1. Introduction

Plasmonic nanostructures have emerged as potential materials for the development of biosensors in both clinical and non-clinical applications ranging from food and water safety, biodefense to point-of-care diagnosis [1]. Surface plasmon-polaritons (SPP) and localized surface plasmon resonances (LSPR) present great qualities for the development of molecular detection and recognition applications [2–8]. LSPR which are collective oscillations arising from the interaction between an electromagnetic wave and conduction electrons in materials, have proved to be sensitive to the surrounding nano-antenna. Particularly, the mid infrared (mid-IR) spectral region is of practical interest for the development of sensing devices since molecules have their characteristic fingerprint vibrational absorption lines in the range of 1800 - 500 cm⁻¹ (5.5 - 20 μm).

Sensing in the mid-IR with LSPR involves the shift of the surface plasmon resonance (SPR) and, the enhancement of the vibrational signal of the sensed molecules when the LSPR and the molecule absorption are in resonance [9]. Thus, SPR and surface enhanced infrared absorption (SEIRA) present compatible performance for mid-IR sensing. The former is intrinsically linked to changes in the bulk refractive index, whereas the latter directly excites the absorption lines of the molecules. In sum, their combination, acting as “chemical fingerprints”, supports a powerful technique for the accurate identification of the surrounding molecules [1,10–13].

Due to their chemical stability, noble metals, particularly gold and silver, are the natural choice for visible and near-IR plasmonic sensing technology [6,14,15]. However, they present some limitations to foresee the mid-IR region and thus, to target molecule sensing. Particularly, gold is almost a perfect conductor and in the infrared region has a very negative value of the real part of the permittivity. As a consequence, it repeals any internal electric field reducing the enhancement of the electric field at its interfaces and thus, dramatically impacts the sensitivity of a biosensor in such spectral range [8]. Highly doped semiconductors (HDSC) have been recently proved as appropriate candidates for the development of sensing devices at longer wavelengths [16–18]. Besides the dependence of the plasma frequency, ${\omega }_p$, on the doping level offers a supplementary degree of freedom to tune the LSPR. This enables to adjust the, ${\omega }_p$ and thus, to tailor the real part of the dielectric function to negative values, therefore the HDSC behave as metals. ${\omega }_p$Setting in the mid-IR requires high doping levels, up to 10¹⁹ - 10²⁰ cm⁻³, achievable using the high quality lattice matched epitaxial system based on Si-doped InAs_0.91Sb_0.09 on (100)-GaSb) [19]. Besides, 1-D InAs_1-xSb_x nano-antennas have proved to outperform comparable gold structures in terms of SPR sensing in this spectral region [20] and to enable the resonance scaling from 500-to-1250 cm⁻¹ (8 – 20 μm) by changing the doping level and the nano-antenna width [19].

In this work, we demonstrate SEIRA using 1-D periodic grating InAs_0.91Sb_0.09/GaSb nano-antennas by analyzing the reflectance spectra before and after coating the nanostructures with ~15 nm-thick layer of polydimethylsiloxane (PDMS). Three samples with different doping levels were nano-engineered to support the LSPR in resonance with the vibrational mode of the PDMS at 800 cm⁻¹. Reflectance in transversal and longitudinal direction to the nano-antennas long-axis were performed in all the samples before and after PDMS deposition. We report a clear dependence of the IR absorption enhancement on the doping level, linked to the plasmonic nano-antennas width. Additionally, finite-difference time-domain (FDTD) simulations are in good qualitatively agreement with experimental data. Furthermore, the influence of the HDSC doping level in the SEIRA sensing is verified also after solving the two coupled harmonic oscillator problem, using the analytical model proposed by Gallinet et al. [21]. It is concluded that the coupling between the plasmonic mode in the nano-antennas and the PDMS vibrational signal increases monotonically with the doping level.

2. Material and methods

The samples were grown by solid source molecular beam epitaxy (MBE) on Te-doped (100)-GaSb substrate. Following oxide desorption, a GaSb buffer layer of a 500 nm-thick was grown at 500°C. Then, a 100 nm-thick layer of Si-doped InAs_0.91Sb_0.09 was grown at 450°C. Three different samples labeled as A, B and C, with plasma frequency at 5.7 × 10¹³ rad·s⁻¹, 5.0 × 10¹³ rad·s⁻¹ and 4.3 × 10¹³ rad·s⁻¹ and doping level 6.6 × 10¹⁹ cm⁻³, 4.0 × 10¹⁹ cm⁻³ and 2.2 × 10¹⁹ cm⁻³ respectively, were selected to study the SEIRA as a function of the doping level (see Table 1). The plasma frequency, and consequently the doping level were determined by using a non-destructive optical technique based on the Brewster mode [22]. The surface was nano-engineered to support the LSPR peak centered at a wavenumber of 800 cm⁻¹. Plasmonic 1-D nano-antennas arrays were fabricated by using photolithography and wet etching [19], being the resulting width 645 ± 10 nm, 490 ± 10 nm and 265 ± 10 nm for samples A, B and C, respectively.

Table 1. Summary of the physical values of the three samples used to study SEIRA in 1-D InAsSb nano-antennas. Column 1 labels the sample and columns 2, 3 and 4 correspond to the plasma frequency, doping level and antenna width, respectively

View Table

The 1-D InAs_0.91Sb_0.09 nano-antennas were structurally and optically characterized by scanning electron microscopy (SEM) and reflection spectroscopy experiments. SEM plan-view images were performed by a SEM-Inspect S-50 with typical electron-beam energy of 20 kV. Cross-sectional images were obtained by using a high resolution (HR) SEM Hitachi S4800. Reflectance measurements were carried out using a Bruker Vertex 70V Fourier transform infrared spectrometer which is equipped with a Hyperion 3000 microscope, a KBr beam splitter, the internal mid-IR optical source and a MCT detector (detectivity, D > 5 × 10⁹ cmHz^1/2W⁻¹). The experiment was performed under quasi-normal incidence in a reduced area of 100 × 100 μm². The microscope was enclosed with a Plexiglas and purged with nitrogen to minimize the environmental fluctuations. The reflectance spectra were measured with 2 cm⁻¹ resolution and 1000 scans from 450-to-5000 cm⁻¹. A gold mirror was used as background reference.

Numerical simulations were performed by 2-D FDTD calculations using the commercial software package from Lumerical Solutions Inc. (Lumerical FDTD Solutions 8.12.631). The optical response and the near electric field distribution were calculated applying the model proposed by Barho et al. [20]. Calculations in the transversal and longitudinal direction to the long-axis of the nano-antennas were conducted assuming the bare sample (without polymer coverage) and the sample coated with a 15 nm-thick of PDMS. Drude function (see Eq. (1)) was used to model the permittivity (ε(ω)) of the InAsSb layer

Figures 1(a)-1(c) show the SEM plan-view images of the tailored nano-antennas to present the LSPR peak centered at 800 cm⁻¹, corresponding to Table 1 describing sample characteristics, where heavily doped layers requires wider nano-antennas compared to that of lighter doped epitaxies [19]. All the samples show a great vertically and small lateral roughness compared to the measured lateral deviation (1%). Figure 1(d) illustrates a SEM cross-sectional image after coating the samples with PDMS. The polymer layer was diluted with hexane (1:100) and spin coated at 6000 rpm and cured at 90°C for 20 min, to achieve a nanometer layer. To increase the contrast of the SEM, a thin layer of about 40 nm of Au was deposited on the PDMS layer. The PDMS layer is homogeneous along the cross-sectional surface of the nano-antennas, as shown in Fig. 1(d). However, the viscosity and porousness of the polymer at the nanoscale make difficult the layer thickness determination [24,25]. HR-SEM experiments were performed to identify and measure the thickness of the PDMS layer. Figures 1(e) and 1(f) show the profile of a nano-antenna covered with PDMS and Au, where the black spacing area between the nano-antenna and the Au layer corresponds to the PDMS layer. The PDMS thickness is ~15 nm in average from the three different samples HR-SEM images.

 

Fig. 1 SEM images before and after PDMS coating of the three samples. (a)-(c) SEM plan-view images of the as-fabricated samples A (w ~645 nm), B (w ~490 nm) and C (w ~265 nm), respectively. (d) Standard and (e)-(f) High-resolution SEM (Hitachi S4800) cross-sectional image of the samples covered with PDMS and a 40 nm Au layer to increase contrast.

Download Full Size | PDF

3. Results and discussion

The experimental reflectance spectra in transversal and longitudinal direction to the long-axis of the resonators are shown in Figs. 2(a) and 2(b). Different color lines and symbols identify the different samples: A (green, square), B (red, triangle) and C (blue, circle). Closed symbols correspond to as-fabricated surface samples, and open symbols to samples covered with the PDMS layer. The three engineered nano-antenna arrays support a LSPR peak wavenumber around ~780 cm⁻¹, a deviation of ~20 cm⁻¹ to the red from the targeted 800 cm⁻¹ (vertical lines in Fig. 2). The amplitude of the LSPR peak decreases with the doping level corresponding to smaller nano-antenna widths, thus scattering area [19]. Upon PDMS coating, the resonance peak shifts towards lower wavenumbers as a consequence of a change in the refractive index of the surrounding from air n = 1 to PDMS n = 1.508 causes the LSPR shift [19,24]. Besides, an absorption band appears in the reflectance spectra in both transversal and longitudinal directions at 800 cm⁻¹, corresponding to the absorption of CH₃ and Si-C [26], in accordance with the attenuated total reflection measurements and literature [24].

 

Fig. 2 Reflectance spectra of the bare (solid symbols) and PDMS-coated (open symbols) samples from experiment in (a) the transversal, (b) longitudinal direction (compared to the antenna long-axis) and, FDTD simulations (c) transversal (d) longitudinal direction. Dashed dotted line indicates the Si-C bond stretching mode vibration at 800cm⁻¹ of PDMS.

Download Full Size | PDF

We compare these experimental results with numerical FDTD simulations. Figures 2(c) and 2(d) show the FDTD simulations of the bare (dash black line) and coated (color solid line) sample in the transversal and longitudinal direction to the resonators, respectively. The Si-C vibration of the PDMS layer was described by a Lorentz oscillator defined by its electrical permittivity, vibrational frequency, strength and damping. At the absorption line of interest, the permittivity and the Lorentz frequency were set to 2.276 [24] and 1.507·10¹⁴ rad·s⁻¹, respectively. To determine the strength and damping parameter we work on the experimental results of reflectance in the longitudinal direction. We find an oscillator strength of 0.5 and a damping of 3·10¹² rad·s⁻¹ as well-suited fitting values. Same qualitatively behavior is found between simulations and experimental results. Nevertheless, some extensive differences are reported in the reflectance amplitudes and the absorption bands at 800 cm⁻¹ between the FDTD simulations and the reflectance measurements in the transversal direction. Such deviations may originate from (i) considering a perfect antenna with uniform losses and a constant width for all the array and (ii) underestimating the complex coupling between the plasmonic mode of the nano-antenna and the Lorentzian oscillator strength and damping [27].

The vibrational signal of the polymer is patent in the reflectance spectra in both transversal and longitudinal directions to the nano-antennas. In the longitudinal direction, the band in the reflectance spectrum corresponds to the pure absorption of the polymer. In contrast, in the transversal direction such band merge two different effects: (i) the typical absorption from molecular vibrations located far from the near-electric field enhancement and (ii) the enhanced IR absorption due to the coupling to the plasmonic modes in the nano-antennas. Both effects show different optical peak shapes, respectively, a Lorentzian-like feature for purely molecular absorption and an asymmetric Fano-type line in the case of the enhanced vibrational signal [28–30]. For both longitudinal and transversal direction reflectance spectra, we calculate the ratio between the optical response of coated and bare nano-antennas according to Neubrech et al. [31]. These ratios are the baseline correction in transversal (black) and longitudinal (red) directions from experimental data and FDTD calculations are shown in Figs. 3(a) and 3(b). As expected, a Fano-like shape and a Lorentzian peak represent the transversal and longitudinal baseline correction reflectance, respectively. FDTD calculated ratios match qualitatively well the experimental ones. The Lorentzian peaks intensity remains constant while the Fano-like increases monotonically with the doping level.

 

Fig. 3 Baseline correction of the reflectance in transversal (black lines) and longitudinal (red lines) direction from (a) experiment and (b) FDTD calculations. SEIRA signal resulting from the ratio of the baseline correction in transversal and longitudinal direction, from (c) experiment and (d) FDTD calculations. A strong dependence of the vibrational signal on the nano-antenna characteristics is found. Heavier Si-doped and wider resonators (sample A) yields to a higher SEIRA. Simulations show a good qualitative agreement with experiments.

Download Full Size | PDF

Since we are interested in the enhanced IR absorption, we divide both baseline corrections in the transversal and longitudinal direction, so to say the ratio of the Fano-like peaks by the Lorentzian ones (as the inset of Fig. 3(d) indicates). This approach considers the same condition at the same point and then the same thickness fluctuation. Figures 3(c) and 3(d) show the enhanced vibrational signal calculated from experiments and simulations. It can be observed that the SEIRA effect depends on the antenna characteristics: increasing the doping level enhances the vibrational signal of the molecules and consequently, the sensing efficiency. FDTD simulations qualitatively support the experimental results. Therefore, considering the total reflection as 1, we estimate an experimental enhancement of the vibrational signal of the 8.1%, 5% and 2.6% for samples A, B and C, in the areas where the E field is confined. The enhancement calculated from FDTD agrees partly well with the experiment.

To better estimate the SEIRA enhancement factor, it is needed to consider the area where the electric field enhancement occurs. Figure 4 illustrates the electric field distribution of half of the nano-antennas coated with 15 nm-thick layer of PDMS. It can be observed that despite the slight invariance of the absolute electric field enhancement, a clear change in the electric field extension and penetration is found. Sample A supports the electric field more extended through the PDMS layer compared to sample B and C, where the electric field is confined in the corners of the interface with the substrate, as it is observed in Figs. 4(d)-4(f). Indeed, considering the total electric field in the PDMS layer, we find that the electric field is 2- and 3-times larger in sample A than in sample B and C, respectively. This result matches well with the SEIRA reported for every sample. Taking into account the effective extension area of the E field and extrapolating the enhanced IR absorption to such area, we define the SEIRA enhancement factor as the ratio of the absorption enhancement percentage and the effective extension area. We obtain a SEIRA enhancement factor of 2 orders of magnitude which is comparable to the enhancement factor reported for Ge nano-antennas [32]. Figures 4 (f)-4(g) show a zoom of the E field in the bottom corner of the nano-antenna. To obtain a good resolution a mesh of 1nm for FDTD simulation is used. This explains the higher E field in Figs. 4(f)-4(f) compared to that of Figs. 4(a)-4(c). The larger extension of the electric field into the PDMS layer and less penetration into the substrate of sample A compared to that of sample B and C enables to reach the top corners. These hot spots are more sensitive to the surroundings, and then showing a higher SEIRA signal. The electric field is mainly confined in the bottom corners as in sample B and C (Figs. 4(b)-4(c)), and it easily penetrates into the substrate (Figs. 4(e)-4(f) and thus, decreasing the sensitivity to external conditions. These results sustain the experimental SEIRA outcomes. Increasing the doping level promotes the electric field extension through the PDMS layer towards the top corners, improving the sensitivity to the surrounding and therefore, resulting in a higher SEIRA.

 

Fig. 4 FDTD simulation of the E field. (a-c) Half-antenna cross-sectional images of the electric field extension and (d-f) zoom, calculated using 1nm mesh, in the bottom corner hot spot of the three samples. The solid and dash lines delimit the nano-antenna and the PDMS layer, respectively. The doping level clearly influences the electric field extension. The presence of electric field into the PDMS layer becomes more important for heavier doped resonators. The electric field locates in the hot-spot arising from PDMS-surface interface. Its extension into the PDMS layer and its penetration in the substrate depends on the doping level of the nano-antenna. The E field is larger extended into the PDMS layer for heavily doped nano-antennas (sample A), reaching even the top corner hot-spot and thus, making more sensitive the system to surrounding conditions.

Download Full Size | PDF

The coupling of the resonator plasmonic mode and the vibrational signal of the molecule determines the optical properties and helps to define optimal applications for the plasmonic platforms. This coupling can be explained by considering the coupled harmonic oscillator model. In this context, incident light interacts intensively with both: the molecular vibration and the plasmonic mode of the nano-antenna [33]. The coupling between these two modes is classified in three different regimes: weak, intermediate and strong. The main difference of these three coupling situations dwells in the sensitivity to the dielectric environment, being the weak regime the most sensitive and the strong coupling the less sensitive [27]. In the intermediate condition, the electromagnetic energy is stored. This enlarges the electric field enhancement and thus, reaches the optimal configuration for surface enhanced spectroscopy applications [27]. From here, it can be inferred that belonging to the weak region and increasing the coupling, in terms approaching the intermediate region, improves the SEIRA sensing.

To evaluate the coupling of the Si:InAsSb/GaAs 1-D nano-antennas to the PDMS vibrational modes, we solve the two coupled harmonic oscillators problems by using the simplified analytical model proposed by Gallinet et al. [27].

Figure 5 shows the optical response of the three samples: bare (solid scatter), coated with a 15nm PDMS layer (open scatter) and the reflectance using the Gallinet approach (solid line). For Gallinet approach calculations, we assume a symmetric modulation (q = 0) since the mode detuning is small compared to the nano-antenna width, and the ${\omega }_0$ centers at 1.507·10¹⁴ rad·s⁻¹ (corresponding to the PDMS absorption line). ${\gamma }_i$ is considered to be constant (3·10¹² rad·s⁻¹, fitting value extracted from FDTD simulations for the Lorentzian PDMS layer) and ${\gamma }_c$ is adjusted to mimic the experimental curve in every sample. $b$, defined as $\frac{{\gamma }_i^2}{{({\gamma }_c+{\gamma }_i)}^2}$, is calculated in every case. The ${\gamma }_c$ optimal values obtained for the samples A, B and C are ${\gamma }_{c (A)}$ = 1 × 10¹¹ rad·s⁻¹, ${\gamma }_{c (B)}$ = 6 × 10¹⁰ rad·s⁻¹ and ${\gamma }_{c (C)}$ = 3 × 10¹⁰ rad·s⁻¹, respectively. A monotonically variation of ${\gamma }_c$ with the doping level and thus, antenna width is found. It increases with increasing the doping level. In terms, the characteristic of the nano-antennas clearly influences the coupling of the PDMS vibrational mode and the plasmonic mode of the resonator. Increasing the doping level linked to an increase of the antenna width, promotes the coupling between the modes approaching to the intermediate region and thus, making the system more suitable for SEIRA spectroscopy sensing. These analytical results match well with the SEIRA experimental outcomes where, we find that increasing the doping level enhances the vibrational signal of the molecules.

 

Fig. 5 Reflectance spectra of the three samples coated with PDMS (solid scatters), bare (open scatters) and Gallinet fitting (black lines) of (a) sample A, (b) sample B and (c) sample C. Adjusting the damping due to coupling, ${\gamma }_c$, the experimental results are well modelled using the approach proposed by Gallinet et al. [27] although the SPR shift is not considered. The depth of the band arising from the interaction between the vibrational and plasmonic modes directly indicates the coupling strength.

Download Full Size | PDF

Strictly speaking, the radiative damping (${\gamma }_c$), arising from the interaction of the plasmonic and vibrational mode, not only depends on the coupling of these two modes but also on the antenna properties [27]. While for metals, aligning the optical response of plasmonic resonators to a wavenumber of interest using the same material means to engineer the nano-antennas with different geometry, for HDSC involves the setting of two entangled parameters: the doping level and the geometry of the nano-antennas [32,34–36]. The doping level places ${\omega }_p$ in the spectral region and thus, defined the range where by tailoring the nano-antennas the extension and enhancement of the electric field can be controlled [19]. Therefore, the tuning of the LPSR to a certain molecular vibrational mode to demonstrate SEIRA spectroscopy sensing lies on the selection of the set of optimal values for the doping level and the nano-antenna size. In the case of 1-D InAsSb plasmonic resonators this alignment is done in terms of adjusting the doping concentration and the width of the nano-antennas.

In 2013, Boltasseva et al. reported a comparison of the electric field enhancement as a function of the free carriers density in the visible and near-IR. It was assumed the same geometry for the nano-antennas and the maximum free carrier concentration (being in metals the intrinsic electron density and in semiconductors the maximum admitted value for the doping level). From here, it can be inferred that increasing the carrier density decreases the electric field enhancement due to the increases of the losses and a weaker penetration depth. The effect of the size directly influences the electric field confinement. Decreasing the size of the nano-antenna increases the confinement of the electric field and consequently, an increase of the SEIRA signal is expected as it has been reported for gold nano-antennas [11,37,38]. Taking into account these premises, the expected SEIRA signal in 1-D InAsSb plasmonic resonators would increase when decreasing the doping level and consequently, the nano-antenna width. Indeed, we have demonstrated previously that reducing the doping concentrations attenuates the radiative losses and consequently, results in an enhancement of the quality factor [19]. In contrast, we have found an unexpected result concerning the maximum SEIRA signal, higher SEIRA effect for heavier doping level. This can be explained in terms of the influence of the doping level on the electric permittivity and the electric field penetration in the resonators. Our results address a change in the electric field extension with the plasmonic resonator characteristics. Light-doped nano-antennas support a high electric field confinement and large penetration. In contrast, heavy-doped nano-antennas result in a reduction of both the penetration and the confinement of the electric field. As a consequence, the electric field surrounding the resonator approaches the upper hot spots, and thus being more sensitive to external changes. These results show the way to engineer the nano-antennas to obtain the maximum SEIRA signal at a target vibrational mode energy.

4. Conclusion

In conclusion, we have demonstrated an efficient SEIRA spectroscopy sensing with 1-D array InAsSb plasmonic nano-antennas. Engineering the nano-antennas to support the LSPR aligned with the vibrational mode of PDMS involves the adjustment of the two linked parameters: doping level and antenna width. Heavier doping levels requires wider nano-antennas than lighter doped resonators. After coating with 15 nm-thick PDMS layer, the system reacts to the new external conditions showing a LSPR redshift and a band due to the polymer absorption. Increasing the doping level and thus, the nano-antenna width, a SEIRA enhancement factor up to 2 orders of magnitude is found. The higher extension of the electric field presented in the case of highly doped nano-antennas along the PDMS layer favors the enhancement of the molecular IR absorption, and consequently improves sensing to environmental changes. This matches well with the results of the two harmonic oscillator problem. The coupling between the plasmonic and the vibrational modes depends on the nano-antenna parameters. An increment of the doping level enhances the coupling between the modes implying wider nano-antennas. Consequently, the SEIRA sensing is improved. In sum, we demonstrate the feasibility of the system based on 1-D InAsSb plasmonic resonators to manufacture chemical fingerprint sensors in the mid-IR and, we pave the way to engineer the nano-antennas supporting the LSPR at the wavenumber of interest to get the maximum SEIRA sensing.