Simulated Raman correlation spectroscopy for quantifying nucleic acid-silver composites

Plasmonic devices are of great interest due to their ability to confine light to the nanoscale level and dramatically increase the intensity of the electromagnetic field, functioning as high performance platforms for Raman signal enhancement. While Raman spectroscopy has been proposed as a tool to identify the preferential binding sites and adsorption configurations of molecules to nanoparticles, the results have been limited by the assumption that a single binding site is responsible for molecular adsorption. Here, we develop the simulated Raman correlation spectroscopy (SRCS) process to determine which binding sites of a molecule preferentially bind to a plasmonic material and in what capacity. We apply the method to the case of nucleic acids binding to silver, discovering that multiple atoms are responsible for adsorption kinetics. This method can be applied to future systems, such as to study the molecular orientation of adsorbates to films or protein conformation upon adsorption.


S1.2. Geometrical optimization and vibrational frequency calculations of cytosine
Upon geometrical optimization, cytosine has 4 potential binding sites to the 20 atom silver tetrahedral structure: N1 to surface, N3 to edge, NH 2 to surface and O to surface as shown in intensity is normalized with respect to the total intensity of the spectrum.

S1.3. Geometrical optimization and vibrational frequency calculations of guanine
Guanine has 5 potential binding sites as shown in Figure S3: ( intensity is normalized with respect to the total intensity of the spectrum.

S1.4. Geometrical optimization and vibrational frequency calculations of thymine
Based on the most common thymine tautomers that exist in water, thymine has 4 potential binding sites as shown in Figure S4: N1 and O2 to edge, N3 and O2 to edge, O2 to edge, and O4 to edge intensity is normalized with respect to the total intensity of the spectrum.

S2. Surface image and characterization of silver films
The SERS substrates used for experimental measurements are random silver films (RSFs) in which an electron beam evaporation system is used to deposit 30 nm of silver onto a silicon substrate. At this thickness, the silver nanoparticles merge together to become irregular shaped islands ( Figure S5). The resulting SERS substrates generate consistent electromagnetic field enhancement across a broad excitation wavelength range. The silver films have a face centered cubic (111) surface which is represented by the silver 20 atom tetrahedral structure surface side.

Supplementary Figure 5. Scanning electron microscope image of random silver films
Random silver films have a broad extinction profile in the visible range which has a weak surface plasmon resonance in the near infrared range. Thus, the electromagnetic effect has a minimal effect on the measured Raman spectra. The extinction profile for silver films is shown in Figure S6, in which two excitation wavelengths (514nm and 785nm) are displayed with their corresponding Raman frequency modes.

Supplementary Figure 6. Extinction spectrum of silver films showing two excitation wavelengths and the corresponding Raman shifts
While an excitation wavelength of 514nm shows a slight increase in the 1335cm -1 mode compared to the 720cm -1 mode, the frequency modes excited by the 785nm wavelength show consistent enhancement based on the extinction profile.
To further demonstrate the consistency of the electromagnetic field enhancement across the substrate, Raman maps were acquired of 1 mM of adenine dissolved on random silver films at excitation wavelengths of 514nm and 785nm. An area of approximately 50 microns by 50 microns was analyzed via a raster scan with spot sizes of approximately 10 by 10 microns. For analysis, the peak intensity ratio of the single stretching mode (~1335cm -1 ) to the ring breathing mode (~720cm -1 ) was calculated at each spot and a colored Raman map based on the intensity ratio was plotted. Figure S7 shows the resulting Raman maps for 514nm (blue) and 785nm (red). The Raman maps show consistent peak intensity ratio at each spot along the map for both wavelengths, with an average 1335cm -1 to 720cm -1 ratio of 0.29 (σ = 0.020) for 514nm and 0.23 (σ = 0.019) for 785nm. Thus, the electromagnetic field enhancement effect is fairly consistent across the substrate.

S3. Mode assignments for SRCS process
Before performing the SRCS process, the calculated vibrational frequency modes are aligned with experimental measurements and are normalized with respect to the total Raman intensity of the system. The computed vibrational frequencies are slightly off-set from each other because the frequency modes for each system vary slightly in location dependent on the orientation of the molecule with respect to the surface, as the molecular strain changes the way in which the system vibrates and leads to small shifts in the vibrational frequencies.
As an example, the adenine silver systems (A-N1, A-N3, A-N7, A-N9, A-NH 2 ,N7) have different calculated frequencies for the ring-breathing-mode (RBM) that are determined by the molecular vibrations.
GaussView is used to visualize the modes, display the force displacements and appropriately assign each frequency value with the corresponding mode. The frequencies for A-N1, A-N3, A-N7, A-N9, and A-NH 2 ,N7 have RBM mode locations of 711cm -1 , 698cm -1 , 714cm -1 , 708cm -1 and 716cm -1 , respectively. The minor variation between the vibrational frequencies for the RBM is caused by the molecular strain of the system which modulates the bond force constants of the structure. Thus, despite the difference in frequency values, the frequencies from 698-716cm -1 are assigned to the RBM by visualizing the force displacements of the modes. The experimental measurements have the RBM slightly red-shifted compared to the simulated measurements due to the scaling factor used in the calculations, with the experimental RBM band ranging from 715cm -1 to 743cm -1 . The strongest 10 to 12 frequency modes for each nucleic acid, such as the RBM, are selected for analysis using GaussView to visualize the force displacement vectors. The list of modes can be found in supplementary section 1.

S4. Coefficient of Determination Calculations for Single Binding Sites
To compare the coefficient of determinations for single binding sites to the SRCS optimized results, the coefficient of determination for each single atom binding site (e.g. A-N1 compared to experimental measurement) is calculated. As described in the main text, the experimental Raman spectrum mode for a nucleic acid NA and a mode of i is defined as and the simulated Raman spectrum mode for a nucleic acid NA, a binding site of b, and a mode of i, is defined as . Thus, the coefficient of determination ( ) is defined as: Here, the coefficient of determination represents the correlation of the simulated Raman spectrum of a single binding site to the experimental Raman spectrum, with a value constituting perfect correlation. The residuals for each mode are shown in Figure S7 Despite the above average correlation for A-N3, C-N3, and G-N3, there are still some simulated modes in these systems that show significant deviation from the experimental values. For example, the C-N stretching mode for A-N3 is greatly enhanced in experimental measurements, but significantly reduced in the simulation results. The C-NH 2 stretching mode for C-N3 is strong in simulations, but very weak in the experimental spectra. Additionally, the C-NH 2 bending mode is absent in the C-N3 simulations. These discrepancies reveal that there is more than one binding site responsible for the Raman signatures in experimental measurements and that the experimental measurements are superimposed spectra of the possible binding sites. To improve the coefficients of determination, the SRCS process is performed in which the coefficient of determination is maximized by finding the optimal weighted coefficient constants for each nucleic acid (main text).