An improved surface enhanced raman spectroscopic (SERS) method operating in the shortwave infrared

Abstract

A method of surface-enhanced Raman spectroscopy includes positioning an analyte in a field-enhancing region of a nanophotonic element, irradiating the analyte with a continuous wave excitation laser operating at a user-defined frequency in a range from near infrared (NIR) to short-wave infrared (SWIR) to generate a signal to cause photon scattering from the analyte and detecting the photons with one or more photodetectors in the SWIR. A surface-enhanced Raman spectroscopy system includes a laser operating at a frequency having a range from near-infrared (NIR) to short-wave infrared (SWIR) that emits light in a spectrum between 700 and 2500 nanometers, a nanophotonic element to receive the light from the laser, and one or more SWIR detector to detect photons scattered by an analyte on the nanophotonic element.

Description

Patent Application U Cal No. BK-2025-010-2 MN No. 407869-0222AN IMPROVED SURFACE ENHANCED RAMAN SPECTROSCOPIC (SERS) METHOD OPERATING IN THE SHORTWAVE INFRAREDTECHNICAL FIELD

[0001] This disclosure relates to surface-enhanced Raman spectroscopy (SERS), more particularly to performing SERS in the shortwave infrared spectral region (SWIR) with a high-fidelity open-cavity enhancement structure combined with source-sweep spectral acquisition and time-correlation analysis.BACKGROUND

[0002] Raman spectroscopy, the inelastic scattering of light by molecular vibrations or solid- state phonons, is a cornerstone technique in chemical analysis, biological imaging, and materials characterization. To increase inherently weak Raman signals, surface-enhanced Raman spectroscopy (SERS) employs plasmonic or dielectric nanostructures that locally amplify the electromagnetic field. Despite widespread use, conventional SERS suffers from two major limitations: (1) poor substrate reproducibility and unstable “hot-spot” enhancement in metallic architectures, and (2) intense fluorescence or carrier-induced background that obscures weak Raman features of target analytes.

[0003] Metallic SERS substrates such as gold, silver, and aluminum offer large field confinement but exhibit significant ohmic loss, limiting the optical quality factor Q Q and introducing local heating and charge-transfer artifacts. As a result, the Purcell-factor-based Raman enhancement Fpoc Q I V Fp ocQ / V requires sub-10 nm gaps to achieve singlemolecule sensitivity, which in turn perturbs the analyte’s native state and precludes analysis of larger biological molecular and particles. Ohmic heating and charge transfer in metallic substrates limits the approach of increasing the Raman signal with the intensity of the pump laser. Dielectric nanostructures - such as metasurfaces made of silicon, silicon nitride, or gallium nitride - provide higher fabrication reproducibility and reduced non-radiative loss but typically provides orders-of-magnitude lower signal enhancement. Their larger mode volumes and lower absorption suppress heating and hot-spot fluctuations, yet the diminished signal often does not allow single-molecule detection and greatly limit the temporal dynamics accessible via frame-by-frame Raman spectroscopy. Increasing laser intensity is furtherconstrained by background luminescence from the Raman enhancement substrate and analyte, particularly in the visible spectral range (400-700nm).

[0004] Consequently, a key unmet need is a high-signal, low-background SERS method capable of probing molecules or particles within larger optical mode volumes, free from substrate-induced perturbations and without fluorescent or plasmonic background. Such a method would extend quantitative Raman spectroscopy to analytes ranging from small molecules to exosomes, viruses, and tissue fragments, and enable advanced assays in protein sequencing and single-particle analytics.

[0005] Cavity-enhanced Raman spectroscopy has been explored as a route to stronger lightmatter interaction, employing closed Fabry-Perot or whispering-gallery resonators to increase the photonic density of states. However, these geometries are poorly suited to biological or heterogeneous samples because the analyte must reside within the optical cavity, limiting sample compatibility and dynamic measurements. Fabry-Perot cavities are further challenging to reliably stabilize while also allowing straightforward analyte integration.

[0006] Cavity-enhanced Raman spectroscopy has been used to improve signal strength by enhancing the electromagnetic field around analyte molecules, including gases (T. T. Nguyen et al., Anal. Chem., 95, 6475 (2023)), and nanomaterials (S. Nie et al., Nat. Commun., 7 , 12155 (2016)), with some application to cells(M. J. Lee et al., Anal. Chem., 89, 9440 (2017)). Such cavities, in principle, allow source-sweep spectral acquisition — that is, acquiring the Raman spectrum by sweeping the frequency of a narrow excitation light source and recording the single-wavelength Raman intensity for different excitation frequencies (See US Patent No. 11,307,092).

[0007] Recent reports have demonstrated dielectric metasurfaces supporting high-Q resonances such as quasi-bound states in the continuum (quasi -BICs) or Fano modes, but their use in SERS remains limited. The narrow bandwidth of such resonances, while advantageous for spectral selectivity, is generally mismatched to broadband Raman emission.

[0008] The monolithic nature of dielectric metasurfaces eliminates the need for active photonic resonance frequency stabilization as needed for Fabry-Perot cavities, while also allowing straightforward analyte positioning in the high-field region, either stochastically, or directed binding e.g., through DNA origami or other surface functionalization.

[0009] Accordingly, there remains a need for an open-cavity, high-Q Raman architecture that combines the spectral purity and low loss of dielectric resonances with broadband Raman accessibility, and that enables quantitative, background-free acquisition from delicate or extended samples without local heating.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] FIG. 1 shows an embodiment of a SWIR SERS apparatus and method.

[0011] FIG. 2 shows the method of acquiring the Raman spectrum and spectral correlation through sweeping of the excitation laser.

[0012] FIG. 3 shows a graph of evolution of background intensity, Raman intensity, and quantum efficiency of superconducting nanowire single-photon detectors sensitive in short wave infrared.

[0013] FIG. 4 shows the measured photon count rate from a substrate covered in Rhodamine 6G, a common Raman dye, recorded with superconducting nanowire single-photon detectors (SNSPDs) and using long pass optical filters with different cutoff wavelength.

[0014] FIG. 5 shows an embodiment of a SWIR SERS apparatus using either a single pixel or array detector with high quantum efficiency in the shortwave infrared (SWIR) to conduct SERS on a biological analyte.

[0015] FIG. 6 shows an embodiment of a SWIR SERS apparatus using a nanophotonic waveguide made of silicon to enhance the Raman signal from analytes outside the waveguide through the evanescent field effect.

[0016] FIG. 7 shows a flow chart of an embodiment of a method of development of a classification process flow based on calibration with known analytes.

[0017] FIG. 8 shows a flow chart of an embodiment of a process for classification of analytes.DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The present disclosure provides such a method by integrating narrow-band SWIR photonic resonances with frequency-swept excitation and single-photon detection. This “source-sweep time-correlation” acquisition approach yields high-resolution Raman spectra and dynamic information on Raman spectral fluctuations, ubiquitous reporters of relevant sample dynamics, without the need for dispersive spectrometers or metal-based enhancement,and with a temporal dynamic range, practically only limited by the signal strength and not the detector temporal response function.

[0019] In contrast to prior cavity-enhanced Raman implementations, which rely on closed resonators and broadband detection, the disclosed open-cavity SWIR metasurface method enables large-mode-volume enhancement, negligible heating, and near-fluorescence-free operation, providing a platform for single-molecule and mesoscale Raman analysis.

[0020] Unlike Fabry-Perot cavities, chip-based nanophotonic resonators do not need active stabilization of the resonance frequency, thus enabling the high-resolution source-sweep Raman spectroscopy and temporal photon-correlation analysis of spectral fluctuations free of artifacts from cavity resonance noise.

[0021] The first object of the disclosure is to use open nanocavities - here used interchangeably with “dielectric metasurfaces”- with high fidelity (spectrally narrow photonic resonances) in the shortwave-infrared in combination with a high-coherence tunable continuous wave (cw) laser.

[0022] Dielectric metasurfaces hosting e.g., bound state in the continuum (quasi-BICs), or Fano resonances can achieve high Q-factor if fabricated in the transparency window i.e., for wavelengths below the bandgap of the dielectric e.g., silicon, gallium nitride, silicon nitride, indium tin oxide, indium phosphide which show low losses from materials absorption in the SWIR.

[0023] Spectrally-stable SWIR nanophotonic resonances can then be addressed by frequency -tuning the cw laser to resonantly enhance the laser itself as excitation source with the advantage of high excitation-field enhancement around the analyte improving the signal strength.

[0024] Spectrally stable SWIR resonances can also be made overlap with Raman scattered photons (Stokes or anti-Stokes bands) by tuning the cw laser off-resonance, enhancing the Raman signal via cavity-stimulated Raman spectroscopy. Unlike method in the near infrared, employing a cw laser in the SWIR beyond the absorption edge of the dielectric substrate eliminates heat dissipation, thereby avoiding detrimental sample heating and permitting higher excitation power (>lmW and up to 100 mW) to be employed for high Raman signals.

[0025] Background fluorescence is dramatically reduced, including from organic dyes, fluorescent proteins, or otherwise emissive motifs in the visible spectral range, as recently shown in non-enhanced tissue samples (See https: / / doi.org / 10.1101 / 2024.06.10.597863).

[0026] Spectral acquisition can be achieved via SWIR array detectors after spectrally dispersive elements e.g., spectrometers or refractive elements. Recording the Raman intensity as a function of laser sweeping at the wavelength of the photonic resonance eliminates the need for SWIR array detectors. The high-Q, spectrally-narrow resonances in the SWIR are enabling to quantify spectral lineshapes in this manner as broader photonic resonances would convolve the Raman spectrum with the field-enhancement spectral density, inducing systematic measurement errors. In some embodiments, the method further quantifies both the spectral amplitude and the characteristic timescales of Raman intensity fluctuations, which serve as sensitive indicators of the analyte’s internal dynamics. These fluctuations arise when the vibrational frequencies of the analyte stochastically shift relative to the narrow resonance of the nanophotonic cavity. By continuously recording the Raman signal as the excitation frequency is swept or held fixed near resonance, temporal correlations in the photon count rate can be computed — typically through an intensity autocorrelation function g(2)(r), where r is the lag time between scattered photons. This analysis reveals molecular or particulate processes such as conformational changes, binding and unbinding, or environmental reorganization. The accessible dynamic range extends from nanoseconds, limited by detector timing jitter, to seconds, limited by overall acquisition time — surpassing the capabilities of conventional SERS or cavity-enhanced Raman approaches. The open-cavity SWIR architecture maintains compatibility with diverse samples while preserving the advantages of low autofluorescence and high spectral resolution enabled by the narrow photonic resonances of the metasurface.

[0027] The embodiments herein involve a method of performing SERS (surface-enhanced Raman spectroscopy) that simultaneously minimizes spurious background emission, minimizes local heating even under high excitation powers, maximizes the Raman signal enhancement of dielectric SERS substrates, that can eliminate the need for expensive array detectors, and measure multi-timescale Raman spectral fluctuations without the need for high spectral acquisition framerates or interferometry.

[0028] Together these advantages render the method a powerful contender for sought after quantitative SERS and reliable analyte and single-molecule detection without substrate- induced fluctuations or other perturbations from SERS substrates. The ability to acquire bothstatic and dynamic information from spectral correlation analysis over many orders of magnitude in time enables commercially relevant usage, particularly in biosciences and diagnostics, exosome classification, other particle classification, DNA / RNA sequencing, protein sequencing, determination of biomolecular binding constants and kinetics, allosteric transduction, enzyme kinetics, photo- synthetic efficiencies, interconversion kinetics between biomolecular conformers, post-translational modifications, determination of molecular folding statuses in natural and synthetic proteins, and classification of different proteoforms. It further has commercial potential in environmental monitoring, food safety, semiconductor inspection, polymer quality control and research, quality control in pharmaceuticals - including vesicles for drug delivery-, materials science, and physical science research.

[0029] The embodiments include a method and apparatus for performing SERS in the shortwave infrared (SWIR) spectral region. Definitions of SWIR may vary slightly as to the range of wavelengths considered as SWIR. SWIR as used here means an optical spectral region between 900 - 2500 nanometers, inclusive. The advantages of performing SERS in the SWIR are multiplicative and include (1) dramatically reduced autofluorescence of common plasmonic and dielectric SERS substrates in the SWIR, including but not limited to silicon nanophotonic SERS substrates, (2) dramatic reduction of optical loss in typical dielectric substrates -in particular silicon-, and (3) dramatic reduction in the autofluorescence of typical organic analytes, discussed below with regard to FIG. 3, other secondary sample constituents, the Raman enhancement architecture (metals and / or dielectrics) or even part of the measurement system employed to conduct SERS.

[0030] SWIR detectors, however, are expensive compared to visible and NIR detectors based on silicon. InGaAs array detectors, for example, are often prohibitively expensive for many applications, possess higher dark counts and lower frame rates compared to silicon detectors, and require cooling, thereby limiting their applicability in Raman applications.

[0031] The dramatically reduced autofluorescence of common dielectric SERS substrates allows for improvements in the SERS signal-to-noise ratio via pump laser power increases. Compared to the visible of NIR, this is enabled by lower optical losses in dielectric SERS substrates in the SWIR. The resulting nanophotonic substrates with high quality factor (Q factor) result in lower losses and increased field-enhancement.

[0032] The method and the apparatus of the embodiments perform SERS spectroscopy as shown in FIG. 1. A tunable near-IR (NIR) or SWIR laser 10 irradiates an analyte on ananophotonic SERS element 12 to cause photon scattering, where the scattered photons have a different energy than the incident photos. The system may operate in one of several Raman spectroscopy modes, including resonant Raman excitation, coherent anti-Stokes Raman spectroscopy (CARS), stimulated Raman spectroscopy (SRS) besides the spontaneous Raman scattering considered herein.

[0033] A SWIR detector e.g., a superconducting nanowire single-photon detector (SNSPD) 14 detects the scattered photons. The detection path may also include optional spectrally dispersive elements, such as spectrometers, interferometers, optical filters, and prisms, as examples and without limitation. The detector may employ time-gating techniques, such as time-correlated photon counting. This assists in separating any residual background from the instantaneous Raman excitation, which would otherwise add noise.

[0034] After detection, the process may apply machine learning models, such as statistical or other deep learning models, to analyze or denoise the SWIR SERS data. This may include classifying the analyte under analysis. The analyte may include viruses, extracellular vesicles, materials, proteins, DNA, RNA, and may be analyzed for biomolecular temporal dynamics, as examples and without limitation. This may further include the classification of biological tissue samples as healthy or diseased, for which the reduced autofluorescence background will improve the classification accuracy from Raman. The machine learning may be implemented on a computer attached to either the tunable laser, the photodetectors, both, or a separate computing device not shown.

[0035] A SERS substrate generally comprises a surface engineered to have subwavelength photonic or plasmonic features that enhance the field at the frequency of the incoming excitation laser or at the frequency of the scattered Raman photons, or both. One particular type of SERS substrate comprises metasurfaces, which contain resonators periodically repeated across the surface. As used here the term “nanophotonic metasurface” or “nanophotonic element” means a metasurface on which the periodic photonic resonators have a scale of nanometers to hundreds of nanometers. Nanophotonic elements may have resonators in various patterns. Nanophotonic elements may be formed from dielectrics, such as silicon, silicon nitride, and gallium nitride or metals including gold, silver, and aluminum. Several nanophotonic architectures can be designed to create high-quality (high-Q) or high- fidelity resonances in the shortwave infrared, using different dielectrics such as silicon, silicon nitride, gallium nitride, and diamond, as examples. The nanophotonic elementarchitectures may include those such as 12 and 14 shown in the diagram, but other architectures and variations are possible.

[0036] In some embodiments, the nanophotonic element may incorporate nanopores for molecular sensing, diagnostics, and - in particular- sequencing. In one embodiment, the nanopores may be selected to feed biomolecular macromolecules into the high-field region of a nanophotonic or nanoplasmonic resonator on the element, including nanophotonic metasurfaces, with the intention of reading the Raman spectrum as the molecule traverses through the high-field region. This can be used for proteins, polypeptides, oligopeptides, RNA, DNA, polynucleotides, and oligonucleotides.

[0037] In some embodiments, photonically active nanowires or nanotubes may be employed with the intention of selectively enhancing the Raman signal from local regions inside of cells or tissue or other materials.

[0038] By red-shifting the resonance to the SWIR, beyond the bandgap of the dielectric e.g., silicon ( ~ 1100 nm), lower photonic losses lead to higher quality (higher fidelity) photonic resonances as indicated by narrower lines such as 18 in the transmission spectrum. In this spectral window of the photonic resonance, the local electric field of either the incoming excitation laser or the outgoing Raman-scattered photons are dramatically enhanced, even if the mode volume is large compared to lossy plasmonic substrates, or resonances in the absorptive window of the dielectric.

[0039] The high quality-factor resonance can be tuned to spectrally overlap with the frequency of the Stokes or Anti-Stokes scattered photons from the analyte, thereby enhancing the Raman scattering signal even if the pump laser is non-resonant with the photonic resonance. This is enabled by the absence of substrate heating for pumping beyond the bandgap of the dielectric. A relatively smaller enhancement of the Stokes or anti-Stokes compared to smaller mode-volume plasmonic SERS substrate can therefore be compensated by high excitation power without sample heating / damage, which is a common concern for plasmonic substrates.

[0040] FIG. 2 shows the method of acquiring static and dynamic Raman spectra in the SWIR. A tunable non-resonant laser is swept in frequency to tune different Raman peaks to spectrally overlap with the photonic resonance. The relative Raman intensity as a function of laser frequency tuning , collected at the photonic resonance frequency, scales with the overlap of the Raman spectrum and the spectrum of the field-enhancement, approximated bythe transmission spectrum. The spectrally narrow photonic resonances made possible by the low photonic losses in the metasurfaces in the SWIR endows this approach with a high spectral resolution, which is determined by the convolution of the spectral shape of the Raman spectrum SR(CO) and the power spectrum of the local photonic intensity enhancement sp(co); SR(CO)* SP(CO), where * is the convolution over frequency co. The narrow width of sp(co), from low-loss resonances in the SWIR relative to the Raman line (>5cm-1) ensures that the convolution approximates the Raman spectrum SR(CO).

[0041] Such open-cavity source sweep methods cannot meaningfully resolve narrow Raman (>5cm-l) lines with lossy resonances in the visible, e.g., plasmonic resonances or from lossy dielectric resonances, where the convolution will reduce the spectral resolution, complicating the discrimination of separate spectral features.

[0042] In some embodiments, the resonance can overlap with the pump laser used for Raman excitation, thereby enhancing the incoming field leading to stronger Raman scattering. Dualresonance metasurfaces may be used in the nanophotonic element to simultaneously enhance the incoming excitation laser and the either Stokes or Anti-Stokes scattered photons.

[0043] Higher quality resonances increase the Purcell factor enhancing the Raman scattering signal. Simultaneously, lower optical losses prevent heating of the nanophotonic substrate, therefore protecting the analyte from heat damage. The Raman signal may further be enhanced via the increase of the laser excitation power enabled by this reduced local heating in the SWIR. Simultaneously, the number of background photons created per unit time from photoluminescence from the nanophotonic substrate or analyte is reduced by orders of magnitude compared to the visible or near-infrared (<900 nm) spectral region due to (1) slower radiative decays for photoluminescence, and (2) faster non-radiative decay through vibrational relaxation (phonon emission) or internal conversion.

[0044] Efficient SWIR detectors with low background counts are new and not yet widely available. Examples of possible detectors for implementations include indium gallium arsenide (InGaAs) avalanche photodiodes, electron-multiplying charge-coupled (EMCCD) cameras, lead sulfur (PbS) quantum dot detectors, superconducting nanowire single-photon detectors (SNSPDs), and their linear and two-dimensional arrays, such as cameras. SNSPDs in particular are an emerging technology not widely employed in biochemical analytics but possess outsized advantages of low photon dark counts (<1 count per second) and high quantum efficiency of detection in the SWIR (>80%).

[0045] FIG. 2 further shows the method of measuring multi-timescale Raman spectral correlations, which are indicative of spontaneous dynamics in analytes e.g., protein conformational switching, binding and un-binding between disparate molecules, etc. At a given laser tuning , single-photon counting detectors are employed to time-tag the Raman signal at the photonic resonance wavelength. Spectrally narrow bandpass filters may be installed before the detector to isolate photons at this wavelength. The intensity autocorrelation g(2)(i) of the signal is then computed, either with one detector, or -after optical signal splicing- with two or more detectors to avoid afterpulsing effects. The Raman-active vibrational modes of the analyte undergo spontaneous frequency fluctuations, e.g., due to conformational switching, intermolecular binding and un-binding, local changes in pH, ionicity, strain, charging, and other processes. For a given , this stochastically changes the spectral overlap with the photonic resonance and therefore the total Raman intensity. The narrow photonic resonance therefore serves to transcribe frequency fluctuations 5co(t) in the Raman spectrum- s(co+ 6co(t)) -into intensity fluctuations I(t), equivalent to the number of detected photons per chosen time interval t+At. Computing the intensity auto-correlation g(2)(r)=<I(t)I(t+ T)> / <I(t)><I(t+ T)> then encodes the frequency fluctuation spectral magnitude, and characteristic fluctuation time-constants. The advantage of g(2)(r) is its wide temporal dynamic range with T practically only limited by photon shot-noise, which can be suppressed by longer data acquisition times.

[0046] For ergodic Raman frequency-fluctuations of the analyte i.e., time-invariant spectral fluctuation statistics, conducting g(2)(i) correlation analysis at different , provides a two- dimensional function g(2)( ,T), which measures the frequency range and characteristic times of the fluctuating Raman spectrum s(co+ 5co(t)). g(2)( ,T) can therefore be used to extract valuable information about the molecular or materials dynamics of the analyte, including the characteristic timescales of protein conformational switching, intermolecular binding and unbinding, etc. g(2)( ,T) can also be used as an additional classifier in Raman-based diagnostic tests, in which the specific temporal dynamics of analytes adds diagnostic confidence by expanding the dimensionality of classification from static ID Raman spectra (state-of-the-art) to 2D spectral correlations.

[0047] FIG. 3 shows a schematic representation of various intensities relative to the percentage of quantum efficiency (QE) of superconducting nanowire single-photon detectors (SNSPD). Curve 20 shows the QE of the SNSPD detectors as a percentage. Curve 22 is the Raman intensity. Curve 24 is the photoluminescence intensity. Curve 26 is the second-ordercorrelation function ^2)curve (the intensity time-correlation function), which is a measure of the statistical relationship between photons emitted from a light source. The subscript PL, PL means photoluminescence, indicates the photon-correlation function from spurious photoluminescence from the sample substrate and analyte molecules combined. Due to its inherent nonlinearity, this correlation function decays quickly with increasing wavelength in the SWIR, where the photoluminescence background is suppressed by orders of magnitude. The disclosed SWIR SERS is therefore particularly well-suited in photon-correlation applications (including the time-correlation analysis via g(2)( Q ,T) explained above), where spurious photoluminescence background photons would otherwise quickly degrade the informative Raman time-correlations. This is a unique feature in the SWIR due to the rapid decay of photoluminescence efficiency of most materials and molecules >900nm.

[0048] One should note that diagram shows the line demarking near-infrared (NIR) from SWIR at 1000 nanometers. As stated above, the embodiments here set that demarcation at 900 nanometers.

[0049] FIG. 4 shows the reduction in the autofluorescence background intensity in units of photons per second for a silicon SERS substrate coated with Rhodamine 6G under excitation with 580 nm light. Different dichroic long pass filters were employed to successively remove background with decreasing spectral bandwidth of detection. The autofluorescence background of these molecules reduces by orders of magnitude into the shortwave infrared (SWIR) > 900 nm, indicating the reduction in the autofluorescence of organic molecules using SWIR detectors. Further reduction in the background photoluminescence is expected for silicon substrates from 1150 nm to 2500 nm, largely eliminating any background photons added to the Raman signal. This low auto-fluorescence enables faithful Raman spectrum acquisition via the source sweep and enables the measurement of the Raman g(2)( Q ,T) without loss of correlation from uncorrelated photoluminescence photons.

[0050] One should note that Rhodamine 6G is merely a model system, common in Raman and resonance Raman studies and that similar reduction in the background fluorescence intensity with increasing wavelength is expected for most alternative analytes, including biological molecules, tissue samples, and materials.

[0051] FIG. 5 shows an embodiment of a SWIR SERS apparatus using either a single pixel or array detector with high quantum efficiency in the shortwave infrared (SWIR) to conduct SERS on a biological analyte on nanophotonic element 46. The laser light travels a first path42 indicated in red by reflecting of a dichroic mirror through the focusing lens 44 to the nanophotonic element 46. Nanophotonic element 46 contains the analyte. After the laser excites the photons in the analyte, those photons travel the path 48 shown as gray. The photons may be scattered by the diffraction grating 50 to then be detected by a multi-pixel SWIR detector 52. In an embodiment, the SWIR detector 52 comprises an SNSPD array. The SNSPD array can be replaced by an InGaAs EMCCD camera.

[0052] In one embodiment, a microfluidic structure 54 could be packaged with the SWIR detector 52. The microfluidic structure 54 would contain one or more analytes 56. This allows the SWIR detector and the microfluidic structure to be packaged as a lab on a chip.

[0053] FIG. 6 shows an embodiment of a SWIR SERS apparatus using a nanophotonic element comprising a waveguide made of silicon to enhance the Raman signal from analytes outside the waveguide through the evanescent field effect. This is similar to the SWIR implementation using metasurfaces because it reduces losses and increases the quality factor and Raman enhancement. A tunable laser 40 produces the excitation light that travels through fiber 62 to a waveguide 64, that resides on the nanophotonic element 66 on which the analyte is surface bound or otherwise located close to the element’s surface. As the laser light excites the analyte, the Raman photons couple into and travel through fiber 68 to the SWIR detectors 72, which may be a single pixel detector or an array detector. The photons may travel through the optical filter or filters 70.

[0054] Machine learning applied in the analysis requires calibration / training. FIG. 7 shows an embodiment of a calibration procedure. The process exposes the SERS nanophotonic element with a known analyte at 80. The classification accuracy is determined at 82. This accuracy is then compared to a threshold accuracy at 84. If the accuracy does not meet the threshold, the process then performs SWIR SERS spectroscopy as discussed above. The results of the spectroscopy are then used to train the machine learning computer model with the time- and frequency-dependent Raman signals and the analyte attributes.

[0055] After the model has been sufficiently trained to meet the accuracy threshold, the apparatus can then be used to classify analytes, as shown in FIG. 8. Using one of the apparatuses discussed above, or using a different apparatus, the process begins with coupling the light into the nanophotonic element that contains the analyte. The Raman photons are then collected in the SWIR at 92. In one path, one part of the Raman spectrum undergoesfiltering at 94 and the SWIR photons are detected over time at 96. In another path, the Raman photons are dispersed spectrally at 98 and then detected over wavelength at time at 100.

[0056] At 102, the detected photons over time, and / or the detected photons over wavelength and time, are checked to see if they match the calibration signature in at least one of the timedependent spectrum, the intensity, and the photon correlation function at 102. If the calibration signature is matched, the attributes of the analyte are classified at 106. If the calibration signature does not match, the determination is made that no classification is possible at 104.

[0057] In one embodiment, the machine learning classification uses either the laser tuningdependent intensity correlation function g(2)( ,T) or the full time-tagged photon arrival time data, prior to time-correlation, and collected at different off-resonant laser tuning as training and classification data.

[0058] The SWIR SERS method allows for single molecule observation with improved signal-to-background ratio. This method minimizes spurious background emissions, minimizes local heating even under high excitation powers, and maximizes the enhancement of the nanophotonic dielectric nanophotonic elements. Possible applications are found in all areas of application of SERS in diagnostics, tissue characterization, assays, and biomolecular analytics. In particular, the high signal-to-background of SWIR SERS is advantageous in biomolecular sequencing e.g., of proteins, polypeptides, or DNA / RNA. Moreover, the high signal-to-background of SWIR SERS is advantageous in single-molecule binding kinetics and affinity measurements by optical means, for example by measuring the Raman spectral shift or fluctuations over time.

[0059] One embodiment comprises integrating one or more microfluidic structures positioned such that analytes in the microfluidic structure can be irradiated by the laser. This allows the SWIR detector and the microfluidic structures to be packaged as a lab-on-a-chip.

[0060] An embodiment comprises a method of surface-enhanced Raman spectroscopy including positioning an analyte in the field-enhancing region of a nanophotonic element; irradiating the analyte with a continuous wave excitation laser operating at a user-defined frequency in a range from the near infrared (NIR) to short-wave infrared (SWIR) to cause photon scattering from the analyte; and detecting the photons with one or more photodetectors in the SWIR. An embodiment comprises the method above wherein irradiating the analyte comprises irradiating the analyte with a frequency tunable excitationlaser at a resonance frequency of the nanophotonic structure. An embodiment comprises the method above, wherein irradiating the analyte comprises irradiating at a frequency so that Raman scattered photons from the analyte of a pre-selected frequency are resonant with the nanophotonic structure.

[0061] An embodiment comprises the above, wherein the method comprises: tuning the frequency of the excitation laser into resonance with the nanophotonic element at different Raman frequencies of the analyte; and obtaining a Raman spectrum of the analyte by measuring the intensity of the SWIR signal at the nanophotonic resonance frequency with SWIR detectors.

[0062] An embodiment comprises the method above, wherein the method further comprises time-tagging the Raman scattered SWIR photons through single-photon counting; and characterizing the analyte by performing temporal photon-correlation analysis of the Raman intensity fluctuations.

[0063] An embodiment comprises the method above, wherein the method further comprises obtaining characteristic time constants of binding of the analyte with secondary molecules in an aqueous solution around the analyte. An embodiment comprises the method above wherein the method further comprises obtaining the sequence of amino acids in proteins, DNA, or RNA, poly- and oligopeptides, and poly- and oligonucleotides, and a chemical nature of any Post-Translational Mutations (PTMs).

[0064] An embodiment comprises the method above, wherein the method further comprises obtaining temporal dynamics of biologically-relevant entities comprising one or more of as intermolecular binding kinetics involving proteins, Enzyme-Substrate Binding, Antibody- Antigen Binding, Receptor-Ligand Binding, Protein-Protein Interactions (PPIs), DNA- Protein Binding, RNA-Protein Binding, Hormone-Receptor Binding, Ion Channel-Ligand Binding, G-Protein Coupled Receptor (GPCR) Binding, Cofactor Binding, Transporter- Substrate Binding, Small Molecule-Protein Binding, Viral Protein-Host Protein Binding, Lipid-Protein Binding, Chaperone-Protein Binding, Cell-Cell Adhesion Molecule Binding, Cytokine-Receptor Binding, Metabolite-Protein Binding, Toxin-Target Binding, Conformational Switching.

[0065] An embodiment comprises the method above wherein irradiating the analyte comprises pulsing the laser, and detecting the photons comprises using a time-gating technique to separate residual background from instantaneous Raman excitation. An embodiment comprises the method above, wherein the time-gating technique comprises atleast one of time-and correlated photon counting. An embodiment comprises the method above, wherein the SWIR photons are used to classify the analyte based on previously collected SWIR Raman responses.

[0066] An embodiment comprises the method above, wherein, further comprising applying machine learning to one of either denoise or analyze the SWIR SERS data. An embodiment comprises the method above, wherein applying machine learning comprises applying one or more of statistical deep learning models, or deep learning models. An embodiment comprises the method above, wherein applying machine learning to the SWIR SERS data comprises applying the deep learning to classify one or more of viruses, extracellular vesicles, materials, proteins, or DNA and RNA molecules.

[0067] An embodiment comprises a surface-enhanced Raman spectroscopy system, including a laser operating at a frequency having a range from near-infrared (NIR) to short-wave infrared (SWIR) that emits light in a spectrum between 700 and 2500 nanometers, a nanophotonic surface to receive the light from the laser, and one or more SWIR detector to detect photons scattered by an analyte on the nanophotonic metasurface.

[0068] An embodiment comprises the system above, further comprising one or more spectrally dispersive elements before the SWIR detector. An embodiment comprises the system above, wherein the one or more spectrally dispersive elements comprises one or more of prisms, optical filters, interferometers, and spectrometers.

[0069] An embodiment comprises the system above, further comprising a microfluidic device integrated into the system and positioned on the nanophotonic surface to be irradiated by the laser.

[0070] An embodiment comprises the system above, wherein the nanophotonic element includes nanopores for the selective Raman excitation of single molecules or parts of single molecules, including proteins, DNA, or small sub-ensembles with high signal-to-background ratio. An embodiment comprises the system above, wherein the nanophotonic element comprises one of dielectric metasurfaces, dielectric pillar dimers, plasmonic metal metasurfaces, plasmonic metal gap cavities, metal nanopores, dielectric photonic crystal cavities, and plasmonic metal bowties.

[0071] An embodiment comprises the system above, wherein the nanophotonic surface comprises a waveguide, the waveguide enhancing the Raman signal of an analyte positioned close to the waveguide’s surface, and the system further includes optical fibers to collect and guide SWIR Raman photons.

[0072] An embodiment comprises the system above, wherein the SWIR detector comprises one of a single-pixel InGaAs photodiode, a single-pixel InGaAs avalanche photodiode (APD), a single-pixel superconducting nanowire single-photon detector (SNSPD), a onedimensional APD array, a one-dimensional SNSPD array, a two-dimensional APD array, a two-dimensional SNSPD array, or an InGaAs electron multiplied charge coupled device (EMCCD). An embodiment comprises the system above, wherein the SWIR detector comprises either one of the one-dimensional arrays or the two-dimensional arrays used in conjunction with a spectrometer to resolve the Raman spectrum in the SWIR.

[0073] An embodiment comprises the system above, further comprising an optical path length interferometer configured to transform Raman spectral fluctuations into intensity fluctuations, and reconstruct an auto-correlation of the Raman spectrum using intensity correlation analysis after SWIR photon detection.

[0074] All features disclosed in the specification, including the claims, abstract, and drawings, and all the steps in any method or process disclosed, may be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. Each feature disclosed in the specification, including the claims, abstract, and drawings, can be replaced by alternative features serving the same, equivalent, or similar purpose, unless expressly stated otherwise.

[0075] Additionally, this written description makes reference to particular features. It is to be understood that the disclosure includes all possible combinations of those particular features. For example, where a particular feature is disclosed in the context of a particular aspect, that feature can also be used, to the extent possible, in the context of other aspects. Also, when reference is made in this application to a method having two or more defined steps or operations, the defined steps or operations can be carried out in any order or simultaneously, except when excluded by the context. Also, when reference is made in this application to a method having two or more defined steps or operations, the defined steps or operations can be carried out in any order or simultaneously, unless the context excludes those possibilities.

[0076] Although specific aspects of this disclosure have been illustrated and described for purposes of illustration, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. Accordingly, the invention should not be limited except as by the appended claims.

Claims

WHAT IS CLAIMED IS:

1. A method of surface-enhanced Raman spectroscopy comprises: positioning an analyte in a field-enhancing region of a nanophotonic element; irradiating the analyte with a continuous wave excitation laser operating at a user- defined frequency in a range from near infrared (NIR) to short-wave infrared (SWIR) to generate a signal to cause photon scattering from the analyte; and detecting photons scattered with one or more photodetectors in the SWIR.

2. The method as claimed in claim 1, wherein irradiating the analyte comprises irradiating the analyte with a frequency tunable excitation laser at a resonance frequency of the nanophotonic element.

3. The method as claimed in claim 1, wherein irradiating the analyte comprises irradiating the analyte with a frequency tunable excitation laser at a frequency so that Raman scattered photons from the analyte of a preselected frequency are resonant with the nanophotonic element.

4. The method as claimed in claim 1, further comprising: tuning the frequency of the excitation laser with the nanophotonic element in resonance with different Raman frequencies of the analyte; and obtaining a Raman spectrum of the analyte by measuring intensity of the signal at a nanophotonic resonance frequency with SWIR detectors.

5. The method as claimed in claim 1, further comprising: time-tagging Raman scattered SWIR photons through single-photon counting; and characterizing the analyte by performing temporal photon-correlation analysis of Raman intensity fluctuations at one or more excitation laser frequencies.

6. The method as claimed in claim 5, further comprising obtaining characteristic time constants of binding of the analyte with secondary molecules in an aqueous solution around the analyte.

7. The method as claimed in claim 5, further comprising obtaining a sequence of amino acids in proteins, DNA, or RNA, poly- and oligopeptides, and poly- and oligonucleotides, and a chemical nature of any Post-Translational Mutations (PTMs).

8. The method as claimed in claim 5, further comprising obtaining temporal dynamics of biologically-relevant entities comprising one or more of as intermolecular binding kinetics involving proteins, Enzyme- Substrate Binding, Antibody-Antigen Binding, Receptor-Ligand Binding, Protein-Protein Interactions (PPIs), DNA-Protein Binding, RNA-Protein Binding, Hormone-Receptor Binding, Ion Channel-Ligand Binding, G-Protein Coupled Receptor (GPCR) Binding, Cofactor Binding, Transporter- Substrate Binding, Small Molecule-Protein Binding, Viral Protein-Host Protein Binding, Lipid-Protein Binding, Chaperone-Protein Binding, Cell-Cell Adhesion Molecule Binding, Cytokine-Receptor Binding, Metabolite- Protein Binding, Toxin-Target Binding, Conformational Switching.

9. The method as claimed in claim 1, wherein irradiating the analyte comprises pulsing the excitation laser, and detecting the photons comprises using a time-gating technique to separate residual background from instantaneous Raman excitation.

10. The method as claimed in claim 9 wherein the time-gating technique comprises time- correlated photon counting.

11. The method claimed as claimed in claim 1, further comprising using SWIR photons to classify the analyte based on previously collected SWIR Raman responses.

12. The method as claimed in claim 1, further comprising applying machine learning to one of either denoise or analyze SWIR SERS data.

13. The method as claimed in claim 12, wherein applying machine learning comprises applying one or more of statistical deep learning models, or deep learning models.

14. The method as claimed in claim 12, wherein applying machine learning to the SWIR SERS data comprises applying deep learning to classify one or more of viruses, extracellular vesicles, materials, proteins, or DNA and RNA molecules.

15. A surface-enhanced Raman spectroscopy system, comprising: a laser operating at a frequency having a range from near-infrared (NIR) to shortwave infrared (SWIR) that emits light in a spectrum between 700 and 2500 nanometers; a nanophotonic element to receive light from the laser; and one or more SWIR detector to detect photons scattered by an analyte on the nanophotonic element.

16. The system as claimed in claim 15, further comprising one or more spectrally dispersive elements before the SWIR detector.

17. The system as claimed in claim 16, wherein the one or more spectrally dispersive elements comprise one or more of prisms, optical filters, interferometers, and spectrometers.

18. The system as claimed in claim 15, further comprising a microfluidic device integrated into the system and positioned on the nanophotonic element to be irradiated by the laser.

19. The system as claimed in claim 15, wherein the nanophotonic element includes nanopores for selective Raman excitation of single molecules or parts of single molecules, including proteins, DNA, or RNA or small sub-ensembles with high signal-to-background ratio.

20. The system as claimed in claim 15, wherein the nanophotonic element comprises one of dielectric metasurfaces, dielectric pillar dimers, plasmonic metal metasurfaces, plasmonic metal gap cavities, metal nanopores, dielectric photonic crystal cavities, plasmonic metal bowties.

21. The system as claimed in claim 15, wherein the nanophotonic element comprises a waveguide to enhance a Raman signal of an analyte positioned close to a surface of the waveguide; and the system further comprises optical fibers to collect and guide SWIR Raman photons.

22. The system as claimed in claim 15, wherein the SWIR detector comprises one of a single-pixel InGaAs photodiode, a single-pixel InGaAs avalanche photodiode (APD), a single-pixel superconducting nanowire single-photon detector (SNSPD), a one-dimensional APD array, a one-dimensional SNSPD array, a two-dimensional APD array, a two- dimensional SNSPD array, or an InGaAs electron multiplied charge coupled device (EMCCD).

23. The system as claimed in claim 22, wherein the SWIR detector comprises either one of the one-dimensional arrays or the two-dimensional arrays used in conjunction with a spectrometer to resolve Raman spectrum in the SWIR.

24. The system as claimed in claim 22, further comprising an optical path length interferometer configured to transform Raman spectral fluctuations into intensity fluctuations and reconstruct an auto-correlation of Raman spectrum using intensity correlation analysis after SWIR photon detection.

Images