954 OPTICS LETTERS / Vol. 24, No.

14 / July 15, 1999

Subdiffraction resolution in far-field fluorescence microscopy

Thomas A. Klar and Stefan W. Hell
High Resolution Optical Microscopy Group, Max Planck Institute for Biophysical Chemistry, D-37070 Göttingen, Germany

Received March 17, 1999

We overcame the resolution limit of scanning far-field f luorescence microscopy by disabling the f luorescence
from the outer part of the focal spot. Whereas a near-UV pulse generates a diffraction-limited distribution
of excited molecules, a spatially offset pulse quenches the excited molecules from the outer part of the focus
through stimulated emission. This results in a subdiffraction-sized effective point-spread function. For a 1.4
aperture and a 388-nm excitation wavelength spatial resolution is increased from 150 6 8 nm to 106 6 8 nm
with a single offset beam. Superior lateral resolution is demonstrated by separation of adjacent Pyridine 2
nanocrystals that are otherwise indiscernible.  1999 Optical Society of America
OCIS codes: 000.3110, 180.2520, 350.5730, 320.7150.

Abbe’s work1 on the role of diffraction greatly in- (Mira 900, Coherent) operating at 766 nm. For this
f luenced the art of microscopy in this century and purpose the laser is split into two beams. Whereas
spurred the development of techniques such as elec- the first beam is frequency doubled to 383 nm and used
tron and scanning probe microscopy in which the use for excitation, the fundamental beam is used for stimu-
of light was abandoned.2 Scanning probe microscopy lated emission. We refer to them as the UV and the
also triggered near-f ield optical microscopy,3 in which STED beams, respectively. Both beams are spatially
a sharp tip is used to conf ine the interaction of a filtered, expanded, and recombined at a dichroic mir-
sample with light to a subdiffraction spot. In spite ror that is transmissive at 383 nm and ref lective at
of their success these techniques have not replaced 766 nm. We ensure that both beams have the same
the far-field light microscope as the most widely ap- polarization by introducing a l兾2 plate into the STED
plied microscope in biology. The reason is that, be- beam. The beams illuminate the back aperture of the
sides involving more-sophisticated specimen handling, objective lens with planar wave fronts.
electron and probe microscopes are conf ined mainly to The UV pulses have a duration of ⬃200 fs, which is
the imaging of surfaces. To date, noninvasive imag- 3–10 times shorter than the vibrational relaxation and
ing in three dimensions has been achieved only with fo- 4 orders of magnitude shorter than the f luorescence
cused light.4 Fluorescence is particularly important, relaxation of an organic f luorophore that is adequately
since it allows specif ic observation of many cellular described by a four-level model.6 Within 1 ps after
compartments and proteins. Evidently, far-f ield f luo- excitation, the molecules relax to a low vibrational
rescence microscopy featuring resolution beyond the level of the excited state S1 , which is the actual
diffraction limit would be highly attractive. f luorescent state. The STED pulses are chosen so
In this Letter we report what we believe to be the that they efficiently quench this state by stimulated
first evidence of pronounced lateral resolution beyond
the diffraction limit. This resolution is obtained by
use of the concept of stimulated-emission depletion
(STED).5 In this concept the f luorescent molecules
that are excited by the outer part of the focus are
deprived of their ability to emit a f luorescence photon
by exposure to a second beam that induces stimulated
emission. The role of the stimulated emission is to
force the molecules into the ground state immediately
after they are excited. As a result the remaining
f luorescence stems from a region that is narrower than
the diffraction-limited excitation focus. This effect
is equivalent to narrowing the effective point-spread
function (E-PSF) of the microscope and to a model case
of point-spread function engineering.5
Our experiment (Fig. 1) is based on a UV-confocal
scanning f luorescence microscope that, owing to the
short wavelength and confocality, already has a high Fig. 1. Setup: frequency-doubled (UV) and fundamental
(STED) beams from a Ti:sapphire laser are cleaned, passed
classical resolution. By employing an oil-immersion
through pinholes (PH’s), expanded, and combined at a
lens of 1.4 numerical aperture (Leica 1003 Planapo), dichroic mirror (HT383/ HR766). After the beams are
we also use the largest aperture available. The im- ref lected from mirror HT510–710, they overfill the rear
ages are obtained by scanning of the sample with a aperture of the lens. The STED beam is shifted with
piezo stage. The excitation and the stimulating beams respect to the UV beam by a piezo-movable tube lens.
both originate from a mode-locked Ti:sapphire laser APD, avalanche photodiode.

0146-9592/99/140954-03$15.00/0  1999 Optical Society of America

 July 15, 1999 / Vol. 24, No. 14 / OPTICS LETTERS 955

emission. First, to allow vibrational relaxation into cited 6 ns after quenching. Fast recovery and nonde-
the f luorescent state we use an optical delay that structiveness are strong evidence of cool depletion of
ensures that the STED pulses arrive at the focus the excited state by stimulated emission.5,7
a few picoseconds after the UV pulses. Second, as Next we investigated whether a laterally offset
the excited molecules are quenched into a higher STED beam would reduce the lateral extent of the
vibrational level of the ground state S0 , reexcitation E-PSF of the microscope. The E-PSF is measured with
by the STED pulse is avoided by stretching the STED a subresolution nanocrystal. The resulting images,
pulses to ⬃54 ps, using a grating. This stretching shown in Fig. 3, are 0.68 mm 3 0.68 mm and consist
allows the fast vibrational relaxation drain to dump of 72 3 72 pixels. The pixel dwell times were 2
the molecules into a low vibrational level of S0 . and 4 ms with the STED beam switched off and on,
The passband of the dichroic filter that couples the respectively. In Figs. 3 and 4 a uniform background
two beams into the objective is 510–710 nm, allow- was subtracted and averaging of adjacent pixels was
ing the back-emitted f luorescence to be detected by a performed. With the STED beam switched off, the
counting avalanche photodiode. The wavelengths of microscope featured a confocal E-PSF, as shown in
the UV and the STED pulses as well as the f luo- Fig. 3(a). The FWHM along the prof ile is 150 6 5 nm,
rescence are shown in Fig. 2(b). The power used is which matches well the predicted value of 145 nm
2.0 mW for the UV beam and 28.3 mW for the STED
beam. We note that the excitation intensity is well be-
low saturation and that the STED intensity is in the
benign energy region for biological imaging. Owing
to the high magnification of the objective, we can pre-
cisely displace the STED beam with respect to the UV
beam by scanning a tube lens on a piezo stage.
To investigate the resolution increase and the STED
effect we prepared a sample consisting of randomly
distributed nanocrystals of the f luorophore Pyridine
2, characterized in Figs. 2(a) and 2(b). We prepared
the nanocrystals by diluting a saturated solution in
ethanol by a factor of 10 and spreading a few drops
on the cover glass. Nanocrystals formed during the
evaporation of the ethanol. The nanocrystals were
covered with a glycerol-based mounting medium
[100-mM Tris-HCl (pH 8.5), 9% Mowiol 4-88 (Hoechst),
25% glycerol]. In addition, we prepared a thick,
f luorescent layer by mixing the solution with the
mounting medium in a 1:1 ratio. We then determined
the eff iciency of the stimulated-emission depletion by
spatially and temporally overlapping the UV and the
STED foci in the layer. Through interruption of the
STED beam by a chopper [Fig. 2(c)] we found that
the STED beam was able to reduce the total f luores-
cence to 3% of its maximum value. In a separate
measurement (data not shown) we verified that for
such strong depletion the relationship between the
depletion efficiency and the STED-beam intensity is
nonlinear. This relationship favors a sharp depletion
edge at the outer part of the focus.
As we wanted to rule out the possibility that f luores-
cence reduction is caused by photobleaching or triplet-
state quenching, we examined whether f luorescence
can be immediately restored after depletion. We di-
vided the UV pulse into two pulses, delayed the sec-
ond pulse by ⬃6 ns, and measured the f luorescence
by use of time-correlated single-photon counting. The
solid curve in Fig. 2(d) shows the signal with only the Fig. 2. (a) Chemical structure and ( b) spectra of Pyridine
two UV beams focused into the sample. The exponen- 2 and (c) f luorescence with overlapping UV and STED
tial decay time is 共725 6 15兲 ps, which corresponds to pulses. Periodic interruption of the STED beam with a
chopper leads to quenching and recovery. (d) Fluorescence
the lifetime of Pyridine 2. The dashed –dotted curve
dynamics on the nanosecond scale: Two UV pulses that
shows the dynamics when the STED pulse coincides are 6 ns apart produce two equal f luorescence decays (solid
with the first UV pulse. Although the f luorescence curve). A STED pulse superimposed upon the first UV
generated by the first UV pulse is quenched, the f luo- pulse quenches the excited molecules, which are, however,
rescence from the second UV pulse remains unaffected, fully reexcited 6 ns later by the second pulse. (The solid
thus demonstrating that Pyridine 2 can be fully reex- curve is shifted upward.)
956 OPTICS LETTERS / Vol. 24, No. 14 / July 15, 1999

390 6 10 nm in the y direction, as depicted in the inset

of Fig. 3(b). As a result, in the y direction the FWHM
is reduced to 106 nm; this corresponds to l兾3.6 and is
clearly beyond the Abbe limit.
The measurement in Fig. 4 demonstrates that the
sharper E-PSF results in better spatial separation of
adjacent nanocrystals. In this case the offset between
the UV and the STED beams was 270 6 10 nm.
Whereas Fig. 4(a) shows that the UV-confocal setup is
not able to discern the crystals, Fig. 4(b) shows that
the STED-confocal arrangement resolves the objects.
A single offset beam proved suff icient for substan-
tial improvement of lateral resolution in one direction.
Extending the STED area with a second beam in the x
or the z direction or applying a doughnut-shaped beam
is expected to improve the resolution in the other direc-
tions also. The large disparity in wavelength between
the excitatory and the depleting pulses, which results
from the use of a near-IR beam and its second har-
monic, adversely affects the resolution improvement.
Visible beams would involve weaker longitudinal chro-
matic aberrations and lead to more-efficient narrow-
ing of the focus. Therefore we expect that employing
Fig. 3. Effective PSF’s of (a) standard UV-confocal and
visible light, for example, from an optical parametric
( b) corresponding STED-confocal microscopes with the oscillator, will make possible a greater relative im-
stimulating beam displaced in the y direction. The pro- provement of the resolution.
files on the right reveal a narrower PSF in ( b) than in (a). In summary, by implementing the concept of STED
microscopy, we have achieved lateral resolution beyond
the diffraction limit. The technique is noninvasive,
and the moderate intensities appear not to cause pho-
todamage to the sample. The STED concept should be
applicable to any organic molecule. In fact, this con-
cept can be considered for any application in which
the lower excited state of a four-level system is effec-
tive. Therefore future applications may well include
subdiffraction resolution in pump–probe spectroscopy,
three-dimensional photochemistry, and data storage.
We thank M. Glatz for developing the time-resolved
detector and D. Ouw for the Pyridine 2 emission
spectrum. This work was partly supported by the
German Ministry of Research and Education (BMBF
project 13N7324/9).

References
1. E. Abbe, Gesammelte Abhandlungen (G. Fischer, Jena,
Germany, 1904).
2. G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel, Phys.
Rev. Lett. 49, 57 (1982).
3. A. Lewis, M. Isaacson, A. Harootunian, and A. Muray,
Ultramicroscopy 13, 227 (1984); D. W. Pohl, W. Denk,
and M. Lanz, Appl. Phys. Lett. 44, 651 (1984).
4. T. Wilson and C. J. R. Sheppard, Theory and Practice
Fig. 4. Images of adjacent nanocrystals in (a) the stan- of Scanning Optical Microscopy (Academic, New York,
dard UV-confocal and ( b) the STED-confocal microscopes. 1984).
Comparison of the images and the profiles reveals signifi- 5. S. W. Hell and J. Wichmann, Opt. Lett. 19, 780 (1994);
cantly improved spatial resolution in the far field. M. Schrader, F. Meinecke, K. Bahlmann, M. Kroug, C.
Cremer, E. Soini, and S. W. Hell, Bioimaging 3, 147
(1995).
for aberration-free focusing, also indicating that the 6. J. R. Lakowicz, Principles of Fluorescence Spectroscopy
observed crystal can be regarded as a nearly pointlike (Plenum, New York, 1983).
source. Figure 3(b) shows the E-PSF with the STED 7. J. R. Lakowicz, I. Gryczynski, V. Bogdanov, and J.
beam switched on. The E-PSF is laterally offset by Kusba, J. Phys. Chem. 98, 334 (1994).