The Thermoptic Camouflage Bench
A cloak that routes light round a body must deliver it late, and a late wave matches free space at one wavelength only: picometres of band for a body, against 370 nm of daylight. A cloak that repaints the front is right for one eye, wears the last frame while it moves, and gives water a surface to land on. And whichever it is, the skin may not be warmer than the night, so a hundred watts go inward, into the core or the ice, for as long as they last. Put the suit on. Calibrate it to your eye. Step aside. Switch the lamp. Switch the camera.
Open the interactive ▸ What you're looking at
The Suit is an alley at night in three dimensions: a lit façade behind, a wet street, a lamp, and a human-sized figure. Three suits go on it. Off is the body. Display is Tachi's optical camouflage: the alley as seen from one calibrated eye, rendered without the body, is projected onto the body, so from that eye it vanishes and from any other it carries a seam; a latency dial slips the image behind the figure when it walks, a pitch dial quantises it. Passive is the metamaterial cloak: the body's opacity is one minus the share of the lamp's light that fits inside the cloak's band, so it vanishes under a laser and stands under a sodium lamp. Four lights, five cameras, a button to calibrate the display to wherever you stand, and rain that splashes at the model's rate. A thermal toggle re-renders the same alley as a white-hot imager sees it, with the suit's outer skin in hexagonal tiles at the night's temperature, the breath puffing every five seconds and the footprints cooling on the model's curve; a window dial does what an imager's operator does.
The Band is the passive bound on a log-log chart: the bandwidth a cloak can hold against the width light must go round, with the sources as lines, daylight to a single-mode laser, and the ladder of what each can hide, a virus to a building; a second faint line is the same bound at 10 µm. Beside it a ledger: the detour, the delay, the slip, the band, the share of each light that fits, and the sum rule as the other half of the account. The Seam is the display from above: the calibrated eye, a second eye, the body, the wall, and the jump at the wall drawn where the two lines of sight land; below it the lag at four speeds and the pixels the suit would need; on the right the water, Marshall & Palmer's drops on the body's top and the wet film's reflectance against angle.
The Heat is where 100 W go when the skin may not glow: the body, the insulation, the tiles and the sink as boxes with the watts on the arrows; a budget clock at three paces, stored only and with the ice; and the imager's noise against the bare skin, the tiles' error, the breath, a bare sole, and ADAPTIV for scale. The File runs thirteen cards, the film, the two routes, the detour, the band, the ladder, the sum rule, the display, the seam, the lag, the water, the heat, the ice, the leaks, then two verdicts at equal size.
Why it's here
This site's signed review of Ghost in the Shell (Oshii, 1995) has a sentence in it: "The thermoptic camouflage, which is what the Major wears in the opening drop, is also the film getting its physics right by instinct. Something that has to match its background moves, and the moment it moves the match slips, so it shimmers, and rain finds it at once." This bench is the bill for that sentence. The film gives the suit behaviour and no numbers: it shows the background where she stands, and a man three metres away does not see her; it ripples when she moves and settles when she stops; in the canal fight the water thrown up and running down shows where she is. The name says it also hides her from thermal imagers; the film never puts one on her. The three behaviours are three pieces of physics: a display that must match its background, a latency that lets the match slip, and a surface for water to land on.
It is here because "invisible" is a sentence that can be computed, in two halves. The optical half has two routes. A passive cloak, what "cloak" usually means in physics, is a shell that routes light round the body and delivers it on the far side as if nothing were there, with no power and no clock; it fails in the frequency domain: the detour is at least (π/2 − 1)·D longer than the straight path, the wave packet cannot go round faster than light in a passive medium, so the phase matches at one frequency only, and a 0.45 m body at 550 nm holds a band of 0.12 pm, a laser line, against 370 nm of daylight. That is the wave-packet argument of Craeye and Bhattacharya (2012) and the causality bound of Monticone and Alù (2016); turned round, the widest thing daylight can passively hide is a virus. An active cloak is what the film draws, a display, which Susumu Tachi's group showed at the University of Tokyo in 2003 and credited to this film; it fails in the viewpoint domain: right for one eye, and for a second eye a metre aside with the wall 1.5 m behind, the background jumps 19 cm; one frame of latency at a walk lags 23 mm; rain at 10 mm/h hits the body 208 times a second. The thermal half is thermodynamics: 100 W with nowhere to go, +1.5 K an hour if stored, 2 kg of ice for 2 h 42 min at rest and 27 minutes running, with breath and bare soles leaking regardless. It is the optical and infrared sister of the Stealth Bench: that one prices why radar does not see, this one why eyes and ATFLIR-class cameras do, and the Tic Tac and the Gimbal were both seen in the infrared. It also sits on the fiction-pricing line after Project Hail Mary: the film is quoted as premise, the device as drawn is priced, and the story is not reviewed; the review is next door, linked, not duplicated.
How it works
Everything on the bench is arithmetic on declared inputs: a body's width and warmth, a light's linewidth, an imager's noise, a rain rate. The film's behaviour comes out of the physics first, and the script that ships with the site re-derives every displayed number before the build passes.
detour = (π/2 − 1)·D · φ = 2 asin(√f / 2) · Δλ = φ λ₀² / (π · detour) · jump = d·B/R · lag = v·τ · hits = (R / V(D₀)) · A_top · t = (m_ice·L + m·c·ΔT) / (P + P_leak / COP)
The band. A ray hugging a cylinder of diameter D travels πD/2 where the straight ray travels D, so the detour is at least (π/2 − 1)·D; a real shell adds to it. In a lossless passive medium the group velocity cannot exceed c, so the extra group delay is at least the detour over c, and the cloak's phase, matched to free space at ω₀ by design, slips at that rate everywhere else. Allow a residual scattered power f relative to the bare body's shadow; the slip that produces it is φ = 2 asin(√f/2), since the residual field is E₀(e^{iφ} − 1). The full band is Δω = 2φ/delay, Δλ = φλ₀²/(π·detour). The share of a source of linewidth W that fits is min(1, Δλ/W), halved for the sodium doublet. Turned round, D_max = φλ₀²/(π(π/2 − 1)W).
The sum rule. For any linear, passive, causal scatterer ∫₀^∞ σ_ext dλ = π²(γ_e + γ_m). The body is a conducting prolate spheroid of its volume and height, orientation-averaged through the depolarisation factors; its visible-band share is the extinction paradox's 2A times 370 nm over the total.
The seam. Two eyes at range R, one calibrated and one d aside, the wall B behind the body: along the body's silhouette the displayed and the true backgrounds differ by d·B/R at the wall, d·B/(R(R+B)) in angle, against the eye's arcminute. The lag is v·τ; the pixel pitch the eye needs is θ·R. The 3D view computes the same seam from where the cameras actually stand.
The water. Marshall & Palmer: Λ = 4.1R^−0.21 per millimetre, median-volume drop D₀ = 3.67/Λ, fall speed 3.78D₀^0.67; the number flux is the rain rate over the drop's volume; hits are flux times the body's 0.12 m² horizontal projection; splashes at once are hits times a 30 ms life. The wet film's glint is Fresnel's unpolarised reflectance at n = 1.333.
The heat. Stored: dT/dt = P/(m·c). Insulated: leak = k·A·ΔT/thickness. Pumped: electric = leak/COP, all of it heat. Sunk: budget = (m_ice·L + m·c·ΔT)/(P + electric). The imager sees ΔT over its NETD. Breath is sensible plus latent on 6 L/min at 34 °C saturated; a bare sole's contact temperature is the effusivity-weighted mean, and the print cools as a flux pulse, (√t − √(t − t_c))/√t_c.
Nothing is fitted. The only choices are the tolerance, the areas, the wall, the frame, the pixels, the splash, the insulation, the COP and the ice, each a dial and each declared; the film's shimmer and its water come out first, at the defaults.
The dials that decide what happens
A width, a wavelength and a tolerance on the band; a wall, a range, an offset, a latency, a speed and a pitch on the seam; a rain rate; a pace, a mass, a limit, an ice mass, a night temperature, an insulation thickness and a COP on the heat; a suit, a light, an imager, a window and a tile error in the alley. Between them they draw every number the film could mean.
- Width light must go round. 1 cm to 5 m on a log dial; 0.45 m by default, the Major at the shoulders. The band scales as one over it.
- Design wavelength. 400 to 800 nm; 550 by default. The band scales as its square; the chart also shows 10 µm.
- Suppression required. 3 to 30 dB; 10 by default, a tenth of the shadow's scattering left. Tighter is narrower.
- The wall, the range, the second eye. 0.5 to 20 m behind the body; 2 to 60 m away; 0 to 3 m aside. The wall dial moves the façade in the alley too; in the alley the offset is wherever you stand from the calibrated eye.
- Latency, speed, pitch. 0 to 100 ms (16.7 by default, one video frame); standing, walking, running, falling; continuous to 50 mm pixels (3 by default).
- Rain. 1 to 50 mm/h; 10 by default, steady rain.
- Heat. 80 to 800 W with three presets; 50 to 100 kg; +0.5 to +3 K of core; 0 to 5 kg of ice; 0 to 35 °C nights; 2 to 30 mm of aerogel; COP 1 to 4.
- The alley. Suit off, display or passive; daylight, a white LED lamp, a sodium lamp or a laser; an uncooled or a cooled imager; a 1 to 40 K window with an operator's button; ±0.02 to ±1 K of tile error; bare soles or soles at background; five cameras and a calibrate button.
The claims, as they stand
The film's three behaviours, the two routes to invisibility, the sum rule's reading, the display's limit and the thermal half, with where each lands when the arithmetic is run.
| A suit that shows its background, ripples when it moves, and is found by water proposed by the film, 1995 | PREMISE | The premise. Three behaviours and no numbers. Each is a piece of physics the bench prices: a display's seam, a display's latency, a wet surface's splashes and glint. |
| A passive cloak can hide a human body from ordinary light proposed by the word "cloak", in the usual physics sense | CLOSED | Closed by causality. The detour round a 0.45 m body is 0.26 m, 857 ps at c, and the band a passive cloak can hold at 550 nm is 0.12 pm for 10 dB of suppression: a laser line, one part in three million of daylight. The widest thing daylight can passively hide is a virus. |
| The sum rule forbids a visible cloak proposed by a common reading of Sohl et al. 2007 | CONTESTED | It does not. The extinction integrated over all wavelengths is fixed at π² times the polarisability, 6.2 m³ for this body, and the visible band holds a hundred-millionth of it; the ledger would let that go to the infrared unnoticed. The sum permits invisibility; the delay forbids it. Monticone & Alù (2013) used the rule the other way: a passive cloak's total scattering must go up. |
| A display cloak makes a body invisible proposed by Tachi 2003; the film | CONTESTED | For one eye. The retro-reflective coat returns the image to the projector, so the illusion holds at the projector's position and nowhere else. A second eye 1 m aside with the wall 1.5 m behind sees the background jump 19 cm, 68 times what the eye resolves; it must stand within 1.5 cm of the first for the seam to vanish. |
| It shimmers when she moves proposed by the film | PREMISE | Reproduced. The suit shows the background of one frame ago; at a walk that is 23 mm behind the body, resolved from 80 m; at a run 84 mm; in the opening's fall a third of a metre a frame. A latency ripples and settles, which is what the film draws and what a passive cloak would never do. |
| Rain finds it at once proposed by the film | PREMISE | Reproduced. At 10 mm/h, 1,730 drops a square metre a second and 0.12 m² of crown and shoulders facing the sky: 208 hits a second, six splashes visible at any instant, a dry patch on the ground beneath, and a water film that reflects 58 % of the lamp at a grazing 85°. |
| It hides her from thermal imagers proposed by the suit's name | CONTESTED | A heat budget, not an optics problem. Bare skin at 33 °C on a 20 °C night is 260 times an uncooled imager's noise. Holding the outer skin at the night's temperature puts 100 W into the core (+1.5 K an hour, 81 minutes to +2 K) or into ice (2 kg for 2 h 42 min at rest, 27 minutes running). The film's uses, a drop of seconds and a fight of minutes, are inside the budget; the pack is what it never shows. |
| Perfect tiles would close the thermal case proposed by ADAPTIV, extended to a body | CLOSED | Breath and feet leak past them. Six litres a minute of 34 °C saturated air is 8 W, bright at 4.3 µm to a cooled imager; a bare sole at 25 °C contact leaves a print above the noise for 26 minutes on this model. Low observability in the infrared is a property of everything a warm thing does. |
Try this
- Start in the alley. The display is on and calibrated to the street camera. Press 'the calibrated eye': the body is gone. Press 'one step aside': read the seam chip.
- Walk. With the display on, press 'Walk' under the seam group and watch the image trail the figure by a frame. Set the latency to 0 and it stops; set it to 100 ms and it is a smear.
- Switch the suit to passive and the lamp to laser. The alley turns red and the body vanishes. Switch the lamp to sodium: it is back at 97 %. Go to the Band and read the ladder.
- Turn the rain on and press 'Close'. Count the splashes on the head and shoulders, then look at the ground beneath the figure.
- Switch the thermal imager on with bare skin. Read the chip: 13 K, 260 times the noise. Switch to tiles: the figure is gone at a 25 K window. Press the operator's window button and it comes back as a mosaic; wait five seconds for the breath.
- Go to the Heat. Press 'Run' and read the clock; slide the ice to zero; slide the aerogel to 2 mm and watch the tiles' bill.
- End on the File. Thirteen cards and two verdicts at equal size.
Accuracy
The honest line between what is reported, what is measured, what is modelled on them, and what is a reading:
| Feature | Status | What that means |
|---|---|---|
| The three ledgers | Exact | The detour (π/2 − 1)·D and its delay at c; the slip φ = 2 asin(√f/2) for a stated residual; Δλ = φλ₀²/(π·detour); the transmission as band over linewidth; the seam d·B/R and its angle; the lag v·τ; the pixel pitch θ·R; the drop flux from the Marshall & Palmer spectrum; heat capacity, latent heat, insulation leak, COP and budget; the contact temperature by effusivities. Closed-form on declared inputs, audited by a re-runnable tune script. In the 3D view the display's image is a true projection of the reference eye's render onto the body, so the seam is geometry, not a texture. |
| The physics and the devices | Measured | The causality bound (Craeye & Bhattacharya 2012; Monticone & Alù 2016) and the sum rule (Sohl, Gustafsson & Kristensson 2007); Tachi's retro-reflective optical camouflage (2003); Marshall & Palmer's drop spectrum (1948) with Atlas & Ulbrich's fall speeds; NETD of 50 mK uncooled and 20 mK cooled; the whole-body specific heat of 3,470 J/kg/K; 100, 300 and 700 W as rounded MET values; ADAPTIV's hand-sized Peltier hexagons (BAE Systems, 2011); the He-Ne mode width, the sodium D lines, ice at 334 kJ/kg. Carried with sources, as published. |
| The dials and the staging | Modelled | The tolerance that turns 'cloaked' into a number (10 dB) and the least detour a cylinder allows, so the bound is generous; the body's width and top area; the wall's distance; one frame of latency; 3 mm pixels; a 30 ms splash; 10 mm of aerogel; COP 2.5; 2 kg of ice; +2 K of core; a flux-pulse footprint on dry concrete. The alley, the façade, the mannequin, the splash sprites and the breath puffs are drawn and enter no ledger. The transmission is a top-hat band against a top-hat line, a coarse share. |
| The fiction, as premise | Reported | The film is quoted only as the picture under audit: a suit that shows its background, ripples when it moves and is found by water, with a name that says it also defeats thermal imagers. The work proves nothing about any real case, and this bench returns the favour: it prices the device without reviewing the story. The signed review next door does that, and is linked, not duplicated. |
In one line: The bound, the sum rule, Tachi's device, the drop spectrum, the imagers' noise and the body's constants are the record's; the band, the seam, the lag, the splashes and the minutes are arithmetic on them at declared dials; the alley, the mannequin and the tiles are drawn and say so. The reading is yours.
Sources
- Ghost in the Shell (攻殻機動隊), dir. Mamoru Oshii, Production I.G, 1995: the opening drop and the canal fight, quoted as premise; the site's signed review for the reading.
- Francesco Monticone & Andrea Alù, "Invisibility exposed: physical bounds on passive cloaking", Optica 3(7), 2016, 718-724: the causality and passivity bound on a cloak's bandwidth against the object's size; and "Do Cloaked Objects Really Scatter Less?", Physical Review X 3, 041005, 2013: the sum rule applied to passive cloaks.
- Christophe Craeye & Aniruddha Bhattacharya, "Rule of Thumb for Cloaking Bandwidth Based on a Wave-Packet Argument", IEEE Transactions on Antennas and Propagation 60(7), 2012, 3516-3520: the delay-times-bandwidth form the bench computes.
- Christian Sohl, Mats Gustafsson & Gerhard Kristensson, "Physical limitations on broadband scattering by heterogeneous obstacles", Journal of Physics A 40, 2007, 11165-11182: ∫σ_ext dλ = π²(γ_e + γ_m).
- Masahiko Inami, Naoki Kawakami & Susumu Tachi, "Optical Camouflage Using Retro-reflective Projection Technology", Proceedings of ISMAR 2003, 348-349: the one-viewpoint display cloak, with the film named as its inspiration.
- J. S. Marshall & W. McK. Palmer, "The Distribution of Raindrops with Size", Journal of Meteorology 5, 1948, 165-166; David Atlas & Carlton W. Ulbrich, "Path- and Area-Integrated Rainfall Measurement by Microwave Attenuation in the 1-3 cm Band", Journal of Applied Meteorology 16, 1977, 1322-1331: the drop spectrum and the fall speeds.
- BAE Systems, ADAPTIV press material, DSEi, September 2011: hand-sized Peltier hexagons, of the order of a thousand on a CV90, effectiveness claimed at 300 to 400 m.
- Raytheon AN/ASQ-228 ATFLIR public description: a cooled mid-wave infrared targeting pod; the Tic Tac and Gimbal benches on this site for the clips.
- Standard physiology and heat-transfer values: whole-body specific heat 3,470 J/kg/K (0.83 kcal/kg/°C); resting, walking and running metabolic rates as rounded MET values for 70 kg; DuBois surface area 1.8 m²; skin emissivity 0.98; latent heats of fusion and vaporisation; effusivities of skin and concrete; the Magnus vapour-pressure form; Fresnel's equations at n = 1.333.
- This site: the signed review of Ghost in the Shell; the Stealth Bench (radar, the other observable); the Tic Tac and Gimbal benches (ATFLIR); the Project Hail Mary and Mothership benches on the fiction-pricing line.