Figure 1

Markers: Re($\varepsilon$) and Im($\varepsilon$) parts for Ag in the 1064- to 400-nm range. The solid, thin black, and dashed green curves correspond to the full Eq. (1) and retrace the data well. The lower thick black curve is only the Re($\varepsilon$) of the Drude portion of Eq. (1). At 450 nm (normalized frequency of 2.5) the discrepancy with the data is substantial. The Im($\varepsilon$) from just the Drude part in Eq. (1) (dashed black curve) still fits the data reasonably well.Reuse & Permissions

Figure 2

Incident TE- and TM-polarized fields. In all our cases the fields overlap but appear spatially separated for clarity.Reuse & Permissions

Figure 3

Conversion efficiency for TM-incident (1064 nm)–TM-detected (532 nm) light for a 125-nm-thick silver layer. The effective mass and densities of both free and bound charges are taken to be $m$${}^{*}$ $=m$${}_e$ and $n$${}_0=5.8\times {10}^{22}$/cm${}^3$, respectively.Reuse & Permissions

Figure 4

(Bottom) Transmission at normal incidence from a five-period Ta${}_2$O${}_5$/Ag multilayer stack, having layer thicknesses of 125 and 20 nm, respectively. (Top) Picture of the incident pulse and the stack.Reuse & Permissions

Figure 5

TM-incident–TM-detected, reflected, and transmitted SH efficiencies, with and without convection. Convection shifts and reduces the amplitude of the reflected signal, while turning the transmission maximum into a minimum. The subscript w/o stands for “without,” while the subscript w stands for “with.” We have assumed that $m^{*}=m_e$ and $n$${}_0=5.8\times {10}^{22}$/cm${}^3$.Reuse & Permissions

Figure 6

TE- or TM-incident and TE-detected, reflected, and transmitted SH conversion efficiencies with (right axis) and without (left axis) convection. Convection shifts the peaks to larger angles but reduces both signals by nearly a factor of 50. The subscript woc stands for “without convection” and wc stands for “with convection.” $m^{*}=m_e$ and $n$${}_0=5.8\times {10}^{22}$/cm${}^3$.Reuse & Permissions

Figure 7

(Right) A TM-polarized 20-fs pump pulse tuned to 800 nm is incident from the left on a silver nanopillar 200 nm on each side. All detected fields are TM polarized. The snapshots are recorded when the peak of the pump pulse reaches the object. The spatial extension of the pulse is hundreds of times larger than the object, so this snapshot mimics what one might expect for plane-wave illumination and steady-state conditions. [(a) and (c)] TM-polarized SH field patterns for free charges only (a) and free and bound charges simultaneously (c). [(b) and (d)] Same as (a) and (c) for the TH fields. Qualitative differences are especially evident for the TH field, where bound charges tend to reinforce the main lobe that points directly backward. In all cases quantitative differences can be read on the side scales. Bound charges enhance SH conversion by a factor of $~2$. TH generation is reduced by approximately the same factor. We have assumed $m^{*}=m_e$ and $n_0=5.8\times {10}^{22}$/cm${}^3$. (Left) Depiction of the setup and pump magnetic field intensity profile recorded at the same time as the harmonics.Reuse & Permissions

Figure 8

(Right) Same as Fig. 7, except for TE-polarized incident pump pulse. The snapshots are recorded when the peak of the pump pulse reaches the metal object. Unlike in Fig. 7, the generated SH signals are TM polarized (the $H$ field is shown for convenience), while the TH fields are TE polarized (the $E$ field is shown for convenience). The TM-polarized SH field patterns from free only (a) and from free and bound (c) charges. TE-polarized TH field patterns obtained from free only (b) and free and bound (e) charges. Qualitative differences are visible in both SH and TH patterns. The polarization of a resulting harmonic field under specific pumping conditions may be assessed by a full decomposition of the nonlinear sources in Eqs. (12) and (29). (Left) Depiction of the set up and electric field intensity profile recorded at the same time as the harmonics.Reuse & Permissions

Figure 9

(Left) Wave incident on a GaAs-filled nanocavity 100 nm deep and 60 nm wide on an infinite grating. Departure from these parameters reduces transmittance and field amplification inside the cavity. (Right) Transmittance vs. wavelength, for a pitch of 570 nm, normalized to the energy that falls directly on the cavity area. Actual maximum total transmittance at the 1064-nm Fabry-Perot resonance is $~30\%$. A barely visible plasmonic transmission peak appears near 570 nm (indicated by the arrow).Reuse & Permissions

Figure 10

Assessing the effects of electron gas pressure. Top (bottom) Snapshot of transverse (longitudinal) pressure compared to its driving field term. The scales show that the transverse pressure is spatially modulated with amplitudes of order $1\%$–$2\%$ compared to the driving field term. The longitudinal pressure is modulated along its entire length but is strongest at the four corners of the cavity. Note: negative pressure simply indicates regions of pressure lower compared to contiguous regions.Reuse & Permissions

Figure 11

SH conversion efficiency under various pumping conditions. The data are taken from Ref. [71], where the average peak intensity is estimated at $~80$ MW/cm${}^2$. The shapes of theoretical curves agree well with the experimental data in all cases. If we choose $m^{*}=0.5$ m${}_e$, and $n_0=3\times {10}^{22}$/cm${}^3$, for both free and bound charges, excellent quantitative agreement emerges with the following choice of incident peak intensities: (a) $~60$ MW/cm${}^2$; (b) $~100$ MW/cm${}^2$; (c) $~300$ MW/cm${}^2$.Reuse & Permissions