Fresnel-Kirchhoff Diffraction Integral

We take from electromagnetism the single fact this whole article rests on, and then leave it behind. The companion piece aq-em-scalar-vector-diffraction-theory frames diffraction the way electrodynamics demands: light is a vector field with boundary conditions, and the "scalar theory" we are about to use — treating the field as one complex number per point — is an approximation that "fails badly" once an aperture shrinks to the wavelength, where polarization and edge currents take over. That article also states the master identity every diffraction formula descends from, $U = \oint_S (G\,\partial_n U - U\,\partial_n G)\,\mathrm{d}a$ , and observes that the entire taxonomy — Kirchhoff versus Rayleigh–Sommerfeld, scalar versus vector — is just the choice of Green's function and of which boundary data you trust. Optics takes that scalar licence and runs the full distance with it: this article owns the complete scalar Kirchhoff development — the integral theorem, the aperture boundary conditions, and the diffraction integral they produce. The prize is a formula that finishes the job the Huygens–Fresnel principle started. Huygens imagined every wavefront point re-radiating a spherical wavelet; Fresnel made the wavelets interfere but had to postulate their amplitude ( $1/\lambda$ ), their quarter-wave phase lead, and a fudge factor to stop them radiating backward. Kirchhoff showed in 1882 that Green's theorem hands you all three for free. The obliquity factor $K = \tfrac12(\cos\theta_0 + \cos\theta)$ , the $-i/\lambda$ prefactor — the guesses become theorems. Everything downstream, the sinc of a slit and the Airy disk of a telescope, is this one integral read out for a particular aperture.

Key takeaways

The Fresnel–Kirchhoff integral gives the diffracted scalar field as an integral over the aperture of the incident field weighted by a spherical wavelet $e^{ikr}/r$ , an obliquity factor $K$ , and the prefactor $-i/\lambda$ .
It is derived, not postulated: Green's second identity applied to the Helmholtz equation with the free-space Green's function $e^{ikr}/r$ gives the exact Helmholtz–Kirchhoff integral theorem; Kirchhoff's aperture boundary conditions then reduce it to the aperture alone.
The derivation supplies what Fresnel guessed — the obliquity factor $K = \tfrac12(\cos\theta_0 + \cos\theta)$ that kills the backward wave, and the $-i/\lambda = e^{-i\pi/2}/\lambda$ that fixes amplitude and the $90°$ phase lead.
At normal incidence $K = \tfrac12(1 + \cos\theta)$ — exactly the factor the Huygens–Fresnel article had to assume — so Kirchhoff derives what Fresnel postulated.
Kirchhoff's boundary conditions are mathematically inconsistent (they over-fix $U$ and $\partial_n U$ together); Rayleigh–Sommerfeld theory repairs this, but all three agree when the aperture is many wavelengths wide.
Key idea: Green's theorem plus the spherical Green's function turns the Helmholtz equation into a surface integral, and Kirchhoff's aperture boundary conditions collapse it to $U(P) = \dfrac{-i}{\lambda}\displaystyle\iint_{\text{ap}} U(Q)\,\dfrac{e^{ikr}}{r}\,K(\chi)\,\mathrm{d}S$ — Huygens' wavelets with a rigorous amplitude, phase, and obliquity.

The core insight

Huygens and Fresnel pictured diffraction as a sum of secondary wavelets; Kirchhoff proved it. Apply Green's second identity to the field $U$ and the outgoing spherical Green's function $G = e^{ikr}/r$ , both solutions of the Helmholtz equation, and the volume integral collapses onto a closed surface. Wrap that surface around the aperture, impose Kirchhoff's boundary conditions, and out drops the Fresnel–Kirchhoff integral — with the obliquity factor and the $-i/\lambda$ prefactor that Fresnel could only guess now determined by the mathematics.

Where this lives today

The Fresnel–Kirchhoff integral, in its paraxial (Fresnel) guise, is a convolution with a quadratic-phase kernel — which means it is an FFT away from being computed. That fact makes it the numerical backbone of modern optical engineering: beam-propagation and angular-spectrum codes step laser fields through space this way; digital holography and computational microscopy reconstruct three-dimensional fields by evaluating it; and diffractive optical elements, from the phase masks in lithography scanners to the metasurface lenses in phone cameras, are designed by inverting it. Every time a photolithography engineer computes the aerial image a mask will cast, they are integrating this formula over the mask aperture. At the frontier the integral marks its own boundary. As feature sizes in extreme-ultraviolet lithography and in nanophotonics drop toward and below the wavelength, the scalar approximation the electromagnetism owner warned about breaks down, and designers switch to full vector solvers (finite-difference time-domain, rigorous coupled-wave analysis). The Fresnel–Kirchhoff integral remains the fast, intuitive first pass — and the yardstick against which the expensive vector calculation is checked.

The physical picture

Tile the aperture with imaginary antennas. Each point $Q$ inside the hole re-emits an outgoing spherical wave $e^{ikr}/r$ whose amplitude and phase match the incident wave arriving there, and the field at any observation point $P$ is the coherent sum of all of them. That is Huygens' picture exactly — but a picture is not a formula. The question Kirchhoff answered is what weight each wavelet carries, and the answer is not "all equal": a wavelet radiates strongly forward and not at all backward, scaled by $1/\lambda$ and phase-advanced a quarter period. The first figure sets up the geometry that decides those weights; the second is the punchline — the obliquity factor that Fresnel invented by hand and Kirchhoff derived.

Figure

Diffraction geometry — a source S illuminates aperture point Q, which re-radiates toward observation point P; θ₀ is the incidence angle and θ the diffraction angle, both from the aperture normal

The two angles are the whole story. As the incident ray and the ray toward $P$ swing away from the aperture normal, each wavelet dims according to how forward-tilted it is. Collapse that dependence onto a single curve and you get the obliquity factor — brightest dead ahead, dark straight back.

Figure

The punchline — obliquity factor K(θ)=½(1+cosθ) at normal incidence: Kirchhoff-derived (solid), equal to Fresnel's postulated factor; K=1 forward, K=½ sideways, K=0 backward kills the return wave

Read the curve from left to right. Straight ahead the wavelet is at full strength; ninety degrees off it is at half; pointing back toward the source it is extinguished. That last value — $K(\pi) = 0$ — is the quiet triumph of the derivation. Huygens' construction, taken literally, radiates a wave backward as readily as forward, an embarrassment optics lived with for two centuries. The Green's-theorem obliquity factor sets that backward amplitude to exactly zero, not as a postulate but as arithmetic.

The core equations

The endpoint is a single integral. It says: to find the field at $P$ , add up a spherical wavelet from every aperture point $Q$ , each weighted by the incident field there, by the obliquity factor, and by a universal prefactor. Read it as Huygens' superposition with the three numbers Fresnel could only estimate now pinned down.

U(P) = \frac{-i}{\lambda}\iint_{\text{ap}} U(Q)\,\frac{e^{ikr}}{r}\,K(\chi)\,\mathrm{d}S \qquad (1)

where: $U(P)$ — complex scalar field at the observation point; $U(Q)$ — incident field at aperture point $Q$ (Kirchhoff sets it equal to the unobstructed field); $r$ — distance from $Q$ to $P$ ; $k = 2\pi/\lambda$ — wavenumber; $K(\chi)$ — obliquity factor; $-i/\lambda$ — prefactor carrying the $90°$ phase lead and the amplitude scaling; $\mathrm{d}S$ — aperture area element. Every diffraction pattern in the rest of optics is this integral evaluated for one aperture and one approximation of $r$ . The obliquity factor is the piece the picture cannot supply. It depends on two angles measured from the aperture normal — where the light came from and where it is going — and its form is fixed by differentiating the Green's function along the normal.

K(\chi) = \tfrac12\big(\cos\theta_0 + \cos\theta\big) \qquad (2)

where: $\theta_0$ — angle between the incident ray (source $\to Q$ ) and the aperture normal; $\theta$ — angle between the diffracted ray ( $Q \to P$ ) and the normal. At normal incidence $\theta_0 = 0$ this collapses to $K = \tfrac12(1 + \cos\theta)$ , precisely the inclination factor the Huygens–Fresnel treatment had to assume — the two theories meet exactly where they should. All of this descends from one field equation. A single-frequency optical field, with each Cartesian component treated as an independent scalar, obeys the Helmholtz equation — the time-stripped wave equation that is the true starting line.

(\nabla^2 + k^2)\,U(\mathbf{r}) = 0 \qquad (3)

where: $U$ — one scalar component of the monochromatic field, $\mathbf{E} \propto U\,e^{-i\omega t}$ ; $k = \omega/c$ — wavenumber in the medium. Reducing the vector $\mathbf{E}$ to this one scalar $U$ is the approximation electromagnetism lends us (and warns fails for sub-wavelength apertures); everything below is exact given this reduction. The engine that turns a field equation into a surface integral is Green's second identity paired with the free-space Green's function — the outgoing spherical wave that is itself a solution of the Helmholtz equation everywhere except at its own source point.

G(\mathbf{r}) = \frac{e^{ikr}}{r}, \qquad (\nabla^2 + k^2)\,G = -4\pi\,\delta^3(\mathbf{r}) \qquad (4)

where: $r$ — distance from the observation point $P$ (the singular point of $G$ ); $\delta^3$ — three-dimensional Dirac delta locating that singularity. The delta is the whole trick: it is what plucks the value $U(P)$ out of the volume integral, converting an identity about volumes into a statement about the field at one point.

Derivation AFrom Green's theorem to the Fresnel-Kirchhoff integral

The goal is to obtain the diffraction integral — prefactor, spherical wavelet, and obliquity factor together — as a theorem rather than a guess, starting only from the Helmholtz equation and the incident field. The strategy is Kirchhoff's: pair the unknown field $U$ with a known Green's function $G$ , use Green's second identity to trade a volume integral for a surface one, exploit the delta-function singularity of $G$ to extract $U(P)$ , and then throw away everything but the aperture using boundary conditions.

Step 1. Write Green's second identity for $U$ and $G$ over a volume $V$ bounded by a closed surface $S$ enclosing $P$ : $\iiint_V (U\nabla^2 G - G\nabla^2 U)\,\mathrm{d}V = \oiint_S (U\,\partial_n G - G\,\partial_n U)\,\mathrm{d}S$ . Why: this identity is the scalar analogue of integration by parts — it moves all the physics onto a surface, and it is the master identity the electromagnetism owner names as the root of every diffraction formula.

Step 2. Both $U$ and $G$ satisfy the Helmholtz equation, so $U\nabla^2 G - G\nabla^2 U = U(-k^2 G) - G(-k^2 U) = 0$ everywhere the functions are smooth. The volume integrand vanishes identically — except at $r = 0$ , where $G$ blows up. Why: this is exactly the cancellation that makes the method work; all that survives is the singularity, which is where the information about $P$ lives.

Step 3. Excise $P$ with a small sphere of radius $\varepsilon$ and use $(\nabla^2 + k^2)G = -4\pi\delta^3$ . Integrating across the tiny sphere and letting $\varepsilon \to 0$ pulls the term $-4\pi\,U(P)$ out of the volume integral, leaving the exact Helmholtz–Kirchhoff integral theorem $U(P) = \tfrac{1}{4\pi}\oiint_S (G\,\partial_n U - U\,\partial_n G)\,\mathrm{d}S$ . Why: the delta function converts "the integrand is zero" into "the field at P equals a surface integral" — this is the rigorous core, and it is still exact.

Step 4. Choose the surface: the aperture plane, the opaque back of the screen, and a large hemisphere closing it off. The hemisphere contributes nothing (the Sommerfeld radiation condition: outgoing waves die as $r \to \infty$ ), so only the screen and aperture remain. Why: we are free to draw $S$ wherever we like, and this choice is the one that isolates the physics we control.

Step 5. Impose Kirchhoff's boundary conditions — on the opaque screen $U = \partial_n U = 0$ ; in the aperture $U$ and $\partial_n U$ equal their unobstructed incident values. Only the aperture integral survives. Why: this is the physical modelling step (and, as the warning below shows, the mathematically shaky one) — it says the screen is perfectly black and the hole perfectly transparent.

Step 6. Insert an incident spherical wave $U_{\text{inc}} = A e^{ik\rho}/\rho$ and differentiate both $U_{\text{inc}}$ and $G$ along the normal. In the far-field regime $kr, k\rho \gg 1$ each normal derivative brings down a factor $ik\cos(\text{angle})$ ; the $1/r^2$ terms are negligible. Collecting them factors the integrand into $ik\,(e^{ikr}/r)\cdot\tfrac12(\cos\theta_0 + \cos\theta)$ , i.e. the obliquity factor $K$ , while regrouping $ik/2\pi \cdot \tfrac12 = -i/\lambda$ recovers the standard prefactor. Why: the two angle-dependent derivatives are the birthplace of the obliquity factor — Fresnel's fudge falls out of $\partial_n G$ and $\partial_n U_{\text{inc}}$ , and the $-i/\lambda$ from the constants. *Verified: prefactor $-i/\lambda = e^{-i\pi/2}/\lambda$ , obliquity $K = \tfrac12(\cos\theta_0 + \cos\theta)$ , the Green's-function identity $(\nabla^2+k^2)(e^{ikr}/r) = -4\pi\delta^3$ , and the normal-incidence reduction $K = \tfrac12(1+\cos\theta)$ in verify_aq-optics-fresnel-kirchhoff-diffraction-integral.py.*

Go deeper · Derivation B · 1 minFrom the integral to the Fraunhofer Fourier transform

Derivation BFrom the integral to the Fraunhofer Fourier transform

The single integral hides every diffraction regime inside one approximation choice — how many terms of the distance $r$ you keep in the fast phase $e^{ikr}$ . The goal here is to show that the far-field limit turns the Fresnel–Kirchhoff integral into a plain Fourier transform of the aperture, the result that makes a lens a spectrum analyser and a telescope's resolution calculable. Take an on-axis plane wave on an aperture and observe far away at small angles.

Step 1. Expand the exact distance $r = \sqrt{z^2 + (x-x')^2 + (y-y')^2}$ binomially for transverse displacements small against the axial distance $z$ : $r \approx z + \tfrac{x^2+y^2}{2z} - \tfrac{xx'+yy'}{z} + \tfrac{x'^2+y'^2}{2z}$ . Why: sorting the phase into constant, observation-only, cross, and aperture-quadratic pieces is what separates Fraunhofer from Fresnel.

Step 2. Impose the far-field condition: the aperture-quadratic term $k(x'^2+y'^2)/2z$ is negligible when $z \gg a^2/\lambda$ , i.e. the Fresnel number $N_F = a^2/\lambda z \ll 1$ . Then only the cross term $-k(xx'+yy')/z$ varies across the aperture. Why: dropping the aperture curvature is exactly the boundary between near and far field — keep it and you get Fresnel diffraction, drop it and you get Fraunhofer.

Step 3. Identify the transverse spatial frequencies $k_x = kx/z = k\sin\theta_x$ and $k_y = ky/z = k\sin\theta_y$ ; the observation angle maps onto a wavenumber. The surviving integral is $U \propto \iint_{\text{ap}} U(x',y')\,e^{-i(k_x x' + k_y y')}\,\mathrm{d}x'\mathrm{d}y'$ — the two-dimensional Fourier transform of the aperture field. Why: this is the structural payoff — a slit transforms to a sinc, a circular aperture to the Airy pattern $[2J_1(u)/u]^2$ , all as Fourier transforms. *Verified: the Fraunhofer condition $z \gg a^2/\lambda$ ( $N_F \ll 1$ ), the angle-to-wavenumber map $k_x = k\sin\theta_x$ , and the Airy first-zero coefficient $1.22$ (first zero of $J_1$ at $3.8317$ , $3.8317/\pi = 1.220$ ) in the verify script.*

Limiting cases

The one integral contains every regime as a successive approximation of the path length $r$ in the exponent — and the obliquity factor as a near-axis simplification.

Case	Result	Meaning
Fraunhofer (far field)	$N_F = a^2/\lambda L \ll 1$	keep only the linear aperture term; the integral is a 2-D Fourier transform of the aperture — sinc and Airy patterns
Fresnel (near field)	$N_F = a^2/\lambda L \sim 1$	retain the quadratic term; Fresnel integrals $C(w), S(w)$ and the Cornu spiral — edge fringes, zone plates
Geometrical optics	$\lambda \to 0,\; N_F \gg 1$	stationary phase pins the integral to the geometric ray; sharp shadows, ray optics recovered
Forward paraxial	$\theta_0, \theta \to 0 \Rightarrow K \to 1$	near the axis the obliquity factor is essentially unity and is routinely dropped in Fourier optics
Backward direction	$\theta = \pi \Rightarrow K = 0$	the return wave vanishes identically — the resolution of the Huygens backward-wave paradox

Common trap“The Fresnel-Kirchhoff integral is an exact solution of the diffraction problem”

It is not, and the flaw is in the boundary conditions, not the arithmetic. Kirchhoff fixes both $U$ and $\partial_n U$ to their incident values in the aperture and both to zero on the screen — but a solution of the Helmholtz equation cannot have $U$ and $\partial_n U$ simultaneously vanish on any finite patch without vanishing everywhere (analytic continuation forbids it). So the integral does not actually reproduce the boundary data it was built from, and it misbehaves within a few wavelengths of the aperture edge. Sommerfeld's repair is elegant: use a Green's function that vanishes (or whose normal derivative vanishes) on the aperture plane, so only one boundary datum is needed and the over-determination disappears — the Rayleigh–Sommerfeld integrals. The electromagnetism owner puts it crisply: the whole taxonomy is just the choice of Green's function and which boundary data you trust. In practice, for apertures many wavelengths across and moderate angles, Kirchhoff and both Rayleigh–Sommerfeld forms agree to leading order — which is why this "wrong" integral remains the workhorse. And remember it is scalar: polarization and sub-wavelength edge physics are simply outside it.

Worked examples

1 · The obliquity factor, direction by direction (confidence builder)

A plane wave hits an aperture at normal incidence ( $\theta_0 = 0$ ). Evaluate the Kirchhoff obliquity factor for a wavelet sent forward, ninety degrees off-axis, and straight back, and confirm it reproduces the factor the Huygens–Fresnel article assumed.

Step 1. Normal incidence sets $\cos\theta_0 = 1$ , so $K = \tfrac12(1 + \cos\theta)$ . Forward ( $\theta = 0$ ): $K = \tfrac12(1 + 1) = \mathbf{1}$ — full strength along the incident direction. This is the maximum; no wavelet radiates more strongly than straight ahead.

Step 2. Sideways ( $\theta = 90°$ ): $K = \tfrac12(1 + 0) = \mathbf{0.5}$ . A wavelet launched along the aperture plane carries half the forward amplitude — a quarter of the intensity, since $I \propto K^2$ .

Step 3. Backward ( $\theta = 180°$ ): $K = \tfrac12(1 - 1) = \mathbf{0}$ . The return wave is extinguished exactly. This is the whole point: the naive Huygens construction radiated backward, and the derived obliquity factor sets that amplitude to zero without any extra assumption.

Step 4. Compare with Huygens–Fresnel: that article had to postulate $K(\chi) = \tfrac12(1 + \cos\chi)$ with $\chi$ the angle from the forward direction. At normal incidence Kirchhoff's $\tfrac12(1 + \cos\theta)$ is identical term for term. Kirchhoff derives what Fresnel guessed: K = 1, ½, 0 forward, sideways, backward.

2 · Where does far field begin? The Fresnel number (exam level)

A circular aperture of radius $a = 1.0$ mm is lit by green light, $\lambda = 550$ nm. Beyond what distance is the diffraction Fraunhofer (Fourier-transform) rather than Fresnel, and what is the Fresnel number at $z = 1$ m and at $z = 10$ m?

Step 1. The far-field (Fraunhofer) condition from Derivation B is $N_F = a^2/(\lambda z) \ll 1$ , i.e. $z \gg a^2/\lambda$ . Compute the crossover scale: $a^2/\lambda = (10^{-3})^2 / (550\times10^{-9}) = 10^{-6}/5.5\times10^{-7} = \mathbf{1.82\ m}$ . Fraunhofer diffraction needs $z$ comfortably beyond a couple of metres.

Step 2. At $z = 1$ m: $N_F = (10^{-3})^2 / (550\times10^{-9} \times 1) = 10^{-6}/5.5\times10^{-7} = \mathbf{1.82}$ . With $N_F \sim 1$ this is squarely the Fresnel regime — retain the quadratic aperture phase; expect zone structure, not a clean Airy pattern.

Step 3. At $z = 10$ m: $N_F = 10^{-6}/(5.5\times10^{-7}\times 10) = \mathbf{0.182}$ . Now $N_F \ll 1$ : the aperture-quadratic phase is small, the integral is essentially the Fourier transform of the disk, and the pattern is the Airy disk with its first dark ring at $\sin\theta = 1.22\,\lambda/2a$ .

Step 4. The lesson is that "far" is set by the aperture, not by intuition: a 1 mm hole only reaches its far field metres away, while a 100 µm hole ( $a^2/\lambda = 18$ mm) does so on a benchtop. Crossover $z \approx 1.8$ m; $N_F = 1.82$ at 1 m (Fresnel), $0.18$ at 10 m (Fraunhofer).

Physics of the portrait

The hero of this article is not a person but a factor — the small trigonometric weight, sitting inside the integral, that turns a suggestive picture into a predictive theory.

Origins · 1 minobliquity factor and how this was discovered

obliquity factor

Fresnel postulate 1818, Kirchhoff derivation 1882 · scalar diffraction theory

The obliquity factor (also called the inclination factor) is the angular weight $K = \tfrac12(\cos\theta_0 + \cos\theta)$ that each Huygens wavelet carries in the diffraction integral. Its biography is a miniature of how physics matures. Augustin-Jean Fresnel, in his 1818 prize memoir, needed the secondary wavelets to interfere and to not radiate backward, so he introduced by hand an inclination factor that fell from unity in the forward direction to zero in the backward one — a physically motivated guess with no derivation behind it, sitting uneasily beside the $1/\lambda$ amplitude and quarter-wave phase he also had to assume. For sixty years it worked but nobody could say why. Then in 1882 Gustav Kirchhoff applied Green's theorem to the Helmholtz equation and the factor simply appeared, as the sum of two normal derivatives — one from the incident spherical wave, one from the Green's function — with the exact form $\tfrac12(\cos\theta_0 + \cos\theta)$ . At normal incidence it reduced to Fresnel's $\tfrac12(1 + \cos\theta)$ , vindicating the guess and, more importantly, showing it was never arbitrary. The factor is thus a marker of the boundary between heuristic and rigorous optics: the same number, first invented to fit the data and then derived to explain it. That its backward value is identically zero — the clean resolution of a paradox Huygens' construction had carried since 1690 — is the sort of quiet consistency that tells you a theory has found its footing.

History

1690

— Huygens publishes the Traité de la Lumière; secondary wavelets explain propagation but leave an unphysical backward wave

1818

— Fresnel's prize memoir adds interference and postulates the inclination factor, the 1/λ amplitude, and the quarter-wave phase

1882

— Kirchhoff derives the diffraction integral from Green's theorem, fixing the obliquity factor and prefactor rigorously

1896

— Sommerfeld solves diffraction by a half-plane rigorously, exposing the inconsistency of Kirchhoff's boundary values

1897

— Rayleigh treats the passage of waves through apertures with a single boundary datum — the germ of the Rayleigh–Sommerfeld integrals

Going further

Texts. Hecht, Optics, 5th ed., §10.2–10.4 (Kirchhoff's scalar diffraction theory and the diffraction integral) · Born & Wolf, Principles of Optics, 7th ed., §8.3 (rigorous Fresnel–Kirchhoff derivation) · Goodman, Introduction to Fourier Optics, 3rd ed., Ch. 3 (Kirchhoff and Rayleigh–Sommerfeld formulations) · Saleh & Teich, Fundamentals of Photonics, Ch. 4 · Sommerfeld, Optics, §34 (the Green's-function critique and repair).

Provenance. The derivations follow the standard Green's-theorem route in Born & Wolf (§8.3), Hecht (§10.2), and Goodman (Ch. 3). Verified in verify_aq-optics-fresnel-kirchhoff-diffraction-integral.py: the prefactor $-i/\lambda$ and obliquity $K = \tfrac12(\cos\theta_0 + \cos\theta)$ (Born & Wolf Eq. 8.3.24; Hecht Eq. 10.28); the Green's-function relation $(\nabla^2 + k^2)(e^{ikr}/r) = -4\pi\delta^3$ ; the Fraunhofer condition $z \gg a^2/\lambda$ ( $N_F \ll 1$ ) and the resulting Fourier-transform form; and the Airy first-zero coefficient $1.22$ (first zero of $J_1$ at argument $3.8317$ , $3.8317/\pi \approx 1.22$ ). All quoted numbers are standard textbook values.

Prerequisites. Huygens-Fresnel Principle · Helmholtz equation and Green's theorem · Spherical waves and the $e^{ikr}/r$ Green's function. Leads to. Rayleigh-Sommerfeld Diffraction Theory · Fresnel Diffraction Regime · Fraunhofer Diffraction Regime · Fresnel Number · Circular Aperture Diffraction and the Airy Pattern. Recalled by name. aq-em-scalar-vector-diffraction-theory (electromagnetism owns the scalar↔vector reduction and the Green's-function taxonomy; optics owns this full scalar development).

Is the Fresnel-Kirchhoff integral the same thing as the Huygens-Fresnel principle?

They describe the same physical picture — the field beyond an aperture as a sum of secondary spherical wavelets — but at different levels of rigour. The Huygens–Fresnel principle asserts the wavelet sum and postulates the amplitude ( $1/\lambda$ ), the quarter-wave phase, and the obliquity factor to make it fit experiment. The Fresnel–Kirchhoff integral derives that same sum from the Helmholtz equation using Green's theorem, and out of the derivation come those three ingredients, no longer assumed. So the Fresnel–Kirchhoff integral is the Huygens–Fresnel principle with its promissory notes paid: same formula, but every factor now accounted for.

Where exactly does the obliquity factor come from in the derivation?

From two normal derivatives taken in Step 6. Green's identity leaves the integrand $U\,\partial_n G - G\,\partial_n U$ . Differentiating the Green's function $G = e^{ikr}/r$ along the aperture normal brings down a factor $\propto \cos\theta$ (the diffraction angle), and differentiating the incident spherical wave $U_{\text{inc}} = Ae^{ik\rho}/\rho$ along the normal brings down $\propto \cos\theta_0$ (the incidence angle). Adding them and factoring out the common $ik\,e^{ikr}/r$ leaves exactly $\tfrac12(\cos\theta_0 + \cos\theta)$ . The factor is thus not inserted — it is the residue of how a spherical wave and a spherical Green's function each vary along the screen normal.

What does the −i/λ prefactor actually mean physically?

It is two facts fused. The $1/\lambda$ says diffraction weakens as the wavelength shrinks: send $\lambda \to 0$ and the diffracted amplitude collapses relative to the geometric beam, which is why short-wavelength light casts sharp shadows and ray optics works. The $-i = e^{-i\pi/2}$ says every secondary wavelet leads the incident field by a quarter period — a $90°$ phase advance. Fresnel introduced both by hand to match measured fringe positions and intensities; Kirchhoff's derivation produces them automatically from the constants $ik/4\pi$ and the geometry, which is the clearest sign the wavelet model was right all along.

If Kirchhoff's boundary conditions are inconsistent, why do we still use the integral?

Because "inconsistent" here is a precise mathematical statement, not a claim that the answers are wrong in practice. The inconsistency — fixing both $U$ and $\partial_n U$ on a surface — corrupts the result only within a few wavelengths of the aperture edge and for sub-wavelength features. For the overwhelmingly common case of apertures many wavelengths across and observation at moderate angles, the Kirchhoff integral and the two mathematically clean Rayleigh–Sommerfeld integrals agree to leading order. So you get a formula that is easy to set up, physically transparent, and numerically accurate where it matters, at the cost of a rigour that only bites in regimes where you would switch to a vector solver anyway.

How do single-slit, Airy, and Cornu-spiral patterns all come from this one integral?

By choosing the aperture and the approximation of $r$ . In the far field ( $N_F \ll 1$ ) the integral is the Fourier transform of the aperture transmission: a slit of width $b$ transforms to $\operatorname{sinc}^2(\pi b\sin\theta/\lambda)$ , a rectangle to a product of two sincs, and a circular disk to the Airy pattern $[2J_1(u)/u]^2$ whose first dark ring at $1.22\,\lambda/2a$ sets every telescope's resolution. Keep the quadratic phase instead ( $N_F \sim 1$ , near field) and a straight edge produces the Fresnel integrals $C(w), S(w)$ traced by the Cornu spiral, with fringes leaking into the geometric shadow. One integral, three famous patterns — the difference is only which terms of $r$ you retain.

Why is this called a scalar theory, and what is lost?

Because it replaces the full electromagnetic vector field $\mathbf{E}$ with a single complex number $U$ per point, as the electromagnetism companion article spells out: light is really a vector field with boundary conditions, and treating it as one scalar throws away polarization and the currents induced at conducting edges. The reward is that a hard vector boundary-value problem becomes a single tractable integral, and the cost is invisible as long as the aperture is much larger than the wavelength — then the polarization coupling and edge effects are tiny corrections. Shrink the aperture toward $\lambda$ , though, and the scalar answer can be wrong by orders of magnitude; that regime belongs to vector diffraction theory, which optics recalls by name and does not attempt here.