D.
Garoli‡
a,
E.
Calandrini‡
a,
A.
Bozzola‡
a,
M.
Ortolani
b,
S.
Cattarin
c,
S.
Barison
c,
A.
Toma
a and
F.
De Angelis
*a
aIstituto Italiano di Tecnologia, Via Morego 30, 16136 Genova, Italy. E-mail: francesco.deangelis@iit.it
bDipartimento di Fisica, Sapienza Università di Roma, Piazzale Aldo Moro, 5, I-00185 Roma, Italy
cIstituto di Chimica della Materia Condensata e di Tecnologie per l'Energia (CNR-ICMATE), Corso Stati Uniti 4, I-35127 Padova, Italy
First published on 7th December 2016
The applications of plasmonics to energy transfer from free-space radiation to molecules are currently limited to the visible region of the electromagnetic spectrum due to the intrinsic optical properties of bulk noble metals that support strong electromagnetic field confinement only close to their plasma frequency in the visible/ultraviolet range. In this work, we show that nanoporous gold can be exploited as a plasmonic material for the mid-infrared region to obtain strong electromagnetic field confinement, co-localized with target molecules into the nanopores and resonant with their vibrational frequency. The effective optical response of the nanoporous metal enables the penetration of optical fields deep into the nanopores, where molecules can be loaded thus achieving a more efficient light–matter coupling if compared to bulk gold. In order to realize plasmonic resonators made of nanoporous gold, we develop a nanofabrication method based on polymeric templates for metal deposition and we obtain antenna arrays resonating at mid-infrared wavelengths selected by design. We then coat the antennas with a thin (3 nm) silica layer acting as the target dielectric layer for optical energy transfer. We study the strength of the light–matter coupling at the vibrational absorption frequency of silica at 1240 cm−1 through the analysis of the experimental Fano lineshape that is benchmarked against identical structures made of bulk gold. The boost in the optical energy transfer from free-space mid-infrared radiation to molecular vibrations in nanoporous 3D nanoantenna arrays can open new application routes for plasmon-enhanced physical–chemical reactions.
Achieving efficient optical energy transfer may be even more difficult in the mid-infrared (mid-IR) region (wavelengths of approximately 2 to 12 microns), where the narrow-linewidth vibrational fingerprints of molecules would allow the selective excitation of selected chemical species, but unfortunately the intrinsic advantages of plasmonics tend to vanish with increasing wavelength. This is due to a progressive and strong increase of the imaginary part of the noble metal permittivity ε2, which is responsible for optical energy losses and plasmon damping, and a simultaneous increase of the absolute value of the real part of the permittivity ε1, whose value close to unity is responsible for field confinement in the visible range. Indeed, a strong field confinement at a metal/dielectric interface of the order of the skin depth can be obtained when the metal and the dielectric permittivity are equal in magnitude and opposite in sign.1–4 In fact, applications of plasmonics are currently limited to the visible/near-infrared region of wavelengths shorter than 1 micron and, although some recent progress has been made using non-metallic conductors,14–18 the research for alternative materials for mid-IR plasmonics is still an open issue. The options that have been explored up to now include conducting oxides, heavily doped semiconductors, phase-change amorphous materials and graphene. Conducting oxides include indium–tin oxide (ITO) and Al-doped ZnO (AZO), which are transparent to visible radiation, are easy to deposit with low-cost processes and feature widely tunable ε1 and ε2 in the entire IR range.14,15 The alloy nature of ITO and AZO, however, makes their optical properties very sensitive to fabrication processes including material removal at an extremely high spatial resolution (e.g. ion milling, reactive ion etching, etc.), which is needed to obtain plasmonic nanostructures. Heavily doped semiconductors with a small effective mass like electron doped InAs16 and Ge17 are definitely more suitable for nanostructure fabrication and integration into microelectronic chips, but, with the doping levels obtained at the present state of the art of materials science, they can only function as plasmonic materials at far-IR wavelengths longer than 10 microns. Nanoantennas made of phase-change materials18 have shown large tunability with the temperature of the resonant wavelength in the mid-IR range from 5 to 8 microns, but, for most chemical catalysis applications, temperature changes are precisely what one would like to induce with the optical energy transfer, rather than having to provide thermal energy to the plasmonic system through a heater. Graphene metasurfaces19 have demonstrated high surface sensitivity, however at the expense of a strongly reduced total sensing volume if compared to standard plasmonic metal meshes.20 In summary, the problem of finding efficient materials and device concepts for mid-IR plasmonics is not solved, and it will be even more urgent in the near future, because of the recent commercialization of quantum cascade lasers (QCLs)21 that are now broadly tunable in the mid-IR range, enabling automated on-chip gas sensors based on vibrational spectroscopy. More importantly for the scope of this paper, QCLs can emit free-space mid-IR radiation beams with a peak power of the order of 1 W available for optical energy transfer to specific molecular vibrational lines, due to their narrow emission linewidths of the order of few GHz or less.21
In this work, we show that the use of porous loadable metal nanostructures can provide a straightforward method for the co-localization of molecules and mid-IR plasmonic hot-spots with an extremely high surface-to-volume density ratio, insofar boosting the total integrated optical energy transfer from the mid-IR free-space radiation to the molecule vibrations. Due to the effective optical properties of the porous material in the mid-IR range, which are found to be very different from those of the bulk metal, the optical fields can penetrate deep into the metal nanostructure, while the molecules can fill the empty spaces created by the nanopore fabrication process. Efficient energy transfer between the free-space optical beam and the loaded molecules then occurs through mid-IR plasmon oscillations of the nanostructure. The plasmonic behavior in the mid-IR range can be optimized at specific wavelengths through the design of nanostructure arrays and their subsequent characterization by IR spectroscopy, as we do here. In particular, here we show that, as in the case of bulk metals, it is possible to pattern the nanoporous metals deposited on-chip along the out-of-plane direction, thus achieving vertical nanoantenna arrays resonating at the desired wavelength. It is shown here that tuning the antenna resonance in the vicinity of the vibrational frequency of the molecules, together with increasing the interaction volume in the out-of-plane nanostructure configuration,22 further strengthens the coupling between the plasmons and the molecules loaded into the metal nanopores.
It is well known28 that, according to the specific porosity level, NPG can behave like an effective conductor with plasma frequency ωp in the mid-IR range. The main hypothesis of the effective-medium approximation, that the structural feature size is much smaller than the wavelength, is acceptable in the mid-IR and near-IR range, however conventional effective medium theories have failed to reproduce the NPG optical constants in our case, probably because of the high degree of connectivity of the nanostructures and more importantly because of their high shape variability.28 This was not the case for single NPG nanowires.30 Therefore, the obtained effective-medium values for the real (Fig. 1(c)) and imaginary (Fig. 1(d)) parts of the effective permittivity in the IR range have to be considered as phenomenological values,31 and the relation of the obtained Drude parameters (free carrier density and mass, carrier relaxation time, screening constant) to the geometric parameters of the material (pore size and distribution) and to the bulk gold dielectric function will be a subject of a forthcoming paper. The observed reduction of the absolute values of the dielectric permittivity of NPG if compared to that of bulk gold has important consequences. First, since the intensity and orientation of the local electric field depends mostly on the geometry of the subwavelength-sized structures and not on material properties, the reduced ε2 enables lower ohmic losses. Second, the reduction of the imaginary part of the refractive index κ (which depends on both the real and the imaginary parts of the dielectric function) implies an increase of the skin depth δ = c/ωκ, which, in turn, enables the IR radiation to penetrate deep into the buried nanopores.
A quantitative estimate of the skin depth is plotted in Fig. 1(e). The skin depth, around 80 nm, is almost four times larger than the value of bulk gold (less than 25 nm in the IR),31 hence the electromagnetic field penetrates into the buried nanopores where both the plasmonic hot spots and the nano-objects of interest are present, ultimately realizing the co-localization of optical energy and molecules.
In order to exploit the co-localization of optical energy and vibrational dipoles, we fabricate mid-IR plasmonic nanoantennas made of NPG. First of all, we simulated different square arrays of vertical nanoantennas made of both bulk gold and NPG. The height of the antennas is set by the present fabrication process to 2.3 μm, while the lattice pitch P varies in the range 2.5–4.0 μm that generates resonances in the mid-IR range as we will now explain. A TM-polarized plane wave (magnetic field H orthogonal to the axis of the antennas) incident at 30° is used in the simulations as well as in the subsequent experiments. A full 3D electromagnetic calculation that takes into account the real geometry of the nanopores is currently a very hard challenge that would require a huge computational effort, therefore we treated the porous metal as a uniform material with the dielectric function given by the effective phenomenological values experimentally determined in Fig. 1. The results for the field enhancement |E/E0| when the antennas are brought at resonance are illustrated in the vertical section of the field map shown in Fig. 2 for P = 3.5 μm. The deeper penetration of the electromagnetic field inside the volume of the vertical antenna can be clearly observed.
We now turn to the demonstration of an enhanced transfer of optical energy to the matter embedded in the NPG pores. First of all, we note that, due to the small size of the pores (tens of nanometers or less), patterning NPG into nanostructures for enhanced sensing/spectroscopy in the VIS spectral regions6 can be difficult,27 because the size of the subwavelength features would be comparable to that of the pores. Instead, we now show that it is possible to fabricate micron-scale antennas for the mid-IR range out of NPG by using properly designed 3D templates. Moreover, as mentioned above, the approximation of a homogeneous effective dielectric function is increasingly valid at longer wavelengths. The mid-IR nanoantenna fabrication process is as follows: we first fabricate polymeric templates of the vertical nanoantennas, namely polymeric pillars protruding from the substrate plane;22 then we deposit the Ag–Au alloy over the templates; finally we leach the Ag from the alloy with a wet etch, thus realizing vertical out-of-plane antennas made of NPG and resonating in the mid-IR range. The fabrication process is described in detail in the Experimental section below. Examples of the fabricated nanoantennas are reported in Fig. 3(a). The same structures, prepared for reference by depositing bulk gold on the same templates are shown in Fig. 3(b).
To match the estimated skin depth value of NPG, the thickness of the porous metal is set to 80 nm, which gives a final geometrical aspect ratio of cylindrical antennas’ height/diameter ∼6. The porous structure of the metal is clearly visible in Fig. 3(a). We characterized the far-field optical response of the antenna arrays by means of Fourier-transform IR (FTIR) spectroscopy using a microscope with a 30°-incidence Cassegrain objective in the reflection mode. At non-normal incidence, the radiation electric field vector has a non-zero component along the TM10 mode, in which the electric field is mostly aligned along the out-of-plane direction, parallel to the main axis of the antenna hence exciting the fundamental antenna mode. The specular reflectance spectrum of the antenna array R(λ) was obtained by using a bulk gold-coated surface as the reference. In Fig. 4, the reflectance spectra of arrays with different pitches (height 2.3 μm, total radius 360 nm, array pitch varying in the range 2.5–4.0 μm) are reported together with the electromagnetic simulations of the reflected intensity, in which the illumination geometry of the FTIR experiment was also simulated. The reflectance dip displayed at the same wavelength by simulations in Fig. 4(a) and data in Fig. 4(b) confirms the validity of both our modelling and fabrication approaches. The interpretation of the reflectance dip is as follows: due to the overlap of the square lattice resonance wavelength with the fundamental longitudinal mode of the antenna,22 a maximum in the intensity scattered at non-specular angle S(λ) is expected. Since in the absence of transmission and for negligible absorption one has R(λ) ∼ 1 − S(λ), a dip is seen in the reflectance spectra shown in Fig. 4. The wavelength at which the dip appears redshifts proportionally to the increasing lattice period by more than 25%, because the fundamental mode of the vertical antenna has a bandwidth of more than 50% of the centre wavelength, therefore it is the array resonance that controls the exact spectral position of the dip and the linewidth (half width at half maximum is less than 10% of the centre wavelength).
The demonstration of the enhanced transfer of optical energy to vibrational energy is done by sensing the mid-IR vibrational modes of a thin layer of amorphous SiO2 deposited on the antenna arrays by means of atomic layer deposition (ALD).32,33 With this technique, exactly the same thickness of an analyte layer can be deposited on different antenna arrays made of either NPG or bulk gold. Moreover, the intrinsic homogeneity of the deposition rate of ALD ensures a uniform, conformal deposition of the analyte over the 3D surface formed by the porous structure.32,33 As a demonstration of the uniformity and conformity of the ALD, we deposited the same analyte (about 10 nm of SiO2, or 90 ALD cycles) on an NPG sample, and we imaged the cross-section by means of scanning electron microscopy (SEM). The image is reported in Fig. 5 and it is possible to verify the uniformity of the SiO2 layer across the entire 3D porous structure. For the sensing experiment, we choose to deposit 3 nm of SiO2 on the nanoantenna arrays (30 ALD cycles).
The sensing results are presented in Fig. 6, where the FTIR reflectance microscopy spectra of the antenna arrays of Fig. 4 after the SiO2 coating are reported in Fig. 6(a) and (b), respectively, as a function of the electromagnetic frequency ν = c/λ expressed in wavenumbers ω = 1/λ. The antenna resonance could be tuned across the characteristic phonon energies of SiO2 by changing the array pitch P as described in Fig. 4. SiO2 is characterized by two distinct active phonons in the IR range: a transverse optical (TO) phonon mode around 1090 cm−1, and a longitudinal optical (LO) phonon mode at 1240 cm−1.34,35 According to the IR spectroscopy selection rules for a thin SiO2 film on a flat surface, the TO phonon oscillator strength is suppressed in thin amorphous films, while the LO phonon becomes infrared-active in amorphous forms of SiO233 and it is more efficiently excited at non-normal incidence.34 In our 3 nm thick amorphous sample, therefore, the LO phonon produces a stronger signature than the TO phonon in the spectra of Fig. 6, so we shall focus on the LO vibration detected at about ωLO = 1240 cm−1. The strength and the asymmetry of the spectral feature at ωLO contain information on the light–matter interaction strength at the nanoscale, although, at the low photon flux of the FTIR blackbody source, the optical energy transfer is still negligible if compared to e.g. the thermal energy at room temperature. We extract the light–matter interaction-strength parameters by performing a multi-peak fit to a broad Lorentzian background for the plasmonic resonance36,37 and a narrow discrete vibrational transition line centred at ωLO with an asymmetric Fano–Breit–Wigner profile.38,39 In the spectral region around the antenna resonance, the ω-dependent reflectance spectra are fitted to:
R(ω) = R0 − RL(ω) + RF(ω) | (1) |
(2) |
(3) |
The method used to fabricate the samples relies on FIB-generated secondary-electron lithography in lithographic resists. The S1813 optical resist (thickness of 2.3 μm) is spun at 3000 rpm on 200 nm thick Si3N4 membranes and soft-baked at 90 °C for 4 minutes. On the back of the membrane a thin layer of gold (about 10 nm) is then deposited by means of sputtering. The membranes are then patterned from the backside using a Focused Ion Beam (Helios Nanolab600, FEI company), operated at 30 keV (current aperture: 40 pA, dwell time: 500 μs, number of repetitions: 850 passes). Due to the high dose of low-energy secondary electrons induced by the ion beam/sample interaction, a 40 nm thick layer of the resist, surrounding the milled hole, becomes highly cross-linked and insoluble to most solvents. After patterning, the sample is developed in acetone, rinsed in isopropanol and dried under a gentle N2 flow. A mild oxygen plasma (process time: 2 minutes, RF power: 100 Watt, gas flow: 25 sccm) reduces the polymeric skin down to 10 nm. A 80 nm thick layer of the silver/gold alloy is deposited by sputtering the sample, tilted 60° with respect to the vertical and rotated, guaranteeing an isotropic coating on both the sidewalls and the base. The deposition follows the same parameters reported above for the preparation of the NPG thin-films. Finally, in order to prepare the sample with the desired nanoporous structure, the nanoantennas were etched in diluted HNO3 for 3 hours at room temperature.
The optical response in the mid-IR range is investigated by means of FT-IR reflectance microscopy measurements by using a Nicolet™ iS™ 50 FT-IR Spectrometer coupled to the Nicolet Continuμm Infrared Microscope by Thermo Scientific™. A gold mirror was used as the reference.
Electromagnetic simulations are performed by means of the Comsol Multiphysics® software. The bulk gold dielectric function was taken from ref. 31, while the dielectric function of nanoporous gold was taken from the FT-IR data of Fig. 1. Periodic boundary conditions are imposed at the lateral facets of the unit cell, and perfectly matched layers are assumed at the top and bottom boundaries of the simulation box.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6nr08231a |
‡ The authors equally contributed to this work. |
This journal is © The Royal Society of Chemistry 2017 |