Layered material platform for surface plasmon resonance biosensing

Plasmonic biosensing has emerged as the most sensitive label-free technique to detect various molecular species in solutions and has already proved crucial in drug discovery, food safety and studies of bio-reactions. This technique relies on surface plasmon resonances in ~50 nm metallic films and the possibility to functionalize the surface of the metal in order to achieve selectivity. At the same time, most metals corrode in bio-solutions, which reduces the quality factor and darkness of plasmonic resonances and thus the sensitivity. Furthermore, functionalization itself might have a detrimental effect on the quality of the surface, also reducing sensitivity. Here we demonstrate that the use of graphene and other layered materials for passivation and functionalization broadens the range of metals which can be used for plasmonic biosensing and increases the sensitivity by 3-4 orders of magnitude, as it guarantees stability of a metal in liquid and preserves the plasmonic resonances under biofunctionalization. We use this approach to detect low molecular weight HT-2 toxins (crucial for food safety), achieving phase sensitivity~0.5 fg/mL, three orders of magnitude higher than previously reported. This proves that layered materials provide a new platform for surface plasmon resonance biosensing, paving the way for compact biosensors for point of care testing.


Transfer protocols
Graphene SLG can used to protect Cu SPR for long (~year) periods of time 1 . SLG is grown by chemical vapour deposition (CVD) on a 35m-thick Cu foil, following the recipe in Ref. 2 . Cu is loaded in the growth chamber and the temperature (T) is raised up to 1000C in~200mTorr H2.
Then T is kept constant for 30mins. Next, 5 sccm of CH4 are added to the 20 sccm H2 flow to enable the growth process, which lasts 30mins. The sample is then cooled at~1 mTorr to room temperature (RT) and unloaded from the chamber.
The material is characterized by Raman spectroscopy using a Renishaw InVia spectrometer equipped with a 50X objective at 514nm. Figure S1 plots the Raman spectrum of SLG as-grown on Cu (red line), with Cu background photoluminescence removed 3 . The D peak is negligible at indicating negligible defects 4,5 . The 2D is a single Lorentzian, fingerprint of SLG 6 . SLG is then wet transferred on another 35m-thick Cu, Fig. S2. A PMMA layer is spin-coated at the surface of SLG/Cu as a mechanical support and then it is dropped at the surface of a solution of ammonium persulfate (APS) and DI water for the etching of Cu. The PMMA membrane with attached SLG is then moved to a beaker filled with DI water for cleaning APS residuals. The PMMA+SLG membrane is lifted with the target Cu substrate.
The material is once again characterized by Raman spectroscopy, Fig. S1, blue curve. The D peak is still absent, indicating that the transfer has not damaged the sample. hBN hBN is a dielectric that has a large band gap 7 but is similar to SLG in crystalline lattice. As a result, hBN can protect metal SPR as efficiently as SLG. We use two transfer methods, P1 and P2.

P1:
1L-h-BN is grown by CVD on 100m Fe foils (99.8%, Goodfellow) as described in Ref. 8 . Fe foils are cleaned in acetone and isopropanol and loaded into a custom cold-wall CVD system with a base pressure~10 -6 mbar. T is ramped to 920C as measured by an internal thermocouple in 1.5x10 -2 mbar Ar (BOC, 99.9995%). The atmosphere is then switched to 10 -2 mbar NH3 (BOC, 99.98%) for 60mins. Borazine (Fluorochem) is added at a leak rate of 2x10 -4 mbar for 2.5 hours to achieve full-coverage. The system is then cooled at a rate of 200C/min. A scanning electron microscope (SEM) image of the as-grown film is in Fig. S4a.
This consists of hBN domains with sizes ~100m as seen from a shorter experiment with incomplete coverage in Fig. S4b. hBN samples are then wet transferred. For this, PMMA is spin-coated on Fe/h-BN foils and the PMMA/h-BN is delaminated using electrochemical bubbling in a 1M-NaOH 8 . The PMMA/h-BN film is then transferred to sequential deionised water baths and lifted onto the target substrate. After drying and baking for 5 minutes at 120C, the polymer is dissolved in acetone. To remove possible surface oxide that could appeared on the Cu substrates during air exposure or during transfer, the Cu/h-BN samples are annealed in 50 mbar Ar/H2 P2: CVD 1L-hBN is grown on Pt foil using ammonia borane as a precursor 9 . The Pt foil is loaded into the centre of a vacuum quartz tube in a furnace, and ammonia borane is placed in a sub-chamber. The furnace is heated to 1100°C under 10sccm H2. The sub-chamber is heated to 150 °C for the decomposition of ammonia borane. Growth is initiated by opening the sub-chamber valve. During growth, for 30min the pressure is maintained at 0.13 Torr.
After growth, the furnace is cooled to RT under H2. 1L-hBN is then transferred onto the target Cu substrate using electrochemical delamination.
Optical images of hBN on quartz/Cr(1.5nm)/Cu(43.5nm) fabricated by e-beam evaporation are shown in Fig. S5a. The surface of hBN-covered Cu is smooth with a small amount of visible defects. These do not influence the SPR resonance shown in Fig. S5b, with almost complete darkness (min~0.8) at 620nm for 45 incidence angle. However, the SPR properties, such as darkness of the resonance and the resonance quality factor, are not stable in air and water. Figure S5c shows the SPR deterioration (with increasing minimum reflection to min~2) measured in the same conditions as in Fig. S5b, after 13 days in air. At the same time, the SPR in water red shift (~2.5nm) with a characteristic time of 1 hour. The poor stability of the hBN protected samples is probably connected with defects.
Analogous results are obtained using P1. An optical image of the final structure is shown in Fig. S5d. Fig. S5e,f indicate excellent SPR amplitude and phase characteristics.
These SPR curves survive for a period of months in air. However, the stability in water is not good. Figure S5g shows an image of the sample after measurements in ATR in water, indicating some surface deterioration. The evolution of SPR curves for hBN protected samples in water is shown in Fig. S5h. This confirms deterioration over ~1 day.

Direct growth
Graphene and hBN Figure S7a shows an optical image of a sample where SLG is directly grown 43.5nm Cu by CVD. This significantly increases the Cu roughness. As a result, the SPR curve becomes broad, Fig. S7b, and Ψmin in the air cannot reach zero, which makes ultra-sensitive biosensing difficult 11,12 . The increased roughness can be due to the high temperatures arising during CVD.
CVD hBN on a SPR film also induces deterioration of the metal surface to such degree that the SPR resonances are of extremely poor quality (not shown here).

CNMs
Cu/glass substrates are introduced in a UHV Multiprobe system (Scienta Omicron) with a base pressure < 2x 10 -10 mbar. The Cu substrates are in situ cleaned by Ar + sputtering (1 keV, 10 mA, Focus) at~2x10 -7 mbar for 10 minutes until no carbon is detected by X-ray photoelectron spectroscopy (XPS, Scienta Omicron, monochromatized Al Kα X-ray source with Argus energy analyser that has a resolution of 0.6 eV). The NBPT SAMs are prepared via vapour deposition for 30 minutes using a Knudsen-type evaporator (TCE-BSC, Kentax) from a quartz crucible heated to 100 °C. Heating of NBPT increases the chamber pressure to~10 -6 mbar, as detected by N2-calibrated vacuum ion gauge. The Cu substrates are kept at RT during vapour deposition. Figure S8a presents high resolution XP spectra (C1s, N1s and S2p) confirming the SAM growth on the Cu/glass substrate. The molecules are covalently bound to Cu by thiolate, as confirmed by the sulphur signal, and form a 1.2 nm film, as calculated from the attenuation of the Cu2p signal before and after growth. The N1s signal consist of two species, attributed to the nitro group of the NBPT molecules at a binding energy (BE)~405.7eV 13 and an amino group (BE 398.8 eV) arising from partially reduced molecules 13 . Afterwards SAMs are crosslinked into CNMs in the same UHV chamber using a low energy electron gun (NEK-150, STAIB Instruments) at 50eV and an electron dose~50 mC/cm² and monitored again by XPS. Due to the irradiation, the C1s signal is broadened as new chemical species are created by crosslinking and the thickness is slightly decreased due to desorption of detached fragments. Furthermore, the nitro signal is completely reduced to amino groups and the sulphur signal is modified by the creation of copper sulphides. These results are in good agreement with previously published CNMs grown via vapour deposition in UHV 14 .
An optical image of a sample is shown in Fig. S8b. The direct growth of CNM films suppresses plasmonic properties of the final structure. Indeed, the SPR curves measured in air for a directly grown sample, Fig. S8c, are not as good as the SPR curves of the samples obtained by CNM transfer, Fig. S6, with minimum Ψmin~5 (which corresponds to the minimum reflection~4%) at 45 incidence. After keeping the sample in air for 6 months, the SPR wavelength redshifts ~26 nm and the corresponding incident angle is~47.9, Fig. S8d.
Thus, some Cu oxidation (or relaxation of the structure) still happens even after the growth of CNM. Fig S8e plots the SPR response in water. The SPR resonance wavelength shifts significantly over 2 days. Thus, a CNM layer grown on a metal substrate cannot be used for effective protection of metals for biosensing. Most likely, roughness is introduced due to the sputtering of the sample before SAM growth and the crosslinking with low energy electrons into CNMs. This might be overcome by optimisation of the fabrication parameters.

Metal oxide
To increase protection, an additional layer of oxide can be added. Even a thin (~10nm) metal oxide layer evaporated on Cu can protect the SPR properties. Fig.S9a shows the SPR curves of 10nm HfO2 covered Cu. We get excellent SPR curves with narrow (quality factor ~10) and Since it is more difficult to arrange bio-functionalization for a dielectric material, these additional oxide layers can be covered with SLG to make use of SLG functionalization as described below.

II. Functionalization for biosensing
Functionalization of SLG-covered Cu SPR sensors is necessary to achieve selectivity and to detect the target molecules (SPR respond to the overall change of local refractive index 15 ). The electrochemical protocol of functionalization with COOH is as follows: 1) Place 0.052 mmol 4-NH2-3,5-F2PhCOOH into a glass vial.
3) Add 25 ml of Milli-Q water and make sure that all solid is dissolved.

10) Once finished, disconnect electrodes, take out the sample and wash it with excess water.
Dry naturally in the air. In case of COOH containing impurities: dip the grafted sample into 1% NaOH, then rinse with water, then dip into 1% acid (e.g. HCl or phosphoric), rinse with excess of water, dry naturally. An analogous protocol can be used to functionalize SLG with NH2 groups.

III. Biosensing of HT-2 1. Graphene protected copper
The functionalization of SLG protected Cu can be done using biochemistry. To detect HT-2, the SPR sensor needs to be functionalized using 1-Pyrenebuturic acid N-hydroxy-succinimide ester as linker and anti-HT-2 toxin Fab fragments as a receptor 16 . First, 1-Pyrenebuturic acid N-hydroxy-succinimide ester linker solution (2 mg/mL) in 100 % MeOH is prepared. After sonication, the linker solution is incubated for 1 h at RT without shaking to ensure saturation.
We then filter the saturated solution with a disposable filter unit attached to a syringe, and then put the sensor chip into the filtered solution. Filtering removes the undissolved linker and the resulting solution is clear. After one-hour incubation, the chip is washed by pure 100 % MeOH and 1 × PBS (pH 7.3). Then, it is transferred to 50 µg/ml of HT2-10 Fab solution in 1 × PBS (pH 5), and incubated for 20 min at room temperature. Next, the chip is moved from the antibody solution to 100 mM Ethanolamine (1M Ethanolamine stock solution (pH 8.5) diluted 1:10 in distilled water), and incubated for 10 mins. The Ethanolamine solution is used to block the unreacted active sites in the linker. Finally, the chip is washed with distilled water and stored in distilled water before SPR measurements.

Biacore T200 benchmarking
We perform benchmarking experiments of HT-2 detection with HT2-10 Fab fragments using a commercial Biacore T200 instrument (GE Healthcare) and Sensor Chip CM5 with carboxymethylated dextran matrix (GE Healthcare) at 25 °C. We use EDC/NHS chemistry to immobilize HT2-10 Fab fragments on an the active surface and anti-mycophenolic acid (MPA) Fab fragment, which does not cross-react with HT-2, on the reference surface. 1 x PBS-P (20 mM phosphate buffer, 2.7 mM KCl, 0.137 M NaCl, 0.05% P20, GE Healthcare) supplemented with 0.1% DMSO is used as running buffer. Six concentrations of HT-2 (0, 0.1, 1, 10, 100 and 1000 ng/ml) in 1 x PBS-P with 0.1% DMSO are injected at a flow rate of 30 µl/min for 3 mins. After dissociation for 6 mins, regeneration is done with 10 mM glycine-HCl, pH 2 (contact time 30 s, flow rate: 30 µl/min) followed by stabilization for 90s. All samples are analyzed in replicates. The results are analyzed with the Evaluation T200software with a 1:1 Langmuir binding model.
The obtained sensorgrams are shown in Fig. S12. This yields an amplitude detection limit~1 ng/ml with the Biacore T200. This is 6 orders of magnitude worse than the phase detection limit of our SLG-protected Cu chips. The extracted constants from a Langmuir model are: the absorption rate ka~10 6 1/(Ms), kd~0.03 1/s and KD~10 -8 M.

Detection of HT-2 toxin in canned beer.
To check that SLG-protected Cu chips can be applied to detection of HT-2 toxin in a commercial beer, we make use of GPC SPR chips functionalized by HT2-10 Fab fragments as described above. Then, we measure the SPR curves in PBS buffer solution before and after pumping a popular canned beer through the flow cell containing SPR chip. Figure S13 shows that the shift of resonance is negligible (<0.1nm) in both amplitude and phase spectra. The change of the phase is <0.8 which, combined with a sensitivity of 9/(pg/mL), gives a HT-2 toxin concentration <100 fg/mL. This is well below 10 ng/mL, which provides a tolerable daily intake for a 70 kg person drinking one pint of beer a day 17 .

Negative control of graphene-protected copper SPR biosensor using neosolaniol
Non-specific binding on the surface of sensor chip can affect the results of biosensing. In