Energy level diagrams of different third order nonlinear induced polarizations. (a) When the energy difference between pump and probe beam is resonant with a vibrational transition of the molecule, a strong resonant CARS signal at the anti-Stokes frequency is emitted. (b) When the energy difference between pump and probe beam is not resonant with any vibrational transitions of the material, a weak but non-vanishing signal, known as the non-resonant background, is still generated at the anti-Stokes frequency. (c) Stimulated Raman loss (SRL) occurring at the pump field frequency has the opposite (180 degree lag) phase compared to the pump field. (d) Stimulated Raman gain (SRG) occurring at the probe field frequency has the same (zero degree lag) phase with that of the probe field.

Principle of stimulated Raman scattering microscopy. (a) Energy diagram of stimulated Raman scattering when the energy difference between pump and probe is resonant with a vibrational transition. Stimulated Raman gain of the probe beam and stimulated Raman loss of the pump beam after interacting with the vibrational oscillators are depicted too. (b) Recorded spectra of the 1595 cm⁻¹ Raman peak of 10 mM retinol in ethanol by spontaneous Raman, CARS and SRS. While the distorted CARS spectrum exhibits a typical peak shift, dispersive shape and nonresonant background, SRS spectrum is identical to that of spontaneous Raman. (c) Linear dependence of SRS signal on concentrations of retinol in ethanol at 1595 cm⁻¹. Modulation depth ΔI_p/I_p <10⁻⁷ can be detected. The detection limit was determined to be 50 μM.

Tissue imaging by SRS microscopy. Distributions of (a) topically applied compound retinoic acid and (b) penetration enhancer dimethyl sulfoxide (DMSO) in a mouse ear skin. These images were acquired when tuned into the Raman shifts (c) of retinoic acid at 1570 cm⁻¹ (blue) and DMSO at 670 cm⁻¹ (green). Skin structures are also highlighted by tuning into the CH₂ stretching vibration at 2845 cm⁻¹ (red). Adapted from Reference (67). (d) A sebaceous gland embedded in a mouse ear imaged at three different Raman shifts corresponding to lipid CH₂, water OH and protein CH₃. The arrows indicate a hair whose keratin is seen the CH₃ image and oil coating in the CH₂ image. The subcellular resolution reveals the water-containing and lipid-deprived nuclei with reverse contrast.

Stimulated emission microscopy. (a) Energy diagram of stimulated emission. (b) A pair of SEM images of toluidine blue O, a drug used as photosensitizer in photodynamic therapy, at two different z-depths (3 and 25 μm, respectively), delivered onto a mouse ear. Optical sectioning is evident. (c) SEM images of genetically encoded non-fluorescent chromoproteins, gtCP and cjBlue, respectively, inside E. coli cells that contain corresponding expression plasmids. Images adapted from Reference (83).

Ground state depletion micro-spectroscopy of single molecules. (a) Energy diagram of ground state depletion. (b) Ground state depletion signal as a function of concentration of aqueous Atto647N solution. The blue inset indicates the data points at lowest concentrations, with estimated mean molecule numbers in the probe volume. Error bars are for 1s integration time, indicating that single-molecule sensitivity is reachable. (c) Simultaneous fluorescence and ground state depletion line scans for a single Atto647N molecule embedded in PMMA film, averaged before (red) and after (blue) photobleaching. The inset shows the one-dimensional fluorescence image constructed from repeated line scans across the molecule, which underwent abrupt single step photobleaching after 45 lines. Figure adapted from Reference (88).

Principle of nonlinear dissipation microscopy and pump-probe microscopy in which a high-frequency modulation transfer scheme is utilized. (a) The generic experimental scheme. Both pump and probe beams are focused onto a common focal spot with a microscope objective. The intensity (or frequency, polarization, phase, etc) of the pump beam is modulated at a high frequency (>1 MHz), and probe beam after interacting with the sample is collected and detected by a photodiode and then demodulated by a lock-in amplifier. (b) Temporal modulation behaviors of the input and output pump and probe pulse trains before and after interacting with the samples. The probe beam could undergo either a gain or a loss in its intensity. (c) Noise spectrum (log-log plot) of a typical laser source as a function of frequency f. In the low frequency range (from DC to kHz), the noise follows the so-called 1/f noise. In the higher frequency, the noise approaches the flat floor of shot noise.

Comparison between CARS and SRS imaging. (a) The theoretical CARS spectrum resulted from interference between non-resonant background and real part of vibrational resonant contribution. (b)-(d) are forward CARS images of 3T3-L1 cells tuned across the C-H resonance. (b) cell imaged at C-H off resonant condition (2086 cm⁻¹); (c) cell imaged at C-H resonant condition (2845 cm⁻¹); (d) cell imaged at the blue dip of the C-H band at 2950 cm⁻¹. Resonant features appear dark against the non-resonant background; (b)-(d) adapted from Reference (31). Simultaneous (e) epi-CARS and (f) SRS images of a live worm, C. elegans, with the Raman shift being set to the lipid band 2845 cm⁻¹. While SRS specifically probes the lipid contribution, CARS contrast is evidently complicated by non-resonant background from non-lipid structures. Simultaneous (g) epi-CARS and (h) SRS images of a layer of 2 μm polystyrene beads spin-coated on a glass coverslip, with Raman shift being at 2845 cm⁻¹. While SRS shows well-behaved round disks for single beads, the corresponding CARS images show a bright ring due to the interference effect occurring at the edge and a bright spot at the center due to the forward going CARS signal being reflected back by the bead/air interface.

SRS imaging of live cells at various spectral regions. (a) SRS image and (b) optical transmission microscope image of an unstained tobacco BY2 cultured cell with the Raman shift being set to 2967 cm⁻¹. The nucleus and cell walls of a tobacco BY2 cultured cell are clearly visualized. Figure adapted from Reference (69). (c) SRS image of unstained human HL60 cells in an aqueous environment with the corresponding Raman shift being 1659 cm⁻¹ on resonance with the C=C stretching vibrations. Figure adapted from Reference (68). (d) and (e) SRS images of a human lung cancer cell incubated with omega-3 fatty acids at 2920 cm⁻¹ and 3015 cm⁻¹, respectively. A clear distinction of saturated and unsaturated lipid distributions is evident. (e) Spontaneous Raman spectrum of oleic acid (with single double C=C bond) and docosahexaenoic acid (with six double C=C bond). The strong peak at 3015cm-1 is characteristic of unsaturated fatty acids. Figures (d)-(f) adapted from Reference (67). (g) Human embryonic kidney cells in metaphase, imaged at three different Raman shifts corresponding to DNA (1090~1140 cm⁻¹), protein (1650 cm⁻¹) and lipids (2845 cm⁻¹), respectively.

Two-photon absorption microscopy. (a) Energy diagram of simultaneous two-photon absorption by a high-lying electronic state through an intermediate virtual state. (b) 3D volume rendering of two-photon absorption signal from human melanoma lesions obtained with femtosecond pulse trains of two different colors. Image adapted from Reference (78). (c) Two photon absorption image of microcapillaries in a sebaceous gland of mouse skin with contrast due to hemoglobin in red blood cells (red). Overlaid are lipid (green) and protein (blue) SRS images, taken with the same picosecond pulse trains, at corresponding Raman shifts, showing lipid-rich gland cells and adipocytes as well as protein-rich structures such as hairs and collagen, respectively.

Excited state absorption microscopy. (a) Energy diagram of sequential two-photon absorption via an intermediate electronic energy state. (b) Bright field image and a series of laser scanning two-color excited-state absorption images from blood at various depths in a mouse ear. Figure adapted from Reference (85).