Spatial Coherence and the Van Cittert–Zernike Theorem

Hold two fingers up against a distant streetlamp and the light between them is, in a precise sense, correlated — the field arriving at your left finger keeps step with the field at your right, even though the lamp's filament is a chaos of independently radiating atoms. Bring the lamp closer, or use a wider one, and that correlation shrinks until the two points are effectively lit by different, unrelated sources. Spatial coherence is the name for this transverse correlation, and its governing law is one of the most surprising results in optics: an utterly incoherent source manufactures coherence purely by letting its light propagate. The field near the source is delta-correlated — neighbouring atoms know nothing of each other — yet a long way off the field is smoothly, partially coherent over a measurable patch. The Van Cittert–Zernike theorem, proved by Pieter van Cittert in 1934 and sharpened by Frits Zernike in 1938, says exactly how much: the complex degree of coherence $\mu_{12}$ between two distant points is the normalized Fourier transform of the source's angular intensity distribution. The same mathematics that turns an aperture into its Fraunhofer diffraction pattern turns a source into its coherence footprint — the source brightness plays the role of the aperture. Two consequences follow immediately and are the spine of this article. First, angular source size and transverse coherence width are conjugate: $\rho_c\,\theta_s\sim\lambda$ , so a smaller or more distant source is coherent over a wider area. Second, and spectacularly, you can run the transform backwards — measure how fringe visibility falls as you separate two collectors, and you have measured the angular size, even the shape, of a source far too small for any telescope to resolve. That is how Michelson and Pease first measured the diameter of a star.

Key takeaways

Spatial coherence measures how well the field at two transverse points stays correlated, captured by the equal-time mutual intensity $J(\mathbf r_1,\mathbf r_2)=\langle E(\mathbf r_1,t)E^*(\mathbf r_2,t)\rangle$ and its normalized modulus $|\mu_{12}|\in[0,1]$ .
Van Cittert–Zernike: for an incoherent quasi-monochromatic source, $\mu_{12}$ is the normalized Fourier transform of the source's angular brightness — propagation creates coherence.
Angular size and coherence width are conjugate, $\rho_c\,\theta_s\sim\lambda$ : a uniform disc of angular diameter $\theta_s$ loses coherence at baseline $d_0=1.22\,\lambda/\theta_s$ (an Airy form, $|2J_1(x)/x|$ ).
Run it backwards — measure visibility versus baseline and invert — to get a source's angular diameter: the principle of the Michelson stellar interferometer and of aperture synthesis.
It is the spatial twin of the Wiener–Khinchin theorem (which ties temporal coherence to the spectrum): broad source ↔ short coherence width, just as broad spectrum ↔ short coherence time.
Key idea: free propagation from an incoherent source Fourier-transforms its angular brightness into the field's transverse coherence, making source angular size and coherence width conjugate ( $\rho_c\,\theta_s\sim\lambda$ ) — so an interferometer's fringe visibility measures the size and shape of an unresolvable source.

The core insight

Free propagation from a spatially incoherent source performs a Fourier transform: the transverse coherence of the field far away is the normalized diffraction pattern of the source's angular brightness. Angular source size and coherence width are therefore conjugate, $\rho_c\,\theta_s\sim\lambda$ , and measuring fringe visibility against baseline inverts the transform to recover the source's angular structure — the basis of stellar interferometry.

Where this lives today

The theorem's forward direction warns you when light is too incoherent to interfere; its inverse direction has become a whole branch of observational astronomy. Radio and optical aperture synthesis arrays — the Very Large Array, ALMA, the CHARA and VLTI optical interferometers, and the Event Horizon Telescope that imaged a black hole's shadow — measure the complex coherence (the "visibility") on many baseline vectors, filling in the Fourier plane of the sky, then inverse-transform to make an image whose resolution is set by the longest baseline rather than any single dish. The Event Horizon Telescope's Earth-sized baseline is nothing but a Van Cittert–Zernike measurement at $d\sim10^7$ m. Closer to the bench, the theorem is the design rule for coherence engineering. Synchrotron and free-electron-laser beamlines quote a "transverse coherence length" set by the electron-beam source size and the distance to the sample, and X-ray coherent-diffraction imaging depends on making that length exceed the sample. Lithography and speckle-suppression systems deliberately reduce spatial coherence — larger effective source, rotating diffusers — to wash out unwanted fringes, using the same $\rho_c\,\theta_s\sim\lambda$ relation in reverse. Whenever an experiment asks "how small must my source be, or how far away, to interfere over this aperture?", the answer is a Van Cittert–Zernike calculation.

The physical picture

Put two pinholes in a screen a long way from a small lamp and treat it as a Young's experiment. Every point of the lamp casts its own two-pinhole fringe pattern on a far screen, and the position of that pattern depends on where on the lamp the light came from — an edge point tilts the incident wavefront and slides its fringes sideways. If the lamp is a true point, all these patterns land on top of one another and the fringes are razor-sharp: full spatial coherence. Widen the lamp and the patterns from opposite edges drift apart; when the drift reaches one full fringe the sum smears into uniform grey. The washout sets in when the extra path from the two source edges to the two pinholes differs by $\lambda$ , which is exactly $\rho_c\,\theta_s\sim\lambda$ . Why a Fourier transform, and not just a blur? Because for an incoherent source you must add intensities of the individual fringe patterns, and each pattern is weighted by a phase that grows linearly with the source point's position — and a sum of contributions weighted by a linear phase is the definition of a Fourier transform. The source's angular brightness is the input; the field's transverse coherence is its transform. This is the same integral that produces Fraunhofer diffraction, with the source brightness standing in for the aperture. And it explains the headline paradox: coherence is not radiated by the source, it is built by propagation. Two distant points end up looking back at the same narrow angular sliver of the source, and so their fields — assembled from the same atoms with nearly equal delays — become correlated.

Figure

Two pinholes far from an extended source: each source point casts a shifted fringe set; a source of angular size θ washes them out once the shift reaches one fringe, giving ρ_c θ_s ≈ λ

The trade-off mirrors temporal coherence exactly. A broad spectrum gives a short coherence time (a wave train that decorrelates quickly along its length); a broad angular source gives a short coherence width (a field that decorrelates quickly across the beam). Making a lamp spatially coherent with a pinhole is the transverse cousin of making it temporally coherent with a filter — and both cost you brightness. The next figure shows the actual coherence curve for the most important source shape, the uniform disc of a star, and marks the two baselines where real starlight and sunlight lose their fringes.

Figure

Coherence of a uniform circular source: |μ| = |2J₁(x)/x|, x = πθ_s d/λ — first zero at d₀ = 1.22 λ/θ_s. Sunlight: d₀ ≈ 72 μm; Betelgeuse: d₀ ≈ 3.07 m

The two callouts span eight orders of magnitude in one law. Sunlight loses coherence across a mere $72\,\mu$ m — which is why you cannot do a wide-slit interference experiment with raw sunlight, and why Verdet estimated this very number in 1865. Betelgeuse, vastly larger but almost unimaginably more distant, has a coherence footprint over three metres wide on the ground — small enough to fit in Michelson's beam, large enough that its finiteness could be measured. Same equation, same coefficient $1.22$ ; only $\theta_s$ changes.

The core equations

Coherence is a statistical property of the field, so the primary object is a correlation. The equal-time mutual intensity records how the analytic field at one point tracks the (conjugated) field at another, averaged over the source's fluctuations; normalizing it strips out the brightnesses and leaves a pure correlation between $0$ and $1$ .

J(\mathbf r_1,\mathbf r_2)=\langle E(\mathbf r_1,t)\,E^*(\mathbf r_2,t)\rangle, \qquad \mu_{12}=\frac{J(\mathbf r_1,\mathbf r_2)}{\sqrt{J(\mathbf r_1,\mathbf r_1)\,J(\mathbf r_2,\mathbf r_2)}} \qquad (1)

where: $J$ — mutual intensity (equal-time mutual coherence); $E(\mathbf r,t)$ — analytic (complex) field; $\langle\cdot\rangle$ — long-time or ensemble average; $\mu_{12}$ — complex degree of spatial coherence, $0\le|\mu_{12}|\le1$ ; $J(\mathbf r_k,\mathbf r_k)=I_k$ — mean intensity at point $k$ . A Young's experiment reads this off directly: fringe visibility $V=\tfrac{2\sqrt{I_1I_2}}{I_1+I_2}|\mu_{12}|$ , so equal-intensity pinholes give $V=|\mu_{12}|$ and the fringe position gives $\arg\mu_{12}$ . The theorem itself states what $\mu_{12}$ is for a spatially incoherent, quasi-monochromatic source. It is the normalized Fourier transform of the source's brightness, evaluated at the spatial frequency set by the point separation and the geometry.

\mu(\Delta\mathbf r)=\frac{\displaystyle\int I_s(\boldsymbol\theta)\,e^{-i\frac{2\pi}{\lambda}\,\boldsymbol\theta\cdot\Delta\mathbf r}\,\mathrm d^2\theta}{\displaystyle\int I_s(\boldsymbol\theta)\,\mathrm d^2\theta}=\hat I_s\!\left(\frac{\Delta\mathbf r}{\lambda}\right) \qquad (2)

where: $I_s(\boldsymbol\theta)$ — angular brightness of the source (angular coordinate $\boldsymbol\theta=\boldsymbol\xi/z$ for source point $\boldsymbol\xi$ at distance $z$ ); $\Delta\mathbf r=\mathbf r_1-\mathbf r_2$ — separation of the two observation points; $\lambda$ — mean wavelength; $\hat I_s$ — normalized 2-D Fourier transform, $\hat I_s(0)=1$ . Each baseline $\Delta\mathbf r$ samples one spatial-frequency component of the sky brightness — the foundation of aperture synthesis. For the single most important source shape — a uniformly bright disc, the model of a star — the transform of a top-hat is an Airy function, and its first zero gives the coherence baseline.

|\mu_{12}|=\left|\frac{2J_1(x)}{x}\right|,\quad x=\frac{\pi\,\theta_s\,d}{\lambda}, \qquad d_0=1.22\,\frac{\lambda}{\theta_s} \quad\Longleftrightarrow\quad \theta_s=1.22\,\frac{\lambda}{d_0} \qquad (3)

where: $J_1$ — Bessel function of the first kind, order $1$ (first zero at $x=3.8317$ ); $\theta_s$ — angular diameter of the disc; $d$ — baseline; $d_0$ — baseline of first coherence zero, where the fringes vanish. The coefficient $1.22=3.8317/\pi$ is the same one that sets the Rayleigh resolution limit — coherence and diffraction are two readings of one integral.

Derivation AVan Cittert–Zernike from the Huygens–Fresnel sum

The goal is to show that an incoherent source, whose neighbouring elements are by definition uncorrelated, nonetheless produces a smoothly correlated field far away, and that the correlation is precisely a Fourier transform of the source brightness. We build the observation field as a Huygens sum of wavelets from every source element, correlate the field at two points, and let the incoherence assumption collapse the resulting double integral to a single one.

Step 1. Write the field at an observation point as a Huygens sum, $E(\mathbf r,t)=\int A(\boldsymbol\xi,t)\,e^{ikR}/R\,\mathrm d^2\xi$ , where $A(\boldsymbol\xi,t)$ is the fluctuating source amplitude at $\boldsymbol\xi$ and $R=|\mathbf r-\boldsymbol\xi|$ . Why: propagation from an extended source is the coherent superposition of spherical wavelets, one per source element.

Step 2. Form the mutual intensity $J(\mathbf r_1,\mathbf r_2)=\iint\langle A(\boldsymbol\xi_1)A^*(\boldsymbol\xi_2)\rangle\,e^{ik(R_1-R_2)}/(R_1R_2)\,\mathrm d^2\xi_1\,\mathrm d^2\xi_2$ — the ensemble average acts only on the source amplitudes. Why: correlation of the observed fields inherits the correlation of the source amplitudes, carried by the propagation phases.

Step 3. Impose spatial incoherence: $\langle A(\boldsymbol\xi_1)A^*(\boldsymbol\xi_2)\rangle=I_s(\boldsymbol\xi_1)\,\delta^2(\boldsymbol\xi_1-\boldsymbol\xi_2)$ . The delta collapses the double integral to a single one over one shared source point, $J=\int I_s(\boldsymbol\xi)\,e^{ik(R_1-R_2)}/(R_1R_2)\,\mathrm d^2\xi$ . Why: "incoherent" means distinct source elements carry independent random phases, so only a single element can contribute in phase to both observation points — the crucial physical input.

Step 4. Expand the paths in the far field, $R_1-R_2\approx\psi(\mathbf r_1,\mathbf r_2)-\boldsymbol\xi\cdot(\mathbf r_1-\mathbf r_2)/z$ , with $R_1R_2\approx z^2$ . The term linear in $\boldsymbol\xi$ is what makes the integral a Fourier transform; the $\boldsymbol\xi$ -independent piece $\psi$ becomes an overall phase that normalization removes. Introducing the angular brightness $I_s(\boldsymbol\theta)$ gives $\mu(\Delta\mathbf r)=\hat I_s(\Delta\mathbf r/\lambda)$ . Why: this is the owned result — an incoherent source's coherence is the diffraction pattern of its own brightness, and it reduces to $\mu\to1$ for a point source (transform of a delta) as it must. *Verified: point-source limit $\mu\to1$ ; disc source $\to|2J_1(x)/x|$ with first zero at $x=3.8317$ ; slit source $\to\mathrm{sinc}$ — all checked in verify_spatial-coherence-van-cittert-zernike.py against Bessel tables.*

Go deeper · Derivation B · 1 minCoherence of a stellar disc and Michelson's diameter formula

Derivation BCoherence of a stellar disc and Michelson's diameter formula

Here the aim is to turn the theorem into a measuring instrument: evaluate the Van Cittert–Zernike integral for a uniformly bright disc, obtain the Airy-type coherence, and read off the baseline at which fringes vanish — the number Michelson actually measured to size a star. We do the disc integral in polar coordinates and locate the first Bessel zero.

Step 1. For a uniform disc of angular radius $\theta_s/2$ , the brightness cancels between numerator and denominator, leaving $\mu(d)=\big(\!\int_{|\boldsymbol\theta|\le\theta_s/2}e^{-i(2\pi/\lambda)\boldsymbol\theta\cdot\mathbf d}\,\mathrm d^2\theta\big)/(\pi(\theta_s/2)^2)$ . Why: a top-hat brightness makes the transform depend only on geometry, not on the (unknown) absolute brightness.

Step 2. Do the angular integral using $\int_0^{2\pi}e^{-i\kappa\rho\cos\phi}\,\mathrm d\phi=2\pi J_0(\kappa\rho)$ and $\int_0^a J_0(\kappa\rho)\rho\,\mathrm d\rho=(a/\kappa)J_1(\kappa a)$ , with $\kappa=2\pi d/\lambda$ and $a=\theta_s/2$ . Normalizing by the disc area $\pi a^2$ gives $\mu(d)=2J_1(\kappa a)/(\kappa a)=2J_1(\pi\theta_s d/\lambda)/(\pi\theta_s d/\lambda)$ . Why: this is the identical integral that produces the Airy disc of a circular aperture — coherence and aperture diffraction share one transform.

Step 3. Set the first zero: $J_1(x)=0$ at $x=3.8317$ , so $\pi\theta_s d_0/\lambda=3.8317$ gives $d_0=3.8317\lambda/(\pi\theta_s)=1.22\,\lambda/\theta_s$ , hence $\theta_s=1.22\,\lambda/d_0$ . Why: this is the owned result — a laboratory recipe for an angular diameter, obtained by widening the baseline until the fringes first disappear and reading off $d_0$ . *Verified: $1.22=3.8317/\pi$ ; Betelgeuse $d_0=3.07$ m at $\lambda=570$ nm $\Rightarrow\theta_s=0.047''$ ; Sun $\theta_s=0.53^\circ\Rightarrow d_0=72\,\mu$ m — in verify_spatial-coherence-van-cittert-zernike.py.*

Limiting cases

Case	Result	Meaning
point source	$\theta_s\to0\Rightarrow\\|\mu_{12}\\|\to1$	transform of a delta is flat — fully coherent everywhere; distant stars and lasers act as coherent points
uniform disc	$\\|\mu_{12}\\|=\\|2J_1(x)/x\\|$ , $x=\pi\theta_s d/\lambda$	Airy coherence; first zero $d_0=1.22\lambda/\theta_s$ sizes a star
uniform slit (1-D)	$\\|\mu_{12}\\|=\\|\mathrm{sinc}(\theta_s d/\lambda)\\|$	top-hat ↔ sinc; first zero at $d=\lambda/\theta_s$
resolved (large) source	$d\gg\lambda/\theta_s\Rightarrow\\|\mu_{12}\\|\to0$	widely separated points see a resolved source; correlations average away, no fringes

Common trap“Spatial and temporal coherence are the same thing, or one implies the other”

They are independent axes, and conflating them is the classic error. A pinhole in front of a white-light bulb produces light that is spatially coherent (a tiny angular source) yet temporally incoherent (a broad spectrum) — it makes crisp two-slit fringes near zero path difference but loses them within a micrometre of delay. A broad sodium lamp is the reverse: temporally coherent enough for thin-film fringes over half a millimetre, yet spatially incoherent, so a wide double slit shows nothing. Coherence area $A_c\sim(\lambda/\theta_s)^2$ is fixed by the source's angular size; coherence length $\ell_c=\lambda^2/\Delta\lambda$ by its bandwidth — different quantities from different properties of the source. And note the theorem's fine print: it applies to incoherent quasi-monochromatic sources, so it says nothing about a laser, which is already spatially coherent no matter how big its gain medium is.

Worked examples

1 · The coherence footprint of sunlight (confidence builder)

The Sun subtends an angular diameter of about $0.53^\circ$ at Earth. Over what transverse distance is sunlight ( $\lambda=550$ nm) spatially coherent — and why can't you run a wide double-slit experiment with raw sunlight?

Step 1. Convert the angular diameter: $\theta_s=0.53^\circ=0.53\times\pi/180=9.25\times10^{-3}$ rad.

Step 2. First coherence zero (uniform-disc model): $d_0=1.22\,\lambda/\theta_s=1.22\times550\times10^{-9}/(9.25\times10^{-3})=7.25\times10^{-5}$ m $=\mathbf{72.5\ \mu m}$ .

Step 3. So sunlight is well correlated only over patches tens of micrometres across. Two slits farther apart than $\sim72\,\mu$ m see essentially uncorrelated fields and make no fringes — which is exactly why Young used a pinhole first to shrink the effective angular source, and why raw sunlight through millimetre-scale slits shows nothing. It is also, historically, the number Verdet estimated in 1865. Coherence width ≈ 72 μm.

2 · Sizing Betelgeuse with a beam across a telescope (exam level)

Michelson and Pease mounted a 20-foot beam on the 100-inch Hooker telescope and found that fringes from Betelgeuse first vanished at a mirror separation of $d_0=3.07$ m, at $\lambda=570$ nm. What angular diameter does this give, and how does it compare to a single telescope's resolution?

Step 1. Apply the stellar formula: $\theta_s=1.22\,\lambda/d_0=1.22\times570\times10^{-9}/3.07=2.26\times10^{-7}$ rad.

Step 2. Convert to arcseconds: $\theta_s=2.26\times10^{-7}\times206265=\mathbf{0.047''}$ — Michelson and Pease's 1920 value, the first measured stellar disc.

Step 3. The 2.54-m Hooker telescope alone resolves only $1.22\lambda/D=1.22\times570\times10^{-9}/2.54=2.7\times10^{-7}$ rad $=0.056''$ — comparable to the star, so it cannot resolve the disc directly. The interferometer wins because its baseline (3.07 m) exceeds the mirror diameter, and by the same $1.22\lambda/d$ law it is the baseline, not the aperture, that sets the coherence-based resolution. θ_s = 0.047″; a 3.07 m baseline sizes a star a single telescope cannot resolve.

Physics of the portrait

The Van Cittert–Zernike theorem carries two names, but it was Frits Zernike who set spatial coherence on its modern statistical footing — and who, chasing the same ideas about phase, went on to reinvent the microscope.

Origins · 1 minFrits Zernike and how this was discovered

Frits Zernike

1888–1966 · Groningen, the Netherlands (Nobel Prize 1953)

Frits Zernike was a Dutch physicist at Groningen whose lifelong subject was the phase of light — the part of a wave that detectors throw away. In 1938 he generalized a 1934 result of Pieter van Cittert into the theorem that now bears both names: the degree of coherence produced by an incoherent source is the Fourier transform of the source's brightness. It was a decisive clarification, showing that "coherence" is a propagated, quantifiable correlation rather than a property a source either has or lacks. Zernike's obsession with phase paid its greatest dividend elsewhere. Noticing that transparent biological specimens — cells, bacteria — are nearly invisible under a microscope because they change only the phase of transmitted light, not its amplitude, he devised in the 1930s a way to convert those phase shifts into visible contrast by inserting a quarter-wave phase plate at the back focal plane, so the diffracted and undiffracted light interfere constructively or destructively. This phase-contrast microscope let biologists watch living, unstained cells for the first time, and it won Zernike the 1953 Nobel Prize in Physics. The polynomials he introduced to describe the aberrations of the optics — Zernike polynomials — remain the standard language of wavefront analysis in everything from telescope adaptive optics to the machine that measures your eye for laser surgery.

History

1865

— Verdet estimates the transverse coherence extent of sunlight, an early spatial-coherence length

1868

— Fizeau proposes using the fringe washout across a baseline to measure stellar diameters

1890

— Michelson develops the interferometric method and the visibility measure of coherence

1920

— Michelson and Pease measure the angular diameter of Betelgeuse, 0.047 arcseconds

1934

— van Cittert derives the coherence of the field from an incoherent source

1938

— Zernike casts the result as a Fourier transform of the source brightness

Going further

Texts. Hecht, Optics, 5th ed., §12.1–12.4 (coherence, Van Cittert–Zernike) · Born & Wolf, Principles of Optics, 7th ed., §10.4 · Goodman, Statistical Optics, 2nd ed., Ch. 5 · Saleh & Teich, Fundamentals of Photonics, 3rd ed., Ch. 11 · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 3.

Provenance. The Van Cittert–Zernike integral is derived from the Huygens–Fresnel sum with the incoherence assumption $\langle A(\boldsymbol\xi_1)A^*(\boldsymbol\xi_2)\rangle=I_s\,\delta^2(\boldsymbol\xi_1-\boldsymbol\xi_2)$ (Born & Wolf §10.4; Goodman Ch. 5). The uniform-disc coherence $2J_1(x)/x$ and its first zero at $x=3.8317$ , giving $1.22=3.8317/\pi$ , are checked against Bessel tables; the point-source ( $\mu\to1$ ) and slit-source (sinc) limits are recomputed. Numbers — sunlight coherence width $72\,\mu$ m at $\theta_s=0.53^\circ$ , Betelgeuse $\theta_s=0.047''$ from $d_0=3.07$ m, the $0.01''$ star giving $d_0\approx13.8$ m — are reproduced in verify_spatial-coherence-van-cittert-zernike.py. The Wiener–Khinchin analogy and the Michelson–Pease 1920 measurement follow Loudon Ch. 3 and Saleh & Teich Ch. 11.

Prerequisites. Mutual Coherence Function and Degree of Coherence · Young's Double-Slit Experiment · Fringe Visibility and Contrast · Two-Wave Interference and the Interference Term. Leads to. Temporal Coherence and Coherence Length · Conditions for Observable Interference · Wavefront-Splitting and Amplitude-Splitting Interferometers.

How can a source that is completely incoherent make coherent light just by being far away?

Because coherence here is a correlation between two observation points, not a property carried by the source. Far from the source, both points look back at essentially the same small angular patch and collect light from the same atoms with nearly the same geometric delays, so their fields — however random in absolute phase — rise and fall together. Mathematically the incoherent (delta-correlated) source intensity gets Fourier-transformed by propagation into a smooth coherence function of width $\sim\lambda z/D$ for a source of size $D$ at distance $z$ . Distance is the machine that manufactures the correlation, which is why every star, though a churning ball of plasma, arrives as a beautifully coherent point of light.

Why is the coefficient 1.22 the same as in the Rayleigh resolution criterion?

Because both come from the first zero of the same Bessel function. The coherence of a uniform disc source is $2J_1(x)/x$ , and the Airy diffraction pattern of a uniform circular aperture is the identical function — the Van Cittert–Zernike integral over a disc source is mathematically the same as the Fraunhofer integral over a disc aperture, with the source brightness in the role of the aperture transmission. The first zero of $J_1$ sits at $x=3.8317$ , and $3.8317/\pi=1.22$ whether you are asking "at what baseline does coherence vanish?" or "at what angle does the diffracted intensity vanish?". Coherence measurement and image resolution are two applications of one transform.

Is the fringe visibility exactly the degree of coherence in a stellar interferometer?

Yes, once the two apertures collect equal intensity. The general relation is $V=\tfrac{2\sqrt{I_1I_2}}{I_1+I_2}|\mu_{12}|$ , so with balanced apertures ( $I_1=I_2$ ) the visibility equals $|\mu_{12}|$ directly. That is what makes the instrument a coherence meter: you widen the baseline, watch $V$ trace out $|2J_1(x)/x|$ , and the baseline where $V$ first hits zero is $d_0=1.22\lambda/\theta_s$ . In practice atmospheric turbulence scrambles the phase and forces you to measure the squared visibility rapidly (intensity interferometry, à la Hanbury Brown and Twiss, sidesteps the phase entirely), but the quantity being measured is still the modulus of the spatial degree of coherence.

What exactly is the difference between coherence area and coherence length?

Coherence area is transverse — the patch across a beam over which the field stays correlated at one instant, $A_c\sim(\lambda/\theta_s)^2$ , set by the source's angular size (this article). Coherence length is longitudinal — the distance along the beam, or the path difference in an interferometer, over which the field stays correlated with a delayed copy of itself, $\ell_c=\lambda^2/\Delta\lambda$ , set by the source's bandwidth. They answer different questions: "how far apart can two slits be?" (area) versus "how unequal can two arm lengths be?" (length). A source can be excellent on one axis and hopeless on the other, and a real interference experiment must satisfy both at once.

Does the theorem require the source to be perfectly monochromatic?

It requires quasi-monochromatic light: a bandwidth narrow enough that the path-difference-induced delays across the region of interest are much less than the coherence time, so that temporal decorrelation can be ignored while you study the spatial part. In practice a filter over a thermal source is enough. If the bandwidth is broad, the mutual coherence function no longer factorizes cleanly into a purely spatial piece, and you must treat the full space-time correlation. The quasi-monochromatic condition is why stellar interferometers work through narrow filters, and why the theorem is stated for a single mean wavelength $\lambda$ .

How does aperture synthesis turn coherence measurements into an image?

Each pair of collectors in an array measures the complex coherence $\mu(\Delta\mathbf r)$ for their particular baseline vector $\Delta\mathbf r$ — and by the theorem that is one Fourier component of the sky's brightness distribution, at spatial frequency $\Delta\mathbf r/\lambda$ . As the Earth rotates and the array reconfigures, many baselines sample many points in the Fourier (" $uv$ ") plane. Inverse-transforming the collected visibilities reconstructs the image, with angular resolution $\sim\lambda/d_\text{max}$ set by the longest baseline. This is how the Very Large Array and the Event Horizon Telescope achieve resolutions no single dish could — they are building the Fourier transform of the sky one coherence measurement at a time.