The physics behind the settings you will choose at the microscope. Every number here comes from a stated formula or a named source; every plot is a model, labeled as such, not a measurement.
A fluorophore absorbs a photon and is promoted from its ground electronic state (S0) to an excited singlet state (S1). Within picoseconds it loses some of that energy as heat by relaxing to the lowest vibrational level of S1. A few nanoseconds later it returns to S0, and the energy it releases leaves as a photon. Because energy was lost in between, the emitted photon has less energy, and therefore a longer wavelength, than the absorbed one. That shift between the absorption and emission peaks is the Stokes shift, and it is what lets a dichroic mirror and an emission filter separate the faint fluorescence from the bright excitation light.[1]
Two numbers describe a fluorophore's usefulness. The quantum yield is the fraction of absorbed photons that are re-emitted as fluorescence; the rest are lost to non-radiative decay. The fluorescence lifetime is the average time a molecule spends in S1, typically 1 to 10 ns for the dyes used in cell biology. Lifetime depends on the molecule's environment, which is why lifetime-based tools can separate two dyes that emit at the same color, or distinguish a labeled structure from autofluorescence.[1]
From S1 the molecule can also cross into a long-lived triplet state. Triplet molecules react readily with oxygen and with neighboring molecules; those reactions are a major route to photobleaching.[1][9]
Jablonski diagram, simplified. The absorption arrow is longer than the emission arrow: the difference is lost as heat inside S1 and S0, so the emitted photon has less energy and a longer wavelength (Stokes shift, right). Energies not to scale; spectra are generic shapes.[1]
Topic 02
The confocal principle: one point lit, one point seen
A widefield fluorescence microscope illuminates the whole field and collects light from every depth at once, so structures above and below the focal plane add a blurred haze to the image. Marvin Minsky's 1957 design removed that haze with two ideas: illuminate a single point, and place a small aperture, the pinhole, in a plane conjugate to the focal plane in front of the detector. Light from the illuminated point focuses through the pinhole; light from above or below the focal plane comes to a focus in front of or behind the pinhole, arrives at the plate spread out, and is mostly blocked. Between the dichroic and the pinhole, an emission filter, or on spectral systems a prism or grating with an adjustable slit, selects the wavelength band to be detected.[2]
Because only one point is observed at a time, the image is built by scanning: two galvanometer mirrors sweep the focused laser spot across the sample line by line while the detector, usually a photomultiplier, records one intensity value per pixel. The result is an optical section, a two-dimensional image of a thin slab, and stepping the focus produces a Z-stack.[4]
Mathematically, the confocal point spread function is the product of the illumination PSF and the detection PSF as seen through the pinhole: PSFconfocal = PSFill × PSFdet. With a very small pinhole the two are nearly identical and the product is narrower than either, which is why an ideal confocal resolves about 1.4 times finer than a widefield microscope in both lateral and axial directions. With a pinhole of one Airy unit or larger, the resolution is set by the illumination spot alone, and the pinhole's job is depth discrimination.[4][5]
Schematic confocal beam path after Minsky and the Zeiss monograph. The objective focuses the collimated laser into a double cone; emission from the focus leaves the objective collimated and the tube lens brings it to a point in the pinhole. Angles are exaggerated.[2][4]
Topic 03
The pinhole: trading signal for sectioning
Move the slider to change the pinhole diameter in Airy units. The module recomputes, from the formulas listed below it, how much in-focus light reaches the detector, how much out-of-focus light is rejected, and how the lateral and axial response change.
0.1 to 4 Airy units. Arrow keys move in 0.05 AU steps.
Used for the "out-of-focus light rejected" readout.
Fraction of a point emitter's light that passes the pinhole, versus its distance from the focal planemodel
Axial response (point object)model
Lateral response (point object)model
One Airy unit here
—object-plane diameter, 1.22 λem/NA
In-focus signal through the pinhole
—relative to a fully open pinhole
Out-of-focus light rejected
—for an emitter 1.0 µm from focus
Optical section, axial FWHM
—point object, computed
Lateral FWHM
—computed; reference values below
Zeiss eq. (4) slice thickness
—fluorescent plane, valid for PH ≥ 1 AU
Abbe limit λexc/(2NA)
—finest resolvable period, see note
Rayleigh 0.61 λexc/NA
—used by this page to define the Airy unit
Formulas used on this page
Airy unit (AU)1 AU = 1.22 λem / NADiameter of the Airy disk (first dark ring) in the object plane, in nm; λem in nm, NA dimensionless. The physical pinhole is this value times the magnification between sample and pinhole, which the software handles.[3][4]
Optical coordinatesv = 2π NA r / λu = (8π nz / λ) sin²(α/2), sin α = NA/nDimensionless lateral (v) and axial (u) coordinates; r, z, λ in nm. The high-aperture form of u follows Sheppard and Matthews.[6]
In-focus signal through the pinholeE(vp) = 1 − J0²(vp) − J1²(vp)Fraction of the Airy pattern's energy inside a circle of radius vp (the pinhole radius in optical units); dimensionless. At 1 AU, vp = 3.83 and E = 0.84.[3]
Out-of-focus transmissionI(u,v) = |2 ∫01 J0(vρ) e−iuρ²/2 ρ dρ|², T(u) = ½ ∫0vpI(u,v) v dvScalar diffraction pattern of a point emitter at defocus u, integrated over the pinhole; T is dimensionless and the "rejected" readout is 1 − T. Computed numerically.[3]
Confocal point response (plotted)axial: Iexc(uexc, 0) · T(uem) lateral: Iexc(0, vexc) · [Airyem ⊗ pinhole](vem)Illumination PSF times detection PSF integrated over the pinhole; normalized to its peak. FWHM in nm is read from these curves. For PH → 0 and PH ≫ 1 AU this reproduces the Zeiss coefficients 0.37/0.64 and 0.51/0.88 below.[4][5]
Optical slice thickness of a fluorescent plane (Zeiss eq. 4)FWHMslice = √[ (0.88 λem / (n − √(n² − NA²)))² + (√2 n PH / NA)² ]PH = object-side pinhole diameter in nm (here, the slider value × 1 AU). Valid for PH ≥ 1 AU, where geometric optics dominates; it describes a thin fluorescent plane, not a point, so it is larger than the point-object axial FWHM above.[4]
Lateral resolution limitsAbbe: d = λ / (2 NA) Rayleigh: r = 0.61 λ / NABoth in nm. Abbe's d is the smallest grating period a lens can transmit; Rayleigh's r is the separation at which two point images are "just resolved", equal to the Airy disk radius. This page uses the Rayleigh/Airy value, because the Airy unit is defined from it (1 AU = 2 × 0.61 λ/NA); Abbe's limit is shown for comparison and is smaller by the factor 0.5/0.61.[7][8][3]
About the model. All curves are computed from the scalar, uniform-pupil diffraction model in Born and Wolf with the high-aperture axial coordinate, for a single point emitter without aberrations, refractive-index mismatch, scattering, or noise. They are illustrative of the trends, not measurements of any Core instrument. Real point spread functions, measured with sub-resolution beads, are always broader. The in-focus signal fraction ignores the illumination side, because a point object is fully illuminated either way.
Topic 04
Resolution and the Airy disk
A lens cannot focus light to a point. Diffraction at the circular aperture of the objective spreads the image of a point source into a bright central disk surrounded by faint rings: the Airy pattern. The radius of the first dark ring is 0.61 λ/NA, where λ is the wavelength and NA the numerical aperture of the objective. Lord Rayleigh proposed that two points are "just resolved" when the center of one Airy disk falls on the first dark ring of the other; at that separation the summed intensity dips to about 74 % between the peaks.[3][8]
Ernst Abbe derived a related limit from a different question: what is the finest periodic structure a lens can transmit? The answer is a grating period of λ/(2 NA). The two criteria differ only in their coefficient (0.50 versus 0.61) and in what they measure. A third, practical descriptor is the full width at half maximum of the point image, about 0.51 λ/NA for a widefield microscope. All three say the same thing: resolution improves with shorter wavelength and larger NA, and nothing else about the lens matters once it is aberration-free.[7][4]
In a confocal microscope the illumination spot and the detection pinhole both shape the point image. With the pinhole closed to a small fraction of an Airy unit the two Airy patterns multiply and the FWHM shrinks to about 0.37 λ̄/NA, an improvement of roughly 1.4×. In practice pinholes are set near 1 AU to keep enough signal, and the lateral resolution is then essentially that of the illumination spot. Move the pinhole slider below to see both cases.[4][5]
Distance between the two emitters in the sample. Compare with the Rayleigh distance in the readouts.
Large pinhole ≈ widefield point image; closing it toward a fraction of an Airy unit sharpens the image, at the cost of signal (see topic 3).
Two point imagesmodel
Intensity profile through both pointsmodel
Rayleigh 0.61 λ/NA
—Airy disk radius
Abbe λ/(2 NA)
—finest grating period
FWHM of the point image
—computed for the chosen pinhole
Dip between the peaks
—midpoint intensity relative to the peaks
About the model. Incoherent sum of two confocal point images, each computed as the excitation Airy pattern times the emission Airy pattern integrated over the chosen pinhole (same model as topic 3, with λexc = λem = λ). The 2D panel is 2 µm across and uses a gamma stretch so the faint rings are visible; the profile is linear. Not a measurement.[3]
Topic 05
Pixel size: the Nyquist criterion
The optics set how fine a detail reaches the detector; the pixel size sets whether that detail survives digitization. The Nyquist–Shannon theorem says a signal is captured without loss only if it is sampled at more than twice its highest spatial frequency, that is, with at least two pixels per smallest resolvable period. Sample more coarsely and fine structures are not merely blurred but aliased: they reappear as false coarser patterns that no processing can undo.[9]
Two conventions are in common use, and they differ by about a factor of two, so it matters which one a protocol means. The practical rule in the Zeiss monograph takes the FWHM of the point image as the resolvable detail and sets the pixel size to half of it: dpix ≤ 0.5 × FWHMlat ≈ 0.25 λexc/NA. The bandwidth rule used by deconvolution software takes the true cut-off frequency of the optical transfer function; for a confocal that cut-off is roughly twice the widefield value, giving Δx = λexc/(8 NA). The same source notes that confocal images can in practice be sampled up to about 1.7× more coarsely than this strict value when they will not be deconvolved.[4][10]
On a point scanner the pixel size is set by the scan zoom and the number of pixels per line, not by a camera, so it can be matched to any objective. Oversampling by a large factor costs time and light without adding information, and that light is paid for in photobleaching.[4]
The size of one pixel in the sample plane.
A bar target blurred by the optics, then sampled at your pixel sizesimulation
Lateral FWHM, 0.51 λexc/NA
—pinhole ≥ 1 AU
Practical rule: 0.5 × FWHM
—Zeiss monograph, Part 2
Bandwidth rule: λexc/(8 NA)
—strict Nyquist for deconvolution
Your pixel
—
About the simulation. The target is a set of line pairs whose period decreases continuously; the blur is a Gaussian with the same FWHM as the point image (an approximation to the Airy core); sampling averages the blurred signal over each pixel. The Nyquist "practical rule" marker uses 0.5 × FWHM. Rows are illustrative, not data.[4][9]
Topic 06
Z-stacks: how thick is a slice, and how far apart should they be
The axial response of a confocal is always worse than the lateral one. For a point object with the pinhole open to 1 AU or more, the axial FWHM is 0.88 λexc/(n − √(n² − NA²)), roughly three times the lateral FWHM for a high-NA oil objective. The same expression shows why NA matters even more axially: for small NA it reduces to 1.77 n λ/NA², a dependence on the square of the aperture.[4]
For an extended fluorescent plane the measured slice is thicker still, because light from the whole plane reaches the pinhole; the Zeiss expression adds a geometric term that grows linearly with pinhole diameter (equation 4 in the pinhole module). Either way, Nyquist applies in z as it does in x and y: step the focus by no more than half the slice thickness, or use the bandwidth rule Δz = λexc/(4 n (1 − cos α)) if the stack will be deconvolved.[4][10]
One caution: when the immersion medium and the mounting medium have different refractive indices, the focus moves through the sample by a different distance than the stage moves, and spherical aberration broadens the axial PSF with depth. Stacks acquired with an oil objective into an aqueous sample are both stretched and blurred in z unless corrected.[13]
Distance between successive optical sections.
An x–z view of the whole stack: 1 µm beads, a tilted filament and a thin sheet, blurred by the optics, sampled at your z stepsimulation
Optical section (axial FWHM)
—
Practical z step: ≤ 0.5 × section
—Zeiss monograph
Bandwidth rule: λexc/(4n(1 − cos α))
—strict Nyquist for deconvolution
Slices for this stack
—
About the simulation. Section thickness uses Zeiss eq. (4) for pinholes ≥ 1 AU (fluorescent plane) and the computed point-object FWHM for smaller pinholes, as in the pinhole module. The x–z panel spans exactly the stack depth you set and blurs idealized beads, a filament and a sheet with a Gaussian of that thickness in z; it ignores refractive-index mismatch and depth-dependent aberration. Not data.[4]
Topic 07
Spectra and bleed-through
Fluorophore spectra are broad, and most emission spectra have a long tail toward the red. When two dyes are imaged in one frame, some light from the shorter-wavelength dye inevitably falls into the detection band of the longer-wavelength one. That contamination is bleed-through (also called crosstalk), and it can masquerade as colocalization.[11]
Three remedies, in order of cost. Narrow the detection band and move it away from the overlap; this sacrifices some signal. Scan sequentially, one laser at a time, so that during the red channel the green dye is barely excited; this doubles the acquisition time. Unmix spectrally: record several bands, measure each dye's spectrum from single-labeled controls, and solve the linear system that assigns each pixel's spectrum to its contributing dyes. Linear unmixing assumes the mixed spectrum is a weighted sum of the pure ones and needs reference spectra recorded under the same conditions.[11]
The module below uses two generic fluorophores with shapes typical of a green and an orange emitter. A tunable laser (as on the Stellaris) lets you place each excitation line at an absorption peak; spectral detectors let you set each band by wavelength.
Start and end wavelength of the band collected by the second detector.
AcquisitionSimultaneous: both lasers on. Sequential: only laser 2 is on while the orange channel records.
Excitation (dashed) and emission (filled) spectra of two generic dyes, laser lines, and the orange detection bandillustrative
What the two detectors record: a cell labeled with the green dye only (left) and one with the orange dye only (right)simulation
Bleed-through into the orange channel
—share of the band's signal that comes from the green dye
Green dye excited by laser 2
—relative to its peak absorption
Green-only cell in the orange channel
—its brightness relative to the orange-only cell
Remedy in effect
—
About the spectra. The curves are generic, asymmetric peaks chosen to resemble a green and an orange emitter. They are not measurements of any specific dye; use the vendor's or the Core's spectra viewer for real values. Bleed-through here assumes equal brightness and abundance of the two dyes and ignores detector sensitivity. Illustrative only.[11]
Topic 08
Photobleaching: a finite photon budget
Every fluorophore survives only a limited number of excitation cycles before a photochemical reaction, often from the triplet state and often involving oxygen, destroys its ability to fluoresce. Bleaching therefore scales with the cumulative excitation dose, the product of intensity and exposure time, in the regime where absorption is linear in intensity. The simplest description is first-order decay, F(t) = F0 e−k t with k proportional to intensity, and it is a useful planning tool; measured kinetics for common dyes deviate from a single exponential and depend on oxygen, mounting medium, and antifade reagents.[9][12]
Because dose, not power alone, does the damage, the practical choices are about how to spend a fixed budget: fewer frames or planes, a lower power with more averaging or longer dwell only where it buys signal that would otherwise be lost to noise, a pinhole no smaller than the question requires, and a pixel size that satisfies Nyquist without oversampling. Multiphoton excitation confines bleaching to the focal plane, which is one of its main advantages for long live-tissue sessions (section 9).[4][14]
Relative to a reference setting in which 100 frames reduce the signal to half.
Remaining fluorescence over the time-lapse, first-order modelmodel
Dose per frame
—power × averages, relative to the reference
Frame at which half remains
—first-order model
Signal left at the last frame
—relative to the first frame
Signal-to-noise gain from averaging
—√(averages), shot-noise limited
About the model. Single-exponential bleaching with a rate proportional to the per-frame dose, in the linear excitation regime. Real fluorophores show multi-exponential kinetics and the rate can change with oxygen depletion and local environment. The √N noise statement assumes shot-noise-limited detection. Illustrative; not a measurement.[4][12]
Topic 09
Multiphoton excitation: a pinhole made of physics
A fluorophore that absorbs blue light can also be excited by two near-infrared photons that arrive almost simultaneously, within a fraction of a femtosecond, each carrying about half the energy. The probability of that coincidence scales with the square of the intensity, so excitation only happens where photons are extremely dense: at the focus of a high-NA objective illuminated by a pulsed laser whose femtosecond pulses pack the photons into brief, intense bursts while keeping the average power low. Above and below the focus the intensity falls off and the squared dependence drives excitation to essentially zero. The excitation volume is therefore confined in three dimensions by the illumination itself, and no pinhole is needed.[14][18]
Two consequences follow. First, all emitted photons are useful, including scattered ones, so the detectors can sit close to the objective (non-descanned detection) and collect light that a pinhole would have thrown away. Second, the near-infrared excitation (typically 700 nm to beyond 1000 nm, depending on the laser) is scattered and absorbed far less by tissue than visible light, so the focus survives to several hundred micrometers in brain and other scattering tissue. Photobleaching and photodamage, which require excitation, are confined to the focal plane rather than the whole illuminated cone.[15][14]
The cost is resolution: the excitation wavelength is roughly twice as long, so the focal spot is larger than in single-photon confocal, only partly compensated by the squared intensity dependence. Multiphoton is the tool for depth and for gentle long-term imaging of living tissue, not for the finest detail in a thin section. The same pulsed laser also drives second-harmonic generation in ordered, non-centrosymmetric structures such as collagen fibrils, producing label-free signal at exactly half the excitation wavelength.[18]
model
Fluorescence generated in each planemodel
Computed from the same scalar diffraction model as the pinhole module (Born and Wolf §8.8), for an NA 1.0 water-immersion objective; the two-photon map uses twice the wavelength and squares the intensity. Single-photon excitation generates the same total fluorescence in every plane, which is why a pinhole is needed; two-photon excitation does not.[14][3]
Topic 10
Enhanced resolution: Airyscan and LIGHTNING
Both Core confocals offer a mode that resolves finer detail than a 1 AU pinhole allows. They get there by different routes, so they are described separately below: Airyscan changes the detector; LIGHTNING changes the processing.
Image scanning microscopy and Airyscan (Zeiss LSM 800)
The pinhole module showed the dilemma of the single pinhole: closing it toward zero improves resolution by up to 1.4× but discards most of the light. Image scanning microscopy escapes the trade-off by replacing the pinhole with an array of small detector elements. Each element behaves as a tiny, slightly off-axis pinhole: it sees a sharp, small-pinhole image, but one that is displaced from the scan position by half the element's own offset. Shifting each element's image back by that amount before summing, a step called pixel reassignment, recovers the small-pinhole resolution while collecting nearly all the light that a 1 AU pinhole would. The idea was proposed by Sheppard in 1988 and demonstrated by Müller and Enderlein in 2010.[16]
Airyscan is Zeiss's implementation of this idea. The detector is a hexagonal array of 32 GaAsP elements that together cover about 1.25 Airy units of the scanned point's image, each element acting as a pinhole of roughly 0.2 AU, according to the manufacturer. Every scan position therefore yields 32 slightly shifted images instead of a single intensity value. Processing shifts each one toward the center and combines them with a deconvolution step. Two gains follow: resolution, because each element is a small pinhole, and signal-to-noise, because the array as a whole rejects nothing within 1.25 AU. Zeiss states 1.7× higher resolution in x, y, and z than a conventional confocal, and a 5× smaller effective volume. [17]
In practice this means three things at the microscope. The image must be sampled at the Nyquist spacing for the enhanced resolution, so pixels are smaller than for a conventional scan. The processing adds a step after acquisition, and its strength setting trades resolution against noise amplification. And because the gain in signal-to-noise is real, Airyscan is as useful for dim samples at ordinary resolution as it is for resolution itself.[17][4]
Image scanning microscopy, the principle behind Airyscan, schematically. The gain comes from reassigning off-center light, not discarding it. Element count and sizes are illustrative.[16]
LIGHTNING (Leica Stellaris)
LIGHTNING takes a different route. The detector and pinhole are those of an ordinary confocal, used with a pinhole smaller than 1 AU and pixels fine enough to satisfy Nyquist for the finer detail the processing aims to recover. The gain comes from deconvolution: the recorded image is modeled as the true object blurred by the microscope's point spread function, and an iterative algorithm estimates the object that best explains the data. What Leica describes as "adaptive" is that the algorithm assesses signal and noise locally and chooses its parameters region by region rather than once for the whole image, which protects dim regions from being sharpened into noise. Leica states lateral resolution "down to 120 nm".
Deconvolution of any kind has conditions: the image must be Nyquist-sampled in all three dimensions, the point spread function must match the objective, wavelength, and pinhole actually used, and the raw data must be kept, because every reconstruction is an estimate. The Stellaris's hybrid detectors and small pinhole make the input to that estimate good; see section 5 for the sampling requirement.[10]
Deconvolution-based enhancement, as used by LIGHTNING, schematically. The curves are illustrative, not measured.
Neither method reaches the resolutions of STED or single-molecule localization, and neither breaks the diffraction limit by a large factor; both are best described as enhanced-resolution confocal with a strong gain in signal-to-noise. For the structures many cell biologists care about, from 100 to 250 nm, that is often exactly what is needed.[19]
Sources
References
Each numbered item is cited in the text above. Figures that come from a manufacturer rather than from the literature are identified as such in the text.
Lakowicz JR. Principles of Fluorescence Spectroscopy, 3rd ed. Springer, 2006. Chapter 1 (Jablonski diagram, Stokes shift, quantum yield, lifetime).
Minsky M. Microscopy apparatus. US Patent 3,013,467, filed 7 November 1957, issued 19 December 1961.
Born M, Wolf E. Principles of Optics, 6th ed. Pergamon, 1980. §8.5.2 (Airy pattern, encircled energy, Rayleigh criterion) and §8.8 (three-dimensional light distribution near focus).
Wilhelm S, Gröbler B, Gluch M, Heinz H. Confocal Laser Scanning Microscopy: Principles. Carl Zeiss Microscopy, Jena, document 60-1-0030/e, 2010, ISBN 978-3-940885-02-9. Part 1 equations (2)–(8); Part 2 (Nyquist theorem, pixel size); Part 3 (optical coordinates).
Wilson T. Resolution and optical sectioning in the confocal microscope. Journal of Microscopy 244(2):113–121, 2011.
Sheppard CJR, Matthews HJ. Imaging in high-aperture optical systems. Journal of the Optical Society of America A 4(8):1354–1360, 1987.
Abbe E. Beiträge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung. Archiv für mikroskopische Anatomie 9:413–468, 1873.
Rayleigh, Lord (Strutt JW). Investigations in optics, with special reference to the spectroscope. Philosophical Magazine, series 5, 8(49):261–274, 1879.
Pawley JB (ed.). Handbook of Biological Confocal Microscopy, 3rd ed. Springer, 2006. Chapter 4 (Pawley JB, "Points, pixels, and gray levels: digitizing image data") and Chapter 39 (Diaspro A, Chirico G, Usai C, Ramoino P, Dobrucki J, "Photobleaching").
Scientific Volume Imaging. "Nyquist rate" and "Nyquist calculator" documentation, svi.nl/NyquistRate and svi.nl/NyquistCalculator, accessed 25 October 2026.
Zimmermann T. Spectral imaging and linear unmixing in light microscopy. Advances in Biochemical Engineering/Biotechnology 95:245–265, 2005.
Song L, Hennink EJ, Young IT, Tanke HJ. Photobleaching kinetics of fluorescein in quantitative fluorescence microscopy. Biophysical Journal 68(6):2588–2600, 1995.
Hell S, Reiner G, Cremer C, Stelzer EHK. Aberrations in confocal fluorescence microscopy induced by mismatches in refractive index. Journal of Microscopy 169(3):391–405, 1993.
Helmchen F, Denk W. Deep tissue two-photon microscopy. Nature Methods 2(12):932–940, 2005.
Sheppard CJR. Super-resolution in confocal imaging. Optik 80(2):53–54, 1988; and Müller CB, Enderlein J. Image scanning microscopy. Physical Review Letters 104:198101, 2010.
Huff J. The Airyscan detector from ZEISS: confocal imaging with improved signal-to-noise ratio and super-resolution. Nature Methods 12:i–ii, 2015 (manufacturer application note).
Zipfel WR, Williams RM, Webb WW. Nonlinear magic: multiphoton microscopy in the biosciences. Nature Biotechnology 21(11):1369–1377, 2003.
Schermelleh L, Ferrand A, Huser T, Eggeling C, Sauer M, Biehlmaier O, Drummen GPC. Super-resolution microscopy demystified. Nature Cell Biology 21:72–84, 2019.