HTML
-
The main contribution to the electromagnetic scattering from a subwavelength obstacle usually originates from a dipolar term in the multipolar decomposition. In other words, under linearly polarized plane wave illumination, the particle becomes predominantly polarized along the direction of the incident electrical field (if the magnetic response can be neglected) and re-radiates the energy according to the dipolar emission pattern. The highest intensity of the dipolar radiation propagates in the directions perpendicular to the dipole and, as the result, along the wave vector of the incident linearly polarized wave. Consequently, scattering in the forward direction (along the incident wave vector) is significant. The common formulation of the optical theorem postulates a direct proportionality between the scattering amplitude in the forward direction and the extinction cross-section1-4. However, the situation can be drastically different for vectorial beams that may carry longitudinal field components, optical angular momentum or transverse spin25. For example, radially polarized beams have a doughnut-like intensity profile for the transverse polarization directions and, most prominently, a strong longitudinal polarization component along the propagation direction at the beam axis26. The intensity map and the electric field structure of a focused radially polarized beam, the scattering of which will be studied below, are presented in Fig. 1.
Fig. 1 Schematic representation of a nanoparticle illuminated by a radially polarized beam.
Evolution of the electric field structure of the incident beam illuminating the nanoparticle from the left (along kinc) is shown by black arrows together with the dipole moment (d) induced in the nanoparticle (green-white arrow).The intensity of the beam is shown by the color map. The directions of the power flow of the scattered field Pscat are shown by red-white arrowsThe optical theorem in its textbook formulation provides a straightforward way to calculate the nanoparticle extinction cross-section ($C_{{\rm{ext}}}^{{\rm{OT}}}$) by relating its magnitude to the value of the far-field component of the normalized scattered electric field amplitude evaluated in the forward direction (along the incident wave vector) ${\bf{e}}_{{\rm{scat}}}^{{\rm{far}}}\left({{\bf{k}} = {\bf{k}}_{{\rm{inc}}}} \right)$ 2-4:
$$ C_{{\rm{ext}}}^{{\rm{OT}}} = \frac{{4\pi }}{{k\varepsilon _{{d}}^{1/2}}}{\Im} \left\{ {{\bf{p}}^ \ast \cdot {\bf{e}}_{{\rm{scat}}}^{{\rm{far}}}\left( {{\bf{k}} = {\bf{k}}_{{\rm{inc}}}} \right)} \right\} $$ (1) where p is the unit vector signifying the polarization of the incident wave, kinc and k are the wave vectors of the incident and scattered waves, respectively, $\left| {\bf{k}} \right| = \left| {{\bf{k}}_{{\rm{inc}}}} \right| = 2\pi \varepsilon _{{d}}^{1/2}/\lambda \, \epsilon_{d}$, is the permittivity of the surrounding dielectric, and ${\bf{e}}_{{\rm{scat}}}^{{\rm{far}}}\left({{\bf{k}} = {\bf{k}}_{{\rm{inc}}}} \right)$ is related to the scattered electric field ${\bf{E}}_{{\rm{scat}}}^{{\rm{far}}}\left({\bf{r}} \right)$ as follows:
$$ {\bf{E}}_{{\rm{scat}}}^{{\rm{far}}}\left( {\bf{r}} \right) = \left| {{\bf{E}}_0} \right|\frac{{e^{{ikr}}}}{r}{\bf{e}}_{{\rm{scat}}}^{{\rm{far}}}\left( {{\bf{k}},{\bf{k}}_{{\rm{inc}}}} \right) $$ (2) where $\left| {{\bf{E}}_0} \right|$ is the amplitude of the incident wave.
Alternatively, the extinction cross-section can be evaluated directly,
$$ C_{{\rm{ext}}}^{{\rm{dir}}} = C_{{\rm{abs}}}^{{\rm{dir}}} + C_{{\rm{scat}}}^{{\rm{dir}}} $$ (3) as the sum of the absorption $C_{{\rm{abs}}}^{{\rm{dir}}}$ and scattering $C_{{\rm{scat}}}^{{\rm{dir}}}$ cross-sections. The absorption cross-section can be calculated as an integral of the absorption losses over the nanoparticle volume V, normalized to the incident power flow:
$$ C_{{\rm{abs}}}^{{\rm{dir}}} = - \frac{{\mathop {\int}\limits_V {\frac{1}{2}{\Re} \left\{ {i\omega {\bf{E}} \cdot {\bf{D}}^ \ast } \right\}{{d}}^3{\bf{r}}} }}{{\frac{{\varepsilon _{{d}}\varepsilon _0c}}{2}\left| {{\bf{E}}_0} \right|^2}} $$ (4) where E0 and E are the incident and total electric fields, respectively, and D is the electric displacement. The scattering cross-section can be calculated as an integral of the intensity of the scattered fields over a surface S enclosing the particle, normalized to the same incident field intensity:
$$ C_{{\rm{scat}}}^{{\rm{dir}}} = \frac{{{\int} {\mathop {\int}\limits_S {{\bf{P}}_{{\rm{scat}}}^{{\rm{far}}}{{d}}s} } }}{{\frac{{\varepsilon _{{d}}\varepsilon _0c}}{2}\left| {{\bf{E}}_0} \right|^2}} $$ (5) where ${\bf{P}}_{{\rm{scat}}}^{{\rm{far}}}$ is the power flow of the scattered waves. We note that other semi-analytical approaches, such as multipole expansion, can also be applied for calculations of extinction cross-sections for linearly or radially polarized beams19.
For the plane wave illumination of a gold nanoparticle, the direct method for the calculation of the extinction cross-section ($C_{{\rm{ext}}}^{{\rm{dir}}}$) and the method based on the optical theorem ($C_{{\rm{ext}}}^{{\rm{OT}}}$) show excellent agreement (Fig. 2a). The extinction in the spectral range λ = 480–550 nm corresponds to the plasmonic dipolar resonance of the particle. The angular distribution of the far-field scattering has the characteristic shape corresponding to a dipolar radiation pattern, namely, cos2φ, where φ is the scattering angle (Fig. 2c). As expected, the scattering dipole is induced along the y-direction along the polarization of the incident plane wave.
Fig. 2 Breakdown of the conventional formulation of the optical theorem for a radially polarized beam.
a, b Extinction cross-section spectra of gold nanoparticles with a radius of 50 nm (the data are normalized to the geometrical cross-section of the nanoparticle Cgeom), calculated using the optical theorem (Eq. (1), solid red lines) and by the direct evaluation of the sum of the absorption and scattering cross-sections (Eq. (3), black lines) for (a) linearly and (b) focused radially polarized illumination. c–f Angular scattering diagrams for the nanoparticle illuminated by (c, e) a plane wave linearly polarized along the y-direction and (d, f) a focused radially polarized beam. The illuminating wave propagates along the z-direction and has the wavelength λ = 530 nm. The particle is located at the center of the diagramThe situation is drastically different for the radially polarized incident beam. The wavelength dependence of the extinction cross-section calculated using the direct integration method shows a distinctive peak at the localized surface plasmon resonance of the particle (Fig. 2b, black line), which is in good agreement with the case of the plane wave excitation (Fig. 2a). At the same time, the extinction cross-section evaluated using the optical theorem given by Eq. (1) is practically zero (Fig. 2b, red line), within a numerical noise defined by the accuracy of the simulations. This means that the optical theorem in its common form cannot be applied for such beams. Here, we note that for the case of optical beams focused to dimensions comparable to the size of the scattering object, reconsideration of the usual notion of the cross-section is required due to the variation of the beam intensity across the object. For example, in the case of scattering of localized electron wave packets, this was done via the definition of the cross-section through the number of the scattering events, normalized to the introduced effective luminosity of the wave packet5. In our case, in analogy to this approach, we used the power extinguished from the beam (equivalent to the number of photons removed from the beam), normalizing it for simplicity to the maximum intensity in the focal plane (Fig. 1).
The apparent contradiction created by the optical theorem described by Eq. (1) can be understood from the angular scattering diagram (Fig. 2d). For the radially polarized beam, the entire nanoparticle is located in the region of space where the longitudinal component of the incident field dominates (Fig. 1) and the transverse polarization is vanishingly small. Furthermore, it has an axisymmetric structure with the transverse fields directed out of the beam axis at the location of the scatterer, prohibiting the excitation of a transverse dipolar mode (Fig. 1). Consequently, the dipole moment in the particle is excited only by the longitudinal components along the beam axis, giving rise to pronounced energy re-radiation directed perpendicular to the latter. The dipolar excitation also causes the related absorption losses in the metal particle. This leads to significant non-zero values of scattering and absorption cross-sections, resulting in a considerable value of the extinction cross-section. However, the optical theorem given by Eq. (1) fails to reflect this: due to the symmetry, the re-radiation of the longitudinal dipole in the forward direction (along the dipole axis) is zero, and therefore, the optical theorem returns a zero extinction cross-section (Fig. 2d). The visual comparison between Fig. 2c and d suggests that they are almost perfect replicas of each other if a 90° rotational transformation is applied to either one of them. This observation enables us to draw an intuitive conclusion regarding the source of the violation of the optical theorem in this formulation. Instead of traveling along the optical axis, the scattering is deflected by 90°, which is the optimal angle for minimizing the forward scattering. Moreover, as seen in Fig. 2e, f, the violation also occurs in the case of a nonresonant excitation at λ = 520 nm for the nanoparticle in oil (the resonance is moved in this case to λ = 600 nm), which was experimentally examined in further studies described in the next section. In the case of the focused Gaussian beam, both transverse (dominant) and longitudinal (which are the consequence of focusing) components are present at the particle position. As was shown above, for the latter components, there is a complete violation of the optical theorem, while for the former components, the optical theorem holds, overall leading to a partial theorem violation. Generally, stronger beam focusing corresponds to a higher ratio between the longitudinal and transverse components and a more significant violation of the optical theorem. As was found in ref.5 for the axisymmetric scattering of vortex electron wave packets, zero forward scattering and yet a non-zero overall scattering cross-section can be observed even for scalar localized incident fields of a complex structure with a vortex phase change around the wave packet axis. However, the vectorial nature of the electromagnetic field essentially amplifies the effect: the zero divergence of the electromagnetic field at the node of the fields at the beam axis leads to the presence of the longitudinal field components, resulting in efficient excitation of the dipolar mode in the nanoparticle along the beam axis and thus efficient scattering, resulting in essential extinction simultaneously with zero forward scattering.
-
To verify the predictions of the numerical modeling, single-particle spectroscopy was performed, allowing the direct imaging of the scattering pattern (Materials and methods section) (Figure 3a). First, linearly polarized excitation was studied with a flat linearly polarized wavefront at the center of the focal spot. These excitation conditions can be directly compared to a plane wave excitation implemented in the simulations. Figure 3b, c show excellent agreement of the experimental and theoretical angular distribution in the back-scattering zone, clearly demonstrating a wave vector distribution that is symmetric with respect to the ky = 0 plane with gradually decreasing magnitude toward higher ky. They directly demonstrate the excitation of a dipole along the excitation polarization direction, as seen from the field map cross-sections of the full scattering diagram (Fig. 2e). Due to the nature of the Fourier plane detection, the number of the wave vectors per (dkx, dky) interval is recorded. Hence, for a uniform angular distribution, the wave vectors scattered at higher angles have higher density than those close to the optical axis. This explains the higher intensities at the sides of the Fourier images in Fig. 3b, c for higher kx.
Fig. 3 Experiments in the visible spectral range.
a Schematic of the Fourier imaging setup used in the experiments. b–e Angular distribution of the far-field back-scattering in the case of a nanoparticle illuminated by (b, c) a linearly polarized plane wave and (d, e) a radially polarized beam. The results obtained in the optical experiments (b, d) are compared with the finite element method numerical modeling (c, e). The parameters of the nanoparticle and illumination are as in Fig. 2For radial polarization, the experimental and numerical observations are again in excellent agreement (cf. Fig. 3d, e). It can be seen that the Fourier intensity in the entire central region of the map around kx = ky = 0 is virtually zero, gradually increasing toward higher kx and ky and clearly possessing a polar angular symmetry. Since the subwavelength size of the sphere only allows the dipolar plasmonic resonance, this provides clear evidence that the direction of the excited dipole is along the z-axis, which can be easily seen by the comparison of wave vector distributions in the Fourier images with the full scattering diagram in Fig. 2f. Such an orientation of the excited dipole inevitably leads to the absolute zero value of the scattering field (and consequently its imaginary part) in the forward direction along the z-axis. Hence, we demonstrated a complete violation of the optical theorem in its conventional form for radially polarized beams both experimentally and numerically: while the optical theorem predicts an extinction cross-section to be zero $C_{{\rm{ext}}}^{{\rm{OT}}} = 0$ (Eq. (1)) on the basis of the zero scattering along the incident wave vector, a considerable extinction cross-section of the particle is observed with strong scattering of the incident radially polarized beam in the direction of large kx and ky wave vectors.
-
The microwave scattering experiments emulate the optical setup and provide direct measurements of the amplitude and phase distributions of the scattered waves23. The distributions of the scattered field amplitude in the forward direction demonstrate very good agreement with the modeling results (cf. Fig. 4b–e). For the linearly polarized illumination, the radiation profile corresponds to the dipolar moment excited along the direction of the incident polarization (Fig. 4b). On the other hand, the radiation profile in the case of the radially polarized illumination provides unambiguous evidence for the dipole moment excitation in the z-direction, along the beam axis (Fig. 4d). Again, the conventional formulation of the optical theorem is violated in this case: an essential scattering signal (and, therefore, considerable extinction cross-section) is evident, while Eq. (1) predicts zero extinction on the basis of the zero field measured in the forward direction (along the z-axis).
Fig. 4 Experiments in the microwave spectral range.
a Near-field scanning setup used in the microwave experiments. b–e Near-field distribution of forward scattering on a 3.5 mm metallic nanoparticle recalculated from the amplitude and phase maps measured at a distance of l2 = 14.5 mm from the particle center in the case of (b, c) a linearly polarized plane wave and (d, e) a radially polarized beam. The microwave radiation frequency is 9.5 GHz, corresponding to λ = 3.2 cm ($l_1 \approx \lambda$). The results obtained in the microwave experiments (b, d) are compared with the finite element method numerical modeling (c, e)The fact that the facet of the probe rectangular waveguide lies in the measurement plane and, therefore, has a lower acceptance of the incident wave at higher angles leads to the smaller measured signal in this region, explaining the related difference between the experimental and theoretical field maps (cf. Fig. 4b with d and Fig. 4c with e). Also, the metallic sphere was held in the setup by a sample holder made of a plastic foam sheet. The sheet is nearly transparent for microwaves (ε~1.1) and, thus, did not essentially influence the results, although its presence can explain minor radial interference fringes observed in the experimental field maps (Fig. 4b, d).
-
These numerical and experimental observations provide strong motivation for the development and application of generalized formulations of the optical theorem27. This can be achieved by considering a relation linking extinction to incident (E0(r), H0(r)) and total (E(r), H(r)) fields valid for any vectorial structure of the field27:
$$ C _{{\rm{ext}}}^{{\rm{GOT}}} = \frac{1}{{\varepsilon _{{d}}\left| {{\bf{E}}_0} \right|^2}} \\ {\Im}\left\{ {\mathop {\int}\limits_V {{{d}}^3{\bf{r}}k\left( {\bf{r}} \right)\left[ {{\bf{E}}_0^ \ast \left( {\bf{r}} \right) \cdot {\bf{E}}\left( {\bf{r}} \right)\eta \left( {\bf{r}} \right) + {\bf{H}}_0^ \ast \left( {\bf{r}} \right) \cdot {\bf{H}}\left( {\bf{r}} \right)\chi \left( {\bf{r}} \right)} \right]} } \right\} $$ (6) where η(r) and χ(r) are the electric and magnetic susceptibilities of the scatterer and the integration is taken over the scatterer volume V. The incident beam in free space is then represented by an arbitrary complex vectorial structure allowed by Maxwell's equations and treated as a superposition of plane waves (generally also allowing the evanescent components) with amplitudes e(k). The extinction cross-section can then be expressed by projecting the amplitudes of the scattered components ${\bf{e}}\left({{\bf{k}}_1} \right){\bf{A}}\left({{\bf{k}}_1, {\bf{k}}_2^ \ast } \right)$ from each of the partial incident waves e(k1) on all other incident components e(k2) and integrating over all possible directional combinations of k1 and k2:
$$ C_{{\rm{ext}}}^{{\rm{GOT}}} = \frac{\omega }{{\varepsilon _{{d}}c\left| {{\bf{E}}_0} \right|^2}} \\ {\Im}\left\{ {{\int} {{{d}}^2k_{1||}{\int} {{{d}}^2k_{2||}e_\beta ^ \ast \left( {{\bf{k}}_2} \right)e_\alpha \left( {{\bf{k}}_1} \right)A_{\alpha \beta }\left( {{\bf{k}}_1,{\bf{k}}_2^ \ast } \right)} } } \right\} $$ (7) Here, $A_{\alpha \beta }\left({{\bf{k}}_1, {\bf{k}}_2^ \ast } \right)$ are the components of the scattering amplitude tensor ${\bf{A}}\left({{\bf{k}}_1, {\bf{k}}_2^ \ast } \right)$, double indices imply summation and || denotes the projection on the z-axis. Applying this approach to the scattering of both linearly and radially polarized light, it was found that the generalized optical theorem (Eqs. (6) and (7)) provides results in excellent agreement with the direct evaluation of the extinction cross-section using Eqs. (3–5) (Fig. 5a, b), revealing its validity even for the vectorial case. We note that in the derivation of the generalized optical theorem, no assumptions regarding the shape of the incident beam and the scattering object were made, e.g., complex fields including evanescent components and objects with an optical response dominated by magnetic dipoles or electric quadrupoles can be considered. The form of the optical theorem given by Eq. (6) is very convenient for approaching the problem with numerical simulations, allowing straightforward integration of the resulting electromagnetic fields. On the other hand, the form given by Eq. (7) is more suitable for analytical and semi-analytical calculations. Finding ${\bf{A}}\left({{\bf{k}}_1, {\bf{k}}_2^ \ast } \right)$ and evaluating the scattering of a plane wave by an object of a given shape are generally within the capabilities of analytical calculations28. Then, the double integration over the wave vectors for simple scatterers can be done analytically; otherwise, numerical evaluation can be performed using standard software.
Fig. 5 Comparison of the conventional and generalised formulations of the optical theorem and numerical modelling for linearly, radially and azimuthally polarized beams.
Extinction cross-section spectra of gold nanoparticles with a radius of 50 nm, calculated using the optical theorem (OT) Eq. (1) (solid red line), generalized OT Eqs. (6, 7) (dashed green line) and direct evaluation of the sum of the absorption Eq. (4) and scattering Eq. (5) cross-sections (black line) for (a) linearly, (b) radially and (c) azimuthaly polarized beamsThe optical theorem was further tested in the case of the scattering of a focused azimuthally polarized beam (with the same parameters as for the radially polarized beam in Fig. 2) on a 100 nm spherical gold nanoparticle. Numerical modeling shows that the generalized version of the optical theorem (Eqs. (6) and (7), dashed green line in Fig. 5c) provides the correct prediction of the value of the extinction cross-section, while the conventional formulation (Eq. (1), solid red line in Fig. 5c) proves to be inadequate. Due to the vectorial structure of the beam signified by the absence of the electric terms in its multipole decomposition29, the electric resonances that play a leading role in the optical response of a spherical plasmonic particle are not excited. This leads to much lower extinction cross-section values compared to linearly and radially polarized excitations (cf. Fig. 5c and Fig. 5a, b). In particular, the peak at the wavelength of 520 nm corresponding to the dipole resonance is no longer present for the azimuthal polarization. The observed small increase in the extinction cross-section toward the shorter wavelengths is related to the increase of the absorption in the metal.