1. Introduction

Stimulated emission depletion (STED) microscopy is a powerful tool for the sub-diffraction visualization of fluorescent structures in the far field [1], able to make observations noninvasively on the nanoscale to explore subjects as diverse as endoplasmic reticulum dynamics in living cells [2] and the properties of single color-center defects in bulk diamonds [3]. Using standard microscope optics, an effective detection focal spot far below the diffraction limit is obtained by switching off the ability of fluorophores to emit fluorescence at the periphery of the excitation focal spot with a beam prohibiting the occupation of the fluorophore’s excited state by stimulated emission. To achieve this, the focal intensity distribution of the stimulating light is rendered to have zero intensity in the center so the de-excitation is saturated at the periphery and the remaining fluorescent spot is reduced to a size many times smaller than the limit imposed by diffraction. Scanning the sample yields a super-resolution image.

However, the application of STED is limited by the availability of light sources capable of saturating the stimulated emission transition within the fluorescence lifetime of the marker, typically 1-5 ns. Within a fraction of this fluorescence lifetime, enough photons must impinge upon the sample to ensure that there is a photon interacting with the excited fluorophore, sending it to the ground state via stimulated emission. So far, STED imaging has been successfully demonstrated using titanium sapphire [4,5], diode [6], continuous wave [7], and supercontinuum laser sources [8], but each source has inherent disadvantages. Femtosecond titanium sapphire sources require pulse stretching, and if they are to yield light in the visible, are used in conjunction with optical parametric amplifiers or regenerative amplifiers, resulting in complex, expensive setups. Diode sources, though convenient, lack sufficient power to reach high resolutions in multiple spatial directions simultaneously. Continuous wave sources can be used to de-excite dyes which exhibit little dark-state buildup and associated bleaching, but may ultimately not reach the resolution provided by their pulsed counterparts. Supercontinuum pulsed sources are attractive for their spectral flexibility but so far have been limited in repetition rate to several MHz and current commercial versions do not produce light of sufficient intensity for STED below 630 nm.

Recently an additional light source, generating a comb spectrum via stimulated Raman scattering (SRS) within a length of standard, polarization-maintaining, single mode fiber was shown to produce light suitable for STED imaging at multiple wavelengths red-shifted from 532 nm [9]. Here we demonstrate a significant improvement over this previous work by employing a different laser pump source. While the previous system produced STED light near 600 nm with a repetition rate of ~20 kHz, measurements with the current system were performed at 1 MHz, increasing the data acquisition rate by a factor of 50. This brings the source into a regime of practical applicability, enabling imaging at speeds comparable to current supercontinuum sources. Additionally, the current pump laser exhibits pulse energy fluctuations of 3σ <10%, much reduced over the pulse energy fluctuations of 50% encountered with the previous pump source, while substantially reducing temporal pulse jitter (~1 ns, in contrast to ~1 µs), and increasing the available pump pulse energy from 0.5 µJ to 10 µJ.

SRS in optical fibers is described in detail elsewhere [10–13]. At a threshold optical intensity in the fiber an appreciable fraction of the photons in the pump beam, with energy ħω_p, begin to scatter inelastically, resulting in Stokes-shifted photons of lower energy ħω_s. The Raman gain spectrum for the propagation medium, g_R(Ω), determines the degree of the Stokes shift, Ω = ω_p - ω_s. For fused silica g_R(Ω) exhibits a maximum near 13 THz, corresponding to a Stokes shift from 530 nm to 543 nm. When a generated Stokes line reaches the threshold intensity for SRS, it becomes a pump source for an additional line. Repeated, this process generates the characteristic comb spectrum. Fiber lengths up to the walk-off length for the employed pulses increase the gain at the Stokes wavelengths until the effects of power depletion of the pump wavelength or fiber losses at the pump and Stokes wavelengths become appreciable.

2. Setup

To test the viability of the light source for obtaining subdiffraction resolution it was incorporated into a STED microscope. A schematic of the setup is show in Fig. 1 (a) . The pump laser for SRS was based on a pulsed master-oscillator, fiber power-amplifier architecture and internally frequency doubled to produce the 530 nm output with 10 µJ pulses available (at maximum power) at 1 MHz, 2 ns in length and with < 0.5 nm spectral bandwidth (G1 + Laser System, Mobius Photonics, Santa Clara, USA) [14,15]. The output laser power could be controlled electronically, but to increase power attenuation below the minimum power level and confer further flexibility to the setup a half wave plate followed by a polarizing beam splitter was included directly after the output. The beam was then directed into a 20 × microscope objective (Olympus, Tokyo, Japan) for launching into the fiber. The fiber used was specified by the manufacturer as having a mode field diameter of 4.1 µm at 630 nm and a nominal 530 nm cutoff wavelength (Fujikura, Tokyo, Japan). With a launch efficiency of 50% at 5% maximum 530 nm pump power, peak optical intensities on the order of 130 MW/cm² were present in the fiber at 530 nm before the transfer of the pump power to Stokes lines. Due to color center formation after the exposure to high-power 530 nm light, however, the launch efficiency soon fell to 20%, where it remained stable for long term operation, equating to optical intensities of 50 MW/cm² in the core. The walk-off length for neighboring Stokes lines was greater than 100 m, allowing each spectral line to pump its Stokes counterpart over the entire length of fiber.

 

Fig. 1 (a) Schematic of setup used for measurements. (b) Measured spectra produced in the standard, single mode, polarization maintaining fiber by the 530 nm pump laser at three different powers (before coupling into the fiber): 360 mW (top), 240 mW (center), and 200 mW (bottom). In each case pump power is chosen to maximize the output power of the last line in the spectrum.

Download Full Size | PDF

An attractive tunability was conferred to the system by the cascaded nature of the spectrum generation. By changing the power of the 530 nm pump source we were able to alter the comb spectrum to maximize power present in a given Stokes line, effectively transferring a larger fraction of the pump power to the desired wavelength. Spectra produced at different pump powers chosen to maximize power in the 585 nm, 600 nm, and 616 nm lines are shown in Fig. 1 (b). By avoiding a transfer power to higher Stokes lines not currently in use, higher optical powers could be obtained at a given wavelength than if the spectrum were non-tunable. Pulse energies up to 50 nJ were observed exiting the fiber within the last spectral line in lines near 600 nm. The pulse length for the SRS light remained similar to that of the pump source, ~2 ns. Up to 616 nm the bandwidth remained < 2 nm, but it increased slightly with each successive line from the < 0.5 nm bandwidth of the pump beam.

Light from the fiber was collimated and the appropriate STED wavelength for the fluorophore marking the sample was selected with a bandpass filter (AHF Analysentechnik, Tübingen, Germany). Light was then directed through a polymeric helical phase delay plate (RPC Photonics, Rochester, USA) and reflected by a dichroic mirror (AHF Analysentechnik) through a quarter waveplate into an 100 × , oil immersion microscope objective (NA 1.4 UPlanSApo, Olympus, Tokyo, Japan), where the helical phase imprint from the phaseplate and the circular polarization yielded the donut PSF in the focal plane.

Excitation light was emitted by a 488 nm diode laser (PicoQuant, Berlin, Germany), with 100 ps pulse length. A small fraction of the 530 nm pump light was redirected by a pellicle onto a photodiode before the SRS fiber to trigger the excitation pulses, with the SRS fiber, electronic delay cables, and an electronic timing box ensuring that the excitation and STED pulses were temporally synchronized in the sample. Fluorescence light emitted by the sample was gathered and collimated by the objective lens, directed through the two dichroic mirrors and focused into a multimode fiber, which, acting as a confocal pinhole selecting 90% of the Airy-disk at 580 nm, guided light into an avalanche photodiode (id Quantique, Carouge, Switzerland) for fluorescence detection. During data acquisition the sample was scanned through the stationary focus via a piezo stage (Nano Block, Melles Griot, Albuquerque, USA), and the image was assembled with in-house image processing software.

3. Results

The STED imaging capabilities of the setup were tested using the 585 nm, 600 nm, and 616 nm Raman lines as STED light. Both yellow-green fluorescent beads (Invitrogen, Carlsbad, USA) and PtK2 cells with vimentin fibers, constituents of the cytoskeleton, tagged

with Chromeo 488 (Chromeon, Tegernheim, Germany) via secondary antibody labeling were used as samples for imaging. STED pulse energies in the back aperture of the objective lens varied from 2 nJ to 8 nJ depending on the STED wavelength employed (larger STED powers were typically used with redder depletion light because the cross section for stimulated emission was smaller at these wavelengths, necessitating more photons in the sample to achieve the same probability of de-excitation). Measurements are show in Figs. 2 , 3 , and 4 . Pixel sizes of 10 nm were used for bead samples while 20 nm pixels were used for the biological images.

 

Fig. 2 Raw STED images using 585 nm with confocal images (far left) for comparison. (a) 43 nm yellow-green fluorescent beads. Images are 4 µm × 4 µm and enlarged areas have dimensions of 1 µm × 1 µm, with an intensity line profile taken at the image position specified by the arrows. (b) Images in both confocal and STED mode of vimentin fibers in PtK2 cells stained with Chromeo 488 via secondary antibody labeling. Images are 10 µm × 10 µm, enlarged areas have dimensions of 3 µm × 3 µm. All images show the complete dynamic range present within each image pane. Photon counts are shown in the color scale in the bottom right of each image pane.

Download Full Size | PDF

 

Fig. 3 Raw STED images using 600 nm with confocal images (far left) for comparison. (a) 43 nm yellow-green fluorescent beads. (b) Vimentin fibers in PtK2 cells stained with Chromeo 488. Images and enlargements have the same dimensions as Fig. 2 showing the complete dynamic range of each image pane.

Download Full Size | PDF

 

Fig. 4 Raw STED images using 616 nm with confocal images (far left) for comparison. (a) 43 nm yellow-green fluorescent beads. (b) Vimentin fibers in PtK2 cells stained with Chromeo 488. Images and enlargements have the same dimensions as Fig. 2, showing the complete dynamic range of each image pane.

Download Full Size | PDF

Pixel dwell times for image acquisition ranged from 0.5 to 1 ms, and total acquisition time for a 4 × 4 µm image with a 0.5 ms pixel dwell time and 10 nm pixels was ~6 minutes. Resolution was determined by superimposing the images of beads from a single recording, aligning their centers, adding their signals and performing a Gaussian fit in the x, y, and diagonal orientations in the summed image. Data used for analysis at each wavelength are shown in Figs. 2-4. 90 or more beads were sampled from a single image by the analysis program, excluding those which were too close to neighbors to be resolved or fitted, or appeared to be multiple beads in an agglomeration. The beads were specified by the manufacturer as having a mean diameter of 43 nm with a standard deviation of 6 nm. Approximating the object, effective point spread function, and resulting image intensity profiles as Gaussians, the full-width-half-maximum (FWHM) of the effective point spread function was determined to be between 20 and 30 nm for all bead images using all STED wavelengths.

FWHM intensity line profiles in the PtK2 vimentin fiber images of 60-100 nm were found in orientations perpendicular to fiber direction in each image. The images shown in Figs. 2-4 were used for analysis. Based on the fact that the minimum diameter of the vimentin fibers, together with secondary antibody labeling and fluorophore markers, was 50 nm and making the above approximation of line profiles being Gaussian, the effective point spread function for these measurements was estimated to have a FWHM value of ~40 nm. We note that the obtained resolution is exclusively due to the molecular optical transitions (i.e. due to

involved physical processes); the resolution can be further enhanced by mathematical deconvolution algorithms which were not applied here.

4. Conclusion

A light source with pulses in the 0.1 ns to 1 ns range with energies on the order of 10 nJ offering a broad spectrum of light in the visible at a repetition rate in the tens of megahertz would be extremely attractive for STED imaging, conferring the flexibility to use fluorescent markers of choice in the visible range with attractive data acquisition speeds. By demonstrating that light created with SRS spectral generation in standard fibers can be used successfully for STED at 1 MHz, we show that SRS spectral generation should be viable for achieving still higher repetition rates, and thus serve as a light source which fulfills the above requirements.