Mirror-Enhanced Plasmonic Nanoaperture for Ultrahigh Optical Force Generation with Minimal Heat Generation

Abstract

Double Nanohole Plasmonic Tweezers (DNH) have emerged as a powerful approach for confining light to sub-wavelength volume, enabling the trapping of nanoscale particles much smaller than the wavelength of light. However, to circumvent plasmonic heating effects, DNH tweezers are typically operated off-resonance, resulting in reduced optical forces and field enhancements. In this study, we introduce a novel DNH design with a reflector layer, enabling on-resonance illumination while minimizing plasmonic heating. This design efficiently dissipates heat and redistributes the electromagnetic hotspots, making them more accessible for trapping nanoscale particles and enhancing light-matter interactions. We also demonstrate low-power trapping and release of small extracellular vesicles. Our work opens new possibilities for trapping-assisted Surface Enhanced Raman Spectroscopy (SERS), plasmon-enhanced imaging, and single photon emission applications that demand strong light-matter interactions.

Graphical Abstract

The ability to precisely trap and manipulate micron/nanoscopic particles using optical tweezers has significantly impacted the field of particle manipulation and has found applications toward trapping and manipulating a wide range of materials, ranging from inorganic dielectric particles,^1,2 to metal nanoparticles^3–5 and carbon nanotubes.⁶ In biological research, optical tweezers have enabled several breakthroughs including, crucial nanoscale force measurements to determine the elastic properties of DNA strands,⁷ determining the refractive index of viruses using well calibrated optical traps,⁸ and in situ single particle Raman spectroscopy via Raman tweezers microspectroscopy.^9,10 Despite the wide range of applications enabled by optical tweezers, the trapping and manipulation of nanoscopic particles requires very high laser powers due to the inherent diffraction limitation of these systems, which limits its applications in biological sciences to larger biospecies like cells. The need for higher laser powers for nanoscale trapping is a consequence of the size-dependent nature of the optical gradient force, which necessitates stronger intensity gradients to trap smaller particles. By leveraging the highly intense electromagnetic hotspots created through the localized surface plasmon resonance (LSPR) of metallic nanostructures, plasmonic nanotweezers enable low-power trapping well beyond the diffraction limit, with ultranarrow potential widths that ensure single particle trapping.^11–17 Although nanoscopic trapping beyond the diffraction limit has been demonstrated using plasmonic nanoantennas, positive thermophoretic effects arise due to the inherent nonradiative losses in metals, which decrease the stability of the trap.^11,17–19 To circumvent unwanted heating effects in plasmonic antennas, the Crozier group demonstrated the trapping of a 110 nm polystyrene particle using a resonance-tuned plasmonic nanopillar coupled with an integrated heat sink that comprised a stack of high thermal conductivity films serving as the substrate.²⁰ In contrast to plasmonic antennas, the use of nanoapertures in metallic films for nanoscale trapping effectively reduces plasmonic heating especially when illuminated off-resonance, as the continuous metallic film facilitates efficient heat dissipation.²¹ Furthermore, the phenomenon of self-induced back action in nanopores, wherein the presence of a particle in a nanoaperture alters the surrounding electromagnetic field, thereby enhancing the trapping mechanism^11,22 makes them more suited for nanoscopic trapping. Double nanohole (DNH) apertures are particularly interesting owing to the highly enhanced electromagnetic hotspots in the “gap” of the nanoaperture.^{14,21,23–25} This presents exciting opportunities in biological sciences to trap and study single nanoscopic bio species. Pang et al.²⁶ trapped and observed the folding and unfolding of a single Bovine serum albumin (BSA) protein using a double nanohole aperture. Trapping of ~20 nm^14,21,24 and ~12 nm²³ dielectric beads have been demonstrated by various groups using DNH apertures in gold films. Despite the success of these nanoapertures, it is important to note that these systems are usually detuned from resonance to avoid substantial heat generation due to the increased absorption of gold at longer wavelengths and the low thermal conductivity of the underlying glass substrate. Resonance-detuned operation results in less intense electromagnetic hotspots, which implies that these destined systems are not fully leveraging the benefits of the plasmonic aperture to maximize the gradient force and enhance spectroscopy of trapped specimens. Finite difference time domain (FDTD) simulations of the electromagnetic field distribution around DNH apertures show that the resonance-detuned hotspots are mostly localized in the glass substrate,²⁵ leaving the hotspots less accessible for applications requiring strong light–matter interaction like surface enhanced Raman spectroscopy (SERS). This is due to the higher photon density of states in the glass substrate relative to that of the water supernatant. The trapping of single small extracellular vesicles (e.g., exosomes) holds great potential for investigating the heterogeneity within specific sub populations of these bio species, leveraging the intense electromagnetic hotspots for enhanced light–matter interaction. In this work, we present a novel DNH aperture design that simultaneously addresses the issue of less accessible hotspots and significant plasmonic heating for resonance tuned DNH apertures. To address these issues, we incorporate an integrated reflector layer into the DNH configuration. Additionally, we experimentally demonstrate low-power single small extracellular vesicle (EV) trapping using this novel DNH design.

The integration of a reflector layer directly beneath the Au film results in a redistribution of the electromagnetic hotspots around the top of the DNH gap. This redistribution of the electromagnetic field makes the hotspots more accessible for applications such as SERS, which rely on an increased light–matter interaction. Efficient heat dissipation away from the hotspots is achieved in-plane via the continuous Au film and axially via the high thermal conductivity substrate, composed of the reflector layer deposited on a sapphire substrate with a higher thermal conductivity compared to the conventionally used silica substrate. The schematic design of this system is shown in Figure 1(a),(b), with the left panel of Figure 1(b) showing the SEM micrographs of DNH on a reflector system. The redistribution of the electromagnetic field due to the integration of the reflector layer results from Maxwell’s boundary condition for metals, which requires the decay of electric fields within the metal at the wavelengths of interest. To investigate the properties of this new design, we first theoretically consider the conventional DNH configuration, an aperture in a 100 nm thick Au film on a glass (silica) substrate. For this system, $d$ (diameter of each circle) is taken to be 100 nm, while cc (which is the center-to-center spacing of the circles making the DNH) is taken as 140 nm. We have considered the effect of the top gap widening caused by focused ion beam milling during sample fabrication. The top DNH gap $g^1$ is taken to be 45 nm, while the bottom gap $g^2$ is taken as 30 nm for all simulations presented in this work. The incident electric field is polarized along the $X$ axis, as depicted in Figure 1(b). The electromagnetic field enhancement distribution in the $ZX$ plane at $Y=0$ is shown for both resonance tuned and detuned scenarios in Figure 2(a),(b). The resonance detuned field enhancement is shown for a wavelength of 973 nm which is the experimental trapping wavelength used throughout this work, while the on-resonance wavelength shown in Figure 2(b) is 1265 nm.

Figure 1.

Schematic design for the double nanohole (DNH) on a reflector system. (a) Schematic design showing the integration of a Cr–Au hybrid reflector layer underneath the gold film. The Cr layer was selected for ease of fabrication and serves as a mask to prevent excessive drilling of the gold film layer during focused ion beam milling. (b) $ZX$ and $XY$ cross sections for the DNH illuminated under $X$ polarized light. The widening of the DNH gap due to the focused ion beam milling fabrication process is shown as $g^1$, while $g^2$ is the bottom gap of the DNH. $d$ and cc represent the diameter and center-to-center distance of the nanoholes. The right panel shows scanning electron micrographs (SEM) of the DNH.

Figure 2.

Electromagnetic and thermal simulations for the DNH on glass configuration. (a) FDTD Electric field enhancement for the resonance detuned DNH on glass system in the $ZX$ plane at $Y=0$, illuminated at 973 nm, while (b) shows the resonance tuned electric field enhancement distribution at the resonant wavelength of 1265 nm in the $ZX$ plane at $Y=0$. The nanohole diameter $d$ is taken to be 100 nm, the center-to-center distance cc between the nanoholes is taken as 140 nm, and $h$ is taken as 100 nm. (c) and (d) show the corresponding FEM thermal simulations for (a) and (b), respectively, in the $XY$ plane at $Z=50\mathrm{nm}$.

For both resonance-tuned and -detuned scenarios, the structure parameters remain constant, while only the illumination wavelength differs. The confinement of the hotspots in the glass substrate can be seen in Figure 2(a), (b), with a maximum field enhancement of 13 for the resonance-detuned configuration as opposed to a maximum field enhancement of 31 for the resonance-tuned configuration. This shows a reduction in the intensity enhancement by over a factor of 5 resulting from resonance detuning of the aperture. Furthermore, we carried out finite element method (FEM) thermal simulations for both scenarios to investigate the plasmonic heating effect, as shown in Figure 2(c),(d). The resonance-detuned scenario in Figure 2(c) shows a temperature rise of up to 15 K, while the resonance-tuned scenario shows a 30 K rise in temperature (Figure 2(d)). This appreciable temperature increase at resonant illumination could give rise to thermal effects like thermophoresis, which could have a negative impact on the trap stability, making resonant illumination less desirable for this system. Next, we theoretically investigate the resonance-tuned DNH on a gold reflector configuration shown in Figure 3(a). For this configuration, we first consider a DNH aperture in a 100 nm thick Au film, on a 150 nm Au reflector with a sapphire substrate. The combination of Au with a thermal conductivity of 317 W/mK and sapphire with a thermal conductivity of 35 W/mK results in efficient heat dissipation in the in-plane and axial directions, giving rise to a minimal plasmonic heating effect for resonance-tuned illumination. For this configuration, the nanohole diameter $d$ is taken to be 170 nm, while the center-to-center distance cc is 200 nm. The gap parameters, $g^1$ and $g^2$, are taken as 45 and 30 nm, respectively, while the incident light is polarized along the $X$ axis. The resonance-tuned electric field distribution in the $ZX$ plane at $Y=0$ is shown in Figure 3(a). The redistribution of the electromagnetic field is apparent from the electric field enhancement distribution, showing more accessible hotspots. Additionally, we show that the maximum achievable electric field enhancement on resonance can be enhanced with the reflector layer by a factor of approximately 1.5 compared to the resonance-tuned DNH on a glass configuration (Figure 2(b)). This results from the interference between the reflected light and the gap plasmon modes in the nanogap. Additionally, this design allows for more predictable trapping by reducing the number of hotspots from 4 as shown in Figure 2(b) to 2 as evident in Figure 3(a). To investigate the temperature increase for this system, FEM thermal simulations were carried out as shown in Figure 3(b). This shows an approximately 6.3 K rise in temperature, which is less than that of the resonance-tuned DNH on glass by a factor of 5, despite having the most intense electric field hotspots. This is attributed to the efficient thermal engineering of the substrate. The DNH design with the integrated reflector layer not only provides more accessible hotspots but also significantly minimized the temperature rise. To experimentally demonstrate nanoparticle trapping using the optimized DNH configuration, an important design consideration for the DNH reflector would be to include a hard mask layer to enable more accurate control of the nanohole depth during the focused ion beam milling process which is employed in this work. For this reason, we simulated the electromagnetic field distribution for the DNH on a reflector system, replacing the 150 nm Au reflector layer with a resonance-tuned hybrid Cr–Au reflector layer consisting of a 50 nm Cr hard mask layer on a 100 nm Au reflector. The diameter $d$ is given as 190 nm, while the center-to-center distance cc is 260 nm. However, it is important to note that the maximum resonance-tuned electric field enhancement for this hybrid reflector configuration is less than that of the Au reflector, as shown in Figure 3(c). This can be attributed to the damping of the LSPR resonance in the Au film by the Cr layer. Haykel et al.²⁷ showed that the incorporation of a Cr adhesion layer within a plasmonic aperture leads to the attenuation and broadening of the LSPR resonance. Moreover, they found that an increased thickness of the adhesion layer corresponds to greater damping effects. To investigate the impact of the Cr layer thickness on resonance damping for the design presented in this work, we carried out electromagnetic simulations, as illustrated in Supplementary Figure 3. Our findings revealed a progressive rise in damping as the Cr thickness increased. Nevertheless, it is essential to consider the Cr layer’s efficacy as a robust hard mask for the FIB fabrication approach. We observed that a greater thickness permits improved control of the aperture’s depth. Through a series of iterative fabrication processes, we have ascertained that a Cr thickness of 20 nm would suffice to serve as a good hard mask while supporting a good field enhancement. However, it is important to note that the extent of damping plateaus for Cr layers exceeds 20 nm. The FEM thermal simulation for the hybrid reflector system in the $XY$ plane at $Z=50\mathrm{nm}$ is presented in Figure 3(d). This shows a maximum temperature increase of 10 K, which is higher than that obtained for the system with a uniform Au reflector. The higher temperature increase is attributed to the lower thermal conductivity of the Cr hard mask layer of 93 W/mK. This is shown in Supplementary Figure 1. To determine the viability of the DNH on a Cr–Au hybrid reflector for the near-field trapping of small EVs, we calculated the optical gradient force and potential. Specifically, we focused on a small EV (100 nm in size) and employed the Maxwell Stress Tensor method defined as $F=\int {〈T〉}_t\cdot \widehat{n}\mathrm{d}S$, where $F$ is the optical force, ${〈T〉}_t$ is the time averaged Maxwell’s Stress Tensor defined as $T=\epsilon \left(EE-\frac{1}{2}{\left|E\right|}^{2}I\right)+\mu \left(HH-\frac{1}{2}{\left|H\right|}^{2}I\right)$ where $\epsilon$ and $\mu$ are the permittivity and permeability of the surrounding medium respectively, while $E$ and $H$ are the electric and magnetic fields, respectively. $\widehat{n}$ is the surface normal and $\mathrm{d}S$ is a bounding surface enclosing the particle. To determine the Maxwell Stress tensor, initial electromagnetic simulations were carried out, taking the refractive index of water as 1.33 and that of the exosome as 1.39.²⁸ The particle center was maintained at $Z=155\mathrm{nm}$, while the $Y$ and $X$ components of the optical force and their associated potentials were computed by translating the particle along the $Y$ (maintaining $X=0$) and $X$ axis (maintaining $Y=0$). Our analysis revealed that the optical potential obtained for the small EV was substantial. Figure 4(e),(f) showcases the results, demonstrating the optical potential energy exceeding $2.5K_{B}T$ in both the $Y$ and $X$ directions, respectively. These findings are significant as they indicate that the generated optical potential possesses sufficient energy to surpass the Brownian motion exhibited by the small EV. In other words, the trapping mechanism can effectively counteract the random thermal motion of the particle, enabling stable confinement within the trap. Furthermore, the effect of thermophoresis was studied theoretically for the DNH on a Cr–Au hybrid reflector as modeled in the SI. We calculated the thermophoretic force and potential considering the temperature rise 5 nm above the Au film (at $Z=105\mathrm{nm}$). For this analysis, we have used the thermal properties of lipid vesicles as reported in literature^29,30 to model small EVs. The data presented in Supplementary Figure 2 show that the thermophoretic potential for a 100 nm exosome is significantly less than the optical potential, resulting in minimal thermal effects for trapped exosomes. The coupling of the reflected light to the gap plasmon mode for the DNH on a reflector configuration allows for easy tunability of the resonance for a given gap size by varying the thickness of the Au film, as opposed to the conventional DNH on glass configuration, which is more stringent with height variations, as shown in Figure 4(a),(b). This expands the parameter space for DNH plasmonic tweezers, making it easier to realize more efficient coupling to a 785 nm Raman excitation laser for SERS experiments (see Supplementary Figure 4). Working at shorter wavelengths at resonance has the advantage of using silicon photodetectors. In contrast, when working at the longer resonance wavelength in the DNH on the glass configuration, silicon photodetectors have lower responsivity. The field enhancement spectra for both the DNH on a reflector and glass configurations are presented in Figure 4(c),(d), showing the resonance and illumination wavelengths. The disparity between the resonance and operation wavelengths commonly used in literature^21,24 for the DNH on glass system is clearly shown in Figure 4(d), underscoring the limitations associated with this off-resonance system. Additionally, Figure 4(b) shows limited tunability for this system beyond 1200 nm, for a 30 nm gap size.

Figure 3.

Electromagnetic and thermal simulations for the DNH on a reflector configuration. (a) Electric field enhancement in the $ZX$ plane for a DNH with an integrated 150 nm Au reflector layer, at $Y=0$ (center of the DNH), tuned to be resonant at an incident wavelength of 973 nm. The top and bottom gaps $g^1$ and $g^2$ are taken to be 45 and 30 nm respectively. The nanohole diameter $d$ is taken to be 170 nm, the center-to-center distance cc between the nanoholes is taken as 200 nm while the Au film thickness $h$ is taken as 100 nm. (b) Corresponding FEM thermal simulations for (a) in the $XY$ plane at $Z=50\mathrm{nm}$. (c) Field enhancement distribution for a DNH in a 100 nm Au film with a hybrid Cr–Au reflector layer tuned to the same resonance wavelength as (a). $g^1$ and $g^2$ are kept constant while the nanohole diameter $d$ is taken to be 190 nm, and the center-to-center distance cc between the nanoholes is taken as 260 nm. The hybrid reflector comprises a 50 nm Cr layer and a 100 nm Au layer. The yellow and white dotted lines show the distinction between various substrate materials. (d) Corresponding FEM thermal simulations for (c) for the same plane as (b). (e) and (f) show the $Y$ and $X$ components of the optical force, respectively (orange curves), exerted on a 100 nm small EV, obtained using the Maxwell stress tensor (MST) method. The optical force calculations were performed by varying the particle center along the $Y=0$ and $X=0$ lines, respectively, while maintaining the particle center position at $Z=155\mathrm{nm}$. The corresponding optical potentials are shown as the blue dotted curves.

Figure 4.

Parametric tuning for the DNH plasmonic cavity. (a) shows the parametric space of the Au film thickness h, for the DNH on a Au reflector while (b) shows the corresponding data for the DNH on glass configuration. (c) Electric field enhancement at $X=0$, $Y=0$, $Z=100\mathrm{nm}$ for the DNH with a gold reflector (blue curve) and the DNH with a Cr–Au reflector (orange curve). (d) Electric field enhancement at $X=0$, $Y=0$, $Z=0$ nm for the DNH on glass configuration showing the resonance wavelength (black dotted line) and its operation wavelength of 1064 nm (blue dotted line), commonly used in literature.^21,24

To experimentally demonstrate trapping using the DNH on a reflector configuration, a 10 nm Cr adhesion layer was first deposited on a sapphire substrate via resistive deposition, followed by electron beam (e-beam) evaporation of a 100 nm Au film. Next, the 50 nm Cr hard mask was deposited resistively followed by a subsequent e-beam evaporation of a 100 nm Au film. All processes were run in a multimode deposition chamber without breaking the vacuum. To define DNH on the Au film, we used a gallium source focused ion beam with a 1.1 pA beam current that ablates the Au film much more efficiently compared with the Cr hard mask layer. Next, we treated the DNH sample using Poly(sodium 4-styrenesulfonate) (PSS) for 10 min, followed by a 5 min rinse with potassium chloride (KCl). This process aimed to prevent the adsorption of negatively charged small EVs to the Au surface. PSS has been shown to passivate Au and quantum dot surfaces with negative charge.^31,32 Afterward, the sample was blow-dried with air. Next, the DNH sample was made into a microfluidic chip, utilizing 120 μm thick dielectric spacers and an ITO-coated coverslip to close the channel. Note that both the DNH sample and the ITO-coated coverslip underwent the same surface treatment. A 973 nm diode laser was used for trapping experiments. To ensure proper alignment, the polarization of the laser was oriented perpendicular to the center-to-center axis of the DNH, using alignment markers milled in the Au film. The laser was focused to achieve a spot size of 1.33 μm, utilizing a 40× 0.75 N.A. objective lens mounted on an inverted microscope. Commercially purchased neon green conjugated small EVs were suspended in deionized water and injected into a microfluidic chamber. Using a 15.4 mW laser, we demonstrated trapping of small EVs as shown in Figure 5. Frames 1 and 2 show a single small EV diffusing in the vicinity of the hotspot, while frame 3 shows the particle trapped in place at the plasmonic hotspot. Using a CCD camera operating at a 200 ms exposure time, we tracked the particle position of the fluorescently labeled small EV in using a custom python script. Supplementary Movie 1 shows a small EV trapped under a 15.4 mW laser, while Supplementary Movies 2 and 3 show the trapping and release of an exosome. The corresponding tracked particle canter displacement data over a duration of 30 s in Supplementary Movie 1 is given in frame 4 of Figure 5. Histograms for the particle center displacement along the $Y$ and $X$ axes are shown in Figure 5(b), (c). To calculate the trap stiffness, we first obtain the experimentally measured variance $Va{r}_m\left(i\right)$ from the particle position displacement presented in frame 4 of Figure 5(a). Using the equipartition theorem, the variance along the ith axis is given by $Var\left(i\right)=\frac{K_{\mathrm{B}}T}{k_i}$, where $K_{\mathrm{B}}$ is the Boltzmann constant, $T$ is the temperature, $Var\left(i\right)$ is the variance, and $k_i$ is the stiffness along the ith axis. Using this approach, the $X$ and $Y$ trapping stiffnesses were determined to be $k_x=1.1820\mathrm{fn}/\text{nm}$ and $k_y=0.8012\mathrm{fn}/\text{nm}$.

Figure 5.

Trapping results for small EVs. (a) Experimental trapping video showing the diffusion of a small EV in the vicinity of the DNH and subsequent trapping (frames 1–3), with frame 4 shown the trapped particle displacement scatter plot, for a 15.4 mW 973 nm trapping laser. (b) and (c) show the particle displacement histograms along the $X$ and $Y$ axes respectively for the data in frame 4 of panel (a), with $X_c$ and $Y_c$ representing the center of the fluorescently labeled small EV.

To validate the performance of the DNH on a reflector configuration, we have also trapped 100 nm polystyrene beads using the DNH on a reflector and the DNH on glass configurations and obtained the normalized trap stiffnesses. We show trapping of a 100 nm polystyrene bead with the reflector configuration in Supplementary Movie 4, while Supplementary Movies 5 and 6 show the particles trapped in place using the DNH on a reflector and DNH on glass configurations, respectively. From this data, we obtained normalized trap stiffnesses of ($K_x=0.1\frac{\mathrm{fN}}{\mathrm{mW}\cdot \text{nm}}$, $K_y=0.067\frac{\mathrm{fN}}{\mathrm{mW}\cdot \text{nm}}$) for the reflector configuration and ($K_x=0.04\frac{\mathrm{fN}}{\mathrm{mW}\cdot \text{nm}}$, $K_y=0.047\frac{\mathrm{fN}}{\mathrm{mW}\cdot \text{nm}}$) for the DNH on glass configuration, thereby validating the DNH on a reflector system as a more optimized plasmonic cavity. Last, we calculated the optical force on sEVs of varying sizes using the Maxwell Stress Tensor method keeping the bottom surface of the particle 5 nm from the surface of the Au film. As illustrated in Supplementary Figure 5, it is evident that the magnitude of the optical force increases with larger particle sizes. This observation diverges from the findings reported in previous literature,³³ where the use of a DNH on a glass substrate resulted in decreased trap stability as particle size increased. This decrease is attributed to the inaccessibility of the hotspots to larger particle sizes for the conventional double nanohole aperture on glass substrate systems. In our proposed design, the redistribution of the hotspot creates more accessible regions for particles, even those of larger sizes, to maintain some degree of overlap with the hotspot.

The integration of a reflector layer in the DNH aperture design offers significant advantages for nanoscale trapping and manipulation. The incorporation of the reflector layer redistributes the electromagnetic hotspots, making them more accessible for applications such as SERS that require a strong light–matter interaction. Additionally, the reflector layer enables efficient heat dissipation, further minimizing plasmonic heating effects. The optimized DNH configuration demonstrates low-power trapping of small extracellular vesicles (exosomes) with improved field enhancement and reduced temperature rise. The experimental demonstration of particle trapping shown here using the DNH reflector system validates its effectiveness, expanding the possibilities for more efficient light–matter interaction applications in areas such as biological sciences and SERS. The ability to trap single small extracellular vesicles using the highly intense electromagnetic hotspots of the DNH aperture holds great promise for single small EV analysis using Raman spectroscopy. In recent studies, the potential of small EVs has been highlighted as a biomarker for the detection of cancer. However, the practical application of small EVs in clinical settings for early cancer detection is limited by the considerable heterogeneity observed among these biological species.^34–36 The optimized DNH on a reflector design proposed in this work allows for more accessible hotspots with minimal heating effects, which can facilitate the investigation of the heterogeneity exhibited by these biological species at the individual particle level via SERS, thereby paving the way for a more comprehensive understanding of their characteristics.