Quantum-optimal nonlinear microscopy with classical light

Abstract

Nonlinear optical processes are used in biological microscopy to surpass the diffraction limit on resolution, image deeper into brain tissues, and identify biomolecules without exogenous labels. These techniques typically require high optical intensities to increase the strength of the nonlinear interactions, which can perturb native biochemistry and damage or kill living samples. Stimulated Raman scattering (SRS) microscopy visualizes the spatial distribution of molecules using a nonlinear interaction between light and chemically specific molecular vibrations. However, the detection of biomolecules at low concentrations is limited by the total photon dose that can be applied before photodamage alters the sample, and photon shot noise sets the minimum achievable noise floor for most microscopes. Here we demonstrate a cavity-enhanced SRS microscope that is more sensitive than an equivalent conventional SRS microscope by up to 8.3(7) dB in spectroscopy and 8.6(1) dB in cell imaging. These results approach quantum limits on sensitivity and demonstrate that quantum states of light are sufficient but not necessary to enhance the sensitivity of microscopy techniques that are limited by photodamage.

Figures (9)

• Figure 1: Cavity-enhanced SRS microscopy. (a) Stokes pulses (red) are enhanced in an optical cavity and are spatially and temporally overlapped with pump pulses (blue) at the cavity waist formed between two microscope objectives to drive SRS. The Stokes (pump) beam center wavelength is 1045 (800). Both pulses are temporally stretched to durations of 2.5 with optical gratings. The cavity $f_\text{FSR}$ is stabilized to $f_\text{rep} \sim$80. GS, grating stretcher; EOM, electro-optic modulator; AOM, acousto-optic modulator for amplitude modulation at $f_\text{mod}$; S, translation stage; PZT, mirror mounted on piezoelectric transducer; M$_{1(2)}$, input (output) coupler with intensity transmission $T_{1(2)}$; D, dichroic mirror; C, concave mirror; Sa, sample; D$_\text{P}$, pump detector; D$_\text{PDH}$, PDH detector; D$_\text{S}$, Stokes detector. (b) Information per average dose for SRS signal estimation with round-trip transmission $\eta$, normalized to the quantum limit (black; QL), for classical single-pass (blue; SP); 1, 10, and 20 squeezed state (orange, blue, purple; dashed; SQ); and cavity-enhanced ($T_1 = 0.004$ and $T_2 = 0.104$) (red; C) measurements. (c) Theoretical SNR gain (red; dashed) over an ESM for the cavity-enhanced SRS microscope. High-frequency modulation signals are attenuated by the cavity, as shown for $f_\text{mod}$ = 210 (orange) and 1.01 (purple). Dashed gray lines are the transmissions measured in this work ($\eta = 0.871$, 0.906). See Methods for details.
• Figure 2: Cavity-enhanced SRS spectroscopy of DMSO. (a) Cavity-enhanced (red) and single-pass (purple) spectra of DMSO measured with the same input power at $M_1$ (Fig. \ref{['fig:setup']}). The Stokes dose is 11 for the cavity-enhanced measurement. The measured single-pass signal is magnified by 40$\times$ for clarity. (b) Cavity-enhanced SRS signal amplitude at 2913 (red) and noise power (purple) as a function of the Stokes dose, normalized to measured values for an 11 Stokes dose. Error bars are 95% confidence intervals. Dashed lines are linear fits. (c) ESM SNR (purple) compared to the cavity-enhanced SNRs with (orange) and without (red) electronic noise, normalized to the ESM limit for each Stokes dose. Error bars are 95% confidence intervals on the measured SNRs. Dashed lines guide the eye.
• Figure 3: Cavity-enhanced SRS imaging. HeLa cells imaged under (a) cavity-enhanced and (b) single-pass conditions for the same input power at $M_1$. Pixel values correspond to relative signal amplitudes. Scale bars are 20. (c) Scatter plot of the cavity-enhanced and ESM SNRs for each pixel, calculated after down-sampling each image onto a 1.5 x 1.5 grid. The SNR enhancement is 8.3(8) (red) relative to an ESM (orange).
• Figure 4: Cavity-enhanced hyperspectral imaging. Cavity-enhanced SRS images acquired sequentially for Raman shifts of (a) 2850, (b) 2934, and (c) and 2967 , corresponding to Raman bands associated with lipids, protein, and DNA, respectively. Pixel values correspond to relative signal amplitudes. (d) Hyperspectral image combining the measurements for lipids (red), protein (green), and DNA (blue). Scale bars are 20.
• Figure 5: Extended Data Figure 1 Cavity characterization. (a) Schematic of the experiment in the frequency domain. Each mode of the Stokes beam (spacing $f_\text{rep} \sim$80) is on resonance with a longitudinal cavity mode (free spectral range $f_\text{FSR} \sim$80), offset for clarity. Phase-modulation sidebands (yellow) are placed on each laser mode by an EOM to stabilize the cavity length. Stimulated Raman scattering transfers amplitude-modulation sidebands at $f_\text{mod}$ (blue) from the pump beam to the Stokes beam, and these sidebands are attenuated by the transfer function of the cavity. (b) Cavity transmission with a DMSO sample inserted, measured as the cavity length is scanned with the piezoelectric transducer. (c) Gaussian fits (dashed, black) to transmission lineouts of RPE-1 cells (Extended Data Fig. \ref{['fig:phase']}c). The mean half width at $1/e^2$ for the fitted Gaussians with half widths at $1/e^2 <$2 is 1.4(2), where the uncertainty is the standard deviation of the width measurements.