Tom
Veeken
*a,
Benjamin
Daiber
a,
Harshal
Agrawal
a,
Mark
Aarts
a,
Esther
Alarcón-Lladó
a,
Erik C.
Garnett
a,
Bruno
Ehrler
a,
Jorik
van de Groep
b and
Albert
Polman
*a
aCenter for Nanophotonics, NWO-Institute, AMOLF, Science Park 104, 1098 XG Amsterdam, The Netherlands. E-mail: t.veeken@amolf.nl
bInstitute of Physics, University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands
First published on 7th January 2022
We present a soft-stamping method to selectively print a homogenous layer of CdSeTe/ZnS core–shell quantum dots (QDs) on top of an array of Si nanocylinders with Mie-type resonant modes. Using this new method, we gain accurate control of the quantum dot's angular emission through engineered coupling of the QDs to these resonant modes. Using numerical simulations we show that the emission into or away from the Si substrate can be precisely controlled by the QD position on the nanocylinder. QDs centered on a 400 nm diameter nanocylinder surface show 98% emission directionality into the Si substrate. Alternatively, for homogenous ensembles placed over the nanocylinder top-surface, the upward emission is enhanced 10-fold for 150 nm diameter cylinders. Experimental PL intensity measurements corroborate the simulated trends with cylinder diameter. PL lifetime measurements reflect well the variations of the local density of states at the QD position due to coupling to the resonant cylinders. These results demonstrate that the soft imprint technique provides a unique manner to directly integrate optical emitters with a wide range of nanophotonic geometries, with potential applications in LEDs, luminescent solar concentrators, and up- and down-conversion schemes for improved photovoltaics.
Resonant nanostructures can help tailor the emission of dipole-like point emitters by controlling the coupling between the resonant modes and the emitter. The spectrum, polarization, and angular distribution of the emission are then determined by the coherent superposition of the scattered fields of the electric and magnetic multipoles and their coupling to the emitter's dipole moment. Initial work in this area focused on coupling noble-metal nanoparticles to optical emitters, where the spectrum and polarization were controlled by coupling emitters to selected plasmonic modes11,12 and directional emission was achieved with nanoparticle antenna's.13–16 More recently, all-dielectric resonant nanostructures have received great interest because of their strongly reduced optical losses and the larger variation of multipoles that can be excited.17–20 This offers more degrees of freedom to design the resonant interaction with the emitter. Moreover, optical emitters can be placed directly inside the resonant nanostructures.4,21Vice versa, nanostructures can be directly placed on top of emitters, as recently shown for resonant silicon (Si) nanowires on monolayers of MoS2,22 to create directional forward and backward emission depending on the complex interplay of the nanowire resonances. In these pioneering first experiments, there was only limited control on the exact placement of emitter with respect to the nanoresonator.23–25 However, to leverage all benefits of the dielectric resonator–emitter coupling, precise control over the placement and coupling between the emitter and resonant nanostructure is of great importance. Selective coating of nanostructures with uniform monolayers of emitters in particular would be highly desirable for applications in e.g. LEDs and luminescent concentrators.
Here, we introduce a new method to achieve selective control over the placement of optical emitters on resonant nanostructures using a soft stamping technique.26 As a demonstration, we fabricate arrays of Si nanocylinders that exhibit strong optical Mie resonances using standard electron-beam lithography (EBL) and reactive-ion etching (RIE) techniques. Then, using a PDMS rubber stamp we selectively place luminescent quantum dots (QDs) on top of nanophotonic structures only in a controlled manner. By directly spin-coating the QDs on the PDMS stamp, we realize a single-step stamping process that is simplified compared to existing pick-and-place stamping techniques.27–30 Using photoluminescence (PL) mapping spectroscopy and lifetime measurements in combination with numerical modeling, we show evidence of strong directional emission of the QDs coupled to the nanoresonators. Our results contribute towards novel routes for improved efficiencies of quantum dot applications in LEDs, wide spectrum emission, and enable up and down-conversion schemes in photovoltaics. The resonant directional light emission demonstrated in this work is generic and can be applied to a wide range of emitters, including semiconductor quantum wells, fluorescent molecules, and perovskite films.
To control the emission anisotropy, high-index resonant nanostructures are placed in the near field of the emitter, as depicted in Fig. 1b. Silicon structures with dimensions in the few-100 nm range support an electric dipole (ED), magnetic dipole (MD), and higher-order multipoles.17,34 The coupling of the emitter to these multipolar modes alters the angular emission pattern through interference of the scattered light in the far-field. The shape and size of the resonator determines the strength of the multipolar modes at the emission wavelength, and as such, acts as a control for the direction of emission. We first use Finite-Difference Time-Domain (FDTD) simulations to calculate the coupling between the emitter and the nanostructure, using known optical constants for Si and a dipole emitting at λ = 760 nm placed 5 nm above the center of a Si cylinder. In Fig. 1e, the fraction of radiation emitted into the top hemisphere is plotted as a function of Si cylinder diameter for different heights (see Fig. S1† for corresponding emission profiles). The upward emission is defined as the total radiation emitted into the upper hemisphere, by a single dipole on top of a single cylinder. The dashed line in Fig. 1e indicates the reference value of 6.5% for the case of a substrate without resonator. We find that for small cylinder diameters between 50 and 250 nm, the upward radiation away from the substrate is enhanced, whereas the downward radiation is enhanced in the range of 250–500 nm. Beyond a diameter of 500 nm, the upward emission approaches the reference; here, the Si cylinder has many resonant modes at the emitter wavelength, but none with significant strength, such that the coupling is similar to that for a planar film. The maximum fraction of downward radiation is achieved for a cylinder diameter of 400 nm and height of 125 nm, at a value of 98%. The corresponding azimuthal angular emission profile is shown in Fig. 1d.
To assess which Mie-like multipolar modes are excited in the Si cylinder, we calculate the normalized scattering cross-section Qscat using FDTD simulations, defined as the scattering cross-section normalized to geometrical cross-section for a normal-incident plane wave.17Fig. 1f shows Qscat as a function of wavelength and cylinder diameter. Many Mie-like resonances are visible for the cylinder without substrate (top), some of which can be attributed to single multipole resonances such as the ED, MD, toroidal dipole (TD), electric quadrupole (EQ), and magnetic quadrupole (MQ). When the substrate is introduced under the cylinder (Fig. 1f, bottom), the lineshape of the multipolar resonances broadens significantly due to strong radiative leakage from the resonant mode into the substrate. However, the Qscat remains well above unity, indicating the strong resonant character of the cylinder.
The broad range of Mie-like modes in a Si cylinder is shown in Fig. 1g, for the cylinder with maximum emission downwards but without the substrate. To obtain insight in which modes contribute to the strongly directional radiation pattern, we use a multipole decomposition to extract the relative contribution of a set of multipolar resonances: ED and toroidal dipole (TD), MD, EQ, MQ, and the electric octupole (EOC).34 Clearly, all five components attribute to the Qscat over a broad range of wavelengths, and notably quite evenly at the target wavelength of 760 nm. Therefore, we attribute the enhanced downwards emission to a combination of multipolar modes in the cylinder and direct emission of the dipole into the far-field. Note that the multipolar resonances we find here are those that can be excited by the normal-incident plane wave. A dipole-like point emitter placed in the near-field of the nanocylinder can couple to the resonant modes with different relative amplitudes, and also excite modes with symmetric modal field profiles that cannot be excited a plane wave. Despite this, the multipole analysis provides valuable insight in the complex combination of multipolar resonances that collectively give rise to strong broadband light scattering.
So far, we analyzed the emission of a dipole placed at the center of the cylinder's top surface. Now, we investigate the angular emission for an ensemble of dipoles homogeneously distributed over the nanocylinder surface. We use FDTD to calculate the angular emission distribution for different radial positions on the nanocylinder surface, as shown in Fig. 1h for the cylinder diameter of 400 nm (which showed the highest downward emission in Fig. 1e). Subsequently, we average these simulation results, weighted by their radial area, to determine the upward emission fraction for the ensemble, which is shown in Fig. 1i. The result shows that the average upward directionality is always higher than that for the flat reference, i.e., the upward emission can be tuned. Similar to the calculation for the single emitter at the center, the curves converge to the reference for large diameter, but there is no cylinder geometry where the emission reaches below the reference. This means that emitters placed in the outer perimeter of the nanocylinder surface couple well to multipolar resonances that promote upward emission (see Fig. S2†). Fig. 1h and i indicate that the placement of quantum dots on the nanocylinders strongly controls the directional emission.
Subsequently, QDs were selectively placed on top of the Si cylinders using our novel one-step soft stamping procedure as depicted in Fig. 2b. We used CdSeTe/ZnS core–shell QDs (QDot 800, Thermo Fisher) dispersed in octane. First, a layer of QDs was directly spin-coated on top of a silanized PDMS stamp. The Si nanocylinder sample surface was activated with a UV-ozone treatment. Then, the PDMS stamp was mounted above the Si sample in a soft printing machine, face down. The stamp was brought into contact with the sample and pressed down with a force of 3 N. Once in contact, the sample was heated to 40 °C for 10 min, after which the stamp was withdrawn. The stamping was implemented with motorized controls. The flexible nature of PDMS facilitated a conformal contact with the substrate, which printed a uniform film of QDs on the sample. Fig. 2c shows a representative SEM image of the top surface of a cylinder after stamping, showing the homogeneous coverage. In the zoomed-in image (Fig. 2d), the individual QDs can be recognized with a diameter of ∼10 nm on the cylinder surface. The QD print on the cylinder and the clean Si surface besides the cylinder in Fig. 2c confirm that the stamping method printed completely and selectively on top of the tallest nanostructured surface.
Fig. 3b shows a photoluminescence (PL) intensity map at λ = 760 nm of the same region imaged in Fig. 3a. A PL spectrum was measured under excitation of a λ = 532 nm laser at each position on the map. The measured PL intensity in the upwards direction is defined by the collection of the objective, i.e. up to angles of ∼64 degrees (NA = 0.9) from the confocal collection spot of 1 μm. Clear QD emission is observed on top of the cylinders, and strong emission is also observed from QDs printed on the Si wafer next to the nanopatterned region. The latter directly results from the use of a flexible stamp that conformally coats the surface. In contrast, in between the cylinders, the signal is low. In Fig. 3c, the PL spectra for three characteristic positions are plotted: on the flat Si surface, on top of a cylinder, and in between four cylinders. The characteristic wide-band QD emission spectrum is observed in all three cases. We attribute the emission observed in the map between the cylinders to the fact that the tail of the laser (diffraction-limited spot size ∼300 nm) excites QDs on top of the cylinders when the spot is centered in between cylinders and indirect excitation by light scattered from small roughness on the etched Si surface.
The PL data in Fig. 3 have been obtained for a rather thick QD layer printed on the sample. This results in a strong signal in the PL map in Fig. 3b and strong contrast between the cylinders and the surface, enabling direct imaging of QDs on top of the cylinders. Atomic force microscopy (AFM) measurements of the printed layer are shown in Fig. 4a. The printed layer is not completely conformal: the top surface of the Si cylinder is clearly visible (dark orange), with the inhomogeneous QD coverage in brighter colors, up to a height of 100 nm. To perform optical experiments on a thinner and more homogeneous QD layer and to enable a comparison with our simulation results, we repeated the printing process with a QD monolayer spin-coated on the PDMS stamp. An AFM map of the resulting QD coverage is shown in Fig. 4b, showing a homogeneous thin film of QDs on the cylinder surface. This imprint of a thin film of QDs corresponds to the SEM images in Fig. 2b and c.
Fig. 5a shows the PL map for the same cylinder array as in Fig. 3, with 425 nm diameter and 1275 nm pitch, but now with a QD monolayer on top. The emission from the individual cylinders is not distinguishable here because of the smaller 20× magnification. To compare the QD emission from the cylinder array with the flat Si beside it, we correct the emission counts for the ratio of the unit cell area to the cylinder top area, assuming that the cylinders are covered with QDs and that there are no QDs between the cylinders. We find that the upward emission intensity from the QDs on the cylinders, as collected by the microscope, is enhanced by a factor 2.6 ± 0.2 relative to the flat Si reference.
To compare the measured results with simulations, we have to account for resonant enhancement of the quantum dot by the cylinders. To do this, we derive the excitation enhancement by using FDTD to obtain the electric field enhancement above the cylinder at the laser wavelength (see Methods and Fig. S3†). The electric field intensity above the cylinder is then used to weigh the contribution to the upward emission intensity of each position on the cylinder's top surface. From this analysis, we find an upward emission enhancement by a factor 1.8, close to the experimental value found above from Fig. 5a. This confirms that the FDTD simulations accurately predict the coupling between the emitter and the nanostructure in the near field and the resulting far field emission.
We now use FDTD to calculate the upward QD emission intensity as a function of cylinder diameter, normalized to the flat Si reference simulation (Fig. 5c). A strong dependence on cylinder diameter is observed, with a ten-fold enhancement for 150 nm diameter cylinders. To test this experimentally, we measured a PL map for a cylinder diameter of 200 nm and pitch of 600 nm (Fig. 5b). The same analysis as for Fig. 5a yields an enhancement factor for upward emission of 6.1 ± 0.4, consistent with the upward trend for smaller diameter shown in the simulations of Fig. 5c, but below the calculated value of 6.8. We explain the discrepancies by small differences in geometry between experiment and calculation. Overall, the experimentally observed enhanced upward emission is well explained by the combination of resonant directional emission and enhanced excitation obtained from simulations. Fig. 5c doesn't compare to the results of Fig. 1 due to the correction for the excitation enhancement; separate curves for the simulated upward emission and excitation enhancement can be found in Fig. S5.†
Flat silicon | On cylinder | |||
---|---|---|---|---|
Fast lifetime | Slow lifetime | Fast lifetime | Slow lifetime | |
d = 425 nm | 1.11 ns | 20.1 ns | 1.22 ns | 21.7 ns |
d = 200 nm | 0.91 ns | 18.8 ns | 0.69 ns | 21.3 ns |
To analyze this in more detail, we use the FDTD simulations of the angular emission profiles to calculate the PL emission rate (see Methods). Fig. 6c shows the simulated PL radiative lifetime for a dipole on a Si cylinder, normalized to the lifetime of a dipole on a flat Si substrate and averaged over all possible dipole positions. Here we also correct for the excitation enhancement by the laser on the cylinder (see Fig. S5†), to ensure that the simulations emulate the experimental conditions. A strongly varying dependence of cylinder diameter is observed, and the measured decay rates (black) correspond well with the simulations: for the 200 nm diameter cylinder, the lifetime is significantly shorter than for the flat Si reference case, while the lifetime of the 425 nm diameter case is just above that for the reference. The variations in the simulated lifetime are almost entirely due to the simulated excitation enhancement (see Fig. S6†), which further corroborates the necessity of the correction.
Finally, we investigate the Purcell factor and LDOS as a function of cylinder diameter. The PL lifetime curve in Fig. 6c results from the convolution of the calculated LDOS enhancement and laser excitation enhancement, which corresponds to the measurement conditions and compares well with the results. In Fig. 6d, we leave out the correction for the laser excitation enhancement and plot only the calculated emission enhancement as a function of the cylinder diameter (but still averaged over polarization and cylinder surface). The left and right y-axes values are the Purcell factor and absolute LDOS, respectively, which directly follow from the FDTD calculations and the free-space LDOS at the emission wavelength (see Methods). We find an increase of the Purcell factor of about 10% for almost all diameters compared to the flat Si reference. The dashed lines in Fig. 6d depict the downward emission fraction for the cylinder and reference case. Comparing the total LDOS to the downward LDOS and both reference values, we can conclude that the enhanced upward emission (Fig. 1i) is due to an increase in the upward LDOS – the downward LDOS is almost equal to the reference. For diameters of 150 nm and smaller, the LDOS increases significantly while simultaneously the upward emission fraction increases, detailing that the cylinder modes increase the upward LDOS more strongly. The LDOS as a function of dipole position on the cylinder is detailed in the ESI (Fig. S7†), which shows a strong radial dependence. Preferential upward or downward emission can thus further be controlled by placing QDs selectively on the center or edge of the cylinder.
To multipole decomposition as in Fig. 1g was performed by calculating the electric field inside the nanocylinder according to the method by Evlyukhin et al.34
Radiative lifetime simulation results were obtained from the same dipole emission simulations by keeping a fixed electric dipole amplitude in each simulation. In this fashion, the total emitted power at constant dipole amplitude is directly proportional to the Purcell factor and the LDOS.31 In turn, the inverse of the total emitted power is a measure for the experimental radiative lifetime. Direct comparison of simulated and measured results was possible because the flat Si reference case normalizes both. The absolute values for the LDOS in Fig. 6d were obtained by multiplying the calculated Purcell factor by the free-space LDOS at the emission wavelength:
The excitation rate enhancement due to resonant coupling at the pump wavelength was simulated, and the results were used to normalize the experimental photoluminescence intensity and lifetime results. For both laser wavelengths, the electric field intensity above the nanocylinder was determined. We used a monochromatic plane-wave source above a nanocylinder, employing periodic boundary conditions in the substrate plane to mimic the array in the experiment and PMLs in the lateral directions. Using a field monitor, we obtained the field intensity 5 nm above the cylinder surface (see Fig. S3 and S4†). Again, normalization to the flat Si reference case allowed for direct comparison with measurements.
The printing experiments were performed with a Universal Testing System model 5965 with 50 kN force capacity from INSTRON. The Si substrate with cylinders was activated for 10 min by UV-ozone treatment. Both the stamp and the substrate were mounted via vacuum on the printing machine.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1na00630d |
This journal is © The Royal Society of Chemistry 2022 |