Integrated optical sensor platform for multiparameter bio-chemical analysis

Peter Lützow; Daniel Pergande; Helmut Heidrich

doi:10.1364/OE.19.013277

1. Introduction

Integrated optical sensors have been extensively investigated during the last decades [1]. They have been shown to exhibit extraordinary sensitivity in chemical as well as biochemical analysis [2]. They feature the potential for mass production on wafer-level and ease of fabrication. Most detection schemes operate label-free, thereby reducing costs and excluding all adverse effects that may be introduced by fluorescent, enzymatic or radioactive labels. Furthermore, strong miniaturization allows for high level integration of complex sensor functions into portable and low cost handheld devices.

MRR are among the most promising transducers in miniaturized integrated optical sensors [3–5]. Usually, analyte detection is based on the measurement of shifts in resonance frequency caused by local changes in ambient refractive index. A chemically selective adlayer is used to specifically promote the accumulation of target molecules on the MRR’s surface leading to an increase in local refractive index with analyte concentration. Highly sensitive and label-free detection of molecular species with concentrations down to the ppb-level has been demonstrated in gaseous [6] environments. MRR have been integrated with microfluidics and shown to be suited even for on chip spectroscopy applications [7]. Depending on the specific application and the receptor coating employed MRR sensors may be operated for reuse or single use only. Alignment tolerant optical coupling schemes [8, 9] enable automated optical interfacing and will allow even for complex integrated optical chips to be used as consumables.

However, little effort has been made so far to address issues related to cross-sensitivity and selectivity. In order to reliably identify a target molecule and to minimize false positives, assay tests including functionalized binding sites as well as reference and control elements with a preferably high level of redundancy are needed. Furthermore, many applications require the simultaneous detection of a certain group of target molecules making multiparameter analysis in highly multiplexed sensor configurations necessary.

With the appropriate multiplexing scheme integrated optical sensors based on MRR can become high value assets in both, fundamental and applied research, as well as commercial sensor applications. We believe that the multiplexing scheme [10] presented in the following will achieve this goal.

2. Concept

The transmission spectrum of a single MRR evanescently coupled to a single bus waveguide, as shown in the inset of Fig. 1(a), is given by [11, 12]:

T (ϕ) = \frac{a^{2} + {| t |}^{2} - 2 a | t | cos (ϕ)}{1 + a^{2} {| t |}^{2} - 2 a | t | cos (ϕ)}

Here, a is the field attenuation coefficient, t is the field self-coupling coefficient and

ϕ = N \frac{2 π}{λ} L

is the optical phase shift per roundtrip. N and L are the effective index of the MRR waveguide mode and the circumference of the MRR, respectively. λ denotes the wavelength. Figure 1(a) shows the calculated transmission spectrum for an MRR with a = 0.964, t = 0.868, N = 1.5 and L = 2π 100 μm. Resonances appear in the transmission spectrum as dips resulting from destructive interference. Sufficiently far from resonance light passes the coupling section between bus waveguide and MRR almost unaffected. Therefore, multiple MRR can be coupled to a common bus waveguide. Resonances belonging to different resonance orders are spectrally separated by the free spectral range (FSR). The spectral window defined by the FSR together with the linewidth of the resonances determines the theoretical maximum of MRR that can be fed by the same bus waveguide without significant spectral overlap between their resonances. In this context, a suitable figure of merit is the finesse of a resonator defined as the quotient of FSR and linewidth. Values of over 100 for the finesse have been reported in literature [13].

Fig. 1 Properties of frequency-modulated MRR. (a) Simulated resonance spectrum of a single MRR. Resonances belonging to different resonance orders are spectrally separated by the free spectral range (FSR). The linewidth is defined as the full width at half maximum (FWHM) of a resonance dip. The inset shows a schematic of a MRR with racetrack shape coupled to a single bus waveguide. (b) Simulated resonance dips of two exemplary MRR with similar resonance frequencies (red and blue lines). If the two MRR are coupled to the same bus waveguide the overlapping resonance profiles lead to the shaded transmission spectrum. The dashed green line represents the calculated and normalized lock-in signal for the MRR with the red resonance profile being modulated. The signals are plotted against the wavelength difference with respect to the resonance frequency of the modulated MRR. Units are given in terms of its unmodulated linewidth Δλ, simulation parameters are stated in the text.

Download Full Size | PDF

Unavoidable fabrication tolerances lead to an uncertainty in the exact resonance frequencies of a fabricated device. Therefore, even with a number of MRR far smaller than the theoretical maximum spectral overlap becomes a challenge.

In Fig. 1(b) the blue and red curve depict the resonance dips of two independent MRR having similar resonance frequencies. If coupled to the same bus waveguide the spectra superimpose resulting in the shaded transmission spectrum. The tracking of individual resonance frequencies becomes unfeasible. The dashed green curve in Fig. 1(b) shows the lock-in signal |F_LOCK–IN(λ)| of the shaded spectrum f (t, λ), where F_LOCK–IN(λ) is idealized as

F_{LOCK–IN} (λ) = \frac{2}{T} \int_{0}^{T} dt sin (ω t) f (t, λ),

for the case that one of the MRR (red spectrum) is modulated by sinosoidally changing it’s effective index with a frequency ω/(2π). Individual MRR parameters used in the calculation behind Fig. 1(b) are the same as used before. The resonance spacing and modulation amplitude are approximately 36 % and 0.7 % of Δλ, respectively.

The lock-in signal at the modulation frequency is proportional to the modulation amplitude, impressed on the optical signal, and depends on the shape of the modulated MRR’s own spectrum. The signal is large near the points of inflection of the resonance dip and vanishes at the minimum. The vanishing of the modulation amplitude allows the tracking of resonance frequencies even in dense optical spectra:

The combined spectrum Fⁿ(λ) of a group of n MRR coupled to the same bus waveguide is the product of the spectra of the individual MRR.

F^{n} (λ) = f (N_{R_{1}}, λ) \cdot \dots \cdot \cdot f (N_{R_{i}}, λ) \cdot \dots \cdot f (N_{R_{n}}, λ)

All waveguides are supposed to be mono-mode with effective index N _R₁ to N_{R_n} for MRR 1 to n, respectively. A modulated MRR’s individual resonance spectrum may be described as

f_{MOD} (N, λ) = f (N + Δ N, λ)

Assuming that the change in effective refractive index ΔN induced by the modulation is small, the first two terms of a Taylor series expansion with respect to N represent a good approximation to f_MOD (N, λ):

f_{MOD} (N, λ) = f (N, λ) + \frac{d}{dN} f (N, λ) Δ N + 𝒪 (Δ N^{2})

With Eq. (6) the combined spectrum with MRR R_i modulated becomes:

\begin{array}{l} F_{MOD : R_{i}}^{n} (λ) & = & F^{n} (λ) \\ + & f (N_{R_{1}}, λ) \cdot \dots \cdot \frac{d}{{dN}_{R_{i}}} f (N_{R_{i}}, λ) Δ N \cdot \dots \cdot f (N_{R_{n}}, λ) + 𝒪 (Δ N^{2}) \end{array}

For ΔN being a harmonic function of time, e.g. ΔN = ɛcos(ω t), spectral filtering at a frequency ω/(2π),i.e., lock-in detection, gives

F_{LOCK–IN : R_{i}}^{n} (λ) = ɛ \cdot f (N_{R_{1}}, λ) \cdot \dots \cdot \frac{d}{{dN}_{R_{i}}} f (N_{R_{i}}, λ) \cdot \dots \cdot f (N_{R_{n}}, λ) + 𝒪 (Δ N^{3})

Considering that the spectrum of a MRR depends on the effective index only via the optical phase (cf. Eq. (2)), some algebraic manipulation yields the final result:

F_{LOCK–IN : R_{i}}^{n} (λ) = ɛ \cdot (- \frac{λ}{N}) \cdot f (N_{R_{1}}, λ) \cdot \dots \cdot \frac{d}{d λ} f (N_{R_{i}}, λ) \cdot \dots \cdot f (N_{R_{n}}, λ) + 𝒪 (Δ N^{3})

F_{LOCK–IN : R_{i}}^{n} (λ)

is proportional to the derivative of f (N_{R_i,λ}) with respect to λ. Since d/d λ f (N_{R_i}, λ) vanishes on resonance, so does

F_{LOCK–IN : R_{i}}^{n} (λ)

. More importantly

F_{LOCK–IN : R_{i}}^{n} (λ)

changes sign at the resonance frequency of MRR i. Therefore, by taking the phase change into account, the correct position of the resonance frequency can be identified even if the optical signal reaches zero elsewhere.

Individual MRR, each of which transducing the signal of a well defined target species, may be modulated sequentially with the same frequency or in parallel using a distinct modulation frequency for each MRR. In the latter scheme the separation between modulation frequencies has to be chosen larger than the inverse of the measurement time in order to be able to accurately resolve each spectral contribution. It should also be noted that in order to avoid higher orders of modulation frequencies to disturb the signals, the maximum frequency should be less than two times the minimum frequency.

3. Design and fabrication

The MRR arrays presented here are composed of four MRR evanescently coupled to a single bus waveguide. With the underlying multiplexing scheme we target sensor applications involving hundreds of MRR. All MRR are equipped with individual heating electrodes for thermo-optic modulation. Figure 2 shows fabricated test structures in three different levels of magnification. A part of a wafer can be seen in the background of the figure. The lower inset shows a microscope image of an exemplary MRR with metal electrodes for thermal modulation. In the scanning electron microscopy (SEM) image in the upper inset a detailed view on waveguide and metal strip is given.

Fig. 2 Fabricated test structures. Background: 4” Si Wafer with MRR arrays. Lower inset: Microscope image of an exemplary MRR. The waveguides are seen as thin lines surrounded by metal structures. Upper inset: SEM image giving a more detailed view on part of the MRR waveguide and metal heater.

Download Full Size | PDF

The optical waveguides forming the MRR array were etched into a 220 nm thin layer of silicon nitride with refractive index (n) of 1.91 deposited by ICP-PECVD using a gas mixture of SiH and NH₃. The 5 μm thick buffer layer of silicon oxide (n=1.46) that optically isolates the waveguide core layer from the silicon handle wafer was also etched by approximately 80 nm during this etch step. After platinum electrodes (90 nm Pt on a 10 nm thick adhesion layer of Ti) had been defined in a lift-off process a second layer of silicon nitride with a thickness of 95 nm was deposited. The gap between waveguides and 10 μm wide metal heaters was fixed to 4 μm in order to guarantee efficient heat transfer without inducing additional optical loss due to metal absorption. The electrical contacts were exposed by partially removing the second silicon nitride layer. All etch steps were performed by reactive ion etching in CF₄ atmosphere. Standard contact lithography patterned photoresist was used for the etchmasks. The MRR have racetrack shape with a radius of 98 μm and straight sections of 20 μm. The coupling distance is 0.9 μm. A waveguide width of 1.1 μm ensures mono-mode operation for TE-polarization at telecommunication wavelengths around 1,550 nm. The actual dimensions may differ by around ± 10 % with respect to the nominal values.

4. Experiment

Light emitted from a tunable laser source was launched into the bus waveguides with a lensed optical fiber (LF). An in-line polarization controller was used to ensure TE polarized optical output from the LF. During characterization the chip was placed on a temperature stabilized stage. At the output side of the chip light was collected with a high numerical aperture microscope objective. The signal was divided by an optical beam splitter and sent to both, an infrared camera and a photodetector connected to a lock-in amplifier. Heating electrodes were electrically contacted using manual probeheads and the electrical signal was supplied by a function generator.

Figure 3(a) shows the measured transmission spectrum of a MRR array. The wavelength scan from 1,553.3 to 1,555.5 nm covers about 108% of the free spectral range. The scan resolution is 10 pm. The resonances are randomly distributed despite of their nominally identical layout and only three dips can be identified. The spectra of two out of the four MRR show strong overlap and cannot be separated. Moderate quality factors of around 8000 are estimated for the two outermost resonances in the spectrum. In order to isolate the contributions from the individual MRR to the combined optical spectrum in Fig. 3(a) we applied the modulation scheme as discussed above. Subsequently, MRR one to four (cf. Fig. 3(b)) were modulated with a frequency of 6 kHz. The effective modulation power was 36 mW. The acquired traces are shown in Fig. 3(b). As expected, sharp dips in the lock-in signals mark the resonances. A zero crossing cannot be observed due to the noise in the system and the limited resolution of the wavelength scan in the experiment. The lock-in signals from MRR three and four are smaller by a factor of about 3-4 compared to the signals from MRR one and two. Due to the strong overlap between their resonance profiles the lock-in signals from MRR three and four are partly suppressed by the near presence of the respective other MRR (cf. Eq. (9)). Nevertheless, the resonance frequencies can be clearly distinguished as seen in Fig. 4. This figure shows a magnified view on the relevant part of the lock-in amplitudes for MRR three and four overlaid by their respective relative lock-in phase. Both lock-in phases change abruptly by 180° at the respective minima of the lock-in amplitudes. As discussed before, this sudden change in lock-in phase is a very precise measure for the resonance wavelength of the modulated MRR. The phase information is therefore extremely useful to accurately identify the exact resonance wavelength. In order to compare the lock-in-signals with the unmodulated transmission spectrum we can use the fact, that according to Eq. (9), the derivative of this spectrum with respect to wavelength is proportional to the phase-corrected sum of the lock-in traces. Figure 5 shows the result (derivative of the original transmission spectrum with respect to wavelength together with the phase-corrected sum of the lock-in traces). Both data have been independently normalized to unity. The phase-corrected sum is shifted to higher wavelength as compared to the original transmission spectrum. This is due to effective heating of the MRR during modulation. The mean blue shift of the phase-corrected sum relative to the original transmission spectrum is 28 pm. This value corresponds well to the expected 30 pm deduced from calibration measurements.

Fig. 3 Transmission spectra. Part (a) of the figure shows the transmission spectrum of four unmodulated MRR. Part (b) depicts the lock-in traces for the same MRR subsequently being modulated. The measured data is represented by the colored dots. Lines are guides to the eye. The inset is a schematic of the MRR array with heating electrodes (yellow) establishing the nomenclature used in the text. The different colors of the MRR refer to the colors of the respective lock-in traces.

Download Full Size | PDF

Fig. 4 Phase response. Zoom-in on lock-in signals for MRR three and four. Also shown is the lock-in phase for both MRR.

Download Full Size | PDF

Fig. 5 Comparison between modulated and unmodulated transmission spectra. Blue dots are the result of a phase-corrected sum of the lock-in traces for MRR one to four subsequently being modulated (S_LI). The light gray background corresponds to the derivative of the unmodulated spectrum with respect to wavelength (d/dλ F). S_LI is blueshifted with respect to d/dλ F due to the mean resistive heating effect of the modulation.

Download Full Size | PDF

5. Conclusion

We have proposed a simple, efficient and reliable scheme for the tracking of resonance frequencies in transmission spectra originating from a multiple of microring resonators coupled to a single bus waveguide. The presented measurement technique is based on distinct frequency modulation of each resonator combined with lock-in detection. We have introduced the basic theory and shown experimental proof of concept. The technique works well even in case of strong overlap between the resonances of different microring resonators. Furthermore, profiting from the additional noise reduction resulting from lock-in detection we expect great potential for all kinds of applications related to integrated optical sensing. In particular we target massively multiplexed label-free detection of biomolecular compounds.

Further investigations have to focus on the exact determination of the maximum number of resonators that can efficiently be operated with one bus waveguide. The value will depend on resonator parameters and fabrication tolerances as well as the desired limit of detection (LOD) and dynamic range of the sensor. While the LOD in multi-MRR arrangements is not expected to differ significantly from the single-MRR case [14,15] as long as spectral overlap is moderate, increasing spectral overlap will influence the signal to noise ratio (SNR) and eventually limit the LOD. Special attention will have to be paid to how the interplay between extinction ratio of the resonances and spectral overlap influences the LOD.

Acknowledgments

The authors acknowledge fruitful cooperation with H. J. W. M. Hoekstra (University of Twente) and H. Venghaus. The work was performed within the MINIMUM project (Grant-No: 101 47 221), partly funded by the Investitionsbank Berlin (IBB) and the European Regional Development Fund (ERDF).

References and links

1. V. M. N. Passaro, F. Dell’Olio, B. Casamassima, and F. De Leonardis, “Guided-wave optical biosensors,” Sensors 7, 508–536 (2007). [CrossRef]

2. A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala, “Label-free, single-molecule detection with optical microcavities,” Science 317, 783–787 (2007). [CrossRef] [PubMed]

3. C.-Y. Chao and L. J. Guo, “Biochemical sensors based on polymer microrings with sharp asymmetrical resonance,” Appl. Phys. Lett. 83, 1527–1529 (2003). [CrossRef]

4. C.-Y. Chao and L. J. Guo, “Design and optimization of microring resonators in biochemical sensing applications,” J. Lightwave Technol. 24, 1395–1402 (2006). [CrossRef]

5. V. M. N. Passaro, F. Dell’Olio, and F. De Leonardis, “Ammonia optical sensing by microring resonators,” Sensors 7, 2741–2749 (2007). [CrossRef]

6. R. Orghici, P. Lützow, J. Burgmeier, J. Koch, H. Heidrich, W. Schade, N. Welschoff, and S. Waldvogel, “A microring resonator sensor for sensitive detection of 1,3,5-trinitrotoluene (TNT),” Sensors 10, 6788–6795 (2010). [CrossRef] [PubMed]

7. A. Nitkowski, L. Chen, and M. Lipson, “Cavity-enhanced on-chip absorption spectroscopy using microring resonators,” Opt. Express 16, 11930–11936 (2008). [CrossRef] [PubMed]

8. G. Roelkens, D. Van Thourhout, and R. Baets, “High efficiency Silicon-on-Insulator grating coupler based on a poly-silicon overlay,” Opt. Express 14, 11622–11630 (2006). [CrossRef] [PubMed]

9. G. Maire, L. Vivien, G. Sattler, A. Kazmierczak, B. Sanchez, K. B. Gylfason, A. Griol, D. Marris-Morini, E. Cassan, D. Giannone, H. Sohlström, and D. Hill, “High efficiency silicon nitride surface grating couplers,” Opt. Express 16, 328–333 (2008). [CrossRef] [PubMed]

10. H. Heidrich, P. Lützow, H. Venghaus, and H. J. W. M. Hoekstra, “Optical sensor and method for detecting molecules,” patent pending (PCT/EP2010/003784).

11. A. Yariv, “Universal relations for coupling of optical power between microresonators and dielectric waveguides,” Electron. Lett. 36, 321–322 (2000). [CrossRef]

12. J. Heebner, R. Grover, and T. A. Ibrahim, Optical Microresonators (Springer, 2008).

13. G. S. Wiederhecker, L. Chen, A. Gondarenko, and M. Lipson, “Controlling photonic structures using optical forces,” Nature 462, 633–636 (2009). [CrossRef] [PubMed]

14. H. J. W. M. Hoekstra, P. V. Lambeck, H. P. Uranus, and T. M Koster, “Relation between noise and resolution in integrated optical refractometric sensing,” Sens. Actuators B 134, 702–710 (2008). [CrossRef]

15. J. Hu, X. Sun, A. Agarwal, and L. C. Kimerling, “Design guidelines for optical resonator biochemical sensors,” J. Opt. Soc. Am. B 26, 1032–1041 (2009). [CrossRef]

Integrated optical sensor platform for multiparameter bio-chemical analysis

Abstract

1. Introduction

2. Concept

3. Design and fabrication

4. Experiment

5. Conclusion

Acknowledgments

References and links

Cited By

Figures (5)

Equations (9)

Optics Express