Holography and Wavefront Recording

A photograph is a confession of failure: it records where light was bright and where it was dim, and throws away everything about the direction each ray was travelling — the phase — which is precisely the information that told your eye how far away things were. Holography is the repair. It borrows two results from earlier in this chapter, the two-beam interference law from Interference and Coherence and the diffraction-integral machinery from Scalar Diffraction Theory, and welds them into a two-step trick. First, superpose the light scattered by an object with a mutually coherent reference wave on a square-law medium; the fringe pattern that develops is a frozen record of the object's phase, encoded as brightness. Second, re-illuminate that pattern with the reference alone and let it diffract: one diffracted order is, term for term, the original object wave rebuilt from scratch. An observer sees a fully three-dimensional virtual image, with parallax and depth of focus, as if looking at the vanished object through a window. The mathematics is entirely wave optics — interference recording followed by diffraction readout — and collapses to multiplying the incident field by the hologram's amplitude transmittance. Because it lives or dies on interference, holography lives or dies on coherence: the object and reference path lengths must match to well within the source's coherence length, which is why the idea waited twelve years, from Gabor's 1948 proposal to the laser of 1960, for a source that could make it work. The same principle then generalizes into volume Bragg holograms, digital holography on a sensor array, and the real-time phase conjugation of nonlinear optics.

Key takeaways

A square-law detector records only intensity and discards phase; holography converts phase into recordable intensity by interfering the object wave with a coherent reference.
The recording is $I=|U_o+U_r|^2$ , whose two cross terms carry the object field; developing gives a transmittance linear in those terms.
Re-illuminating with the reference regenerates four beams: a transmitted bias, a halo, a virtual image $\propto U_o$ , and a conjugate real image $\propto U_o^{*}$ .
Gabor's in-line geometry overlaps the twin images; the off-axis Leith–Upatnieks geometry puts a spatial carrier on the cross terms so the four beams fan apart.
Holography needs a path mismatch far below the coherence length $\ell_c=\lambda^2/\Delta\lambda$ (hence lasers) and a medium resolving thousands of lines per millimetre.
Thin holograms cap at 6.25 percent (amplitude) or 33.9 percent (phase) efficiency; thick Bragg-regime volume holograms reach nearly 100 percent.
Key idea: a square-law detector cannot record phase directly, but interference with a known reference converts phase into a recordable intensity modulation, and diffraction of that reference from the recorded pattern regenerates the original wavefront.

The core insight

A square-law detector cannot record phase directly, but interference with a known reference converts phase into a recordable intensity modulation, and diffracting the reference wave back off that recorded fringe pattern regenerates the original object wavefront in full.

Where this lives today

The wavefront-recording principle now runs mostly on sensor arrays and computers. In digital holography the plate is a CCD or CMOS chip, the interference field is captured as pixels, and the object wave is reconstructed by numerically propagating it with a Fresnel or angular-spectrum integral. Because both amplitude and phase are recovered, one can refocus a single exposure to any plane after capture and read out quantitative phase — the optical thickness of a transparent object to a few nanometres — which is why digital holographic microscopy images living cells without stains and inspects the flatness of wafers and MEMS. The thick-hologram branch is just as alive in hardware. Volume Bragg gratings in photo-thermo-refractive glass lock and narrow high-power diode lasers, and photopolymer volume holograms are the diffractive combiners in augmented-reality headsets, steering a projector's image into the eye while the world shows through. The same coupled-wave physics underlies the diffractive waveguides shipping in wearable displays. Meanwhile holographic interferometry — comparing an object's wavefront before and after a stress — remains a standard whole-field, non-contact metrology tool, detecting delamination in turbine blades and mapping vibration modes to a fraction of a wavelength.

The physical picture

Treat the reference beam as a radio carrier and the object beam as the signal riding on it. Interference multiplies the two, imprinting the object's phase onto the position and spacing of the recorded fringes just as amplitude modulation stamps an audio signal onto a carrier. Reconstruction is demodulation: the reference re-illuminates the fringes and the object wave reads straight back out. Nothing here is optical magic — a hologram is a diffraction grating whose fringes merely happen to be complicated, and if you understand a grating splitting a plane wave into orders, you understand the scheme. The encoding is radically distributed: each object point scatters a wave covering the entire plate, and each plate point receives light from every object point. That is the sense in which a hologram is holos, whole — cut it in half and each half still reconstructs the complete scene, dimmer and at lower resolution, as masking half a lens still forms the whole image. The first figure shows the recording geometry and its fringe spacing.

Figure

Two mutually coherent beams meeting at 30° write a fringe pattern into the plate: with λ = 633 nm the fringes have spacing d = λ/(2 sin(θ/2)) = 1.22 µm, a carrier νc = 818 lines per mm that only a high-resolution emulsion resolves

The four beams on reconstruction come straight from squaring a sum: the detector registers $|U_o+U_r|^2$ , the self-intensities $|U_o|^2$ and $|U_r|^2$ plus the cross terms $U_oU_r^{*}$ and $U_o^{*}U_r$ . On readout these become a transmitted beam, a halo, the reconstructed object wave, and its phase-conjugate twin. Gabor's in-line arrangement let all four pile up along one axis — the twin-image problem. The off-axis fix is purely geometric: tilt the reference, and the cross terms acquire a spatial carrier that flings their orders off in separate directions, as the second figure shows.

Figure

Reading the hologram with the reference splits the light into three: a straight-through zero order, a diverging wave that reconstructs the object (the virtual image an eye sees in 3-D), and a converging conjugate wave forming a real image — the off-axis tilt fans them apart

The core equations

A square-law medium responds to the time-averaged intensity, so we superpose the object field $U_o$ and reference field $U_r$ and take the squared modulus. Expanding the square is the entire content of holographic recording: two self-intensity terms and two cross terms, and only the cross terms know the object's phase.

I(x,y)=|U_o+U_r|^2=|U_o|^2+|U_r|^2+U_oU_r^{*}+U_o^{*}U_r \qquad (1)

where: $U_o(x,y)$ — complex object field at the recording plane; $U_r(x,y)$ — complex reference field at the recording plane; $I(x,y)$ — time-averaged intensity registered by the medium. The first two terms are a slowly varying bias; the pair $U_oU_r^{*}+U_o^{*}U_r=2|U_o||U_r|\cos(\phi_o-\phi_r)$ is where the object phase $\phi_o$ survives, written as a brightness modulation. A developed silver-halide emulsion (or photopolymer, or dichromated gelatin) has an amplitude transmittance roughly linear in exposure $E=I\tau$ . Passing the recorded intensity through that linear response, with a constant bias $t_b$ absorbing the self-intensities, leaves a transmittance carrying exactly the two cross terms.

t(x,y)=t_b+\beta\,\tau\,\big(U_oU_r^{*}+U_o^{*}U_r\big) \qquad (2)

where: $t(x,y)$ — amplitude transmittance of the finished hologram; $t_b$ — bias transmittance from the $|U_o|^2+|U_r|^2$ terms; $\beta$ — slope of the amplitude-transmittance-versus-exposure curve, taken linear; $\tau$ — exposure time. The object information now lives entirely in the modulation about $t_b$ . Reconstruction is a single multiplication: shine the reference $U_r$ on the hologram and the field just behind it is $U_{\text{out}}=U_r\,t$ (thin-hologram limit). Distributing and using $U_rU_r^{*}=|U_r|^2$ sorts the light into a bias beam, a term $\propto U_o$ , and a term $\propto U_o^{*}$ .

U_{\text{out}}=U_r\,t=\underbrace{t_bU_r}_{\text{bias}}+\underbrace{\beta\tau|U_r|^2\,U_o}_{\text{virtual image}}+\underbrace{\beta\tau\,U_r^{2}\,U_o^{*}}_{\text{real image}} \qquad (3)

where: $U_{\text{out}}$ — field emerging on readout; $\beta\tau|U_r|^2U_o$ — a scaled copy of the object wave, diverging as if from the absent object, an upright virtual image; $\beta\tau U_r^{2}U_o^{*}$ — the phase-conjugate wave, converging to a real, depth-inverted (pseudoscopic) image. This one line is why holography works. To keep those three beams from overlapping, tilt the reference. Two waves crossing at the plate write fringes whose spatial frequency is the difference of their direction sines; for an on-axis object and a reference tilted by $\theta_r$ this is a pure carrier the medium must resolve.

\nu_c=\frac{\sin\theta_r-\sin\theta_o}{\lambda}\;\approx\;\frac{\sin\theta_r}{\lambda} \qquad (4)

where: $\nu_c$ — carrier (fringe) spatial frequency; $\theta_r,\theta_o$ — angles of reference and object waves at the plate; $\lambda$ — recording wavelength. Larger tilt means a finer carrier and a more demanding medium. The corresponding fringe period, for a total angle $\theta$ between the beams, follows the same geometry.

d=\frac{\lambda}{2\sin(\theta/2)} \qquad (5)

where: $d$ — fringe period, measured normal to the fringe planes; $\theta$ — full angle between object and reference beams. As $\theta\to180^\circ$ (reflection holograms) the period shrinks toward $\lambda/2$ , demanding the highest-resolution media. If the emulsion is thick compared with the fringe spacing, it becomes a stack of partially reflecting planes: light adds in phase only at the Bragg angle, giving sharp wavelength and angular selectivity and routing nearly all the light into one order.

2d\sin\theta_B=\lambda \qquad (6)

where: $\theta_B$ — Bragg angle from the fringe planes inside the medium; $d$ — fringe period; $\lambda$ — reconstruction wavelength in the medium. It is the same relation that governs X-ray diffraction from crystal planes, applied to a light-written grating. Every step above assumed high-contrast interference — a statement about coherence. Fringe visibility survives only while the path mismatch stays far below the coherence length, which fixes how monochromatic the source must be.

\ell_c=\frac{c}{\Delta\nu}=\frac{\lambda^2}{\Delta\lambda} \qquad (7)

where: $\ell_c$ — longitudinal coherence length; $\Delta\nu,\Delta\lambda$ — spectral linewidth in frequency and wavelength; $c$ — vacuum speed of light. For a filtered lamp $\ell_c$ is micrometres; for a laser, centimetres to metres — the entire reason holography waited for the laser.

Derivation AFrom recorded fringes to a reconstructed wavefront

The goal is to prove the central claim of holography: that multiplying the incident reference wave by the developed hologram's transmittance regenerates the original object wave exactly, alongside three unwanted terms, and to identify which is the virtual image and which the real one. We assume linear amplitude recording, read out with the same reference $U_r$ , distribute the product $U_r\,t$ , and interpret each term. The payoff, equation (3), comes from squaring a sum and multiplying it back by one of its parts.

Step 1. Record the superposition on the square-law medium, $I=|U_o+U_r|^2=|U_o|^2+|U_r|^2+U_oU_r^{*}+U_o^{*}U_r$ . Why: the detector responds to intensity, not field, so the object's phase can only enter through the interference cross terms — everything downstream is a consequence of this squaring.

Step 2. Develop to a transmittance linear in exposure, $t=t_b+\beta\tau\,(U_oU_r^{*}+U_o^{*}U_r)$ , with the constant $t_b$ soaking up the self-intensities. Why: only the modulation about the bias carries object information, and linearity keeps it an undistorted copy of the cross terms.

Step 3. Illuminate with the reference and multiply, $U_{\text{out}}=U_r\,t=t_bU_r+\beta\tau\,U_rU_oU_r^{*}+\beta\tau\,U_rU_o^{*}U_r$ . Why: in the thin-hologram limit the plate just imposes a local amplitude factor, so reconstruction is literally reference times transmittance.

Step 4. Simplify with $U_rU_r^{*}=|U_r|^2$ to get $U_{\text{out}}=t_bU_r+\beta\tau|U_r|^2U_o+\beta\tau\,U_r^{2}U_o^{*}$ . Why: the middle term is a scalar multiple of $U_o$ — the object wave rebuilt, diverging as if from the vanished object, a virtual image; the last carries $U_o^{*}$ , a converging conjugate forming a real, pseudoscopic image.

Step 5. Put in a plane-wave reference $U_r=Ae^{ikx\sin\theta_r}$ , so $U_r^{2}=A^2e^{i2kx\sin\theta_r}$ . Why: the conjugate term now carries twice the carrier and the object term the reference's own tilt, so the two images and the bias leave the plate at three different angles — the Leith–Upatnieks separation. *Verified: four-term structure and off-axis angular separation match Goodman, Introduction to Fourier Optics, Ch. 9, and Hecht, Optics, Ch. 13.*

Go deeper · Derivation B · 1 minAngular separation condition for off-axis holography

Derivation BAngular separation condition for off-axis holography

The goal is to find how far the reference must be tilted so that the zero order, virtual image, and real image occupy non-overlapping bands of the angular spectrum. We model the object wave as a band of half-width $B=\mathrm{NA}/\lambda$ , give the reference a carrier $\nu_c=\sin\theta_r/\lambda$ , track where each recorded term lands, and demand that the shifted object bands clear the widened bias band. The result is a clean rule of thumb, $\nu_c\ge 3B$ , plus a resolution demand on the medium.

Step 1. Locate the bias band. The $|U_o|^2$ term is the autocorrelation of the object spectrum, so it occupies $|\nu_x|\le 2B$ — twice the object's own bandwidth, centred on zero. Why: autocorrelation convolves the spectrum with itself, doubling its width, and this is the band the shifted images must avoid.

Step 2. Locate the image bands. Each cross term is the object spectrum rigidly shifted to $\pm\nu_c$ , so it occupies $\nu_c-B$ to $\nu_c+B$ (and its mirror). Why: multiplying by the tilted reference $e^{i2\pi\nu_c x}$ is a pure frequency translation, carrying the object's information up to the carrier untouched.

Step 3. Impose non-overlap: the lower edge of the shifted band must sit above the bias band, $\nu_c-B\ge 2B$ , hence $\nu_c\ge 3B$ , or $\sin\theta_r\ge 3\,\mathrm{NA}_{\text{obj}}$ . Why: below this tilt the halo swamps the image; at or above it the three orders separate cleanly.

Step 4. Read off the medium's job: it must resolve up to $\nu_{\max}=\nu_c+B\ge 4B$ . For $\mathrm{NA}=0.1$ at $\lambda=0.5\ \mu\text{m}$ , $B=200\ \text{mm}^{-1}$ and $\nu_{\max}\approx 800\ \text{mm}^{-1}$ . Why: ordinary film resolves only ~100 lines/mm, so off-axis holography forces ultra-fine-grain emulsions resolving several thousand. *Verified: bandwidth accounting and the factor-of-three carrier requirement follow the standard treatment in Goodman, Ch. 9.*

Go deeper · Derivation C · 1 minCoherence-length constraint on the path mismatch

Derivation CCoherence-length constraint on the path mismatch

The goal is to quantify how monochromatic the source must be, by relating fringe visibility to the object–reference path difference and the source linewidth. We identify fringe contrast with the modulus of the temporal coherence function, convert coherence time to a length, recast it in wavelength linewidth, and put in numbers for a helium–neon laser and for white light. The result, $|\Delta L|\ll\ell_c=\lambda^2/\Delta\lambda$ , is the single inequality that made lasers indispensable to holography.

Step 1. Write fringe visibility as the modulus of the complex degree of coherence at delay $\tau=\Delta L/c$ , $\mathcal{V}=|\gamma(\tau)|$ , which for any real line decays over $\tau_c\sim 1/\Delta\nu$ . Why: two beams with a path mismatch sample the field at times separated by $\tau$ , and their fringes wash out once $\tau$ exceeds the phase-steady time.

Step 2. Convert coherence time to length, $\ell_c=c\,\tau_c=c/\Delta\nu$ . Why: a length is the natural yardstick for a path mismatch — the distance over which interference survives.

Step 3. Recast in wavelength using $\Delta\nu=c\,\Delta\lambda/\lambda^2$ , giving $\ell_c=\lambda^2/\Delta\lambda$ . Why: sources are specified by their wavelength linewidth $\Delta\lambda$ , so this is the form an experimenter evaluates.

Step 4. Put in numbers. A multimode HeNe with $\Delta\lambda\approx 2\times10^{-3}\ \text{nm}$ at $633\ \text{nm}$ gives $\ell_c\approx 20\ \text{cm}$ , ample for a tabletop scene; white light ( $\Delta\lambda\approx 300\ \text{nm}$ ) gives $\ell_c\approx 1\ \mu\text{m}$ , hopeless. Why: the two arms differ in length by centimetres, so only a laser keeps the fringes visible — why the field lay dormant from 1948 to 1960. *Verified: standard coherence relations, with the HeNe coherence length reproduced as a textbook order of magnitude (Loudon; Demtröder).*

Limiting cases

Holographic behaviour is set by three knobs: how thick the emulsion is relative to the fringe spacing, how far off-axis the reference sits, and how coherent the source is.

Case	Result	Meaning
Thin (plane) hologram	$t_{\text{em}}\ll d^2/\lambda$	Flat grating, many diffraction orders, no Bragg selectivity (Raman–Nath); efficiency capped at 6.25 percent (amplitude) or 33.9 percent (phase).
Thick (volume) hologram	$t_{\text{em}}\gg d^2/\lambda$	Bragg regime: one strong order, sharp wavelength and angular selectivity $\Delta\theta\sim d/t_{\text{em}}$ , efficiency approaching 100 percent; basis of reflection holograms, HOEs, filters.
In-line (Gabor)	$\theta_r\approx 0$	Reference collinear with the transmitted beam; virtual and real images overlap — the twin-image problem — so it works only for sparse, transparent objects.
Off-axis (Leith–Upatnieks)	$\sin\theta_r\ge 3\,\mathrm{NA}_{\text{obj}}$	Reference tilted enough that the four terms fan to distinct angles; the medium must resolve $\nu_c=\sin\theta_r/\lambda$ .
Coherence limit	$\|\Delta L\|\ll\ell_c=\lambda^2/\Delta\lambda$	Path mismatch far below the coherence length; sets fringe visibility and hence diffraction efficiency.

Worked examples

1 · Off-axis recording geometry (confidence builder)

A HeNe laser ( $\lambda=633\ \text{nm}$ ) records a hologram with the object and reference beams arriving symmetrically at $\pm15^\circ$ about the plate normal, a total crossing angle $\theta=30^\circ$ . Find the fringe spacing, the carrier spatial frequency, and say what recording medium this demands.

Step 1. Fringe spacing from equation (5): $d=\lambda/[2\sin(\theta/2)]=633\ \text{nm}/[2\sin15^\circ]$ . With $\sin15^\circ=0.2588$ the denominator is $0.5176$ , so $d=633\ \text{nm}/0.5176=\mathbf{1.22\ \mu m}$ , about twice the wavelength.

Step 2. The carrier spatial frequency is the reciprocal of the period, $\nu_c=1/d=8.18\times10^{5}\ \text{m}^{-1}=\mathbf{818\ lines/mm}$ — equivalently $\nu_c=2\sin15^\circ/\lambda$ .

Step 3. Interpret. Ordinary film resolves roughly 100 lines/mm and would smear these fringes into grey; holographic emulsions (Agfa 8E75, Slavich PFG-01) resolve 3000 to 5000 lines/mm and capture 818 easily. A reflection hologram with counter-propagating beams ( $\theta\to180^\circ$ ) would demand $d\approx\lambda/2\approx 0.3\ \mu\text{m}$ , near 6000 lines/mm. d = 1.22 µm, νc = 818 lines/mm — a job for a high-resolution holographic emulsion, not ordinary film.

2 · Coherence budget for a real scene

The same multimode HeNe has a linewidth $\Delta\lambda\approx 2\times10^{-3}\ \text{nm}$ . Over what depth of scene will the fringes stay sharp, and what happens if you swap in a filtered mercury lamp with $\Delta\lambda\approx 1\ \text{nm}$ ?

Step 1. Coherence length from equation (7): $\ell_c=\lambda^2/\Delta\lambda=(633\times10^{-9})^2/(2\times10^{-12})\ \text{m}=\mathbf{0.20\ m}$ . The path difference must stay well inside this.

Step 2. A depth difference $\Delta z$ in the scene contributes a path mismatch of order $2\Delta z$ , so keeping $2\Delta z\ll\ell_c$ limits the usable scene depth to roughly $\ell_c/2\approx\mathbf{10\ cm}$ — fine for a tabletop object, marginal for a deep scene.

Step 3. The mercury lamp: $\ell_c=(633\times10^{-9})^2/(1\times10^{-9})=\mathbf{0.40\ \mu m}$ , a fraction of a wavelength. No practical arm match holds the fringes, so transmission holography is impossible with it. HeNe gives ~10 cm of usable depth; the lamp gives ~0.4 µm — this is why holography needs a laser.

3 · Diffraction efficiency across the regimes

A hologram is read with 1.0 mW of reference power in the ideal first order. Compare the reconstructed image power for a thin amplitude hologram, a thin phase hologram, and a thick phase (volume) hologram.

Step 1. Thin amplitude hologram: the maximum first-order efficiency of a sinusoidal amplitude grating is $\eta=1/16=6.25\%$ (half the light absorbed by the bias, only $1/16$ of the rest into each first order). Image power $=0.0625\times1.0\ \text{mW}=\mathbf{62.5\ \mu W}$ .

Step 2. Thin phase hologram: bleaching turns absorption fringes into phase fringes; the first-order power is $J_1^2$ , maximized at argument $1.84$ where $J_1=0.582$ , so $\eta=33.9\%$ . Image power $=\mathbf{339\ \mu W}$ , over five times the amplitude case.

Step 3. Thick phase (volume) hologram: in the Bragg regime coupled-wave theory couples all the light into the single phase-matched order, $\eta\to100\%$ , so image power approaches $\mathbf{1.0\ mW}$ . 62.5 µW → 339 µW → ~1.0 mW: this is why bright display holograms and holographic optical elements are thick and phase-type, not thin absorption gratings.

Physics of the portrait

Holography is a rare case where one man wrote down the whole idea more than a decade before the technology existed to show it off, then lived to collect the Nobel Prize once it did.

Origins · 1 minDennis Gabor and how this was discovered

Dennis Gabor

1900–1979 · Hungarian-British electrical engineer and physicist, inventor of holography

Dennis Gabor conceived wavefront reconstruction in 1947 while trying to improve the electron microscope, whose lenses were too aberrated to reach atomic resolution. His insight, published in Nature in 1948 as "A New Microscopic Principle", was to split imaging into two stages — record the interference of the scattered wave with a coherent background, then reconstruct optically — sidestepping the bad lens. He coined hologram, from the Greek holos for whole. But Gabor had no coherent source: mercury-arc light gave only millimetres of coherence, and his in-line geometry piled the twin images atop each other, so reconstructions were dim. The idea languished until Maiman's laser (1960) supplied coherence in abundance, and until Emmett Leith and Juris Upatnieks (1962), drawing on side-looking radar work, introduced the off-axis reference that fanned the twin images apart and gave the first crisp three-dimensional holograms. In parallel Yuri Denisyuk (1962) invented the white-light reflection hologram, and Herwig Kogelnik (1969) built the coupled-wave theory predicting the near-unit efficiency of thick holograms. Gabor received the 1971 Nobel Prize in Physics for the holographic method — an idea that had waited twenty-three years for its light source.

History

1948

— Gabor publishes holography in Nature as a new microscopic principle

1960

— Maiman demonstrates the ruby laser, giving holography a coherent source

1962

— Leith and Upatnieks introduce off-axis holography and defeat the twin-image problem

1969

— Kogelnik publishes coupled-wave theory of thick holograms

1971

— Gabor is awarded the Nobel Prize in Physics for holography

Going further

Texts. Hecht, Optics, 5th ed., Ch. 13 · Goodman, Introduction to Fourier Optics, 4th ed., Ch. 9 · Saleh and Teich, Fundamentals of Photonics, 3rd ed. · Collier, Burckhardt and Lin, Optical Holography (1971) · Hariharan, Optical Holography, 2nd ed. · Kogelnik, "Coupled Wave Theory for Thick Hologram Gratings", Bell System Technical Journal (1969) · Loudon and Demtröder for the coherence relations.

Provenance. Equations and numbers cross-checked against Hecht Ch. 13 and Goodman Ch. 9. A verifier should confirm: (i) the four-term expansion of $|U_o+U_r|^2$ and reconstruction $U_{\text{out}}=U_r\,t$ ; (ii) thin-hologram efficiency limits 6.25 percent (amplitude) and 33.9 percent (phase, the $J_1^2$ maximum at argument 1.84) and the Bragg near-100 percent limit (Kogelnik); (iii) $d=\lambda/[2\sin(\theta/2)]=1.22\ \mu\text{m}$ and $\nu_c=818\ \text{lines/mm}$ at $\theta=30^\circ$ , $\lambda=633\ \text{nm}$ ; (iv) $\ell_c=\lambda^2/\Delta\lambda$ giving 0.20 m for the HeNe and 0.40 µm for the $\Delta\lambda=1\ \text{nm}$ lamp; (v) the carrier condition $\nu_c\ge 3B$ , with $\nu_{\max}\approx 800\ \text{mm}^{-1}$ at $\mathrm{NA}=0.1$ , $\lambda=0.5\ \mu\text{m}$ . Dates follow Gabor's 1948 Nature paper, Leith–Upatnieks and Denisyuk 1962, Kogelnik 1969, and the 1971 Nobel Prize.

Prerequisites. Interference and Coherence · Scalar Diffraction Theory · Fourier Optics · Complex Amplitude of Optical Fields. Leads to. Phase Conjugation and Nonlinear Optics · Volume Holograms and Bragg Diffraction · Digital Holography and Coherent Imaging · Integrated Photonic Circuits.

Ask the tutor: Ask me to walk through why reading a hologram with a phase-conjugate reference plays the object wave backwards and cancels the aberrations of whatever it passed through.

Why can a photograph never show depth but a hologram can?

A photograph records only the intensity $|U_o|^2$ , and intensity is blind to phase — yet phase is what encodes the direction each ray arrived from, and hence depth. Discard it and you keep a flat map of brightness with no way to recover the three-dimensional arrangement that produced it. A hologram interferes the object wave with a reference so that the phase $\phi_o$ is written into the fringe pattern through the cross term $U_oU_r^{*}$ . On readout, diffracting the reference off those fringes regenerates the full complex wavefront, direction included, so your eyes receive light as if the object were still there, with parallax and the need to refocus near and far.

If I cut a hologram in half, why do I still see the whole object?

Because the recording is completely delocalized. Every object point scatters light across the entire plate, so each patch captures the whole scene from its own vantage point — much as you see an entire room through any part of a window. Break the plate and each fragment still holds a full, if partial-view, record. What a smaller piece loses is not content but capability: the reduced aperture diffracts less sharply, so resolution drops, and the smaller window narrows the parallax, so you can no longer look "around" the object as far. A cut photograph, by contrast, simply loses the part of the image that was there.

What is the twin-image problem, and how does off-axis recording cure it?

In Gabor's in-line geometry the reference travels along the same axis as the transmitted object beam, so on reconstruction the virtual image ( $\propto U_o$ ) and the conjugate real image ( $\propto U_o^{*}$ ) emerge collinearly, and a viewer sees one image contaminated by a blurred version of the other — the twin images. Leith and Upatnieks fixed this by tilting the reference by $\theta_r$ : the tilt stamps a spatial carrier $\nu_c=\sin\theta_r/\lambda$ onto the cross terms, so on readout the two images and the zero order diffract off at three distinct angles. Provided $\sin\theta_r\ge 3\,\mathrm{NA}_{\text{obj}}$ , the three bands do not overlap and the virtual image is seen cleanly on its own.

What actually distinguishes a thin from a thick (volume) hologram?

The criterion is the emulsion thickness $t_{\text{em}}$ against $d^2/\lambda$ . When $t_{\text{em}}\ll d^2/\lambda$ the fringes act as a flat 2D grating (Raman–Nath regime): many orders at once, no angular selectivity, efficiency capped at 6.25 percent (amplitude) or 33.9 percent (phase). When $t_{\text{em}}\gg d^2/\lambda$ the fringe planes form a 3D stack (Bragg regime): light reinforces only at the Bragg angle $2d\sin\theta_B=\lambda$ , so it diffracts into essentially one order, becomes sharply wavelength- and angle-selective, and can reach nearly 100 percent. That selectivity is what lets one thick emulsion store thousands of multiplexed pages and makes bright, white-light-viewable reflection holograms possible.

Can I reconstruct a hologram with a different wavelength or with white light?

Yes, with consequences. Reading a thin transmission hologram at $\lambda_2\ne\lambda_1$ rescales the image: lateral $M_{\text{lat}}=[1-z_o/z_r-\mu z_o/z_c]^{-1}$ with $\mu=\lambda_2/\lambda_1$ , longitudinal $M_{\text{long}}=M_{\text{lat}}^2/\mu$ , so a colour change stretches and distorts depth; under white light every wavelength reconstructs at a different angle and scale, smearing into a rainbow. Thick holograms differ: Bragg selectivity admits only the wavelength and angle satisfying $2d\sin\theta_B=\lambda$ , so a reflection volume hologram in white light returns just the one colour it was recorded to diffract — exactly why Denisyuk holograms are viewable under an ordinary lamp.

How is digital holography different from making a hologram on film?

The recording physics is identical — object and reference still interfere on a detector — but the detector is now a CCD or CMOS array and reconstruction is done in software rather than by shining light back through a plate. The pixel array is numerically propagated with a Fresnel or angular-spectrum integral, yielding the complex field from which amplitude and quantitative phase are read out. The trade-off is resolution: sensor pixels are 2 to 6 µm, far coarser than film grain, so the interference angle must be small (limiting numerical aperture and field of view) to keep the carrier fringes below Nyquist. In return you gain instant reconstruction, digital refocusing, phase unwrapping for nanometre metrology, and numerical removal of the twin image and zero order.