Photothermal spectroscopy on-chip sensor for the measurement of a PMMA film using a silicon nitride micro-ring resonator and an external cavity quantum cascade laser

2.1 Fabrication

The air cladded MRRs were fabricated using a thermally oxidized 4″ bulk silicon wafer (BOx thickness of 2.2 µm) with a Si₃N₄ layer (300 nm thick), deposited with plasma-enhanced chemical vapor deposition (PECVD). First, a layer of ZEP 520A resist (∼450 nm thick) was spin-coated on the wafer. The device layouts were patterned on the resist by means of 100 kV electron beam lithography (EBL) and developed in a bath of n-amyl acetate solution for 90s and rinsed with IPA. The patterns were transferred to the PECVD Si₃N₄ layer through inductively coupled plasma (ICP) etch step with CHF₃:O₂ chemistry in a 21:4 ratio (etch rate of ∼1.5 nm/s). The residual resist was later removed through an O₂ plasma ashing step and a bath in 1165 Remover for 10 min. For the analyte deposition, a layer of PMMA-A4 (160 nm) was spin-coated on top of patterned Si₃N₄ layer. The thickness of the PMMA film was achieved following the spin curve of the polymer and experimentally measured with a profilometer. Among the different MRRs design and geometry, better described in one of our work [35], [36], the one reported in the Scanning Electron Microscope (SEM) image (Figure 1A and B) have been employed to perform the here-presented PTS experiment. The ring presents four partially transmitting elements (PTE) aiming for enabling the realization of a Fano-line shape, steeper, and sharper to maximize the sensitivity of the PTS signal [37]

Figure 1:

SEM images of the MRR employed for the PTS experiment. (A) The ID code indicates respectively: the gap of 400 nm in the coupling region between the straight WG and the ring itself, the number of PTE in the coupling region, in this case 4, the ring radius (33 µm); the S indicates the shape of the PTE, in this case, rectangular slot. (B) SEM image of a magnification of the coupling region with partially transmitting elements.

2.2 Experimental setup and measurement procedure

For optical characterization of WGs, an End-Fire (EF) setup [38] was home-built. It allows for butt-coupling the light to one facet of the WG and retrieves the light in output from the other facet. The setup includes a polarizing beam splitter that transmits only TE-polarized light to avoid optical higher order modes propagation. The sample holder is an aluminum block that includes a Peltier element for temperature stabilization (20 °C). An Optical Spectrum Analyzer (Yenista Optics, OSA20) 20″ was used to acquire the device’s transmission spectrum in a range between 1560 and 1610 nm. The highest sensitivity allowed from the instrument was employed for spectrum trace acquisition (0.5 nm/s). A NIR photodiode sensor (Thorlabs, S155C) was used to measure the output power. The alignment was performed using an InGaAs camera (Electrophysics, MicronViewer 7290A) for monitoring of light propagation along the WG.

One resonance was picked from the transmission spectrum of the WG, and the inflection point was roughly found from OSA trace. The probe laser is a tunable laser (Yenista Optics TLS, tunable laser between 1500 and 1630 nm) tuned at the inflection point. The wavelength was finely adjusted while performing PTS experiment using a resolution of pm range in fine-tuning mode.

To perform the PTS measurement, depicted in Figure 2, the NIR imaging system, prior used for butt-coupling, was replaced by a visible camera (Thorlabs, DCC1645C) in combination with a beam-combiner (Thorlabs, CM1-BP145B4 45:55 (R:T), 3–5 µm). In this way, it was possible to align the mid-IR laser (DRS Daylight Solutions, MIRCAT) on top of the MRR to which the probe was coupled. This is made possible by means of a visible laser embedded in the mid-IR source head. The red laser was positioned at the center of the ring resonator. Due to a possible mismatch between the alignment laser and the proper IR beam, finer adjustment of the beam position on top of the ring was made along the x–y–z axis while performing PTS experiment.

Figure 2:

EF variation to perform PTS experiment. The NIR camera is substituted by visible imaging system, used to align the mid-IR excitation laser on top of the ring via visible laser embedded in EC-QCL head.

The collimated mid-IR beam was focused on the MRR by means of a Cassegrain reflector (Thorlabs, LMM25X-P01) with a magnification of 25×, a numerical aperture N.A. of 0.4, and a focal length f of 8 mm. The reflective microscope objective ensured the excitation laser beam radius ω _e to be shrunk (measured w _0e at the focal point ∼16 µm) such that the mid-IR radiation could be concentrated on top of the MRR. This in turn maximizes the temperature rise in the PMMA deposited on the responsive region: probe–pump beam volume of interaction.

The probe output signal from the WG was sent to a fiber-coupled InGaAs photodetector (Thorlabs, DET08CFC/M). The signal from the detector was sent as an input to a LIA (Anfatec, USB LockIn 250), which demodulates it, with a time constant of 100 ms, taking the internal pulse trigger from the mid-IR laser as a reference. The raw demodulated output from LIA constitutes the actual measured PTS signal.

The excitation source was operated in pulsed mode, with a forward sweep scan resolution of 5 cm⁻¹/step. Pulse width and rate were experimentally optimized, as described in “Results and discussion” section.

The magnitude of the signal vector of the LIA demodulated PTS signal was normalized to the mid-IR source optical power spectrum profile recorded with a mercury cadmium telluride (MCT) detector (DRS Daylight Solutions, Amplified MCT detector) at the sample position. This allows to reconstruct the PMMA absorption spectrum for evaluation of peak shape and positioning. A Savitzky–Golay filter was used to remove a prominent water vapor contribution. See the recorded spectra reported in the “Results and discussion” section.

3.1 The transduction scheme

In this work, a Si₃N₄ MRR is employed as a transducer. It is important to underline that the design/geometry optimization and the theoretical physical background of ring resonators are out of the scope of this study. More details can be found in one of our works [36]. The main goal of our work is to demonstrate that Photonic Integrated Circuits (PICs) can be exploited and further explored for PTS sensing in a compact format, sensitive sample detection, and fast response time, using a MRR as an example. Similar conclusions and comments also apply to other waveguide-based transducers, such as Mach–Zehnder Interferometers (MZIs), Photonic Crystal Cavities (PhCs), surface-plasmon resonance (SPR) waveguides [39], [40], etc., depending on the corresponding working principle.

In this scenario, using a MRR as a transducer, the detection variable is the resonant wavelength shift Δλ _Res. By tuning the probe laser at the IP, we aim for retrieving highly sensitive shift of the resonant wavelength photoinduced by the periodical heating and cooling of the MRR, as a function of excitation laser power P ν ̃ . The transduction scheme is depicted in Figure 3.

Figure 3:

Transduction scheme. The probe laser is tuned at the inflection point, aiming for highest sensitivity, linearity, and PTS signal maximization. The excitation source is enabled to sweep, inducing a λ _Res, due to temperature change ΔT (proportional to the absorption of the analyte α) and to a consequent overall effective refractive index change Δn _eff.

After the MRR is covered with PMMA, the refractive index of the cladding n _clad changes from n _air = 1 to n _PMMA = 1.48 and a Δλ _Res occurs.

Once we perform a PTS experiment, the PMMA-cladded WG resonances’ position become a baseline and a new shift will occur as soon as the excitation source is enabled, and a temperature change happens consequently.

3.2 The photothermal spectrum recorded with MRR

Using EF setup, the transmission spectrum of the MRR was recorded. One resonance was picked for performing the PTS experiment (λ _Res = 1593.4 nm) and the IP was roughly identified from the OSA acquisition. The probe laser was coupled to the WG and tuned to the selected IP. By removing the NIR camera and sliding in the Cassegrain reflector system, the PTS measurement could be performed. Before sweeping the EC-QCL, the mid-IR laser was tuned at the absorption peak of PMMA (∼1730 cm⁻¹) for signal optimization. It consisted in (I) finding the optimal excitation pulse rate and pulse width, (II) fine adjustment of the probe laser wavelength at the IP, and (III) fine adjustment of the mid-IR beam position on top of the ring by changing the x–y–z position of the Cassegrain after switching from visible alignment laser to IR beam. An optimal duty cycle of 1.25 % was found for the first optimization, meaning a pulse rate of 25 kHz and a pulse width of 500 ns, the highest allowed from excitation source settings (see Figure 4A and B). The optimal duty cycle corresponds to the full device thermal relaxation time to reach the equilibrium temperature. The different layers get heated up during the mid-IR laser pulse and a longer duration time is needed to bring the MRR to the initial temperature before enabling for the next excitation pulse.

Figure 4:

PTS excitation laser tuning for optimal pulse width and pulse rate. (A) PTS signal at PMMA absorption peak (∼1730 cm⁻¹) recorded by fixing the repetition rate at 50 kHz and tuning the pulse width. (B) PTS signal at PMMA absorption peak (∼1730 cm⁻¹) recorded by fixing the pulse width at 500 ns and tuning the pulse rate.

Using the optimal parameters and alignment, we recorded 10 PTS spectra using forward sweep scan mode with a resolution of 5 cm⁻¹/step and subsequently averaged (acquisition scan ∼ few seconds each). Using an MCT detector, we recorded the optical power spectrum of the excitation laser at the sample position, immediately before sample interaction. The PMMA absorption feature was reconstructed, by normalizing PTS signal to MCT acquisition, and compared against FTIR spectrometer (PerkinElmer Spectrum 400) absorbance spectrum. In Figure 5A, the spectra are reported. Water vapor lines are pronounced in both acquisitions, and they are overlapped across the whole spectral range. When performing normalization between the PTS and the power spectrum, a Savitzky–Golay filtering was additionally performed to remove water artefact. The qualitative absorption spectrum of PMMA, obtained via photothermal spectroscopy on-chip sensor, is depicted in Figure 5B. The PMMA absorption peak is at about 1730 cm⁻¹ and it is in good agreement with the reference one retrieved with commercial FTIR (see Figure 5C). The reference measurement was performed by using a Si₃N₄ air-cladded wafer piece as a background and a Si₃N₄ PMMA-coated wafer.

Figure 5:

Mid-IR spectra of PMMA recorded with PTS on-chip setup. (A) In orange, the raw LIA PTS signal of the PMMA-coated Si₃N₄ MRR (averaged of 10 scans). In dark green, the optical power spectrum of the EC-QCL between 1580 and 1770 cm⁻¹ recorded with an MCT detector at the sample position. (B) Normalized PTS signal obtained by division of the orange and the green acquisition. (C) FTIR reference spectrum of PMMA-coated Si₃N₄ MRR.

3.3 The photothermal effect in the MRR

The periodical heating and cooling generated by the excitation laser in the sample can be translated into a change of the overall effective refractive index Δn _eff of the WG defined as follows [41]:

where Γ _x represent the confinement factors of the guided mode in the substrate (SiO₂), in the core (Si₃N₄), and in the cladding (PMMA), respectively, δ n δ T x are the thermo-optical coefficients (TOCs) of the different materials, and δT _x are the temperature changes in each of the sections, where also the absorption α ν ̃ contribution is taken into account. However, in such a complex configuration, it is difficult to predict the overall temperature gradient since each layer has different thickness, thermal properties, absorption coefficient, and in turn an optical power decay of excitation due to absorption from a layer to the following must be included as well. The excitation source, with a Gaussian radial profile, is focused on the center of the ring, as optimally aligned in the experimental measurement. The sensitive area is at the ring annular perimeter, where probe and pump beam lasers overlap. The optical intensity I r of the beam with an optical power P can be described with a Gaussian function I r = 2 P / π ω e 2 e − 2 r 2 / ω e 2 . Figure 6 shows the optical intensity at the ring position (r = R = 33 μm) as a function of the excitation beam radius (ω _e ). Assuming that the beam is centered, it is observed that the maximum optical intensity at the ring position occurs for ω e = 2 R ∼ 46.7 μ m . Therefore, in this condition, a maximum temperature variation is obtained within the ring and surroundings, maximizing the amplitude of the PTS signal. After the excitation pulse, the heat absorbed mainly by the PMMA and partially by the core (Si₃N₄) and the substrate (SiO₂) diffuses throughout the sensor, returning to the thermal equilibrium condition. During the measurement, as described in the section “The photothermal spectrum recorded with MRR,” optimization of the PTS (III) consisted in adjustment of the Cassegrain system position on top of the ring along the three axes. By moving the z position, the beam width changed consequently being off-focus, and a maximized PTS signal was achieved. Figure 7 depicts the interaction region between the mid-IR excitation laser beam and the MRR when operating at the maximum optical intensity at the ring position. For an excitation beam with a radius much smaller than the ring, the center of the ring will have a high temperature, but the WG region will not be heated considerably. If the excitation radius is much larger than the ring radius, the light intensity is spread over a large area, heating a region much larger than the ring. The optimized condition occurs for the excitation beam radius equal to 2 of the ring radius (see Figures 6 and 7).

Figure 6:

Optical intensity [W/m²] at the ring position (r = R = 33 μm) as a function of the excitation beam waist radius.

Figure 7:

The pulsed mid-IR excitation beam arrives collimated at the Cassegrain reflector. The Gaussian beam penetrates the sample. The optimal beam radius condition is achieved when the Cassegrain position is moved along the z-axis from ω _0e (at the focal length f) to ω e = 2 R (off-focus).

A resonance shift dλ _Res can be approximately estimated following eq. (4) [42]:

The approximated shift dλ _Res due to the PTS thermal change in the sample is in the order of few pm. It mainly depends on the refractive index change Δn that occurs because of the thermal change δT _x . It scales with the TOCs slope of each layer. The overall change in the WG translates in an effective refractive index change Δn _eff (see eq. (3)). An increase in temperature also leads to an expansion proportional to the thermal expansion coefficient of each material α _THx due to the respective δT _x . Eq. (4) takes into account both the optical change of the guided mode and the physical thermal expansion photoinduced by PTS in the WG. The thermal change can be seen as a weighted contribution of involved TOCs of the WG layers. In fact, the radiation, arriving from the top, reaches the PMMA layer that has a certain thickness L ₁ (160 nm). The intensity gets absorbed and decay exponentially due to absorption within L ₁ ( I 1 = I 0 e − α PMMA L 1 ₎. Since PMMA has a negative TOC, the contribution due to increasing temperature in the cladding shifts the resonance toward blue. The remaining optical intensity that reaches the core, immediately below L ₁, slightly heats up the 300 nm (L ₂) Si₃N₄ as well ( I 2 = I 1 e − α S i 3 N 4 L 2 ), leading to a resonance shift to the opposite direction (red) due to the positive TOC of Si₃N₄. The same effect applies to the substrate based on the residual intensity. It is important to address that the WG materials, Si₃N₄, as well as the substrate (SiO₂ and Si), have an optical absorption coefficient α much lower than the absorption peak of PMMA (at 1730 cm⁻¹) and can be neglected (see Table 1). The absorption coefficients α have been taken from database available in literature [43]. The PTS method presented in this work allows qualitative measurement of the absorption spectrum of the thin film in contact with the MRR. However, it cannot provide a direct absolute absorption coefficient measure. The PTS signal is proportional to the absorbance, meaning that the optical absorption could be obtained only after validation against reference methods and sensor calibration. A critical PMMA thickness L ₁ could even lead to an athermal condition where the blue shift from the analyte is fully compensated by the red shift from the core. In this scenario, no detectable shift could be measured. Temperature insensitive devices’ design is widely exploited when PICs thermal shifts must be avoided and either negative TOCs for the substrate are used or certain polymers with negative TOCs are coated on top of the WG by thickness optimization [42], [44], [45], [46].

3.4 Normalized Noise Equivalent Absorption coefficient

Apart from the excitation source parameters’ optimization, what matters the most in PTS sensitivity enhancement is the shape of resonance used in the transduction scheme, the greater the steepness, and the more sensitive the measurement. Moreover, the choice of materials and their thermal sensitivity, meaning their TOCs, must be cleverly engineered to promote target analyte excitation amplification.

Considering a polymer, with concentration c = 100 %, a noise equivalent thickness (NET) can be estimated as a figure of merit. It could be expressed as the ratio between the PMMA thickness L ₁ (160 nm) and the calculated signal-to-noise-ratio (SNR) of the system, considering the 1σ retrieved from the averaged replicates of the PTS spectra acquisition. A minimum detectable absorption coefficient α _min is evaluated as:

A Normalized Noise Equivalent Absorption coefficient (NNEA) can be estimated as:

The obtained NNEA (1730 cm⁻¹) is found to be 6.11 × 10⁻⁶ cm⁻¹ W/Hz^1/2. The achieved value is comparable to the one obtained in Vasiliev et al. work [30]. It is worth noting that in our case, dealing with Si₃N₄ as a core, rather than Si [30], less power limitations due to thermal nonlinearities are encountered meaning that NNEA value could be further improved even increasing the probe laser power aiming for higher SNR. For a summary of the most relevant parameters of the MRR and of the PTS experiment, please refer to Table 1.