Plasmonics–Nanofluidics Hydrid Metamaterial: An ... - ACS Publications

Sep 25, 2017 - Innovative Photon Manipulation Research Team, RIKEN Center for Advanced Photonics, Wako, Saitama 351-0198, Japan. ‡. Metamaterials ...
0 downloads 0 Views 3MB Size
www.acsnano.org

Plasmonics−Nanofluidics Hydrid Metamaterial: An Ultrasensitive Platform for Infrared Absorption Spectroscopy and Quantitative Measurement of Molecules Thu H. H. Le† and Takuo Tanaka*,†,‡,§,⊥ †

Innovative Photon Manipulation Research Team, RIKEN Center for Advanced Photonics, Wako, Saitama 351-0198, Japan Metamaterials Laboratory, RIKEN, Wako, Saitama 351-0198, Japan § Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Kanagawa 226-8503, Japan ‡

S Supporting Information *

ABSTRACT: One of the most attractive potentials of plasmonic metamaterials is the amplification of intrinsically weak signals such as molecular infrared absorption or Raman scattering for detection applications. This effect, however, is only effective when target molecules are located at the enhanced electromagnetic field of the plasmonic structures (i.e., hot-spots). It is thus of significance to control the spatial overlapping of molecules and hot-spots, yet it is a long-standing challenge, since it involves the handling of molecules in nanoscale spaces. Here a metamaterial consisting of a nanofluidic channel with a depth of several tens of nanometers sandwiched between plasmonic resonators and a metal film enables the controllable delivery of small molecules into the most enhanced field arising from the quadrupole mode of the structures, forming a plasmon−molecular coupled system. It offers an ultrasensitive platform for detection of IR absorption and molecular sensing. Notably, the precise handling of molecules in a fixed and ultrasmall (10−100 nm) gap also addressed some critical issues in IR spectroscopy such as quantitative measurement and measurement in aqueous solution. Moreover, a drastic change in the reflectance characteristic resulting from the strong coupling between molecules and plasmonic structures indicates that molecules can also be utilized as triggers for actively switching the optical property of metamaterials. KEYWORDS: infrared absorption spectroscopy, metamaterials, plasmons, hot-spot, nanofluidics, SEIRA, quantitative measurement

V

plasmonic resonators. The coupling results in a large change in the far-field spectral response of the system, in which signals of molecules are apparently amplified. This is the mechanism for the enhancement of sensitivity.7−10 This effect, however, is only effective when molecules are located in the vicinity of hot-spots; thus it is common sense that positioning target molecules exactly at the hot-spots is crucial to effectively utilize the plasmonic−molecular coupling.11−13 Despite recent efforts in this field, most of the studies have focused on either engineering or tailoring the plasmonic structures, while molecules are randomly adsorbed on sensing surfaces by chemisorption or physisorption.14−16 In such cases, the small spatial overlapping of hot-spots and molecules restricts the

ibrational spectroscopy, including IR absorption and Raman spectroscopy, is one of the most powerful biochemical analysis tools, as it extracts essential information on chemical bonds and molecular structures in a label-free fashion. These techniques are especially useful in tracing subtle changes of the conformational structures in response to surrounding environment or probing the kinetics of chemical events.1−3 Although vibrational spectroscopies can offer a large information content, the low sensitivity originating from the intrinsically low absorption/scattering cross sections severely limits their practical applications. Surface-enhanced and the later developed plasmonic metamaterials-based IR absorption and Raman spectroscopies have emerged as promising approaches that improve the sensitivity by several orders of magnitude, owning to the localized enhanced electromagnetic field (hot-spots) in plasmonic materials.4−6 Vibrational modes of molecules, which do not interact strongly with the incident light, nevertheless can strongly couple with © 2017 American Chemical Society

Received: April 20, 2017 Accepted: September 25, 2017 Published: September 25, 2017 9780

DOI: 10.1021/acsnano.7b02743 ACS Nano 2017, 11, 9780−9788

Article

Cite This: ACS Nano 2017, 11, 9780-9788

Article

ACS Nano

Figure 1. (a) Concept of the plasmonics−nanofluidics hybrid metamaterial: a nanofluidic channel of several tens of nanometers in depth sandwiched between Au nano square-disks and a Au metal film. (b) Configuration of a unit cell. (c) The numerical calculation result of the corresponding model indicates that at resonance the electric and magnetic field are being accumulated inside the nanochannel. The color scale bar is commonly used for the two images.

system carried out by the finite-element method (FEM) reveals that there is a strong resonant mode, at which the electric dipole on the Au nano square-disk induces an opposite dipole on the mirror, and the two dipoles form a quadrupole mode (i.e., magnetic dipole mode). Figure 1(b) illustrates the conceptual diagram of a unit cell, in which the width w of the square-disk, the periodicity p, and the lattice structure are controlled to tune the resonant condition. To standardize the structures of different lattices, we define the filling factor f as the coverage ratio of the nano square-disk in a unit cell in terms of surface area. Figure 1(c) shows the electric E and magnetic H field distributions under x-polarized irradiation, indicating that at resonance, E field and H field are accumulated inside the gap (i.e., inside the channel). This distribution suggests that when molecules are introduced into the channel, their large spatial overlapping with the hot-spots can induce a profound rise in the plasmonic−molecular coupling. This resonant mode is easily affected by the spacer layer thickness; therefore it is essential to control the nanogap thickness g (or the channel depth) at a fixed value to realize the structure. The fabrication of a micro/nanofluidic device, so far, relies on soft substrate materials such as dimethylpolysiloxane (PDMS), yet controlling the nanogap in such devices is quite challenging since they easily deform. Although hard wafers such as SiO2 or Si can solve this problem, they have low compatibility with plasmonic metamaterials, especially due to the bonding process that is the most critical step in the fabrication of fluidic chips. The bonding between the SiO2 substrates is usually performed at extremely high temperatures, for example, around 1000 °C for fused-silica glass, after cleaning in piranha solution.23−25 Unfortunately, most of plasmonic materials cannot tolerate high temperatures nor harsh chemical conditions. This has been a technical bottleneck for integrating metallic objects into the nanofluidic chip. Here we developed a room-temperature (RT) bonding technique for SiO2 fluidic chips to fabricate the proposed device. We also confirmed its validity for the CaF2 substrate, which is transparent to IR light, but weak to thermal and chemical treatments. The fabrication process is described in Figure 2(a). On a SiO2 or CaF2 substrate coated with resist, nano square-disk arrays with different patterns (different width w, periodicity p, and lattice geometry) were written by electron beam (EB) lithography. They were then deposited with a 5 nm Cr/65 nm Au film and subsequently lifted-off to obtain Au nano square-

sensitivity. This problem gets much more difficult in complex 3D or multilayer metamaterials, in which hot-spots are embedded inside the structures. In addition, the inhomogeneity and the uncertainty in number of adsorbed molecules are serious obstacles for the reproducibility and determination of molecular quantity. Those issues can be addressed, in part, by using hollow-core waveguides,17 resonant cavities,18 or microfluidics19 for continuously introducing molecules onto the plasmonic surface; however the improvement was still limited because the sample volume in such cases is much larger than the hot-spot sizes. The challenge lies behind the precise handling of molecules at the nanoscale hot-spots. On the other hand, nanofluidics, which is the science and technology of fluids confined within nanoscale structures, has recently attracted much attention in the field of single molecule and single cell analysis, as it enables the precise and automatic handling of molecules in ultrasmall spaces at will.20−22 Here we proposed and demonstrated a metamaterial that leverages the resonant coupling between molecules and plasmonic structures to its fullest, by combining molecules and plasmonic resonators in a heterogeneous configuration. It is composed of plasmonic resonators and a metal film sandwiched by a nanofluidic channel. This structure enables a controllable and effective delivery of molecules into the most enhanced electromagnetic field of the resonance arising from the interference between the top resonators and bottom metal mirror, which is known as a “MIM absorber” structure. This configuration induces a strong coupling between molecules and the plasmonic structure; thus it is expected to significantly enhance the IR absorption signal of molecules with respect to conventional approaches. Notably, the manipulation of molecules by fluidic operation in 10−100 nm spaces also allows us to quantify the number of molecules with good reproducibility, as well as to measure biomolecules in aqueous solution, since herein the disturbance from large absorption of water is effectively suppressed.

RESULTS AND DISCUSSION Concept and Device Fabrication. The concept of the device is depicted in Figure 1(a). A nanofluidic channel is embedded with gold (Au) nano square-disk arrays on the top wall and a Au mirror on the bottom one. The nanogap g between two structures is controlled at several tens of nanometers. The numerical calculation of the corresponding 9781

DOI: 10.1021/acsnano.7b02743 ACS Nano 2017, 11, 9780−9788

Article

ACS Nano

Figure 2. (a) Details of the fabrication process. (b) Fabricated device, microscopic image of the nanochannel, and representative SEM images of nano square-disk arrays with different width w, periodicity p, and lattice geometry.

disk arrays. On another SiO2 substrate, the microchannels used for guiding fluidics into the nanochannels were fabricated by UV photolithography followed by reactive ion etching (RIE). Nanochannels (designed width: 200 μm; depth: 180 nm; and length: 400 μm) were fabricated by UV lithography and wetetching in buffered hydrofluoric acid. After that, a Au film thicker than 90 nm (5 nm Cr/ >85 nm Au) was deposited inside the nanochannels by UV photolithography and a lift-off technique. A careful position adjustment was necessary in all steps to ensure the matching between upper and lower patterns. The two substrates then underwent a UV/O3 treatment to activate the surface −OH groups, before they were aligned and bonded together at RT. To reinforce the bonding strength, in the case of the CaF2 chip, the chip was applied with a pressure of 250 Ncm−2 in 1 h right after the bonding. Details of the bonding process and possible bonding mechanism are described in the Supporting Information SI-2. Figure 2(b) shows the fabricated device, the microscopic image of the nanochannel, and the representative scanning electron microscopy (SEM) images of Au nano square-disk arrays. The bright part in the image represents the bottom Au mirror, while the eight darker patterns indicated by an arrow are the squaredisk arrays. In one device, several arrays of various patterns can be simultaneously integrated to finely tune the resonant frequency. In this experiment, the nanogap between two structures was designed at 15 nm (experimental value: 14.3 ± 0.9 nm). The nanogaps were well controlled from batch to batch, while they varied from experiment to experiment at several nanometers (see Supporting Information SI-5). The bonding strength was tested by introducing water under pressure-driven flow. The device was verified to tolerate an applied pressure up to ∼360 kPa, which corresponds to a flow velocity of ∼26 μm/s through the nanogap. Characterization of Optical Properties. The device embedded with different patterns was measured by an FT-IR spectrometer under nonpolarized irradiation. As shown in the reflectance spectra in Figure 3(a), each pattern exhibited several resonant dips. Among them, the strongest one can be attributed to the quadrupole resonance, at which the reflected light is significantly eliminated. The origin of this mode will be discussed further below. Since the bottom Au film is thick enough to block the transmitted light, the combined zero transmission and cancellation of reflection implies a nearly

perfect (∼90%) absorption of light. The result also confirms the possibility of finely tuning the resonance in the mid-IR regime. Since a larger field enhancement effect can be expected in structures with large absorbance (i.e., low reflectance), several parameters were taken into consideration to optimize the absorbance, in conjunction with tailoring the resonant frequency to the mode of interest. According to the spectra of devices with different widths w and filling factors f (Figure 3(a)), at a certain gap thickness, even with the same width w, the resonant frequency and the absorbance slightly varied with the filling factor. Under a fixed gap thickness of 15 nm, by controlling the width w and filling factor f, the fine-tuning of resonant frequencies with an optimum reflectance dip was demonstrated as shown in Figure 3(a). Next, we verified the interaction of the plasmonic structure and molecular vibrational modes by introducing water into the nanochannel. The structure was designed to exhibit the resonance at 3200−3500 cm−1, which is overlapped with the O−H stretching band of H2O molecules (detailed design in Supporting Information SI-4). Heavy water, D2O, was chosen as reference sample, since it shares the same refractive index with H2O (real part of the complex refractive index), while its absorption band (imaginary part of the complex refractive index) is shifted to a much lower energy region (2400−2600 cm−1). The IR reflectance spectra of the device filled with D2O and H2O are shown in Figure 3(b). A strong resonance dip exhibiting a symmetric Lorentzian line at 3200−3600 cm−1 was clearly observed in the spectrum of the “square-disk−D2O−Au film” system. This dip can be attributed to the light absorption by the plasmonic resonance of the structure itself, since D2O and the structure are in off-resonance. When the filling fluidic was changed from D2O to H2O, the spectrum transformed to a Fano-like resonance line shape, revealing the typical characteristic of coupled plasmonic systems.26,27 The drastic change in reflectance spectra in the 3000−3600 cm−1 range is solid proof of a strong coupling between H2O molecules and the plasmonic structure. A numerical calculation was carried out to study the mechanism of the above systems (Supporting Information SI6). The calculated electric field and electric current profiles in Figure 3(c−f) have elucidated the origin of the observed spectra. Under x-polarized light illumination, at resonant frequency (indicated by the broken line in Figure 3(b)) an 9782

DOI: 10.1021/acsnano.7b02743 ACS Nano 2017, 11, 9780−9788

Article

ACS Nano

Figure 3. (a) Reflectance spectra of devices filled with air measured under nonpolarized light revealing a strong resonance at which the reflected light is extremely canceled out (quadrupole mode). Structures perform as absorbers. Their absorbance and resonant frequencies were finely tuned in the mid-IR regime by controlling the widths w and filling factors f. (b) Reflectance spectra of devices filled with D2O and H2O, respectively. The drastic change of reflectance from 12.9% to 50.4% indicated a strong coupling of plasmonic structures and vibrational modes of H2O molecules. (c) Profiles of electric field Ez and (d) the current density Jx indicated by red arrows in the “square-disk−D2O−Au film” structure revealing the quadrupole mode resonance. (e) Electric field Ez and (f) current density Jx disappear in the “square-disk−H2O− Au film” structure, indicating the breaking of the quadrupole mode as it couples with vibrational modes of H2O. The result clarifies the mechanism for the recovery of reflected light in the “square-disk−H2O−Au film” system. The color scale bar is common for four images (c− f).

range of interest. The device was purposely designed to exhibit the resonance at 2800−3000 cm−1 that covered the C−H vibrational bands of C18H38 molecules. Figure 4(a) shows the reflectance spectra of a channel filled with CCl4 solvent and C18H38/CCl4 solution, respectively. It unveiled three C−H vibrational modes including the asymmetric, symmetric stretching of −CH2, and asymmetric stretching of −CH3 as distinct peaks in the broad reflectance dip of the plasmonic structure. Although the shape of those peaks were different from the case of water (Figure 3(b)) due to the difference in Qfactor of the intrinsic vibrational modes of molecules, this result agreed well with the mechanism discussed above. In order to quantitatively analyze the coupling between molecules and plasmonic structure, we proposed the following theoretical model. The observed spectrum is a result of the coupling between three resonators: (a) the plasmonic structure whose resonance was measured as the channel was entirely filled with CCl4, (b) the asymmetric stretching of −CH2 at ∼2930 cm−1, and (c) the symmetric stretching of −CH2 at ∼2850 cm−1. The asymmetric stretching of −CH3 was omitted in this model, as its peak intensity was negligibly small. The three-resonator coupled system was fitted by employing the temporal coupled theory model (TCTM), which is generally used in coupled plasmonic systems.19,30,31 This model allows us to numerically analyze the underlying coupling coefficients between reso-

electric dipole is resonantly excited in the upper Au square-disk with the dipole oriented along the electric incident field. This localized plasmon mode induces an opposite dipole, which is its own mirror image, in the Au film. They appear as antiparallel currents running in the top Au square-disk and the bottom Au mirror in the Jx distribution map displayed in Figure 3(c,d). The two dipole modes interfere and form a quadrupole mode (i.e., magnetic dipole mode), which is the nonradiative mode, leading to the antireflection.28,29 The opposite dipoles are clearly observed in the electric field Ez profile. When H2O is introduced into the gap, their vibrational modes couple with this quadrupole mode and break the antireflection condition. This is observed as the disappearance of both the quadrupole mode and antiparallel currents Jx in Figure 3(e,f). As a consequence, the light is reflected back at the frequency range corresponding to the vibrational absorption bands of H2O, and the obtained spectrum exhibits a Fano-like line shape. Detection and Tracing the Quantity of Molecules. Exploiting this mechanism, even a weak molecular absorber whose coupling rate with the incident light is low can be effectively detected. To demonstrate the performance of our device in detecting and tracing the quantities of molecules, we used octadecane (C18H38) dissolved in carbon tetrachloride (CCl4) solvent as target molecules. CCl4 solvent was chosen here because it is free of vibrational absorption in the frequency 9783

DOI: 10.1021/acsnano.7b02743 ACS Nano 2017, 11, 9780−9788

Article

ACS Nano

Figure 4. (a) Reflectance spectra as the device was filled with octadecane C18H38/CCl4 solution confirming three distinct peaks of C−H stretching bands in the reflectance dip of the plasmonic structures. The fitted spectrum based on TCMT matched well with the experimental one, strongly supporting the validity of our analysis model. (b) Reflectance spectrum of the device filled with 7.8 mM C18H38 solution clearly showing detectable signals of two C−H stretching peaks, proving the sensitivity at a molecular density of ∼10−4 molecules Å−2. This value demonstrated the sensitivity improvement by 2 orders of magnitude compared to reported techniques using metamaterials. (c) Dependency of molecular density on the molecular signals expressed as (μab + μac) showing a linearity at the low molecular density regime, verifying the possibility of quantitative measurement of molecules. The LOD was estimated to be 4.4 × 10−4 molecules Å−2.

experimental one, as shown in Figure 4(a), strongly supporting the validity of our analysis model. Furthermore, we assume that the vibrational mode of every single molecule is an independent resonator that shares the same coupling coefficients μ with the plasmonic structure. According to this assumption, for example, μab also presents ∼2930 cm−1 vibrational modes involved in the coupling, or in other words it corresponds to the number of molecules contributing to the coupled system. To fairly evaluate the sensitivity of our device, we employed the molecular density, which is defined as the number of sensing molecules per surface area of the nanostructure. Most of the studies on sensing the self-assembled monolayer (SAM) of the same analyte species on engineered plasmonic structures reported the detection limit (LOD) at a molecular density of ∼10−2 molecules Å−2 (e.g., ∼4.6 × 10−2 molecules Å−2).9,16,32 We define the molecular density in our device as the total number of molecules inside the nanogap (i.e., concentration × gap volume) divided by the gap surface area. For example, the reflectance spectrum when the device was filled with 7.8 mM C18H38 in Figure 4(b) confirmed detectable peaks with sufficient signal-to-noise ratio (SNR). This concentration in the 15 nm gap corresponds to a molecular density of 7.1 × 10−4 molecules Å−2. It should be noted that this value is 2 orders lower than previous reports, indicating the enhancement of sensitivity up to 2 orders of magnitude. The achieved sensitivity

nators, which can be used to extract the molecular IR absorbance. We assume that resonator (a) couples with resonators (b) and (c) independently, while (b) and (c) do not interfere with each other, and the coupled mode equations for the configuration become da = jνaa − (γae + γa0)a + jμab b + jμac c + κS1 + dt

(1)

db = jνbb − γb0b + jμab a dt

(2)

dc = jνcc − γc 0c + jμac a dt

(3)

In these equations, a, b, and c represent the mode amplitudes of resonators (a), (b), and (c), respectively; νa, νb, and νc are the intrinsic center frequencies of the resonators; γae is the radiative loss rate of plasmonic resonator (a), while γa0, γb0, and γc0 are the nonradiative damping rates of resonators (a), (b), and (c), respectively. μab and μac represent the coupling coefficients of (a)−(b) and (a)−(c), while S1+ is the incoming wave. Solving those equations to extract the reflectance and fitting it with the experimental data (see Supporting Information SI-1), we obtained the coupling coefficients μab and μac for each spectrum. The fitted spectrum matched perfectly with the 9784

DOI: 10.1021/acsnano.7b02743 ACS Nano 2017, 11, 9780−9788

Article

ACS Nano

Figure 5. (a) Reflection spectra of the device filled with water and lysozyme 2 mg mL−1 demonstrating the detection of protein in aqueous solution. (b) Differential spectrum revealing two distinct bands of amide I and amide II in a typical Fano-like line shape, verifying the measurement in aqueous solution with a prominent SNR.

of our device firmly verified our proposal that exactly positioning molecules at the hot-spots has contributed to the enhancement of plasmonic−molecular coupling and thus improved the sensitivity. Taking advantage of precisely controlling the number of molecules inside the nanogap, our device can simultaneously detect and trace the quantity of molecules. Solutions with different concentrations were introduced into the device, and the corresponding spectra were recorded and fitted using the fitting procedure mentioned above. We experimentally confirmed a linear relationship between coupling coefficients and molecular density in a narrow concentration range. The calibration curve in Figure 4(c) in which the summation of coupling coefficients μab and μac was plotted against the molecular density (molecules Å−2) revealed a good linearity at a molecular density lower than 3.0 × 10−2 molecules Å−2. Note that it is essential to compare the summation of the two peak strengths since the tail of the first peak may affect the height of the second one and vice versa.32 The relationship between coupling coefficients and number of molecules in a large range of concentrations is discussed in more detail in Supporting Information SI-1. The small deviation of signal intensities standing for a prominent reproducibility supports that our device performs as an excellent platform for quantifying the number of molecules. The detection limit is defined as the analyte density corresponding to the signal of the sample blank value adding three times the standard deviation. According to this definition, we estimated the LOD to be 4.4 × 10−4 molecules Å−2, showing an improvement of 105-fold compared to previous studies.9,16,32−34 It is also worth mentioning that the minimum detectable volume is estimated to be ∼0.5 fL, which is the smallest one to the best of our knowledge. Our device not only offers a pronounced improvement of sensitivity but also allows us to address some critical issues in conventional and surface-enhanced IR absorption (SEIRA) spectroscopies. The first demonstration is the quantitative measurement of molecules that has been verified above. The later one is the measurement of biomolecules in aqueous solution. Measurement of Protein in Aqueous Solution. In bioanalysis, measurement in water solution is an inevitable process, but the comparatively large absorption of water becomes a crucial obstacle in IR spectroscopy in aqueous

solution. Specifically, the O−H bending band of water overwhelms the amide I and II bands, which carry essential information on protein secondary structures. Various efforts have been made so far to limit the optical path length (