Cornu Spiral and Fresnel Integrals

Near-field diffraction from a straight edge is notoriously fiddly: the intensity neither drops cleanly to zero in the shadow nor holds flat in the light, but ripples and overshoots in a way a brute-force integral only grudgingly gives up. Cornu's insight, in 1874, was that all of it lives on one fixed curve. Slice the wavefront into strips, drop a phasor arrow for each, and lay them head to tail; because the Fresnel phase advances as the square of position, the chain curls into a single universal spiral — the Cornu spiral, or clothoid. The complex field at any observation point is then just the straight chord joining the two spiral points that map to the aperture edges, and the intensity is that chord's length squared. A delicate calculation becomes a measurement with a ruler. The spiral coils into two fixed points — its "eyes" — at $(\tfrac12,\tfrac12)$ and $(-\tfrac12,-\tfrac12)$ , and every straight-edge and slit result is a chord swung between moving endpoints. On the geometric shadow boundary the chord runs from the spiral's centre to one eye, half the full diagonal, so the amplitude is halved and the intensity is exactly one quarter of the free field — the light does not switch off at the edge, it fades through a quarter. A little into the bright side the chord reaches its longest and the intensity overshoots to about $1.37$ times the unobstructed beam; deep in the shadow the endpoint winds into an eye and the field decays smoothly with no fringes. This article builds the spiral from the Fresnel integrals, proves it is the Euler clothoid, and reads the straight-edge pattern off the chord.

Key takeaways

Fresnel near-field diffraction from straight edges and slits reduces to two universal functions, the Fresnel integrals $C(v)$ and $S(v)$ , which accumulate the Huygens phasors across the aperture.
Plotting $S(v)$ against $C(v)$ traces the Cornu spiral (a clothoid): its arc length equals the coordinate $v$ and its curvature grows linearly, $\kappa=\pi v$ .
The field at any point is proportional to the chord joining the two spiral points that map to the aperture edges, with $C(\infty)=S(\infty)=\tfrac12$ fixing the eyes at $(\pm\tfrac12,\pm\tfrac12)$ .
A straight edge pins one endpoint at an eye and slides the other: at the geometric boundary $I/I_0=\tfrac14$ , the first bright fringe overshoots to $\approx1.37$ near $v\approx-1.22$ , and the shadow side decays monotonically to zero.
The construction is exact within the Fresnel (paraxial) approximation and passes into the Fraunhofer sinc as the Fresnel number goes to zero.
Key idea: within the Fresnel approximation the diffracted amplitude is a chord of one universal curve — the Cornu spiral $\big(C(v),S(v)\big)$ — drawn between the points that map to the aperture edges, so intensity is a chord length squared, giving $I/I_0=\tfrac14$ at a straight edge and a $1.37$ overshoot in the first fringe.

The core insight

Within the Fresnel approximation the diffracted complex amplitude equals a chord of a single universal curve — the Cornu spiral $\big(C(v),S(v)\big)$ — drawn between the points that map to the aperture edges; the intensity is the chord length squared. For a straight edge one endpoint is pinned at an eye $(\tfrac12,\tfrac12)$ and the other slides with observation position, giving $I/I_0=\tfrac14$ at the geometric shadow boundary, a $1.37$ overshoot in the first bright fringe near $v\approx-1.22$ , and a smooth, fringe-free decay into the shadow.

Where this lives today

The Cornu construction is still the quickest route from geometry to a near-field pattern, and its calibrated straight-edge curve underlies real measurement. Knife-edge tests — the Foucault test that shaped every large telescope mirror, and modern beam-profiling — locate a focus or beam waist from the position of that first $1.37$ fringe, and the edge-ringing the spiral predicts sets the contrast limit at aperture stops in cameras and mask edges in lithography. Wherever a coherent beam meets a hard edge, the ripple it leaves is a chord of this curve. The clothoid's second life is even more visible. Because its curvature rises linearly with distance, it is the transition curve that eases railway track and highway ramps from straight into circular arc without a jump in lateral acceleration, and it forms the entry and exit of a well-designed roller-coaster loop. It has migrated into digital design too — the Euler spiral is the backbone of curvature-continuous path planning for robots and self-driving cars, and of font tools that fair a stroke by holding curvature smooth. One curve, drawn once by Euler for a bent rod and by Cornu for a diffracted wave, now steers trains, coasters, and cursors.

The physical picture

Walk along the open wavefront and drop a phasor arrow for each strip. Near the geometric-ray direction the Fresnel phase $\tfrac{\pi}{2}v^2$ changes slowly, so consecutive arrows point almost the same way and the running sum races ahead — the gently curving inner part of the spiral. Move outward and the quadratic phase spins the arrows faster, curling them into ever-tighter loops that wind onto the two eyes; those far strips contribute almost nothing net, which is why the spiral converges. The whole diffraction pattern is the tip of this phasor chain, tracked as the aperture edges slide its endpoints along a curve that never changes.

Figure

The Cornu spiral (clothoid): plotting S(v) against C(v). Arc length = v, curvature κ = πv; the eyes sit at (½,½) and (−½,−½). The chord from the centre to an eye is the straight-edge amplitude at v = 0, of squared length ½ → I = I₀/4

The chord picture makes edges intuitive. A knife edge pins one endpoint at an eye — the blocked half of the wavefront contributes nothing — and lets the other slide with the observation position. Into the light, the sliding endpoint climbs the loose outer turns, so the chord rises and falls: fringes, and an overshoot before it settles. Into the shadow, it winds tightly into the near eye, so the chord shrinks smoothly and the field fades without a ripple. All of straight-edge diffraction is one moving chord on one fixed curve — and the next figure tells that story as an intensity plot, with the two defining numbers marked.

Figure

Straight-edge diffraction: I/I₀ versus the Fresnel coordinate v. On the geometric edge (v = 0) the intensity is exactly ¼ (−6 dB); the first bright fringe overshoots to 1.37 near v ≈ −1.22; the shadow side (v > 0) decays monotonically toward zero

Read left to right, the plot is the whole physics of an edge. Far into the light ( $v\to-\infty$ ) the endpoint reaches the far eye and the intensity settles to $I_0$ , fringes averaging out. Approaching the edge from the bright side it climbs to the $1.37$ overshoot at $v\approx-1.22$ — a $37\%$ excess over the open beam, the bright rim hugging any sharp shadow. On the geometric boundary it passes through exactly $I_0/4$ , and beyond it, in the shadow, slides monotonically to zero with no oscillation. The asymmetry — ringing on one side, silence on the other — is the signature of an edge, not a slit.

The core equations

Everything is built from two tabulated integrals: they accumulate the real and imaginary parts of the running phasor sum across the wavefront, with the quadratic Fresnel phase $\tfrac{\pi}{2}w^2$ in the integrand, and their limits fix where the spiral winds up.

C(v)=\int_0^{v}\cos\!\Big(\tfrac{\pi}{2}w^2\Big)\,\mathrm{d}w,\qquad S(v)=\int_0^{v}\sin\!\Big(\tfrac{\pi}{2}w^2\Big)\,\mathrm{d}w,\qquad C(\pm\infty)=S(\pm\infty)=\pm\tfrac12 \qquad (1)

where: $v$ — dimensionless upper limit (the aperture edge in Fresnel units); $w$ — integration variable running along the wavefront; $C,S$ — real and imaginary parts of the accumulated phasor. Both are odd functions, and both tend to $\pm\tfrac12$ as $v\to\pm\infty$ — the coordinates of the spiral's two eyes. The dimensionless $v$ compresses the geometry — wavelength, source and screen distances — into one number, so a single curve serves every straight-edge problem. It comes from the quadratic (Fresnel) expansion of the path length; for a plane wave it reduces to a clean square-root scaling.

v=y\sqrt{\dfrac{2}{\lambda}\Big(\dfrac{1}{r_0}+\dfrac{1}{r}\Big)}\;\xrightarrow{\;r_0\to\infty\;}\;y\sqrt{\dfrac{2}{\lambda r}} \qquad (2)

where: $y$ — physical transverse distance of the observation point from the geometric shadow edge; $\lambda$ — wavelength; $r_0,r$ — source-to-aperture and aperture-to-screen distances. Inverting the plane-wave form, a feature at coordinate $v$ sits at $y=v\sqrt{\lambda r/2}$ , so the fringe scale grows as $\sqrt{\lambda r}$ . Plotting $S(v)$ against $C(v)$ traces the complex function $Z(v)=C(v)+iS(v)$ . Differentiating it exposes the curve's geometry: it is traced at unit speed, so $v$ is arc length, and its curvature grows linearly with $v$ — the definition of a clothoid, or Euler spiral.

Z(v)=\int_0^{v}e^{\,i\pi w^2/2}\,\mathrm{d}w,\qquad \big|Z'(v)\big|=1,\qquad \theta(v)=\tfrac{\pi}{2}v^2,\qquad \kappa(v)=\pi v \qquad (3)

where: $Z$ — point on the Cornu spiral in the complex $C$ – $S$ plane; $Z'$ — unit tangent; $\theta$ — tangent angle, equal to the accumulated Fresnel phase; $\kappa$ — local curvature, linear in arc length. Because $\theta=\tfrac{\pi}{2}v^2$ is exactly the physical phase, the geometry is the phasor diagram — no separate bookkeeping is needed. The physical content is one line: the diffracted amplitude is the chord spanning the two aperture edges, normalized so the fully open wavefront (edges at the two eyes) reproduces the unobstructed field. The normalization constant is the eye-to-eye chord — the complete Fresnel integral.

U(P)=\frac{U_0}{1+i}\big[\,Z(v_2)-Z(v_1)\,\big],\qquad \int_{-\infty}^{\infty}e^{\,i\pi w^2/2}\,\mathrm{d}w=1+i \qquad (4)

where: $U_0$ — unobstructed complex amplitude at $P$ ; $v_1,v_2$ — Fresnel coordinates of the two aperture edges; $1+i=\sqrt2\,e^{i\pi/4}$ — the eye-to-eye reference chord, $|1+i|^2=2$ . For a fully open wavefront ( $v_1\to-\infty$ , $v_2\to+\infty$ ) the bracket becomes $1+i$ and $U(P)=U_0$ , as it must. For a straight edge one aperture edge sits at infinity (its eye) while the other slides with the observation point. The chord from the moving $Z(v)$ to the eye $Z(+\infty)=(\tfrac12,\tfrac12)$ , squared, gives the intensity in a form evaluable from a table or from `scipy.special.fresnel`.

\frac{I}{I_0}=\tfrac12\Big\{\big[\tfrac12-C(v)\big]^2+\big[\tfrac12-S(v)\big]^2\Big\} \qquad (5)

where: $v$ — Fresnel coordinate of the observation point relative to the geometric edge, with $v<0$ in the illuminated region and $v>0$ in the shadow. At $v=0$ this gives $\tfrac12(\tfrac14+\tfrac14)=\tfrac14$ ; its maximum $\approx1.37$ falls near $v\approx-1.22$ ; and it decays monotonically to $0$ as $v\to+\infty$ .

Derivation AFrom the Fresnel–Kirchhoff integral to a chord of the spiral

The claim that a delicate near-field integral is just a chord of a fixed curve deserves a proof. Our aim is to reduce the one-dimensional paraxial diffraction integral over an aperture strip to a difference of Fresnel integrals, $Z(v_2)-Z(v_1)$ , and fix the normalization so a fully open wavefront returns the unobstructed field. We keep only the quadratic phase, substitute the Fresnel coordinate, and recognise the definite integral as a chord.

Step 1. Start from the paraxial field of a wavefront limited to $y\in[y_1,y_2]$ and translationally invariant along the edge: $U(P)\propto\displaystyle\int_{y_1}^{y_2}\exp\!\Big[i\tfrac{\pi}{\lambda}\Big(\tfrac1{r_0}+\tfrac1r\Big)y^2\Big]\,\mathrm{d}y$ . Why: over the small transverse scales that matter, only the quadratic (Fresnel) term of the path length survives; the slowly varying obliquity and $1/r$ factors are constant across the strip and are absorbed into the prefactor $U_0$ .

Step 2. Substitute the Fresnel coordinate $y=v\big[\tfrac{\lambda}{2}(\tfrac1{r_0}+\tfrac1r)^{-1}\big]^{1/2}$ from $(2)$ . The exponent becomes $i\tfrac{\pi}{2}v^2$ and $\mathrm{d}y$ a constant times $\mathrm{d}v$ , folding into $U_0$ . Why: this is the point of the dimensionless coordinate — every straight-edge geometry becomes the same integrand $e^{i\pi v^2/2}$ , so one curve serves all.

Step 3. Recognise the definite integral as a spiral chord: $\displaystyle\int_{v_1}^{v_2}e^{\,i\pi w^2/2}\,\mathrm{d}w=Z(v_2)-Z(v_1)$ , the straight vector connecting two points of $Z(v)=C(v)+iS(v)$ . Why: the running integral $Z(v)$ is by definition the tip of the phasor chain, so the field from a finite aperture is the displacement between the chain's start and end — a chord.

Step 4. Fix the normalization by the fully open wavefront: $\displaystyle\int_{-\infty}^{\infty}e^{\,i\pi w^2/2}\,\mathrm{d}w=1+i$ , so demanding $U(P)=U_0$ there forces the prefactor $1/(1+i)$ , giving $U(P)=\tfrac{U_0}{1+i}[Z(v_2)-Z(v_1)]$ . Why: the complete Fresnel integral equals the eye-to-eye chord $1+i$ (a Gaussian/Fresnel integral evaluation), and dividing by it calibrates the chord so that "no obstruction" reproduces the incident field. *Verified: the eye-to-eye reduction $Z(+\infty)-Z(-\infty)=(\tfrac12+\tfrac i2)-(-\tfrac12-\tfrac i2)=1+i$ and $|1+i|^2=2$ in verify_aq-optics-cornu-spiral-fresnel-integrals.py.*

Go deeper · Derivation B · 1 minThe straight-edge intensity and the 1.37 overshoot

Derivation BThe straight-edge intensity and the 1.37 overshoot

The two numbers everyone quotes for a knife edge — a quarter-intensity on the shadow line and a first fringe overshooting the free field by about $37\%$ — both fall straight out of the chord construction. Our aim is to evaluate the edge chord at the geometric boundary and at its longest, using the eye coordinates and the tabulated integrals. We pin one endpoint at an eye, slide the other, and read off the special values.

Step 1. For a half-plane, the blocked edge sits at $v_1\to-\infty$ , so $Z(v_1)=-\tfrac12-\tfrac i2$ ; the open region runs to $+\infty$ . The intensity is $\tfrac{I}{I_0}=\tfrac12\big|Z(+\infty)-Z(v)\big|^2=\tfrac12\{[\tfrac12-C(v)]^2+[\tfrac12-S(v)]^2\}$ . Why: only the transmitted strip from the observation-dependent edge $v$ out to the far eye contributes, so the chord runs from $Z(v)$ to $Z(+\infty)=(\tfrac12,\tfrac12)$ .

Step 2. Evaluate on the geometric boundary, $v=0$ : $C(0)=S(0)=0$ , so $\tfrac{I}{I_0}=\tfrac12(\tfrac14+\tfrac14)=\tfrac14$ . Why: at the edge the chord runs from the spiral's centre to one eye — exactly half the eye-to-eye diagonal — so the amplitude is halved and the intensity is quartered; the light fades through a quarter, it does not switch off.

Step 3. Slide into the light: the chord is longest at $v\approx-1.217$ , where $C,S$ put the endpoint on the first outward turn farthest from the eye, giving $\tfrac{I}{I_0}\approx1.37$ . Why: the outer loops swing the endpoint beyond the eye, so the chord briefly exceeds the eye-to-eye reference — a genuine overshoot above the open beam.

Step 4. Slide into the shadow, $v\to+\infty$ : $C,S\to+\tfrac12$ , the endpoint winds into the eye $(\tfrac12,\tfrac12)$ , the chord shrinks to zero, and $I/I_0\to0$ monotonically. Why: on the shadow side the endpoint is trapped in the tightening spiral around the eye, so the chord only ever shortens — no loops are swung, hence no fringes. *Verified with `scipy.special.fresnel`: $I/I_0=0.25$ at $v=0$ , maximum $1.370$ at $v\approx-1.217$ , and $I/I_0\approx0.012$ at $v=+2$ in verify_aq-optics-cornu-spiral-fresnel-integrals.py.*

Go deeper · Derivation C · 1 minClothoid geometry: unit speed and linear curvature

Derivation CClothoid geometry: unit speed and linear curvature

Why is this spiral — and not an Archimedean or logarithmic one — the curve of diffraction? Our aim is to show $Z(v)$ is the Euler clothoid: unit speed, curvature proportional to arc length, tangent angle equal to the accumulated Fresnel phase. We differentiate the spiral function and read the geometry off its tangent.

Step 1. Differentiate: $Z'(v)=e^{\,i\pi v^2/2}$ , so $|Z'(v)|=1$ . Why: unit modulus means $v$ is the natural arc-length parameter — equal strips of wavefront map to equal lengths of spiral, so the drawing tracks accumulated phase faithfully.

Step 2. Read the tangent direction: $\theta(v)=\arg Z'(v)=\tfrac{\pi}{2}v^2$ . Why: the direction each phasor points is precisely the Fresnel phase it has accumulated, so the geometry and the physics are the same object — the tangent angle is the phase.

Step 3. Differentiate along the arc: $\kappa=\dfrac{\mathrm{d}\theta}{\mathrm{d}s}=\dfrac{\mathrm{d}\theta}{\mathrm{d}v}=\pi v$ , since $s=v$ . Curvature grows linearly with arc length — the defining clothoid property, and the loops tighten ( $1/\kappa=1/\pi v$ ) as $|v|$ grows. Why: linear curvature is what makes the spiral wind ever faster onto its eyes, and it is the same property that makes the clothoid the ideal transition curve for roads and railways. *Verified: $|Z'(v)|=1$ , $\theta(1)=\pi/2$ , and $\kappa(1)=\pi$ in verify_aq-optics-cornu-spiral-fresnel-integrals.py.*

Limiting cases

Case	Result	Meaning
Unobstructed wavefront	chord $=1+i$ , $\|1+i\|^2=2\Rightarrow I=I_0$	both endpoints at opposite eyes; the full free field
Straight edge, geometric boundary ( $v=0$ )	$I=\tfrac14 I_0$	chord = centre-to-eye, half the diagonal → amplitude $\tfrac12$ , intensity $\tfrac14$ ( $-6$ dB)
Straight edge, first bright fringe ( $v\approx-1.22$ )	$I\approx1.37\,I_0$	chord swings to the first outer turn — the maximum overshoot
Deep shadow ( $v\to+\infty$ )	$I\to0$ monotone	endpoint winds into an eye; the chord shrinks with no fringes
Fraunhofer limit, $N_F\to0$	inner arc $\approx$ straight $\to$ sinc	only the nearly-straight inner arc is used; the slit collapses to the far-field sinc

Common trapReading numbers off the spiral without checking the convention or the light/shadow sign

Two traps sit here, and both flip your answer. First, the convention: the "normalized" integrals used throughout — with $\cos(\tfrac{\pi}{2}w^2)$ in the integrand — put the eyes at $(\pm\tfrac12,\pm\tfrac12)$ , but some texts and libraries define $C,S$ with a bare $t^2$ argument, moving the eyes to $\pm\tfrac12\sqrt{\pi/2}$ and rescaling every length. Locate the eyes before measuring a chord. (In code, `scipy.special.fresnel(x)` uses the $\pi/2$ convention and returns the pair as $\big(S(x),C(x)\big)$ — $S$ first.) Second, the light/shadow sign: the straight-edge formula must be $I/I_0=\tfrac12\{[\tfrac12-C(v)]^2+[\tfrac12-S(v)]^2\}$ with $v<0$ in the light, so $v\to-\infty$ returns $I_0$ and $v\to+\infty$ returns $0$ . Pair the $[\tfrac12+C(v)]$ form with a " $v>0$ is light" labelling and you place the fringes on the wrong side of the edge and predict a bright shadow. Fix the formula and the sign of $v$ together, and check both limits before trusting any middle value.

Worked examples

1 · Where the first bright fringe of a knife edge falls (confidence builder)

A collimated beam of green light, $\lambda=546$ nm, is cut by a straight edge; the pattern is observed on a screen $r=2$ m beyond the edge. Find the physical position of the first bright fringe and its intensity.

Step 1. Set the coordinate scale. For a plane wave, equation $(2)$ gives $v=y\sqrt{2/(\lambda r)}$ , so a feature at coordinate $v$ sits at $y=v\sqrt{\lambda r/2}$ . Here $\sqrt{\lambda r/2}=\sqrt{(546\times10^{-9})(2)/2}=\sqrt{5.46\times10^{-7}}=7.39\times10^{-4}$ m $=0.739$ mm. Each unit of $v$ is about three-quarters of a millimetre on the screen.

Step 2. The first bright fringe is the intensity maximum at $v\approx-1.217$ (negative, so on the illuminated side of the geometric edge). Its position is $y=-1.217\times0.739\ \text{mm}=\mathbf{-0.90\ mm}$ — nine-tenths of a millimetre into the light from the shadow line.

Step 3. Its intensity comes from equation $(5)$ evaluated at $v=-1.217$ : $I/I_0=\mathbf{1.37}$ . The brightest edge fringe is $37\%$ brighter than the fully unobstructed beam — the luminous rim that hugs any sharp shadow in coherent light.

Step 4. Later maxima crowd toward the edge (roughly $\sqrt{}$ -spaced, $v\approx-1.87,-2.35,\dots$ ), each weaker, so the fringe spacing shrinks and contrast fades outward; move the screen back and the whole pattern scales as $\sqrt{\lambda r}$ . First bright fringe at $0.90$ mm inside the edge, $I/I_0=1.37$ .

2 · The intensity at the geometric edge and into the shadow (exam level)

For the same knife edge, find the intensity exactly on the geometric shadow boundary, express it in decibels, and describe how the intensity behaves as the observation point moves into the shadow.

Step 1. On the geometric boundary the observation point maps to $v=0$ , where $C(0)=S(0)=0$ . Equation $(5)$ gives $I/I_0=\tfrac12\{(\tfrac12)^2+(\tfrac12)^2\}=\tfrac12\cdot\tfrac12=\mathbf{1/4}$ . The amplitude is exactly halved at the edge, so the intensity is a quarter — not zero, and not a half.

Step 2. In decibels, $10\log_{10}(1/4)=\mathbf{-6.0\ dB}$ : the geometric shadow edge always sits $6$ dB below the free field, a handy calibration point for locating an edge or a focus in a beam scan.

Step 3. Moving into the shadow ( $v>0$ ), the endpoint winds into the eye $(\tfrac12,\tfrac12)$ and the chord shrinks smoothly. By $v=+2$ the intensity has fallen to $I/I_0\approx\mathbf{0.012}$ — roughly $1\%$ of the free field — with no oscillation on the way down.

Step 4. The contrast with the previous example is the whole diagnostic: the light side overshoots and rings, the shadow side decays in silence. That one-sided fringe pattern tells you at a glance that a straight edge — not a slit or wire — cast it. $I/I_0=1/4$ ( $-6$ dB) at the edge; monotonic decay to $\approx0.012$ by $v=+2$ .

Physics of the portrait

The portrait is a curve, not a person — but a curve with a rich double life. The same clothoid Cornu drew to tame diffraction is the transition spiral engineers lay between straight track and circular curve, so a train's lateral acceleration builds up linearly rather than in a jolt. Constant rate of change of curvature is what a passenger's comfort and a Huygens phasor's bookkeeping both require.

Origins · 1 mincornu spiral and how this was discovered

cornu spiral

the graphical phasor engine of near-field diffraction · Fresnel (near-field) diffraction

The spiral is named for Marie Alfred Cornu (1841–1902), the French physicist who in 1874 recast Fresnel's near-field integrals as a single graphical construction, turning the evaluation of a diffraction pattern into the drawing of a chord on a fixed curve. Cornu is better remembered by the public for a different feat — his 1878 refinement of Fizeau's rotating-toothed-wheel measurement of the speed of light, one of the most precise of the century — but the spiral is his most enduring gift to optics teaching. The curve is older than its optical use: Leonhard Euler studied it in 1744 while analysing the elastica, the shape of a bent elastic rod, so it is also called the Euler spiral. Its defining property, curvature proportional to arc length, was later seized upon by railway and highway engineers — Arthur Talbot's Railway Transition Spiral made it the standard easement between tangent and curve — and today the same clothoid smooths motorway interchanges, roller-coaster loops, and the strokes of digitally designed typefaces. In optics it remains the fastest way to see, rather than compute, what a straight edge does to a wavefront.

History

1744

— Euler finds the spiral whose curvature grows linearly with arc length, studying the elastica

1818

— Fresnel's prize memoir founds near-field diffraction on the integrals

C(v)

S(v)

1874

— Cornu recasts the Fresnel integrals as one universal graphical spiral

1901

— Talbot's Railway Transition Spiral applies the same curve to track easements

Going further

Texts. Hecht, Optics, 5th ed., §10.3 (Fresnel diffraction, the Cornu spiral) · Born & Wolf, Principles of Optics, 7th ed., §8.7 · Saleh & Teich, Fundamentals of Photonics, 3rd ed., Ch. 4 · Goodman, Introduction to Fourier Optics, 3rd ed., §4.5 · Abramowitz & Stegun, Handbook of Mathematical Functions, Ch. 7 (Fresnel integrals) · Cornu, A., J. Phys. (Paris) 3, 5 (1874).

Provenance. Symbols and conventions follow Hecht §10.3 and Born & Wolf §8.7. A verifier should, independently of any export: (1) confirm the Fresnel-integral definitions and the eye limits $C(\pm\infty)=S(\pm\infty)=\pm\tfrac12$ against Abramowitz & Stegun Ch. 7; (2) recompute the straight-edge intensity giving $I/I_0=\tfrac14$ at $v=0$ and the maximum $\approx1.37$ near $v\approx-1.22$ ; (3) verify the clothoid relations $|Z'|=1$ , $\theta=\tfrac{\pi}{2}v^2$ , $\kappa=\pi v$ ; (4) check the normalization $\int_{-\infty}^{\infty}e^{i\pi w^2/2}\mathrm{d}w=1+i$ . All numbers are standard textbook values; every one is reproduced in verify_aq-optics-cornu-spiral-fresnel-integrals.py using `scipy.special.fresnel` (never a truncated numerical integral of the oscillatory tail, which under-resolves the eyes). Convention note: the straight-edge formula is written with $[\tfrac12-C(v)]$ and $v<0$ in the light, so that $v\to-\infty$ returns $I_0$ and $v\to+\infty$ returns $0$ ; the equivalent $[\tfrac12+C(v)]$ form requires the opposite sign of $v$ .

Prerequisites. Huygens–Fresnel Principle · Fresnel–Kirchhoff Diffraction Integral · Fresnel Diffraction Regime · Fresnel Number. Leads to. Single-Slit Diffraction · Fraunhofer Diffraction Regime · Fresnel Zones and Zone Plates · Rectangular Aperture Diffraction.

Why is the intensity a quarter at the edge, and not a half?

Because intensity is amplitude squared, and it is the amplitude that halves. On the geometric boundary the observation point maps to $v=0$ , so the phasor chain runs from the centre $(0,0)$ to the eye $(\tfrac12,\tfrac12)$ — exactly half the eye-to-eye diagonal. Half the amplitude means one quarter of the intensity: $I/I_0=\tfrac14$ . An edge blocks half the wavefront, so you might guess the field halves — and it does, in amplitude. Squaring turns "half the light gets through" into "a quarter of the intensity survives," a factor students routinely lose.

Why do fringes appear only on the bright side of a straight edge?

It is the geometry of the spiral. Moving into the light, the endpoint climbs the loose outer turns, so the chord to the fixed eye lengthens and shortens repeatedly — bright and dark fringes, with the $1.37$ overshoot the first and largest. Moving into the shadow, it winds into the near eye along tightening loops, so the chord only ever shrinks — a smooth, monotone fade. The asymmetry is intrinsic: one side samples the spiral's open arcs, the other its collapsing coils. Fringes on one side and silence on the other is how you identify an edge pattern by eye.

Is the 1.37 overshoot real, or an artefact of the approximation?

It is real and routinely observed. In monochromatic, spatially coherent light a straight edge casts a bright rim just inside the illuminated side — brighter than the open beam — followed by fainter fringes, exactly the $I/I_0\approx1.37$ maximum at $v\approx-1.22$ and its successors. It is the same physics as the bright fringe on the lit side of any sharp shadow under a coherent source. The Fresnel (paraxial, quadratic-phase) approximation is excellent whenever the Fresnel number is around unity and angles are small — essentially all tabletop edge diffraction. Only for grazing geometries or sub-wavelength edges does the scalar picture need replacing.

Which convention should I use, and where exactly do the eyes land?

Use the "normalized" integrals with the $\tfrac{\pi}{2}w^2$ argument, as Hecht, Born & Wolf, and this article do: then $C(\pm\infty)=S(\pm\infty)=\pm\tfrac12$ and the eyes sit at $(\pm\tfrac12,\pm\tfrac12)$ . Beware the alternative with a bare $t^2$ integrand, common in pure-math references, whose eyes land at $\pm\tfrac12\sqrt{\pi/2}\approx\pm0.627$ and whose coordinate is rescaled by $\sqrt{\pi/2}$ . Mixing the two is the classic Cornu-spiral error. Before reading any chord off a plot, check where the curve winds up — eyes at $\tfrac12$ means the optics convention, and every number here applies directly.

Why do engineers use the same curve for railway track and roller-coasters?

Because the clothoid is the curve of gradually changing curvature. On straight track curvature is zero; on a circular curve it is fixed; jumping between them would jerk a train with a sudden step in lateral acceleration. A clothoid transition, with $\kappa\propto$ distance travelled, ramps the sideways acceleration up linearly — the same $\kappa=\pi v$ that winds the Cornu spiral. Roller-coaster loops use it to spread the g-force onset rather than deliver a spike. It is a deep coincidence that the curve summing Huygens phasors most conveniently is also the one that eases a body through a turn most gently: both ask curvature to build at a constant rate.

How do I actually compute C(v) and S(v) in practice?

Call a library special function — do not numerically integrate the definition to a finite bound. In Python, `scipy.special.fresnel(x)` returns the pair $\big(S(x),C(x)\big)$ ( $S$ first) in the $\tfrac{\pi}{2}$ convention, accurate into the deep tail. Naive quadrature fails because $\cos(\tfrac{\pi}{2}w^2)$ oscillates ever faster as $w$ grows, so integrating to a large finite $v$ under-resolves those oscillations and wobbles around $\tfrac12$ instead of converging. The special-function routines use the asymptotic form $C(v)\sim\tfrac12+\tfrac1{\pi v}\sin(\tfrac{\pi}{2}v^2)$ for large $v$ , capturing the eyes exactly. This article's verifier is computed this way.