STED microscopes are commonly built using separate optical paths for the excitation and the STED beam. As a result, the beams must be co-aligned and can be subject to mechanical drift. Here, we present a single-path STED microscope whose beams are aligned by design and hence is insensitive to mechanical drift. The design of a phase plate is described which selectively modulates the STED beam but leaves the excitation beam unaffected. The performance of the single-beam setup is on par with previous dual-beam designs.
©2009 Optical Society of America
Fluorescence microscopy is among the most frequently used methods in the biosciences. It allows to image specimen under ambient conditions in two- or three-dimensions and at high speed. Since the introduction of stimulated emission depletion (STED) microscopy [1, 2] its most limiting drawback – the finite spatial resolution – has vanished. Once a privilege of electron and scanning probe instruments, resolutions of 25 nm and less can be routinely achieved with modern far-field optical microscopes [3–5].
In a typical beam scanning STED microscope, molecules excited by a focused excitation beam are instantly sent back to the electronic ground state such that fluorescence can originate from the focal center only [1–5]. To this end, the focal spot of excitation light is superimposed with the doughnut-shaped spot of a second (STED) beam at a wavelength which does not excite the molecules but, rather, efficiently induces stimulated emission. Any fluorescent molecule subject to a STED beam intensity I ≫ Is is incapable to fluoresce due to its instant quenching by stimulated emission of a photon. I S is a characteristic intensity at which a fraction 1/e of the molecules are deactivated (i.e. switched off) by the STED beam (I S ≈10–30 MW/cm2). Since the STED wavelength is spectrally excluded from detection neither the bright STED beam, nor the few stimulated photons are detected. To be spatially selective the STED beam is focused such that it features a zero-intensity minimum in the center (e.g. in the shape of a torus). The diameter d around the focal center within which molecules remain fluorescent defines the resolution and is approximately given by [3,6]
In (1), I represents the maximum intensity of the STED beam, i.e. the intensity at the doughnut crest. The image is recorded by moving both beams simultaneously through the sample and detecting the fluorescence. This directly renders, without further processing, a high-resolution image of the structures.
The wider use of STED microscopy has recently been stimulated by the introduction of a simplified STED microscope [7, 8] which employs a single turn-key supercontinuum laser source instead of a complex and difficult to operate laser system. The microscope is capable to deliver equal if not superior spatial resolution compared to a system using a mode-locked Ti:sapphire master oscillator, a regenerative amplifier, and an optic-parametric amplifier . In addition, the microscope can be configured for high-resolution 3D imaging . While the introduction of a single, inexpensive laser source has increased the acceptance of STED microscopy the layout of the microscope unit itself has not changed significantly. In particular, the excitation and STED beams still require alignment in all three directions as well as in time. Therefore, experience in experimental optics is advisable to assemble a STED microscope and – unless automated – to perform occasional realignment of the beams during operation.
Here we report on a new STED microscope which entirely eliminates one beam path and which is aligned by design. The proposed STED setup sketched in Fig. 1 shares many components with a standard scanning (confocal) microscope: the combined excitation and the STED beams are coupled into a single polarization preserving single-mode optical fiber whose output is collimated and directed to the objective lens. Since both laser beams travel the same optical path right from the exit of the optical fiber they are automatically focused to the same position in the sample. Only chromatic errors in the lenses can lead to small focal offsets. Nevertheless, the common optical path ensures that their relative position is insensitive to mechanical drift in the setup.
The fundamental challenge in the implementation of this single-beam STED microscope is to design a phase filter which acts on the STED beam but leaves the excitation beam unaffected.
One approach based on diffractive optical elements (DOE) was suggested for and applied in absorbance-modulation lithography [9, 10]. However, the concept was only demonstrated at relatively low numerical apertures (NA ≤ 0.83) and the wide diameter of the doughnut mode (800 nm at λ = 532 nm) afforded a resolution increase of only 20%. Alternatively, the polarization dependence of birefringent materials or liquid crystal spatial light modulators (LC-SLM) might be used to modulate only one of two orthogonally polarized laser beams. However, as SLMs introduce a fair amount of complexity and the selection of birefringent materials is rather limited and expensive, we decided to pursue an approach based on the spectral dispersion properties of different optical materials.
2. Phase filter design
The conceived phase filter gains its wavelength-selective effect by a careful combination of two optical media whose refractive indices are matched at the excitation wavelength λexc but which are notably different at the STED wavelength λSTED . As illustrated in Fig. 2 a fictitious phase plate of two materials A and B for which n A(λexc) = n B(λexc) but n A(λSTED) < n B(λSTED) appears as a homogeneous plane parallel plate to the excitation beam (yellow) but introduces a step onto the wavefront of the STED beam (red). The induced phase shift ϕ is a function of the geometrical path length d along which different parts of the wavefront cross different materials and of the refractive index difference Δn at the STED wavelength λSTED:
In the design process we first considered the combination of a glass and an optical fluid which, due to significantly different dispersion properties, allows a quite large refractive index difference Δn to be realized at the STED wavelength. As the temperature coefficient of the refractive index (dn/dT) of optical fluids (as e. g. sold by Cargille Laboratories) is typically 2–3 orders of magnitude higher than for optical glasses, the temperature can be used as an additional tuning parameter to adjust the dispersion properties of the phase filter. However, the temperature sensitivity also requires that the temperature be homogenously controlled to about 0.1 K across the whole aperture which proved rather impractical. Therefore, we did not further pursue the approach of a solid-liquid phase filter but set out to identify suitable combinations of solid media. To this end, we wrote a MATLAB script (The Mathworks, Inc., Natick, MA, USA) which, by using the Sellmeier equation, calculates the dispersion curves for all optical glasses available from SCHOTT AG (Mainz, Germany):
The dispersion coefficients B 1–B 3 and C 1–C 3 are tabulated for a temperature of 20 °C in the Optical Glass Datasheets provided by SCHOTT. Because our phase plate design temperature (21 °C) is close to the reference temperature and dn/dT ≈10−6 K–1 a temperature correction to n(λ) was not necessary. We specified a wavelength interval for fluorescence excitation (typically 20–30 nm around the absorption maximum of the dye to be used) and, for any combination of two optical glasses, checked whether their dispersion curves intersect within this interval. For the pairs which fulfilled this condition we calculated the difference in refractive index Δn(λSTED) at the design STED wavelength λSTED. For example, the search with an excitation wavelength window of 625–635 nm and a STED wavelength of 750 nm gave N-KZFS4 and N-SK4 as a suitable pair. Both glasses have n(630.5 nm) = 1.61062 but their refractive indices differ by Δn(750 nm) = 1.24×10−3 at the STED wavelength (see Fig. 3a ). The screening was repeated for different wavelength combinations which were chosen to match the properties of some common fluorescent dyes. The results are summarized in Tab. 1.
The next step was to identify a preferable phase filter geometry for the STED beam. It has been shown that the optimal phase filter  for focal (xy) plane resolution enhancement is a charge-1 vortex which is characterized by a complex transmission function p(r, φ) = exp(iφ), i.e. a linear increase in phase from 0 to 2π with increasing azimuthal angle φ (Fig. 4a ). Here, (r, φ) are the transverse cylindrical coordinates. Upon focusing, a circularly polarized beam prepared in this way destructively interferes in the focal center and generates a zero-intensity minimum which is surrounded by a torus of high intensity [13, 12]. With currently available laser sources which provide 10–20 nJ laser pulses this torus of STED light confines the fluorescence emission to a spot with lateral (x, y) dimensions of ~20 nm; since no quenching occurs along the optical axis the extent in z remains at the confocal level of ~600 nm. Commercial vortex phase filters are produced by stamping an array of helical patterns into a thin polymer film on a glass substrate where each of the vortices is designed for a specific wavelength. The resulting optical element can be easily embodied into a STED microscope and has made this configuration the most frequently used one in STED microscopy.
The manufacture of a vortex phase filter from two solid media is significantly more challenging due to the requirement that two pieces with identical spiral surface reliefs must be produced and accurately bonded together. While the relatively small refractive index difference at the STED wavelength [Δn(λSTED) ≈10−3] entails a macroscopic slope of the spiral on the order 0.5–3° the 0/2π discontinuity and the singularity in the center of the vortex render the production with conventional methods highly difficult. As a simplification, the continuous spiral can be discretized into segments of constant phase. A version consisting of six segments already approximates the vortex very well, see the resulting PSF shown in the inset of Fig. 4a. Yet, the production of twelve individual segments with well-defined heights and their assembly is nontrivial and also introduces many glued surfaces which may deteriorate the performance of the phase filter.
Finally, we decided to implement a phase filter which consists of two antiparallel phase gradients (0…π and π…2π) in a side by side arrangement [14–16] (Fig. 4b). The PSF due to this filter is very similar to that of the vortex the only difference being a slight intensity modulation along the ridge of the torus (Fig. 4b). The filter can be manufactured from four wedges made of two optical glasses which are cemented together as illustrated in Fig. 5a . The wedge angle α is derived by first calculating the geometrical path length h at which a phase shift of ϕ is accumulated:Fig. 5b. For Δn(λSTED) ≈10−3 the wedge angle amounts to 0.5–3° which can be fabricated with sufficient accuracy. The exact wedge angles were calculated for the glass combinations found in the previous screening and for a NA 1.4 oil objective lens (PL APO 100×/1.40–0.7 OIL, Leica Microsystems, Wetzlar, Germany) with a back aperture of d = 5.6 mm. The results are summarized in Tab. 1.
3. Results and discussion
For the experimental verification of our concept we custom-produced two of the phase filters listed in Tab. 1 (#3, #6). The attained accuracy in terms of the wedge angle was on the order of a few arcsecs, corresponding to a phase error of approx. 2–5%. High precision was also required in the cementing of the wedges as any deviation from a parallel mounting translates into a parasitic phase gradient, in particular as the refractive index between the glasses and the adhesive is large (Δn ≈0.1) compared to the difference between the two glasses (Δn ≈10−3).
To test the wavelength-selective phase filter, we set up a single path STED microscope using a supercontinuum laser source which provided both the excitation and the STED wavelengths, similar to the setup described previously [7, 8]. However, both beams were coupled into the same polarization preserving single mode fiber, (see Fig. 1). According to the manufacturer datasheets, these fibers support single-mode operation over an extended range and we were able to identify suitable fibers for both wavelengths pairs employed (631 nm and 750 nm: PM-630-HP, Thorlabs GmbH, Dachau, Germany; 569 nm and 650 nm: PMC-532-4-NA012-3-OPC-200-P, Schäfter & Kirchhoff GmbH, Hamburg, Germany). The fiber output was collimated and was sent to the objective lens. In order to avoid transverse chromatic aberrations particular care was taken to meet the collimation lens and the objective lens perpendicularly and in the center. The fluorescence collected from the sample was separated from the laser beams with a dichroic mirror and was detected with a confocalized detection unit. An additional nonconfocal detector (not shown in Fig. 1) was used to probe the point-spread functions by scanning gold nanoparticles (80 nm) in scattering mode.
We started our experiments with a phase filter produced from N-KZFS4 and N-SK4 (SCHOTT AG, Mainz, Germany) with a wedge angle of 3.095° (Tab. 1, #6) which was designed for an excitation wavelength of 630.5 nm and a STED wavelength of 750.0 nm. We first characterized the point-spread functions of the excitation and the STED beams and checked whether they were properly aligned. To this end, we measured the PSFs by scanning a gold nanoparticle first with the excitation and then with the STED beam. The results shown in Fig. 6 confirm that the two PSFs are perfectly aligned in the focal plane (as expected) whereas they are offset by 104 nm along the optical axis (z). This mismatch was anticipated due to the longitudinal chromatic aberrations introduced by the collimation lens and the objective lens. In fact, the measured value is in good agreement with a calculation which yields Δz≈90–115 nm assuming f 630nm = 40.272 mm and f 750nm = 40.341 mm (manufacturer data) for the collimation doublet lens and Δf 630nm/750nm ≈50–75 nm for the objective lens. Since the PSFs are quite extended in the z direction this ~10% offset does not significantly affect the STED measurements.
Besides the axial shift, the torus is also modulated along its ridge by 48% which is significantly stronger than predicted by the calculation (8%). This modulation can be explained in terms of an astigmatism introduced by imperfect optical surfaces or by the polarization maintaining optical fiber . The modulation is equally visible when a vortex phase plate is used, see Fig. 3 in . Due to the modulation the spatial resolution attained in the focal plane is not equal in all directions within the focal plane but, based on the resolution formula for STED microscopy [3, 6], a factor of in resolution can be expected between the directions of highest and lowest STED intensity.
We repeated the PSF measurements with phase plate #3 (N-PSK53A / N-SSK8, α = 3.464°) and found similar results as in the previous experiments: the excitation and the STED PSFs were well aligned in space but the torus is significantly stronger modulated than predicted by theory.
Next, we examined the robustness of our single-beam configuration with respect to thermally induced mechanical drift. In order to be able to compare the drift to the conventional two-beam setup we added a reference laser beam which was co-aligned with the combined excitation and STED beams but which originated from a separate optical fiber, like in conventional STED setups. The setup was further enclosed in a box which could be heated or cooled. All three foci were initially aligned on top of each other and lateral (xy) sections of all three PSFs were acquired (Fig. 7a ). Then, the air in the enclosed setup was either heated or cooled by ~20 °C, the heat source / sink was removed and the setup was allowed to settle for ~15 min. Afterwards the PSFs were acquired again. While after heating and cooling the reference focus was offset by 140 nm and 233 nm with respect to the combined excitation and STED beams, respectively, the latter two remained perfectly on top of each other (Fig. 7b, c). While, even after these drastic temperature changes, the single-beam setup remains operational without any realignment, a slight deformation of the STED focus is noticeable indicating a shift of the combined beams on the back aperture of the objective lens.
Finally, we applied the single-beam setup to the nanoscale imaging of the microtubular network of mammalian PtK2 cells in our single-beam setup. The cells were stained using an immunofluorescence protocol involving a primary antibody [anti β-tubulin mouse IgG (monoclonal), Sigma, Saint Louis, USA] and a secondary antibody (sheep anti-mouse IgG, Dianova, Hamburg) labeled with the rhodamine dye KK114. The excitation and STED wavelengths were (630±5) nm and (750±15) nm, respectively; the repetition rate was 1 MHz. Excitation was performed with time-averaged powers between 160 nW (confocal measurements) and 530 nW (STED measurements) whereas the STED beam had 2.3 mW of optical power at the back aperture of the objective lens, corresponding to a pulse energy of 2.3 nJ. For comparison, the imaging was performed in STED and in confocal mode (Fig. 8a, b ). As can be seen in the figure, the STED image resolves single tubulin fibers down to a separation of ~80 nm whereas bundles of fibers are completely blurred in the confocal image. The smallest fiber diameters observed are ~60 nm which is in accordance with previous experiments using a vortex phase plate and comparable pulse energies  and with expectations considering the fiber diameter and the extent of the primary and secondary antibody sandwich. The magnification shown in Fig. 8c even discloses a discontinuous staining as evidenced by the dot-like structures.
We presented a new STED microscope featuring a common beam path for the excitation and the STED beam. Hence, the two foci are spatially aligned by design and their relative position is insensitive to thermal drift in the setup. The common beam path was made possible by the development of a wavelength-selective phase plate which acts on the wavefront of the STED beam but not on the excitation beam. While alternative approaches are conceivable, our design rests on the use of two optical media which are index-matched at the excitation wavelength but notably differ in refractive index at the STED wavelength. We verified our concept experimentally and found comparable performance as for setups with separate beams. The setup is mostly immune to thermal drift even under drastic (> 20 °C) temperature changes.
In the future, the concept introduced in this work could be further refined. It would be desirable, for example, to use the optical fiber not only for laser beam delivery but also as a confocal pinhole for fluorescence collection which would entirely eliminate any alignment requirements in the microscope unit. Comprising very few optical components, such a microscope head could possibly be implemented as an endoscope simplifying in vivo imaging at high resolution.
The concept may also be modified to find material combinations which allow the construction of an achromatic phase plate for an extended wavelength region. These could be used to implement a two-color STED microscope with a common beam path for both STED beams. Such a configuration would enable reliable high-resolution colocalization measurements over extended periods of time which are not compromised by thermal drift.
We are confident that the ongoing simplifications in instrumentation will allow a growing number of researchers to take advantage of STED microscopy in their applications.
We thank M. Kleinhans and W. Kluge for producing the phase plates. The dye KK114 was kindly provided by V. Below. The research leading to these results has received funding from the European Community's Seventh Framework Programme FP7/2007-2011 under grant agreement no. 201837. D. W. acknowledges a doctoral fellowship by the German National Academic Foundation.
References and links
1. S. W. Hell and J. Wichmann, “Breaking the diffraction resolution limit by stimulated emission: stimulated emission-depletion fluorescence microscopy,” Opt. Lett. 19(11), 780–782 (1994). [CrossRef] [PubMed]
2. T. A. Klar, S. Jakobs, M. Dyba, A. Egner, and S. W. Hell, “Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission,” Proc. Natl. Acad. Sci. U.S.A. 97(15), 8206–8210 (2000). [CrossRef] [PubMed]
4. G. Donnert, J. Keller, R. Medda, M. A. Andrei, S. O. Rizzoli, R. Lührmann, R. Jahn, C. Eggeling, and S. W. Hell, “Macromolecular-scale resolution in biological fluorescence microscopy,” Proc. Natl. Acad. Sci. U.S.A. 103(31), 11440–11445 (2006). [CrossRef] [PubMed]
10. R. Menon, P. Rogge, and H.-Y. Tsai, “Design of diffractive lenses that generate optical nulls without phase singularities,” J. Opt. Soc. Am. A 26(2), 297–304 (2009). [CrossRef]
11. L. Kastrup, and V. Westphal, “Wavelength or polarisation sensitive optical assembly and use thereof,” German Patent DE102007025688A1, 2007/06/01.
13. V. V. Kotlyar, A. A. Almazov, S. N. Khonina, V. A. Soifer, H. Elfstrom, and J. Turunen, “Generation of phase singularity through diffracting a plane or Gaussian beam by a spiral phase plate,” J. Opt. Soc. Am. A 22(5), 849–861 (2005). [CrossRef]
14. G.-H. Kim, J.-H. Jeon, K.-H. Ko, H.-J. Moon, J.-H. Lee, and J.-S. Chang, “Optical vortices produced with a nonspiral phase plate,” Appl. Opt. 36(33), 8614–8621 (1997). [CrossRef]
15. V. G. Shvedov, Y. V. Izdebskaya, A. N. Alekseev, and A. V. Volyar, “The formation of optical vortices in the course of light diffraction on a dielectric wedge,” Tech. Phys. Lett. 28(3), 256–259 (2002). [CrossRef]
16. X.-C. Yuan, B. P. S. Ahluwalia, S. H. Tao, W. C. Cheong, L. S. Zhang, J. Lin, J. Bu, and R. E. Burge, “Wavelength-scalable micro-fabricated wedge for generation of optical vortex beam in optical manipulation,” Appl. Phys. B 86(2), 209–213 (2007). [CrossRef]