• Home   /  
  • Archive by category "1"

Laser Infrared Sample Analysis Essays

s-SNOM and nano-FTIR set-up

The s-SNOM and nano-FTIR set-up (Neaspec GmbH, Germany) used in this work is based on an AFM where the tip is oscillating vertically at the mechanical resonance frequency Ω of the cantilever. Infrared near-field imaging and spectroscopy are performed by interferometric detection of the light scattered from a gold-coated AFM tip, which is illuminated by the radiation from a tunable quantum cascade laser (QCL) or a broadband mid-infrared laser beam, respectively (Fig. 1a,b).

For s-SNOM imaging, the tip is illuminated with radiation from the wavelength-tunable QCL at individual wavelengths. The tip-scattered light is detected by operating the interferometer in pseudoheterodyne mode34. Demodulating the detector signal at a higher harmonic nΩ of the tip oscillation frequency yields infrared amplitude and phase images, sn and ϕn (see Methods section).

In nano-FTIR spectroscopy, broadband infrared amplitude and phase spectra are obtained by employing the interferometer for Fourier transform spectroscopy of the scattered light14,15,16,17,18,19. In this case, the tip is illuminated with a broadband laser of a total power of about 100 μW (refs 17, 35). As the tip and the sample are located in one of the interferometer arms (in contrast to standard FTIR), we can measure both local amplitude and phase spectra, sn(ω) and ϕn(ω). These spectra are normalized to the references, sref,n(ω) and ϕref,n(ω), obtained on a clean area on the sample support (marked R in the topography images presented in the work). We thus obtain normalized amplitude and phase spectra form which we can calculate the nano-FTIR absorption spectrum an(ω)=sn/sref,n sin (ϕnϕref,n). As demonstrated in refs 17, 19, the nano-FTIR spectrum an(ω) reveals the local infrared absorption in the sample.

Setting the foundations for protein analysis with nano-FTIR

We first verify the ability of nano-FTIR to measure nanoscale broadband infrared spectra of the amide I and II bands over the range from 1,400 to 1,800 cm−1. To that end, we studied TMVs (Fig. 2a). These well-defined protein complexes have a diameter of 18 nm and a length of 300 nm, and consist of 2,140 identical proteins assembled helically around an RNA strand31. Figure 2c shows near-field infrared phase ϕ3 images taken at 1,660 and 1,720 cm−1, that is, on and off resonance with the amide I vibration. At 1,660 cm−1, the phase image exhibits strong contrast owing to the amide I absorption25. As expected, the phase contrast vanishes when the illumination is tuned to 1,720 cm−1 where the protein does not absorb. From nano-FTIR amplitude s3(ω) and phase ϕ3(ω) spectra recorded on top of the virus (position marked by the red dot in the topography image) and normalized to those taken on a clean silicon area, we obtained the local infrared absorption spectrum a3 (red spectrum in Fig. 2d). For comparison, we recorded a FTIR spectrum of a large TMV ensemble (blue spectrum in Fig. 2d) by employing p-polarized-grazing incidence (GI-FTIR, schematics see Fig. 1c). We find a good agreement between the GI-FTIR and nano-FTIR spectra. Both reveal the amide I and amide II bands associated with combinations of C=O and C-N stretching with C-N-H-bending vibrations, respectively. We note that the agreement can vary among different amide bands, since the exact peak position and shape of an absorption band depend on the band strength and the employed FTIR technique36 (see also Supplementary Fig. S1).

In Fig. 2e–h, we perform nano-FTIR studies of ferritin, a globular protein complex of 12 nm in diameter. Ferritin comprises 24 subunits that form a cage around a ferrihydrite nanoparticle (Fig. 2e). Each ferritin subunit is composed of six α-helixes and one β-sheet30. The topographical image in Fig. 2g shows two particles of about 10 and 8 nm, and one of about 6 nm height, which appear much broader because of the convolution with tip apex that has a radius of about 35 nm (see Fig. 2f and Supplementary Fig. S2). The infrared phase image ϕ3 at 1,660 cm−1 reveals a strong absorption for the 10- and 8-nm particles, indicating that these are ferritin complexes. No infrared contrast is seen for the 6-nm particle, which we interpret as ferrihydrite core that does not absorb at 1,660 cm−1. Figure 2h shows the nano-FTIR spectrum (red curve) of the ferritin marked by the red dot in Fig. 2g in comparison with a GI-FTIR spectrum of a large ferritin ensemble (blue curve). As demonstrated with the TMV (vide supra), the nano-FTIR spectrum reveals the amide I and II bands in excellent agreement with GI-FTIR. Thus, we conclude that nano-FTIR allows for measuring infrared spectra of individual protein complexes, which can be interpreted by comparison with standard GI-FTIR absorbance spectra. The analysis provided in Supplementary Fig. S2 and Supplementary Note 1 indicates that the marked object is one ferritin complex.

To explore whether and how nano-FTIR absorption spectra depend on the protein orientation, we studied the PMs of Halobacterium salinarum (Fig. 3), which is composed of a double layer of polar and neutral lipids, and the integral membrane protein bacteriorhodopsin. The secondary structure of bacteriorhodopsin comprises seven transmembrane α-helices and an extracellular β-sheet33. As the helices are predominantly oriented perpendicular to the membrane plane, the infrared amide I vibration is normal to the membrane surface, whereas the amide II vibration is parallel to the membrane surface37 (see illustration in Fig. 3b). The well-defined directions of the amide vibrations render PM an excellent protein structure for studying the sensitivity of nano-FTIR to protein orientation. In the topography image (Fig. 3a), the membranes appear as 6 nm high flat layers. As before with the TMV, the infrared phase contrast (Fig. 3c) at 1,660 cm−1 exhibits the typical amide I absorption, which vanishes at 1,720 cm−1 because the archaeobacterial PM is devoid of any ester lipids that usually give rise to carbonyl stretching vibrations in lipid membranes of eubacteria and eukarya. A nano-FTIR spectrum a3 (Fig. 3d, red curve) is obtained by placing the tip atop the PM (marked by red dot in Fig. 3a), analogous to the TMV experiment. The GI-FTIR spectrum (p-polarized incident field) of horizontally adsorbed PMs was recorded for comparison (Fig. 3d, blue curve; for schematics see Fig. 1c). Interestingly, the amide II band does not appear in both spectra. In the GI-FTIR experiment, the electric field is perpendicular to the PM surface and thus perpendicular to the direction of the amide II vibrations. For that reason, amide II vibrations are not excited37. The absence of the amide II band in the nano-FTIR spectrum is explained by the field distribution at the tip apex, which essentially is that of a vertically oriented point dipole located in the tip apex (Fig. 3b)20. The strongest field, located directly below the tip apex, is vertically oriented, and efficiently couples the protein vibrations normal to the PM surface (amide I) but not the protein vibrations parallel to the PM surface (amide II). This orientation-dependent effect is well known from surface-enhanced IR absorption spectroscopy38. Thus, we conclude that nano-FTIR primarily probes molecular vibrations that oscillate perpendicular to the sample surface.

In Fig. 4, we evaluate the spatial resolution and reproducibility of nano-FTIR protein spectroscopy by recording 20 spectra, while the tip is scanning in steps of 10 nm across the PM. The spectral line scan (Fig. 4b) along the line of red and blue dots in the topography image (Fig. 4a) reveals the amide I peak on top of the membrane until the edge is reached (at position x≈80 nm). Plotting the nano-FTIR absorption a3 at 1,660 cm−1 as a function of the position x (Fig. 4c) shows that absorption vanishes within three scan steps. This demonstrates a spatial resolution of about 30 nm, which is an improvement by more than 100 compared with micro-FTIR mapping39. We further note that the eight local infrared spectra on the PM (P1–P8, between x=0 and 70 nm) exhibit a stable peak position and peak height, which illustrates the high reproducibility of nano-FTIR (Fig. 4d). On the other hand, these results show that the infrared absorption of PM does not exhibit significant spatial variations, confirming the homogeneity of the protein structure within the PM.

Nanoscale mapping of structural protein heterogeneity

To establish nanoscale identification of α-helical and β-sheet structure in proteins, and to demonstrate heterogeneity in a sample at the nanometre scale that is only resolvable by nano-FTIR, we studied insulin aggregates deliberately contaminated with a low amount of TMV (Fig. 5). While α-helices dominate in TMV, the insulin aggregates are composed of essentially two β-sheets32. The TMV can be thus considered a diluted α-helical structure within a β-sheet sample. Note that for sample preparation, we used a 2-year old insulin solution (see Methods section) because we found that such a solution yields insulin aggregates rather than well-defined fibrils. For such a sample, we expect that the orientation of the infrared-active β-sheet dipole of the amide I band is random, thus yielding a signal in nano-FTIR. Note that only vertically oriented dipoles yield a signal in nano-FTIR (see Fig. 3).

In Fig. 5a, we show attenuated total reflectance (ATR-)FTIR spectra of pure insulin aggregates (black curve) and of the insulin/TMV mixture (blue curve). Both pure insulin and insulin/TMV samples were deposited on a silicon support. The ATR-FTIR spectrum of the insulin/TMV mixture closely matches the ATR-FTIR spectrum of pure insulin and does not reveal the presence of TMV. The topographical image (Fig. 5b) of the mixture shows rod-like structures and aggregates of up to 32 nm in height. Still, TMV and insulin cannot be clearly discriminated. Figure 5c shows the near-field infrared phase images at 1,634 and 1,660 cm−1, corresponding to the centre frequencies of the amide I band for β-sheet and α-helical structure, respectively11. In both infrared images, all structures show a distinct absorption contrast. As expected, the phase contrast to the substrate (that is, absorption) increases with particle height (that is, volume). Comparing the two infrared images, we find a significant increase of the phase contrast from 1,634 to 1,660 cm−1 for the smooth rod in the lower left corner of the image, while for all other structures the phase contrast did not change with respect to the substrate. From this observation, we conclude that α-helices form this rod-like structure, which is thus most probably a TMV. For identification of the secondary structure, we recorded nano-FTIR spectra (Fig. 5d) on the smooth rod at the position marked by the red dot in the topography image (red spectrum), and on the aggregate marked by the green dot (green spectrum). The red spectrum exhibits the absorption maximum at 1,660 cm−1, which corresponds to the amide I resonance frequency of α-helices. Further, the shape fits well to the infrared spectrum of TMV (Fig. 2d). These results lead us to identify this rod-like structure as a TMV particle. In contrast, the green spectrum significantly differs from the red, exhibiting two peaks at 1,660 and 1,634 cm−1, which are attributed to the presence of α-helices and β-sheets, respectively. Thus, we can identify this particle as an insulin aggregate, as its nano-FTIR spectrum agrees with the ATR-FTIR spectrum of the pure insulin sample (Fig. 5a, black curve). Note that we compare the nano-FTIR spectrum with an ATR-FTIR spectrum, in contrast to Fig. 2, as the infrared signal was too weak to obtain a GI-FTIR spectrum. As ATR-FTIR spectra are typically red shifted relative to GI-FTIR spectra36 (see Supplementary Fig. S1), the ATR-FTIR spectrum is slightly red shifted compared with the nano-FTIR spectrum.

To visualize the nanoscale distribution of TMV and insulin aggregates, that is, the nanoscale structural heterogeneity of the sample, we calculated the ratio between the infrared phase images at 1,660 and 1,634 cm−1. From the spectra displayed in Fig. 5d, we know that the absorption of TMV is stronger at 1,660 than at 1,634 cm−1. For insulin, the absorption is stronger at 1,634 cm−1. Every image pixel where we obtain a ratio that is significantly larger than 1 thus reveals TMV. Pixel where we obtain a ratio that is smaller than 1 reveals insulin. Figure 5e shows a map where pixel with ratio larger than 1.5 are depicted in purple, and pixel with ratio smaller than 1 in yellow. The map clearly reveals the TMV (purple) and shows that all other protein structures (yellow) have a strong signal at 1,634 cm−1, which can be attributed to the presence of β-sheets. This allows us to identify them as insulin aggregates.

Nano-FTIR studies of individual insulin fibrils

Having demonstrated the capability of mapping secondary structure on the nanometre scale, nano-FTIR is well prepared for applications in biochemical and biomedical research. We explore as a first application example the protein conformation in individual insulin fibrils (Fig. 6). Insulin can form amyloid-like fibrils and fibres composed of filament-shaped protein aggregates11,32,40, which renders it an excellent and widely used model system for neurodegenerative disease research (that is, Alzheimer and Parkinson). Studies32,41 show that the filaments have a core composed of β-sheets (Fig. 6c). It is assumed that this core is surrounded by randomly oriented secondary structures, including α-helices, β-turns and unordered structures11. However, the exact structure of the shell is still an open question of high biological relevance40. Recent studies by tip-enhanced Raman spectroscopy indicate the presence of α-helices/unordered structures at the surface of the fibrils42,43. In the following, we apply nano-FTIR to study the protein conformation in insulin fibrils.

In contrast to Fig. 5, insulin fibrils were grown by incubating insulin protein at 60 °C in a pH 2 buffer for 30 h, thus representing a biologically relevant model system. In Fig. 6a, the topography of a sample area is shown where we found type I (3 nm high, consisting of two protofibrils, respectively four protofilaments, marked I) and thicker insulin fibrils (with increasing thickness the number of protofilaments increases) with heights ranging from 5 to 10 nm11,44,45. In the nano-FTIR spectrum of a 9-nm-thick fibril (Fig. 6b, red curve, taken at the position marked by the red dot in the topographical image), we find the strongest peak at 1,669 cm−1. By comparison with a GI-FTIR spectrum of monomeric insulin (predominantly α-helical structure, blue dashed curve), we can assign this peak to the presence of α-helices. The peak at 1,638 cm−1 is assigned to β-sheets. However, it is relatively weak compared with the nano-FTIR spectrum of the insulin aggregates (Fig. 5). We explain this observation by the well-aligned β-sheet structure forming the core of the fibrils seen in Fig. 6, where the dipole orientation of the amide I band is mostly parallel to the filament axes (that is, to the substrate surface)13,46. As seen before in Fig. 3 with PM, protein vibrations parallel to the substrate couple only weakly to the near field at the tip apex and thus are suppressed in the nano-FTIR spectra.

Standard band decomposition of the nano-FTIR spectrum (Fig. 6d) reveals two major bands at 1,639 and at 1,671 cm−1, confirming that β-sheet and α-helical structures are the predominant contributions. The band at 1,697 cm−1 could be caused by β-turns or even antiparallel β-sheet structure and the weak band at 1,609 cm−1 might indicate side chains. Most importantly, no band is observed in between the β- and α-peaks where typically disordered structure is located10. Thus, we conclude that disordered structures are almost absent in the fibrils. Note that the dipole of the amide I band is isotropically oriented in disordered structures and thus cannot be the cause for this observation.

To explore the presence of α-helices in type I insulin fibrils (which are only 3 nm thick), we performed infrared s-SNOM imaging at different frequencies provided by our QCL. For nano-FTIR of such thin fibrils, the broadband laser source does not yet provide enough infrared power. Figure 6e shows infrared near-field phase images at four different wavelengths, exhibiting clear contrast for both type I and 9 nm thick fibrils. From altogether 12 images, we extracted local infrared spectra at the positions marked by the red, green and blue symbols in Fig. 6a. The spectra are plotted in Fig. 6f using the corresponding symbols. For both the 9-nm-thick fibril (red and green symbols) and type I (blue symbols) fibril, we find the same spectral signature as observed in the nano-FTIR spectrum (thick red curve in Fig. 6f). The spectrum acquired at the position marked by the blue symbol in Fig. 6a thus provides experimental evidence that α-helices are also present in type I fibrils. The current resolution of about 30 nm does not allow for concluding whether the α-helices are inside the core or forming a shell. Assuming that the core is formed of purely β-sheets (according to current models41), our findings suggest that the shell is highly structured (mainly α-helical structure) and not randomly organized. The presence of α-helices in the shell could explain the tendency of fibrils to associate.

Set-up and methodology for hyperspectral infrared nanoimaging

We developed hyperspectral infrared nanoimaging (Fig. 1) using a commercial nano-FTIR set-up (Neaspec GmbH). It is based on an AFM, where a standard Au-coated tip is vertically vibrating at the mechanical resonance frequency Ω of the cantilever. The tip is illuminated with a DFG-generated mid-infrared laser continuum of about 350 cm−1 spectral bandwidth, which centre frequency is tuneable between 1,200 and 1,600 cm−1 (see Fig. 1a and Supplementary Note 1). The tip-scattered light is recorded with a Michelson interferometer. To perform background-free nano-FTIR spectroscopy14,38,39, the detector signal is demodulated at a higher harmonic n of the tip’s oscillation frequency, nΩ, and recorded as a function of the reference mirror position d, yielding the interferogram I(d). Throughout this work, the demodulation order was n=3. Because tip and sample are located in one interferometer arm, Fourier transform of I(d) yields amplitude ss(ω) and phase ϕs(ω) spectra14,27. To obtain normalized nano-FTIR spectra s=ss/sref and ϕ=ϕs−ϕref, the tip is positioned on a reference area on the sample (typically a clean gold or silicon surface)14,23 to record the reference spectra sref(ω) and ϕref(ω).

For hyperspectral nanoimaging we record interferograms at each pixel (x, y) of a 2D area of the sample surface (Fig. 1b). Subsequent Fourier transform and normalization to reference spectra yields a 2D array of nano-FTIR spectra with a spectral bandwidth determined by the output spectrum of the DFG laser source, that is, a hyperspectral data cube A(x, y, ω) where x and y represent two spatial dimensions and ω the frequency (spectral dimension). For the present work, we recorded and studied nano-FTIR phase spectra ϕ(ω), as they are related to the sample’s infrared absorption16,27, the later being typically analysed when infrared spectroscopy of organic materials is performed. To increase the spectral bandwidth, we record data cubes Ak=ϕk(x, y, ω) at three different DFG output spectra (indicated by the index k=I,II,III and shown in the upper panel of Fig. 1c) and stitch together at each pixel the corresponding normalized nano-FTIR phase spectra (Fig. 1c, lower panel). As a result, a hyperspectral data cube A is obtained. In Fig. 1d we show the hyperspectral infrared data cube of a three-component polymer blend on a silicon substrate, exhibiting a spatial resolution of about 30 nm (further details see below). Cutting the cube at different infrared frequencies ω yields monochromatic infrared images. They clearly reveal a rich variety of spectrally and spatially varying features, indicating that neither individual point spectra nor individual monochromatic images provide the full information content contained in the hyperspectral data of this sample.

In the following we describe the key implementations of our technique (for more technical details see Supplementary Notes 1–6).

We increased the data acquisition speed by improving the signal-to-noise ratio (SNR) of the individual nano-FTIR spectra. To that end, the output power of our mid-IR laser continuum (previously reported in refs 14, 23) was increased from about 100–250 μW to 600 μW. Further, because of the asymmetric interferometer set-up (that is, the sample is located in one of the interferometer arms), we record only one half of the interferogram (Fig. 1b), as the other one does not contain spectroscopic information about the sample (see Supplementary Note 2 and Supplementary Fig. 1). With our improved laser source we succeeded to obtain about 350 cm−1 broad nano-FTIR spectra of organic materials in 1.66 s, which is more than one order of magnitude faster than what has been previously achieved with DFG and synchrotron radiation14,18,23.

For accurate and reliable normalization of the nano-FTIR spectra, the reference spectra need to be recorded under the same experimental conditions. Due to slight spectral fluctuations of the laser continuum and drift of the interferometer arms (see discussion below, Supplementary Note 5 and Supplementary Fig. 3), however, the conditions may vary between sample and reference measurements. For that reason, we regularly acquire interferograms of a reference area while recording the data cube. This can be achieved, for example, by recording the data cube such that each line contains a clean reference area (for example, silicon, marked Ref. in Fig. 1a where the line scans are parallel to the y-axis). The nano-FTIR spectra of each line are then normalized to the reference spectrum included in this line.

Depending on the number of individual spectra, the acquisition of a data cube Ak may still take several minutes to few hours. To avoid artifacts due to sample drift, we adapted concepts known from other imaging techniques (for example, electron energy loss spectroscopy mapping or AFM40). In brief, after n spectroscopic line scans (that is, recoding nano-FTIR spectra along n complete lines of the image), the sample is repositioned, yielding a data cube in which sample drift has been compensated (see Supplementary Note 3 and Supplementary Fig. 2). To this end, the spectroscopic data acquisition is stopped after each block of m line scans and a topography image of the sample including a reference point is recorded. The position of the reference point relative to its previous position is measured. The sample scanner is accordingly repositioned. We chose the number m such that the sample drift (determined essentially by the temperature stability of the setup) during the m line scans is smaller than the spatial resolution of about 30 nm. In the experiments presented in Figs 3 and 4 we set m=2 and m=4, respectively.

To combine the individual bandwidth-limited data cubes Ak, we need to combine at each position (x, y) the individual phase spectra ϕk(x, y, ω) obtained with the different DFG settings. A key for achieving this task is the sample drift correction during the recording of each bandwidth-limited data cube. As outlined above, the sample drift correction ensures that the position uncertainty is less than the spatial resolution, thus ensuring that spectra of the same position (x, y) are combined.

The remaining challenges and solutions for combining the individual spectra are shown in Fig. 2. We recorded sample-drift-corrected data cubes of the same sample area using the three different laser outputs shown in Fig. 2b (numerated k=I, II, III). In Fig. 2a (bottom) we show the bandwidth-limited nano-FTIR phase spectra ϕk(x, y, ω) of six subsequent pixels on the polymer blend sample. For better comparison, and to demonstrate the reproducibility of the individual spectra, the upper panel of Fig. 2a displays all spectra plotted on top of each other. Within the overlapping spectral regions (marked by grey areas) we clearly observe the same spectral features in the adjacent bandwidth-limited nano-FTIR phase spectra. However, the spectra can be significantly offset against each other by up to 12 degrees, although all spectra are normalized to a reference spectrum (see above). We explain this phase offset Δϕ (marked in Fig. 2a) by a small unavoidable drift of the interferometer paths (LR and Lt in Fig. 1a), which occurs between the acquisition of the individual nano-FTIR spectra and the acquisition of the reference spectrum (see Supplementary Note 5 and Supplementary Fig. 3). A drift as small as 100 nm of LR relative to Lt shifts the normalized nano-FTIR phase spectrum by about Δϕ=6 degree, which is in the same order of magnitude as the phase shift produced by absorption in the sample. To reduce offset fluctuations below 1 degree, the path lengths LR and Lt (of about 6 cm) need to be stabilized with a precision better than 20 nm, which, however, will require sophisticated technology development in the future. Here we tackled the problem by correcting the phase offset. To that end, we shift at each pixel (x, y) the phase spectra ϕI(ω) and ϕIII(ω) by the constant phase values and (the average values are evaluated in the corresponding spectral overlap regions marked grey in Fig. 2), to match ϕII(ω). The offset-corrected phase spectra are shown in Fig. 2c. We find that the spectral features (marked by dashed black circles) in the overlapping regions are now well matched for all pixels, thus verifying the validity and reliability of our rather simple offset-correction procedure. Finally, we combine the three offset-corrected spectra at each pixel to obtain a single-broadband nano-FTIR phase spectrum. To achieve a smooth transition between the spectral ranges, we multiply the phase spectra ϕk by the functions Fk(ω) shown by the red, green and blue graphs in Fig. 2d. Subsequently, the spectra are summed up. The final broadband nano-FTIR phase spectra are shown in Fig. 2e. Altogether, this methodology enables an automatized combination of bandwidth-limited data cubes to one hyperspectral data cube.

Analogue to far-field FTIR spectroscopic imaging, we finally apply a baseline correction41,42 to each broadband (composite) nano-FTIR phase spectrum ϕ(ω). We select two frequencies ωb1 and ωb2 (marked in Fig. 2f, where we know or can assume that the sample absorption is negligible, see Supplementary Fig. 5) and subtract the linear baseline defined by the phase values ϕ(ωb1) and ϕ(ωb2). A comparison of baseline-corrected spectra (Fig. 2f) with the corresponding uncorrected spectra (Fig. 2e, upper panel) clearly shows the reduction of fluctuations between neighbouring spectra. Hence, the baseline correction significantly improves the SNR in the images extracted from the hyperspectral data cube (see Supplementary Note 6 and Supplementary Fig. 4).

The recording of a bandwidth-limited data cube Ak consisting of 5,084 nano-FTIR spectra (that is, 82 × 62 pixels, Fig. 1d) took (5,084 × 1.66 s)/3,600=2.3 h while the accumulated additional time needed for the sample repositioning was of less than 10 min (for more details see Supplementary Note 3). Thus, the recording of the hyperspectral data cube A shown in Fig. 1d required 7.4 h.

Hyperspectral chemical nanoimaging of a polymer blend

In a first application example, we demonstrate hyperspectral IR nanoimaging with a three-component polymer blend, showing that this novel tool can meet the strong demand for highly sensitive nanoscale chemical mapping of the spatial distribution and local chemical interaction of the components. The model system studied in this work is based on a fluorine copolymer (FP), an acrylic copolymer (AC), and a polystyrene latex (PS; for details and fabrication see ‘Methods’ section). Figure 3a shows the topography image of the about 170 nm thick spin-coated polymer blend on silicon, while the hyperspectral infrared data of this sample area are displayed in Fig. 1d. The good reproducibility of the individual spectra (see Fig. 3b showing four sets of neighbouring point spectra at the sample positions A to D marked in Fig. 3a,d) allow for valuable multivariate analysis based on established procedures known from far-field IR spectroscopy, as we demonstrate in the following.

We first applied inter-spectral distance mapping, where we calculate for each pixel spectrum S(x, y) the multivariate spectral distance D(x, y) between the point spectrum and a reference spectrum (see ‘Methods’ section). The distance values are converted into colour scales and plotted as specifically coloured pixels at the positions (x, y). In the resulting distance maps, high colour intensity denotes a small distance (high similarity) with the reference spectrum (and vice versa). With nano-FTIR spectra obtained from reference samples made of pure AC and FP components (black spectra in Fig. 3b) we obtained the distance (similarity) maps for the AC (red) and FP (blue) components shown in Fig. 3c. Superposition of the two maps (see ‘Methods’ section) yields a compositional map (Fig. 3d), which highlights areas with the highest relative content of each polymer. We observe homogenous but distinct red and blue areas for the two different references, indicating that the AC and FP components are not fully mixed but rather separated. On the red areas, representative spectra (B) match well with the AC reference (black reference spectrum in Fig. 3b), indicating the presence of pure AC (illustrated by situation B in Fig. 3g). Within the blue areas, interestingly, representative spectra (D in Fig. 3b) show FP peaks at low frequencies (ω<1,500 cm−1) but also the C=O peak of AC at 1,740 cm−1. It can be shown (Supplementary Fig. 6) that the spectra at position D are a linear superposition of the pure FP and AC spectra, which lets us conclude that both FP and AC are present within the volume probed by the near field below the tip apex (represented in Fig. 3g by the reddish elliptical area below the tip apex). We conclude that FP forms cluster-like nanostructures, while AC is spread all over the sample surface. For that reason, the near field below the tip apex probes both the FP cluster and AC layer below the FP cluster (illustrated by situation D in Fig. 3g). We explain this finding by the differences in viscosity, molecular weight and/or chain stiffness. On the other hand, the clearly visible spatial dispersion of AC and FP—avoiding the formation of percolation networks—indicates that AC and FP are miscible even at the nanoscale. Note that any improvement of film formation or film homogenization by mixing or optimizing drying conditions was beyond the scope of this study.

In Fig. 3d we also observe black areas on the polymer blend, indicating that the similarity of the local spectra (A in Fig. 3b) is low compared with the FP and AC reference spectra. We explain these areas by the dominating presence of PS (illustrated by situation A in Fig. 3g). Note that we did not perform distance mapping with PS references, as the peaks in nano-FTIR spectra of pure PS reference sample are comparably weak.

Interestingly, Fig. 3d reveals purple areas (marked C in Fig. 3d), which indicate local spectral differences compared with both the AC (red) and FP (blue) regions. Indeed, representative spectra (marked C in Fig. 3b) of the purple areas cannot be reproduced by a linear superposition of AC and FP reference spectra (see Supplementary Fig. 6). Compared with the red AC spectrum (B), the C=O peak at 1,740 cm−1 is reduced, indicating that the amount of AC in the probing volume of the tip is reduced. On the other hand, the peak at 1,155 cm−1 is significantly increased. We attribute this finding to the presence of FP and its chemical interaction with AC. Note that chemical interactions are known to cause peak shifts33 and increase of peak heights43. Specifically, we assume a spectral shift of the CF2 stretching vibration of FP from 1,195 cm−1 towards lower frequencies, most likely due to the formation of hydrogen bonding between the C-F bonds of FP and the acrylic polymer chains44,45. Further, the C-O stretching of the ester bond of AC at 1,155 cm−1 may be enhanced due to the chemical interaction. The concerted action of both effects could thus explain the enhanced peak at 1,155 cm−1 that is found in the purple regions, which consequently indicate the areas where the AC and FP components are well mixed.

We note that similar compositional maps as the ones of Fig. 3c,d can be obtained with far-field infrared reference spectra of the pure AC and FP samples, as we demonstrate in the Supplementary Fig. 7 with the help of attenuated total reflectance FTIR (ATR-FTIR) spectra. This possibility enables rapid hyperspectral data analysis (that is, identification of specific target components) based on standard references without the need of nano-FTIR reference spectra.

Next we demonstrate that valuable multivariate data analysis can be also performed without any reference spectra. To this end, we apply unsupervised hierarchical cluster analysis46 (see ‘Methods’ section), which was used to segment the hyperspectral data of Fig. 1d into five distinct clusters (denoted as cl1 to cl5, see Fig. 3e). We justify the choice of at least four clusters by our findings from inter-spectral distance mapping (Fig. 3d). On the other hand, segmentation into more than five clusters resulted essentially in further segmentation of the PS-rich (dark) areas. In the future, a more detailed analysis could be applied to determine the optimal number of clusters47, which, however, would go beyond the scope of this work. The cluster map is shown in Fig. 3e, and the cluster spectra (coloured) in Fig. 3f. The spectra and areas of clusters cl1, cl2, cl4 and cl5 agree well with Fig. 3b,d, respectively, corroborating the robustness and reliability of the data and the strategy of data analysis. Interestingly, unsupervised cluster analysis reveals another significant area (cluster cl3, green) which is not recognized in Fig. 3d, and which average spectrum (green curve in Fig. 3f) can be reconstructed by a linear superposition of AC and FP reference spectra. We explain it by either a lateral (illustrated by situation E1 in Fig. 3g) or vertical (illustrated by situation E2 in Fig. 3g) arrangement of AC and FP within the volume probed by the tip’s near field. For that reason, the green areas indicate interfacial areas without chemical interaction, in contrast to the purple regions where peak shifts indicate significant chemical interaction. Note that polymer chain interactions depend on several factors such as distance between interacting groups, orientation or steric hindrance43,44, and thus may occur only partially at the interface between AC and FP.

In situ analysis of native melanin in human hair medulla

In Fig. 4 we apply hyperspectral IR nanoimaging to perform the first in situ infrared-vibrational chemical analysis of native melanin in human hair medulla. Melanin is a polymer pigment, present in the human hair and skin, and responsible for tissue colouring and ultraviolet photoprotection48. Due to its photo absorbing properties, melanin has attracted large attention from cosmetic and solar energy industries49,50. Unfortunately, it has revealed impossible to analyse human melanin without extracting it from hair or tissue, which comes along with potential damage and modification51,52.

Figure 4a shows the topography of a resin-embedded cross-section of a hair. In the infrared near-field image taken with a quantum cascade laser at 1,660 cm−1 we observe an enhanced infrared absorption of the cuticle and cortex regions compared with the resin, owing to the strong amide I absorption of the hair proteins (α-keratin microfibrils). Within the cortex region we find disk-shaped areas of about 300 nm diameter, where the infrared absorption is reduced (that is, the protein content is reduced). Their size and distribution corresponds to that of melanin granules observed in electron microscopy images53. However, we found that nano-FTIR spectra of the individual granules can differ significantly from each other. To elucidate the spectroscopic variations, we performed hyperspectral IR nanoimaging of the area marked by dashed black line in Fig. 4a. From the hyperspectral data cube (Fig. 4b) we extracted spectra at different positions (Fig. 4c). Within the cortex region (position C, green spectrum in Fig. 4c) we observe the well-known amide I and II bands being typical for protein (α-keratin). For particle A (blue curve in Fig. 4c) we observe four distinct peaks that are characteristic for melanin: at 1,290 cm−1 (–C–OH phenolic stretching), 1,454 cm−1 (C–C aliphatic stretches), 1,563 cm−1 (indole N–H bending) and 1,638 cm−1 (C=C, C=O, and/or COO- stretching in aromatic cycle)54. On particle B—supposed to be a melanin granule—the nano-FTIR spectrum significantly differs from that of particle A. Surprisingly, spectrum B shows more similarity to the keratin spectrum of the cortex region (C), although the amid I and II peaks are shifted by several cm−1. Further, while both particles A and B are seen in the monochromatic image at 1,670 cm−1, only particle A exhibits a contrast at 1,580 cm−1 (see corresponding slices of the data cube in Fig. 4b). Obviously, a set of images and local spectra is not sufficient to clarify the identity of particle B. However, having acquired a full hyperspectral data cube, we can take advantage of multivariate data analysis. Since there are no reference spectra available for natural melanin in hair, we performed unsupervised hierarchical cluster analysis of the hyperspectral data. The segmentation map resulting from cluster analysis with three clusters (Fig. 4d) reveals well-defined features, thus corroborating the applicability and robustness of cluster analysis to the nano-FTIR spectra. We note that cluster analysis with more than three clusters does not reveal well-defined new features (see Supplementary Fig. 8). In both maps the particles A and B appear as homogenous, although as distinct clusters (circular blue and red areas, respectively). Most important, we find a red ring (D) around the blue central area of particle A, revealing that the corresponding spectra belong to the same cluster as those of particle B. It can be shown that the spectra of the red cluster (B and D) are a linear superposition of spectra A and C (see Supplementary Fig. 9). For the width of the red ring (D) we measure about 50 nm, which is in the range of the lateral spatial resolution. We thus conclude that the red ring highlights a steep interface between melanin and keratin (see illustrations D in Fig. 4e). The red area B, in contrast, is a closed disk-shaped area of about 200 nm diameter (that is, larger than the spatial resolution). We thus conclude that spectra B of this area are due to a vertical arrangement of melanin and keratin, that is, a horizontally oriented interface between them. Indeed, near-field probing can be sensitive to subsurface components55, which lets us conclude that particle B is either a subsurface melanin granule or a thin slice of a melanin granule (see illustrations B in Fig. 4e; note that melanin granules are elliptical vesicles of an aspect ratio of about 3:1).

One thought on “Laser Infrared Sample Analysis Essays

Leave a comment

L'indirizzo email non verrà pubblicato. I campi obbligatori sono contrassegnati *