Radiative decay engineering 5: metal-enhanced fluorescence and plasmon emission Joseph R. Lakowicz* Center for Fluorescence Spectroscopy, Department of Biochemistry and Molecular Biology, University of Maryland at Baltimore, 725 West Lombard Street, Baltimore, MD 21201, USA Received 15 April 2004 Available online 13 December 2004 Abstract Metallic particles and surfaces display diverse and complex optical properties. Examples include the intense colors of noble metal colloids, surface plasmon resonance absorption by thin metal films, and quenching of excited fluorophores near the metal surfaces. Recently, the interactions of fluorophores with metallic particles and surfaces (metals) have been used to obtain increased fluores- cence intensities, to develop assays based on fluorescence quenching by gold colloids, and to obtain directional radiation from flu- orophores near thin metal films. For metal-enhanced fluorescence it is di���cult to predict whether a particular metal structure, such as a colloid, fractal, or continuous surface, will quench or enhance fluorescence. In the present report we suggest how the effects of metals on fluorescence can be explained using a simple concept, based on radiating plasmons (RPs). The underlying physics may be complex but the concept is simple to understand. According to the RP model, the emission or quenching of a fluorophore near the metal can be predicted from the optical properties of the metal structures as calculated from electrodynamics, Mie theory, and/or Maxwell��s equations. For example, according to Mie theory and the size and shape of the particle, the extinction of metal colloids can be due to either absorption or scattering. Incident energy is dissipated by absorption. Far-field radiation is created by scattering. Based on our model small colloids are expected to quench fluorescence because absorption is dominant over scattering. Larger col- loids are expected to enhance fluorescence because the scattering component is dominant over absorption. The ability of a metal��s surface to absorb or reflect light is due to wavenumber matching requirements at the metal���sample interface. Wavenumber matching considerations can also be used to predict whether fluorophores at a given distance from a continuous planar surface will be emitted or quenched. These considerations suggest that the so called ������lossy surface waves������ which quench fluorescence are due to induced electron oscillations which cannot radiate to the far-field because wavevector matching is not possible. We suggest that the energy from the fluorophores thought to be lost by lossy surface waves can be recovered as emission by adjustment of the sample to allow wavevector matching. The RP model provides a rational approach for designing fluorophore���metal configurations with the desired emissive properties and a basis for nanophotonic fluorophore technology. �� 2004 Elsevier Inc. All rights reserved. The effects of metallic surfaces on fluorescence have a long scientific history perhaps starting with the classic reports of Drexhage [1,2]. These papers showed that a fluorophore placed within wavelength-scale distances from a reflecting metallic surface, in this case a thick sil- ver film (P100 nm) or mirror, resulted in oscillations of the emissive lifetime with distance from the metal sur- face. This effect could be explained by the reflected far-field radiation from the fluorophore back on itself, which depends on the distance from the metal surface. When the reflected field amplitude at the fluorophore was increased the lifetime decreased. When the reflected field opposed the fluorophore��s field the lifetime in- creased. Agreement of the data with this reflective model was adequate at most distances, except when the fluoro- phore was close to the metal. At distances below 20 nm 0003-2697/$ - see front matter �� 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ab.2004.11.026 * Corresponding author. Fax: +1 410 706 8408. E-mail address: lakowicz@cfs.umbi.umd.edu. ANALYTICAL BIOCHEMISTRY Analytical Biochemistry 337 (2005) 171���194 www.elsevier.com/locate/yabio

the lifetime dropped dramatically and the emission was strongly quenched. This quenching effect was attributed to lossy surface waves (LSWs),1 dissipated loses, ohmic loses, and similar terms, all of which implied the nonra- diative dissipation of energy within the metal. These studies [1,2] resulted in numerous theoretical studies of the interactions of an oscillating dipole with metallic surfaces [3���9] and particles [10���12] and a num- ber of experimental studies [13���16]. This citation list is far from complete. These papers have been summarized in several classic reviews which provide accurate summa- ries of the theory and experimental results [17���21] but are rather di���cult to read. In these reports the short-range quenching is attributed to lossy surface waves or some similar dissipative process. An extensive search of the lit- erature revealed no additional details about the mysteri- ous LSW quenching mechanism. For clarity we note that we are considering only the electromagnetic interactions of the fluorophore at a short distance above the metal surface. We are not considering chemical or other effects occurring upon direct fluorophore���metal contact. We use the term ������metal������ to describe any conducting metallic particle or surface and not the ionic species. The effects of metals on fluorophores led us to use these interactions for increased detectability of fluoro- phores. We studied the interactions of fluorophores with metallic particles [22] and surfaces [23]. We found that proximity of fluorophores within about 10 nm of silver island films (SIFs) resulted in increased emission intensi- ties and decreased lifetimes [24���27]. SIFs are surfaces coated with subwavelength-sized silver particles which have a heterogeneous size distribution. Similar enhance- ment effects were also observed with silver colloids [28] and fractal silver surfaces [29]. The emissions from a me- tal���ligand complex [30] and a lanthanide luminophore [31] were also found to be enhanced. The results were uniformly consistent with an increase in the radiative de- cay rate of the fluorophores. This is an unusual effect be- cause the decay rate of a fluorophore is determined by its extinction coe���cient and the local refractive index [32���35]. The radiative decay rate is not changed in most fluorescence experiments [22]. In more recent studies we examined fluorophores near continuous thin silver ( 50-nm) films [23]. Gold films of a similar thickness are used for surface plasmon resonance (SPR). Gold and silver films both display a plasmon resonance absorption. We found that excited fluorophores near the silver films radiated into the underlying glass substrate with a sharp angular distribu- tion [36���38]. The directionality of the emission in the prism and its unique polarization properties indicated that the radiation was from surface plasmons induced in the metal by the nearby excited fluorophores. How- ever, the wavelength distribution of the emission matched precisely with the usual emission spectra of the fluorophores. We call this phenomenon surface-plas- mon-coupled emission (SPCE). We also observed SPCE with gold [39] and aluminum [40] films. SPCE was ob- served using a gold film with electrochemiluminescence [41], eliminating the possibility that the radiating plas- mons were created by incident light. We were surprised by the observation of SPCE on gold and aluminum, which are known to strongly quench fluorescence [42��� 48]. Quenching has also been observed with some silver particles [49���51]. We were puzzled by the observation of SPCE, espe- cially with metals known to quench fluorescence. From the thickness of the samples we knew that the fluo- rophores were at short distances from the metal (10���80 nm) and not at the more distant Drexhage reflec- tive-field conditions (100���500 nm). Our SPCE results suggested that the excited fluorophores at these short distances induced electron oscillations in the metal film which radiated into the glass prism. This was surprising because the literature indicates that the emission is quenched at these short distances by lossy surface waves [17,18] and thus would not be observable. We felt intu- itively that if a fluorophore induces oscillations in the metal when at the larger reflective distances then it would continue to induce oscillations as the fluorophore entered the short-range quenching zone. Additionally, we knew that metallic particles enhanced fluorescence at short fluorophore-to-metal distances of 5���10 nm [52], so that a fluorophore is not necessarily quenched at 5���10 nm from a planar metal surface. These disparate results led us to ask why metal surfaces and particles have different effects on fluorescence. Why is there quenching at the shorter distances from a planar surface but not near the surface of a metal particle or a thin (50- nm) continuous metal film? We also questioned the physical meaning of LSWs. Being perplexed because some fluorophore-induced oscillations resulted in far-field radiation and other in- duced oscillations resulted in quenching, we examined the theory [17,18] in more detail. This led to the follow- ing conclusion. The statement that metals quench at the short distances is misleading, hides the actual origin of quenching, and prevents the effective use of fluoro- phore���metal interactions. We believe that metallic sur- faces do not necessarily quench fluorescence, except when there is some underlying absorption not due to electron motions. These absorptions are sometimes re- ferred to as interband absorption [53]. We suggest that oscillations created in metals at short fluorophore���metal distances cannot radiate because of optical constraints at the metal���sample interface. The observed quenching 1 Abbreviations used: LSWs, lossy surface waves MEF, metal- enhanced fluorescence RP, radiating plasmon SIF, silver island films SP(R), surface-plasmon (resonance) TIR, total internal reflection SPCE, surface plasmon-coupled emission NRP, nonradiative plas- mon NSOM, near-field scanning optical microscopy. 172 Radiative decay engineering / J.R. Lakowicz / Anal. Biochem. 337 (2005) 171���194

at short distances from the metal may not be due to a typical chromophoric absorption process, and the short range interactions may not necessarily result in quench- ing. Plasmons created at short fluorophore���metal dis- tances may be trapped because of optical properties of the interface, and as a result they decay as heat. We now suggest that the plasmons will radiate the energy from a fluorophore whenever allowed by the optical conditions. For continuous surfaces the plasmons will radiate if wavevector matching occurs at a metal���dielec- tric interface. For colloids the induced plasmons will radiate whenever the scattering cross section of the col- loid is dominant compared to the absorption cross sec- tion of the colloid. The recognition that lossy surface waves are trapped plasmons is not a trivial observation. The term LSW im- plies that the energy cannot be recovered as a useful sig- nal. The concept of trapped plasmons suggests that changes in the optical conditions can allow the energy to radiate into the far field. We suggest that many of the different effects of thick mirrors, thin metal films, and metallic particles can all be understood with regard to the ability or inability of the plasmons to radiate. We refer to this concept as the radiating plasmon (RP) mod- el. The radiating plasmon model has implications for the practical applications of fluorophore���metal interactions. Metallic structures can be selected by consideration of whether far-field plasmon radiation can occur from flu- orophore-induced plasmons. Strong interactions of flu- orophores with the surface can be desirable, rather than something to be avoided, because even low-quan- tum-yield fluorophores can transfer their energy quickly to the metal, which may then radiate with a higher e���- ciency than the fluorophore in free space. The far-field radiation pattern and e���ciency can be calculated from electrodynamic theory, allowing the rational design of fluorophore���metal nanostructures with the desired opti- cal properties. In the following sections we expand on the radiating plasmon model within the context of fluorescence detec- tion. It is not practical to describe all the theory for an oscillating dipole interacting with a conducting metal surface, and to do so would obscure the essence of the model. We summarize those aspects of the theory which are required to understand how the radiating plasmon model can be used to predict the effects of nearby metals on fluorophores. We also show how consideration of the radiative strength of plasmons can provide new oppor- tunities for the use of fluorescence in the biological and medical sciences. Metal���dielectric interfaces and surface plasmons The optical properties of metallic surfaces are com- plex and cannot be completely described in this report. Instead we will describe these properties in a way which provides an intuitive understanding of surface plasmons. The term plasmon or surface plasmon is used to describe the collective oscillations of a group of electrons in a me- tal [54���57]. The term plasmon indicates that the elec- trons are free to migrate in the metal in a manner similar to that of ions in a gaseous plasmon. The term surface plasmon polariton (SPP) has been suggested to describe optically induced electron oscillations. How- ever, the term surface plasmon is now commonly used with the same meaning. In the case of optical excitation the frequency of electron oscillation is the same as the frequency of incident light. Surface plasmons can occur on planar metal surfaces or in metallic particles. Scheme 1 illustrates surface plas- mons on a flat metal surface. Surface plasmons can be created by illumination of the metal surface with p-po- larized light. However, surface plasmons are induced by incident light only under special optical conditions (below). Plasmons are not created when silver surfaces or mirrors are illuminated, and hence these surfaces re- flect rather than display the plasmon absorption. Crea- tion of surface plasmons requires illumination of a thin metal film through a glass prism or some higher dielectric-constant material. Surface plasmons are also created by direct illumina- tion of metal colloids (Scheme 2), resulting in rapid oscillation of the spatially bound electrons [58]. Colloids display vibrant colors because of a combination of absorption and scattering [59]. Unlike planar metal sur- faces, no special conditions are required to observe the surface plasmon absorption in colloids. The term ������ab- sorption������ is often used to describe colloids, but the more correct term is extinction because there are both absorp- tion and scattering contributions to the colors. In discussing fluorophore���metal interactions we use the terms far-field or far-field radiation to indicate a wave propagating in space away from its source. We use the term near-field to indicate the field around an Scheme 1. Schematic of surface plasmons on a metal surface. Adapted from [54]. Radiative decay engineering / J.R. Lakowicz / Anal. Biochem. 337 (2005) 171���194 173