Confocal Microscopy: Recent Developments

Confocal Microscopy: Recent Developments


3MB Sizes 1 Downloads 31 Views


Confocal Microscopy: Recent Developments ERNST HANS KARL STELZER and FRANK-MARTIN HAAR Light Microscopy Group. Cell Biology and Biophysics Programme. European Molecular Biology Laboratory ( EMBL) Meyerhofstrasse I [email protected] 10.2209, 0-6911 7 Heidelberg. Germany



1. Resolution in Light Microscopy . . . . . . . . . . . . . . . . . . . . . 11. Calculating Optical Properties . . . . . . . . . . . . . . . . . . . . . .

A . Point-Spread Functions . . . . . . . . . . . . . . . . . . . . . . .

111. Principles of Confocal Microscopy . . . . . . . . . . . . . . . . . . . .

A . Light Paths in a Confocal Microscope . . . . . . . . . . . . . . . . . B. Technical Aspects of a Confocal Microscope . . . . . . . . . . . . . . C. Applications of Confocal Microscopy . . . . . . . . . . . . . . . . . D . Alternatives to Confocal Microscopy . . . . . . . . . . . . . . . . . E. Optimal Recording Conditions . . . . . . . . . . . . . . . . . . . . F. Index Mismatching Effects . . . . . . . . . . . . . . . . . . . . . . IV . improving the Axial Resolution . . . . . . . . . . . . . . . . . . . . . A . Standing-Wave Fluorescence Microscopy . . . . . . . . . . . . . . . B. 4Pi-Confocal Fluorescence Microscopy . . . . . . . . . . . . . . . . C . Confocal Theta Microscopy . . . . . . . . . . . . . . . . . . . . . V . Nonlinear Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Two-Photon Excitation . . . . . . . . . . . . . . . . . . . . . . . B. Multiphoton Excitation . . . . . . . . . . . . . . . . . . . . . . . C . Stimulated-Emission-Depletion Fluorescence Microscopy . . . . . . . . D . Ground-State-Depletion Fluorescence Microscopy . . . . . . . . . . . VI . Aperture Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII . Axial Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Three-Dimensional Measurements and Qualitative Analysis . . . . . . . . VIII . Spectral Precision Distance Microscopy . . . . . . . . . . . . . . . . . . IX . Computational Methods . . . . . . . . . . . . . . . . . . . . . . . . X . Spinning Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI . Perspectives of Confocal Fluorescence Microscopy . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

293 299 299 301 301 307 307 308 309 309 311 311 314 317 321 322 324 324 325 327 329 331 333 334 335 336 331


The modern light microscope is usually operated in a mode that is close to the diffraction limit [Abbe. 18731. This means that the resolution is determined by the wavelengths of the incoming and outgoing light. the refractive index of the medium. the focal length of the lens. and the diameter of the aperture (Figs. 1 and 2). The optical systems. in particular the 293

Volume 106 ISBN 0-12-014748-3

ADVANCE3 I N IMAGING AND ELECTRON PHYSICS Copyright ‘c) 1999 by Academic Press All rights of reproduction iii m y form reserved ISSN 1076-5670/99 $30 00



FIGURE1. Characteristic properties of a telecentric system. In a telecentric optical system all beams pass the aperture diaphragm as a planar wave. In the optical path the tilt angle p determines the distance of the focus from the optical axis. The focal length f of the objective lens and the diameter of the entrance aperture 2a determine the opening angle a. The numerical aperture is the product of the angular aperture sin CI and the refractive index of the medium n. A planar wave tilted by an angle fi has a focus in the object plane at a distance s = J’.tanB from the optical axis.

objective lenses, have their limits, of course. They will not transmit outside a certain spectral range, they have a finite working distance and a finite field of view, and they show slightly different lateral and axial magnifications depending on wavelength, position in the field, polarization, and temperature. Assuming the usual working conditions, such limits are not encountered. The consequence is that it makes sense to calculate system functions (point-spread functions), to discuss their properties, and to describe the image formation process as an accumulation of point images.



w j



29 5




FIGURE2. Three-dimensional imaging in a telecentric system. A pair of angles always encodes the lateral positions (see Fig. 1). A divergence or convergence angle defines the position of the emitter along the optical axis. The lateral distance M ' s in the image plane i is independent of the position of the emitter along the optical axis. In every telecentric system the lateral magnification is the ratio of the focal length of the tube and the objective lens. The axial magnification is the square of the lateral magnification M. If the objects are located in different planes with a distance z then the images have a distance along the optical axis M 2 .z.

A thorough and readable description on how to calculate the intensity distribution in the focus of a lens is found in the famous book by Born and Wolf [1980, pp. 435-4411. The essence of their calculations and of many other authors is that the lateral resolution (Ax, Ay, i.e., the resolution in the focal plane) is proportional to the wavelength 1 and inversely proportional



to the numerical aperture N . A . = n.sin a. The axial resolution (Az,i.e., the resolution along the optical axis) is proportional to the wavelength and the refractive index n and inversely proportional to the square of the numerical aperture. The effects of the light distribution in the aperture are only apparent in some factors (ki,mi,yi):

AX = k;k,. ..:

A kp*N.A.

- k, . k , -

n.i N .A .


i 1 n [email protected]

..: k; -. -

Az = m , ~ m , - , . : r n ; ~= m,.m,. ..:m

A 1 n sin2ci

The minimal extent of a volume element Au is, therefore, proportional to the third power of the wavelength and inversely proportional to the fourth power of the angular aperture. The volume element [Lindek et al., 1994b1 is still not a popular way of looking at resolution, but it is reasonable. Microscopy, and in particular confocal microscopy and two-photon excitation, provide a three-dimensional resolution; therefore, a three-dimensional resolution criterion is necessary, which documents that a resolution improvement along one axis is not imposed at the expense of a decrease along another axis. Some techniques using, for example, annular apertures [Wilson and Hewlett, 19901 improve the lateral resolution but at the same time degrade the axial resolution. In this case the three-dimensional (volume) resolution will be worse. Before going into the details one can consider how the lateral and/or the axial resolution can be improved by decreasing the illumination wavelength, increasing the refractive index, or increasing the numerical aperture [Stelzer, 19981. Other means to modify the resolution are nonlinear effects, multilens arrangements, or computational efforts that take the actual illumination/ detection process into account. The lateral and the axial resolutions can be improved simultaneously by decreasing the wavelength. The wavelengths, currently used span a range from 300 nm to about 1100 nm. Due to technical limits the wavelengths are probably closer to a range from 350 nm to 900 nm. Wavelengths in the UV may provide the optimal resolution, but biological objects tend to suffer and eventually die [Carlsson et al., 1992; Montag et al., 19911. In fluorescence microscopy the excitation wavelength has to be adapted to the fluorophore. The choice of the dye thus determines the wavelengths of the light source [Tsien and Waggoner, 19951.



However, longer wavelengths have their virtues [Fischer, Cremer, and Stelzer, 19951. Short pulses of high intensity can be used to induce a two-photon absorption process [Denk, Strickler, and Webb, 19901. Applying two-photon excitation in fluorescence microscopy has three effects. First, the wavelength is increased by a factor of two and the resolution is a factor of two worse along all three axes [Stelzer et al., 19941. Second, the proportionality of the fluorescence emission to the square of the excitation intensity introduces the same factor 1/$ encountered in confocal fluorescence microscopy [Sheppard and Gu, 19903. This partially compensates worse, but the higher wavelength and the resolution is only a factor of the two-photon microscope has the same properties as a confocal fluorescence microscope. Third, almost all dyes can be excited with two photons [Fischer, Cremer, and Stelzer, 1995). Increasing the refractive index will improve the resolution in the same manner as a decrease of the wavelength. Lenses for refractive indices up to 1.7 (using Xylol as immersion medium) have been available, but common oil-immersion media have a refractive index of 1.518 at 23°C and a wavelength of 546 nm. Since the observation of living samples becomes more important in the life sciences, water immersion lenses corrected for the refractive index of 1.33 replace oil immersion lenses. Increasing the angular aperture (sin a ) will also increase the resolution. Angles a up to about 70" are technically feasible. The largest numerical aperture ( in air is 0.94, in water 1.23, and in oil 1.4. Another limit is the angle of total reflection, which is encountered when oil immersion lenses are used to observe samples mounted in an aqueous medium [Hell et al., 19931. The numerical aperture is then at most 1.3. Another problem is the refractive index mismatch, which induces important spherical aberrations at large depths [Hell et al., 1993; Torok, Varga, and Booker, 1995). However, magnification, working distance, and numerical aperture of an objective are not independent. An oil immersion lens with a numerical aperture of 1.4 and a magnification of 63 (abbreviated 1.4/63x) will have a working distance around 0.250 mm, while a water immersion 0.9/63x lens will have a working distance around 1.5 mm. High numerical aperture lenses are, therefore, only available for thin objects. A good method to reduce the stray light coming from thick samples is to reduce the field of view. The confocal microscope reduces the illumination field and the detection field to the physical limit determined by diffraction and hence discriminates all light emitted outside the focal volume. Using a confocal microscope [Brakenhoff, 19793 increases the lateral resolution by 1/$. It affects only the factors ki,mi,qi. However, as we will explain later, it has a depth discrimination capability like that of the two-photon excitation microscope.




Another important method to improve the resolution is to change the intensity distribution in the aperture plane. The best-known case is the annular aperture [Airy, 1841; Sheppard, 19771. It has a low transmission in the center and a high transmission on the edges. The higher angles thus contribute more to the image formation process. This improves the lateral resolution up to 73% but causes extensive ringing and increases the depth of field to infinity. Special apertures can be designed that maintain the lateral resolution intact but improve the axial resolution (Martinez-Corral, Andres, and Zapata-Rodriguez, 1995b1. While all the methods previously mentioned leave the instrument basically intact, others require severe modifications to the microscope, as it is traditionally known. Recently introduced methods that improve the numerical aperture use two or more lenses that illuminate a sample coherently or detect the emitted fluorescent or scattered light coherently. In a standing-wave fluorescence microscope [Lanni, 1986; Lanni, Waggoner, and Taylor, 19861 a whole field is illuminated coherently using planar waves in the focus of two opposing lenses, which produces a fringe pattern along the optical axis. Images are recorded as a function of the phase and the position of the object along the optical axis and later reconstructed using appropriate algorithms. In a 4Pi(A) confocal microscope [Hell and Stelzer, 1992a,b], two opposing lenses use spherical waves to illuminate a focal spot coherently and produce a standing wave that modulates the intensity along the optical axis. Two minima are slightly more than i/2n apart. Adjusting the phase and moving the object relative to the focal spot while recording the fluorescence intensity as a function of the position generates the images. Two lenses in an orthogonal arrangement provide another interesting method since in this case the lateral resolution dominates the extent of the point-spread function along all three directions [Stelzer and Lindek, 19941. The resolution becomes isotropic, that is, the lateral and the axial resolution are almost identical. In the confocal theta microscope the axial resolution can be improved by a factor of three, and low-NA systems may have axial resolutions, which are better than those achieved with confocal high-NA systems. In tomographic methods [Bradl et al., 1994; Shaw et al., 1989; Skaer and Whytock, 19751, the sample is mounted on a rotating stage and observed from different angles. This can be accomplished using conventional or confocal microscopes. The data sets are then normalized and an attempt is made to construct an improved view of the object. Such a system can be applied to increase the axial resolution to almost the lateral resolution, and three-dimensional distances can be measured with an improved accuracy. Another nonlinear method for fluorescence microscopy is to prevent the excitation of fluorophores by depleting their ground state [Hell and Kroug,



1995; Hell and Wichmann, 1994) or by stimulating the emission [Hell and Wichmann, 19941. By generating appropriate patterns with one light source and observing the remaining fluorophores with another light source, a higher resolution can be achieved. If the imaging process can be reconstructed using samples with a well-characterized spatial dye distribution, attempts to estimate the fluorophore concentration in the sample by computational means can be applied [Agard and Sedat, 1983; Carrington et al., 1995b; Shaw, 19951. Such methods can be used with data recorded using wide-field methods as well as data recorded with confocal microscopes. The former actually competes with the confocal microscope. The latter, however, is usually regarded as a means to relax the confocal imaging conditions, for example, using larger pinholes or noisier images. The first description of a confocal microscope is found in a patent by Marvin Minsky [1961; 19881. It was probably Charles McCutchen [1967] who first appreciated and described the properties of a combined point illumination and detection device. Petran et al. [1968] built one of the earliest instruments, and the first experimental verification of the confocal principle is due to Godefridus Jacobus Brakenhoff, P. Blom, and C. Bakker [1978] and Brakenhoff, Blom, and P. Barends [1979]. The first investigators to realize the depth discrimination capability of the confocal fluorescence microscope were Ingemar Cox, C. J. R. Sheppard, and T. Wilson [1982]. Most of the literature on the theory of confocal microscopes has been written by Tony Wilson and Colin Sheppard [198l].


In a microscope the field of a pointlike light source in the image plane is equivalent to the system response. It is referred to as the amplitude pointspread function (PSF) and is used to describe the properties of the optical components for a given wavelength [Born and Wolf, 1980, pp.435-4491. The most basic approach to calculating an amplitude PSF is to apply Huygens’ principle, that is, to assume an aperture with an area A as the source of waves. A complete description of this process is the following equation

which can be solved numerically for each point p in the object volume [Born and Wolf, 1980, p. 4361. Appropriately phrased, it will take polariz-



ation effects into account [Hell et al., 19931. However, for most purposes it is sufficient to restrict all calculations to a domain in which a linear approach holds. The amplitude PSF extends in all three dimensions. Due to the cylindrical symmetry of lenses the two lateral components can be regarded as equal. It is, therefore, in most cases sufficient to be able to describe an amplitude PSF in a plane containing the optical axis. Applying appropriate simplifications [Born and Wolf, 1980, p. 436ff1, the amplitude PSF becomes a solution of

h(u, v) = - i

2nnA sin2cr ei&



Jo(vp)e - + U P z p dp

v = 2nnr sin ~ J I . A u = 2nnz sin2u//1 A r =


where v and u are normalized optical units perpendicular and parallel to the optical axis, respectively [Hopkins, 19431, while r is the distance from the optical axis. Neither the spatial distribution of the amplitudes nor the variation as a function of time can be measured directly in the optical frequency range. However, the intensity PSF (i.e., an image) can be visualized by placing a piece of paper into the optical path or after recording it using a camera oriented normal to the optical axis. The intensity PSF jh(u, u)I2 is calculated as the product of the amplitude PSF and its complex conjugate

Ih(u, Y ) ( 2

= h(u, v ) . h*(u, v).

The intensity PSF also describes the spatial absorption pattern of a uniform fluorophore solution in the vicinity of the focus. The concept of transfer functions [Goodman, 1968, pp. 111-120; Frieden, 19671 has many advantages. However, in this paper we avoid them and use PSFs instead. Coherent transfer functions are the Fourier transform of the amplitude PSF. Optical transfer functions are the autocorrelation function of the coherent transfer function. Therefore the two descriptions are equivalent. There are three questions concerning the concept of PSFs or system functions:

1. Can they be measured? The concept of PSFs may seem straightforward, but under ordinary conditions a recording system based on an objective lens has to be able to cope with aberrations, noise, and a number of nonlinear effects. Spherical aberration, for example, causes a decrease of the energy under the main maximum and its shift into the higher-order terms [Hell et al., 19931. The maximal intensity is lower than in the unaberrated PSF. A number of conditions have to be met by an optical system: The optical system must be linear, that is, no light must be absorbed or scattered and no object must be in the shadow of another object.



The optical system must be invariant, that is, features in the image of an object must be independent of the position of the object in the field of view. Since these conditions are never perfectly fulfilled, the concept of PSFs is only approximately correct. Therefore, the existence of PSFs is not obvious. 2. How can one derive PSFs from real images? The shape of the analyzed object has to be taken into account. This is often forgotten when regular patterns are used to determine the resolution of an optical system. It is probably a common error to confuse the transfer function of line pairs that are rectangular (or square wave) functions with those of sine waves. 3. How are they calculated? As previously pointed out, two conditions must be met for PSFs to exist. There is a rich literature on how to include effects such as high numerical aperture, absorption, and refractive-index mismatch. They usually provide insight into how to calculate images of point sources, but these are in general not identical to PSFs. It usually turns out that invariance is not maintained and that taking into account absorption effects requires the exact path of light through the object [White et al., 19961. Although we are aware of the theory’s deficiencies we will follow it as Born and Wolf present it in our calculations of intensity distributions. We will also not take into account effects such as polarization, absorption, and scattering, unless they are mentioned explicitly. In this paper we investigate images of point objects. These objects are so small that their features cannot be resolved. We define point objects as objects whose diameters are much smaller than the diameter of the intensity PSF. In fluorescence microscopy we look at objects in which, for example, the diameter of the area over which the fluorescent molecules are scattered, or the maximum distance between two fluorescent molecules, is very often only half the diameter of the Airy disk [Stelzer, 19981. Each of the molecules creates an intensity PSF in the image, but they are so close to each other that their sum becomes smeared and indistinguishable from a single intensity PSF. 111.



A. Light Paths in a Confocal Microscope

A confocal fluorescence microscope (CFM) is usually based on a conventional microscope. It contains an objective lens, a stage for the sample, a light source, and a detector. If we work with fluorescent light, two filters are



instrumental. The dichroic deflector separates the illumination light with which the fluorophore is excited from the fluorescent light that is emitted by the fluorophore, while the second filter separates the emitted fluorescence from scattered excitation light (Fig. 3). The light path is best understood if one first looks at the excitation light and then at the emission light. The light source is usually a laser. The laser light is focused into a pinhole, deflected by a dichroic mirror into the objective lens, and focused inside the specimen. Most of the light will pass the specimen, but a small fraction is absorbed by the fluorophore, which will emit fluorescent light. The fluorescent light is emitted in all directions with an almost equal probability.' The lens will, therefore, collect only a small fraction of the fluorescent light. This fraction passes the dichroic mirror and is focused into a pinhole in front of a detector. The detector will convert the flux of photons into a flux of electrons, which is converted to a number proportional to the intensity of the fluorescent light [Stelzer, 199.51. As pointed out, the excitation light passes the sample. It will excite not only the fluorophores that are in the plane of focus but also those that are either in front or behind the plane of focus. However, their images are either in front or behind the plane in which the point detector is located (Fig. 2). In the plane of the detector these images are expanded, hence only a small fraction of the light will pass the pinhole and enter the detector. The detector pinhole thus discriminates against the light that is not emitted in the plane of focus. The importance of the pinhole may become clearer if the detector pinhole is removed and a detector with a large sensitive area is used. The discrimination does not occur anymore. Instead, all the fluorescent light that is collected by the objective lens contributes to the signal. Such an optical arrangement behaves essentially like any conventional fluorescence microscope. Another view is to regard the objective lens as the device that forms an image of the illumination pinhole and the detection pinhole in their common conjugate image plane, the object plane 0.Only the fluorophores that are in the volume shared by the illumination and detection PSFs are excited and detected (Figs. 3 and 4). Therefore, in order to calculate the confocal PSF, one calculates the illumination intensity PSF and the detection intensity PSF and multiplies the two PSFs: The PSFs can also be viewed as being proportional to probability density 'Each Buorophore behaves as a dipole, but in general its orientation is not fixed and many fluorophores are observed at the same time. On average the polarization and the orientation are lost. Fluorophores attached to bio-polymers behave differently (Marriott, G. Zechel, K., Jovin, T. M. (1988). Spectroscopic Biochem. 27(17) 6214-6220).


7 &\dichroic

\/ I-y




FIGURE3. Principal layout of a beam/object scanning confocal fluorescence microscope. A laser beam is focused into an illumination pinhole (in the plane i’), collimated and deflected toward a microscope objective lens, which focuses the light inside the object. The emitted light is collected by the objective lens, passes the dichroic mirror, and is focused into the detection pinhole in the image plane i. This pinhole allows the light emitted in the plane of focus o to pass and discriminates against all out-of-focus light (also see Fig. 2).

functions. An integral over a volume ui in the illumination intensity PSF describes the probability of illuminating the fluorophores in that volume. In order to operate confocally, both events -the illumination event and the detection event -have to occur. The probabilities have to be multiplied. In many cases the illumination and detection intensity PSFs are quite similar, and a reasonable first approximation of the confocal intensity PSF is to assume it is the square of the illumination intensity PSF: Ihc,(x, Y , z)12

Ihi,(x, Y , z)12


Y , z)I’ = (Ihiii(X* Y , 211’)~.

Figure 4 shows three intensity PSFs. Figure 5 shows their components




2 N



a: 4 \9






0 0.8


.-3 0.6



4 0.4

4 0.4


8 0.2 0.1



0.4 r [PI









FIGURE 5. Comparison of lateral (a) and axial (b) intensity point spread functions. (a) The components of the intensity PSFs in the focal plane ( z = 0) for an illumination at 488 nm, a detection at 530nm, a refractive index of 1.518, and a numerical aperture of 1.4. The illumination and detection curves describe the Airy disk. The confocal curve is the product of the illumination and detection curves. (h) The components of the intensity PSFs along the optical axis ( r = 0) at the same conditions. The illumination and detection curves follow the behavior of the function (sin(u/4)/(~/4))~.The confocal curve results from the product of the illumination and detection curves.

along a lateral direction (z = 0) and along the optical axis fr = 0). If one looks at the full width at half maximum (FWHM) value the CFM has an improved lateral resolution and an improved axial resolution by about a factor of 1/& The zero crossings (location of the first minimum) are of course identical in the PSFs. Using this definition, the CFM has no improved resolution. It should not be forgotten that a conventional fluorescence microscope has an axial resolution for pointlike objects, which is not much worse than that of the CFM. To fully appreciate the CFM one should look at the integrated intensities of the illumination and confocal intensity PSFs:







(r, z)122nrdr.

This function is constant, which reflects the conservation of energy. The square of the integrand, however, is not conserved. Thus the integral

jr=, r=m



(lhi1I(r, z)12)22nr dr

has a maximum in the focal plane (Fig. 6a). This is the explanation for the depth discrimination capability of a CFM [Cox, Sheppard, and Wilson, 1982; Wijnaendts-van-Resandt et al., 19851. The best illustration for this effect is to record the intensity as one focuses through the cover slip into a











I\ \ 1u 0.4




.a 0.6









= trml


FIGURE6. Integrated intensities and sea-response. The parameters for the calculations of the intensity point-spread functions are identical to those for the previous figures. (a) Due to the conservation of energy, the integrated intensity of the illumination PSF is constant. Since the square of the intensity is not conserved, the confocal PSF has a maximum in the geometrical focus. (b) The sea-response for a confocal microscope indicates the resolution of an axial edge.

thick layer of fluorophore dissolved in the immersion medium of the lens. This sea-response EcJ,sea(Zo) =


jz=zn z = - x

(Ihin(r,z)I2)’2nr dr dz

is plotted in Fig. 6b. It shows the intensity recorded by the photodetector behind the detection pinhole in a CFM. The light distribution described by the confocal intensity PSF in the layer of fluorophore experiences no fluorophores outside the layer and finds fluorophores everywhere in its immediate environment once deep inside the thick layer. The slope and intensity variations in the shape of the sea-response can be used to characterize the resolution of many confocal microscopes. The sea-response is unique to the CFM. A conventional fluorescence microscope has no such property and, so long as no phase information is available, no computational methods are able to reconstruct the transition into the fluorophore layer from wide-field images. It may also become clear that not all contrasts apart from fluorescence will show depth discrimination in a confocal arrangement. Transmission contrasts (implemented using two lenses) [Brakenhoff, Blom, and Barends, 1979; Marsman et al., 19831 usually depend on absorption and on scattering. Only those in which the signal is a t least partially due to scattered light will have an improved lateral resolution (e.g., phase contrast and differential interference contrast). An axial resolution as defined through the searesponse is only available in fluorescence, reflection, and scattering light microscopy.



B. Technical Asprcfs of a Confocul Microscope The confocal microscopes described so far observe only a single point in an object. Such an instrument, therefore, does not record an image. To get an image one must either move the beam relative to the object [Wilke, 19831 or the object relative to the beam while recording the intensity as a function of their relative position [Stelzer, Marsman, and Wijnaendts-van-Resandt, 1986; Voort et al., 1985; Wijnaendts-van-Resandt et al., 19851. In a practical instrument the beam is moved laterally in the focal plane of the instrument while the sample is moved along the optical axis. The lateral movement can be achieved by two mirrors, which control the direction of the beam in two orthogonal axes [Slomba et al., 1972; Wilke, 19851. The mirrors are mounted on very accurate motors (galvanometers) that allow almost arbitrary changes of the angle as well as the speed at which an angle is reached. The optical system assigns a position in the object to every angle and allows the beam to address every point in the object. A large angle is equivalent to a large field. Thus, changing the angle opening controls the field size [Stelzer, 1994; Stelzer, 1995; Stelzer, 19971. The axial movement is achieved by moving either the lens relative to a fixed stage (in most inverted microscopes) or the stage relative to a fixed optical system (in most upright microscopes). Since the axial displacement moves a larger mass, it is in general much slower than the lateral movement. A serious alternative is the use of scanning disks [Nipkow, 1884) that are located in an image plane. These have a number of holes (usually wellspaced) that transmit the light of ordinary lamps [Petran et al., 1968; Petroll et al., 19923. In its simplest form the same holes are used in the illumination and the detection process [Kino, 19951. One rotation of the disk covers the whole field of view, which is observed either directly or recorded using a camera, at least once. When a laser is used instead of the lamp, lens arrays can replace the holes in a disk and provide a very efficient and fast confocal microscope [Yin et al., 1995). Apart from the fact that certain compromises are made to allow for an efficient observation and the background is not as well discriminated, the properties of such systems are described in the same way as explained previously [Sheppard and Wilson, 1981). C. Applications of Confocul Microscopy

The main reason to use a CFM is to get rid of the background haze. The sample should be thick, that is, extend along the optical axis, and in the conventional fluorescence microscope the image should suffer from out-offocus contributions. Using the CFM the images become crisper, and features



that were invisible become observable. The new information one gets by analyzing the images contributes significantly to what was known before. Some typical good applications are the observation of small but densely labeled structures such as chromosomes [Agard and Sedat, 1983; Merdes, Stelzer, and De Mey, 1991; Stelzer, Merdes, and De Mey, 19911 or the observation of large, thick objects such as mouse embryos [Palmieri et al., 19941. The confocal microscope is of no use when the samples are flat. The slightly higher lateral resolution usually cannot be used because the signalto-noise ratio is not sufficient [Stelzer, 19981. A really bad application for CFM is to study flat, fluorescent in situ hybridized samples (for the technique, not the application, see Speicher, Ballard, and Ward [1996]). This also applies to samples in which fluorescent objects are sparse, well separated, or hardly overlapping, which is very often the case when less abundant proteins are observed. Problems are also raised with very dense objects such as those encountered in many medical samples. The dye concentrations are too high, and absorption effects prevent a penetration beyond that available in conventional fluorescence microscopy. That a new instrument ever becomes widely accepted depends on several factors. The new instrument must provide information that was not available until then or only available at outrageous costs or efforts. The instrument must be reliable. Most importantly, there have to be scientists who are willing to invest time in the sample preparation. The latter is usually underestimated. In confocal microscopy one of the main efforts is to make sure that the three-dimensional structure is preserved. This is a tricky task that is very often not accomplished and is the main reason why only living specimens were observed initially [Van Meer et al., 19871. However, considerable progress has been made [Bacallao, Kiai, and Jesaitis, 1995; Reinsch, Eaton, and Stelzer, 19981, and fixed as well as living specimens are now relatively easily observable [Zink et al., 19981.

D. Alternatives to Confocal Microscopy The simplest alternative to CFM is to work with small fields of view. This is particularly useful when thick specimens are observed, where the production of stray light is avoided by restricting the illumination to the actually observed field of view. In fact, the confocal fluorescence microscope can be regarded as the instrument that implements the lower limit for the field size by decreasing it to a single focal volume. The diameters of the illumination and detection pinholes are determined by the diffraction limit. Perhaps the most serious contender for CFM is conventional recording



and a subsequent deconvolution [Cox and Sheppard, 1993; Holmes et al., 1995; Krishnamurthi et a]., 1995; Sandison et al., 1995a; Shaw, 19951, which has been described many times and is available through many software manufacturers.

E. Optimul Recording Conditions In conventional microscopy the magnification of the lens determines the field size, and since ordinary film has the extremely high resolution of several thousand lines, a full field of view can be photographed at the resolution of the lens. The disadvantage of film is a low sensitivity. CCD-based cameras have a good sensitivity but a limited number of picture elements [Hiraoka, Sedat, and Agard, 19871. Working with appropriate oversampling requires a reduction of the field of view. On the other hand, extensive oversampling reduces the number of photons per picture element, and the images tend to become noisy. Scanning microscopes suffer in principle from the same problems as microscopes that use cameras, but the amplitude of the scanner and, therefore, the field of view can be changed, the pixel-pixel distance can be very small, and the dwell time per pixel can be adapted. The main disadvantage of scanning microscopes is that they are sampling devices. They observe one picture element at a time, whereas cameras record the intensities of all picture elements in parallel. From the point of view of sampling, one requires between 8 and 16 picture elements per Airy disk diameter to record a fully resolved data set [Stelzer, 19983. In an image with a size of 500 elements per line and 500 lines per image the field area will be reduced to about 5%. Obviously, the higher the resolution the smaller the field will be. A serious problem in CFM is that the dyes are essentially consumed during the illumination process. Fluorophores can only be excited a certain number of times before they become nonfluorescent and in some cases even toxic. This limits the number of photons one can get from a sample and the resolution that can be achieved. But this of course works in both directions. Provided an image has been recorded, it should be possible to estimate the number of photons and the resolution actually achieved. Good methods that estimate the perfect recording conditions and take all these aspects into account have not been implemented until now. F. Index Mismatching Efects

A serious problem that cannot be neglected is the spherical aberration due to mismatching of the refractive indices. One problem is that high-resolution

3 10


FIGURE7. Calculation of point-spread functions in optically mismatched systems. There are at least four elements in the optical path of a microscope that can have different refractive indices: the objective lens (nl), the immersion medium (nJ, the cover slip (n,). and the sample (n4). Ideally, all four are identical. In many cases in biology, the values for the refractive indices are n, = nz = n , = 1.518, and n4 = 1.33. This mismatch will cause a change in the position of the focal point. The actual position (AFP) is closer to the cover slip than the nominal focal position (NFP).

oil immersion objective lenses are used to observe specimens that are embedded in an aqueous medium (Fig. 7 ) . Another problem is that the refractive index varies inside large specimens, and recording conditions that may be valid in one spot may not work in others. This problem is important for quantitative microscopy. The usual case (high NA oil immersion lens, aqueous embedding medium) causes a shift of the actual focal plane toward the lens, hence a decrease of the axial distances. A decrease of the maximal intensity as well as an increase of the axial FWHM as one moves the focal plane further away from the refractive index transition plane degrades the image quality. For example, ten microns below the transition plane the axial FWHM is twice as large as under perfect conditions [Hell et al., 19931. The literature on this topic is quite rich [Gibson and Lanni, 1991; Torok, Hewlett, and Varga, 1997; Torok et al., 19961, and the effects are quite well



understood. But, despite a number of efforts [Visser, Groen, and Brakenhoff, 1991; White et al., 19961, it is unlikely that such effects will ever be correctable. The only reasonable solution to this serious problem is to use water immersion lenses. This attempts to evade the problem. The disadvantage is a lower resolution close to the cover slip (i.e., transition plane) but the advantage is of course a uniform, undistorted view of the complete specimen [Hell and Stelzer, 19951.

IV. IMPROVINGTHE AXIAL RESOLUTION The important role confocal fluorescence microscopy has in modern research is entirely due to its axial resolution, that is, its depth discrimination capability, which allows three-dimensional imaging. However, since a typical microscope objective covers only a small fraction of the full solid angle of 47t and thus focuses only a small segment of a spherical wave front, the axial resolution of a confocal microscope is always poorer than the lateral resolution. Hence the observation volume in any single-lens microscope is an ellipsoid elongated along the optical axis (Figs. 4 and 5). A large extent of the observation volume in that direction is equivalent to poor axial resolution. This elongation gives rise to certain artifacts [Stelzer et al., 19951, and any attempt to improve CFM should address the axial resolution and try to decrease its extent. A . Standing- Wave Fluorescence Microscopy

In standing-wave fluorescence microscopy (SWFM) two coherent, counterpropagating planar waves cross each other in the specimen volume [Bailey et al., 1993; Lanni, 1986; Lanni, Waggoner, and Taylor, 19861. The fluorophore in the specimen is excited by a series of axially spaced planar interference fringes, which are parallel to the focal plane of the microscope (Fig. 8). The fluorescence images are recorded as a function of the position of the object relative to the focal plane or relative to the phase of the two planar waves using sensitive cameras. SWFM is, therefore, not a confocal method. Although the SWFM can be regarded as an image-forming device since it produces an image, the full information must be reconstructed from a series of images. While the first instruments were based on total internal reflection (TIR) [Lanni, 19863 or a setup with a mirror [Bailey et al., 19931, a more powerful design uses two opposing microscope objectives [Bailey, Krishnamurthi, and Lanni, 1994; Lanni et al., 19931. If the two fields are polarized normal to their common plane of incidence






, * upper objective lens


[lower objective lens]

FIGURE8. Setup of a standing-wave fluorescence microscope with two objective lenses. Two coherent counter-propagating planar waves overlap in the specimen on the optical axis of the microscope. They create an axial interference fringe field consisting of nodal and anti-nodal planes with a spacing As. Using complementary angles of incidence H with respect to the optical axis, the nodal and anti-nodal planes are parallel to the focal plane. Fluorescence in the specimen is excited at anti-nodal planes. One of the objectives is used conventionally to form an image of the specimen in a camera.

(s-polarization), are of equal amplitude, and cross at complementary angles (0,n - 6) relative to the axis of the microscope, the resulting excitation intensity field varies sinusoidally along the microscope axis

I,,, Here k

= 2nn cos 6/A

= ZO[l - cos(2kz

+ $)I.

where I is the wavelength and n is the refractive



index, and 4 specifies the shift of the pattern relative to the specimen. The nodes and anti-nodes of this field, which are planes parallel to the focal plane, are AS = ____ 2n cos 8 apart. By controlling the angle 8, the node spacing Asmincan be varied down to a minimum value of

The relative position of the nodes and anti-nodes within the specimen can be adjusted without changing the node spacing by shifting the relative phase of one of the beams. The PSF of a SWFM is calculated by multiplying the PSF of a conventional epifluorescence microscope with Zexc(z): IhSWFhdX,

Y, Z)l2



y , z)I2 . I3 - cos(kz + 411.

The lateral resolution of a SWFM is determined by the conventional lateral properties of the objective lens. The enhanced axial resolution in SWFM is due to the modulation of the excitation field and thus not directly limited by diffraction. It can be estimated by noting that two small objects will be differentially excited by a 180" shift in the field if their axial separation is half the node spacing: Asmin/2= L/4n. Using ultraviolet light with a wavelength 1 = 365 nm, and considering a refractive index of water ( n = 1.33), the axial resolution limit is around 68 nm independent of the numerical aperture of the lens. However, this is only correct in the case of a very thin specimen that falls entirely within the depth-of-field of a high-NA objective lens and has a thickness t < ;1/4n. The sample thickness is less than half the node spacing. A controlled movement of a single node or anti-node within the object alternately illuminates stratified structures (optical subsectioning). In this case the axial resolution can be better than 1/8n, or 40-50 nm, which is one order of magnitude better than a confocal microscope. SWFM is particularly useful when the specimen is so thin that only one or two nodal planes cover its entire depth. In thicker objects several planes are illuminated at the same time, and their separation becomes very complicated. The SWFM has an axial discrimination, which is determined by the sum of detection PSFs as in any image-forming device. All layers, which are axially separated by half a wavelength, are observed simultaneously and there is no obvious way to resolve this ambiguity in a general manner. No successful reconstruction of a thick specimen has been reported until now.

3 14


A very interesting improvement is due to developments by Gustafson, Agard, and Sedat [1995; 19963, who use “white light.” It means the coherence length becomes extremely short and the axial interference pattern extends only over a few micrometers, which makes the reconstruction process much simpler. A serious problem in all high-resolution methods using interference is wave-front uniformity. Its effect is obvious in SWFM, and little can be done to account for the refractive-index heterogeneity inside a specimen and the resulting light scattering or wavefront distortion. Defocusing, aberrations, or irregularities on reflecting surfaces cause deformations in all planes. An optimization of the SWFM optics permits nodal plane flatness better than one-tenth of a wavelength peak-peak [Freimann, Pentz, and Horler, 19971 so long as the specimen does not change any phase relationships. Excitation field synthesis [Lanni et al., 1993) can be used to further improve the power of SWFM. Coherent light sets up an interference pattern that is most intense at the in-focus plane of the specimen and is sharply attenuated over sub-wavelength distances above and below this plane. As a result, the axial resolution can be improved to well below the wave-optical depth of field of the objective lens. B. 4Pi-Confocal Fluorescence Microscopy In a 4Pi-confocal fluorescence microscope (Fig. 9), a sample is illuminated and/or observed coherently through two coaxial objective lenses opposing each other but having a common focus [Hell and Stelzer, 1992b; Hell et al., 1994c; Lindek, Stelzer, and Hell, 19951. This technique is, in effect, an increase of the angular aperture, hence an improvement of the axial resolution. It was given the name 4Pi-confocal microscopy since the technique tries to come close to a perfect spherical wave with a solid angle of 471. In contrast to SWFM, the lenses focus the light into the focal volume. Lanni [1986], who introduced the SWFM, already noted that the convergent beam of an objective lens damps the axial lobes. As in SWFM, coherent illumination wave fronts can interfere in the focal volume, and the illumination intensity PSF is modulated along the optical axis. Depending on the phase difference cp, the interference in the geometrical focal plane is constructive (cp = 0, Zn,. . .), destructive (cp = TC, 371,. . .), or something intermediate (e.g., cp = n/2). The calculation of the 4Pi-illumination intensity PSF requires two counter-propagating amplitude PSFs, which are independently shifted in phase (Fig. 10). This is indicated by two functions whose difference is proportional to the phase: @,(a) - a2(p) cc eiq.




ldichroic bcamsplitter




... . .._... ..

t I


Fieurn 9. Schematic diagram of a 4Pi-confocal fluorescence microscope. Laser light is appropriately split into two coherent beams, which are deflected into two opposing objective lenses and thereafter interfere in the focal region. A phase-compensating device in the illumination path adjusts the relative phase of the beams to allow for constructive or destructive interference in the geometrical focus for 4Pi(A) contrasts. The same lenses collect the light, which now passes dichroic beamsplitters that separate the illumination from the detection path. A phase-compensating device in the detection path is needed to adjust the phase for 4Pi(B) and 4Pi(C) contrasts. The two beams are combined and interfere in the pinhole in front of the detector. For 4Pi(A) contrasts, either one lens is used to collect the detection light or both lenses collect the light incoherently.

In the case of constructive interference, the modulation of the PSF leads to a central main maximum with a full width at half maximum (FWHM) that is four to five times smaller than the FWHM of the envelope (Fig. lob): lh4Pi,ill(X,

y,2)I2 = [email protected]~fa)hillfx, Y ,2)

+ @,(B)hill(X,

Y , -z)l2.

Unfortunately, secondary maxima are also present along the optical axis, and their contribution or loss has to be taken into account when calculating the total illumination volume (Fig. 10). For reasons of symmetry, these considerations can also be extended to the detection of light. Provided the


[WI 1

4 0 0


0 LA w





0 -1.5

-0.6 -0.4 -0.2 0

0.2 0.4 0.6


FIGURE 10. Intensity PSFs of a 4Pi(A)-confocal fluorescence microscope. (a, b, c) The parameters for the calculations of the intensity point-spread functions are identical to those for the previous figures. (a) The two-dimensional intensity PSF for a 4Pi(A)-confocal fluorescence microscope. (b) The z-component of the illumination intensity PSF forms an envelope for the z-component of the 4Pi-illumination PSF. (c) The multiplication by the detection PSF damps the axial side lobes. The result is the z-component of the 4Pi(A)-confocal fluorescence microscope. (d) Using the higher numerical aperture of 0.9 but otherwise identical conditions produces considerably more fringes. Note that the z-axis scale has changed. ( e ) For the confocal 4Pi(A) microscope the multiplication by the detection PSF reduces the axial side lobes.



path length difference is smaller than the coherence length of the emitted light, the signal collected by the two objective lenses interferes in the detector, and the detection PSF presents an analogous interference pattern along the optical axis: lh4Pi.det(X?


= tyl(t)hdet(xj

Y , z>+ y2(t)hdet(x,

Y? -z)12.

In any confocal arrangement, the PSF of the microscope is the product of the illumination and detection PSFs. Consequently, combining the 4Pi-method with the confocal principles leads to several microscopies [Hell and Stelzer, 1992bl. In the 4Pi(A)-confocal microscope, illumination occurs coherently through two lenses, while the fluorescently emitted light is gathered incoherently using probably only one of the two lenses [Hell and Stelzer, 1992b1: lh4Pi(A)(X,

Y , z)12

= lhLPh.ill(X, Y,



Y , z)12.

In a 4Pi(B)-confocal microscope [Hell et al., 1994a], the illumination occurs through one lens, while both lenses are used to collect the emitted light coherently and have it interfere in the pinhole in front of the detector: lh4Pi(B)(X,

y , z)12 = lhilI(& y , z)lz ’ Ih4Pi,det(X, y , Z)l’.

In a 4Pi(C)-confocal microscope [Hell et al., 1994~1,illumination and detection are both coherent through two objective lenses: JkPi(C)(x,

Y , z)12 = t h 4 P i , i d ~ Y, , z)12 . I h P i , d e t ( x , Y , z)Iz.

These effects have been extensively verified in a series of papers [Lindek, 19931. An application that makes use of this technique has not been reported until now. An important improvement has been the use of two-photon excitation [Gu and Sheppard, 1995; Hanninen et al., 1995; Hell and Stelzer, 1992a; Hell et al., 1994b; Hell, Lindek, and Stelzer, 1994d; Lindek, Stelzer, and Hell, 19951. The resolution of a two-photon 4Pi(A)-confocal microscope is of course worse than its single-photon counterpart, but the main effect is that node spacing will only change in the illumination intensity PSF, and the subsequent multiplication by the intensity detection PSF reduces the axial lobes much more efficiently. The second major improvement has been the use of de-convolution and other more sophisticated computational methods that take the imaging process into account and numerically remove the axial side lobes [Hell et al., 1996b; Hell, Schrader, and Van Der Voort, 19971. C. Confocal Theta Microscopy

Confocal theta microscopy has been proposed [Stelzer and Lindek, 19941 to overcome the elongation of the observation volume and to achieve an



almost isotropic resolution. The idea is to change the spatial configuration of the illumination and detection volumes by using two different axes (Fig. 11). One objective lens is used to illuminate the sample. The other objective lens has its optical axis at an angle 9 to the illumination axis and is used to collect the emitted light. The improvement in axial resolution stems from an arrangement in which the detection axis is nearly orthogonal to the illumination axis. Then, the good lateral resolution in the detection path compensates the poor axial resolution in the illumination path and vice versa. The lateral resolution dominates the overall resolution, and the resulting observation volume is nearly isotropic [Lindek and Stelzer, 1994; Stelzer and Lindek, 19941. It can be shown that an azimuth angle of 9 = 90" results in the smallest confocal volume [Lindek et al., 1994bl. Angles between 70" and 110" result in still acceptable small volumes [Lindek, Pick, and Stelzer, 1994a1. There are different possibilities to realize such an optical arrangement. The most apparent way is the use of two (or even more) microscope objectives, which are positioned in such a way that an angle 9 of nearly 90"

FIGURE11. Principles of theta microscopy. Two optical axes are used for illumination and detection. The two PSFs are centered and tilted by 90" relative to each other. Since the axial extents of the PSFs are larger than their lateral extents, they overlap in a volume whose size is dominated by the lateral extents. Fractions of the volumes encircled by the PSFs are illuminated but not detected and vice versa.



between illumination and detection axis can be achieved [Lindek, Pick, and Stelzer, 1994a; Stelzer et al., 19951. Two more practical solutions that can be adapted to any standard confocal microscope are the theta double objective (TDO) [Stelzer and Lindek, 1996a) or a configuration where a single microscope objective lens can be used (single-lens theta microscope, SLTM) [Lindek, Stefany, and Stelzer, 1997; Stelzer and Lindek, 1996bl. In Fig. 12 one possible SLTM design is shown. The beam of the microscope objective is reflected by the surface of a horizontal mirror, and the fluorescence signal is detected by the same objective via a coated rectangular prism that is glued to the flat mirror (Figs. 12 and 13). This mirror unit is placed between the objective lens and its focal plane. It is deflecting illumination and detection light in such a way that their foci coincide and the detection axis is perpendicular to the illumination axis. Since the

FIGURE 12. Single-lens confocal theta microscope. The SLTM is based on a confocal fluorescence microscope. The light is focused into the sample that is above a horizontal mirror in the object plane 0.A second mirror deflects the light that is emitted at an angle of 90". While the illumination is on the optical axis, the detection is off-axis.



FIGURE 13. Mirror unit in a single-lens confocal theta microscope. The theta mirror unit consists of a horizontal mirror with an attached prism. The incident laser beam is deflected off the plane mirror and forms a focus above the front mirror surface. The fluorescence is emitted in all directions, but only a fraction can be detected due to the constraints of the optical system. As indicated by the dashed lines, it is deflected off the hypotenuse of the prism toward the microscope objective lens and is focused into an off-axis point in the image plane.

illumination light is reflected by the surface of the horizontal mirror, the resulting focus is above the mirror surface. The detection light is reflected by the hypotenuse face of the rectangular prism so that the detection axis is horizontal. Using further prisms on the flat mirror, different kinds of microscopies can be realized [Lindek, Stefany, and Stelzer, 19971. For example, a 4Pi(A)confocal fluorescence microscope can be built by arranging two prisms opposite to each other. The 4Pi-illumination is performed using the prisms, and the fluorescence light is detected using the flat mirror. The investigation of physical or biological specimens in a confocal theta microscope is performed by mounting them onto glass capillaries, which are scanned through the coincident foci of illumination and detection axes.


32 1

The extent and the overlap of the illumination PSF and the detection PSF determine the shape of the PSF. Therefore, the resulting observation volume can be considerably reduced by choosing an angle 9 % 90' between the axes used to illuminate the sample and to detect the emitted light. This means the good lateral resolution of the optical detection system compensates for the inferior axial resolution of the illumination system. Additionally, the good lateral resolution of the illumination system compensates for the inferior axial resolution of the detection system. The result is an almost spherical observation volume that results in an almost isotropic resolution (Fig. 14):

The confocal fluorescence-4Pi(A) theta intensity PSF is calculated by replacing the illumination PSF by the 4Pi(A) illumination PSF. The excellent lateral resolution of the detection intensity PSF damps the axial side lobes considerably (Fig. 14c, d). Since the lateral and axial resolution are proportional to the inverse of the numerical aperture and its square, respectively, theta microscopy is very well suited for optical systems that use low-NA, large-working-distance objective lenses [Stelzer et al., 19951. There are also some technical constraints, since two lenses have to come quite close, and although a water immersion lens with a numerical aperture of 0.9 has a better axial resolution than an oil immersion lens with NA = 1.4 its volume resolution is still worse [Stelzer and Lindek, 1994; Sheppard, 19953. More about theta microscopy is found in remarks by other authors [Gu, 1996; Hell, 1997; Shotton, 19951.

V. NONLINEAR IMAGING The methods described so far assume that the intensity of the fluorescence emission is linearly proportional to the intensity of the absorbed light. This assumption is usually correct so long as most of the fluorophore molecules are in the ground state. If the excitation intensity becomes too high, the linear response fails and less fluorescent light is emitted than expected. The other inherent nonlinear effect is that the fluorophores are consumed by bleaching. There is always a certain probability that fluorophores react with each other or fall victim to free oxygen radicals, that is, they are photobleached. Although such effects have been described and taken advantage of [Brakenhoff, Visscher, and Gijsbers, 1994; Sandison and Webb, 1994; Sandison et al., 1995b], they are not yet common.





-0.5 ___a)











-0.5 0


-1 -0.5




0.6 0.8



x [pml

FIGURE14. Intensity point-spread functions of confocal theta microscopes. All PSFs are calculated for a numerical aperture of 0.9, a refractive index of 1.518, an excitation wavelength of 488 nm and an emission wavelength of 530 nm. (a) Two-dimensional confocal intensity PSF for a confocal theta fluorescence microscope. (b) Comparison of the z-component of the PSF in a confocal and a confocal theta fluorescence microscope. The axial resolution is improved by a factor of 3.5. (c) Two-dimensional confocal intensity PSF for a 4Pi-confocal theta fluorescence microscope. (d) Comparison of the z-component of the PSF in a 4Pi-confocal and a 4Pi-confocal theta fluorescence microscope. The higher side lobes are well suppressed.

The most popular nonlinear imaging methods take advantage of absorbing more than a single photon. Another method is to deplete the ground states in certain areas with one laser and thereby prevent their observation with another probing laser. A . Two-Photon Excitation

During the research for her Ph.D. thesis, Maria Goppert-Mayer [1931] was the first to realize that the transition from the ground state into an excited



state can be accomplished by absorbing two photons, each having half the energy of the gap. She also realized that the probability for such a process is quite low and that high intensities are required to induce it. Thus it was not until 1961 that Kaiser and Garrett [1961] proved the existence of this effect in an experiment using lasers. The important aspect for microscopy is that the fluorescence emission Flhv after two-photon excitation (TPE) is proportional to the probability of absorbing two photons within a short period of time. This probability is proportional to the square of the excitation intensity, hence the fluorescence intensity is proportional to the square of the excitation intensity: Flhv




The PSF of a microscope that is based on TPE is thus the square of the illumination intensity PSF [Sheppard and Gu, 19901. The TPE-microscope has the same properties as a CFM but does not require a detection pinhole. In a CFM, having point illumination and point detection create the volume. In a TPE-microscope the volume is created by the intensity squared dependence of the fluorescence emission: IhZhv(X,


= (Ihill(X,


By adding a point detector the resolution is further improved [Stelzer et al., 19941: Ih2hv,r/(X,

Y , z)12






2 '

Denk, Strickler,and Webb [1990; 19911were the first to describe a microscope based on TPE. They used a colliding pulse mode-locked (CPM) laser to demonstrate the effect in test samples and biological specimens. More recent microscopes usually use lasers with short pulses in the femtosecond range and peak pulse powers in the kW range. TPE has been reported with picosecond- and cw-lasers [Hanninen, Soini, and Hell, 1994; Hell et al., 1994b), but their advantage for microscopy is not clear at the moment. Basically, all important dyes can be excited with laser lines between 700 and 1100 nm [Fischer, Cremer, and Stelzer, 1995; Xu and Webb, 19961.The wavelengths are thus about twice as long as those used for single-photon excitation. The longer wavelength is also the reason why the resolution of the TPE-microsco is worse than that of a CFM, which is only partially compensated by the l/$erm due to the squaring effect. A discussion of the possibilities to use two-photon emission has also been reported [Hell, Soukka, and Hanninen, 19951. TPE-microscopy will also not evade the problem of the refractive index mismatch just discussed. This has been verified theoretically [Hell et al., 1993; Hell and Stelzer, 1995;



Jacobsen et al., 1994; Jacobsen and Hell, 1995) and experimentally [Hell et al., 1993; Jacobsen et al., 1994; Hell and Stelzer, 1995; Jacobsen and Hell, 19951. (1) The main advantage of TPE-microscopy is that an illumination volume is created. This has been demonstrated in bleaching experiments using single- and two-photon excitation. Only in TPE, a hole (a volume in which the fluorophore has been bleached and is no longer excitable) is found in the location of the geometric focus [Denk, Strickler, and Webb, 1990; Stelzer et al., 19941. In single-photon excitation the fluorophore is bleached all along the optical axis, so only those volumes that are observed (Az = 50.45 pm) are actually excited in TPE-microscopy. (2) Caged fluorescent dyes and compounds such as caged FITC [Mitchison, 19891 and caged ATP [Kubitscheck, 19951 can be activated in a volume well defined in three-dimensional space. ( 3 ) The excitation light is well beyond the glass barrier of 380nm. (4) Fluorophores excited at higher wavelengths are usually quite efficient, and photomultipliers have a higher quantum efficiency in the blue region than in the red region. (5) Biological objects are less sensitive to near-infrared than to blue and ultraviolet light. (6) Since the excitation wavelength is higher, less light is scattered [Stelzer et al., 19941. B. Muhiphoton Excitation A further extension of TPE-microscopy is the use of three or more photons

to bridge the gap from the ground state to an excited state in a fluorophore. Several papers report results in this direction [Davey et al., 1995; Gryczynski, Malak, and Lakowicz, 1996; He et al., 1995; Hell et al., 1996a; Nakamura, 1993; Sheppard, 19961. However, an application has not been reported and an advantage is not obvious: Ihnhv(x,

Y , z)12 = (Ihilj(X, Y , z)121fl

Ihnhv,c/(X,Y ,

412= (Iil,(X, Y , 412)”. Ihdet(Xr y , 412.

The resolution will be further improved but the intensities are much higher (the excitation efficiency decreases), and the likelihood of inducing artifacts, for example, to damage the samples or the fluorophores, is also much higher. C. Stimulated-Emission-DepletionFluorescence Microscopy After the absorption of an excitation photon, the fluorescent molecules undergo a transition from a low vibronic level of the ground state So to a



vibrationally excited level of a higher singlet state S , . Within picoseconds this level decays to a low vibronic level of the first singlet state S , , which has a lifetime of 1-5 ns. This state is now susceptible to stimulated emission, provided that the wavelength of the light is in the emission spectrum of the dye. Hell and Wichmann first proposed stimulated-emission-depletion (STED) fluorescence microscopy [Hell and Wichmann, 19943. In STED microscopy the diffraction resolution limit is overcome by employing the effect of stimulated emission to inhibit the fluorescence process in the outer regions of the illumination PSF. Therefore, the spatial extent of the PSF in the focal plane is reduced, and as a consequence the resolution is increased. The stimulated emission is induced by an additional beam of light (STED beam), which depletes the excited singlet state S , of the fluorophores before fluorescence can take place. For stimulated emission it is advantageous to use pulsed lasers with pulses significantly shorter than the average lifetime of the excited state, that is, in the picosecond range. STED can thus be realized by two subsequent pulses, one for excitation and one for stimulated emission. This results in a temporal separation of excitation and stimulated emission. A set-up of a STED microscope should be possible by using two STED beams symmetrically offset by v = 1 . 2 2 ~with respect to the geometric focus. With this offset the first minimum of hSTED(u)coincides with the maximum of h,,(u). The resulting effective excitation PSF of the STED microscope is

where n , ( x , y, z ) is the spatial distribution of the fluorophore molecules in the S , state. The lateral resolution of STED microscopy is about 3-5 times higher than that of confocal microscopes. On the other hand, the increase in lateral resolution is associated with a reduction of the detectable intensity. D. Ground-State-Depletion Fluorescence Microscopy

Hell and Kroug [19951 introduced the ground-state-depletion (GSD) microscope. The idea is to deplete the ground states of the fluorophores in the outer regions of a PSF in such a way that no excitation-emission process is possible and that all emissions come from the innermost region of the PSF. In contrast to the STED microscope, the GSD microscope can be used with low-power, continuous-wave illumination. A lateral resolution in the range of 10-20nm, which would be an improvement by an order of magnitude compared to confocal microscopy, has been predicted. The calculation of the intensity PSF relies on three or more PSFs (Fig. 15). The





central PSF is responsible for the actual detection process while the outer PSFs cause the depletion of the outer regions: IhGdx, Y,z)12 = Ihildx, Y, z)12 .(1 - Ihill(x + a, Y , z)12Y (1 - IhiIl(x - P, Y,~11'). '

If the intensity of the laser used for GSD is higher than 10 MW/cm2, the first triplet state T, of the fluorescent molecules has to be taken into account. The molecules undergo a recycling process from the ground state So to the first singlet state S , and back to So. During each loop, a fraction is caught via intersystem crossing from S , to T, into the long-lived triplet state. This leads to the depletion of the ground state. The ground state remains depleted so long as the excitation beams are switched on. Considering an arrangement of GSD and excitation laser beam as pointed out for STED microscopy, one can estimate the resolution of GSD microscopy. For an offset of Au, = 1.2271, the first minima of the GSD beams coincide at the geometrical focus, hdepl(Ux) =

h,(U, - AU,)

+ h2(u, + Au,),

whereas the main maximum of one beam partly overlaps with the first side maximum of the other. Their intensity can be given by IhGSD(x9

Y , z)12 = Ihill(X,

Y,z)12 .(1 - nz(x, Y , z)),

where n,(x, y , z ) is the spatial distribution of the fluorophores in the triplet state. GSD fluorescence microscopy is limited by the relaxation of the dye from the triplet state, since this determines the maximum pixel scan rate. To record the neighboring point, one has to wait until all the molecules are back in the ground state again. Therefore, the maximum recording speed is about 200 kHz, which is of the same order as that of a confocal laser scan microscope.


A well-established field is the modification of the PSF by changing the light distribution in the illumination and/or detection aperture. In such arrangements, absorbing or phase-shifting plates are placed into the Fourier plane of the optical system. They will affect all beams, independent of the location of their focus in the object plane. Well known in conventional microscopy are phase contrast and differential interference contrast (DIC or Nomarski). Both can be used in laser scanning microscopy but are of no importance to CFM.



A main issue in any fluorescent microscopy is to have an efficient detection system. All fluorescence emitted by an excited molecule should be detected. Apertures in the illumination path can, however, have any desired transmittance so long as simply increasing the illumination intensity can compensate it. Apertures in the detection path must have a high transmission and are thus best avoided. The idea of using apertures was discussed extensively by Francia [1952], who realized that super-resolution can be pushed to an arbitrarily high level at the expense of signal. He also noted a conflict with some basic physical principles. Boivin [1952] offered a calculation more relevant to the theory of microscopy in which he determined the diffraction due to concentric arrays of rings. Finally, McCutchen [19643 discussed the three-dimensional intensity and phase distribution in the focus and presented a result for the smallest achievable diameter of a focal spot. Since 1982 a whole series of papers have been published that discuss the effects of the aperture modification on the lateral and axial resolution in CFM. In essence, it is probably fair to say that annular apertures will improve the lateral resolution but at the same time tend to decrease the axial resolution. An exception is the confocal theta microscope because its lateral and axial resolution are determined by the lateral extents of the illumination and the detection PSFs. Annular apertures will thus improve the lateral as well as the axial resolution in confocal theta fluorescence microscopy. More complicated apertures with a central and an annular opening have been proposed. A special class of these filters, where the area of the central opening and the outer annular opening are identical, have the interesting property of leaving the lateral resolution intact but improving the axial as shown by Martinez-Corral et al. resolution by at most a factor of [1995a]. An axial resolution gain results in loss of transmission. Other types of apertures with gradient intensity changes and super-resolving characteristics have been proposed, but their fabrication is somewhat difficult. To summarize, the resolution gains achievable with technically feasible pupil plane filters are at most a factor of two [Hegedus and Sarafis, 19861. A breakthrough has been the proof that any rotationally symmetric pupil filter can be approximated by a binarized set of concentric rings. This has pushed the field somewhat and resulted in special apertures for improved resolution in confocal microscopy. By this method the implementation of gradient intensity changes has become possible, since the fabrication of binarized versions of these filters with similar properties is easier. There are various different methods for achieving this binarization [Hegedus, 19851. The calculation of PSFs is identical to the methods we have mentioned so far. A pupil function P(p) defines the contributing areas. Intensity PSFs are calculated for the detection and the illumination path and then



multiplied: Il,(U,


= --i


2 z n sin2cc ~




P ( p ) J , ( o p ) e f i U p 2dp p - 1 d P ( p ) < 1.

In Fig. 16 we present an example described by Martinez-Corral et al. [1995a], which shows an axial resolution improvement by a factor of Finally, it should be mentioned that these apertures are located in conjugate Fourier planes. So, even with a single lens for illumination and detection, the pupil functions can be different.


VII. AXIALTOMOGRAPHY If the specimen is tilted in the focal plane it can be imaged from different directions, and this should result in an improved axial resolution. Skaer and Whytock [1975] first described the tilting of objects by a few degrees. Shaw [1990] and Shaw et al. [1989] presented a method in which the specimens were tilted by an angle of k90”. This allowed an investigation of the object from different sides. But the complexity of internal movements of structures in biological specimens and internal rearrangements during the tilting process seem to have limited the resolution improvement. Axial tomography is a microscopic technique first presented by Bradl et al. [1992] that tilts objects by any angle in the range 0 to 2z. Twoand three-dimensional images become recordable from any desired perspective. A special tilting device is used in which the specimens are adapted to a rotatable mounted capillary or a glass fiber (Fig. 17). The rotation axis of the object is usually parallel to the capillary axis. The resolution along all three axes depends on the technique that is used to observe the object. Using a CFM will provide the highest resolution. The method can be used to generate views from different directions and to use computational methods to reconstruct an improved view of the object [Larkin et al., 1994; Satzler and Eils, 1997; Shaw, 19901. However, a common problem in many fields in biology is the quantitative distance measurement of adjacent objects. In a conventional microscope the depth of field region of a 3D object is projected onto a two-dimensional image. Therefore, distances mostly appear to be shorter than they are. For the determination of the distance d of two points P and Q in the object space, one has to determine their coordinates. d

= J(x,


+ ( Y , - Y J 2 + ( z p- z*)”

The error Ad of the distance measurement depends on the spatial location










B -I


-0.6 -0.4 -0.2


0.2 0.4








FIGURE 16. Intensity point-spread function for a confocal fluorescence microscope with a multislit aperture. The PSFs are calculated for a numerical aperture of 1.4, a refractive index of 1.518, an excitation wavelength of 488nm, and an emission wavelength of 530nm. The central opening and the outer ring have the same area. Half of the illumination aperture is obstructed while the detection aperture is completely open. (a) Two-dimensional confocal intensity PSF. (b) Comparison of the r-component of the PSF with that of a confocal microscope. The lateral resolution has not changed significantly. (c) Comparison of the z-component of the PSF with that of confocal. The axial resolution has been improved.



FIGURE17. Setup of a capillary-based tilting device used for axial tomography. A capillary attached to a mounting block is placed into the focal region of the microscope objective lens. The capillary axis is chosen perpendicular to the optical axis. The capillary is located between the mounting block and the cover glass and is embedded in a buffer medium. The freely rotating axis is pointing out of the image.

of the two points and is determined by the measurement with the lowest precision. This is always the measurement along the optical axis. With the help of a tilting device as in axial tomography, the object can be moved into the focal plane of the microscope, and the distance between the two points can be measured accurately (Fig. 18). A. Three-Dimensional Measurements and Qualitative Analysis

Optical sectioning of the object from different views can perform the determination of object volumes. In one 3D data set the resolution along the optical axis is inferior compared to the resolution in the focal plane. Additional information from data sets acquired at different angles can be



d FIGURE 18. Distance measurement using axial tomography. (a) Two objects inside the capillary are in different planes but laterally indistinguishable. (b) By rotating the capillary around its axis and a careful translation both objects are moved into the focal plane. The distance d , between the two objects is thereby maximized.

included in the analysis to achieve the best possible resolution (ideally the lateral resolution). After the segmentation of distinct domains from each data set, their volumes can be determined. Moreover, the tilting of the sample allows observation of only one part of interest in the object from different views by angular sectioning. As in computer tomography, only one image is recorded at each angle. This



results in a number of image data sets of the same object in cylindrical coordinates, which have to be transformed into a Cartesian coordinate system. But it should be considered that after each tilting step the focal plane has to be readjusted (Fig. 18), so that in general there is no common point of reference in the different images. Nevertheless a qualitative visualization of the data is indeed possible after the alignment of the image data in such a way that for all images the same coordinates are allocated to the center of mass (bary center) in the object. Animated sequences from these “corrected” images provide a first impression of the three-dimensional organization of the object.

VIII. SPECTRAL PRECISION DISTANCE MICROSCOPY In this paper we have so far looked at resolution, that is, at two objects that emit light of the same wavelength. Therefore, an image can only show the sum of both objects. However, if two pointlike objects emit at distinct wavelengths, two independent images can be recorded that will each show a single object. The exact location of each object can be calculated using the intensity-weighted center of mass equivalents, and the distance of any two objects can be determined with a noise-limited precision [Burns et al., 19851. The same idea also applies to single-particle tracking in video sequences [Saxton and Jacobson, 19971. The issue is to determine distances from intensity-weighted center of mass equivalents in independently recorded images. Such distances can be below 20nm. Another example is the photonic force microscope [Florin et al., 19971, which uses the position of a single bead to determine a three-dimensional structure. The position of the bead inside the focal volume can be determined with a precision that is most likely below 10 nm. The distance of topological structures in an object can thus be determined with a resolution around 15 nm. An important example is the determination of the surface topology of integrated circuits using, for example, confocal reflection microscopes [Wijnaendts-van-Resandt, 19871. The height differences of planes that are sufficiently far apart can be determined with an unlimited precision. Here the surface roughness, the precision with which the position of the object can be measured along the optical axis, the reflectivity of the surface, and the coherence of the light source [Hell et al., 19911 limit the resolution. In the case of distance determination of small objects, the localization accuracy of these objects is given by the error of the coordinates for the intensity maximum. This intensity maximum corresponds to the intensity bary center. Therefore, the standard deviation of the intensity bary center



coordinates of a series of measurements can be used to express the localization accuracy of an object. In a biological specimen it was found that it can be estimated to about a tenth of the corresponding PSF-FWHM. Thus the accuracy of distance determination for objects that are more than 1 FWHM apart and possess the same spectral signature is considerably better than the optical resolution (as low as f20 nm) [Bradl et al., 1996a; Bradl et al., 1996b1. In order to measure distances of objects that are beyond 1 FWHM, “spectral precision distance microscopy” can be used [Bornfleth et al., 1998; Burns et al., 1985; Hausmann et al., 1998; Hausmann et al., 19971. As a prerequisite, pointlike objects have to carry a different spectral signature (e.g., different emission spectra or different fluorescence lifetimes). Diffraction limited images can be recorded independently for each object, and their intensity bary centers are determined independently from each other with a localization accuracy valid for targets of the spectral signature. Applying digital image analysis, the Euclidean distances between the intensity bary centers can be calculated. The resolution equivalent, that is, the smallest distance between targets of different spectral signature, determines the precision with which these distances can be measured. It depends strongly on the localization error, which is influenced by the optical resolution, the signal-to-noise ratio, the detector sensitivity, and the digitization [Bornfleth et al., 1998; Manders et al., 1996; Manders, Verbeek, and Aten, 19931. In particular, chromatic shifts between the objects of different spectral signatures have to be taken into account [Bornfleth et al., 1998; Manders, 19971. By using a combination of spectral precision distance microscopy with one of the other high-resolution microscopy techniques, a further improvement of localization accuracy down to the nanometer range seems possible.

IX. COMPUTATIONAL METHODS The idea behind deconvolution is best understood when one looks at transfer functions. Higher frequencies are less efficiently transferred than lower ones. Deconvolution essentially divides the Fourier transform of the image by the Fourier transform of the PSF and thereby amplifies the higher frequencies [Agard, 1983; Agard, 1984; Agard and Sedat, 19831. Due to noise this procedure is not straightforward, thus the noise has to be estimated as a function of the frequency [Shaw and Rawlins, 19911. In addition, the PSF must be estimated. This is done in a separate experiment [Carrington, 1994; Carrington et al., 1995a; Hiraoka, Sedat, and Agard, 19901 or calculated during the deconvolution process [Holmes, 19921. A



perfect deconvolution would produce a transfer function in the shape of a rectangle. Its Fourier transform is a sinc-function, which causes ringing visible in the edges of the reconstructed image. The solution is to employ additional filters that give the transfer function of the image a smoother shape. It has been mentioned that conventional microscopy has a constant integrated intensity. It is, therefore, unable to resolve axial edges. Deconvolution of conventional images works well with pointlike objects such as spheres and collections of spheres. Using integrating CCD cameras it can start with images having a high dynamic range, but since it produces information about a small volume this is given up during the computational process. Computational methods that claim to reassign the photons to the location from which they were emitted have also been used on images recorded with CFM [Van Der Voort and Strasters, 19951 and 4Pi-CFM [Hell, Schrader, and Van Der Voort, 1997; Schrader and Hell, 1996; Schrader, Hell, and Van Der Voort, 19963.They should have an even higher resolution [Shaw, 19951. A method that has been discussed several times but has never had a really strong impact is the use of more than one detector in the image plane [Bertero, Brianzi, and Pike, 1987; Reinholz et al., 19891. By detecting not only the central spot that is usually observed in a confocal microscope but also the complete pattern using either a CCD camera or several point detectors, the damping of the transfer functions toward higher frequencies can be compensated to a certain extent [Bertero et al., 1990; Reinholz and Wilson, 19941. By using a square instead of a spherical aperture the signal can be directed to a small number of detectors [Barth and Stelzer, 19941, which greatly simplifies the data collection procedure. The method is computationally intensive because several signals have to be combined to calculate an image. Since it reconstructs an almost square optical transfer function, extensive ringing occurs and has to be corrected.

X. SPINNING DISKS An alternative to beam scanning devices are Nipkow disks [Nipkow, 18841 that rotate either in primary or conjugated image planes [Kino, 1995; Petriin et al., 1968; Xiao and Kino, 19871. Their optical properties are identical to those of any other CFM, with a few minor exceptions [Wilson and Sheppard, 1984, pp. 157-168; Wilson, 19901. The advantage is that images can be observed directly through an eyepiece or integrated using CCD cameras. A very important development is to replace the disk by



arrays of microlenses and to use lasers instead of white light sources [Yin et al., 19951. It also works with TPE [Bewersdorf et al., 19981. This field has become interesting due t o developments found by Juskaitis et al. [1996], who realized that by subtracting certain patterned images from a bright field image a confocal image can be found on top of a constant background signal. XI. PERSPECTIVES OF CONFOCAL FLUORESCENCE MICROSCOPY This review concentrates on the scientific and not the technical developments that push CFM. New lasers, improved detectors, better scanners, faster computers and so forth will influence the application of the developments and may even make certain currently important developments obsolete. Improved computer interfaces that guide the user, perform many tasks automatically, and thus relieve the user from routine corrections will of course have a major impact. Such programs can help with proposals concerning the optimal wavelength and the objective lens. They provide hints concerning the optimal recording conditions and the effective resolution. Of course, this development is found in every scientific instrument. The development of lasers will have a dramatic influence. Since singlephoton excitation requires only laser power in the mW regime, small diode lasers that cover the range from blue to infrared will replace helium-neon and argon-ion lasers. Small, solid state diode pumped lasers will provide enormously high powers and extremely short pulses in the 10-femtosecond range and thus make two- and three-photon excitation much more widely available. In a few years simple laser light sources will cover the whole range from UV to IR with three or maybe even only two lasing units. Filters and shutters will be replaced to some extent by pulsing and switching lasers. The least progress can be expected in the field of detectors. Quenched avalanche photodiodes will push the operating frequency of solid state detectors and may even surpass photomultipliers, but this development is currently not too clear. Two-photon excitation is one of the most fascinating developments of the past few years. Unfortunately, it has generated a lot of hype, and this makes it somewhat difficult to estimate the actual impact on the application side. The situation is a bit similar to the early days of CFM, when many applications made use of it but in unconvincing ways. It took a number of years to identify the really useful areas of applications. TPE-microscopy is most likely useful for thick specimens ( > 100pm) that require a good resolution. It is of no advantage for the study of single cells. It is also the technique that can make best use of the many new lasers that already have



appeared and will continue to appear on the market. A further impact may result from the development of new dyes with an increased two-photon excitation cross section [Cheng et al., 19981. An important development will be the application of computational methods. They will certainly supplement the purely technical attempts to improve the resolution. Again, it is difficult to assess the impact since a direct comparison with C F M and TPE-microscopy is very time-consuming. While some of the techniques mentioned at this point could disappear, the computational methods will not. They will either be used to improve the quality (whatever that may be) of conventional images or they will be applied to images recorded with confocal-4Pi and theta microscopes. The problem with all methods that claim to achieve a higher resolution is the signal-to-noise ratio (SNR). A higher resolution is always equivalent to a decreased volume, and (assuming the dye concentration remains the same) this means that fewer fluorophores are observed. Hence the signal decreases, and to maintain the SNR the observation time must be extended [Stelzer, 19981. Another point is the significance of a higher resolution in biological specimens. Apart from y = “green fluorescent fusion proteins (GFPs) [Cubitt et al, 19951,’’ most methods are indirect and require a homogeneous penetration of the sample to guarantee a complete labeling of the sample. This, and of course the fact that target and ligand have finite sizes, put a lower limit to the actual distances that can be resolved. There is no doubt that SWFM, Theta-CFM, and 4Pi-CFM will be further developed and will take advantage of many technical achievements. Their impact in terms of applications is a different issue and depends on the acceptance by potential users and their direct advantage. This in turn depends on how those fields develop. One should not underestimate the power of conventional microscopy and the power that is provided by relatively simple techniques such as fluorescence resonant energy transfer [Bastiaens et al., 19961. 111 roto, confocal fluorescence microscopy has seen a rapid development since 1979 [Brakenhoff, Blom, and Barends, 19791. Instrumentation may have matured, but there is no reason to believe that the evolvement of new ideas has stopped. REFERENCES Abbt, E. (1873). Beitrage zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung. Arch. Mikroskopische Anutomie 9 41 1-468. Agard, D. A. (1983). A least-squares method for determining structure factors in threedimensional tilted-view reconstructions. J Mol Biol 167(4) 849-52. Agard, D. A. (1984). Optical sectioning microscopy: cellular architecture in three dimensions. Atin. Rev. Biophys. Bioeny. 13 191-219.



Agard, D. A. and Sedat, J. W. (1983). Three-dimensional architecture of a polytene nucleus. Nature 302 676-681. Airy, G. B. (1841). On the diffraction of an annular aperture. London. Edinburgh, und Dublin Philosophicul Magazine, Series 3 18 1- 10, 132- 133. Bacallao, R., Kiai, K., and Jesaitis, L. (1995). Guiding principles of specimen preservation for confocal Ruorescence microscopy. In Handbook of Bioloyicul Confocal Microscopy, J. B. Pawley, ed. Plenum Press, New York. pp. 311-325. Bailey, B., Farkas, D. L., Taylor, D. L., and Lanni, F. (1993). Enhancement of axial resolution in fluorescence microscopy by standing-wave excitation. Nuture 366 44-48. Bailey, B., Krishnamurthi, V., and Lanni, F. (1994). Optical subsectioning and multiple focal-plane imaging in the standing-wave fluorescence microscope. In Proceedings ofthe 22nd Annual Meeting of the Microscopy Society of Americu, G. W. Bailey and A. J. Garratt-Reed, eds.) San Francisco Press, Inc., New Orleans, Louisiana, pp. 158-159. Barth, M. and Stelzer, E. H. K. (1994). Boosting the optical transfer function with a spatially resolving detector in a high numerical aperture confocal reflection microscope. Optik 96(2) 53-58. Bastiaens, P. I. H., Majoul, I. V., Verveer, P. J., Soling, H.-D., and Jovin, T. M. (1996). Imaging the intracellular trafficking and state of the AB, quaternary structure of cholera toxin. EMBO 1.15(16) 4246-4253. Bertero, M., Boccdcci, P., Brakenhoff, G . J., Malfanti, F., and Van Der Voort, H. T. M. (1990). Three-dimensional image restoration and super-resolution in fluorescence confocal microscopy (of biological objects). In 1st International Conference on Confocal Microscopy und 2nd International Conference on 3-0Image Processing in Microscopy (Netherlands Soc. Electron Microscopy; Royal Microscopical Soc.; Int. Soc. Stereology; Univ. Amsterdum, 15-1 7 March 1989), Amsterdam, Netherlands, pp. 3-20. Bertero, M., Brianzi, P., and Pike, E. R. (1987). Super-resolution in confocal scanning microscopy. Inverse Probl. 3 195-212. Bewersdorf, J., Pick, R., Hell, S. W. (1998). Multifocal multiphoton microscopy. Opt. Lert. 23(9) 655-657. Boivin, A. (1952). On the Theory of Diffraction by Concentric Arrays of Ring-Shaped Apertures. J . Opt. Soc. Am. 42( 1) 60-64. Born, M. and Wolf, E. (1980). Principles of Optics. Pergamon Press, Oxford. Bornfleth, H., Satzler, K., Eils, R., and Cremer, C. (1998). High-precision distance measurements and volume-conserving segmentation of objects near and below the resolution limit in three-dimensional confocal Ruorescence microscopy. J . Microsc. 189(2) 118- 136. Bradl, J., Hausmann, M., Ehemann, V., Komitowski, D., and Cremer, C. (1992). A tilting device for three-dimensional microscopy: application to in situ imaging of interphase cell nuclei. J . Microsc. 168 47-57. Bradl, J., Hausmann, M., Schneider, B., Rinke, B., and Cremer, C. (1994). A versatile 2n-tilting device for fluorescence microscopes. J . Microsc. 176(3) 21 1-221. Bradl, J., Rinke, B., Krieger, H., Schneider, B., Haar, F.-M., Durm, M., Hausmann. M., and Cremer, C. (1996a). Comparative study of three-dimensional distance resolution in conventional, confocal-laser-scanning and axial tomographic fluorescence light microscopy. In BIOS Europe '96 Optoelectronic Reseurch and Techniques Optical Microscopic Techniques, Wien. Bradl, J., Rinke, B., Schneider, B., Edelmann, P., Krieger, H., Hausmann, M., and Cremer, C. (1996b). Resolution improvement in 3-D microscopy by object tilting. Europ. Microsc. A n d . 44 (November) 9- 11. Brakenhoff, G. J. (1979). Imaging modes in confocal scanning light microscopy (CLSM). J . Microsc. 117(2) 233-242.



Brakenhoff, G. J., Blom, P., and Bakker, C. (1978). Confocal scanning light microscopy with high aperture optics. In Opticu Hoy Y Mununu (Optics Present and Future). jhternat. Commission f o r Optics: Ministerio de lrr Defensa: et id. 10-17 Sept. 1978) J. Bescos, A. Hidalgo, L. Plaza, and J. Santamaria, eds. Instituto de Optica ‘Daza de Valdes,’ CSIC, Sociedad Espanola de Optica, Madrid, Spain, Madrid, Spain, pp. 215- 18. Brakenhoff, G. J., Blom, P., and Barends, P. (1979). Confocal scanning light microscopy with high aperture immersion lenses. J . Microsc. 117(2) 219-232. Brakenho5, G. J., Visscher, K., and Gijsbers, E. J. (1994). Fluorescence bleach rate imaging. J . Microsc. 175(2) 154-161. Burns, D. H., Callis, J. B., Christian, G. D., and Davidson, E. R. (1985). Strategies for attaining superresolution using spectroscopic data as constraints. Appl. Opt. 24(2) 154-161. Carlsson, K., Mossberg, K., Helm, P. J., and Philip, H. (1992). Use of UV excitation in confocal laser scanning fluorescence microscopy. Micron and Microscopicu Acta 23(4) 413-428. Carrington, W. A. (1994). Advances in computational fluorescence microscopy. In Proceedings 41 the 52nd unnular meeting of the Microscopy Society of Americu, G. W. Bailey and A. J. Garratt-Reed, eds. San Francisco Press, Inc., New Orleans, LA, pp. 926-927. Carrington, W. A,, Lynch, R. M., Moore, E. D. W., Isenberg, G., Fogarty, K. E., and Fay, F. S. (1995a). Superresolution three-dimensional images of fluorescence in cells with minimal light exposure. Science 268 1483-1487. Carrington, W. A,, Lynch, R. M., Moore, E. D. W., Isenberg, G., Fogarty, K. E., and Fredric, F. S. (1995b). Superresolution three-dimensional images of fluorescence in cells with minimal light exposure. Science 268(5216) 1483-1487. Cheng, P. C., Pan, S. J., Shih, A,, Kim, K.-S., Liou, W. S., and Park, M. S. (1998). Highly efficient upconverters for multiphoton fluorescence microscopy. J . Microsc. 189(3) 199-212. Cox, G. and Sheppard, C. (1993). Effects of image deconvolution on optical sectioning in conventional and confocal microscopes. Bioim. 1 82-95. Cox. I. J., Sheppard, C. J. R., and Wilson, T. (1982). Super-resolution by confocal fluorescent microscopy. Optik 60(4) 391-396. Cubitt, A. B., Heim, R., Adams, S. R., Boyd, A. E., Gross, L. A. and Tsien, R. Y.: (1995). Understanding, improving and using green fluorescent proteins. Trends Biochem. Sci. 20( 11) 448-455. Davey, A. P., Bourdin, E., Henari, F., and Blau, W. (1995). Three photon induced fluorescence from a conjugated organic polymer for infrared frequency upconversion. Applied Physics Letters 67(7) 884-885. Denk, W., Strickler, J. H., and Webb, W. W. (1990). Two-photon laser scanning fluorescence microscopy. Science 248(4951) 73-6. Denk, W., Strickler, J. P., and Webb, W. W. (1991). Two-photon laser microscopy. In Unired Stures Patent. Cornell Research Foundation, Inc., Switzerland, New York, p. 12. Fischcr, A,, Cremer, C., and Stelzer, E. H. K. (1995). Fluorescence of coumarins and xanthenes after two-photon absorption with a pulsed titanium-sapphire laser. Appl. Opt. 34(12) 1989-2003. Florin, E.-L., Pralle, A,, Horber, J. K. H., and Stelzer, E. H. K. (1997). Photonic force microscope based on optical tweezers and two-photon excitation for biological applications. Journul of Structural Biology 119 202-21 1. Francia, G. T. D. (1952). Super-Gain Antennas and Optical Resolving Power. Supplemento del Nuouo Cimento 9(9) 426-437. Freimann, R., Pentz, S., and Horler, H. (1997). Development of a standing-wave fluorescence microscope with high nodal plane flatness. J . Microsc. 187(3) 193-200. Frieden, 9. R. (1967). Optical transfer of the three-dimensional object. J . Opt. Soc. Am. 57(1) 56-66.



Gibson, S. F. and Lanni. F. (1991). Experimental test of an analytical model of aberration in an oil-immersion objective lens used in three-dimensional light microscopy. J . Opt. Soc. Am. A 8(10) 1601-13. Goodman, J. W. (1968). Introduction to Fourier Optics. McGraw-Hill, San Francisco. Goppert-Mayer, M. (1931). Uber Elementarakte mit zwei Quantenspriingen. Ann. Phys. 9 273-294. Gryczynski, I., Malak, H., and Lakowicz, .I.R. (1996). Multiphoton Excitation of the DNA stains DAPI and Hoechst. Bioim. 4 138-148. Gu, M. ( 1996). Principles of three-dimensional imaging in confocal microscopes. World Scientific Publishing Co., Singapore, p. 337. Gu, M. and Sheppard, C. J. R. (1995). Optical transfer function analysis for two-photon 4n confocal fluorescence microscopy. Opt. Commun. 114( 1-2) 45-49. Gustafsson, M. G . L., Agard, D. A,, and Sedat, J. W. (1995). Sevenfold improvement of axial resolution in 3D widefield microscopy using two objective lenses. S P I E Proc. 2412 147. Gustafsson, M. G. L., Agard, D. A,, and Sedat, J. W. (1996). 3 D widefield microscopy with two objective lenses: experimental verification of improved axial resolution. S P I E Proc. 2655 62. Hanninen, P. E., Hell, S. W., Salo, J., Soini, E., and Cremer, C. (1995). Two-photon excitation 4n confocal microscope: enhanced axial resolution microscope for biological research. Applied Physics Letters 66( 13) 1698-1700. Hanninen, P. E., Soini, E., and Hell, S. W. (1994). Continuous wave excitation two-photon fluorescence microscopy. J. Microsc. 176(3) 222-225. Hausmann, M., Esa, A,, Edelmann, P., Trakhtenbrot, L., Bornfleth, H., Schneider, B., Bradl, J., Ben-Bassat, I., Rechavi, G., and Cremer, C. (1998). Advanced precision light microscopy for the analysis of 3D-nanostructures of chromatin breakpoint regions towards a structurefunction relationship of the BCR-ABL region. In Fundamentals for rhe Assessment qf Risks from Environmentd Radiation, G. Horneck, ed. Plenum Press, NY. Hausmann, M., Schneider, B., Bradl, J., and Cremer, C. (1997). High-precision distance microscopy of 3D-nanostructures by a spatially modulated excitation fluorescence microscope. S P I E Proc. 3197 217-222. He, G. S., Bhawalkar, J. D., Prasad, P. N., and Reinhardt, B. A. (1995). Three photon absorption induced fluorescence and optical limiting effects in an organic compound, O p t Left. 20( 14) 1524- 1526. Hegedus, Z. S. (1985). Annular pupil arrays- application to confocal scanning. Opt. Acta 32 815-826. Hegedus, Z. S. and Sarafis, V. (1986). Superresolving filters in confocally scanned imaging systems. J . Opt. SOC.Am. A 3 1892- 1896. Hell, S. (1 997). Increasing the resolution of far-field fluorescence light microscopy by pointspread-function engineering. In Topics influorescence spectroscopy Nonlineur and two-photonindured.puorescence, J. Lakowicz, ed. Plenum Press, NY, pp. 361-426. Hell, S., Reiner, G., Cremer, C., and Stelzer, E. H. K. (1993). Aberrations in confocal fluorescence microscopy induced by mismatches in refractive index. J . Microsc. 169(3) 391 -405. Hell, S. and Stelzer, E. H. K. (1992a). Fundamental improvement of resolution with a 4n-confocal fluorescence microscope using two-photon excitation. Opt. Cornrnun. 93 277.282. Hell, S. and Stelzer, E. H. K. (1992b). Properties of a 4n-confocal fluorescence microscope. J . Opt. Soc. Am. A 9(12) 1259-2166. Hell, S., Stelzer, E. H. K., Lindek, S., and Cremer, C. (1994a). Confocal microscopy with an increased detection aperture: type-B 4n-confocal microscopy. Opt. Lett. 19(3) 222-224.


34 1

Hell, S., Witting, S., Schickfus, M. V., Resandt, R. W. W. V., Hunklinger, S., Smolka, E., and Neiger, M. (1991). A confocal beam scanning white-light microscope. J . Microsc. 163(2) 179- 187. Hell, S. W., Bahlmann, K., Schrader, M., Soini, A., Malak, H., Gryczynski, I., and Lakowicz, J. R. (1996a). Three-photon excitation in fluorescence microscopy. Journal of Biomedical Optics l(1) 71-74. Hell, S. W., Hanninen, P. E., Salo, J., Kuusisto, A., Soini, E., Wilson, T., and Tan, J. B. (1994b). Pulsed and cw confocal microscopy: a comparison of resolution and contrast. Opt. Cornmun. 113 144-152. Hell, S. W. and Kroug, M. (1995). Ground-state-depletion fluorescence microscopy: a concept for breaking the diffraction resolution limit. Appl. Phys. B 60 495-497. Hell, S. W., Lindek, S., Cremer, C., and Stelzer, E. H. K. (1994~).Measurement of the 4n-confocal point spread function proves 75 nm axial resolution. Applied Physics Letters 64( 11) 1335-1337. Hell, S. W., Lindek, S., and Stelzer, E. H. K. (1994d). Enhancing the axial resolution in far-field light microscopy: two-photon 4a-confocal fluorescence microscopy. J . Mod. Opt. 41(4) 675-68 1. Hell, S. W., Schrader, M., Hanninen, P. E., and Soini, E. (1996b). Resolving fluorescence beads at 100-200nm axial distance with a two-photon 4x-microscope operating in the near infrared. Opt. Commun. 128(4-6) 394. Hell. S. W., Schrader, M., and Van Der Voort, H. T. M. (1997). Far-field fluorescence microscopy with three-dimensional resolution in the 100-nm range. J . Microsc. 187( I ) 1-7. Hell, S. W., Soukka, J., and Hanninen, P. E. (1995). Two- and multiphoton detection as an imaging mode and means of increasing the resolution in far-field light microscopy: a study based on photon-optics. Bioim. 3(2) 64-69. Hell, S. W., and Stelzer, E. H. K. (1995). Lens aberrations in confocal fluorescence microscopy. In Hundbook ef biologicul confocal microscopy. J. B. Pawley. ed. Plenum Press. NY, pp. 347-354. Hell, S. W., and Wichmann, J. (1994). Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy. Opf. Lett. 19( 1 1) 780-782. Hiraoka, Y., Sedat, J. W., and Agard, D. A. (1987). The use of a charge-coupled device for quantitative optical microscopy of biological structures. Sciences 238 36-41. Hiraoka, Y., Sedat, J. W., and Agard, D. A. (1990). Determination of three-dimensional imaging properties of a light microscope system. Partial confocal behavior in epifluorescence microscopy. Biophys. J . 57(2) 325-33. Holmes, T. J. (1992). Blind deconvolution of quantum limited incoherent imagery: maximumlikelihood approach. J . Opt. Soc. Am. A 9(7) 1052-61. Holmes, T. J., Bhattacharyya, S., Cooper, J. A,, Hanzel, D., Krishnamurthi, V., Lin, W. C., Roysam, B., Szarowski, D. H., and Turner, J. N. (1995). Light-microscopic images reconstructed by maximum likelihood deconvolution. In Hundbook of Biological Confocul Microscopy, J. B. Pawley, ed. Plenum Press, NY, pp. 389-402. Hopkins, H. H. (1943). The Airy disk formula for systems of high relative aperture. Proceedings of the Physical Society 55 116- 128. Jacobsen, H., Hiinninen, P. E., Soini, E., and Hell, S. W. (1994). Refractive-index-induced aberrations in two-photon confocal fluorescence microscopy. J . Microsc. 176(3) 226-230. Jacobsen, H. and Hell, S. W. (1995). Effect of the specimen refractive index o n the imaging of a confocal fluorescence microscope employing high aperture oil immersion lenses. Bioim. 3 39--47. Juskaitis, R., Wilson, T., Neil, M. A. A., and Kozubek, M. (1996). Efficient real-time confocal microscopy with white light sources. Nature 383 804-806.



Kaiser, W. and Garrett, C. G . B. (1961). Two-photon excitation in CaF,:Eu*+. Phys. Rev. Lett. 7 229-231. Kino, G. S . (1995). Intermediate optics in Nipkow disk microscopes. In Handbook of Biological Confocal Microscopy, J . B. Pawley, ed. Plenum Press, NY, pp. 155-165. Krishnamurthi, V., Liu, Y.H., Bhattacharyya, S., Turner, J. N., and Holmes, T. J. (1995). Blind deconvolution of fluorescence micrographs by maximum likelihood estimation. Appl. O p t . 34(29) 6633-6647. Kubitscheck, U., Pratsch, L., Passow, H., Peters, R. (1995). Calcium pump kinetics determined in single erythrocyte ghosts by microphotolysis and confocal imaging. Biophys. J . 69( 1) 30-41. Lanni, F. (1986). Standing-wave fluorescence microscopy. In Applications offluorescence in the biomedical sciences, D. L. Taylor, A. S. Waggoner, R. F. Murphy, F. Lanni, and R. R. Birge, eds. Alan R Liss, Inc., pp. 505-521. Lanni, F., Bailey, B., Farkas, D. L., and Taylor, D. L. (1993). Excitation field synthesis as a means for obtaining enhanced axial resolution in fluorescence microscopes. Bioim. 1 187-196. Lanni, F., Waggoner, A. S., and Taylor, D. L. (1986). Standing wave luminescence microscopy, USA. Larkin, K., Cogswell, C. J., O’Byrne, J. W., and Arnison, M. R. (1994). High-resolution, multiple optical mode confocal microscope: 11. Theoretical aspects of confocal transmission microscopy. In SPIE Proc., San Jose, CA, pp. 55-63. Lindek, S. (1993). Auflosungsmessungen mit dem 4n-konfokalen Mikroskop. Universitat Heidelberg, lnstitut fur Angewandte Physik. Lindek, S., Pick, R., and Stelzer, E. (1994a). Confocal theta microscope with three objective lenses. Rev. Sci. Instr. 65( 11) 336773372, Lindek, S., Salmon, N., Cremer, C., and Stelzer, E. H. K. (1994b). Theta microscopy allows phase regulation in 4n(A)-confocal two-photon fluorescence microscopy. Optik 98( 1) 15-20. Lindek, S., Stefany, T., and Stelzer, E. H. K. (1997). Single-lens theta microscopy-a new implementation of confocal theta microscopy. J . Microsc. 188(3) 280-284. Lindek, S. and Stelzer, E. H. K. (1994). Confocal theta microscopy and 4n-confocal theta microscopy. In Three-dimensional microscopy image acquisition and processing. SPIE Proc., San Jose, CA, pp. 188-194. Lindek, S., Stelzer, E. H. K., and Hell, S. (1995). Two new high-resolution confocal fluorescence microscopies (4Pi, Theta) with one- and two-photon excitation. In Handbook of Biological Confocal Microscopy, J. B. Pawley, ed. Plenum Press, NY, pp. 417-430. Manders, E. M. M. (1997). Chromatic shift in multicolour confocal microscopy. J . Microsc. 185(3) 321-328. Manders, E. M. M., Hoebe, R., Strackee, J., Vossepoel, A. M., and Aten, J. A. (1996). Largest contour segmentation: a tool for the localization of spots in confocal images. Cytometry 23 15-21. Manders, E. M. M., Verbeek, F. J., and Aten, J. A. (1993). Measurement of co-localization of objects in dual-colour confocal images. J . Microsc. 169(3) 375-382. Marsman, H. J. B., Wicker, R., Resandt, R. W., Brakenhoff, G. J., and Blom, P. (1983). Mechanical scan system for microscopic applications. Rev. Sci. Instr. 54(8) 1047- 1052. Martinez-Corral, M., Andres, P., Ojeda-Castaiieda, J., and Saavedra, G. (1995a). Tunable axial superresolution by annular binary filters. Application to confocal microscopy. Opt. Commun. 119(5-6) 491-498. Martinez-Corral, M., Andres, P., and Zapata-Rodriguez, C. J. (199513). Improvement of three-dimensional resolution in confocal scanning microscopy by combination of two pupilplane filters. Optik lOl(3) 1-4.



McCutchen, C. W. (1964). Generalized Aperture and the Three-Dimensional Diffraction Image. 1. Opt. Soc. Am. 54(2) 240-244. McCutchen, C. W. (1967). Superresolution in microscopy and the Abbe resolution limit. J . Opt. Sot. Am. 57(10) 1190-1192. Merdes, A., Stelzer, E. H. K., and De Mey, J. (1991). The three-dimensional architecture of the mitotic spindle, analyzed by confocal fluorescence and electron microscopy. J Electron Microsc Tech l8(l) 61-73. Minsky, M. (1961). Microscopy apparatus. In United States Parent Ofice, US, p. 9. Minsky, M. (1988). Memoir on inventing the confocal scanningmicroscope. Scanning 10 128-138. Mitchison, T. J. (1989). Journal Cell Biology 109: 637. Montag, M., Kukulies, J., Jorgens, R., Gundlach, H., Trendelenburg, M. F., and Spring, H. (199I). Working with the confocal scanning UV-laser microscope: specific DNA localization at high sensitivity and multiple-parameter fluorescence. J . Microsc. 163(2) 201-210. Nakamura, 0. (1993). Three-dimensional imaging characteristics of laser scan fluorescence microscopy: two-photon excitation vs. single-photon excitation. Optik 93( 1) 39-42. Nipkow, P. (1884). Elektrisches Teleskop. In Kaiserfische.7 Patentamt (Deutsches Patentamr), Germany. Palmieri, S. L., Peter, W., Hess, H., and Scholer, H. R. (1994). Oct-4 transcription factor is differentially expressed in the mouse embryo during establishment of the first two extraembryonic cell lineages involved in implantation. Developmental Biology 166(1) 259-267. Petran, M., Hadravsky, M., Egger, M. D., and Galambos, R. (1968). Tandem-scanning reflected-light microscope. J. Opt. Soc. Am. 58 661-664. Petroll, W. M., Cavanagh, H. D., Lemp, M. A,, Andrews, P. M., and Jester, J. V. (1992). Digital image acquisition in in vivo confocal microscopy. J . Microsc. 165 61 -69. Reinholz, F., Schiitt, W., Griimmer, G., Kuhlmann, F., and Kraft, S.-K. (1989). A new powerful mode of laser scanning microscopy. Optik 82 165-168. Reinholz, F. and Wilson, T. (1994). Image enhancement by tracking and sampling in the detection plane. Optik 96(2) 59-64. Reinsch, S., Eaton, S., and Stelzer, E. (1998). Confocal microscopy of polarized epithelial cells. In Cell biology: a laboratory handbook, J. E. Celis, ed. Academic Press, pp. 170-178. Sandison, D. R., Piston, D. W., Williams, R. M., and Webb, W. W. (1995a). Quantitative comparison of background rejection, signal-to-noise ratio, and resolution in confocal and full-field laser scanning microscopes. Appl. Opt. 34( 19) 3576-3588. Sandison, D. R. and Webb, W. W. (1994). Background rejection and signal-to-noise optimization in confocal and alternative fluorescence microscopes. Appl. Opt. 33(4) 603-61 5. Sandison, D. R., Williams, R. M., Wells, K. S., Strickler, J., and Webb, W. W. (1995b). Quantitative fluorescence confocal laser scanning microscopy (CLSM). In Handbook qf Biological Confocal Microscopy, J. B. Pawley, ed. Plenum Press, NY, pp. 39-53. Satzler, K. and Eils, R. (1997). Resolution improvement by 3-D reconstructions from tilted views in axial tomography and confocal theta microscopy. Bioim. 5 171-182. Saxton, M. J. and Jacobson, K. (1997). Single-particle tracking: applications to membrane dynamics. Annu Rev Biophys Biomol Struct 26 373-399. Schrader, M. and Hell, S. W. (1996). 4n-confocal images with axial superresolution. J . Microsc. 183(2) 189-193. Schrader, M., Hell, S. W., and Van Der Voort, H. T. M. (1996). Potential of confocal microscopes to resolve in the 50- 100 nm range. Applied Physics Letters 69(24) 3644-3646. Shaw, P. J. (1990). Three-dimensional optical microscopy using tilted views. J . Microsc. 158(2) 165-172. Shaw, P. J. (I 995). Comparison of wide-field/deconvolution and confocal microscopy for 3D imaging. In Hundhook ofBiologica1 Confocal Microscopy,J. B. Pawley, ed. Plenum Press, NY, pp. 373-387.



Shaw, P. J., Agard, D. A,, Hiraoka, Y., and Sedat, J. W. (1989). Tilted view reconstruction in optical microscopy. Three-dimensional reconstruction Drosophila melanogaster embryo nuclei. Biophys. J . 55(1) 101-1 10. Shaw, P. J. and Rawlins, D. J. (1991). The point-spread function of a confocal microscope: its measurement and use in deconvolution of 3-D data. J . Microsc. 163(2) 151-165. Sheppard, C. J. R. (1977). The use of lenses with annular aperture in scanning optical microscopy. Optik 48(3) 329-34. Sheppard, C. J. R. (1995). Fundamental reduction of the observation volume in far-field light microscopy by detection orthogonal to the illumination axis: confocal theta microscopy Comment. Opt. Commun. 119 639-695. Sheppard, C. J. R. (1996). Image formation in three-photon fluorescence microscopy. Bioim. 4 124- 128. Sheppard, C. J. R. and Gu, M. (1990). Image formation in two-photon fluorescence microscopy. Optik 86(3) 104-6. Sheppard, C. J. R. and Wilson, T. (1981). The theory of the direct-view confocal microscope. J . Microsc. 124 107-17. Shotton, D. M. (1995). Electronic light microscopy: present capabilities and future prospects. Hisrochem. Cell. Biol. 104 97- 137. Skaer, R. J. and Whytock, S. (1975). Interpretation of the three-dimensional structure of living nuclei by specimen tilt. J . Cell Sc. 19 1-10. Slomba, A. F., Wasserman, D. E., Kaufman, G. I., and Nester, J. F. (1972). A laser flying spot scanner for use in automated fluorescence antibody instrumentation. Journal qf the Assoc. for the Adv. of Med. Instrumentation 6(3) 230-234. Speicher, M. R., Ballard, S. G., and Ward, D. C. (1996). Karyotyping human chromosomes by combinatorial multi-fluor FISH. Nature Genetics 12 368-375. Stelzer, E. H. K. ( 1994). Designing a confocal fluorescence microscope. In Computer-assisted multidimensional microscopies, P. C. Cheng, T. H. Lin, W. L. Wu, and J. L. Wu, eds. Springer, NY, pp. 33-51. Stelzer, E. H. K. (1995). The intermediate optical system of laser-scanning confocal microscopes. In Hundbook of Bioloyical Confocal Microscopy, J. B. Pawley, ed. Plenum Press, NY, pp. 139- 154. Stelzer, E. H. K. (1997). Three-dimensional light microscopy. In Handbook of Microscopy: Applications in materials science, solid-state physics and chemistry, S. Amelinckx, D. v. Dyck, J. v. Landuyt, and G. v. Tendeloo, eds. VCH-Verlag, Weinheim, NY, pp. 71-82. Stelzer, E. H. K. (1998).Contrast, resolution, pixelation, dynamic range and signal-to-noise ratio: fundamental limits to resolution in fluorescence light microscopy. J . Microsc. 189( I ) 15-24. Stelzer, E. H. K., Hell, S., Lindek, S., Stricker, R., Pick, R., Storz, C., Ritter, G., and Salmon, N. ( I 994). Nonlinear absorption extends confocal fluorescence microscopy into the ultraviolet regime and confines the illumination volume. Opt. Commun. 104 223-228. Stelzer, E. H. K. and Lindek, S. (1994). Fundamental reduction of the observation volume in far-field light microscopy by detection orthogonal to the illumination axis: confocal theta microscopy. Opt. Commun. 111 536-547. Stelzer, E. H. K., and Lindek, S. (1996a). Konfokales Mikroskop mit einem DoppelobjektivSystem. In Deutsches Patentumt. EMBL, Germany. Stelzer, E. H. K., and Lindek, S. (1996b). Strahlumlenkeinheit zur mehrachsigen Untersuchung in einem Mikroskop, insbesondere Rasterlichtmikroskop. In Deutsches Putentamt. EMBL, Germany. Stelzer, E. H. K., Lindek, S., Albrecht, S., Pick, R., Ritter, G.. Salmon, N. J., and Stricker, R. (1995). A new tool for the observation of embryos and other large specimens: confocal theta fluorescence microscopy. J . Microsc. 179(1) I - 10.



Stelzer, E. H. K., Marsman, H. J. B., and Resandt, R. W. (1986). A setup for a confocal scanning laser interference microscope. Optik 73( 1) 30-33. Stelzer, E. H. K., Merdes, A,, and De Mey, J. (1991). Konfokale Fluoreszenzmikroskopie in der Zellbiologie. Biol. in uns. Z . 1 19-25. TorGk, P., Hewlett, S. J., and Varga, P. (1997). The role of specimen-induced spherical aberration in confocal microscopy. J . Microsc. 188(2) 158-172. Torok, P., Varga, P., and Booker, G. R. (1995). Electromagnetic diffraction of light focused through a planar interface between materials of mismatched refractive indices: structure of the electromagnetic field. 1. J . Opt. Soc. Am. A 12(10) 2136-2144. Torok, P., Varga, P., Konkol, A., and Booker, G. R. (1996). Electromagnetic diffraction of light focused through a planar interface between materials of mismatched refractive indices: structure of the electromagnetic field. 11. J . Opt. Soc. Am. A 13(11) 2232-2238. Tsien, R. Y. and Waggoner, A. (1995). Fluorophores for confocal microscopy. In Hundbook of Biologicul Conjbcul Microscopy, J. B. Pawley, ed. Plenum Press, NY, pp. 267-279. Van Der Voort, H . T. M., Brakenhoff, G. J., Valkenburg. J. A. C., and Nanninga, N. (1985). Design and use of a computer controlled confocal microscope for biological applications. Srunniny 7 66-78. Van Der Voort, H. T. M., and Strasters, K. C. (1995). Restoration of confocal images for quantitative image analysis. J . Microsc. 178 165-181. Van Meer, G., Stelzer, E. H. K., Wijnaendts-van-Resandt, R. W., and Simons, K. (1987). Sorting of sphingolipids in epithelial (Madin-Darby canine kidney) cells. J . Cell Biol. 105 1623- 1635. Visser, T. D., Groen, F. C. A,, and Brakenhoff, G . J. (1991). Absorption and scattering correction in fluorescence confocal microscopy. J . Mioosc. 163 189-200. White, N. S., Errington, R. J., Fricker, M. D., and Wood, J. L. (1996). Multidimensional fluorescence microscopy: optical distortions in quantitative imaging of biological specimens. In Fluorescence Microscopy und Fluorescent Probes. J. Slavik, ed. Plenum Press, NY, pp. 47-56. Wijnaendts-van-Resandt, R. W. ( 1 987). Application of confocal beam scanning microscopy to the measurement of submicron structures. In S P I E Proceedings, pp. 101- 106. Wijnaendts-van-Resandt, R. W., Marsman, H. J. B., Kaplan, R., Davoust, J., Stelzer, E. H. K., and Stricker, R. (1985). Optical fluorescence microscopy in three dimensions: microtomoscopy. J . Microsc. 138(1) 29-34. Wilke, V. (1983). Laser scanning in microscopy. S P I E Proc. 397 164-172. Wilke, V. (1985). Optical scanning microscopy- the laser scan microscope. Scunniny 7 88-96. Wilson, T. (1990). Detector arrays in confocal scanning microscopes. Scunning Microscopy 4( 1) 21 -4. Wilson, T. and Hewlett, S. J. (1990). The use of annular pupil plane filters to tune the imaging properties in confocal microscopy. J . Mod. Opt. 37( 12) 2025-46. Wilson, T. and Sheppard, C. J. R. (1984). Theory and practice of scanning optical microscopy. Academic Press, London. Xiao, G . Q. and Kino, G . S. (1987). A real-time confocal scanning optical microscope. In Scanning Imaging Technology (SPIE; Comiie Brlgr Opt.; IEEE; el al., 2-3 April 1987). The Hague, Netherlands, pp. 107- 13. Xu, C. and Webb, W. W. (1996). Measurement of two-photon excitation cross sections of molecular fluorophores with data from 690 to 1050nm. J . Opt. Soc. Am. B 13(3) 481-491. Yin, S., Lu, G., Zhang, J., Yu, F. T. S., and Mait, J. N. (1995). Kinoform-based nipkow disk for a confocal microscope. Appl. O p f . 34(25) 5695-5698. Zink, D., Cremer, T., Saffrich, R., Fischer, R., Trendelenburg, M. F., Ansorge, W., and Stelzer, E. H. K. (1998). Structure and dynamics of human interphase chromosome territories in vivo. Hum. Genet. 102 241-251.