The Optics Laboratory
Group of Hans Hallen, Physics Department, North Carolina State University
The page is about using plasmonics to enhance Raman spectra using a nano-antenna or nano-particle. It can also be done using a scanning probe tip, known as nano-Raman or tip-enhanced Raman (TERS) or by many other names, as we were the first to do in the mid 1990's with both spectra and images, see the near field Raman page, and its cousin gradient-field Raman (GFR), also observed with nano-antennas.
Plasmonics use resonances of the density of surface electrons with an incoming field to locally enhance the electric field strength. This increases the optical interaction in that small volume of space where the resonance is taking place, and thus the signal from a few molecules placed there can be detected. Our efforts have focussed on: (1) pushing the nano-antenna approach into the deep UV optical range , . This allowed dual-resonance Raman, with one resonance the plasmonic antenna, and the second an absorption resonance in benzene to enable resonance Raman spectroscopy. The joint gain in signal is in the millions -- within that small volume of the plasmon resonance. We see a GFR signal as part of this work -- it is due to the proximity to metal; (2) Improving the performance of upconversion nano-particle (UCNP) sensors, which don't blink or bleach like other probes, but suffer from weak excitation. The plasmon resonances improve the excitation level, and the metal proximity changes the spectral characteristics  compared to far from the metal; (3) Understanding plasmonic particles: the physical effects of defects such as lumps of metal on realistically fabricated plasmonic particles. The qualitative understanding of the process does not agree with naive expectations, as  describes; (4) Building a mass-producible nanoparticle sensor that has a larger enhanced volume that is located on the outside of the particle. Both aspects are important. The larger volume means more efficient detection of trace species in a solution within which the particles are immersed (we find that the real signal gain when the particles are spread out may be very limited unless a very dense array of nano-plasmonic devices is used). The location on the outside of the particle (as opposed to squeezed between two neighboring particles) is advantageous since it makes diffusion much more likely to bring a trace species into that volume.
Our aim is a deep UV nano-antenna that can be used for plasmon excitation light focussing and DUV resonance Raman. More detail is here ,  . It extends the wavelength range dramatically compared to earlier work, as seen in Fig. 1.
Figure 1. An extrapolation from the literature [Crozier, K. B., Sundaramurthy, A., Kino, S., and Quate, C. F., J. Applied Physics 94, 4632 (2003).], we show that our antenna resonance is consistent with the simple model used to determine the resonance of a metal triangle. Our antenna has two triangles to form a bowtie, but the same plasmon resonance holds for each.
We expect the maximum enhancement to be when the incident light polarization lines up with the long axis of the bowtie. This is tested in Fig. 2, from which we can also extract the enhancement (assuming no enhancement with the polarization perpendicular to the bowtie long axis, and find that it the irradiance enhancement is ~30,000 ).
Figure 2. (a) An overlay of the angles of incident polarization overlayed on an SEM of the nano-bowtie antenna are color-matched to the spectra in part (b). (b) The excitation-power normalized spectra from repeated measurements. The top spectrum was taken when antenna was in the focused laser spot. The middle one was also taken from a bow-tie site with a different polarization angle. The bottom spectrum was measured after moving the focus spot laterally to a region without bow-tie nano-antenna. The polarization was rotated an angle of α from the long bow-tie axis. The laser wavelength was 258.8nm. The excitation power used was in the range of 0:05–0:13 mW. The image integration time was 120 s.
It is interesting to compare three spectra: the far-field off resonance, the far field with excitation tuned to an absorption (resonance Raman), and the near-field (bowtie) plasmon enhanced resonance Raman. Figure 3 shows that the resonance Raman has many more excited vibration lines than the nonresonant case, and that different vibrations are excited. This is explained by the fact that vibrations which are strongly affected by the bond breaking of the absorption will be enhanced, and that several of these phonon types may be created at once (overtones) or mixtures with these phonons (combinations) are often created, as discussed on the resonance Raman page. The addition of the plasmonic nano-enhancement does not dramatically change the observed overtone and combination bands compared to the just-resonance-Raman case. New, unexpected vibration lines, which are large enough that should have been observed if they were present, are found, however, as shown in Fig. 3 (b).
Figure 3. (a) Raman spectra of various types are compared. The upper plot contains a non-resonant Raman spectrum and a resonance Raman spectrum taken without any nearby metal. The lower plot contains a resonance Raman spectra in the presence of the metal nano-antenna. The plots are shifted so that the horizontal energy axes align, and the green dotted lines relate stronger Raman features in the plots. The label colors are black: observed in both resonant Raman experiments (with and without metal), blue: likely too weak to be observed without metal enhancement, and red: infrared vibrations, the GFR effect, so only in the metal nearby, bowtie spectrum. (b) The bowtie spectrum repeated with zoom-ins to show the presence of the infrared peaks. The same color scheme is used.
To give further evidence that these bands observed in the bowtie, near-metal spectrum but not in the far-field spectrum are gradient-field Raman, GFR, related, we must show that infrared spectra are strong at these energies (wavenumber shifts). This is done in Fig. 4. It is interesting that we easily observe the low energy combination mode line in the far-infrared specteal region. The consistency between the strong infrared lines from the literature and our 'new' peaks is good evidence that the gradient field effect is present here, as expected.
Figure 4. The resonance Raman peak with nearby metal (nano-antenna) shows GFR effects, as indicated by comparison to the infrared spectrum [NIST WebBook entry for benzene,
The behavior of Up Conversion Nano Particles (UCNP) have their own web page because the nearby metal impacts the interaction of the light with material (but differently from the GFR effect). In short, it increases the local photon density of states, which increases the interaction of light with the material, or the photon transitions within the material, while the phonon (energy loss to vibrations rather than light) transitions are largely unchanged. This changes the relative de-excitation transition probabilities, which we observe in this system and are described on that page. It is also in the paper , along with the more classical plasmonic field strength enhancements.
More detail is in the paper . Plasmonic nanoparticles can be fabricated by coating a silica core with metal. If it is coated around the entire surface, a simple particle is formed, and the resonance depends upon the size of the particle and the thickness of the coating. Very small particles show only effects near the wavelength at which the real part of the metal dielectric constant is near -2 (metals tend to have negative real parts to the dielectric constant above the plasma frequency). This is due to the polarizability of a sphere. As the size increases, a dipole resonance emanates from this wavelength and moves to longer wavelengths. Next would comes a magnetic dipole (not permitted for a spherically symmetric particle), then a quadrupole mode, et cetera. We see all these effects as the metal coating on a core is increased from only a tiny fraction to a large fraction of a fairly large silica core. Figure 5 shows the progression after the magnetic dipoles have moved to longer wavelengths and a quadrupole can now 'fit' into the metal on the particle, since the size of the metal is large enough to accomodate the nodes at the wavelengths available (above the dielectric constant ~ -2 wavelength, ~500 nm in gold). These are calculations of the scattered light as a function of wavelength. The larger metal part also drives the magnetic dipole (seen) and the electric dipole (off scale to right) to longer wavelengths so that those modes fit into the larger space. We identify the modes by examining the rms current and electric field volume images of our finite element calculations.
Figure 5. Dependence of calculated far field scattering on the partial shell fraction. The numbers are the fraction of the total width of the particle covered by metal.
Real, fabricated particles do not usually follow the 'perfect partial coating' that was used in these calculations. A combination of grain growth and diffusion usually leaves protrusions or bumps near the edges, particularly when they are grown by evaporative coating in a vacuum system, as ours are. We then transfer them to another substrate and the orientation is changed. Examples of electron microscope images of real particles are shown in Fig. 6. Particle (a) is close to an ideal one, while (b) and (c) each have a protrusion on the side and the top, respectively. Particle (d) has a 'crown' of protrusions ringing its metal rim. One would naively expect the latter particle to behave simply as a particle with a larger fractional coating. After all, the spacing between and size of the bumps are much smaller than the wavelength. This is not the case, however, as is evident in the spectra of Figs. 6 (scattered measurements) and 7 (near-field calculations).
Figure 6. Measured far field spectra and corresponding SEM images, comparing a) control: no protrusion, b) top protrusion, c) side protrusion, and d) crown of protrusions configurations. The silica cores shown in the SEM images are of 90 nm in diameter.
What has happened for the protrusions is a 'splitting' of the magnetic dipole peak into two parts. The electric dipole also splits, but is off scale. We can understand this by recalling that the magnetic dipole peak is a resonant sloshing (current <-> magnetic) of electrons back and forth across the bowl of the partial shell up to the edges on opposite sides. The 'bowl' has different lengths when the protrusion is included or not, so two peaks makes sense. There is more complication in the actual modes, of course, since simply on or off the protrusion are not 'normal modes.' Rather, a combination of those is required, as can be seen in the field and current maps in the paper and supplementary material. Continuing this trend, the crown particle has many peaks, some only visible as shoulders in the data. The difference between the top and the side crown arises from how easily the light can interact from the protrusion. It is arriving mainly from near the top side, so if the protrusion is on the top, it is not shadowed at all and can be effectively scattered (strong scattered signal measured).
Figure 7. Calculated near field spectra comparing a) no protrusion, b) top protrusion, c) side protrusion, and d) crown of protrusions configurations.
You might have noticed that the relative strength of the spectra changes in the near and far field. The strongest scattered mode is on the perfect partial shell, but that also has the weakest local electric fields. This is easily understood via conservation of energy. Either the incident energy is scattered, which is is effectively for the perfect case, or it is trapped in the near-field of the metal particle, where it can be used to couple energy into nearby material or is eventually absorbed by the metal.
The eukaryotic living cell is compartmentalized into functional and structural domains. The mammalian cell consists of the chromatin packed nucleus and cytoplasmic organelles such as mitochondria, ribosomes, golgi bodies, lysosomes and others. Intracellular structures regulate cellular functions such as, signaling, internalization, and transport, which enable cellular growth, differentiation, division, use and storage energy, interaction with its external environment and gene expression. Many such biological processes occurring in the subcellular organelles within a cell results in changes in temperature, pH, electrical potentials, and concentration of ions, to name a few. Abnormal changes in intracellular pH for example, have been linked to cancer. Thus, it is useful to have noninvasive (so no fluorescent dyes or other chemicals inserted, use a push in, pull out probe as we plan, so use label-free optical spectroscopy, enhanced as above.).
We propose to address these issues with a novel nanostructure -- more info coming after the paper is out. The proposed 3D nanostructure retains the small spaced antennas but increases the availability and volume of the detection volume compared to conventional plasmon-based detection such as between two particles. The detection is at the outer edge of the structure, maximizing coupling to the environment. Advantages over other nanoparticle sensors for facile sensing applications are (1) consists of a single particle, so does not require positioning of two objects with nanometer relative accuracy (as particle pairs require), (2) has the enhanced volume exposed on the surface of the particle so that when the particle is brought close to a cellular compartment, the membrane and contents of the compartment are within the enhanced volume (which would not be the case for the enhanced volume of two or more spherical particles, (3) is capable of exciting dark plasmons, which show large local field enhancements as is useful for this application, and do so when illuminated by ‘plane wave’ excitation, (4) is easy to manipulate both in position and rotation, (5) possesses enhancement primarily on one side, so that (a) a combination of position and rotation control can localize sensing to a surface or single cell compartment, and (b) the variability in time of a signal can be used to deduce motion of a species into and out of this region, and (6) has a high figure of merit compared to many other particles.
The figure of merit is equal to the integral of the difference between the local electrical field magnitude squared and the incident field squared over the volume occupied by the sensor and its surroundings. This is easy to calculate in computational tools (finite element or FDTD), is independent of the volume chosen (in a simulation using proper boundary conditions), and has the qualitative meaning of the increase in signal (linear response) in the presence of the sensing particle compared to its absence. The figure of merit is not applicable for isolated single molecule imaging, and neglects the impact of strong absorption of spectroscopic signal by metal within a couple of nanometers of the metal surface. The subtraction of the incident field over the entire volume, including that occupied by the sensor volume, is important as it corrects for the sensing particle size, making comparisons between different sized and shaped particles meaningful, and approaches zero as the sensing particle disappears. It is a significant improvement over using simply the average field, since that depends upon the volume averaged over (it is very volume-dependent), and does not have the quantitative connection to signal strength. Further, it is not a peak field, which has application for single particles when the particle is in the proper place. The problem with peak fields is that they often occur so close to the metal surface that any spectroscopic energy tends to be absorbed by the metal, producing electron-hole pairs at small distances for which the evanescent photon momentum is sufficient, rather than scattering the energy to the far field, as noted above. Finally, our figure of merit naturally decreases as the excitation leaves resonance, so spectroscopic comparisons are meaningful.
When the sensing application is already enhanced, by resonance Raman, for example, and the material to be studied is uniformly spread, the raw signal gain by inserting a single nano-plasmonic particle within the focussed laser volume may be very modest. We show in another upcoming paper that a nanobowtie antenna with a gain (in the active region) of >20,000 only gives a real signal gain of ~20%, in part due to the low occupancy within the antena gap region (even with a relatively strong solution), and in part due to the volume ratio of the laser spot size to the volume between the antennas. The conclusion is that nano-plasmonic sensors are appropriate for nanoscale studies, but need to be in very dense arrays if they are to produce real signal gains in dilute (or even not so dilute) solutions of molecules of interest.
More info is in the papers.
NC State University | Physics | Optics Home
Copyright ©2000-2017, Hallen Laboratory, NCSU, Raleigh, NC. www.physics.ncsu.edu/optics
Comments or questions? Hans_Hallen@ncsu.edu