1. Introduction

Spectroscopic detection of molecular vibrations presents an effective way to measure and identify molecular spices, with a wide range of applications in material science and sensor technologies [1]. Traditional infrared absorption spectroscopy usually requires a sufficient amount of target molecules to ascertain molecular absorption signatures and is generally limited in sensitivity [2,3]. To address this limitation, surface enhanced infrared spectroscopy (SEIRA) has emerged as an attractive option for molecular detection with high sensitivity [4–6]. Using optical antennas and metasurfaces such as metallic rods [7], arrayed disks [8], and split-ring resonators [9], a number of molecular sensors have been demonstrated with field enhancement factor more than 10⁵ and detection sensitivity up to single molecule [10–17]. The high sensitivity in SEIRA depends on local field enhancement associated with plasmonic resonances in nano-structured surfaces, whose fabrication involves complex top-down or bottom-up nano-fabrications. Correspondingly, the controllability and reproducibility of sample manufacturing in SEIRA presents a challenge for large-scale applications.

Planar thin-film cavity, on the other hand, is simple and compatible with standard thin-film deposition techniques. Thin-film structures, made of one or more films of dielectric or metallic materials, have been commonly used in optical coatings [18], and in recent demonstrations of broadband absorbers [19,20], structural colors [21,22], and boosted light absorbance in ultrathin materials [23]. Very recently, Ayas et al. also demonstrated a molecular sensor using a CaF₂ quarter-wavelength cavity [24]. They measured a signal intensity of 2% according to CH₂ vibration in octadecanethiol (ODT) molecules at normal incident. Indeed, the Fabry-Perot resonance in a thin-film cavity depends on both polarization and incident angle, which could be harnessed for sensing. In this paper, taking ODT as a representative target molecule, we demonstrate a thin-film cavity enhanced molecular sensing and exploits its dependence of both angle and polarization. The sensor structure is made of a quarter wavelength thick ZnSe on top of a reflecting metal plane. A maximum signal intensity of 8.54% per monolayer of ODT is measured at 75° angle for s-polarization, which is four times of the value demonstrated in [24] at normal incident and is comparable to those obtained in more sophisticated plasmonic nanostructures. Whereas the structure reported here is simple, without involving lithography or nano-patterning technique, the underlying amplification mechanism operates on both the enhanced surface field in and the coupled-oscillators’ energy transfer between the molecules and the cavity underneath.

2. Design and simulation

The planar thin-film cavity is designed to be a simplest, as illustrated in Fig. 1(a) and consisting of a ZnSe dielectric layer atop a Cu reflecting base. The thickness of the ZnSe is set to be a quarter wavelength to ensure that the field maximum of the Fabry-Perot resonant mode lies on the top surface of the ZnSe where the target molecules immobilize. We take wavenumber of 2918 cm⁻¹ as the target wavelength for field enhancement, which matches the CH₂ vibration mode of ODT molecule, and the corresponding thickness of the Fabry-Perot cavity is thus d = 362 nm. The resonant frequencies of such cavity are determined by the phase relationship given as [25]

 

Fig. 1 (a) Schematic of designed thin-film cavity for ODT detection. Phase shifts upon reflection from (b) ZnSe to air (i.e. α₁₀) and (c) ZnSe to Cu (i.e. α₁₂). Amplitudes of reflection coefficients from (d) ZnSe to air (i.e. |r₁₀|) and (e) ZnSe to Cu (i.e. |r₁₂|) at 2918 cm⁻¹ for different polarizations. (f) Dependence of fundamental resonant mode on incident angle and polarization. Blue and red curves are for s- and p-polarizations, respectively. The bars denote spectral linewidths. The black dashed line marks the CH₂ absorption line of ODT molecules at 2918 cm⁻¹.

Download Full Size | PDF

Figure 1(f) shows the resonant frequency and linewidth of the fundamental mode (j = 0) calculated with Eqs. (1) and (2), as functions of both the incident angle and polarization. As the incident angle increases, for s-polarization, the resonant frequency gradually blue-shifts towards the target absorption line of CH₂ at 2918 cm⁻¹ plotted in black dashed line, falling well within the spectral linewidth of the resonance for all incident angles. For p-polarization, the resonant frequency shows similar blueshift first, but jumps to higher order mode (j = 1) at 5946 cm⁻¹ (out of our interested wavelength range) at the Brewster angle due to the sudden phase change of α₁₀. It is worthy to note that a small difference exists between the resonant frequencies of the s- and p-polarizations for angles less than 68°, which is not visible in Fig. 1(f) due to a large scale in the vertical wavenumber axis. In addition, linewidth of the resonance decreases with increasing angle for s-polarization, while for p-polarization it increases first for angles less than the Brewster angle of the ZnSe-air interface at 68°, and then decreases for larger angles. The quality factor Q for s-polarization is generally larger than that for p-polarization at large angles, which is expected to lead to an enhanced light-molecule interaction with the increased round-trip the resonant light makes inside the dielectric cavity.

The electric field distribution of the resonant mode can be found by using either the transfer matrix method [29] or a finite-difference time domain simulation (Lumerical FDTD Solutions). Results from the analytical matrix method agree with those obtained from FDTD calculations. Figure 2(a) shows the simulated spatial field distributions inside the cavity for the fundamental mode. We can see that for s-polarization, the field enhancement at the cavity top surface maintains a high value of 1.87 for both 0° and 75° angles at the targeted CH₂ absorption frequency, as desired for detecting the molecules. Meanwhile, the cavity’s spectral linewidth becomes narrower when the angle changes from 0° to 75°, consistent with the mode quality factors discussed earlier. However, for p-polarization, the electrical field enhancement drops from 1.87 at 0° angle to nearly zero at 75°, which follows the Fraunhofer’s diffraction relations but renders the p-polarization at large angles unsuitable for ODT detection. Figure 2(b) shows the field enhancement as a function of incident angle for both s- and p-polarizations. It is seen that the electrical field is enhanced by a factor of 1.87 at normal incidence. As the incident angle increases, the field enhancement of s-polarization slightly increases and maintains the value around 1.87 before a sudden drop at the angle around 80°. For p-polarization, however, the field enhancement quickly drops as the angle starts to increase. The field and quality factor analysis suggests that for a greater sensitivity in molecular detection it should operate with incident s-polarization at large angles to benefit from the stronger light-molecule interaction.

 

Fig. 2 (a) Spatial distributions of optical field inside the cavity (d = 362 nm) at different wavenumbers at 0° and 75° angles and for s- and p-polarizations. The electrical field E is normalized to the incident field E₀. The black dashed line marks the CH₂ absorption line of ODT molecules at 2918 cm⁻¹. (b) Electrical field enhancement at top surface of the cavity (air side) as a function of incident angle at 2918 cm⁻¹.

Download Full Size | PDF

For verification and greater insights, we simulated molecular sensing properties of the cavity by using the transfer matrix method. A 2.5 nm thick ODT film is added on the top surface of the cavity. Refractive index of the ODT molecule was taken from [14], whose real part and imaginary part are plotted in Figs. 3(a) and 3(b), respectively. Figure 3(c) shows the calculated reflection spectra of s-polarization. In the presence of the ODT molecules, the system exhibits ODT’s signature absorption peaks at 2850 cm⁻¹, 2918 cm⁻¹, and 2954 cm⁻¹, corresponding to the CH₂ symmetric, CH₂ asymmetric and CH₃ asymmetric vibration modes, respectively. Their absorption amplitudes increase as the incident angle increases. A signal intensity of 8.83%, obtained as the difference between maximum and minimum reflectance at the vibration feature, is observable at 75° angle. As expected, for p-polarization, the ODT molecule absorption is much smaller and its amplitude decreases with increasing angle as shown in Fig. 3(d).

 

Fig. 3 (a) Real part n and (b) imaginary part k of the refractive index of ODT molecules. Calculated reflection spectra of the quarter-wavelength cavity with d = 362 nm for (c) s- and (d) p-polarizations. (e) and (f) are the results of a reference cavity with a smaller thickness of d = 50 nm. The numbers next to ODT absorption lines indicates their signal intensities.

Download Full Size | PDF

It is interesting to observe in Fig. 3(c) the rather large difference between the absorption peaks at low (15°) and large (75°) angles, with the latter being ~360% greater. Recall that the electrical field at the top surface is enhanced by about the same factor of 1.87 at both angles. It hints another factor or underlying mechanism at play. The enlarged optical path at oblique incidence is one likely factor. However, the optical path of light in the ODT layer for 75° incident angle is only 1.3 times of that for 15°. This small optical path difference is insufficient to explain the large molecular detection signal at 75° angle. Indeed, there is also the light energy stored in the cavity under the ODT that can be coupled and transferred to the ODT. This can be better seen by treating the Fabry-Perot resonance of cavity and the ODT molecular vibration mode as two coupled oscillators, whose motions can be described by [30]

By using a trial solution of x_i = C_ie^iωt (i = 1, 2) for x₁ and x₂, the amplitude of the ODT “phonon oscillator” at resonance matching condition ω = ω₁ = ω₂ can be obtained as

Equation (5) derived from the coupled oscillator model can explain the molecular absorption behaviors observable in Figs. 3(c) and 3(d). For s-polarization, as the angle increases, the driving force F₀, proportional to the field amplitude shown in Fig. 2(b), barely changes, and meanwhile the damping rate of the cavity “photon oscillator” γ₁ (proportional to linewidth ∆k of resonant mode shown in Fig. 1(f)) decreases, which leads to a larger |C₂| or an increased energy transfer to the molecule at large angles for s-polarization. For p-polarization, the damping rate γ₁ generally increases with incident angle up to around the Brewster angle. Meanwhile, the significantly dropped driving field F₀ (proportional to the field amplitude shown in Fig. 2(b)) dominates and gives rise to a smaller |C₂|, which explains the decreased molecular absorption as the angle increases. For angles further beyond the Brewster angle, the “photon oscillator” is moved out of the resonant coupling condition and the decreasing surface field leads to a monotonically decreased molecular absorption.

To illustrate the enhancement effect of the thin film cavity in molecular sensing, we also designed a reference sample, a cavity with a much thinner thickness of d = 50 nm, whose resonance is shifted out of our interested wavelength range to at around 1.35 × 10⁴ cm⁻¹. The simulated results for the reference structure are shown in Figs. 3(e) and 3(f). For s-polarization at 15°, the absorption features of ODT are barely seen (less than 0.06%). At 75°, there are no ODT absorption features at all. While for p-polarization shown in Fig. 3(f), the ODT absorption signatures are present and similar to the thicker cavity case in Fig. 3(d) with exception that the absorption features increase with incident angle.

The molecular absorption behavior in the reference cavity case can be understood from the optical field at the cavity top surface, where the ODT molecules are located. Figure 4(a) shows the calculated field at the air-ZnSe interface. For s-polarization, the field is nearly zero for both 0° and 75° angles because of the close proximity to the metal base where the s-polarized field vanishes as dictated by the boundary condition. However, for p-polarization, the field increases from 0.3 at 0° to 1.74 at 75°. This field behavior at the ODT monolayer is bettered viewed in Fig. 4(b). The field decreases from 0.3 to a negligible small value for s-polarization, explaining the extra small absorption features in Fig. 3(e). For p-polarization, the field increases from 0.3 to a maximum value of 1.77 at 71°, which explains the increasing ODT absorption features for p-polarization shown in Fig. 3(f). From the comparison between the resonant (d = 362 nm) and non-resonant (d = 50 nm) cavities, it is also clear that molecular detection sensitivity can be amplified through both enhanced local field and energy transfer from a simple resonant structure that stores energy, e.g. a Fabry-Perot cavity. To keep the structure and fabrication simple, therefore more readily adaptable to large-scale applications, we opted for a simplest design of having just a dielectric layer deposited on a metal base instead of a more sophisticated or more symmetric one of a higher Q.

 

Fig. 4 (a) Spatial distributions of optical field inside the reference cavity (d = 50 nm) at different wavelengths at 0° and 75° angles and for s- and p-polarizations. The electrical field E is normalized to the incident field E₀. The black dashed line marks the CH₂ absorption line of ODT molecules at 2918 cm⁻¹. (b) Electrical field enhancement at top surface of the cavity (air side) as a function of incident angle at 2918 cm⁻¹.

Download Full Size | PDF

3. Experimental results and discussion

For validation of the proposed mechanism of molecular sensing with amplified sensitivity in a simple cavity, we conducted a series of experiments in fabrication and measurements. First, a 300 nm thick Cu was deposited by electron beam evaporation onto a silicon wafer, and then a 363 nm thick ZnSe was evaporated on the Cu base to form the cavity. For comparison, we also fabricated a reference cavity with much thinner ZnSe of 43 nm. The thicknesses of the ZnSe films were measured by ellipsometry on reference silicon samples prepared in the same growth runs of the cavity samples. To add a monolayer of ODT molecules, the two samples were immersed in 1 mM/L ODT (Sigma-Aldrich, >95%) in ethanol solution for 24 hours at room temperature, and then rinsed thoroughly by ethanol and dried with nitrogen gas. In the self-assembling process, the hydrophobic surface of ZnSe favors the formation of monolayer ODT moleclues [31]. Figure 5 shows a photo of the two fabricated samples. The cavity sample size is 1 cm × 1.5 cm and the reference sample size is 1.5 cm × 2 cm.

 

Fig. 5 Photo of fabricated cavity samples with two different ZnSe thicknesses of d = 363 nm and d = 43 nm.

Download Full Size | PDF

Next, we measured the reflection spectra of the two cavity samples using angle-variable Fourier transform infrared (FTIR) reflection spectroscopy. The sample and the detector were mounted on two co-axial rotational stages, which allowed for a continuous scan of the incident angle from 15° to 85°. Figures 6(a) and 6(b) show the reflection spectra of the cavity sample with ZnSe thickness of 363 nm. We can see that for s-polarization at 15°, three absorption peaks of 1.42%, 2.24% and 0.22% are visible at 2850 cm⁻¹, 2919 cm⁻¹ and 2962 cm⁻¹, which correspond to the CH₂ symmetric, asymmetric stretching modes and CH₃ asymmetric stretching mode, respectively. These signal intensities at 15° angle are close to those measured at normal incidence in [24], where a monolayer ODT was immobilized on a 1.5 nm thick Au surface. This comparison suggests the formation of monolayer ODT on our ZnSe cavity samples. As the incident angle increases, the signature absorptions become stronger. At 75°, signal intensities of the three vibration modes reach 5.07%, 8.54% and 0.75%, respectively. These measured maximum fingerprint signal intensities of ODT are four times of the values reported in [24] under normal incidence. It can be attributed to the use of a Fabry-Perot cavity and the improved quality factor of the resonance mode at large incident angles. As the polarization changes to p-polarization as shown in Fig. 6(b), the ODT absorption features are still visible, but with much smaller feature signals. The CH₃ asymmetric stretching mode at 2962 cm⁻¹ is absent for angles larger than 15°. These measured results are in general agreement with the calculated ones presented in Figs. 3(c) and 3(d). Small deviations in absorption positions and relative intensities exist between theory and experiment. In particular, position shifts of 1 cm⁻¹ and 8 cm⁻¹ were observed for the absorptions of CH₂ asymmetric and CH₃ asymmetric stretching modes, respectively. These discrepancies could arise from uncertainties in our used ODT optical parameters. In addition, anisotropy of the ODT monolayer is another likely factor, which was not accounted for in our calculations. In fact, we measured spectra of the cavity samples by changing the sample orientations, and no anisotropic spectral features of ODT were observed. The ZnSe film is amorphous in our structure, which likely causes some randomness in ODT molecule orientations and therefore suppresses anisotropic effects from ODT molecules.

 

Fig. 6 Measured reflection spectra of the quarter-wavelength cavity with d = 363 nm for (a) s- and (b) p-polarizations. (c) and (d) are the results of the reference cavity sample with d = 43 nm. The spectra are shifted vertically for better visualization of the spectral features. The numbers next to ODT absorption lines indicates their signal intensities.

Download Full Size | PDF

Figures 6(c) and 6(d) show the measured results of the reference cavity with ZnSe thickness of 43 nm. The ODT absorption features are absent in the case of s-polarization. From the comparison between Fig. 6(a) and Fig. 6(c), the planar Fabry-Perot cavity obviously enhances the ODT detection for s-polarization. For p-polarization, the spectra of the reference sample in Fig. 6(d) shows very weak absorption signature for 15° incidence and relatively larger signature for 75° incidence. Its signal intensities are slightly smaller than the p-polarized case in resonant cavity sample (d = 363 nm), which agree qualitatively with the simulated results presented in Figs. 3(d) and 3(f).

For a calibration of the sensing performance against the state-of-the art, we compared the measured vibration signal intensity in our simple planar cavity with those reported in plasmonic nanostructures [9–15] as given in Table 1. It is seen that our demonstrated maximum signal intensities are 5.07% and 8.54% per monolayer for the CH₂ symmetric and asymmetric modes, respectively. These values are comparable to those obtained in plasmonic nanostructures. Although the local field enhancement is much smaller in our case, but the coupling to and energy transfer from the cavity below made up the difference. It is also noted that the comparison here does not account for the number of molecules per unit area. In our cavity structure the whole surface is covered by ODT, but in plasmonic nanostructures, only the sharp edges or nano-gaps with much less total surface area are covered.

Table 1. Comparison of measured vibration signal intensities per monolayer ODT between this work and those in literature.

View Table

4. Conclusion

In conclusion, we have demonstrated an amplification mechanism for molecular sensing implementable in a simplest two-layer cavity structure. In a proof-of-concept experiment, the two-layer cavity was made of a quarter wavelength thick ZnSe on top of a reflecting Cu film. While ODT was used as a model molecule for the sensitivity tests, the mechanism, structure design and implementation are extendable to others. The vibrational absorption feature of CH₂ symmetric, asymmetric and CH₃ asymmetric stretching modes in ODT molecule were resolved in both theory and experiment, which can be explained by the combined effect of enhanced surface field with energy transfer between coupled-oscillators. For s-polarization at large incident angles, the measured vibration signal intensity reached 8.54% per monolayer, which is larger than previously reported CaF₂ cavity values in [24] and is comparable to those obtained in nanostructured plasmonic devices [9–15]. In comparison, with the detection amplification built in, our detection structure can be much simpler and is compatible with standard thin-film coating technologies, free of lithography and patterning, and thus advantageous in fabrication and large-area applications. Differentiating from a standard amplifier, the amplification mechanism here is self-powered and harvests the energy that is not yet absorbed by the molecules and delivers it back for re-absorption. Also differentiating is the fact that it is a spectral selective amplification that amplifies the signal/noise ratio instead of the whole band. It is adaptable to other designs, e.g. more symmetric and high-Q cavities [32], and to other spectral band for sensing a wide range of molecules including organic compounds and biological species.