2.1.

Materials

Caucasian untreated natural white hair, obtained from a commercial source (International Hair Importers, New York), was formed into swatches ( $16\mathrm{cm}$ , $1.5\mathrm{g}$ ). White hair was chosen to avoid putative photodamage induced by linear absorption of the laser light by melanin. The hair was cut into $\sim 1\text{-}\mathrm{cm}$ segments and mounted on a glass coverslip using double-sided adhesive sheets (Grace Biolab). Because glass has no resonances in the spectral region of interest, a small coverslip ( $\sim 170\mu \mathrm{m}$ thick) was placed next to the hair to serve as an internal calibration for normalizing the CARS signal. ${\mathrm{D}}_2\mathrm{O}$ and d-glycine (d5) were obtained from Sigma-Aldrich. In this study, d-glycine is used as a model for glycine because of the enhanced CARS contrast of provided by the deuterium label.

In the d-glycine uptake experiments, the hair was immersed for at least $4\mathrm{h}$ in an aqueous solution of $3.3\mathrm{M}$ d-glycine and $0.52\mathrm{M}$ ammonium carbonate, adjusted to pH 9.0 for efficient uptake into the hair.¹⁹ The hair was kept in the solution during the imaging experiments.

2.2.

Imaging

A $76\text{-}\mathrm{MHz}$ mode-locked laser source (PicoTrain, High-Q Lasers), which produced $7\text{-}\mathrm{ps}$ -pulsed $1064\text{-}\mathrm{nm}$ light, was used as the Stokes beam in the CARS process. The source also produces a frequency-doubled, $532\text{-}\mathrm{nm}$ beam, which was used to pump a picosecond optical parametric oscillator [(OPO), Levante Emerald]. The OPO delivers radiation tunable in the $650–970\mathrm{nm}$ range, which was employed as the pump beam in the CARS process. The two beams were collinearly overlapped on a dichroic mirror and directed to a galvanometric scanner (Fluoview 300V, Olympus) before entering an inverted microscope (IX-71, Olympus). A $40\times$ , NA 1.15 water-immersion objective lens (UplanApo, Olympus) was utilized in these studies. Average illumination dose was $15\mathrm{mW}$ per beam. The forward-propagating CARS signal was collimated by a $0.55\text{-}\mathrm{NA}$ condenser, passed through two bandpass filters, and detected by a red-sensitive photomultiplier tube [R3896, Hamamatsu]. All images $(512\times 512\text{pixels})$ were collected in $1.5\mathrm{s}$ per frame and averaged three times. CARS spectra were obtained by scanning the OPO wavelength while keeping the $1064\text{-}\mathrm{nm}$ Stokes beam fixed. All CARS signals were normalized relative to the nonresonant signal from a glass coverslip.

2.3.

Image Processing

To convert the CARS intensity images to concentration maps, we have employed the enhanced contrast obtained when using ratio images. For a given vibrational band, ratio images between the maximum CARS signal and the minimum CARS signal were generated. Because of the dispersive line shape of CARS spectral features, the resonance peak was taken as the maximum and the dip on the blue side of the Raman resonance frequency was chosen as the minimum CARS signal. Ratio images were constructed in the following way. (i) An image was taken at the CARS spectral maximum, followed by a background reference image obtained by blocking the Stokes beam. (ii) An image at the CARS spectral minimum was recorded, and supplemented with a background reference image. (iii) The backgrounds were subtracted from the maximum and minimum CARS images. (iv) The resulting images were corrected for the intensity dependence of the field of view by dividing each by a prerecorded CARS field-of-view intensity profile. The field-of-view correction profile was obtained for each spectral setting and obtained by recording a field profile of the nonresonant signal of the supporting glass coverslip. Each correction profile was corrected for any background contributions. (v) The CARS ratio image was obtained by dividing the maximum by the minimum image. Because all ratio images are scaled relative to the nonresonant glass signal, a direct quantitative comparison between the processed images is possible. All image processing was performed within the ImageJ environment,²⁰ using home-written software. The total processing time for generating a calibrated concentration map is $<1\min$ .

2.4.

Concentration Dependence

At low concentrations of the Raman scatterer, the ratiometric CARS signal is more sensitive than the regular CARS intensity to changes in the concentration. CARS imaging based on a comparison of signals measured at different Raman shifts is a common procedure for improving the image contrast.^{21, 22} The ratio signal employed here enhances the contrast obtained from small concentration differences. Under the assumption that the nonresonant electronic contribution is invariant among the sample components, the ratiometric signal $R\left(c\right)$ can be written as

Concentration maps were determined by converting the ratio scale to a concentration scale using a calibration graph. For both water and d-glycine, the maximum and minimum CARS signal was obtained as a function of the concentration of the target compound. A calibration curve was acquired by plotting the corrected CARS ratio between the maximum and minimum against concentration. A polynomial fit was used to describe the concentration dependence of $R\left(c\right)$ . Note that the glass-normalized CARS ratio is always independent of such factors as image-to-image laser intensity variation, which enables a direct conversion from CARS ratio images to concentration with the aid of the ratiometric calibration curve.

3.1.

CARS Microscopy of Endogenous Hair Components

Before studying the concentration of externally applied compounds to the hair, we first examined the CARS response of several major endogenous components of the hair. The CARS spectrum of ${\mathrm{D}}_2\mathrm{O}$ -immersed hair measured in both the cortex and medulla is given in Fig. 1 . Relative to a dry hair, immersion in ${\mathrm{D}}_2\mathrm{O}$ leads to a better refractive index match between the slide and sample. The ${\mathrm{D}}_2\mathrm{O}$ was chosen because it is minimally vibrationally resonant in the $2800–3000{\mathrm{cm}}^{-1}$ range, the spectral region of the $\mathrm{C}{\mathrm{H}}_2$ - and $\mathrm{C}{\mathrm{H}}_3$ -stretching vibrational modes.

Fig. 1

CARS spectrum of the cortex (black dots) and the medulla (red dots) of human hair. The $\mathrm{C}{\mathrm{H}}_2$ -symmetric stretch $\left(2840{\mathrm{cm}}^{-1}\right)$ indicates the presence of lipids in the medulla, whereas the cortex is rich in $\mathrm{C}{\mathrm{H}}_3$ -symmetric stretch $\left(2920{\mathrm{cm}}^{-1}\right)$ associated with keratin structural proteins. (Color online only.)

As is evident from Fig. 1, the spectra of the cortex and the medulla are different. In the medulla, a higher contribution from the symmetric $\mathrm{C}{\mathrm{H}}_2$ stretching mode $\left(2840{\mathrm{cm}}^{-1}\right)$ is observed, an indication of a higher density of lipid aliphatic chains. A higher relative lipid concentration in the medulla was also observed in infrared microspectroscopy studies.²³ Importantly, compared to infrared microscopy techniques, the higher resolution in CARS allows for a much finer assessment of the distribution of lipids in the medulla. It is seen in Fig. 2 that the lipids are nonhomogeneously distributed. The lipid clusters resemble the globular structures previously observed with SEM.⁴ The spectral profile of the cortex mimics the CARS spectrum of structural proteins.²⁴ This is an expected result, as the density of keratin in the cortex is higher than in the porous area of the medulla.

Fig. 2

CARS images of human hair at different Raman shifts. Images were taken at the equatorial plane of the hair: (a) nonresonant background contribution from both cortex and medulla at $2787{\mathrm{cm}}^{-1}$ , (b) lipid contrast of the $\mathrm{C}{\mathrm{H}}_2$ -symmetric stretch probed at $2832{\mathrm{cm}}^{-1}$ . Note that the medulla exhibits lipid rich areas not seen in the cortex, and (c) protein contrast based on the $\mathrm{C}{\mathrm{H}}_3$ symmetric stretch at $2920{\mathrm{cm}}^{-1}$ . A strong signal from the keratin-dense cortex is observed.

3.2.

Water Concentration in the Hydrated Hair

The vibrational sensitivity of CARS permits the registration of externally applied compounds in the hair. To visualize water, the OH-stretching vibrational range is used for generating contrast. The CARS spectrum of water is depicted in Fig. 3, showing the broad band of the OH-stretching vibrations. To quantitatively analyze the distribution of water within the hair, ratio images of the hydrated hair were constructed by taking images at 3100 and $3550{\mathrm{cm}}^{-1}$ . Note that both vibrational energies exhibit minimum overlap with the Raman vibrational energies of the hair’s endogenous components. A calibration curve was generated by using solutions of varying ${\mathrm{H}}_2\mathrm{O}∕{\mathrm{D}}_2\mathrm{O}$ ratios to record the dependence of the ratio signal as a function of ${\mathrm{H}}_2\mathrm{O}$ volume percent. In determining the calibration curve, we have ignored the subtle spectral changes as a consequence of HOD formation. The calibration curve for the ratio signal ratio is depicted in Fig. 4 . The concentration dependence of the ratiometric signal exhibits a smooth profile and is well described by the polynomial fitting function plotted in Fig. 4.

Fig. 3

CARS spectrum of water in the OH-stretching region. The CARS signal is measured relative to the nonresonant response from a glass coverslip. The signal around $3100{\mathrm{cm}}^{-1}$ is taken as the maximum and the dip at $3550{\mathrm{cm}}^{-1}$ is taken as the minimum in the CARS ratio images for water.

Fig. 4

Dependence of the CARS ratio signal as a function of water volume concentration in a ${\mathrm{H}}_2\mathrm{O}:{\mathrm{D}}_2\mathrm{O}$ mixture. The data points are fitted to a polynomial.

In Fig. 5a, the water concentration map of a hair immersed in water is shown. The water concentration in the hair is homogenously distributed, and the spatially averaged concentration was found to be 34% by volume. At full hydration, hair can absorb $\sim 30\%$ of its dry weight in water.¹⁸ The density of dry hair is $\sim 1.32\mathrm{g}∕{\mathrm{cm}}^3$ , and the hair can absorb $\sim 0.40\mathrm{g}∕{\mathrm{cm}}^3$ of water.² Assuming a 30% volume increase of the fully hydrated hair,²⁵ the maximum water concentration in the hair amounts to $\sim 31\mathrm{vol}\%$ , which is in close agreement with the water concentration found in the CARS measurement.

Fig. 5

Water concentration maps determined from CARS ratiometric images $(3100{\mathrm{cm}}^{-1}∕3550{\mathrm{cm}}^{-1})$ at the OH-stretching vibration: (a) water concentration in a hair immersed in pure ${\mathrm{H}}_2\mathrm{O}$ and (b) water concentration in a hair immersed in pure ${\mathrm{D}}_2\mathrm{O}$ . The lack of contrast in the ${\mathrm{D}}_2\mathrm{O}$ image indicates that the ratiometric contrast of the hair itself is negligible.

Note that in the ratio images, the signal-to-noise near the boundary of the hair is rather low, due to the fact that at the hair/water interface the CARS signal is severely reduced as a consequence of light scattering. Nonetheless, while the signal from these regions cannot be used for a quantitative analysis, they provide a clear demarcation of the hair’s surface. To facilitate the analysis, we have chosen a hair that is devoid of a highly scattering medulla. As a control, the water concentration in a hair immersed in pure ${\mathrm{D}}_2\mathrm{O}$ was determined. The ratio image of the ${\mathrm{D}}_2\mathrm{O}$ -immersed hair is shown in Fig. 5b. As expected, no significant ratiometric contrast was observed under these conditions, which illustrates the feasibility of the CARS ratio images as a probe for water concentration.

3.3.

Dynamic Water Uptake

When hair fibers are fully hydrated in pure water, a significant swelling of up to $\sim 30\%$ can be observed.²⁶ It is, however, unclear whether the swelling of the hair is linearly dependent on the absorbed water concentration or whether other dynamic structural changes during the hydration process are responsible for the increased diameter of the hair. Because of the fast imaging capability of the CARS microscope, a direct correlation between swelling and intrahair water concentration can be obtained.

We have performed dynamic water measurements in hair by probing the CARS water signal at $3100{\mathrm{cm}}^{-1}$ during the immersion of a dry hair in pure water. Representative results are depicted in Fig. 6, which shows the diameter and CARS signal as a function of time after immersion of the hair in water. The hair diameter is seen to expand by 15% (volume increasing by 31%) within $2.5\min$ after application of water. The CARS water signal increase follows the same temporal trend as the expansion of the hair. Note that in this dynamic measurement, the CARS intensity is monitored rather than the ratiometric signal. Although the relative CARS intensity is not a direct measure of the water concentration, it provides clear evidence for the direct temporal correlation between the water increase and the swelling of the hair fiber. After $2.5\min$ , the increase of the CARS signal in the cortex subsides, while the water signal measured in the medulla continues to rise. The different trend in water uptake between the cortex and the medulla is reproducible $(n=3)$ and indicates that saturation of water absorption in the keratin-dense cortex is reached before the maximum water concentration is attained in the medulla.

Fig. 6

Dynamic recording of the relative CARS signal at $3100{\mathrm{cm}}^{-1}$ as a function of time after immersion of hair in water. Black dots indicate the change of the hair diameter, red dots the temporal dependence of the relative CARS signal in the cortex, and the blue squares the gradual increase of the CARS signal in the medulla. (Color online only.)

3.4.

Uptake of Glycine by the Hair

Glycine is a small molecule of direct relevance to hair cosmetic products, yet its absorption by the hair and the subsequent distribution of the chemical in the cortex on a micrometer scale has hitherto not been characterized. Glycine has a Raman spectrum with insufficient spectral specificity for a proper discrimination in the hair matrix using CARS microscopy. However, spectral selectivity can be attained by using deuterated glycine. When fivefold deuterated d-glycine is dissolved in water, three of the substituted deuterium atoms readily exchange with hydrogen at the amido and hydroxy groups. Glycine remains deuterated at the $\mathrm{C}{\mathrm{D}}_2$ group in aqueous solution, and the single $\mathrm{C}{\mathrm{D}}_2$ unit of the molecule exhibits a unique spectral Raman signature with limited overlap from the vibrational response of endogenous hair components.

The CARS spectrum of a d-glycine crystal in the $\mathrm{C}{\mathrm{D}}_2$ stretching vibration range is given in Fig. 7 . Two clear peaks can be recognized, corresponding to the symmetric $\left(2145{\mathrm{cm}}^{-1}\right)$ and asymmetric $\left(2248{\mathrm{cm}}^{-1}\right)$ stretching vibration of the $\mathrm{C}{\mathrm{D}}_2$ unit. Upon dilution in water, these signatures can still be resolved, albeit at reduced contrast due to the relatively large contribution of the water nonresonant background. To improve contrast, the ratio between 2140 and $2160{\mathrm{cm}}^{-1}$ is taken as a measure for the presence of d-glycine. Figure 8 shows how the ratio signal scales with the concentration of d-glycine in water. A linear relation is found. Following Eq. 1, this linear dependence is expected for a weak Raman scatterer. The different concentration dependence of the ratiometric signal for d-glycine compared to water (Fig. 4) is largely due to the difference in the relative strength of the vibrationally resonant contribution of these different molecular agents.

Fig. 7

CARS spectra of d-glycine in the $\mathrm{C}{\mathrm{D}}_2$ stretching region. Black dots indicate the spectrum obtained from a d-glycine crystal, whereas red dots represent the spectrum of a saturated aqueous d-glycine solution $\left(3.3\mathrm{M}\right)$ . CARS signals are measured relative to the nonresonant response of a glass coverslip. (Color online only.)

Fig. 8

Dependence of the CARS ratio signal $(2140{\mathrm{cm}}^{-1}∕2160{\mathrm{cm}}^{-1})$ as a function of d-glycine concentration in an aqueous solution.

Figure 9 depicts d-glycine concentration maps obtained from ratio images in combination with the calibration curve of Fig. 8. When the hair fiber is immersed in a saturated solution of d-glycine, a significant fraction of the compound can be found inside the hair. It is seen that the distribution of d-glycine throughout the hair is virtually homogeneous. The concentration map shows that the d-glycine concentration outside the hair is $3.3\mathrm{M}$ , which corresponds to the expected concentration for a saturated solution. In the cortical regions of the hair, the average d-glycine concentration was found to be $0.22\pm 0.08\mathrm{M}$ $(n=3)$ . We found that the concentration varied from hair to hair. To test the accuracy of the d-glycine measurements, we have also performed reference measurements (at the d-glycine resonances) in a water immersed hair, showing a background contrast of $0.08\pm 0.16\mathrm{M}$ $(n=3)$ in the cortex. The difference between the d-glycine measurement and the background reference is evident, indicating that the concentration contrast observed in the hair fiber, as shown in Fig. 9, can be attributed to the presence of d-glycine.

Fig. 9

D-glycine concentration map of a hair immersed in a saturated solution of d-glycine, as determined from CARS ratio images $(2140{\mathrm{cm}}^{-1}∕2160{\mathrm{cm}}^{-1})$ . The concentration outside the hair is $3.3\mathrm{M}$ and the average concentration in the cortical regions of the hair is $0.22\pm 0.08\mathrm{M}$ .