Confocal or multiphoton laser scanning microscopyoffers the possibility of getting three-dimensionalhigh-resolution images of thick specimens. One ob-tains three-dimensional resolution by confining themeasured signal to the small focal volume of a tightlyfocused laser beam. In the case of one-photon excita-tion, this is achieved by confocal detection.1 Whereasin the case of nonlinear excitation, localization isachieved by the excitation process.2 One acquires animage of the object point by point by scanning thefocal point through the sample. One drawback of la-ser scanning microscopes is that the frame rate islimited by the time-consuming pointwise scanningprocess. Imaging speed can be dramatically improvedif simultaneous multipoint excitation is employed.3,4
In this case the sample is scanned simultaneously bymany beams that are focused to different points. Thisreduces the number of points that have to be scannedby one individual beam and speeds up the imaging
process accordingly. The highest imaging speed be-comes possible if the distances of the focal points ofthe individual laser beams are so small that thefocal volumes start to overlap. In this case, a wholearea of the sample is covered by the beams, and noscanning is necessary. Unfortunately, it is a funda-mental problem that the axial sectioning capability,which is the main advantage of laser scanning mi-croscopy, is lost owing to interference of the over-lapping beams if the distance between the beams istoo close. For nonlinear excitation, it has been sug-gested that interference of the beams can be pre-vented by the introduction of small temporaldelays.5 In this paper we demonstrate for the firsttime two-dimensional imaging by multibeam two-photon microscopy without scanning. Images of flu-orescent latex beads and stained yeast cells aregiven as an example. The axial resolution is inves-tigated by our measuring the sea response.6
This new imaging technique could be used to ob-serve the diffusion of fluorescent molecules on mem-branes with a high frame rate.
2. Experimental Setup
The experimental setup is shown schematically in Fig.1. The beam of a Ti:Sa laser (Coherent Vitesse,800 nm, 800 mW) is used to excite the fluorescence inthe sample. The beam can be attenuated by a rotatablehalf-wave plate and a polarizer. The pulse length inthe sample is controlled by a dispersion-compensationunit of two SF10 prisms. The pulse duration inside the
When this research was performed, the authors were with theDepartment of Applied Laserphysics and Laserspectroscopy, Uni-versity of Bielefeld, Universitätsstrasse 25, D-33615 Bielefeld,Germany. T. Nielsen is now with Philips Research, D-22315 Ham-burg, Germany. The e-mail address of M. Fricke [email protected].
Received 18 May 2004; revised manuscript received 3 November2004; accepted 4 November 2004.
sample is approximately 41 fs. The beam is split bymultiple reflections into an array of 8 � 8 beams thatconverge to one point. This intersection point of allbeams is imaged by a telescope onto the back apertureof the microscope objective lens (Olympus, 60��1.2N.A. water immersion) that focuses the beams into thesample. All beams are focused to different locationsowing to their angles relative to each other. The objec-tive lens is mounted to a piezoelectric transducer(Physik Instrumente, PIFOC) to control the axial focusposition. An inverted microscope (Olympus IX70) isused for the imaging experiments. The laser beamsenter the microscope by a special side port and arereflected by a dichroic mirror toward the objective lens.The fluorescence from the sample is collected with thesame objective lens, passes the dichroic mirror, and isimaged onto a CCD camera (Imager3, LaVision). Anedge filter (2�mm BG39, Schott) in the detection beampath is used to suppress laser stray light.
The beam multiplexer (see Fig. 2) has been dis-cussed in greater detail earlier.7 Here only the mainfeatures are summarized. The beam enters from thelower left and is split into two beams by a dichroic50% beam splitter. Both beams are then reflectedback to the beam splitter by two high-reflectivity mir-rors (S0, R0) and are split into four beams. This isrepeated to obtain eight beams. The eight beams arethen turned by 90 deg by a periscope and are eachsplit again into eight beams in a second beam-splitterstage that is constructed the same way as in Fig. 2. Inthis way, an array of 8 � 8 beams is generated fromthe laser beam. By tilting the mirrors Si by an angle
of �Si� 2i�1�, we find that all beams have the same
angle of � relative to each other. Because the distanceof the foci is determined by the angle between thelaser beams, an evenly spaced array of foci is gener-ated.
Owing to the different optical paths of the beamsinside the beam-splitter arrangement, the beamshave a small temporal delay of approximately 10 pswith respect to each other. As a consequence of this,there is no interference of the beams inside the sam-ple because the delay between the beams is muchlarger than the duration of the femtosecond pulses.This is especially of importance if the foci are close toeach other because interference between the beamsinside the sample severely reduces the axial resolu-tion.3 To achieve small focal distances, the angle ofthe beams relative to each other must be small. At theexit of the beam splitter, the beams are separated by3.3 mm in the x direction (by 3.7 mm in the y direc-tion). Because the distance of the beams at the exitof the beam splitter cannot be reduced easily with-out the beams starting to overlap or part of a beambeing cut off, the only way to get small angles is tointroduce a large optical path between the beamsplitter and the microscope. To obtain focal dis-tances of 0.47 �m �y: 0.53 �m�, we set the distancebetween the beam splitter and the microscope to 7 m.Additionally, the beams are magnified by a telescopeby a factor of 3, which further reduces the anglebetween the beams to 160 �rad �y: 180 �rad�. Thearea of the sample that is covered by the foci is3.3 �m � 3.7 �m.
3. Two-Dimensional Imaging Capabilities
To demonstrate the two-dimensional imaging capabil-ities of the setup, we show in Fig. 3 an image of thefluorescence that is obtained from a homogeneous so-lution of Rhodamine 6G in water with the foci 10 �mdeep inside the solution. From this image the unifor-mity of the illumination of the area that is covered bythe 8 � 8 beams can be judged. Intensity profiles atseveral positions indicated by the arrows in Fig. 3 aredisplayed in Fig. 4 along with a profile of the fluores-cence of one individual beam (this was obtained by ourblocking some of the mirrors in the beam-splitter ar-
rangement). The intensity distribution in the focalplane is strongly dependent on the alignment of thebeam-splitter mirrors. There is some remaining inho-mogeneity in the illumination resulting from the mul-tipoint excitation. The variation between maximumand minimum fluorescence intensities is approxi-mately 30%. This is due to the fact that the distancebetween the beams is slightly larger than the fullwidth at half-maximum of the fluorescence of a singlebeam �0.45 �m�. The second reason is that not allbeams hit the aperture of the objective lens equallywell, resulting in different intensities of the individualbeams. Because the diameter of the beams is approx-imately 2.5 times larger than the aperture (owing todivergence of the laser beam and the magnifying tele-scope), the beams still fill the whole aperture. Bothreasons lead to a variation of the excitation efficiencyover the field of view. But, since this is independent ofthe object that is studied, it can be calibrated for.
As examples of two-dimensional imaging, Fig. 5displays an image of fluorescent latex beads (Molec-ular Probes) with 400�nm diameter. Figure 6 shows
the fluorescence image of yeast cells stained withRhodamine 6G along with a bright-field image. Inthat the cells are larger than the area that is coveredby the beams, only part of the cells is visible in thefluorescence image.
4. Axial Sectioning
The axial sectioning capability is one important as-pect of two-photon microscopy. To demonstrate thatit is not affected by the multibeam setup, we mea-
Fig. 3. Fluorescence from a homogeneous solution of Rhodamine6G in water. The arrows indicate the lines along which the profilesof Fig. 4 were taken.
Fig. 4. Intensity profiles of the image displayed in Fig. 3. Theprofiles belong to the lines identified by the arrows and symbols inFig. 3. From this, the uniformity of the illumination can be seen.For comparison, the profile of a single beam is also included in theplot.
Fig. 5. Image of fluorescent latex beads.
Fig. 6. Fluorescence image of yeast cells stained with Rhodamine6G and a bright-field image of the cells. The region that is coveredby the laser foci is indicated by the white rectangle in the bright-field image.
sured the so-called sea response, i.e., the fluorescencesignal if the foci are scanned axially from a nonfluo-rescent medium (cover slip) into a homogeneouslyfluorescent medium (in this case a solution of Rhoda-mine 6G in water). From theoretical calculations, onecan derive the response curve3:
I(z) �A� ��
2 � arctan�z � z0
zR��
1zR
z � z0�
z � z0
zR
�,
(1)
with
zR � 1.169� n
N.A.2�� 0.86 �m, (2)
where n � 1.33, N.A. � 1.2, and � 800 nm.Figure 7 shows the result of the measurement if the
signals of all beams are averaged and if only onebeam is used along with the prediction of the model.Two things can be seen from this plot. First, bothmeasured curves have the same general behavior asthe theoretical curve. It should be noted in particularthat at both ends of the scanned region (i.e., far fromthe edge) the curves are flat. This is in contrast toother measurements of the sea response that have bedone with multifocal two-photon microscopy.3 Thesemeasurements were carried out at interfocal dis-tances of 4.5 �m, but still interference between thebeams led to a rising signal even if the foci were farfrom the edge. This contribution was found to in-crease if the focal distance was reduced. That thisrising signal is missing here—even though the inter-focal distance is ten times smaller—shows that thefoci of the individual beams are truly independentowing to the temporal delays.
Second, the response curve of a single beam is asexpected from the theory. It is steeper than the re-sponse curve that is obtained from all beams. This isdue to the fact that the axial position of the focus is
not exactly the same for all beams. By averaging overthe response curves from all beams that are shiftedslightly with respect to each other, we find that theeffective response curve is not as steep as the curvesof the individual beams. This has been confirmed bymeasurement of the sea response of all 64 beamsindependently (data not shown). The width of therising edge [zR in Eq. (1)] was found to be essentiallyindependent of the beam. The axial position of theedge [z0 in Eq. (1)] has a distribution approximately0.2 �m wide. Possible reasons for these deviationsare aberrations of the optics or defects of the beam-splitter mirrors.
If the plane mirrors in the beam multiplexer arenot exactly flat, they would act as a focusing or defo-cusing element. Because each beam hits a differentcombination of mirrors in the multiplexer, they wouldbe differently focused or defocused, leading to a vari-ation of the axial focal spot position in the focal planeof the microscope. E.g., a tiny curvature of the mirrorwith a radius of 10 m would be enough to cause afocal spot shift of 0.1 �m.
If an image stack is recorded with this setup to geta three-dimensional reconstruction of the object, itdoes not matter that the foci are not exactly in aplane. By measuring the exact position of the foci(e.g., by the sea response) and using a deconvolutionalgorithm that takes into account the actual focalpositions, one can realize the axial resolution of asingle-beam microscope.
For better appreciation of the axial resolution ofthis imaging technique, Fig. 8 shows the sea responseof the multifocal excitation in comparison with asingle-beam excitation covering the same area of ap-proximately 4 �m � 4 �m. Obviously, the multibeamexcitation results in much better axial resolution.
5. Discussion
The field of view of the current setup is limited to4 �m � 4 �m. In the following, we estimate which
Fig. 7. Response curves that are obtained if the foci are scannedaxially into a fluorescent sea: all beams (solid curve), single beam(dashed curve), and theoretical limit (dashed–dotted curve).
Fig. 8. Response curves that are obtained if the foci are scannedaxially into a fluorescent sea: multiple foci covering an area of4 �m � 4 �m (solid curve) and a single focused beam with anequivalent area of the focus (dashed curve). Obviously themultibeam approach results in much better axial resolution than asingle beam.
field of view could be achieved by an optimized setup.The field of view can be enlarged by use of more than8 � 8 laser beams, but this requires higher laserpower. A factor of 2 can be gained with a more pow-erful laser system. One can gain another factor of 5 byoptimizing the beam-splitter optics by better match-ing the diameter of the beams to the aperture of themicroscope objective lens. This would also lead to amore homogeneous excitation. Because the fluores-cence signal is inversely proportional to the pulselength, it could also be enhanced by use of shorterpulses. But with the pulse length already being ap-proximately 45 fs, a significant improvement can beachieved only with much effort. This means, in con-clusion, the number of beams could be enlarged by afactor of 10, leading to an increased field of view ofapproximately 10 �m � 10 �m. This is sufficient formany live cell studies.
Even with the current field of view, interestingproblems can be studied; e.g., the diffusion of fluores-cent molecules on membranes can be observed with ahigh frame rate. One advantage of the technique pre-sented here is that fluorescence excitation is limitedto a thin section of the sample. Uptake of fluorescentmolecules by cells can be investigated without a prob-lem with fluorescence from the medium outside thecells. Tracking the diffusion of fluorescent moleculesin three dimensions requires only scanning in theaxial direction, which works fast and can be imple-mented easily.
The frame rate of this imaging technique is, inprinciple, limited only by the repetition rate of thelaser �80 MHz�. This is a fundamental differencefrom scanning systems, which are always limited bythe speed of the scanning system. A practical limita-tion is, of course, the signal strength, but this is alsothe case for scanning systems.
A drawback of this imaging technique is that it willwork well only if the emitted fluorescence is not scat-
tered strongly. This means it is limited to specimensonly a few cell layers thick. Measurements withinthick tissue will have poor lateral resolution becausethe fluorescence emission from the plane of excitationcannot be imaged properly onto the detector.
6. Conclusion
With the imaging technique presented here, it is pos-sible to get high-resolution images of thin axial sec-tions of specimens without the need for scanning inthe lateral directions. This allows the highest-speedimaging because only the amount of fluorescencelight limits the acquisition time that is needed for oneframe.
We thank Peter Andresen, who died in February2001, for his support in this study. We gratefullyacknowledge support from the Bundesministeriumfür Bildung, Forschung und Technologie (Verbund-projekt “Nicht-lineare Laser-Raster-Mikroskopie,”13N73075) and LaVision GmbH, D-37081 Göttingen,Germany, for funding.
References1. J. B. Pawley, ed. Handbook of Biological Confocal Microscopy
(Plenum, New York, 1995).2. W. Denk, J. H. Strickler, and W. W. Webb, “Two-photon laser
scanning fluorescence microscopy,” Science 248, 73–76 (1990).3. A. H. Buist, M. Müller, J. Squier, and G. J. Brakenhoff, “Real
time two-photon absorption microscopy using multi point exci-tation,” J. Microsc. 192, 217–226 (1998).
4. J. Bewersdorf, R. Pick, and S. W. Hell, “Multifocal multiphotonmicroscopy,” Opt. Lett. 23, 655–657 (1998).
5. A. Egner and S. W. Hell, “Time multiplexing and parallelizationin multifocal multiphoton microscopy,” J. Opt. Soc. Am. A 17,1192–1201 (2000).
6. S. Hell, G. Reiner, C. Cremer, and E. H. K. Stelzer, “Aberationsin confocal fluorescence microscopy induced by mismatches inrefractive index,” J. Microsc. 169, 391–405 (1993).
7. T. Nielsen, M. Fricke, D. Hellweg, and P. Andresen, “Highefficiency beam splitter for multifocal multiphoton microscopy,”J. Microsc. 201, 368–376 (2001).
