated at the LSPR wavelength and focused at the sharp tips of anisotropic NPs (here at 785 nm). Note that the LSPR wavelength for an Au sphere is around 520 nm. 6 Reproduced with permission of: ref. 4 CC-BY, 2016 Intech; ref. 5. CC-BY 2012 SpringerOpen; ref. 6 Copyright r2012 IOP Publishing. each shape class, we briefly describe the optical properties, and discuss some recent works where the morphology-dependent SERS performance can be appreciated. 2. Brief theoretical aspects 2.1. Plasmonics The unusual optical properties of metal NPs are related to the presence of strong plasmon resonances in the visible and near-IR (NIR) ranges of the electromagnetic spectrum. Plasmon resonances exist in all metals and refer to the excitation of coherent, collective oscillations of delocalized electrons in the conduction band by an external electromagnetic (EM) field as the driving force. Similar to the illustrative picture of a driven oscillator in classical mechanics, here the frequency of the incident light wave is in resonance with the characteristic frequency of the free electrons in the metal (plasma frequency) driving the electron movement with the amplitude of the light field. In metal NPs, plasmon resonances exist at the interface between the metal how strongly the LSPR and near-field are damped, e.g. due to intraband or interband transitions, electron–phonon collisions and electron scattering with the surface. Ag exhibits the lowest losses in the visible and NIR frequency ranges, whereas Au shows significant plasmon damping at wavelengths below 600 nm. These dielectric properties render Ag the material of choice for applications in the visible (350–650 nm). In practice however, the use of Ag NPs is limited due to higher reactivity and susceptibility to oxidation. On the other hand, Au is an eminently inert material (though also more expensive), so that the choice of material ultimately depends on the framework requirements for each specific application. 2.2. Surface enhanced Raman scattering In Raman scattering, molecules interact with an incident EM field E 0 (o 0 ), e.g. from an intense laser beam. Similar to plasmons in NPs, depending on the polarizability of the molecule, an oscillating dipole is induced exciting the molecule into a virtual state. In the classical picture, scattering can be described by the modulation of incident field through a molecule vibrating with an eigenfrequency o vib by irradiating an EM field with shifted frequency E 0 (o R ), as schematically shown in Fig. 2 (top). Raman scattering can be dramatically improved when the molecule of interest is close to a metallic surface or preferentially in contact with the same. As the EM field strength attenuates as 1/r 2 ,a typical value rfor the distance between molecule and NP surface ranges between 0–10 nm. The presence of the plasmonic nearfield at the NP surface upon LSPR excitation E loc (o 0 ) increases a ofthemolecule(ormorecorrectlyofthemolecule-NPensemble) by 1 to 3 orders of magnitude compared to free molecules, due to mutual excitation between induced dipole in the molecule and induced dipole in the NP. In other words, NP and molecule scatter the impinging field whereupon the NP irradiates a locally enhanced EM field, which can be re-scattered by the molecule leading to enhanced Raman scattering. In turn, the molecule irradiates the Raman-scattered field, which is then re-scattered by the NP upon LSPR excitation and re-emission of an enhanced near-field E loc (o R ) (with shifted frequency). This situation is pictured in Fig. 2 (bottom). The SERS intensity I SERS thus depends on both incident (incoming) E 0 (o 0 ) and scattered (outgoing) fields E loc (o R ), and SERS enhancement is achieved when incoming and outgoing fields are in resonance with the LSPR of the NP. As molecular vibration frequencies o vib are small Fig. 2 Basic SERS electromagnetic mechanism. The AuNP (1) enhances both (a) the incident laser field and (b) the scattered field, greatly boosting the Raman scattered signal from the nearby molecule (2). Reproduced with permission of ref. 7. Copyright r2012 IOP Publishing. This effect is exemplarily shown by experimentally measured Raman and SERS spectra of rhodamine 6G (right-hand figure, measured onto assembled Au nanostars substrate). parallel to the main axis). The wavelength of the longitudinal mode is here significantly red-shifted with respect to that of the transverse mode, due to its higher a (Fig. 1B). Upon resonant excitation of strong LSPRs, a greatly enhanced local EM field (near-field) E loc (o 0 ) relative to the incident excitation field E 0 (o 0 ) is generated at the NP surface. This near-field in turn couples to the incident field and re-reemits radiation with the same wavelength (resonant scattering). As the scattering cross section scales with a 2 also the near-field strongly increases for anisotropic NPs due to higher a, as compared to spheres (Fig. 1D). Elongated NPs or those with edges and corners, such as nanorods, nanocubes, and nanotriangles, thus generate strong near-fields concentrated at vertices. Trapping the oscillating metal charges in highly confined spaces, e.g. at the sharp tips of nanostars, can lead to even stronger near-field enhancements. These areas of locally high near-field are known as intrinsic hotspots. A different type, extrinsic hotspots, are generated when NPs are positioned at very short distances (1–10 nm) or on top a solid support. The coupling of plasmon modes between individual NPs or between a NP and a flat metal surface leads to new modes with highly localized, intense EM fields at the narrow gaps within the junctions. Fig. 1D also shows that, when the NP is irradiated under off-resonance conditions, the resulting near-field is significantly lower and less confined than under resonance conditions. Resonant scattering becomes generally more dominant over absorption as the size of the NP increases. Note that absorption as well as resonant scattering by NPs, are both strongly increased at the LSPR wavelength. Therefore, the measurement of extinction (the sum of absorption and scattering) in the far-field fairly mirrors the LSPR excitation and the resulting near-field spectrum. In large NPs or those with high AR, LSPRs with multipolar character can also be excited, such as longitudinal quadrupolar and hexapolar modes in long NRs. These modes differ in near-field strength and distribution. Apart from the two hotspots at the tips, additional high-field nodes form along the main axis, as shown in Fig. 1C. The efficiency of LSPR excitation (and hence the near-field intensity) is largely influenced by the dielectric constant of the metal e(o), defined as a function of the angular frequency of the applied field o: e(o) = e0(o)+ie00(o). The real part e0(o) reflects how strongly the material is polarized upon exposure to an external EM field, while the imaginary part e00(o) represents I RS must be weighed by the numbers of molecules involved in RS (N RS ) and SERS (N SERS ). Estimation of EF then results in: EF = I SERS N RS /I RS N SERS . Measuring EFs from colloidal solutions leads to the determination of so-called analytical EF. Indeed, they are typically much smaller compared to those on dry surfaces but instead we record a statistically relevant number of molecules. Thus, the obtained results are more reproducible than for single- or few-molecule events whose values can strongly scatter from event to event. For a detailed understanding of the theory, practical advantages and drawbacks of alternative approaches, the reader is directed to the comprehensive review by Le Ru and Etchegoin. 8 In order to facilitate the understanding of the characterization methods used for selected examples of plasmonic nanoparticles and adsorbed molecules, Table 1 summarizes the most common experimental and theoretical methods that are relevant within this tutorial. 3. Nanorods Nanorods (NRs) and nanowires (NWs) are nanostructures where one dimension is longer than the other two (ref. 9 and references therein). The term nanorods is commonly referred to relatively short elongated particles (length o100 nm), while nanowires are longer, not always perfectly straight, typically with AR above 10 and length up to 5 mm. NWs are known since around 1960, when the vapor–liquid–solid approach was developed. During the 1990’s electrochemical and photochemical methods were used for the synthesis of Au and Ag NRs and NWs, but it was not until the early 21st century that the so called seed-mediated growth was developed, which has become the most popular approach. 9 This method comprises the synthesis of small metal NPs, ‘‘seeds’’, which are then overgrown into larger particles with different shapes by reduction of a metal salt in the presence of a ‘‘shape-directing’’ surfactant. Thus, nucleation and growth are temporally separated, leading to uniformity and precise control Table 1 Experimental and theoretical methods for the characterization of NPs and molecules relevant to this tutorial Experimental methods Characterization Ultraviolet-visible (UV-Vis) spectroscopy Extinction, LSPR and molecular resonance wavelength (or frequency, energy) Dark-field (DF) optical microspectroscopy Scattering, LSPR wavelength (or frequency, energy) Scanning electron microscopy (SEM) Morphology of NPs Transmission electron microscopy (TEM) Morphology and crystallinity of NPs Scanning TEM (STEM) Morphology, crystallinity and orientation of NPs Electron energy loss spectroscopy (EELS) LSPR mapping: near-field (or local electromagnetic field) intensity and spatial distribution Energy dispersive X ray (EDX) absorption Elemental analysis: distribution and chemical composition Annular dark field (ADF) detection Scattered electron mapping High-angle ADF (HAADF) Incoherently scattered electrons mapping (z-contrast) Cathodoluminescence (CL) LSPR mapping Surface enhanced Raman scattering (SERS) Chemical mapping: structure, orientation, interaction of molecules with metal surface, relative near-field (or local electromagnetic field) intensity and distribution (not relevant in this review) Theoretical simulations Discrete dipole approximation (DDA) LSPR modes, near-field (or local electromagnetic field) intensity and distribution, EELS, CL, SERSFinite-difference time domain (FDTD) Boundary element method (BEM) Surface-integral equations discretized by the method of moments (SIE-MoM) compared to the excitation frequency o 0 , the shift (o 0  o vib = o R E o 0 ) of the outgoing field is also small and the enhancement factor (EF) can be approximated to the fourth power of the local EM field enhancement at NP surface EF = I SERS /I Raman = (E loc (o 0 )) 2 (E loc (o R )) 2 /(E 0 (o 0 )) 2 (E 0 (o R )) 2 E (E loc (o 0 )/E 0 (o 0 )) 4 . This effect can be illustrated by a simple numerical example. Even a moderate enhancement of plasmon-supported EM field with a factor 10 would increase the intensity by 10 4 , but with a field enhancement of 100 the intensity would rise by 10 8 . In the context of practical applications, this conclusion has an important impact on how NPs should be tailored to maximize the EM field enhancement and hence the SERS output. Here, the NP morphology occupies a key position. The above described mechanism of SERS is known as electromagnetic mechanism. Note that it requires the presence of a molecule but is not dependent on its nature. If the molecule binds to the NP surface, the electronic interaction between molecular orbital and NP conduction band can give rise to new charge-transfer resonances, which can couple to vibrational states of the (adsorbed) molecule. This can give rise to a significant reallocation of electron density within the molecule and a change of a during the vibrational motion leading to additional SERS enhancement with EFs of 10 1 –10 3 . This effect, known as chemical mechanism, is generally much smaller than the electromagnetic effect but can improve the SERS sensitivity when charge-transfer resonances, LSPR and incident excitation wavelength are matched. If electronic resonances of the molecule or rather the adsorbate-NP complex are involved, the Raman cross section and hence the signal intensity further increase due to surface enhanced resonant Raman scattering (SERRS). To evaluate the performance of different NP morphologies and structures, their SERS response must be compared by using an adequate reference. To date, the most established method to evaluate the SERS efficiency is the determination of EF. The EF is defined as the ratio between the intensities of SERS (I SERS )and ordinary Raman (I RS ) for a given molecular mode, both measured under the same experimental conditions. The intensities I SERS and Both calculations indicate high EM field enhancement at the NR tips (hotspots) under longitudinal excitation, but moderate field enhancement on the NR lateral sides for transverse excitation. Similar results were obtained from EELS experiments (Fig. 4A(III)) and also for AgNRs. Fig. 4B(III) shows the localized field enhancement at NR tips for the dipolar longitudinal mode (top image). Furthermore, at higher energies new multipolar modes appear (lower images), with enhancement at intermediate nodes. Although these multipolar modes can in principle be used for SERS, the highest near-field intensity produced at the longitudinal dipolar mode with intense hotspots at the tips, still render this mode the preferred choice. NRs are undoubtedly the most widely used anisotropic nanoparticles as SERS substrates. Tailoring the LSPR relative to the laser excitation wavelength, to produce an on-resonance measurement, is one of the most important keys to achieve a high Raman enhancement. This has been confirmed by a sizedependent SERS study on AgNRs and AuNRs with AR ranging from 1.7 to 16 and under 633 nm excitation. 17 The authors demonstrated three aspects: (1) AgNRs generate higher EFs than AuNRs at 633 nm. (2) The EF for AgNRs increases with increasing AR whereas for AuNRs the EF decreases. This is explained by the LSPR red-shift with increasing AR driving the resonance towards excitation wavelength in case of Ag and away from excitation wavelength in case of Au. (3) Although nanorods with LSPRs not matching the excitation wavelength (off-resonance) can show relatively high EFs (10 4 –10 6 ), 1–2 orders of magnitude greater enhancements are estimated for NRs on-resonance. El-Sayed’s group compared SERS on AuNRs vs. spherical AuNPs Fig. 3 Characterization of gold nanorods. (A) Representative UV-Vis-NIR spectra of AuNRs with longitudinal LSPR bands at 650 nm, 760 nm, 840 nm, 900 nm, 1000 nm, 1036 nm, 1060 nm and 1100 nm (from left to right). (B) Relationship between longitudinal LSPR and AR (measured from TEM images). (C–F) Representative TEM images of AuNRs with longitudinal LSPR at (C) 650 nm, (D) 760 nm, (E) 920 nm, and (F) 1000 nm. Scale bars: 100 nm. Reproduced with permission of ref. 14. CC-BY, 2015 Ivyspring. over particle size and aspect ratio. AgNRs and AgNWs were successfully prepared in 2001 by Murphy et al., by reducing silver nitrate (AgNO 3 ) with ascorbic acid in the presence of Ag seeds. 10 The use of cetyltrimethylammonium bromide (CTAB) as surfactant and different amounts of NaOH promoted the formation of NRs at high NaOH concentrations, or NWs at low concentrations. In the case of Au, high-yield (490%) colloidal synthesis of monodispersed single crystal AuNRs was introduced by Nikoobakth and El-Sayed in 2003. 11 Since then, several approaches have been reported, further improving quality and yield, reaching values higher than 99% shape yield. 12 The addition of low amounts of Ag + ions was essential to obtain high NR yield in this approach, whereas Ag-free NR growth leads to mixtures of pentatwinned NRs, spheres and other shapes. Very recently, yields above 85% have also been obtained for pentatwinned NRs, by optimization of pentatwinned seeds. 13 LSPR wavelengths in NRs can be readily tuned, by simply varying the aspect ratio (Fig. 3). Typical UV-Vis spectra feature two plasmon bands, corresponding to transverse and longitudinal modes. Whereas the intensity of the transverse LSPR band is relatively low and nearly independent of the AR, the longitudinal LSPR is much more intense and strongly dependent on AR (Fig. 3B). 14 Interestingly, the plasmon modes in NRs give rise to regions with intense EM fields, in the proximity of the NP. This is illustrated in Fig. 4, which shows theoretical (DDA) and experimental (EELS) near-field mapsforpureAuNRsandAgNRs. 15,16 DDA simulations for AuNRs (Fig. 4A) have shown the equivalence between plasmon excitation by light (extinction cross section in Fig. 4A(I)) and by electrons (loss probability in Fig. 4A(II)). under off-resonance conditions. 18 Interestingly, the measured EFs (10 4 –10 5 ) were two orders of magnitude higher than expected from the electromagnetic mechanism, indicating significant chemical enhancement. The SERS activity was higher for AuNRs than for AuNPs, which was explained by the presence of more reactive facets in AuNRs. The same group recently reported a comparative study between AgNRs and AuNRs with comparable ARs and longitudinal LSPR peak position (Fig. 5). 19 The results confirmed that AgNRs are more efficient than AuNRs: EF values of 0.1 to 2.0  10 14 for AgNRs vs. 0.3 to 2.2  10 12 for AuNRs were reported. This was attributed to the higher plasmon field intensity at the tips of AgNRs as observed in DDA simulations (Fig. 5B–E), and to higher Rayleigh scattering for AuNRs. The extraordinarily high values of EF differ from other works on NRs, which highlights the difficulties to compare results obtained in different labs. It should be noted that, apart from LSPR position and composition, other factors can also affect SERS performance. A recent comparative study showed that AuNRs with different sizes but equal LSPR wavelength provide different SERS under the same conditions (see Fig. 6). 20 It was found that larger NRs show stronger scattering intensity but weaker SERS. The larger SERS signal at smaller sizes was attributed to a synergistic effect between the stronger lightning rod effect (charge concentration at the tips), and to the weaker radiation damping resulting in stronger local field enhancement. The same effect was observed when extrinsic hotspots were generated by formation of NR dimers with equal LSPR but different sizes. It should be noted Fig. 4 (A) AuNRs: comparison of excitation with light vs. electrons, of the NR longitudinal and transverse LSPR modes. (A.I) Calculated extinction cross section for an individual NR (diameter/length, 27/85 nm), using parallel (red curve) or perpendicular (blue curve) incident light-electric field to the rod axis, leading to excitation of bright LSPR modes. Inset: Calculated near-field maps for both modes. The positive and negative signs denote the induced charge oscillation patterns of the modes. (A.II) Calculated electron energy loss probability for locations I (red curve) and II (blue curve) of the electron beam relative to the NR under grazing incidence. The loss probability is given per incoming electron and per electronvolt for a given lost energy. Inset: Calculated EELS excitation-intensity maps. (A.III) Experimental counterpart of (A.II). Inset: Measured EELS excitation-intensity maps for modes a and b. Color scale bar represents the linearly normalized image intensity. 15 (B) AgNRs: (B.I) electron microscopy (ADF) image of a high AR AgNR: length = 192 nm, diameter = 20 nm, AR = 9.6. (B.II) Summed electron energy loss spectrum (B.III) starting from the left column, multivariate statistical analysis (MVSA) score images and loading spectra interpreted as plasmon maps and energies, respectively. Right, DDA calculated electric field plots displaying the field generated by a plane wave optical excitation at the energies and polarizations given on each panel. 16 Reproduced with permission of ref. 15 and 16 Copyright r2009, 2011 American Chemical Society. that the employed NRs are larger and rougher than those usually obtained by wet chemistry methods. When comparing the SERS efficiency of different NRs, special care should be taken to maintain similar experimental conditions. For example, NR concentration in the colloid is an important parameter when dense suspensions are used. 21 This could be of special importance in biological applications where a high variability of NP concentration exists (cell cytoplasm, endosomes, etc.). Murphy and co-workers showed, using concentrated AuNR solutions of different AR, that the effectively measured SERS intensity can be attenuated by extinction of both the incident and scattered light during propagation (Fig. 7). 21 The maximum SERS signal (taking the number of Raman reporter molecules per Au NR into account) was obtained from NRs with plasmon resonances that are far blue-shifted from the excitation wavelength at 785 nm, since the intensity of on-resonance SERS was quenched, and in striking contrast to the electromagnetic enhancement mechanism. On the basisofthisresult,theauthors developed a model for extinction-corrected EFs which explained the maximum EF at AR = 2.75 instead of 4 as expected from the Fig. 5 (A) SERS of 4-nitrothiophenol adsorbed on the surface of AuNRs and AgNRs with similar ARs (AgNRs: 3.45 0.35 and 4.09 0.44, AuNRs: 3.20 0.34 and 3.50 0.41) excited at 785 nm. Plasmon maps simulated by DDA for AuNRs and AgNRs of 22 nm diameter and different AR: (B) AuNRs, AR = 3, (C) AuNRs, AR = 3.5 (D) AgNRs, AR = 4, (E) AgNRs, AR = 3.5. Color code for the spectra in panels B–E: extinction (black), scattering (red), and absorption (white). Reproduced with permission of ref. 19. Copyright r2013 American Chemical Society. Fig. 6 (A) Single nanoparticle dark field scattering spectra of three AuNRs showing different scattering intensities. (B) Normalized DF spectra. (C) Single nanoparticle SERS spectra of malachite green isothiocyanate (MGITC) on the three AuNRs. (D) Background-subtracted SERS spectra. Both white light and laser polarizations were parallel to the nanorods. (E) Representative SEM images. (F) Simulated local electric field scattering cross section maps. Reproduced with permission ref. 20. Copyright r2016 American Chemical Society. Fig. 7 (A) Extinction spectra for AuNRs of different AR (pink: 2; purple: 2.5; blue: 2.75; green: 3; brown: 3.5; orange: 4.5), normalized to concentration. The Raman excitation wavelength is indicated by a red-dotted vertical line and the Raman spectrum in black. (B) SERS spectrum of methylene blue (MB) on the various AuNRs bearing a polyelectrolyte layer, normalized for AuNR and reporter molecule concentrations. (C) Comparison of Raman spectra acquired from the suspensions bearing polyelectrolyte layers plus MB reporter in methanol. The variation in Raman intensity for the methanol bands is illustrated by the peak at 1030 cm 1 , which varies as a function of AR. AuNR suspensions are normalized for concentration and the number of reporter molecules per AuNR. Reproduced with permission of ref. 21 Copyright r2013 American Chemical Society. Fig. 8 (A) EELS maps of an Ag NT at different energies, showing the in-plane dipolar and quadrupolar plasmon modes. 22 (B) BEM simulation of EELS for the same NT. 22 (C) Extinction spectra of Ag NTs at different sizes from 20 to 128 nm (DDA simulations). 24 (D) Extinction spectra of truncated Ag NTs at different snip sizes (DDA simulations). 1 Reproduced with permission: ref. 22 Copyright r2007 Macmillan Publishers Ltd; ref. 1 and 24 Copyright r2003, 2007 American Chemical Society. when illuminating at 785 nm (Fig. 17C and D). As expected, the EFs correlated with the proximity between the resonance frequency and the excitation laser. When the authors used nanospheres instead of NSs, EFs of only 10 4 to 10 5 were obtained. It is worth mentioning that, despite of NSs being deposited on a substrate, a further increase of SERS enhancement is not expected for dense AuNSs arrays, as recently demonstrated by Solı ´set al. using numerical simulations based on SIE-MoM. 45 They observed a relatively constant enhancement with increasing surface coverage due to the initially high number of intrinsic hotspots in AuNSs and damping in the case of valley to tip contacts. This trend was radically different for nanospheres and NRs, where an increase of coverage resulted in a dramatic increase of enhancement, close to three orders of magnitude. Despite this increase in EF, the order of enhancement was still NSs 4NRs 4nanospheres when illuminating at the maximum extinction wavelength. It should be noted however, that for NTs the opposite effect has been experimentally observed, with better performance in solution than for a deposited monolayer. 25 The effect of NS size on SERS enhancement has been addressed by Khoury et al., using an illumination wavelength far away from the LSPR maximum, and taking advantage of the breadth of AuNS plasmon bands. 47 They synthesized PVP-coated AuNSs with 6 different sizes (from 45 to 116 nm) by varying the seeds to HAuCl 4 ratio during the synthesis, resulting in increased number and length of tips, as well as an LSPR red-shift Fig. 17 (A) TEM image of AuNSs synthesized with three different sizes. (B) Absorbance spectra of the corresponding NSs of (A) tested in SERS detection. (C) Quantitative SERS analysis facilitated by the SERS substrates (three different NSs and two spherical NPs), with varying 4-MBA concentration and represented as the relative intensity at the 1077 cm 1 peak. Below 10 8 M no signal could be detected for the nanospheres (incident laser was 785 nm for NSs and 633 nm for nanospheres). (D) Characteristic SERS spectra of 4-MBA at different concentrations. Reproduced with permission of ref. 43 Copyright rRoyal Society of Chemistry 2014. the tips and the central body (Fig. 16C). 41 Such a plasmon mode hybridization produces an antenna effect resulting in EM field enhancement of up to 4 times the one that would be exclusively produced by the tips. In addition, due to the polydispersity in tip shape and orientation, plasmon bands from AuNS colloids are usually broader than those for NRs, thereby facilitating the use of Raman excitation lasers that do not precisely match the absorbance maximum, but are still on-resonance with a significant proportion of plasmon modes. Branched nanoparticles display particularly high EFs at the resonance wavelength, even higher than those for rod or sphere dimers, 44 thus enabling unprecedentedly low detection limits for single particles. Rodrı ´guez-Lorenzo et al. achieved zeptomole detection limits when AuNSs were used to measure 1,5-naphtalene- dithiol (1,5-NDT) that was located between a Au substrate and a AuNS tip covalently bonded to both surfaces. 42 In this case, an additional enhancement was obtained due to the plasmon coupling between the flat Au substrate and the NS tip, thereby creating a hotspot with even higher SERS enhancement. Indrasekara et al. used AuNSs on a deposited substrate (Fig. 17A) to detect 4-mercaptobenzoic acid (4-MBA) with limits of detection as low as 10 fM, corresponding to enhancement factors of 10 9 . 43 The authors also compared three different AuNS sizes with plasmon bands around 800, 700 and 600 nm (for large, intermediate and small NSs, respectively), and obtained enhancement factors of 4.9  10 9 , 1.6  10 9 , and 5.3  10 8 , respectively, signal was obtained for the RE-SERS measurement (Fig. 19G), which was attributed to lower cell damage (and therefore lower toxicity), as well as a lower amount of stacked molecules at the AgNW tip. Interestingly, not only nanoparticle anisotropy could be useful to solve specific problems, but a combination of anisotropic metal NPs with different materials can result in probes with superior features. Fales et al. showed that, upon partial coverage of a AuNS with Ag, the SERS signal could be increased up to one additional order of magnitude. 49 Ag coating occurred mainly on the AuNS core and the maximum SERS signal was achieved for the highest coating that still left the Au tips free. This enhancement in AuNSs@Ag was attributed to the remaining AuNS hotspots, while the NP plasmon was shifted towards an off-resonance situation, which decreased self-absorption. The authors further covered the NPs with a silica shell (AuNS@Ag@ SiO 2 ), thereby trapping the dye reporter at the NP surface and Fig. 18 Influence of gold nanoparticle shape on SERS enhancement. (A) Scanning electron microscopy images of NPs with different morphologies: aggregated nanospheres, NTs and NSs. (B) UV-Vis spectra from the different nanoparticles at constant NP concentration (N=610 8 particles per mL). (C) SERS spectra from 5 mM R6G, using nanoparticles of different shapes (3 10 9 particles per mL, l 0 = 785 nm). Reproduced with permission of ref. 46 Copyright rRoyal Society of Chemistry 2014. (from B700 to B850 nm). EFs around 5  10 3 were achieved for 4-MBA, but no big differences were observed among the different sizes, probably due to a similar chemical enhancement, and absorbance off-resonance with respect to the illumination. Note that EF values obtained from different works are not fully comparable (see above), since parameters such as concentration, NP coatings, or SERS substrate preparation play an important role in the final values. One of the few examples where NP shapes were compared in solution involved AuNSs, NTs and nanospheres of similar sizes, as well as small nanosphere aggregates (Fig. 18). 46 They found that SERS of R6G was negligible for nanospheres with a resonance peak at 532 nm, when compared to other shapes. Upon aggregation, the nanospheres displayed a red-shifted LSPR at 660 nm, while NTs and NSs peaked at 800 and 785 nm, respectively. Using R6G as the analyte and exciting at 785 nm, the resulting SERS enhancement sequence was: nanosphere aggregates o NTs o NSs, which was attributed to theincreaseofhotspotsandthemagnitude of LSPR absorbance at the excitation wavelength. However, differences in organic coating and zeta potential between the different particles, which may obviously affect the adsorption of the organic dye on the NP cores, cannot be discarded as a potential source of SERS variation. 7. More complex systems Although the morphologies discussed above (NRs, NTs, NCs, and NSs) are by far the most frequently used anisotropic NP shapes for SERS sensing, many other shapes containing intrinsic hotspots have been or could potentially be used. These include, e.g. regular or highly symmetric shapes such as octahedra, decahedra and bipyramids, as well as a myriad of other irregular shapes containing intrinsic hotspots at sharp edges or tips. Additionally, the anisotropic morphology of NPs not containing intrinsic hotspots could still be advantageously used in certain cases, whereas extrinsic hotspots can be generated by means of directed assembly strategies. This is the case for nanoshells, where the core–shell configuration (dielectric core@metallic shell) allows LSPR tuning into the red or NIR regions, where illumination is most appropriate for biological applications. Hotspots have been created with these NPs via formation of monolayers, resulting in NP–NP and NP–substrate hotspots. Anisotropy in NPs can also be used to solve the specific requirements for detection in a complex system. Uji-i and co-workers devised a strategy by which an AgNW could be used for so-called SERS endoscopy, penetrating inside a living cell and probing the chemical composition at different locations by SERS. 48 The AgNW (50 nm in diameter) was attached to a tungsten tip used for micromanipulation and two different NW excitation methods were used: directly at the NW tip and far away from the tip (remote excitation SERS or RE-SERS), making use of the propagation of surface plasmon polaritons along the AgNW (Fig. 19A). These authors recorded a higher signal inside the cell nucleus, due to the presence of DNA and proteins, as compared to the cytoplasm (a more crowded environment) or the buffer at the cell exterior (Fig. 19B). Interestingly, a much clearer making the signal less sensitive to other molecules, which was proposed as a suitable probe for SERS bioimaging. Another example of composite NPs was demonstrated by Reguera et al., who prepared Janus magnetic nanostars (JMNSs) comprising an asymmetric AuNS attached to an epitaxially attached iron oxide nanosphere. Such multifunctional NPs could be used as SERS substrates, allowing magnetic concentration of NP and analyte, thereby increasing the Raman signal. 50 Fig. 20A shows an electron tomography image of one such nanostructure, where the Janus configuration of AuNS and iron oxide NP can be clearly appreciated, allowing selective functionalization of both components. These NPs were used to measure crystal violet (CV) in solution, at a sub-micromolar concentration (Fig. 20C in red). Interestingly, a 5-fold increased enhancement was recorded when a hand-held magnet was used to concentrate the NPs after incubation with the analyte (Fig. 20C, blue lines). By using this strategy, it was possible to detect CV concentrations down to 15 nM, well below the limit of detection in solution. Therefore, the NPs could be used to improve the ‘‘capture’’ of analyte in solution (small diffusion distances) prior to magnetic concentration, producing a local accumulation of the analyte at the spot where the Raman excitation laser was focused. Although more complex systems are being developed to solve particular problems, or with higher sensitivity, there is also a need to develop quantitative and reliable SERS. For the case of anisotropic nanoparticles this includes the intra- and inter-lab reproducibility of samples, the stability of the substrates and a reliable calibration. Although this analysis is out of the scope of the current review, many articles have tackled this topic in recent years for different analytes and systems, contributing to the advancement in the use of the technique. Fig. 19 SERS endoscopy of a living cell using an AgNW. (A) Schematic illustrations of plasmonic waveguide SERS endoscopy using an AgNW in direct and remote excitation modes, (top and bottom cartoons, respectively). (B) Scheme of the positions chosen for direct excitation SERS spectroscopy insidea live HeLa cell during AgNW probe endoscopy, and corresponding SERS spectra (the HBSS buffer, the cell nucleus, and the cell cytoplasm). (C) After direct excitation SERS endoscopy measurement, cell components adhere to the AgNW probe and come out with the probe during release (indicated by the arrow). (D–F) Images of an AgNW probe using RE-SERS endoscopy of a live HeLa cell, by optical transmission (D), combination of optical transmission and RE-SERS (E) and RE-SERS only (F). (G) RE-SERS spectrum from the nucleus of a live HeLa cell. Reproduced with permission of ref. 48 Copyright r Wiley-VCH Verlag GmbH & Co. 2014. 8. Conclusions NP synthesis is blooming into plenty of new anisotropic morphologies of mono- and multi-component materials. Better protocols have been designed that produce NPs with high uniformity and yield. Modern Raman instruments provide higher sensitivities and versatility, lower cost, higher portability, etc. All these rapid advances offer a promising future for SERS detection in fields where analytes of diverse nature are to be detected. In this situation of high interest on SERS the availability of anisotropic morphologies is a clear advantage toward obtaining ultrasensitive detection and to the design of ad hoc configurations that allow the detection of the selected analytes in complex environments such as those in biological or environmental disciplines. In this tutorial review, we highlighted several comparative works that help elucidating the key morphological parameters to achieve high EFs for SERS. The highest impact on SERS performance comes from the effective near-field enhancement which is in general achieved when the plasmon oscillation is minimally damped, the LSPR wavelength matches the excitation wavelength (on-resonance condition) and the NP exhibits preferentially a large number of protrusions with high curvature giving rise to highest polarizability. The importance of LSPR efficiency regarding plasmon damping becomes evident when comparing Au and Ag NPs. An additional chemical effect favors the application of anisotropic NPs because of the presence of highly reactive facets with higher binding affinities for analytes as compared to nanospheres. However, an overall evaluation of the NPs as SERS substrates, e.g. via EF is not simple due to the difficulties of controlling certain parameters such as the type and quality of organic coatings or limitations regarding the available laser wavelengths. The evaluation is even more complicated when EFs are compared among different laboratories, where the lack of standardized protocols makes it virtually impossible. Many factors influence the SERS output, the establishment of such procedures and standardization are absolutely desirable, and within the rising commercial interest of NPs for SERS it will be indispensable. Besides the confinement and strength of the local EM field and the material, the surface chemical sensitivity of SERS results in parameters such as nature of the analyte, the chemical affinity to the nanoparticle surface, concentration of NP and analyte (or the ratio of both), surface coverage, coating/protection layer etc. playing important roles. Further, the stability of the analyte-NP ensemble under given conditions should be taken into account. On one hand, the interplay among these factors offers plentiful possibilities but at same time a rational design of the NP for a specific application is often required and limits the generalization of the SERS substrate. Acknowledgements This work has been funded by the European Research Council (ERC Advanced Grant #267867, Plasmaquo). References 1 K. L. Kelly, E. Coronado and L. L. Zhao, J. Phys. Chem. B, 2003, 107, 668–677. 2 P. K. Jain, K. S. Lee, I. H. El-Sayed and M. A. El-Sayed, J. Phys. Chem. B, 2006, 110, 7238–7248. 3 M. Fleischmann, P. J. Hendra and A. J. McQuillan, Chem. Phys. Lett., 1974, 26, 163–166. 4 L. Sun, P. Chen and L. Lin, in Applications of Molecular Spectroscopy to Current Research in the Chemical and Biological Sciences, ed. M. T. Stauffer, InTech, 2016, pp. 383–404. 5 Z. Yong, D. Y. Lei, C. H. Lam and Y. Wang, Nanoscale Res. Lett., 2014, 9, 187. 6 M. Li, S. K. Cushing, J. Zhang, J. Lankford, Z. P. Aguilar, D. Ma and N. Wu, Nanotechnology, 2012, 23, 115501. 7 N. C. Lindquist, P. Nagpal, K. M. McPeak, D. J. Norris and S.-H. Oh, Rep. Prog. Phys., 2012, 75, 036501. 8 E. C. Le, Ru, E. Blackie, M. Meyer and P. G. Etchegoin, J. Phys. Chem. C, 2007, 111, 13794–13803. 9 H. Chen, L. Shao, Q. Li and J. Wang, Chem. Soc. Rev., 2013, 42, 2679–2724. 10 N. R. Jana, L. Gearheart and C. J. Murphy, Chem. Commun., 2001, 617–618. 11 B. Nikoobakht and M. A. El-Sayed, Chem. Mater., 2003, 15, 1957–1962. 12 X. Ye, C. Zheng, J. Chen, Y. Gao and C. B. Murray, Nano Lett., 2013, 13, 765–771. 13 A. Sa ´nchez-Iglesias, N. Winckelmans, T. Altantzis, S. Bals, M. Grzelczak and L. M. Liz-Marza ´n, J. Am. Chem. Soc., 2017, 139, 107–110. Fig. 20 SERS detection using JMNSs. (A) Schematic representation of the nanoparticle morphology (on the left) and electron tomography image (HAADF-STEM + EDX) of the Janus hetero-NP. (B) Diagram showing the SERS measurement in solution and after magnetic separation of NPs with ‘‘trapped’’ analyte molecules. (C) SERS spectra of CV-containing JMNSs in solution (red) and after magnetic concentration (blue) for two different dye concentrations [CV] = 450 nM (upper spectra) and [CV] = 15 nM (lower spectra). Concentration of JMNSs was [Au 0 ] = 0.2 and 0.1 mM, respectively. The spectra were shifted vertically for clarity, the scale bar corresponds to an intensity of 200 counts. Reproduced with permission of ref. 50 Copyright rRoyal Society of Chemistry 2014. P. G. Kotula, M. Varela, G. C. Schatz, S. J. Pennycook and J. P. Camden, Nano Lett., 2011, 11, 3482–3488. 17 C. J. Orendorff, L. Gearheart, N. R. Jana and C. J. Murphy, Phys. Chem. Chem. Phys., 2006, 8, 165–170. 18 B. Nikoobakht, J. Wang and M. A. El-Sayed, Chem. Phys. Lett., 2002, 366, 17–23. 19 M. A. Mahmoud and M. A. El-Sayed, J. Phys. Chem. Lett., 2013, 4, 1541–1545. 20 K.-Q. Lin, J. Yi, S. Hu, B.-J. Liu, J.-Y. Liu, X. Wang and B. Ren, J. Phys. Chem. C, 2016, 120, 20806–20813. 21 S. T. Sivapalan, B. M. DeVetter, T. K. Yang, T. van Dijk, M. V. Schulmerich, P. S. Carney, R. Bhargava and C. J. Murphy, ACS Nano, 2013, 7, 2099–2105. 22 J. Nelayah, M. Kociak, O. Ste ´phan, F. J. Garcı ´a de Abajo, M. Tence ´, L. Henrard, D. Taverna, I. Pastoriza-Santos, L. M. Liz-Marza ´n and C. Colliex, Nat. Phys., 2007, 3, 348–353. 23 A. Losquin, L. F. Zagonel, V. Myroshnychenko, B. Rodrı ´guez- Gonza ´lez, M. Tence ´,L.Scarabelli,J.Fo ¨rstner,L.M.Liz-Marza ´n, F. J. Garcı ´ adeAbajo,O.Ste ´phan and M. Kociak, Nano Lett., 2015, 15, 1229–1237. 24 A. Brioude and M. P. Pileni, J. Phys. Chem. B, 2005, 109, 23371–23377. 25 L.Scarabelli,M.Coronado-Puchau,J.J.Giner-Casares,J.Langer andL.M.Liz-Marza ´n, ACS Nano, 2014, 8, 5833–5842. 26 Y. Yang, X. L. Zhong, Q. Zhang, L. G. Blackstad, Z. W. Fu, Z.-Y. Li and D. Qin, Small, 2014, 10, 1430–1437. 27 T. Tan, C. Tian, Z. Ren, J. Yang, Y. Chen, L. Sun, Z. Li, A. Wu, J. Yin and H. Fu, Phys.Chem.Chem.Phys., 2013, 15, 21034–21042. 28 V. S. Tiwari, T. Oleg, G. K. Darbha, W. Hardy, J. P. Singh and P. C. Ray, Chem. Phys. Lett., 2007, 446, 77–82. 29 S. E. Skrabalak, L. Au, X. Li and Y. Xia, Nat. Protoc., 2007, 2, 2182–2190. 30 M. N. O’Brien, M. R. Jones, K. A. Brown and C. A. Mirkin, J. Am. Chem. Soc., 2014, 136, 7603–7606. 31 M. B. Cortie, F. Liu, M. D. Arnold and Y. Niidome, Langmuir, 2012, 28, 9103–9112. 32 B. J. Wiley, S. H. Im, Z.-Y. Li, J. McLellan, A. Siekkinen and Y. Xia, J. Phys. Chem. B, 2006, 110, 15666–15675. 33 O. Nicoletti, F. de la Pen ˜a, R. K. Leary, D. J. Holland, C. Ducati and P. A. Midgley, Nature, 2013, 502, 80–84. 34 J. M. McLellan, Z.-Y. Li, A. R. Siekkinen and Y. Xia, Nano Lett., 2007, 7, 1013–1017. 35 H.-L. Wu, H.-R. Tsai, Y.-T. Hung, K.-U. Lao, C.-W. Liao, P.-J. Chung, J.-S. Huang, I.-C. Chen and M. H. Huang, Inorg. Chem., 2011, 50, 8106–8111. 36 J. M. Romo-Herrera, A. L. Gonza ´lez, L. Guerrini, F. R. Castiello, G. Alonso-Nun ˜ez,O.E.ContrerasandR.A.Alvarez-Puebla, Nanoscale, 2016, 8, 7326–7333. 37 M. Rycenga, K. K. Hou, C. M. Cobley, A. G. Schwartz, P. H. C. Camargo and Y. Xia, Phys. Chem. Chem. Phys., 2009, 11, 5903. 38 A. Guerrero-Martı ´nez, S. Barbosa, I. Pastoriza-Santos and L. M. Liz-Marza ´n, Curr. Opin. Colloid Interface Sci., 2011, 16, 118–127. 39 H. Yuan, C. G. Khoury, H. Hwang, C. M. Wilson, G. A. Grant and T. Vo-Dinh, Nanotechnology, 2012, 23, 075102. 40 P. S. Kumar, I. Pastoriza-Santos, B. Rodrı ´guez-Gonza ´lez, F. Javier Garcı ´a de Abajo and L. M. Liz-Marza ´n, Nanotechnology, 2007, 19, 015606. 41 F. Hao, C. L. Nehl, J. H. Hafner and P. Nordlander, Nano Lett., 2007, 7, 729–732. 42 L. Rodrı ´guez-Lorenzo, R. A. Alvarez-Puebla, I. Pastoriza- Santos,S.Mazzucco,O.Ste ´phan, M. Kociak, L. M. Liz-Marza ´n and F. J. Garcı ´a de Abajo, J. Am. Chem. Soc., 2009, 131, 4616–4618. 43 A. S. D. S. Indrasekara, S. Meyers, S. Shubeita, L. C. Feldman, T. Gustafsson and L. Fabris, Nanoscale, 2014, 6, 8891–8899. 44 R. Alvarez-Puebla, L. M. Liz-Marza ´n and F. J. Garcı ´ade Abajo, J. Phys. Chem. Lett., 2010, 1, 2428–2434. 45 D. M. Solı ´s, J. M. Taboada, F. Obelleiro, L. M. Liz-Marza ´n and F. J. Garcı ´a de Abajo, ACS Photonics, 2017, 4, 329–337. 46 F. Tian, F. Bonnier, A. Casey, A. E. Shanahan and H. J. Byrne, Anal. Methods, 2014, 6, 9116–9123. 47 C. G. Khoury and T. Vo-Dinh, J. Phys. Chem. C, 2008, 112, 18849–18859. 48 G. Lu, H. De Keersmaecker, L. Su, B. Kenens, S. Rocha, E. Fron, C.Chen,P.VanDorpe,H.Mizuno,J.Hofkens,J.A.Hutchison and H. Uji-i, Adv. Mater., 2014, 26, 5124–5128. 49 A. M. Fales, H. Yuan and T. Vo-Dinh, J. Phys. Chem. C, 2014, 118, 3708–3715. 50 J. Reguera, D. Jime ´nez de Aberasturi, N. Winckelmans, J. Langer, S. Bals and L. M. Liz-Marza ´n, Faraday Discuss., 2016, 191, 47–59. 14 S. Wang, W. Xi, F. Cai, X. Zhao, Z. Xu, J. Qian and S. He, Theranostics, 2015, 5, 251–266. 15 M.-W. Chu, V. Myroshnychenko, C. H. Chen, J.-P. Deng, C.-Y. Mou and F. J. Garcı ´a de Abajo, Nano Lett., 2009, 9, 399–404. 16 B. S. Guiton, V. Iberi, S. Li, D. N. Leonard, C. M. Parish,