Make your paper interactiveCreate

Text from the interactive explanation

Anti-counterfeiting white light printed image multiplexing by fast nanosecond laser processing

Nicolas Dalloz, Van Doan Le, Mathieu Hébert, Bálint Éles, Manuel Figueroa, Christophe Hubert, Hongfeng Ma, Nipun Sharma, Francis Vocanson, Stéphane Ayala, Nathalie Destouches

Original paper

Dalloz et al. · Advanced Materials 2021 · an interactive reading

Could a printed ID photo or a patch on a banknote hide several pictures that anyone can switch between simply by changing how they hold it to the light — no polarizer, no lens, no UV lamp? This paper shows how, with a thin silver-loaded film on glass that a fast industrial laser rewrites pixel by pixel.

What the eye sees

Between modes — the pictures fade ψ = +18°

Drag the lamp around its orbit. 0° = lamp beside the eye; ±180° = lamp behind the slide.

simulated viewPixel colours approximate the measured colour tree of the paper's three-image sample (Fig. 6c); lamp positions follow the measurement geometries in the Methods; transitions between them are illustrative. The three hidden pictures here are stand-ins — the real sample shows photographs.

Why would anyone want a print that changes with the light?

Why hide pictures in a print?

Banknotes, passports and luxury packaging carry security features that are checked at three levels. The paper starts from a simple observation: the features that matter most in daily life are the ones nobody needs equipment to check.

Level 1

Overt

Seen directly by eye — colour-shifting inks and coatings, holograms.

Anyone can check it: no tool at all.

Level 2

Covert

Hidden until you use a revealing tool.

Needs a lens, a polarizer or a UV lamp.

Level 3

Forensic

Only an expert with laboratory equipment can confirm it.

Needs lab instruments.

Overt features are the first line of defence and often the only one actually used. But the familiar ones — holograms, micro-lens arrays, diffractive foils — are well known, and they can be counterfeited. The authors propose a richer overt feature: image multiplexing , “encoding on a single metasurface the complete information of multiple images that can be displayed successively by switching the observation conditions.” Hold the print one way and you see one picture; tilt it to the light differently and another appears.

Metasurface — a surface patterned at scales smaller than the wavelength of light, so that it acquires optical properties ordinary inks do not have.

Here every printed square is a random plasmonic metasurface : a film filled with silver nanoparticles whose size, density and orientation are controlled statistically by the laser — not placed one by one. “Plasmonic” refers to the way electrons in metal particles respond collectively to light (more on that in the physics section).

Before this work

Multiplexing existed — but not as an overt feature

Printed multiplexing under white light needed polarizers to read it — a tool, so it stayed covert.

The laser route to such prints used femtosecond lasers , costly and slow: about 10 mm/s scan speed.

This paper

Naked-eye pictures, written fast

Two and three pictures that switch under ordinary white light, no tool needed .

An industrial nanosecond laser with scan speeds up to 2000 mm/s — about 50× faster on average ; each demonstrator took 5–10 minutes.

Customisable: the pictures can be, for example, an ID-card holder's own portrait.

Uses the authors envisage:

If one printed square can change colour with the light, how many different looks can it have?

Seven ways to look at one square

The team printed small squares, each with a single laser recipe, and photographed them under ordinary (incoherent) white light in seven observation modes — seven arrangements of lamp, sample and camera. Five need nothing but your eyes; the last two need a pair of polarizers.

Compare two squares

Tap any two squares in the sheet. The bars show their colour difference ΔE in each mode (CIE 1976 distance in L*a*b* — about 2 is barely visible, above 10 is obvious).

Tap two squares in the sheet to compare them.

Challenge 1

Find two squares that look alike in unpolarized transmission but not in diffraction .

Not solved yet — compare pairs above.

Challenge 2

Find a square whose colour flips between polarized 1 and polarized 2 (a dichroic square).

Not solved yet — tap squares in the sheet.

measured coloursSquare colours approximated from Fig. 2 of the paper (printed figure; displayed sRGB). Each square is 0.864 mm wide, printed with one laser recipe (laser polarization along the squares' vertical side). Scattering and diffraction are very dark and are shown 2.5× brighter; all ΔE values use the unboosted colours.

What makes those twelve squares so different from one another?

What the laser writes into each square

The starting material is a microscope glass slide coated with a mesoporous (sponge-like) amorphous titanium-dioxide film, soaked with a silver salt. UV light turns the silver ions into tiny silver nanoparticles. Then a laser reshapes them.

Five laser parameters were varied: fluence (energy per area), repetition rate , the orientation of the laser's linear polarization , scan speed , and the interline spacing or exposure time (plus the number of passes in one demonstrator). 72 µm pixels are filled with overlapping laser lines parallel to the pixel diagonal; 20 µm pixels are written with a static spot. Pick one of the six recipes the paper examined under the electron and atomic-force microscopes (Fig. 3, repetition rate 300 kHz):

Laser recipe · fluence (mJ/cm²) × speed

15 and 125 mm/s correspond to about 1080 and 129.6 effective pulses per spot.

illustrativeSketch inspired by the SEM images in Fig. 3, not data. Scale bar 500 nm.

Measured colours of square B4

Why does one nanostructure give a different colour in each mode?

Why one square has many colours

Two mechanisms do most of the switching. A surface grating sends a narrow band of colours into one direction only — and can be “turned off” by rotating it. Rows of silver particles absorb light differently depending on its polarization. Try each one.

A grating you can switch off by turning it

A ridge pattern with period Λ splits white light. In the plane of incidence the −1st diffraction order leaves at an angle given by sin θ₋₁ = sin θᵢ − λ/Λ. With Λ ≈ 300 nm only short, blue wavelengths have a propagating −1st order, and it is thrown back toward the light source — so a camera on that side sees blue. Rotate the grating in its own plane and the diffracted light leaves the plane of incidence (conical diffraction): the camera catches nothing and the pixel looks black. The paper uses exactly this: same fluence, laser polarization rotated by 60°, and the grating “diffracts out of the observation plane”.

computedGrating equation for an ideal grating; colours from the CIE 1931 colour-matching functions (analytic fit). The second panel is the “direction map”: the disc holds every direction light can leave in; points outside it are evanescent (they do not propagate). The dashed blue circle is the path the 450 nm order follows as the grating turns in its plane.

Camera at θc = −34.6° (aperture ±4°)

Camera sees: …

ρ is the angle between the grating vector and the plane of incidence (ρ = 0: lines perpendicular to that plane).

Dichroism: particles in a row talk to each other

Each silver particle has a localized surface plasmon resonance (LSPR) : its free electrons oscillate together and absorb or scatter a band of colours. An oscillating dipole radiates most strongly perpendicular to its axis. In the paper's simplified, ordered model — Ag particles 80 nm across, aligned with period 297 nm in a 90 nm film of index 2 on glass — light polarized along the grating lines ( TE ) makes the dipoles radiate along the row toward their neighbours. Their waves add up into a collective lattice mode near the Rayleigh anomaly (a diffraction order grazing along the surface), which hybridizes with the broad LSPR. For TM (field across the lines, along the row) the dipoles barely radiate toward each other.

illustration — not an electromagnetic simulationTop view of one row of the model, looking down on the film: the light comes toward the film perpendicular to the screen, and each dashed line marks one particle line. The ripples are a cartoon of waves from point dipoles (no film interfaces, no exact fields), meant only to show the in-step idea. The authors' rigorous electromagnetic simulations (their own multiple-scattering code, Fig. 4e–l) are not reproduced here. Back-of-envelope estimate: a diffraction order grazing inside the film appears near λ ≈ n·Λ ≈ 2 × 297 ≈ 594 nm.

Measured transmission (Fig. 4e)

measuredTransmission traced approximately from the measured curves of Fig. 4e. TM: one broad dip near 468 nm; TE: dips at 468 and 658 nm. Where to look for the lattice mode: the green band (546–600 nm) is where the paper's simulations show the lattice mode excited for TE, attenuating TE extinction relative to TM — and indeed TE transmits more than TM there. The dotted line marks the grazing-order estimate n·Λ ≈ 594 nm. The extra TE dip at 658 nm (▼) is the particles' dipole resonance, narrowed and strengthened by coupling to that lattice mode; TM shows no such dip. Swatches: white light × T(λ) through the CIE colour-matching functions; T(λ) = cos²φ·TTE + sin²φ·TTM.

If colours can switch independently, how do we turn them into pictures?

The trick: a logical colour tree

To hide M pictures with K colours each, every pixel must show picture 1's colour in mode 1, picture 2's colour in mode 2, and so on. So the printer needs one metasurface for every combination of colours — a logical colour tree with KM branches. With two colours: 2 pictures need 4 metasurfaces, 3 pictures need 8. The bit depth of the multiplexed image is the sum of the pictures' bit depths.

illustrativeAfter Fig. 5 of the paper: a diagonal split and a disc, then a third picture (a triangle).

How many metasurfaces?

2² = 4 metasurfaces

The paper's colour picture (results card d) used 7 colours in backside reflection and 2 in diffraction, yet needed only 8 metasurfaces, not 14: the logos sit only where one colour (MS7) appears, and MS8 duplicates that colour with a rotated grating.

Puzzle: build a two-picture tree from the 12 real squares

Pick a square in the tray, then tap a slot. Slot xy means: colour x in the first mode, colour y in the second. The preview paints both pictures with the squares you chose — wrong choices show the other picture's ghost (crosstalk) or no contrast.

Tray · top half: first mode, bottom half: second mode

Fill the four slots to see if your tree is clean.

A tree is clean when, in both modes, the two colour groups differ by ΔE ≥ 15 (signal) and squares that must match differ by at most 0.6 × that signal (ghost).

With thousands of measured squares, how do you find a set that forms a clean tree?

Searching 9,072 squares for a tree

The paper's colour search algorithm (Section 2.4) needs no physical model — only measured colours. Here it runs step by step on a small synthetic library.

Measure each square's colour in the chosen modes and express it in CIE 1976 L*a*b* , a colour space where distance (ΔE) roughly matches perceived difference.

In each mode, run maximum-distance (complete-linkage) hierarchical clustering with a cut-off: every pair inside a cluster is closer than the cut-off.

For each pair of mode-1 clusters Ci, Cj, find mode-2 clusters holding members of both. Two such clusters Ck, Cl give a tree: Ci∩Ck, Ci∩Cl, Cj∩Ck, Cj∩Cl.

Keep the solutions with the largest colour and lightness differences in each mode; the user picks among them.

Best tree · top: mode 1, bottom: mode 2

Larger cut-offs → more solutions, but members of one cluster differ more, so ghosts appear. Smaller cut-offs → cleaner trees, but fewer of them.

syntheticSynthetic library of 360 squares for illustration, drawn from a toy model of laser recipes — the real database holds 9,072 measured squares. The method works for any technology whose colours depend on the observation mode, and it can suggest non-obvious combinations, which adds to security.

Each pixel shows one of only two colours — so where do faces and shading come from?

From two colours to a portrait

Shading comes from halftoning , or dithering: “printing two or more colours side by side to render the average colour”. Smaller pixels make the trick invisible.

How to read the picture. Left: the two-colour image, hugely enlarged so you can see its pixels. Top right: a zoom on 18 × 18 pixels; the orange square is a single pixel — one square of laser-written metasurface, 72 µm or 20 µm on a side (the paper prints square pixels of these two sizes). Bottom right: the whole printed pictures at their reported sizes, drawn to one common scale with a 10 mm bar — 13.4 × 15.9 mm with 72 µm pixels (butterfly sample) and 5.4 × 8.0 mm with 20 µm pixels (portrait sample). They are rectangles simply because the printed images are. At normal viewing distance such tiny squares are at or below what the eye can separate, so neighbouring pixels blend into an average colour.

Two colours approximated from the backside-reflection tree of the paper's portrait sample (Fig. 6b); the face is a stand-in drawing, not the authors' portrait.

Writing it with light

Each 72 µm pixel is filled with overlapping laser lines parallel to its diagonal. The recipe changes from pixel to pixel — here the laser polarization (arrow) is turned by 60° for two of the four metasurfaces, as in the transmission + diffraction demonstrator. The two plaques are written at a 50 : 1 rate: ≈50× faster on average; each demonstrator took 5–10 min.

illustrativeSpot path and timing are schematic; colours approximated from the transmission tree of Fig. 6a.

So what did the team actually print?

What the team printed

Five demonstrators, each built from a logical colour tree of measured metasurfaces. Each card shows the measured tree colours and a stand-in picture you can switch between modes.

measured treesTree colours approximated from Figs. 6 and 7 of the paper. Stand-in pictures — the samples show photographs. Smartphone and camera recordings of each sample are in Supporting Movies 1–5 at the publisher.

Card e · four pictures with polarizers

Three portraits and a blank in one print

Unpolarized transmission shows portrait 1. Between two polarizers 85° apart, the sample shows portrait 2 at azimuth α = 45° and portrait 3 at α = 135°. In frontside reflection all eight metasurfaces look alike — a blank. All eight recipes are dichroic (self-organized particle gratings parallel to the laser polarization); two scan speeds (50 and 200 mm/s) give the two unpolarized colours, and four polarization orientations (30°, 60°, 120°, 150°) give the four combinations in the polarized modes.

measured treeColours approximated from Fig. 7 at the recorded angles α = 45° and 135°; between them colours are blended for illustration. Stand-in portraits replace the photographs.

So — can one print hold several pictures that anyone can switch?

Back to the bench

Yes. Under ordinary white light, with nothing but a change of lamp position, this print shows two or three different pictures — an overt security feature that is hard to replicate, not least because no physical model yet predicts which laser recipe gives which set of colours.

The authors' outlook: industrial security printing read with a smartphone or checked by eye, with customised overt features such as the portrait of an ID card's owner.

Source

Nicolas Dalloz, Van Doan Le, Mathieu Hébert, Bálint Éles, Manuel A. Flores Figueroa, Christophe Hubert, Hongfeng Ma, Nipun Sharma, Francis Vocanson, Stéphane Ayala, Nathalie Destouches. “Anti-counterfeiting white light printed image multiplexing by fast nanosecond laser processing.” Advanced Materials (2021), article 2104054.

DOI: 10.1002/adma.202104054 · Read from the authors' manuscript version of the published article deposited on HAL: ujm-03409454v1. Supporting Movies 1–5 accompany the article at the publisher.

All visuals on this page are original drawings and simulations made with code; colours marked “approximated” were read from the paper's figures; stand-in pictures replace the paper's photographs.

Topics in this iPaper

Physics
Electromagnetic simulations are proposed to explain one of the mechanisms that are at the origin of the singular optical properties of such metasurfaces.

Introduction, printed page 3

Engineering
This high speed and flexible laser process, based on energy efficient nanoparticle reshaping and self-organization, produces centimeter-scale customized tamper-proof images at low cost

Abstract, printed page 1

plasmonic metasurfaces
Passive plasmonic metasurfaces enable image multiplexing by displaying different images when altering the conditions of observation.

Abstract, printed page 1

image multiplexing
the implementation of image multiplexing is extended, thanks to a color search algorithm, to various observation modes compatible with naked-eye observation under incoherent white light

Abstract, printed page 1

laser processing
processed with a focused nanosecond laser beam at 532 nm wavelength

Section 2.1, printed page 3

anti counterfeiting
can implement printed image multiplexing for high-end anti-counterfeiting applications

Introduction, printed page 2

localized surface plasmon resonance
hybridization between the localized surface plasmon resonance of metallic nanoparticles and a lattice resonance

Abstract, printed page 1

hierarchical clustering
a maximum distance hierarchical clustering is performed by using the chosen perceptual color distance

Section 2.4, printed page 11

Anti-counterfeiting white light printed image multiplexing by fast nanosecond laser processing — Interactive explanation | iPaper.ai