Bloch surface waves biosensing in the ultraviolet wavelength range – Bragg structure design for investigating protein adsorption by in situ Kretschmann-Raether ellipsometry

We designed a Bragg mirror structure with an SiO 2 top layer to create a resonance in the ultraviolet wavelength range, near the absorption peak position of various proteins. We demonstrate that the wavelength of enhanced sensitivity can be adjusted by proper design of the 1D photonic structure. The possibil-ity to design the wavelength of enhanced sensitivity supports measurements of better selectivity, optimized for the absorption of the target material. Since the width of the resonant peak in the reﬂectance spectra can be sharper than those of plasmonics, and they can be positioned at more favourable regions of the in-strument and material (e.g., in terms of intensity or selectivity), the sensitivity can exceed those of plasmon-enhanced measurements. In this study we demonstrate the main features of the concept at the example of in situ spectroscopic ellipsometry of ﬁbrinogen adsorption in the Kretschmann-Raether conﬁguration. We realized a resonant peak with a full width at half maximum of 3 nm near the wavelength of 280 nm, which coincides with the absorption maximum of ﬁbrinogen. The inﬂuence of depolarization and surface roughness on the measurements, and the potential for improving the current experimental detection limit of 45 pg/mm 2 is also discussed.


Introduction
Optical biosensors are of fundamental role in their field of label-free characterization of various processes related to biomolecules due to the outstanding sensitivity and non-destructive characteristic [1,2]. Among the numerous optical sensing approaches surface plasmon resonance (SPR) spectroscopy [3,4] 5 is one of the most widely used technique for capturing the typically minute changes in the signal, related to e.g., protein adsorption or conformation changes of biomolecules. Biological changes are accompanied with a change in the optical properties and thus biological processes can be studied in SPR approach by measuring the reflectance of a p-polarized probe light. In case of SPR spec- 10 troscopy usually a thin Au film is used as a sensing layer. With the help of the so-called Kretschmann-Rather configuration [5] propagating surface plasmon oscillation can be excited by incident light at the interface of Au layer and aqueous ambient. If appropriate conditions are fulfilled, the incident light couples with surface plasmons, thus a dip appears in the reflectance spectrum. 15 The exact wavelength (λ) value of this dip is highly dependent on the thickness (d) and optical properties of the Au layer [6], the angle of incidence (θ) of the light beam, the optical properties of the configuration and most importantly the optical properties of the investigated ambient near the Au surface.
Enhanced sensitivity can be reached by various approaches in SPR spec-20 troscopy. One may not only monitor the reflectance, but also the phase information provided by novel measurement setups [7] or by spectroscopic ellipsometry (SE) [8]. SE is a method with outstanding sensitivity to the optical properties of a solid surface that makes it ideal for biosensing applications [9].
The combination of SE and SPR spectroscopy can be realized reasonably [10]. 25 This combined configuration is usually referred in the literature as total internal reflection ellipsometry (TIRE) [11]. TIRE has several advantages, such as the large freedom in terms of the θ and λ range compared to a traditional SPR configuration of either a λ-or a θ-tracking principle. Compared to the biological measurements through the liquid ambient with SE, TIRE is far more sensitive 30 as well as the available λ range is wider since the absorbance of the aqueous ambient is not present anymore.
Constructing novel layer structures can also contribute to the enhanced sensitivity [12]. By using not only a bare Au layer but also one or more 2D layers (e.g., graphene, molybdenum-disulfide) on top of the Au film may lead to a 35 superior sensing performance [13,14]. Another layer structure with improved sensitivity utilizes the so-called long range surface plasmons (LRSPRs) [15,16].
LRSPRs are special surface modes that are usually enhanced when using a thin metal layer positioned between two dielectric media with similar refractive indices (n). 40 It is also possible to realize a TIRE biosensor without a thin Au layer and the absence of any SPR related material (usually metal) in a sensing structure has already been proposed [17,18,19,20]. As an example, a new configuration has been introduced recently for biosensing applications, the so-called Braggmirror structure (BMS) [21,22,23,24,25]. Similar to SPR, electromagnetic 45 waves (the so-called Bloch surface waves) are confined to the surface of the layer structure which show an exponential decay of the field inside the layered medium and in the liquid ambient. These tailored periodic layer structures have several improved features compared to the usually used Au layers [26].
One of the most important advantages is the large freedom they provide in 50 terms of operating wavelength (OW). Carefully choosing the optical properties and thickness of the layers in the BMS one can achieve basically any OW that is aimed. The resonance peaks are usually narrower -due to the small absorption of the dielectric materials constituting the structure -leading to an improved performance over the SPR sensor, and in case of BMS, s-polarized light can 55 also be used for surface wave excitation. Surface chemistry can also be more convenient, since dielectric materials (e.g. SiO 2 ) are allowed instead of a Au layer on the top at the reaction interface.
In this work, a novel BMS of alternating SiO 2 and ZrO 2 layers on fused silica substrate was introduced that can be used as a biosensor in the ultraviolet (UV) 60 range of the wavelength spectrum. Other techniques [27] working with a plasmonic structure have also been proposed [28,29] for biosensing in the UV range, however, BMS can offer additional attractive properties. The tailored OW can be chosen in the range of λ = 265 − 365 nm (near to the absorption peak of several proteins [30]), depending on the angle of incidence. The sensor perfor-65 mance of the proposed BMS was demonstrated by investigating its response to glycerol solutions in a wide %(w/w) range as well as to bovine fibrinogen (Fgn) solution in phosphate buffered saline (PBS). The results are compared to the performance of an SPR-based structure consisting of a single Au layer. The characterization of these structures was performed using spectroscopic ellipsometry 70 (SE), thus the phase information was also measured providing more information and enhanced sensitivity. The SE measurements with BMS (BMS-SE) and SPR (SPR-SE) were evaluated by constructing appropriate optical models. The effect of various imperfections (e.g., surface roughness, angular spread of the light beam) to the sensitivity were numerically analyzed. 75

Preparation of model solutions
For characterizing the optical biosensing performance of BMS-SE glycerol solutions (from VWR, glycerol bidistilled 99.5%) of various concentrations (ranging from 0 to 29% (w/w)) in ultrapure deionized (DI) water as well as Fgn (from 80 Sigma-Aldrich) solution in prefiltered 10-mM PBS with a concentration of 0.5 mg/mL were prepared at room temperature (RT).

Refractometry
A standard automatic refractometer (J157 Automatic Refractometer) was used to measure the refractive index (RI, n) of glycerol solutions at RT with 85 an accuracy of RI ± 0.0001 and thus to obtain an independent measurement to compare with the SE results both on the BMS-SE and SPR-SE structures.

Spectrophotometry
Proteins usually have an absorption maximum at 280 nm due to the absorbance of two aromatic amino acids tryptophan (Trp) (max. at 280 nm) and 90 tyrosine (Tyr) (max. at 275 nm) and to a smaller extent also cystine (i.e., disulfide bonds) [31].
The peptide groups of the protein main chain absorbs light with a maximum at about λ = 190 nm. The aromatic side-chains of Tyr, Trp and phenylalanine (Phe) also absorb light in this region and besides, they also absorb in the λ = 95 240 − 300 nm range. Disulfide bonds that form between two cysteine residues also show an absorbance band near λ = 260 nm.
To obtain the UV and visible absorbance spectra of the glycerol and protein solutions spectrophotometric measurements were carried out. The spectrophotometer (Agilent 8453) used in this study had two light sources, a tungsten-100 halogen and a deuterium lamp, both for covering a wide λ range from the UV to NIR (ca. between 190 and 1100 nm). The light from the sources passed through a monochromator and was focused into the fused silica (FS) cuvette filled with the investigated solution. Subsequently, the transmitted light was detected by a photomultiplier.

105
The absorbance (A) was calculated from the transmittance (T ) given as T = I/I 0 , where I is the transmitted light intensity and I 0 is the intensity of the light beam before the l = 10 mm long cuvette. The absorbance was then calculated as A = − log 10 T .
All the measured absorbance spectra were measured at RT and the spectrum 110 of ultrapure deionized (DI) water with a resistivity of 18.2 MΩcm was used as a background measurement.

Spectroscopic ellipsometry
A Woollam M-2000DI rotating compensator spectroscopic ellipsometer was used in the range of λ = 191 − 1690 nm at variable θ utilizing the  Raether (KR) geometry that allows θ up to 75 • when using the focus extension.
The dual-source equipment allows high intensities in the UV spectral range, which is of primary importance in the current study. We also utilized an improved hemisphere for the KR ellipsometry (KRSE) setup that contributed to the outstanding signal-to-noise ratio in the crucial spectral range below 300 nm.

120
The KR cell can be mounted on the mapping stage of the ellipsometer [32], and the optical adjustment of the system is supported by the control of the mapping stage. The optical parameters of the KR setup (focusing lenses, hemisphere, glass slide, index matching liquid) enabled us to use the λ = 200 − 1690 nm spectral range of the ellipsometer. The spectral resolution bandwidth is around 125 5 nm and 10 nm in the UV/VIS and in the near infrared wavelength ranges, respectively. The spectral density of the experimental data points is about 1.6 nm and 3.4 nm in the UV/VIS and in the near infrared wavelength ranges, respectively. The angular divergence is smaller than 0.3 • without focusing, that can be significantly higher while using KRSE.

130
It is important to point out that the depolarization caused by angular spread and spectrometer bandwidth is hard to separate, since their effect on the measured spectra is similar. Thus, a bandwidth value specified by the manufacturer was used in this analysis and only the angular spread was fitted simultaneously with the ellipsometric angles.

Flow cell design
In order to exploit the potential of SE for in situ TIRE measurements, a 10-µl flow cell has been realized with a KR configuration consisting of an FS 145 hemicylinder (Fig. 1A). This configuration makes the investigation of the optical properties possible in a liquid ambient in the range of λ = 200 − 1690 nm, and also in a wide angle of incidence range of θ = 45 − 75 • . For ensuring the best performance (e.g., due to ensuring a normal incidence at the air/hemisphere interface over the whole illuminated spot), a focused light beam is used during 150 the measurements with a spot size below 1 mm.
It is not exploited in this work, however, there is room for miniaturization [34] using the same concept. Although with the current hardware the beam cannot be focused below a diameter of approximately 300 micron, if the scanning capability is not used, the lateral size of the flow cell can theoretically be as small 155 as the spot itself, with a depth also smaller than a millimeter, which results in a microliter-size cell. With a restricted wavelength range the spot size can also be smaller. This approach can also be combined with imaging ellipsometry having a lateral resolution down to one micron.  Note that electron beam evaporation may produce porous oxide layers [35] that can adsorb water from the ambient causing a drift in the measured signal. In this study we found, however, that after a relaxation time of a day, all these drifts were eliminated and the signal was stable.

Results and discussion
The freshly prepared samples were then cleaned using a rinse of DI water and blown by nitrogen stream. The characterization was carried out in λ = It was supposed that all the layers with the same composition have the same complex refractive indeces (n = n + ik, where k is the extinction coefficient), thus their values were coupled in the analysis. The optical properties of the alternating SiO 2 and ZrO 2 layers were described using the Cauchy term: where λ corresponds to the incident wave in vacuum in unit of µm, the parameter A C is dimensionless, while B C and C C are in the units of µm 2 and µm 4 , respectively. For describing the absorption of ZrO 2 layers the Urbach-tail was also included in the optical model: where D C is the amplitude, F C is an exponent factor and γ is the band edge in the unit of µm. Since it is correlated with the other parameters, the band edge was not fitted, and its value was fixed at the lowest measured wavelength value of λ = 0.2 µm. The dielectric optical properties of the FS substrate was described by using the Sellmeier term: where ε ∞ , A S , B S and E S are the offset, amplitude, center energy and position ε ∞ was fitted. The relation between the complex dielectric function (ε) andn is described asε = ε 1 + iε 2 =n 2 .
During the fitting process, the root mean square error (RMSE) was minimized and we accepted the calculated values of the parameters as the true physical values at lowest value of the RMSE [36]: where p is the number of the measured λ values, m is the number of the unknown 175 parameters in the model, 'exp' and 'cal' denote the measured and calculated N = cos(2Ψ), C = sin(2Ψ) cos(∆) and S = sin(2Ψ) sin(∆) values, while σ is the standard deviation of the measured data. The depolarization is given in % and defined as Depol. = (1 − P 2 ) · 100%. Here P denotes the polarization calculated The spectra of the measured ellipsometric angles and the fitted curves are shown in Fig. 2A. In this analysis not only the measured values of Ψ and ∆ were fitted but simultaneously the depolarization and the measured transmission intensity (the latter from an independent measurement on the same sample) were also taken into account (Fig. 2C). The depolarization emerges from the 185 back-side reflection of the light beam due to the transparency of the relatively thin FS substrate. If the depolarization is zero, there is no back-side reflection.
In case of depolarization, however, incoherent interference modeling must be included in the model. This effect is present mainly in the range of λ = 200 − 1000 nm for all investigated θ.

190
The optical properties and thicknesses of the layers shown in Fig. 2D and Table 1 were calculated from the fitted spectra of Fig due to its n similar to the substrate); ε ∞ was also fitted to describe the optical properties of the FS substrate. It is notable that in spite of the relatively large number of fitted parameters, the confidence limits (as shown by the confidence limit values in Table 1) and parameter correlations are small. This is partly due to the large differences in the values and spectral distributions of n and k of  Table 1. This step was vital, since a significant depolarization was evident from previous measurements in the KR configuration due to the angular spread of the focused beam and to the spectrometer bandwidth [37].

210
This depolarization has usually a huge effect on the measured spectra in KRSE (Fig. 1B).
As the first step, intensity spectra for both polarizations were calculated for λ = 200 − 400 nm at θ = 74 • using TMM and FE (Fig. 1). An excellent agreement was found ensuring that the calculations are physically relevant in terms  Supporting measurements were also carried out using SE in the same configuration. The solution was measured in the same quartz suprasil cuvette (from Hellma Analytics) and the transmission intensity data were collected by SE.

245
The baseline was DI water for this investigation, and the absorption coefficients

305
The phase information, represented by the ∆ ellipsometric angle was also investigated for both structures (Fig. 7). The most sensitive wavelengths were identified for each glycerol solution transition (∆n ij = n j − n i @ 633 nm, where i and j denote the regions marked in Fig. 6A) and the maximum of absolute changes in ∆ were plotted at a given wavelength in Fig. 7B. For demonstrating 310 this method four curves were plotted in Fig. 7A for transitions between glycerol samples '5'→'6' and '6'→'7' at four different wavelengths. From this analysis the measurement of BMS-SE shows an enhanced sensitivity compared to SPR-SE.
Note that in the case of BMS-SE we suppose that due to its sharp-resonance manner we are not necessarily able to find the biggest change in ∆ ellipsometric

Investigation of Fgn adsorption
The spectral range of in situ bioellipsometry is usually limited either by the transparency of the water [40,41], the optical components or the lack of information of the dispersion of protein in the UV range. The n and k spectra of 335 protein can usually be fitted using a polynomial [40] and an exponential function, respectively. However, as the transmission and absorption results in Fig. S1 show, the polynomial and exponential dispersions must be completed with an oscillator model for an accurate description of the features below λ ≈ 280 nm. to the thickness of an adsorbed monolayer. From this analysis two peaks were identified in k, which were fitted by two Kramers-Kronig consistent Gaussian oscillators. ε 2 for the jth oscillator is given by where E is the photon energy of incident light in eV, σ j = Br j /2 ( ln 2). Here,

340
A G is the amplitude, E n is the center energy in eV and Br is the broadening in eV. Γ G is a convergence series that produces a line shape for ε 1 in a Kramers-Kronig consistent manner [42]. An additional parameter, a constant value from KK-integration ε G∞ was also fitted, and was found to be ε G∞ = 1.27 ± 0.03. k is in turn the imaginary part of (ε 1 + iε 2 ) 1/2 . The calculated results are shown 345 in Fig. 8C and in Table 2, wheren and the oscillator parameters of the Fgn layer are presented. Note that the calculated optical properties may be valid only for an Fgn layer with a given volume fraction, since numerical random sequential adsorption models showed that there is a maximum coverage that can be achieved during protein adsorption [43].

350
During protein adsorption then values of the Fgn layer (n F gn ) were fixed and only the thickness of the layer (d F gn ) was fitted. The surface mass density where n P BS is the refractive index of the PBS ambient, and a denotes the refractive index increment of the Fgn solution (n s ) with the Fgn concentration (dn s /dc F gn ) at the wavelength value of 632.8 nm. The value of a was fixed at 0.18 mL/g. Γ F gn is calculated in the unit of ng/mm 2 .
The temporal evolution of Γ F gn is shown in Fig. 8D for both the BMS-SE Au, respectively). The calculated SMD is in good agreement with several other results published before [45,46,47]. The absolute changes in Ψ and ∆ during 360 the adsorption are also presented for the most sensitive λ in Figs. 8A and 8B.
The variations corresponding to the BMS are comparable with those of the SPR, revealing an excellent sensitivity in both cases. Apart from the amplitude ratio (tan(Ψ) = |ρ| = |r p /r s |) the phase (∆) of ρ is also measured by SE showing a variation that is six times larger than that of Ψ (Fig. 8B vs. 8A) for both 365 the BMS and SPR structure, which leads to a sensitivity that is significantly larger than that of simple amplitude and intensity measurements [6]. Also note that apart from the capability of the accurate sensing at the selected λ and θ, SE adds a modeling opportunity due to the large number of data in a broad spectral range. Although most of the spectral regions do not offer a high 370 sensitivity, the models can be used to have an insight in the layer structures and inhomogeneities, to have a better understanding of the complex processes that occur during the high-sensitivity variation of the signal at the most favorable λ.
The limit of detection was also calculated for the smallest detectable surface There is also room for the improvement of σ in other areas of instrumentation that increases the signal-to-noise ratio and the stability.

385
The biosensing capability and properties of BMS in terms of bulk and thin film sensitivity and adsorption monitoring was discussed in the above sections.
Although it was not investigated here in detail, it is important to point out the also from the layer structure. The sensitivity in this case is given as the variation in n or SMD corresponding to the smallest detectable change in ∆ defined as five times the noise of a ∆ spectra at a given angle of incidence. The simulation of 0.05 • . The same calculation was also performed with/without a 6.5 nm thin Fgn layer that has the optical properties as shown in Fig. 8C. The results are presented for n (bulk) and SMD (layer) sensitivity in Fig. 9. It is remarkable that in the range of λ < 300 nm completely different structures appeared in

Conclusion
Three pairs of SiO 2 and ZrO 2 layers were evaporated on FS slides to create a multilayer structure with a sharp (FWHM=3-4 nm) absorption feature in surface-enhanced internal reflection Kretschmann-Raether configuration for SE.     The inset in A shows the schematic arrangement of the spectrophotometric measurement.