Metallic photonic crystal-based sensor for cryogenic environments.

We investigate the design, characterization, and application of metallic photonic crystal (MPC) structures, consisting of plasmonic gold nanogratings on top of a photonic waveguide, as transducers for lab-on-chip biosensing in cryogenic environments. The compact design offers a promising approach to sensitive, in situ biosensing platforms for astrobiology applications (e.g., on the "icy moons" of the outer solar system). We fabricated and experimentally characterized three MPC sensor geometries, with variable nanograting width, at temperatures ranging from 300 K to 180 K. Sensors with wider nanogratings were more sensitive to changes in the local dielectric environment. Temperature-dependent experiments revealed an increase in plasmonic resonance intensity of around 13% at 180 K (compared with 300 K), while the coupled plasmonic-photonic resonance was less sensitive to temperature, varying by less than 5%. Simulation results confirm the relative temperature stability of the plasmonic-photonic mode and, combined with its high sensitivity, suggest a novel application of this mode as the sensing transduction mechanism over wide temperature ranges. To our knowledge, this is among the first reports of the design and characterization of a nanoplasmonic sensor specifically for low-temperature sensing operation.

sensors is that they are compatible with on-chip integration via the guided modes, facilitating the design of "spectrometer-free" biosensing platforms [25][26][27][28]. However, these and other plasmonic-based biosensors have been primarily applied at ambient conditions.
In order to expand the range of operation to emerging life detection/astrobiology applications in more extreme environments (e.g., cryogenic temperatures), thorough characterization under relevant conditions is required. It is especially desirable to make biosensing measurements in situ [29][30][31][32], to minimize sample processing and contamination risks, thus requiring cryogenic-temperature operation for many astrobiology applications. However, the optical response of plasmonic nanostructures can vary significantly with temperature, depending on both the material properties and device geometry [33,34].
The low-temperature optical properties of engineered plasmonic nanostructures have not been thoroughly studied in the literature, with the existing body of literature being primarily concerned with reducing ohmic losses at low temperature and improving performance of plasmonic devices for which these losses are problematic (e.g., plasmonic circuits). Bouillard et al. studied two nanoplasmonic device designs: a plasmonic nanorod metamaterial, and a one-dimensional plasmonic crystal [33]. They showed that the temperature-dependent changes in spectra are strongly dependent on the device nanostructure, with a much greater response for the nanorod metamaterials compared to the plasmonic crystal. Not only did the magnitude of the transmission spectra change considerably in each case between room temperature and liquid nitrogen temperature, it did so in the opposite direction: the metamaterial showed significantly decreased transmission (by a factor of 10), while the plasmonic crystal exhibited slightly increased transmission (up to approximately 20%). Liu et al. studied gold bipyramids in an effort to understand the specific mechanisms responsible for the temperature-dependent optical properties, in particular the reduced losses at lower temperatures [34]. At low temperature, they observed a narrowing of the plasmonic line width, as well as a shift in the resonance location. Ultra-thin plasmonic films (i.e. randomlyformed metallic nanoislands on a substrate) for surface enhanced Raman scattering (SERS) and metal induced fluorescence (MEF) have also been shown to exhibit enhanced optical properties (increased SERS signals, and increased fluorescence of nearby fluorophores) at low temperatures [35,36]. Brandt et al. showed that such ultra-thin gold films exhibited either increased or decreased reflection intensity at low temperature, depending on the film thickness/nanoparticle geometry, further illustrating the complicated temperature-and geometry-dependent effects in nanoscale plasmonic systems [37]. Transmission intensity, reflection intensity and resonance location are all important characteristics of a nanoplasmonic sensor, and have all been shown to vary with temperature, as discussed above. Therefore, careful characterization of proposed sensor designs in relevant environments is critical. To our knowledge, the nanoplasmonic sensing application has not been studied at low temperature.
In this work, we study the optical properties of a proof-of-concept MPC-based sensor in low-temperature environments, with potential application to life detection on the "ocean worlds" of the outer solar system (e.g., Europa, Enceladus, and Titan), and other cold locations (e.g., Mars, and distant asteroids/comets) [38]. We demonstrate the temperaturedependent optical response of the plasmonic nanogratings both experimentally and theoretically. In contrast to the plasmonic response, the coupled plasmonic-photonic resonance is shown to be essentially insensitive to temperature (both in terms of intensity and resonance location), enabling stable operation over wide temperature ranges. Using this coupled mode as the sensor transduction signal therefore has two benefits: (i) it can be made more sensitive to changes in the local dielectric environment than the photonic or plasmonic responses alone [21], and (ii) it is insensitive to wide temperature variations. Together, these contribute to a viable sensor candidate for the sensitive detection of biochemical targets in extreme environments, where stable operation is required over wide temperature ranges.

Metallic photonic crystals
The unique properties of plasmonic nanoparticles coupled to photonic waveguides were first reported by Linden et al. [18]. They showed that selective suppressed (or enhanced) extinction (Ext = -ln(T), where T is the transmission spectra) may be achieved through appropriate tuning of the geometrical parameters of the nanoparticles and the waveguide. Two-dimensional periodic arrays of nanoparticles on top of the waveguide enabled momentum matching of the incident light and excitation of the guided modes, producing an extinction minimum superimposed on the plasmonic resonance peak. The relevant geometrical parameters in these structures are nanoparticle size, spacing, and waveguide thickness. Christ et al. [19] later showed that one-dimensional nanograting structures provided enhanced coupling compared with the two-dimensional counterparts, attributed to the stronger optical response of the nanogratings. For our design, we used one-dimensional nanogratings and aimed to tune the plasmonic and waveguide responses in the visible spectrum where plasmonic enhancement of gold nanoparticles is maximized [6]. This spectral region (ca. 600 nm to 1000 nm) is also readily accessible through available light sources (e.g., light emitting diodes) and detectors (e.g., photodiodes and charge coupled devices), for eventual on-chip integration. The nanogratings were fabricated on top of an indium tin oxide (ITO) photonic waveguide deposited on a glass substrate. A schematic of our sensor is shown in Fig. 1. Next, we quantify the conditions necessary to excite the plasmon and waveguide resonances in the region of interest, and to form the coupled plasmonic-photonic modes. Our design parameters were: nanograting width w, nanograting thickness, grating period d x , and waveguide thickness h ITO . The first design consideration was the spectral region of interest. For the plasmon resonances, the 600 nm to 1000 nm region was targeted. Nanogratings of approximately 200 nm in width and 35 nm in height excite a plasmonic resonance in this range to fulfill this requirement. While both of these parameters influence the optical properties of the MPC structures, the grating width plays a much larger role (see Appendix A), especially towards exciting the coupled plasmonic-photonic modes used for our sensing application. Besides exciting the broad plasmon resonance within the region of interest, two conditions must be satisfied to produce these coupled plasmonic-photonic modes: (1) The waveguide must support guided modes in the spectral region of interest (i.e. it must not be in cutoff); (2) The momentum of incident light must match the momentum of guided modes, in order to facilitate efficient excitation. For our experimental values (h ITO = 175 nm, ε sub = 2.3, ε ITO = 3.8, ε air = 1, and c = speed of light in vacuum), the cutoff frequency was calculated to be f co = ω co /2π = 167 THz, which corresponds to 1795 nm. Our spectral region of interest of 600 nm to 1000 nm (300 THz to 500 THz) is well above the cutoff frequency and is thus supported by the waveguide.
Coupling free-space light to the guided mode requires momentum matching (Condition 2) [18]: .
Here, wg k  is the wavevector of the guided mode, x k  is the component of the incident wavevector along the x axis, j is an integer, and x g  is the reciprocal grating vector: . Assuming plane (collimated) waves at normal incidence, x k  goes to zero, and wg k  is determined by the grating period alone. As discussed later in this section, for our experimental setup, light is focused onto the sample using an objective lens, so that a range of angles are present and x k  does not vanish completely. The interplay between the plasmonic and photonic resonances is quite complicated and cannot be completely decoupled in the simplified design process outlined here. This relationship may be studied in more detail with full wave electromagnetic simulations using CST Microwave Studio, as discussed below. Simulations were conducted with open boundary conditions in the z direction, and unit cell (Floquet port excitation) in the x and y directions. In addition to the mesh convergence tests, additional tests were run to ensure the Floquet ports had a sufficient number of modes to absorb energy at the boundaries to prevent spurious reflections. The nanograting structure was illuminated using plane waves from the negative z direction at different angles of incidence θ. As shown in Fig. 2(a), at normal incidence (top curve), only one dip in the reflection is present. This dip may be understood in terms of destructive interference between the plasmon resonance and the waveguide modes [18,20]. In fact, there are two counterpropagating modes inside of the waveguide which degenerate into a single dip under normally incident illumination [18]. As the angle of incidence increases, this dip separates into two branches: a lower energy mode and a higher energy mode, as shown in the simulations in Fig. 2(a). Initial experimental measurements of these structures reported by Linden et al. showed that even at normal incidence, two dips were present [18]. A later paper clarified that this was the result of the finite numerical aperture of the lens used to illuminate the sample [19]. We therefore explored the role of numerical aperture for our samples and showed that a larger numerical aperture resulted in a larger splitting between the modes (even at normal incidence), consistent with increasing the angle of incident illumination [18,19]. In order to estimate the range of possible angles introduced by our experimental setup, we calculated the maximum angle in the light cone, based on the numerical aperture of our objective lens (see Appendix B). For our simulations, we chose the angle of incidence to best approximate the finite numerical aperture effect of our lens (ca. 4° to 8°). A more quantitative study of this phenomenon is beyond the scope of this paper; however, our results are consistent with literature [18][19][20] and theoretical expectations.
It should also be noted that these resonance dips are not simply the result of spectral overlap between the plasmon and waveguide modes, rather they are formed by strong coupling between the two modes [19,20]. Due to the resulting highly enhanced fields, these coupled modes are especially useful for sensing applications [21,22]. For the purposes of our investigation here and throughout the rest of this paper, we focus on the higher energy mode (resonance dip ca. 710 nm in Fig. 2(b)). Addressed in in more detail in the next section, this mode (as well as the lower energy mode) is sensitive to changes in the local dielectric environment and is therefore used as our sensor transduction mechanism. The electric and magnetic field profiles corresponding to the resonance dip minimum of the higher energy mode (710 nm) are shown in Figs. 2(c) and 2(d), respectively. Enhanced electric fields near the nanogratings (responsible for transducing changes in the local dielectric environment) may be seen in Fig. 2(c). The dipolar nature of the grating resonances, and the out-of-phase relation to the standing wave in the waveguide, result in destructive interference and produce the observed dips in reflection [20]. Figure 2(d) shows the magnetic field distribution on resonance (710 nm), indicating a quasi-guided mode inside of the waveguide. In both Figs. 2(c) and 2(d), significant transmission (decreased reflection) is evidenced by propagation of energy through the waveguide into the substrate.

Sensor fabrication
Samples were prepared using electron beam lithography (EBL) and lift-off processes. ITOcoated glass slides were purchased from Delta Technologies (GC-4IN-0107), with nominal ITO thickness of ca. 175 nm. The grating width w was tuned from 200 nm to 300 nm by varying the charge dose during electron beam exposure from 150 μC/cm 2 to 350 μC/cm 2 . Scanning electron micrographs confirm the approximate dimensions of the nanostructures, and Fig. 1(b) shows fabricated gratings with nominal width w = 250 nm on top of ITO waveguide. All of the fabricated samples have grating period d x = 500 nm, thickness of 35 nm (30 nm Au on top of 5 nm Ti adhesion layer) and cover an area of 50 µm × 50 µm. The nanograting widths w were chosen such that plasmon resonance would be excited in the 600 nm to 1000 nm range, while the combination of the ITO waveguide thickness and period d x enable coupling to the waveguide mode in this same range [18,19].

Optical measurements
The nanograting sensors of different geometries were first characterized at ambient conditions. All measurements were made using a Nikon OptiPhot epi-fluorescence microscope, fiber-coupled to an Ocean Optics Flame-T UV-VIS spectrometer. Polarized light was focused at normal incidence onto the sample using a long working distance objective lens (Mitutoyo 20×, numerical aperture NA = 0.28), and the reflected light was directed to the spectrometer (see Fig. 3). Figure 3(a) shows how the reference and signal spectra are obtained from the sample.
where R sig is the raw spectrum reflected from the nanogratings, R ref is from an area absent of nanogratings (see Fig. 3(a)), and N is the noise spectrum (dark signal) obtained by blocking all light incident on the spectrometer entrance slit.

Role of nanograting width
Previous reports on waveguide-coupled MPC sensors did not discuss the role of nanograting width, however, this offers another degree of freedom in tailoring the optical response towards high sensitivity biosensing applications. In order to explore this avenue, onedimensional nanograting samples were fabricated with three different nanograting widths: 200 nm, 250 nm, and 300 nm. All samples have the same period (d x = 500 nm) and waveguide thickness (h ITO = 175 nm), and therefore meet the theoretical requirements for exciting the plasmonic-photonic mode as discussed in Section 2.1. Reflection spectra for both experiment and simulation, in transverse magnetic (TM) and transverse electric (TE) polarizations, are shown in Fig. 4. TM polarization (perpendicular to nanogratings) excites the transverse localized surface plasmon resonance, enabling coupling between this strong optical response and the waveguide modes [19], as shown by the strong resonance dips in Figs. 4(a) and 4(c). In contrast, TE polarization does not excite the plasmonic resonance and only the waveguide resonance peaks are seen (primary waveguide resonance at ca. 680 nm in Figs. 4(b) and 4(d)). The waveguide resonance wavelengths are dependent on the waveguide thickness and the period of the gratings, which facilitate excitation of this mode according to Eqs. (1) and (2). The plasmon resonance itself (and therefore the width of the nanogratings) does not affect excitation of the waveguide modes. This can be seen in the TE curves in Figs. 4(b) and 4(d) by noting that the resonance peak wavelengths do not change with grating width. In fact, the waveguide modes may be excited in purely dielectric structures [19]. For the TM curves, as the nanograting width increases, there is a nearly linear redshift in both coupled plasmonicphotonic resonance dip locations. This redshift may be interpreted in terms of increasing effective index within the evanescent fields above the sensor surface, due to an increased gold fill fraction [27]. The higher energy dip (ca. 700 nm) appeared to be more sensitive to changing the nanograting width, suggesting higher sensitivity to the local environment.

Bulk refractive index sensitivity
Next, we characterized the bulk refractive index sensitivity of sensors having different nanograting widths, in order to obtain the baseline sensor performance. As noted above, the higher energy resonance dip appeared to be more sensitive to changes in the surrounding environment. Therefore, we used this feature as the sensor transduction element throughout this study. For the bulk refractive index experiments, we placed the sensors in a fluidic channel and introduced the following mixtures of glycerol and water, ranging from pure water to pure glycerol: {0:100, 20:80, 50:50, 80:20, 100:0}; with corresponding refractive indices: {1.33, 1.36, 1.40, 1.45, 1.47}. In between each mixture, the sample and fluidic channel were both rinsed with ethanol and blow dried with compressed air. In order to quantify the spectral response to changing the refractive index, we tracked the location of the resonance minimum. The center of mass method may also be considered. However, using the resonance minimum is sufficient for our purposes, since the resonances are relatively symmetrical. The results of the bulk refractive index characterization are shown in Fig. 5. The wider nanogratings were more sensitive to changes in the bulk refractive index as shown in Fig. 5(b): 146 nm RIU −1 (w = 300 nm) vs. 110 nm RIU −1 (w = 250 nm), both of which are comparable to other nanoplasmonic sensors [4]. This trend is confirmed qualitatively with bulk refractive index sensitivity simulations (see Appendix C). The reason for the increased sensitivity may be due to stronger coupling between plasmonic and photonic modes for the wider gratings. Christ et al. noted a positive correlation between stronger plasmonic response and stronger plasmonic-photonic coupling [19]. To explore this in more detail, we simulated the electric field profiles (see Fig. 6) for our MPC structures with different grating widths w, at the respective resonance minimums (700 nm to 720 nm, depending upon w; cf. Fig. 4(c)). The resonance minimum was chosen because it corresponds to maximum plasmonicphotonic coupling in MPCs [19,20]. As shown in Fig. 6, the more strongly coupled modes associated with wider gratings exhibit enhanced near fields which are ultimately responsible for the higher sensitivity to the local dielectric environment [39]. These experimental and simulation results indicate that the nanograting width should also be considered as a design parameter for MPC-based sensors, in addition to the nanograting period, waveguide thickness, and angle of incidence previously studied [23,24]. Sensor designs may therefore be further optimized by incorporating this additional degree of freedom in tailoring the optical response.

Low-temperature results
We explored temperature dependence using a cryogenic module with optical access from MMR Technologies [40]. The cryogenic stage was held under high vacuum (10 −6 Torr), to limit thin film formation (adsorption of residual vapors [33]) and to allow efficient operation of the Joule-Thompson refrigerator which cools the stage [41]. The stage was gradually cooled to the lowest temperature attainable and then ramped up to each setpoint of interest, being allowed to stabilize for at least five minutes at each setpoint before capturing a spectrum. A live spectrum view was used to monitor dynamic behavior to confirm that the system had stabilized at each setpoint. (See Appendix D for more details on the low temperature experimental methods.) Figs. 7(a) and 7(b) show the temperature-dependent experimental and simulation results, respectively. As the temperature decreases, an increase in intensity of the plasmonic resonance can be seen in Fig. 7(a), corresponding to reduced plasmonic losses at lower temperature [33,34]. For the plasmonic structures studied experimentally in this work, the change in intensity is modest, only increasing by 13% at 180 K compared with ambient temperature (300 K). This is on the same order of magnitude as plasmonic crystal structures reported previously [33]. The coupled plasmonic-photonic modes (the resonance dips ca. 700 nm and 800 nm) were even less affected by temperature, varying by less than 5% in intensity. (See Appendix E for calculations.) In both cases, the resonance wavelength positions were essentially unaffected by temperature.
In order to explore these phenomena in more detail, we conducted temperature-dependent simulations in CST Microwave Studio (see Fig. 7(b)). The temperature-dependent dielectric permittivity for gold was calculated using the Drude model (see Appendix F). This single oscillator model is only valid for wavelengths above 550 nm, which is sufficient for our application, because the region of interest is above 600 nm [33]. Qualitatively similar behavior is seen between Fig. 7(a) and Fig. 7(b) (solid curves). In both cases (experiment and simulation), lower temperature increases the plasmonic response (i.e., reflection intensity), while the dips corresponding to the coupled plasmonic-photonic modes are relatively unaffected. To better understand why these resonance dips were less affected by temperature, we also simulated the temperature-dependent response of the TE mode (see Fig. 7(b), dashed curves). As discussed in Section 2, the resonances in the TE curves correspond to waveguide modes and do not depend strongly upon the plasmonic nature of the nanogratings. This behavior is confirmed in the dashed curves in Fig. 7(b), where one can see that there is a much weaker temperature dependence in the TE response, compared with the TM curves. This is plausible, since the ITO waveguide, which dominates the TE response, was assumed to be independent of temperature in the simulation. This assumption is reasonable, because, compared to gold, ITO has significantly lower optical losses (which are primarily responsible for the temperature dependence in gold) [42]. We also conducted temperature-dependent simulations for different plasmonic geometries to determine whether enhanced plasmonic response would influence the temperature stability of the resonance dips. Additionally, these simulations covered a wider range of temperatures (73 K to 300 K), to span the maximum range expected for in situ astrobiology applications, as shown in Fig. 8. It was found that even for much a stronger plasmonic response, and wider temperature range, the dips corresponding to the coupled plasmonic-photonic modes remained essentially unaffected by temperature (see Fig. 8(a)). The feature near 610 nm (peak in TE response, dip in TM response) is nearly unchanged by temperature, while plasmonic peak (located at 700 nm in TM response) varies significantly (ca. 24% increase in reflection at 73 K, compared with 300 K). For reference, the temperature response of the plasmonic nanogratings alone was also calculated (see Fig. 8(b)), showing the relatively strong temperature dependence of the gold nanostructures. These results illustrate the benefits of using the coupled plasmonicphotonic mode as the sensor transduction mechanism over wide temperature ranges and demonstrate the suitability of this design for potential astrobiology applications.

Conclusion
We have designed, fabricated, and experimentally characterized metallic photonic crystal (MPC)-based sensors for application to lab-on-chip sensing in cryogenic environments. In addition, we explored the role of nanograting width and showed that wider gratings provided enhanced sensitivity to the surrounding dielectric environment, likely due to resonantly enhanced fields from increased coupling between the plasmonic and photonic modes. The sensitivity enhancement trends were confirmed qualitatively using full wave electromagnetic simulations in CST Microwave Studio. These results suggest that nanograting width should also be considered as a design parameter for the optimization of MPC-based sensors, in addition to grating period and waveguide thickness. The unique optical properties of the coupled plasmonic-photonic modes in these sensors were also shown to be beneficial for the targeted astrobiology sensing applications in extreme environments. Importantly, the coupled plasmonic-photonic resonance location and intensity were shown to be stable across wide temperature ranges (73 K to 300 K in simulation, 180 K to 300 K in experiment). The combination of high sensitivity and stability over wide temperature ranges shows the promise of this sensor design for future in situ lab-on-chip astrobiology applications.

Appendix A: Role of nanograting width and thickness
The simulations in Fig. 9 show that the nanograting width plays a larger role in determining the resonance position of the hybrid plasmonic-photonic modes (the resonance dips in the TM curves). The reflection peak near 550 nm that appears in Fig. 9(b) is due to the bulk metallic response of gold for thicker gratings.

Appendix B: Role of numerical aperture
Objective lenses with higher numerical aperture (NA) introduce angular dependencies at normal incidence, similar to those seen when illuminating the sample off-axis at a larger angle of incidence. The light cones of two 20× objectives having NA = 0.28 and NA = 0.46 were calculated to have maximum angles of incidence θ max = 11° and θ max = 18°, respectively (NA = n sin(θ max ); n = 1.50 for glass). See Fig. 10. To first order, we assume that the average angle incident upon the sample, focused onto the sample through the objective, is approximately θ max /2. This approximation is further justified by the plot in Fig. 2(a), where we vary the angle of incidence from 0° to 12° (the minimum angle to approximately the maximum expected angle in our NA = 0.28 lens). One can see that the middle curves of Fig.  2(a) (i.e. 4° to 8°) bear the most similarity to the experimental measurements. Therefore, the choice to use angles of incidence in this range for comparison of simulation to experiment is reasonable.

Appendix C: Bulk refractive index sensitivity
Wider gratings are more sensitive to changes in bulk refractive index than thinner gratings; simulation results in Figs. 11 and 12 qualitatively confirm experimental results (cf. Fig. 5).

Appendix D: Temperature-dependent experiments
Ultra-high purity nitrogen at 500 psi was purged through the Joule-Thompson (J-T) refrigerator for 30 minutes before each test, while a heater was used to control the stage temperature to 300 K. An initial spectrum was taken at this point. The heater was then turned off and nitrogen pressure was increased to 1800 psi, allowing the J-T cycle to cool the stage to the lowest possible temperature. The heater was then used to balance the cooling of the J-T refrigerator, allowing arbitrary temperature setpoints to be met. Temperature was measured using an integrated silicon diode temperature sensor. Our nanograting sample substrates were relatively large (25 × 25 × 1 mm glass slides) and therefore cooling was inefficient compared with smaller (and thinner) samples we have tested. To reach lower temperatures, the slides may be diced into smaller pieces before definition of the nanograting structures. Hysteresis was also checked each run by taking spectra at each temperature setpoint for both cooling down and warming up, and negligible differences were observed.

Appendix E: Quantification of temperature dependence
For sensor operation, we are primarily concerned with how the plasmonic and coupled plasmonic-photonic features change with temperature. Relative differences in reflection intensity are therefore calculated near the features of interest (see Fig. 13). Change in peak intensity P Δ = |P 180K -P 300K |, and change in dip intensity D Δ = |D 180K -D 300K | are both referenced to peak height P taken at 300 K. The same method is used to quantify temperature dependence in simulations, where 180 K data points are replaced with lowest temperature tested in simulation and therefore maximum difference from ambient (300 K) data. In all cases the dip intensity is less sensitive to temperature variations than the peak. Furthermore, the resonance wavelengths (peak maximum, dip minimum) show negligible temperature dependence with respect to our application. These results are summarized in Table 1.