Photothermal microscopy, a pump-probe method sensitive to local thermal gradients, offers a powerful approach for mapping nanostructures and molecules. While visible and mid-infrared (MIR) photothermal microscopy techniques exist, they have limitations. Visible methods often lack chemical specificity, while MIR techniques suffer from significant light attenuation due to water absorption. This necessitates the development of a new technique that combines the advantages of both approaches while mitigating their respective shortcomings. The shortwave infrared (SWIR) window (900 nm to 2 µm) offers a potential solution, possessing a significantly smaller water absorption coefficient compared to the MIR region and allowing for greater penetration depth due to reduced light scattering. This makes SWIR spectroscopy an attractive analytical tool for bioimaging, especially when exploiting the rich chemical information encoded in overtone absorption bands. However, existing SWIR imaging methods, such as hyperspectral reflectance/transmittance imaging, diffuse optical spectroscopic imaging (DOSI), and photoacoustic microscopy (PAM), often compromise either spatial resolution or sensitivity to achieve sufficient penetration depth. This study introduces overtone photothermal (OPT) microscopy to address this limitation by combining the chemical specificity of SWIR excitation with the high sensitivity and resolution of visible photothermal detection. The technique aims to provide a powerful tool for high-resolution, high-sensitivity chemical imaging across a variety of applications, from materials science to biological systems.

Several SWIR imaging methods have been previously developed, each with its own strengths and weaknesses. Hyperspectral reflectance/transmittance imaging, while providing qualitative spectral characterization, generally sacrifices spatial resolution for a larger field of view. Diffuse optical spectroscopic imaging (DOSI), a wide-field approach, offers sub-millimeter lateral resolution but is limited in depth penetration. Photoacoustic microscopy (PAM), using ultrasonic transducers to detect acoustic waves generated by optical absorption, has shown promise for imaging lipids and water but its sensitivity is heavily dependent on the quality of the transducer. Existing photothermal microscopy techniques, based on either visible or mid-infrared excitation, also have limitations. Visible photothermal microscopy, often relying on gold nanostructures for signal enhancement, may lack chemical specificity. Mid-infrared photothermal (MIP) microscopy, while offering bond-selective imaging, suffers from significant water absorption, limiting its applicability in biological systems. OPT microscopy aims to overcome these limitations by leveraging the advantages of SWIR excitation and visible detection for improved sensitivity, resolution, and penetration depth.

A custom-built OPT microscope was developed, utilizing a tunable femtosecond SWIR pump laser (1080-1280 nm) and a 520 nm probe beam (generated by second harmonic generation of a 1040 nm laser). Both beams were chirped to picosecond durations to minimize photodamage. The beams were collinearly focused onto the sample using a high-numerical aperture (NA) water immersion objective. The axial positions of the pump and probe beams were offset to optimize photothermal signal generation. The thermal lensing effect induced by selective absorption of the SWIR pump beam altered the propagation of the probe beam, which was then detected by a photodiode and processed by a lock-in amplifier to extract the OPT signal. Spectroscopic OPT imaging was performed by tuning the SWIR wavelength and recording images via laser scanning. For depth-resolved imaging, the sample was moved incrementally along the axial direction using a piezo stage. Spectral unmixing, using the least absolute shrinkage and selection operator (LASSO) algorithm, was employed to resolve the overlapping spectral contributions of different chemical components (e.g., protein, fatty acids, and water) in biological samples. Reference spectra for spectral unmixing were obtained from pure samples of bovine serum albumin (BSA), triglyceride (TAG), and water. The LASSO algorithm incorporated a sparsity constraint to improve chemical specificity by minimizing signal cross-talk. Various samples were imaged including polymer nanostructures, cancer cells, C. elegans worms, and mouse brain tissue. In some cases, heavy water (D₂O) was substituted for regular water in the imaging buffer to reduce background signal from water absorption.

OPT microscopy demonstrated excellent spectral fidelity, as confirmed by comparison of OPT spectra with those obtained using a UV-Vis-NIR spectrophotometer. The technique achieved a lateral resolution of 405 nm, allowing for the clear visualization of individual 200-nm PMMA beads. The OPT signal intensity showed a linear relationship with both pump and probe power. The detection limit for DMSO in glycerol-d8 was determined to be 0.3%. Three-dimensional (3D) imaging capability was demonstrated through depth-resolved imaging of a 1-µm PMMA bead, revealing an axial resolution of 2 µm. OPT microscopy successfully imaged phase separation patterns in PS-PMMA polymer blends, clearly distinguishing between PS and PMMA domains based on their distinct spectral signatures. The technique also imaged fabricated PMMA nanostructures with high fidelity. In biological samples, OPT microscopy was used to perform depth-resolved chemical mapping of protein and fatty acids in OVCAR-5 ovarian cancer cells. Using LASSO spectral unmixing, protein and fatty acid distributions were distinguished even in the presence of overlapping spectral features and water absorption. The depth-resolved imaging revealed varying distributions of protein and fatty acids within the cells. Similar depth-resolved chemical mapping was successfully performed on C. elegans worms, distinguishing between protein and fatty acid distributions in different tissues. Finally, OPT microscopy successfully achieved depth-resolved imaging (up to 100 µm) in highly scattering mouse brain tissue, differentiating protein and lipid distributions in the corpus callosum region.

OPT microscopy successfully addresses the limitations of existing SWIR and photothermal microscopy techniques. By utilizing SWIR excitation and visible detection, it achieves both high resolution and high sensitivity, surpassing the capabilities of other SWIR imaging methods and overcoming the water absorption limitations of MIP microscopy. The high resolution of OPT microscopy, coupled with LASSO-based spectral unmixing, enables accurate identification and quantification of chemical species even in complex biological environments. The demonstrated ability of OPT microscopy to image deep into scattering tissues like brain slices expands its potential for applications in biomedical research. While the current implementation has a limited depth penetration (~100 µm), this is significantly improved compared to MIP microscopy. The use of a high-NA objective contributes to the high resolution and sensitivity of the technique. Future improvements could include expanding the spectral range, increasing imaging speed, and incorporating epi-detection to further enhance its capabilities.

This study presents overtone photothermal (OPT) microscopy as a powerful new technique for label-free, bond-selective imaging. OPT microscopy bridges the gap between visible and mid-infrared photothermal microscopy, offering superior sensitivity and resolution in the SWIR region. Its successful application in various materials and biological systems demonstrates its broad utility. Future research will focus on expanding the spectral range, improving imaging speed, and exploring the use of epi-detection to enhance its versatility and applicability.

One limitation of OPT microscopy is the overlapping nature of overtone bands in the SWIR region, leading to potential cross-talk between chemical species. While LASSO spectral unmixing effectively mitigates this issue, improvements in spectral resolution and unmixing algorithms could further enhance chemical specificity. Another limitation is the relatively limited penetration depth in highly scattering samples, although this is considerably better than mid-infrared methods. The current implementation is also relatively slow for hyperspectral imaging. Addressing these limitations will be the focus of future studies.