Tunable Fluorescence from Dye-Modified DNA-Assembled Plasmonic Nanocube Arrays

Abstract

Colloidal crystal engineering with DNA on template-confined surfaces is used to prepare arrays of nanocube-based plasmonic antennas and deliberately place dyes with sub-nm precision into their hotspots, on the DNA bonds that confine the cubes to the underlying gold substrate. This combined top-down and bottom-up approach provides independent control over both the plasmonic gap and photonic lattice modes of the surface-confined particle assemblies and allows for the tuning of the interactions between the excited dyes and plasmonically active antennas. Furthermore, the gap mode of the antennas can be modified in situ by utilizing the solvent-dependent structure of the DNA bonds. This is studied by placing two dyes, with different emission wavelengths, under the nanocubes and recording their solvent-dependent emission. It is shown that dye emission not only depends upon the in-plane structure of the antennas but also the size of the gap, which is regulated with solvent. Importantly, this approach allows for the systematic understanding of the relationship between nanoscale architecture and plasmonically coupled dye emission, and points toward the use of colloidal crystal engineering with DNA to create stimuli responsive architectures, which can find use in chemical sensing and tunable light sources.

Plasmonic nanoparticles (NPs)^[1] give rise to strong electric field (E-field) confinements, and therefore, when coupled with nearby emitters such as dye molecules or quantum dots, act as antennas to enhance their emission rates and intensities.^[2,3] These so-called plasmonic nanoantennas can be used in lasing,^[4] molecular sensing,^[5] and quantum information processing.^[6] Although single NPs can function as plasmonic nanoantennas,^[7] typically researchers exploit dimers of NPs,^[8,9] particles coupled with a metallic surface,^[10–12] or arrays of particles lithographically placed into configurations that lead to optimum plasmonic coupling effects.^[9] The nanoscale patch antenna geometry,^[10–12] where an NP is positioned a short distance above a metallic film with a dielectric spacer, gives rise to a plasmonic gap mode that enhances light absorption and subsequent fluorescence. When plasmonic NPs are placed in an array with a periodicity comparable to the wavelength of light, long-range radiative coupling between localized surface plasmons of single NPs give rise to diffraction modes in the plane of the array known as surface lattice resonances (or lattice plasmons), producing even stronger field enhancements and higher quality resonances.^[13,14] Consequently, particle arrays have found use in both lasing^[15] and sensing.^[16,17]

Colloidal crystal engineering with DNA is emerging as a robust method for controlling crystal composition, lattice parameter, and habit.^[18–23] In addition to being useful for preparing 3D assemblies, the technique can be combined with top-down lithography (e.g., e-beam, EBL) to position NPs on substrates, which allows one to synthesize 2D structures not attainable through conventional lithographic techniques.^[24,25] For example, arrays of single crystalline nanocubes that display strong coupling between plasmonic gap modes and photonic lattice modes^[26] have been assembled on gold-coated silicon substrates. Here, we explore how such structures, combined with DNA structural sensitivity to solvent,^[24,27] can be used to modify the emission properties of dyes strategically placed into the E-field hotspots of the particle array structure. These studies reveal four novel and attractive characteristics of such assemblies which enable precise control over structural parameters that affect the plasmonic resonances, coupling strength, and emission enhancement. First, the emission of the dye can be controlled through lattice spacing, which controls the lattice mode resonance. Second, the gap mode resonance, tunable through DNA length and solvent,^[27] can be used to control dye emission. This is important because it allows one to do so without irreversibly changing lattice structure. Third, the position of the dye, which can be controlled through chemical attachment to specific bases, can be fine-tuned with sub-nm precision and used to influence and optimize emission.^[28,29] Finally, using multiple different dyes, emission of each can be regulated through an understanding of these effects and coupling to specific structures. Researchers have preliminarily explored active tunability with plasmonic patch antennas using thermoresponsive polymers as gap materials,^[30,31] but such methods do not enable precise emitter coupling. Methods for coupling dye molecules into antennas include the use of macrocyclic molecules^[32] or DNA origami^[33] to position the dye molecules inside the gaps. These approaches are neither scalable nor allow for active tuning. The method described herein enables precise positioning of the dye molecules in a repeatable and scalable manner, as well as reversible tuning using solvent-responsive DNA.

To explore these structure–function relationships, dye-functionalized gold nanocube arrays were assembled using template-confined DNA-mediated assembly^[24–26] (Figure 1a). First, EBL was used to pattern poly(methyl methacrylate) (PMMA) pores at desired lattice positions on gold-coated Si substrates, with the bottom of each pore consisting of exposed gold. Both the nanocubes and the exposed gold in the pores were densely functionalized with specific sequences of propylthiolated DNA. Dye-labeled linker strands complementary to the substrate strands were then hybridized to the substrates. Note that these DNA strands were synthesized with amino-modified dT groups at the desired base, where dye-labeled NHS esters can be covalently linked. 80 nm cubes were assembled in the pores by designing their single-stranded binding region to be complementary to the binding region of the linkers. Previous work using this technique hybridizes linker strands to both the gold substrate and the NPs and then assembles the NPs within the pores using complementary binding regions on the two linkers.^[24–26] However, in this approach, only a single linker was attached to the substrate, which was directly hybridized to the binding region of the thiolated DNA on the cube to allow one to position the dye near the cube (Figure 1a, top). Cube-functionalized DNA was not directly labeled with dyes because only dyes underneath the cube are inside the plasmonic hotspots. Due to the short length of the DNA strands on the cube (15 bases compared to conventionally used 28 bases^[18,24–26]), assembly was done in aqueous 0.3 M NaCl, rather than 0.5 M, and at lower temperatures, in order to keep the cubes from aggregating. The successful implementation of this new three-strand DNA design, which asymmetrically incorporates dye molecules underneath the nanocubes, highlights that we can change DNA design and assembly conditions to create structures required for a desired optical response not achievable with previously studied designs. After assembly, the PMMA template was removed, and the samples were transferred to 0.5 M NaCl buffer for optical characterization.

Figure 1.

a) Schematic showing dye-functionalized nanocube arrays. Scale bar is 1 μm. Thiolated DNA is functionalized onto the substrate (red) and the nanocube (blue), a complementary linker strand (gray) is hybridized to the substrate DNA, which leaves a single-stranded binding region that is complementary to the nanocube binding region. The dye is attached to the linker strand at one of three binding sites (blue dots, denoted d1, d2, and d3) on the linker strand. b) Schematic of a sample with cubes and a control sample (DNA modified with dye without cubes).

Dye-labeled nanocube arrays with periodicities from 400 to 880 nm in 20 nm increments were assembled onto the same substrate (Figure 2). ATTO 680 was chosen as the dye since it emits at 702 nm, within the range of experimentally accessible lattice and gap modes. A control sample was made by hybridizing dye-labeled linker DNA to the substrate DNA without the final step of assembling nanocubes in order to quantify the enhancement in dye photoluminescence (PL) from the nanocube arrays (Figure 1b). The control sample can also account for differences in dye density between different array periodicities and slight variation of dye emission in different solvents. PL measurements were taken by illuminating the samples with a 633 nm HeNe laser coupled to an inverted microscope and spectrometer. PL maps were collected at 700 nm, near the dye emission maximum (Figure 2c,d). To obtain fluorescence enhancement factors, the fluorescence of the sample was normalized to that of the control (Figure S4, Supporting Information). A significant enhancement was observed (maximum of ≈20), with certain periodicities having larger enhancements due to the presence of lattice or gap modes. Importantly, the unique ability to place dyes only in plasmonic hotspots allowed for large-area characterization, whereas other work on similar geometries only measured PL from under individual nanocubes.^[11,12]

Figure 2.

a,b) Optical images of the control sample (DNA modified with ATTO 680 without cubes) and sample with cubes, respectively. The periods are labeled on each corresponding array. c,d) Corresponding PL maps. Scale bars are 200 μm.

To explain the trends in PL at different periodicities, lattice and gap mode dispersions were measured experimentally at 0% and 60% ethanol in water at constant 0.5 M NaCl and calculated using finite-difference time-domain (FDTD) simulations. To experimentally characterize the interaction of nanocube arrays with dye emission and laser excitation, optical and PL measurements were performed between 500 and 800 nm wavelengths using an inverted optical microscope in reflection mode. In order to correlate ethanol concentrations to gap distances, experimental reflectivity spectra were compared to FDTD spectra at different gap distances. From this comparison, 0% ethanol correlates with an 18 nm gap and 60% ethanol with a 7 nm gap, falling within the range of expected gap sizes^[27] (Figure 3a–d). The lattice modes can be recognized through their dispersive nature, whereas gap modes depend on gap size, which changes with solvent.

Figure 3.

a,b) FDTD simulation of reflection spectra for 7 and 18 nm gap arrays, respectively. The white, dashed lines indicate lattice modes and dotted lines indicate gap modes. c,d) Corresponding experimental reflection spectra at 60% and 0% ethanol, respectively. The reflection minima near 570 nm in all spectra (a–d) are from individual nanocube resonances. e,f) Experimental PL spectra at 60% and 0% ethanol, respectively. g) E-field plot of the excitation enhancement at a 7 nm gap. h) E-field plot of the emission enhancement at a 7 nm gap.

PL intensity was examined as a function of lattice periodicity (Figure 3e,f). At 60% ethanol, the dye emission and laser excitation only overlap with lattice modes, which depend on periodicity. The PL maximum occurs at a periodicity of 560 nm, where the strong enhancement is attributed to coupling between the dye emission and lattice mode around 700 nm. A smaller PL peak occurs for a periodicity of 460 nm and, in this case, is due to coupling between the laser and the lattice mode around 630 nm. These peaks occur at periodicities for which the lattice mode overlaps with either the incident laser or dye emission wavelengths and are known as excitation and emission enhancement, respectively. Although in both cases coupling with lattice modes occurs, enhancement is dependent on the E-field inside the gap, shown in simulations to be higher at the emission energy for 560 nm periodicity than at the excitation energy for 460 nm periodicity (Figure 3g,h). At other periodicities, the lattice mode does not cross with either the laser excitation or dye emission, yet still produce increased PL relative to the reference sample, attributed to nonresonant E-field enhancements inside the gap. At 0% ethanol, PL maxima occur for 400 and 500 nm periodicities. In both cases, emission enhancements are observed, but this time, in the presence of the gap mode occurring near 700 nm. Only a gap mode exists for 400 nm periodicity, but for 500 nm periodicity the dye emission is in the regime of lattice-gap coupling and could be interacting with both modes simultaneously. The ability for two spectrally separate, yet spatially overlapping, tunable modes, to interact with the dye could be used to simultaneously produce excitation and emission enhancements. Furthermore, the gap mode can be tuned in situ by changing solvent polarity (Discussion and Figure S5, Supporting Information).

To study the distance dependence of the plasmon-exciton interactions, PL was measured at three dye-to-cube distances between the bottom of the cube and the film (5, 10, and 15 nm, termed d1, d2, and d3) in 40% ethanol (≈16 nm DNA length) and a 400 nm period array. Here, the dye experiences emission enhancement from a gap mode at 700 nm. We found that the PL intensity decreased from d1 to d2 to d3 (Figure 4a). This trend can be explained by the calculated E-field intensities at the excitation and emission wavelengths (Figure 4b,c). In both cases, the E-field is strongly localized in the gap near the nanocube edges, characteristic of a gap mode, explaining why d1 has the highest PL. When the dye is too close to either the cube or the substrate, the PL intensity could be diminished, due to nonradiative quenching.^[7]

Figure 4.

a) PL for three locations of the dye. In all cases, nanocubes are assembled in a 400 nm lattice and measurements are performed in buffer with 40% ethanol. b) Calculated E-field intensity profile at the dye excitation wavelength (630 nm) for a nanocube separated from a gold film by a 16 nm gap. c) The same calculation done at the dye emission wavelength (700 nm).

Finally, to explore the potential for such lattices in tunable optics, 400 nm period array samples with two dyes under the nanocubes were prepared. These assemblies contained one dye that emits at 609 nm (ATTO Rho101) in the d1 position in addition to the original dye; both dyes are physically in the same proximity to the cubes but randomly mixed within the d1 plane (Figure 5a). Since the dyes have different excitation wavelengths, they can be addressed in a wavelength-dependent manner. Irradiation with 635 nm light (red laser, Figure 5b) leads to single dye emission, while exciting with 543 nm light (green laser, Figure 5c) leads to emission from both dyes. In both cases, one can clearly see that ethanol can contract the DNA and shorten the gap distance, which in turn diminishes the ATTO 680 emission with little effect on that of ATTO Rho101. Indeed, at 0% ethanol, the gap mode is spectrally matched with the emission maximum of ATTO 680 (Figure S5a, Supporting Information), causing enhanced emission near 700 nm, while at 60% ethanol, the gap mode moves away from the dye emission, resulting in little fluorescence. In contrast, the emission of ATTO Rho101 is relatively unaffected by ethanol since it is spectrally mismatched with the gap mode at all ethanol concentrations. It has a large peak near 609 nm, at all ethanol concentrations, because of its inherently high quantum yield coupled to off-resonant enhancement. Therefore, by changing the concentration of ethanol, the intensity ratio of the two dyes can be tuned, yielding a tunable light source.

Figure 5.

Samples prepared with two dyes (ATTO Rho101 and ATTO 680) under the cubes randomly dispersed in the d1 plane. a) Schematic. b,c) PL measurements at varying ethanol concentrations when the sample is excited with a,b) red and c) green laser.

In conclusion, we have shown that template-confined colloidal crystal engineering with DNA can be used to construct plasmonic nanoantennas with individual control over in situ tunable plasmonic gap modes and in-plane photonic lattice modes, both capable of enhancing the emission behavior of fluorophores strategically placed into the antenna hotspots. Additionally, the molecular-level control over the placement of dye molecules afforded by using DNA as linkers allows for fine-tuning of plasmon-exciton interactions. The geometry of the nanoantennas can be optimized in the future for different emission responses by using other anisotropic NPs such as elongated rods and flat disks, or by arranging the NPs in arrays of oligomers. In principle, this novel technique will allow for the fabrication of tunable plasmonic nanoantennas that can be tailored to achieve the desired emission responses in chemical sensing, lasers, and other forms of tunable light sources.