1. Introduction

The future of healthcare will be driven by personalized medicine and artificial intelligence [1–3]. This requires a high amount of information and data about the patient’s health condition. Molecular biomarkers such as proteins or metabolites are used for this purpose. Multiplex point of care biosensors are a promising technology to detect multiple biomarkers from small volumes of body fluid. They need to cover a wide range of different concentrations and perform reliably.

So far, various multiplex optical biosensors have been introduced for label-free detection. Surface plasmon resonance (SPR) is a standard laboratory detection approach. It shows high sensitivities to refractive index changes. However, freely propagating light cannot couple to the SPR modes resulting in bulky optical setups. Recently, a low-cost combined SPR and localized SPR (LSPR) sensor system was introduced based on transmission measurements excited by a freely propagating infrared LED light [4]. Especially LSPR show strong localized fields resulting in strong light-matter interactions with the analyte and low detection limits. They are very attractive for multiplex measurements too. Nanorods assembled in an array shape show real-time and parallel measurement of 32 detection sides with remarkably small footprint and a detection limits of 500 pg/ml [5]. Here the LSPR are excited by freely propagating light but still require a microscope and spectrometer setup. A similar system was employed for the detection of multiple cytokines [6]. Also multiplex silicon photonic microring resonators have been used for microRNA [7] and DNA detection with low bulk refractive index detection limits of 7.6E-7 RIU. High quality factor whispering gallery mode (WGM) sensors possess very narrow resonance linewidth and allow even single molecule detection for example when a microsphere is coupled with a plasmonic nanorod [8]. Such opto-plasmonic sensors benefit from very high Q-factor cavities combined with a small mode volume and high field enhancement from the plasmonic counterpart. Two drawbacks of these sensors ranging from LSPR to microrings and WGM resonators are the coupling of light in and/or out of these structures. Requiring couplers, tunable lasers and spectrometers. Also their fabrication costs are still quite high. Another class of optical signal transducers are photonic crystal slabs (PCS). They have shown their potential for compact and simple to fabricate multiplex biosensors [9–11] in case of a nanoimprint replication processes. Phase measurements of the resonant light reflected from the PCS allow a lower limit of detection (LOD) than intensity-based systems [12–14]. The detection limit can be as low as 1 pg/ml, realized with a PCS based common path interferometer that has a phase sensitivity of 732 rad/RIU [14]. To allow multiplex measurements with a PCS based interferometric sensor we developed a common-path interferometric setup based on multi-pinhole interferometry [15,16]. Pinholes [16] or microdisks [17] select parts of the optical wavefront to diffract the light of the guided mode resonance and let it interfere. The phase difference between two pinholes is retrieved from the complex value computed by the Fourier transform of its recorded far-field diffraction pattern [15,16]. By functionalizing one site with specific capture molecules and using another site as reference, time-resolved molecular binding is observed from the phase difference signal [13]. This approach is very promising because it has the potential of low detection limit measurements and has already proven its multiplex capability. Also, it allows for a simple optical measurement setup and PCS fabricated by nanoimprint replication process. Recently, we have shown that the dynamic range of this system is extended by combining intensity and phase based sensing of different modes [18]. For this purpose, we integrate a near-field intensity measurement into the interferometric setup. This allows to measure the wavelength shift of the TE guided mode resonance in addition to the phase shift of the TM mode to increase the dynamic range of the sensor system.

In this publication we further examine the process of multiple biomarker detection by simultaneous intensity and phase measurements, based on the photonic crystal microdisk biosensor setup [18]. The intensity and phase shift dynamics of the microdisk signal transducer are characterized for the TM mode theoretically and experimentally for different refractive index changes. Different effects observed during combined phase and intensity measurements are examined and explained by simulation and experiments. The results allow to align the setup to its sensitive working point and sets tolerances for the fabrication process of the signal transducer. Besides, we deduce benefits of the combined intensity and phase measurements. Combining two different measurement approaches increases the information about the measured binding processes.

The paper is structured as follows. In section 2 the principles of intensity and phase detection in reflection are discussed. After a description of the measurement setup, the fabrication method for the microdisks and the functionalization process are explained. The last part of the second chapter is on microdisk pattern design and simulation. In section 3, the sensor’s intensity and phase shift dynamic are investigated by simulations and refractive index sweeps. The consequences of inhomogeneous microdisk properties, which typically accompany fabrication by replication, are examined. Then simultaneous intensity and phase measurements of human-alpha thrombin and streptavidin are presented. In section 4 conclusions are given.

2. Methods

2.1 Intensity and phase detection for biosensing

The guided mode resonance shift in a photonic crystal slab induced by molecular binding may be measured by an intensity-based [10,19,20] or phase-based [12,14,16] approach. In our setup a fixed laser wavelength of 632.8 nm is used. As signal transducer a one-dimensional photonic crystal slab is micro-structured to form microdisks as shown in Fig. 1 and Fig. 2(f). The parameters of the photonic crystal slab are given in section 2.3. The inset graph in Fig. 1 shows the simulated behavior of the reflected light intensity and phase around the guided mode resonance. The angle of incidence is changed to align the microdisk to the resonance position [19]. Hence, for simulation we use the fixed laser wavelength and simulate the intensity and phase for different incidence angles. Exemplarily the curves are shown for two different refractive indices n₀= 1.332 and n₁= 1.333 of the cover medium. Latter shows the effect of a shifting guided mode resonance (GMR). The change in the GMR position induces a shift in intensity and phase when we measure with a fixed angle at a fixed wavelength. Hence, by measuring the intensity or phase shift of the reflected GMR light the resonance shift can be measured. In section 2.2 the setup for these measurements is explained in detail.

 

Fig. 1. Sensor setup for simultaneous intensity and phase measurements of microdisk guided mode resonances. A laser excites guided mode resonances in periodically-nanostructured (photonic crystal) microdisks. The intensity of the reflected resonance light is detected in the near field and in the far field of the multi microdisk diffraction pattern by two cameras. From the far field interferogram the relative phase shift between each pair of two pindisks is computed. Upon binding of molecules to one of the microdisks the guided-mode resonance wavelength is shifted for this pindisk. This is observed by a change in the reflected laser intensity as well as by a phase change in the interferogram relative to a reference microdisk.

Download Full Size | PDF

 

Fig. 2. Microdisk fabrication process. A planar photonic crystal slab is structured with lithography to obtain microdisks. First, the microdisk structure is transferred from the mask to a positive photoresist. In the second step the structure is etched into the photonic crystal slab by ion beam etching (IBE). At the end the residual resist is removed from the microdisks. (f) Scanning electron image of a 75 µm microdisk.

Download Full Size | PDF

2.2 Setup

The biosensor measurement system is shown in Fig. 1. The signal transducer is a one-dimensional photonic crystal slab that is micro-structured to form microdisks with a diameter of 250 µm. The microdisks form a circular pattern and are fabricated on a glass substrate. For biomarker detection the substrate is placed in a fluid cell [16]. The quasi guided transverse magnetic (TM) mode in the microdisks is excited with a HeNe-laser at 632.8 nm (Thorlabs, HNL008L). To illuminate the full microdisk pattern, the laser spot is magnified with a telescope setup, built by two aspheric lenses with focal length f₁ = 4.6 mm and f₂ = 50 mm. To fulfill the Bragg equation (λ_res = Λ (n_eff ± sinα)) the angle of incidence is aligned around 7.0°. For the alignment the fluid cell with the signal transducer is mounted on a rotation stage. The polarization of the light exciting the GMR is converted into a circular polarization by a circular polarization filter to suppresses the non-resonant light that is reflected back from the air-glass interface of the backside of the glass substrate. The part of the circular polarization that is orthogonal to the photonic crystal grating lines, excites the TM mode. The back reflected GMR light passes through a second circular polarization filter and is divided by a 50:50 beam splitter into two orthogonal paths. The light reflected at the beam splitter is projected onto a CMOS camera chip in the near field of the multi-microdisk diffraction pattern (in Fig. 1). For the projection two lenses with focal lengths f₃ = 125 mm and f₄ = 25.4 mm are used. This allows to perform intensity-based measurements of the GMR shift [21]. The transmitted path is focused by a Fourier lens (f = 150 mm) into the far field. The interferogram of the far field diffraction pattern is captured by a second CMOS camera (Fig. 1). The diameter of the center bright spot is approximately 460 µm (Ø_Airy= 1.22*λ*f/d_MD). It is formed by the superposition of the Airy intensity distributions of the microdisks with diameter d_MD at the focal plane f of the Fourier lens. By computing the Fast-Fourier Transform of the image the phase difference between each pair of two microdisks is obtained [16]. The distance of the interference fringes is determined by the focal length f, and the distance d between the microdisks. (s_F = fλ/d). This measurement setup allows to perform multiplex biomarker detection by measuring simultaneously the intensity and phase shift induced by molecular binding to the different microdisk sites on the sensor surface.

2.3 Photonic crystal and microdisk fabrication

The fabrication of microdisks includes two major fabrication steps. At first, a photonic crystal slab (PCS) is produced by nanoimprint lithography and ion beam etching [22]. Afterwards it is micro-structured by a lithography process to produce the circular microdisk (Fig. 2(f)) pattern.

For the PCS fabrication a glass substrate is sputtered with a 150 nm thick Nb₂O₅ layer (Kurt J. Lesker, EJUNBOX353TK4) that acts as a dielectric waveguide. For nano-structuring the waveguide layer, AMOPRIME (AMO GmbH) is spincoated on top, followed by softbaking for 2 min at 115°C. Afterwards, AMONIL (AMO GmbH) is spincoated on top. Then a PDMS (Dow, Sylgard 184 and curing agent in a ratio of 8:1) stamp with a one-dimensional grating structure is manually pressed into the soft AMONIL layer [9]. The grating has a period of 370 nm and a grating depth of 60 nm. For curing, the AMONIL layer with the stamp is illuminated with a UV LED for 80s. Afterwards, the PDMS stamp is removed.

To transfer the nanostructure of the grating from the AMONIL layer into the dielectric waveguide, the stack is etched by an ion beam (PC3000, Oxford Instruments, Abingdon, UK). The etching rates of AMONIL r_A = 27.5 nm/min and of Nb₂O₅ r_N = 15 nm/min were measured to determine the resulting grating depth of 33 nm in the waveguide. Residual photoresist is removed with P1316 (TechniStrip P1316, MicroChemicals) in an ultrasonic bath (20 min at 75°C). The height of the waveguide is etched down to 135 nm.

For the microdisk fabrication (Fig. 2), the positive photoresist AZ 1518 (MicroChemicals) is spincoated onto the PCS. After a softbake step (60s at 110°C) the two-dimensional microdisk pattern is transferred from a lithography mask into the photoresist by UV light exposure (82 mJ/cm²). To remove the exposed photoresist the sample is develop for 75s in AZ 726 MIF (MicroChemicals). At the end of the lithography process the sample is hardbaked (60s at 110°C). The photoresist covers and protects the PCS at the areas where microdisks are supposed to be formed. During ion beam etching the PCS is removed at all non-covered areas. At the end, residual photoresist is removed with TechniStrip P1316 (MicroChemicals) in an ultrasonic bath (20 min at 75°C).

2.4 Functionalization

To investigate multiplex biomarker detection thrombin (CellSystems GmbH, human alpha-thrombin, HCT-0020-MG) and streptavidin (Sigma Aldrich, 189730-1 MG) are used as test molecules. Thrombin binding aptamers [9] (Fig. 3, left microdisk) and BSA-biotin (Sigma Aldrich, A8549-10 MG) (Fig. 3, right microdisk) are immobilized on the microdisks as capture molecules according the protocol described in [9]. The capture molecules are placed in form of a small volume drops (1 µl - 2 µl) on the microdisks. The rest of the substrate surface is passivated with BSA.

 

Fig. 3. Functionalization of the photonic crystal microdisk substrate. On the left microdisk thrombin-binding aptamers are immobilized and on the right BSA-biotin to bind streptavidin. The surface is passivated with BSA.

Download Full Size | PDF

2.5 Simulation methods

The computation of the far field diffraction pattern, spatial frequencies and phase shifts (Fig. 4, Fig. 7b-e) is performed in Matlab [16,23]. The phase curve (Fig. 7a) around the guided mode resonance is simulated with the finite-element method (COMSOL Multiphysics). As well as the guided mode resonance shift relative to the change in the high index waveguide layer thickness.

 

Fig. 4. Intensity and phase computation of circular microdisk pattern. (a) Real-space microdisk pattern with different assigned phase values. (b) Intensity of far field diffraction pattern. (c) Frequency space: Magnitude of Fourier-transform of far field diffraction intensity. (d) Phase of frequency space (only inner circle with lowest frequencies is shown). Regions with clearly defined phase caused by the interference signal of two microdisks are observed. The background exhibits a rapidly changing phase.

Download Full Size | PDF

2.6 Microdisk pattern design

For the experiments seven microdisks, each with a diameter of 250 µm, are aligned equidistantly on a circle with a radius of 2.5 mm. Figure 4(a) shows the pattern with different phase values assigned to each microdisk. In Fig. 4(b) the computed intensity of the far field diffraction pattern is depicted. In the experiment, this intensity distribution is measured by the camera. To demodulate the spatial frequencies and phase information the Fast-Fourier transform is computed. Its magnitude is shown in Fig. 4(c) and the phase of the inner circle with the lowest frequencies in Fig. 4(d). Each combination of two neighboring microdisks produces two spatial frequencies on the inner circle. The absolute phase value is the phase difference between both microdisks in real space. The two spatial frequencies have the opposite sign and lie on the opposite sites of the circle (180° shift).

3. Results and discussion

3.1 Dynamics of intensity and phase shifts (TM mode)

For the sensor performance it is important to know how much the intensity or phase changes due to a shift in the refractive index (RI) of the cover medium. This is expressed by the bulk refractive index sensitivity S. Intensity and phase shifts of the TM mode are simulated for different positive bulk RI shifts Δn (as illustrated in Fig. 5(a)) between 1E-5 RIU and 2E-2 RIU relative to the initial RI of water n₀. This is done for four different incidence angles α, β,γ and δ (Fig. 5(b)-(f)) which are initially positioned around the resonance peak (Fig. 5(a)) to present different alignment scenarios. The intensity sensitivity S_I = ΔI/Δn and the phase sensitivity S_P = ΔΦ/Δn are computed from the simulation data exemplarily for an incidence angle of β = 7.024° in Fig. 5(d)).

 

Fig. 5. (a) Simulated reflection intensity and phase of the photonic crystal at 632.8 nm excitation wavelength around the incidence angle of 7°. Exemplarily intensity and phase are shown for two different RI of the cover medium. The simulated phase (blue) and intensity (orange) shifts for different refractive index shifts are plotted for angles of incidence of (b) 7.005°, (c) 7.024, (e) 7.030°, and (f) 7.041°. The sensitivity for phase and intensity at an incidence angle of 7.024° is shown in (d). The combined phase shifts for the different angles are shown in (g) and for the intensity shifts in (h).

Download Full Size | PDF

The simulations of the four angles show that for all angle alignments around and at the resonance peak the phase shift increases monotonously relative to increasing refractive index shifts. In contrast, for the intensity change this behavior cannot be observed for all alignment angles. For alignment angles (α,β), where a shift occurs from one side of the peak to the other one (Fig. 5(b)-(d)), the intensity sensitivity (Fig. 5(d)) can reach a value of zero when the RI change is large enough, so that the reflectivity value after the shift is the same as the one before but positioned on the other peak side. In this case the intensity change and the sensitivity is inverted from negative to positive or the other way around. For the alignment angles γ (peak position) (Fig. 5(e)) and δ (Fig. 5(f)) there is no shift to the other peak side, hence the intensity change increases monotonously too. For sensing applications care has to be taken to choose an initial angle alignment such that a monotonous intensity change with refractive index change is obtained in the refractive index range of interest. For non-monotonous intensity change curves in the range of the “zero dip” the same intensity shift is observed for two different RI changes rendering the result ambiguous.

To compare the performance for the different alignment angles Fig. 5 g summarizes all phase shift curves and Fig. 5 h shows the intensity curves. The highest phase sensitivity S_P = 800 rad/RIU is reached when the angle is aligned to the steep and linear slope of the phase curve at angle β (Fig. 5(c)-(d)). It stays high until the peak is shifted out of this steep range relative to the excitation. Hence, this alignment would be beneficial for phase shift measurements in a narrow refractive index range with low LOD. However, the corresponding intensity shift curve has a “zero-sensitivity dip” due to shift over the peak maximum. Hence, the shift curve is ambiguous for RI shift smaller than 1E-3 RIU. If the goal is to measure both intensity and phase over the whole RI shift range, then the excitation should be aligned to the side of the peak where no shift over the peak maximum occurs. In case of only increasing RI shifts relative to the initial alignment RI, this is the higher angle side of the peak. For high sensitivity phase performance, the alignment angle should be close to the peak maximum γ. The highest intensity sensitivities are at the steepest part of the peak. On the higher angle peak side the maximum sensitivity in the simulation is S_I =353%/RIU. It is achieved when the initial angle is aligned to 7.039° and stays constant for RI shifts smaller than 2E-4 RIU. This is close to the angle δ which shows to have the highest intensity change for small refractive index shifts in Fig. 5 h, when we neglect the ambiguous curves of angle of incidence aligned to the lower angle side of the resonance peak. In general, instead of aligning the angle also a tunable coherent light source could be used.

For experimental validation refractive index sweeps with different glucose concentrations in water are performed. Two different alignment angles θ1 (Fig. 6(a)) and θ₂ > θ₁ (Fig. 6(d)) close to the resonance peak γ are chosen (|θ_1/2-γ | < 0.04°). The substrate with the microdisks is placed in a fluid chamber. Sequentially different glucose concentrations are injected. Since all microdisks perceive the same RI change, no phase shift measurements are shown which only detect relative phase shifts between a pair of two microdisks that are near zero in this experiment setup. Table 1 shows the injection sequence and corresponding RI changes. The normalized intensity shift over time is shown in Fig. 6(a) (θ₁) and Fig. 6(d) (θ₂) for three different microdisks (MD1-MD3), aligned on one circular microdisk pattern. The measured normalized intensity shift relative to the RI change is plotted for microdisk MD1 in Fig. 6(b) and Fig. 6(e) respectively. And the corresponding simulated intensity change in Fig. 6(c) and Fig. 6(f). From the different microdisk intensity shift curves it is concluded that all microdisks have a slightly different resonance peak position relative to the excitation wavelength. Hence, the sensitivity for different RI changes and the resulting intensity shift differs between the microdisks. Some show positive intensity shifts and some negative ones, indicating that they are positioned on opposite sides of the resonance peak.

 

Fig. 6. Experimental intensity shift examination. (a) RI sweep at an alignment angle θ₁, three different microdisks MD1, MD2 and MD3 are shown. (b) Measured intensity shift of MD1 at θ₁ and (c) simulated intensity shift curve at 7.010°. (d) RI sweep at an alignment angle θ₂, for MD1, MD2 and MD3. (e) Measured intensity shift of MD1 at θ₂ and (f) simulated intensity shift curve at 7.028°

Download Full Size | PDF

Table 1. Injection sequence of glucose concentrations in water for alignment angles θ₁ and θ₂

View Table | View all tables in this article

At alignment angle θ₁ microdisk MD1 shows small positive intensity shifts for RI changes smaller than 6.4E-3. Then at minute 27 there is no shift at all, even so the other microdisks indicate the injection of the 5.2% glucose solution. The intensity change crossing the zero line can be seen in Fig. 6(b) too. For larger RI changes a monotonous increasing negative intensity change is observed. This matches well to the simulated behavior shown in Fig. 6(c). Looking at the intensity change observed at the slightly higher alignment angle θ₂, we see a complete monotonous increasing negative shift. This can be explained by an initial alignment to the other side of the peak, where for increasing RI shifts no peak maximum crossing is possible. This is confirmed by the simulated intensity change in Fig. 6(f) at an angle of incidence of 7.028°. It fits approximately to the measured curve. The main difference is, that the simulated curve has a shorter linear range. This is probably due to the higher quality factor of the simulated resonance compared with the real microdisk that has higher losses, for example through scattering and the edges.

An intensity limit of detection of LOD_I = 2E-4 RIU is computed for the angle θ₂. With similar setup we have already shown limits of LOD_I = 3E-5 RIU [16]. The maximal signal to noise ratio is SNR = 159 and the highest measured intensity sensitivity is 11E3%/RIU. The sensitivity values of the simulation cannot be compared quantitively with the measured ones. Since in our simulations only the photonic behavior of the resonance shift is modeled, without the consideration of the camera detection system.

The inhomogeneity of the resonance position is approximately 1 nm within one microdisk circle (see section 3.2). The typical TM resonance width of our sensor is not smaller than 2 nm. Hence, it is usually possible to excite all microdisks, but at slightly different positions. At the end of section 3.4 reasons for the inhomogeneity and solutions to reduce them are discussed.

3.2 Effect of inhomogeneous microdisk resonances on the phase detection

The technique of nanoimprint lithography combined with ion beam etching produces slightly varying waveguide layer thicknesses. The GMR shifts about 0.55 nm per nm thickness change (simulated with COMSOL, section 2.5). This results in a wavelength offset of the guided mode resonance between the microdisks on the circle. The change in the GMR wavelength over the substrate is usually between 0.1 and 0.15 nm/mm using our fabrication method. It leads approximately to a maximum GMR offset of 1 nm between microdisks on a circle. When the microdisks are excited with a narrowband laser peak at 632.8 nm, each microdisk has a slightly different phase and sensitivity. To simulate the effect of this inhomogeneity on the measurement, the computation of the far field and the spatial frequencies described in sections 2.5 and 2.6 are employed. To estimate the phase variance of the different microdisks due to inhomogeneities, each microdisk is simulated with a slightly different GMR wavelength. The resonance offset is smaller than 1 nm. Figure 7(a) shows the resulting phase curves of three microdisks. The markers with the numbers show the phase of the selected microdisks, all at the fixed laser wavelength of 632.8 nm. Figure 7(b) shows the microdisk pattern with the phase distribution derived from the simulated phase curve. Each pair of two neighboring microdisks produces a spatial frequency in the far field (Fig. 7(c)). The position of the frequencies is derived from the difference vector between the two microdisks (red arrows). To simulate a measurement situation the refractive index at the surface of the PCS is shifted about 1E-4 RIU. Each microdisk perceives a slightly different phase shift. Figure 7(d),(e) show the difference in the phase of the spatial frequencies between the two steps. With a constant GMR of the different microdisks we would observe a phase shift of zero because the relative phase between two microdisk would stay constant. However, in the inhomogeneous case each frequency shows a slightly different phase shift and drift over time (Fig. 7(d)). This effect is observed during measurement too. Figure 7(f) shows the measured drift of seven neighboring spatial frequencies over time.

 

Fig. 7. Drift simulation of microdisk pattern. (a) Simulated phase of three photonic crystal slabs with three different GMR with offsets ≤ 1 nm. The stars mark the different phases of the microdisks shown in (b). (c) The lowest spatial frequencies show different phases two. When the GMR is shifted also the phase of the spatial frequencies is shifted (d). It results in an inhomogeneous drift of the microdisks (e) that is also observed in experiment (f).

Download Full Size | PDF

The phase drift of the different spatial frequencies is partly removed during post measurement data processing. A linear drift is subtracted from the measured signal by fitting the baseline of the reference signal that is supposed to have a constant value. The calculated drift is then subtracted from the measured signal. For the process of drift compensation, the measurement presented below in chapter 3.4 is divided into two parts since the phase curve around the GMR (Fig. 7(a)) shifts out of the area of constant slope and sensitivity during the measurement (see Fig. 5), resulting in different slopes of the drift.

The wavelength change of the guided mode resonances within the microdisk pattern prevents to achieve the full potential of self-noise cancelation [1,6]. In case of constant-wavelength guided mode resonances, each pair of two microdisks would perceive (depending on the noise source) a very similar phase noise which would be canceled out in the phase difference signal of the spatial frequency.

If we want to ensure that the difference of all GMR within one microdisk circle (d = 5 mm) stays below a quarter of the resonance width, the high index layer thickness should not vary more than 0.1 nm/mm over the circle area. The smaller the variance, the better the noise suppression works and the lower the drift.

3.3 Experimental phase measurement from spatial frequencies

In contrast to simulation (Fig. 7(c)), the computed spatial frequencies (Fig. 8(b)) from the measured far field diffraction pattern (Fig. 8(a)), do not have a constant phase difference value within one frequency spot. Therefore, a selection of multiple pixels (red box) is used to determine the phase difference. As an example, we observe the phase difference shift produced by the binding of streptavidin (Fig. 8(d)). The binding is measured by detecting the phase difference between a microdisk functionalized with streptavidin binding biotin and a neighboring reference microdisk. Figure 8(b) shows the magnitude of the corresponding spatial frequencies (red arrows) in the Fourier space. In Fig. 8(c) the phase shift difference corresponding to the magnitude image is plotted. It is computed by subtracting the phase difference from the image at t₁ from t₂. Figure 7(d) shows the mean phase difference (solid line) and the standard deviation (shaded area) of all selected pixels from the frequency F1 (yellow) and -F1 (blue). As expected from the phase simulation (Fig. 7(d)), each pair of two microdisks is represented by two frequencies with identical phase value but opposite sign (Fig. 4(d), Fig. 7(c),(d)).

 

Fig. 8. Process of phase measurement from (a) the detected far field diffraction pattern originating from the circular microdisk pattern (Fig. 7(b)). (b) Inner circle with magnitude of lowest spatial frequencies produced by neighboring microdisks. The spatial frequency F1 (red arrow) is used for processing. The red square indicates the pixels used for the computation of phase shift. (c) Corresponding phase shift image computed by subtracting the phase image at t₁ from t₂. (d) Phase shift due to binding of streptavidin. The inset shows the mean phase shift (dark line) and the standard deviation of all selected pixels (red box) from the frequency F1 (blue) and -F1 (yellow).

Download Full Size | PDF

3.4 Multiplex biomarker detection based on intensity and phase measurements

As proof of principle for multiplex biomarker detection based parallel intensity and phase measurements, human-alpha thrombin and streptavidin are chosen as test biomarkers. Two microdisks are functionalized with the thrombin binding aptamer and three with BSA-biotin to bind streptavidin. The corresponding areas are marked with the dotted line in Fig. 9. At the beginning of the experiment the fluid chamber is filled with a buffer solution (DPBS D8537, Sigma-Aldrich). At minute 12, DPBS spiked with Thrombin (5 µg/ml) is injected into the chamber. To stop binding of thrombin to the sensor surface, buffer is injected at minute 22. Next, at minute 27 a 2 ng/ml streptavidin in DPBS solution is injected. At minute 68, streptavidin in DPBS (8 µg/ml) is injected. The binding process is stopped at minute 73 by the buffer solution. As reference a 1% NaCl in water solution is injected at minute 83 and 88. The measurement is finished with a buffer injection at minute 90.

 

Fig. 9. Multiplex detection of thrombin and streptavidin based on simultaneous intensity and phase measurements. (a) Near field intensity image of microdisks. Colored dotted line shows the area functionalized with the thrombin binding aptamer (blue) and BSA-biotin (orange). The colored circles indicate the microdisks used for the intensity measurement (b),(c) and each arrow marks the microdisk pair producing the spatial frequency shown in (d) for the phase measurement (e),(f). The sequence of buffer (B), thrombin, streptavidin 2 ng/ml, buffer, streptavidin 8 µg/ml, buffer, 1% NaCl in water (NaCl) and buffer injection is given in the graphs

Download Full Size | PDF

For the intensity measurement three microdisks are selected in the different areas of interest – a BSA-biotin functionalized microdisk (orange), a microdisk functionalized with thrombin binding aptamer (blue) and a reference microdisk with no functionalization (green). The arrows indicate the pairs of microdisks used to track the relative phase shifts between microdisks. The corresponding spatial frequencies recorded with the far field camera are shown in Fig. 9(d). The frequency pattern looks as expected from the theoretical computation in Fig. 4(c). As reference phase difference signal, the spatial frequency produced by two microdisks of the same kind is used. Here, we choose the two microdisks functionalized with the thrombin binding aptamer (green arrow). They are expected to have the same phase shift behavior and hence no relative phase shift over time.

Due to the injection of thrombin (minute 12), both the intensity signal in Fig. 9(b) of the microdisk functionalized with the aptamer, and the phase signal in Fig. 9(e) of the spatial frequency produced by the reference microdisk and aptamer functionalized microdisk show a binding signal. The reference signal stays almost constant in both cases. The orange curves representing the binding site for streptavidin show only a weak unspecific binding compared with the signal produced at the aptamer microdisk.

At minute 22, the chamber is flushed with DPBS. Both the intensity and phase signals show that the binding is stopped and that some thrombin molecules are removed from the sensor surface. Afterwards the signal stabilizes again. Indicating a much higher number of thrombin molecules remaining on the aptamer microdisk compared with the reference and biotin microdisk.

At minute 27 a 2 ng/ml streptavidin in DPBS solution is injected, however there is no binding detected. But at minute 68, a higher streptavidin concentration of 8 µg/ml is injected. This time the intensity (Fig. 9(c)) and phase (Fig. 9(f)) signal related to the biotin functionalized microdisk show a strong increase. In case of the intensity measurement the aptamer and reference microdisk intensity signal are roughly reduced by half. Both show the same behavior and indicate that the signal does not result from binding of streptavidin but could be explained by a drifting bubble at the location of the reference and aptamer microdisk. In contrast, the phase signals of the reference and aptamer microdisks stay almost constant They only show the typical noise, smaller than 10 mrad. This shows one benefit of simultaneous intensity and phase measurement. Some noise processes effect the intensity more than the relative phase change, and the other way around. This results in a more reliable measurement providing more information that can be used for interpretation of the processes.

At minute 83, 1% NaCl in water is injected as reference step for the intensity measurement. Followed by the injection of DPBS at minute 91. The induced change in bulk refractive index is supposed to have the same effect on all microdisks except if part of the bonded molecules are removed. Therefore, the NaCl step is barely visible in the phase difference signal compared with the intensity signal from the single microdisks. The phase signal shows that only few streptavidin molecules are removed from the biotin. This is expected due to the strong binding between biotin and streptavidin.

Another benefit of the phase measurement is observed when we compare the signal to noise ratio (SNR) to the one of the intensity signals (Table 2). The thrombin binding signal shows a three times higher SNR in case of the phase (SNR = 42) compared to the intensity measurement (SNR = 14). In case of the streptavidin injection, both the intensity and phase signal are strong and the SNR with 205 and 190, respectively, high. The phase has only a slightly higher SNR compared to the lower thrombin signal. In case of the intensity the signal increases from the lower thrombin step to the higher thrombin step by a factor of 14 whereas for the phase signal, it increases only by a factor of 5. The noise stays on the same order for both cases. This is probably due to the angle alignment of the microdisks relative to the angle of incidence. Figure 5(d) shows the special case, for which larger refractive index shifts leads to an increase in the intensity sensitivity whereas the phase sensitivity decreases. Hence the intensity- and phase-SNR would move towards each other.

Table 2. Signal to noise ratios of intensity and phase measurements of signal steps caused by addition of thrombin and streptavidin, respectively

View Table | View all tables in this article

To achieve higher SNR it is important to carefully align the angle of incidence (Fig. 5) and it is necessary to further reduce the noise [24]. As discuss before, the microdisks have an offset in their guided mode resonance due to the fabrication process. Hence, noise suppression can be further improved by improving the homogeneity of the microdisks in the fabrication process. Currently the high index waveguide layer is nanostructured by manually pressing a PDMS stamp into the AMONIL layer followed by ion beam etching to transfer the nanostructure into the waveguide. The manual process of imprinting results in a varying AMONIL layer thickness and after etching into inhomogeneous waveguide layer thickness. One solution could be to use for photonic crystal slab fabrication a modified nanoimprint process described in [9]. The described process fabricates the waveguide by sputtering the high index layer on top of the nanostructured AMONIL layer. Therefore, the waveguide layer is not influenced by the inhomogeneous AMONIL layer. However, during the cleaning process, after ion beam etching, the AMONIL layer would be partly dissolved by the cleaning solvent and hence the microdisks would be flushed away or be damaged. This can be prevented by the fabrication of an intermediate protection layer or by finding another photoresist solvent system that does not remove the AMONIL layer. Besides, the nanoimprint process could be performed with a fixed and homogeneous force instead of manual pressing to produce a uniform photoresist layer thickness. Much better results would be achieved by producing the structure directly with e-beam or interference lithography.

4. Conclusions

A microdisk based biosensor system has been introduced that combines simultaneous intensity and phase measurements to detect multiple biomarkers in parallel. This system approach has been analyzed in detail regarding the expected phase and intensity sensitivity.

It was shown that the measurements allow for a simple optical setup to detect the phase and intensity shift of the guided mode resonance. No spectrometer is needed. Only two cameras positioned in the near- and far field are used. To achieve a compact system, a Fourier lens is used to project the far field onto the camera. The microdisk pattern is designed for multiplex biomarker measurements based on phase detection. Latter has the potential of high SNR and low detection limits. The microdisk signal transducer has a small footprint. This makes it suitable for small sample volume measurements.

Simulations and experiments of the intensity and phase sensitivity (S_P = 800 rad/RIU) for different RI shifts have shown that a proper alignment angle has to be chosen to guarantee high phase sensitivity and unambiguous intensity measurements. The intensity detection part itself can be used to align the sensor systems excitation light relative to the resonance peak position and its reflection minima by tracking the intensity of the reflected resonant light for different angles of incidence or the intensity change relative to a refractive index change.

Various effects originating from inhomogeneous microdisk resonances were explained by simulation and experiments. The investigations emphasized the importance of homogeneous microdisk fabrication. They provide fabrication tolerances in regard to the resonance inhomogeneity. The simulated maximum phase sensitivity S_P = 800 rad/RIU is comparable to the one in literature [14] based on a similar measurement approach. However, to reach detection limits on the order of 1E-7RIU the phase noise is with 3E-3 rad too high. The reason is found in the inhomogeneity of the different microdisks within the circular pattern. The thickness of the photonic crystal high index layer should not vary more than 0.1 nm/mm over the substrate. The smaller this variance the higher the homogeneity of the GMR position and the better the noise cancelation of the phase measurement works. This problem can be faced by a modified nanoimprint process or much better results would be achieved by producing the structure directly with e-beam or interference lithography.

As proof of principle the detection of human-alpha thrombin and streptavidin has been demonstrated. Different sources of noise result in different phase and intensity signals. The combined measurement allows a more reliable interpretation of the binding dynamics. Already with the current nonuniform microdisks, the SNR is increased up to a factor of 3 for phase detection compared to the intensity measurement. Once the microdisk fabrication process is improved the characterization of the setup by the detection of various biomarkers over a larger concentration range is the next important step to show the full potential of the measurement approach.