Spectroscopy and electron microscopy imaging of a single metal nano-object

The new optical resonances appearing with size reduction of an object down to the nanometric scale have motivated considerable interest in the academic and industrial domains. From different origin in different materials, e.g. electron quantum confinement in semiconductors, or dielectric confinement in metals, they offer a wide range of possibilities for designing the linear and nonlinear optical properties of nano-structured materials, and for adapting them to perform new function. In metal nanoparticles, the main feature is the surface plasmon resonance (SPR) whose characteristics, wavelength, line shape and amplitude, depend on the size, shape, structure and environment of the particle [1–4]. It is a consequence of the resonant interaction of the metal electrons with the electromagnetic field that leads to a large local field enhancement in and around a particle [1–5]. This local field effect modifies not only the light absorption or scattering efficiencies of the nano-object, but also all the responses depending locally on the field, such as the nonlinear optical response of the particle or the luminescence or Raman scattering of surrounding molecules [1–5]. Optimization of the SPR characteristics and their adaptation to specific applications have thus led to a considerable amount of work during recent decades, with, in particular, the goals of adapting the SPR wavelength to the required application and reducing the SPR width to enhance the amplitude of the local field.

With the advance of fabrication techniques, metal nanoparticles with different shapes and sizes can now be created by chemical or physical means [6–9]. Though a good control of their morphology is now possible, nanosystems still exhibit sufficiently large particle-to-particle variations to impact the optical response measured on an ensemble of particles. Furthermore, local variation of the environment of particle, both in matrix or in solution, leads to additional dispersion of the characteristics of their optical spectrum. This is a consequence of the large sensitivity of the surface plasmon resonance frequency and width on the particle shape and environment. It, for instance, shows-up in a SPR width much larger than the computed homogeneous width in ensemble measurements [10–13]. This dispersion is preventing not only precise comparison of the experimental and computed spectra, but also quantitative analysis of efficiency in nanosensor applications. These limitations can be overcome by investigating a single metal nanoparticle, provided the geometry of the studied object is known. For simple objects, this can be performed fully optically [12–15], but for more complex ones it requires combining optical measurement and electron microscopy of the same object as done with transmission electron microscopy [16–18]. In this purpose, though its spatial resolution is more limited, scanning electron microscopy (SEM) offers many advantages as it does not require an electron transmitting substrate. We discuss here correlation of the spatial modulation spectroscopy technique with SEM of the same single metal nano-object.

Optical investigations of a single nano-object were performed using the spatial modulation spectroscopy (SMS) technique [13–15, 19]. The position of the sample, formed by metal nanoparticles deposited on the surface of a substrate, is periodically modulated under the focal spot of a laser beam, tightly focused by ^a×100 microscope objective with numerical aperture 0.8 (figure 1). The sample position is scanned by an XY scanner. When a particle is in the focal spot, the transmitted light energy is modulated, with an amplitude proportional to the particle extinction cross-section, σ ^_ext (λ)=σ ^_abs (λ)+σ ^_sc (λ). The transmitted light is collected by a microscope objective and detected by a silicon photodiode. The modulated part of the transmission ΔT is demodulated by a lock-in amplifier and the total sample transmission T is simultaneously measured using a digital voltmeter. After calibration, the relative transmission change ΔT/T yields the absolute value of the single-particle extinction cross section σ ^_ext , with a sensitivity in the few ^{nm 2} range [15]. For the measurements presented here, a Ti:sapphire oscillator tunable in the λ=640–1080 nm range has been used as the laser source.

Zoom In Zoom Out Reset image size

As for all far-field techniques, the limited spatial resolution (roughly half the optical wavelength) is overcome by using diluted samples. The surface density of the particles has to be small enough so that only one particle is under the focal spot, i.e. their separation has to be larger than about 1 μ^m. Such large separations can easily be performed using electron beam lithography. We have investigated here individual gold nanorings of nominal external diameter D=120 ^nm, thickness T=25 ^nm and height H=30 ^nm (inset of figure 2). They were grown on a ^Si ₃ ^N ₄ substrate to form a square network with lateral size 2 μ^m.

Zoom In Zoom Out Reset image size

The extinction cross-section spectra of individual gold nanorings computed using the finite element method is shown in figure 2. The nominal sizes of the object are used and the light beam is assumed to be incident normal to the ring plane. Calculations were performed with or without taking into account the substrate (free nanoring in vacuum). As expected, because of the large refractive index of ^Si ₃ ^N ₄ (about 2 for a wavelength around 1 μ^m), the latter substantially influences the position of the surface plasmon resonance. This induces a red shift of the resonance by more than 200 nm (figure 2). Due to the isotropy of the system, no light polarization dependence is predicted.

The extinction spectra measured on different single nanorings are shown in figure 3. For most of the tested objects, similar extinction spectra have been observed indicating a SPR shifted to the near infrared part of the spectrum, i.e., out of the tunability range of our laser. Only a few objects exhibit a maximum of their extinction cross-section in the 700 ^nm–1 μ^m range. Furthermore, for all these objects the results show a large dependence on the direction of the linear polarization of the incident light. Though the amplitude of σ ^_ext is consistent with the computed one (figure 2), both observations are in contrast to the computed spectra. This suggests that the fabricated nano-objects significantly differ from the nominal shape. The polarization dependence indicates anisotropy of the objects, with an elliptical rather than cylindrical shape. The SPR red shift suggests either a thinner ring thickness, T, or a reduced height, H, increasing the object aspect ratio D/H (leading to a red shift as in ellipsoids or nanorods).

Zoom In Zoom Out Reset image size

A better insight into the connection between the size and shape of a nano-object and its optical response can be obtained by imaging it by electron microscopy. We used here scanning electron microscopy (SEM), which permits observation of a non-electron transparent substrate at the expense of a limited spatial resolution. The SEM images of the three objects whose spectra are shown in figure 3 are shown as insets. The first two (figures 3(a) and (b)) exhibit a ring-like shape with a significant anisotropy, as suggested by the observed light-polarization dependence. The fact that their SPR is shifted to the near-infrared, is most probably due to a reduced height as compared to the nominal one. The actual height cannot be easily inferred from the SEM image which provides only a two-dimensional image in the ring plane. In contrast, the third object is strongly deformed with a crescent-like shape, probably at the origin of its blue shifted spectrum as compared to the other objects. In principle, more precise discussion of the optical response in correlation with the observed object shape could be performed using the latter as an input parameter in the numerical model. However, the object height is still unknown here, and is also probably slightly changing from object to object, making this comparison non-parameter free.

The optical extinction spectra of individual nanoring-like gold nano-objects fabricated on a ^Si ₃ ^N ₄ substrate were investigated using the spatial modulation spectroscopy technique. The measured frequency of their surface plasmon resonance is found to be red-shifted as compared to the computed one, even when taking into account the presence of the substrate, with a large light-polarization dependence. This clearly stresses the difficulty in properly interpreting optical spectra, even at a single object level. Imaging the same single nano-object by scanning electron microscopy shows that these discrepancies are due to shape and size deviations of the objects from their nominal geometry. As only 2D electron images are obtained, comparison of the computed and experimental spectra still requires introduction of a parameter (i.e. the dimension in the direction perpendicular to the surface) only permitting a qualitative discussion of the impact of the object geometry on its optical spectrum. Quantitative comparison could be achieved using electron tomography to obtain a 3D image of the investigated nano-object.

N Del Fatti acknowledges support by Institut Universitaire de France.