Infrared and X-Ray Imaging of Paintings

What Infrared Reflectography Records

A painting is a stack of materials with different optical habits. Varnish and many coloured paints scatter visible light. Longer infrared wavelengths scatter less, so more of the beam reaches the ground and returns. Carbon-rich drawing (charcoal, lamp black, carbon ink) absorbs that returning light and reads dark on the image. A light ground reflects it and reads pale. The resulting picture is a reflectogram. Conservators assemble many frames into a reflectogram mosaic when the camera’s field of view is small.

Infrared photography and infrared reflectography share that logic and part company on wavelength. Silicon sensors in ordinary cameras, even after the infrared-blocking filter is removed, stop near 1100 nm. Copper-containing blues and greens (azurite, malachite) stay opaque in that band. IRR uses detectors that reach further: indium gallium arsenide (InGaAs) cameras, historic vidicon tubes, platinum silicide and indium antimonide systems. Van Asperen de Boer, who named the method, argued from paint-film physics that hiding thickness peaks near 2.0 micrometres. National Gallery of Art (NGA) work in the 1990s put the useful window at about 1.0 to 2.5 µm, with the best band depending on the pigment and the drawing material together.

IRR does not “see through” paint as a window. It records contrast between an absorbing line and a reflecting ground after the upper layers have become more transparent. Thick umber, carbon-rich darks in the paint itself, and some modern pigments still block the view. Metalpoint, red chalk and some iron-gall lines absorb little infrared, so the reflectogram stays blank even when a drawing exists. Read a dark line as carbon until microscopy or another method says otherwise.

X-radiography answers a different question. The beam goes through the sandwich. Lead white, vermilion (mercury), metal hardware and some grounds absorb it and print light on the radiograph. Organic lakes, indigo and varnish print almost nothing. MA-XRF maps where lead, copper, calcium, iron, mercury, arsenic and other elements sit, including some material below the surface. A chalk-rich sketch that infrared misses can still appear on a calcium map, as on Rembrandt’s The Night Watch.

Physics: Wavelengths, Pigments and Density

Conservators choose a method by matching radiation to material. Visible light (about 400 to 700 nm) shows the finished surface. Near infrared (NIR, about 700 to 1100 nm) begins to penetrate some paints. Short-wave infrared (SWIR, about 1000 to 3000 nm) is the working range of infrared reflectography. X-rays used on paintings sit at much higher energy: National Gallery, London practice in the film era used about 35 kilovolts for canvas and 35 to 50 kV for panels. MA-XRF uses a collimated or polycapillary X-ray beam and reads the fluorescent photons that elements emit, not the shadow of the whole object.

The same pigment can be a help in one method and a wall in another. Lead white is a gift on an X-radiograph and a problem for infrared if the layer is thick. Carbon black is a gift for IRR and almost invisible to X-rays. Quartz-rich grounds, common in Rembrandt, barely register on radiographs; the NGA and Rijksmuseum therefore lean on infrared and elemental maps for those pictures. You read the methods as a set.

Infrared photography and infrared reflectography

Infrared photography of paintings dates to the 1930s. R. A. Lyon and Marie Farnsworth published early work on how paint materials behave beyond the visible. Film sensitive to about 700 to 900 nm, later 1100 nm, recorded underdrawing on thin Early Netherlandish panels with more success than on thick Italian oil. A restorer with a modified digital camera and an IR-pass filter (for example a 760 nm or 1000 nm cut-on) still uses this band. Azurite and Prussian blue remain stubborn at those wavelengths, which is why labs still buy specialised cameras that reach toward 1700 nm and beyond.

Infrared reflectography is the name van Asperen de Boer gave to imaging around 2 µm, first with a Barnes lead-sulphide camera (1966–68 papers), then with infrared vidicon television systems (Studies in Conservation, 1969). A vidicon is an electron tube that formed a television image. Early IRR practice photographed the monitor, printed the frames, and pieced them by hand. Resolution was on the order of 800 by 800 samples per frame, so a large panel became a mosaic of many small tiles. Illumination, viewing angle and tube lag produced seams. The method still beat infrared photography on copper greens and blues.

Solid-state detectors changed the trade. Elizabeth Walmsley, John K. Delaney and colleagues at the NGA compared vidicon, platinum silicide and other cameras on test panels with six drawing materials under eight paints (Studies in Conservation, 1994). Peak visibility sat at different wavelengths for different pairs. A camera with even response from 1.0 to 2.5 µm served more combinations than a vidicon that rolled off near 1.8 µm. InGaAs arrays (about 900 to 1700 nm) became the workhorse because they are stable, portable and commercial. They do not reach van Asperen’s 2.0 µm optimum. For malachite-heavy passages, an InSb or filtered PtSi system still earns its keep. The NGA’s later multi-band cameras collect separate frames (for example 1100–1400, 1500–1800, 2100–2400 nm) and assign them to red, green and blue channels. The false-colour reflectogram separates pigments that a single greyscale band muddles.

X-radiography: density through the sandwich

Wilhelm Röntgen discovered X-rays in 1895 and tested wooden blocks painted with lead white. Walter König in Frankfurt made the first radiograph of an actual painting in 1896. The beam that leaves the tube passes through stretcher, canvas or panel, ground, paint, varnish, and any lining or cradle, then hits film or a digital plate. Every absorbing layer superimposes. The image is a sum, not a tidy slice of one buried layer.

Lead white in flesh, highlights and many grounds prints light. Vermilion prints light because of mercury. Metal tacks, keys, nails and lead seals print as hard white marks. The canvas cloth itself is transparent to X-rays; the weave appears because radio-absorbent ground pushed into the interstices between threads. Vermeer’s lead-rich grounds make thread count, cusping and loom faults readable. Rembrandt’s quartz grounds do not. A wooden stretcher prints as a paler band. Cradles on panel paintings cast a grid that can hide the design; some studios pack the cradle with radio-transparent material to reduce that interference.

Kilovoltage sets contrast against penetration. Counting Vermeer’s technical notes put the useful range for paintings at about 15 to 40 kV, with the lowest kV that still gives a reasonable exposure preferred. The National Gallery, London, published film settings of 35 kV and 0.9 mA for canvas, 35-40 kV for thin panels, and 45-50 kV for thicker wood, with 50-second exposures on 300 by 400 mm industrial film. Digital plates now replace much of that workflow. E-RIHS Italian mobile radiography quotes about 144 shots and three hours for a 1 by 1 metre area, with software stitching.

MA-XRF: elemental maps, not a second radiograph

Macro X-ray fluorescence scans a small X-ray spot across the painting and records a spectrum at each point. Software turns those spectra into maps: lead, copper, iron, calcium, mercury, tin, antimony, arsenic, cobalt. A lead map can resemble an X-radiograph because lead white dominates both. A calcium map can show a chalk ground or a chalk-rich sketch that neither IRR nor a radiograph recorded. A copper map can show a verdaccio, a copper drier, or azurite that infrared treated as a transparency problem.

Joris Dik, Koen Janssens and colleagues visualised a lost head under Van Gogh’s Patch of Grass at the DESY synchrotron in 2008, combining antimony and mercury maps. Matthias Alfeld and the Antwerp-Delft group then built mobile tube-based scanners so the work could happen in a gallery. Bruker’s M6 Jetstream (described in 2013) was the first commercial painting scanner; XGlab’s Crono followed. Scan areas on the order of 50 by 50 cm or 80 by 60 cm take hours. The Night Watch required a custom gantry and many days.

Light elements (carbon, nitrogen, oxygen) give weak fluorescence in air, so carbon underdrawing remains IRR’s job. Overlapping layers still mix in a map: a copper signal can come from the surface, from below, or from both. Cross-sections and reflectance imaging spectroscopy (RIS) settle those arguments.

How a Technical Imaging Study Is Run

A serious campaign is a sequence, not a single photograph. Van Asperen de Boer taught that sequence at Groningen from 1976: unframed inspection in visible light, ultraviolet, measurements, binocular microscopy, then IRR and X-radiography when the kit was available, then samples if the question still needed a cross-section. The NGA’s Dutch catalogue “Notes to the Reader” still follows that order. Operation Night Watch added hyperspectral RIS and MA-XRF at gallery scale, with the public watching through glass. The sequence stayed the same: start with what the eye and the microscope can state, then add radiation that answers a named question.

You X-ray a painting to test panel joins, a suspected reuse, a lead-white modelling habit, or a structural repair. You run IRR to test for carbon drawing, a transfer grid, or a change in a contour. You run MA-XRF when the question is which element, or when infrared and X-ray both stayed quiet, as with Rembrandt’s chalk sketch. The question governs the dose, the band, the scan step and the time the object spends off the wall.

Preparing the painting and capturing frames

The painting comes off display. Frames, glazing and some backing boards leave, because they add density and reflections. Conservators photograph the recto and verso in visible light, then under ultraviolet, then in raking light. A stereomicroscope pass notes craquelure, fills, retouch and pigment mixtures at the surface. Those notes later stop you from reading a modern fill as an artist's change.

Infrared capture needs a stable easel, even illumination, and a detector matched to the question. Tungsten halogen lamps are the historic source; some studios now use filtered LEDs to cut heat and ultraviolet. The camera sits on a tripod or a motorised easel. For a vidicon or a small InGaAs array, the operator steps across the surface, overlapping each tile. Opus Instruments’ Osiris, built with the National Gallery, London, moves an InGaAs line sensor inside the camera and writes a mosaic up to about 16 megapixels. The successor Apollo uses a cooled 128 by 128 InGaAs area sensor, stitches tiles in software, and outputs 16-bit files up to about 26 megapixels; one full scan cycle takes about 20 minutes. Filter sets split 900-1250, 1250-1510 and 1510-1700 nm. The NGA’s InSb and PtSi cameras, with astronomy J, H and K filters, go further into SWIR. Rijksmuseum RIS line scanners (VNIR 400-1000 nm and SWIR 900-2500 nm) ride a 3-D frame that keeps focus as the painting warps.

X-radiography needs a controlled room or a licensed portable tube. Film or a digital detector goes behind or against the painting; the tube faces the other side. Overlaps of 50 mm or more let the mosaic close. Lead numbers mark plate position. For a cradled panel, the team may X-ray at an angle (stratiradiography) so features on the reverse separate from those on the front, as Yvonne Szafran’s Getty team did on Heemskerck’s Ecce Homo. MA-XRF needs a scanner that can sit a few centimetres from the paint, a clear floor, and time. Dwell times below 10 ms per pixel are now possible on commercial machines; a whole altarpiece still eats a week.

Mosaicking, registration and reading the images

Raw tiles are useless until someone registers them to a colour photograph. Early IRR mosaics were cut prints on a light box. The NGA and George Washington University later published automatic registration (Conover, Delaney and Loew, SPIE 2013). The Rijksmuseum registered 200 hyperspectral swaths of The Night Watch to a high-resolution colour image at sub-pixel accuracy. False-colour IRR assigns three infrared bands to RGB; the colours are a code, not a reconstruction of the original palette. The NGA Vermeer papers warn readers of that point in the methodology notes.

Reading starts with materials. A dark IRR line that follows a contour and stops at a reserve is underdrawing. A dark patch that matches a painted shadow is carbon in the paint. A pale X-ray shape that does not match the surface may be a buried composition, a lead-white sketch, or a repair. A calcium MA-XRF map that traces spears later painted out is a sketch in chalky paint.

Conservators check the same passage under the microscope before they publish a claim. The NGA’s 2020-22 Vermeer campaign is a model: MS-IRR, XRF maps and RIS together, then a written argument about brushwork in the underpaint. Wavelength range, detector, filters, kV, mA, scan step and software belong in the report. A vidicon reflectogram from 1980 and an Apollo mosaic from 2024 are not interchangeable evidence.

Historical Development of the Methods

X-radiography arrived first. After König’s 1896 painting radiograph, Alexander Faber, a Weimar medical radiologist, patented the X-radiography of oil paintings in 1914, unaware of König’s work. The patent’s practical effect was small until museums wanted machines. During the First World War, doctors with portable tubes (Leo Gerard Heilbron in Amsterdam, André Chéron in Paris, Guido Holzknecht in Vienna) examined paintings as a sideline. In 1924 the Bayerische Staatsgemäldesammlungen in Munich installed the first museum X-ray set and then faced a royalty fight with Faber. Philips and Siemens-Reiniger-Veifa bought the rights in 1931 and sold a unit designed for pictures.

Alan Burroughs (1897–1965) turned radiographs into a research archive. After supervising the X-ray of a mummy case in Minneapolis in 1923, he joined Edward Forbes’s Fogg Museum circle in the mid-1920s, beside George Stout and Rutherford John Gettens. From 1925 to 1944 he built a library of “shadowgraphs” of securely attributed paintings, travelling with a portable Picker unit to the Louvre, the Kaiser-Friedrich-Museum, Belgian and English collections, and the National Gallery, London (1929-30). In 1939 he commissioned a 60-pound shock-proof Art-X from Campbell X-ray Corporation; it stayed in use at the Fogg until 1971. The Harvard Art Museums now hold on the order of 5,400 of those plates. Burroughs’s 1938 book Art Criticism from a Laboratory already warned that X-ray evidence still needs interpretation. Wilhelm von Bode had called the method nonsense. The fight was about who reads the plates.

Infrared photography of art objects took hold in the 1930s, the same decade Farnsworth measured infrared absorption of paint materials. Johannes Taubert and others showed that blue and green draperies stayed closed to IR film. Van Asperen de Boer (Johan Rudolph Justus, known as Dolf; 1935-2020) answered that limit at Amsterdam’s Central Research Laboratory, named infrared reflectography, and in 1970 took his doctorate at the University of Amsterdam on the method. From 1976 he taught a Groningen minor in technical examination; in 1987 he held a chair there in the technical examination of works of art.

Hamamatsu vidicons, NGA PtSi and InSb cameras, then the National Gallery, London’s SIRIS scanner (2005) and Opus Osiris/Apollo, moved IRR from photographed television monitors to portable mosaics. Hyperspectral RIS treats every pixel as a spectrum from the visible into SWIR. MA-XRF, from the 2008 synchrotron experiment to gallery scanners after 2011-13, added elemental specificity that neither infrared nor a radiograph supplies.

Who Developed and Used the Toolkit

Van Asperen de Boer defined IRR, chose the vidicon, and trained Dutch technical art historians at Groningen. Molly Faries and others carried IRR expeditions into collections that had no laboratory. Elizabeth Walmsley, John Delaney, Paola Ricciardi, Kathryn Dooley and Melanie Gifford made the NGA a centre for multi-band infrared and imaging spectroscopy. David Saunders, Joseph Padfield, John Cupitt and Rachel Billinge at the National Gallery, London, turned scanning cameras, digital X-ray mosaics and Early Netherlandish reflectograms into collection practice.

On the X-ray side, König, Faber and the Munich museum made the method institutional. Burroughs made it comparative. Martin de Wild in The Hague, Johannes Wilde in Vienna and Kurt Wehlte in Germany supplied plates to his archive. Padfield, Saunders, Cupitt and Atkinson (2002) remain the reference for kV, overlap and archiving.

MA-XRF is a younger cast. Dik (Delft), Janssens (Antwerp), Geert Van der Snickt, Alfeld, and Bruker Nano’s Michael Haschke and Klaus Erler moved the method from a synchrotron beamline to a motorised stage. Petria Noble’s team at the Rijksmuseum ran those scanners on The Night Watch in public. Private firms sell time on the same classes of camera. A museum technical study still starts with microscopy, documented bands, a named question, and a reader who has looked at a hundred reflectograms.

Named Campaigns and What the Methods Recorded

The useful case studies show a protocol under pressure: a large canvas in a public gallery, a reused oak panel, a small tronie whose underpaint handling is in dispute, an altarpiece with paint on both faces. Each campaign chose radiation to match a material. The NGA paired multi-band infrared with lead maps because the buried man in Girl with the Red Hat was carbon in the hat and lead in the flesh. The Rijksmuseum used a calcium map because Rembrandt’s first sketch on The Night Watch was chalk-rich paint on a quartz ground. The Getty needed an angled X-ray because Heemskerck had painted both sides of a wing. For the vocabulary of artist’s changes, use the pending pentimenti guide. For the forensic story of fakes, use the pending detection guide. The paintings below illustrate how the radiation was chosen, what it recorded, and what it could not.

National Gallery of Art: Vermeer campaign, 2020-22

Covid closure let the NGA move all four Washington pictures by or attributed to Vermeer into the conservation studio: A Lady Writing, Woman Holding a Balance, Girl with the Red Hat, and Girl with a Flute. The team unframed each work, examined recto and verso, used a stereomicroscope to 100x, ran ultraviolet, took X-radiographs, and collected infrared with several cameras. Older files labelled “Vidicon” used a Hamamatsu C1000-03 with a lead-sulphide tube and a Wratten 87A filter. Newer files name micron ranges and one of four cameras: Kodak 310-21X PtSi (1.5-2.0 µm), Mitsubishi M600 PtSi (1.2-2.5 µm), Indigo/FLIR InGaAs, and a Santa Barbara Focalplane InSb SBF187 with band filters between 1.1 and 2.5 µm. Reflectograms were mosaicked and registered to colour with the NGA-GWU algorithm.

Multispectral IRR combined three infrared bands into false colour so pigments separated more than in greyscale. X-ray fluorescence imaging spectroscopy mapped elements across the whole surface. On Girl with the Red Hat (oil on a single oak plank, about 1669), infrared at 1.1 to 2.5 µm recorded a bust-length man, rotated 180 degrees to the girl: broad-brimmed hat, curling hair, cape, vigorous background strokes. The X-radiograph and the lead map showed his face, whitish collar and garment.

The process point is the pairing: IRR caught carbon-rich darks in the hat and hair; lead mapping caught the flesh. Either method alone would have been a fragment. The same campaign used underpaint handling on Girl with a Flute (awkward copper-map brushwork against Vermeer’s sleeve in Woman Holding a Balance) as technical evidence in an attribution argument. The imaging contribution was a map of how the paint was put on.

Rijksmuseum: Operation Night Watch

Rembrandt’s militia portrait (1642; 378.4 by 453 cm) had been studied with infrared photography, X-radiography and samples in the 1970s. Those methods showed changes and radio-opaque pigments. They did not map the first sketch. In summer 2019 the Rijksmuseum built a glass chamber in the Gallery of Honour and began Operation Night Watch: MA-XRF, macro X-ray powder diffraction, RIS, optical coherence tomography, high-resolution photography and 3-D scanning, plus micro-samples.

RIS used two line-scanning hyperspectral cameras (VNIR 400–1000 nm at 2.54 nm sampling; SWIR 900–2500 nm at 6 nm) on a computer-controlled 3-D frame with distance sensing to hold focus. Two hundred swaths, spatial resolution about 168 micrometres, were mosaicked to a colour master. Scans ran after closing because stray gallery light would contaminate the reflectance cube. SWIR false colour (bands near 1603, 1201 and 1040 nm) made upper pigments more transparent and showed a change in the length of Willem van Ruytenburch’s spear tip, plus dark sketch lines at a boot and a helmet.

The beige underpainted sketch, in a chalk-rich paint on a brown quartz ground, appeared on the MA-XRF calcium map, announced on 7 December 2021. Infrared had little to grip: the sketch was not a carbon drawing. X-radiography had little to grip: quartz grounds are radio-transparent. Calcium mapping was the method that matched the material. The same campaign recorded extra spears in the upper right, feathers on Claes van Cruijsbergen’s helmet later painted out, a shift in Rombout Kemp’s leg, and signs of a sword between captain and lieutenant. Those facts describe how Rembrandt found the composition on the canvas.

Getty and Met: two faces, a reused panel, a transfer grid

Maerten van Heemskerck’s Ecce Homo triptych (1544, National Museum in Warsaw) spent eighteen months at the Getty. Conservators used X-radiography and an Osiris camera (InGaAs, 0.9–1.7 µm). The X-ray of the left wing’s interior showed Saint John the Evangelist’s head from the reverse, loaded with lead white, printing through the panel. Infrared recorded a change in the coat of arms: swans’ heads from open and vertical to closed bills facing down. Stratiradiography, spinning the panel at an angle, separated those overlapping densities. The process lesson is double-sided panel painting: one radiograph is a sandwich of both faces until geometry or infrared pulls them apart.

Rembrandt’s An Old Man in Military Costume (Getty, about 1630–31) had long shown a second head-and-shoulders figure, inverted, under the surface. MA-XRF mapped lead white and a mercury pigment in the buried face, copper in the cloak. Infrared reflectance imaging spectroscopy and targeted cross-sections then showed more than one attempt to place the lower figure, and a blocking layer before the panel was rotated. Again the methods split the labour: elemental maps for pigment identity, infrared for sequence, samples for stratigraphy.

At the Met, IRR of Jacques-Louis David’s The Death of Socrates recorded carbon underdrawing, including drapery folds, that follows the squared transfer from the Bayonne studies. An X-radiograph showed a stylus hole at the vanishing point. MA-XRF mapped further shifts (Socrates’ forearm, the cup) that infrared missed. The Met’s note on two Adoration of the Magi panels states a further limit: a first dry charcoal outline may have been brushed off after liquid lines, so the reflectogram records the second drawing.

Failure Modes

Infrared reflectography goes silent when the drawing is not carbon. Metalpoint, red chalk, some iron-gall inks and pale earths absorb too little. Vermeer’s underpaint, in the NGA’s account, can be invisible in traditional IRR when it lacks carbon black; false-colour MS-IRR and XRF maps then do the work. Absence of a reflectogram line is not proof of autograph status.

Carbon in the paint layer mimics drawing. A dark contour that is a painted outline, a restoration in lamp black, or a carbon-rich glaze will read as underdrawing if you skip the microscope. Thick lead white, dense umber and some modern blacks keep even 2 µm radiation from reaching the ground. InGaAs cameras that stop at 1700 nm still struggle with malachite, the problem van Asperen de Boer already flagged when he compared Heimann vidicons (to about 1.8 µm) with the Barnes 2.0 µm detector.

X-radiographs fail in the opposite direction. Organic colourants, quartz grounds and varnish print almost nothing. Everything that remains is stacked: a cradle bar across a face, a lining adhesive, a lead mend on the reverse, two compositions on one panel. Burroughs already said the plate does not give an exact answer. A reused canvas with a lead-white sketch can look like a “hidden masterpiece” when it is a blocked-out study. Thermography (mid- and long-wave infrared, heat) maps temperature, not underdrawing. Search results that mix heat cameras with reflectography are answering a different question.

MA-XRF misses carbon, struggles with overlapping layers, and takes time that a loan schedule may not allow. A copper map does not tell you whether the copper is azurite, malachite, verdigris or a drier until a spectrum or a sample says so. Scan step and dwell time set whether a fine drawing appears or smears. False colour in MS-IRR is a visualisation. It is not the original colours returning.

Over-reading is the human failure mode. A reflectogram is evidence of materials and sequence. It is not a verdict on authorship, date or fraud. Pair it with dendrochronology, thread count, documentary history and connoisseurship, or leave the claim unmade.

Costs and Access Today

A modified full-spectrum DSLR and filters cost as much as a serious camera body: hundreds to a few thousand pounds, within reach of a private conservator. That kit records the 700-1100 nm band. An InGaAs reflectography camera is capital equipment. A European retail listing for the Opus Apollo in 2026 sat near €49,000; museums that bought Osiris in the 2010s already hold the previous generation. Filter sets, macro lenses, illumination and a motorised easel add more. MA-XRF scanners (Bruker M6 Jetstream, Crono and lab-built gantries) sit at the scale of a conservation-science budget, not a freelance invoice. Hyperspectral RIS systems of the Night Watch class are project infrastructure.

Time is the other cost. An Apollo scan cycle of about 20 minutes covers one view; a large altarpiece needs many positions, filters and a day of registration. E-RIHS mobile digital radiography quotes about three hours per square metre. MA-XRF of a half-metre passage can take an afternoon; a Rembrandt-sized canvas takes a campaign.

Access routes split three ways. In-house museum labs image their own collections and, by partnership, loans. E-RIHS and MOLAB offer competitive access to mobile infrared, radiography and XRF for research proposals; market work goes on a contract with the host lab. Private conservation studios advertise IRR to about 2200 nm, IR photography to 1100 nm, and portable X-ray. They do not publish a standard price list. A collector’s path is to hire an accredited conservator who already has a relationship with a lab. Radiation licences, insurance and object handling dominate the real cost.

Published technical images are the open route for students: NGA catalogues, JHNA’s Vermeer viewers, Rijksmuseum calcium-map releases, Getty Heemskerck pages, Met Perspectives essays. Use those files with the stated wavelength and detector. A greyscale JPEG labelled “infrared” without a band is a hint, not a measurement.

FAQ

Museum labels and technical essays often name infrared reflectography without explaining what the camera actually records. The questions below cover how IRR works, how it differs from X-radiography, what underdrawing looks like, and when a study fails. Answers stay on methods and limits. Pentimenti as a vocabulary topic and forgery narratives belong in their own articles.

How does infrared reflectography work?

The operator lights the painting with a near-infrared source (halogen lamps are the historic choice) and records the return with a detector beyond a normal camera’s range, InGaAs (about 900-1700 nm) or a system that reaches toward 2 µm. Carbon drawing on a light ground absorbs the infrared and prints dark. The file is a reflectogram; many frames become a mosaic. Van Asperen de Boer’s 1969 paper remains the physics reference: hiding thickness of paint films peaks near 2.0 micrometres.

What is the difference between infrared photography and infrared reflectography?

Infrared photography uses film or a silicon digital sensor, so it stops near 1100 nm. It is fast, high in pixel count, and enough for some underdrawings on thin paint. Infrared reflectography uses specialised detectors at longer wavelengths, so copper blues and greens that block IR photographs become more transparent. The National Gallery, London glossary states that difference in those terms.

What does an X-ray of a painting show?

It shows a density sum of everything in the beam: lead white modelling, vermilion, tacks, stretcher bars, panel joins, cradles, some repairs, and sometimes a buried composition in dense paint. Carbon underdrawing and organic colourants stay quiet. Canvas weave appears when ground filled the gaps between threads. Quartz grounds, as on The Night Watch, stay quiet. Painting settings sit around 15-50 kV depending on support thickness.

Can infrared see every underdrawing?

No. The drawing has to absorb infrared. Carbon does. Metalpoint, red chalk and some inks do not. Paint that is itself carbon-rich, or a thick lead-white blanket, hides a drawing that exists. Rembrandt’s chalky beige sketch on The Night Watch needed a calcium MA-XRF map, not IRR. Absence of a reflectogram line is not absence of a design.

How does MA-XRF differ from an X-radiograph?

An X-radiograph is one shadow of all layers at once, weighted toward heavy atoms. MA-XRF is a raster of spectra turned into separate element maps. A lead map can look like a radiograph. A calcium, copper or mercury map can show a sketch or a pigment that the radiograph never isolated. Carbon still belongs to infrared. Overlapping layers still mix.

How much does technical imaging cost, and who can commission it?

A modified camera for infrared photography is a studio purchase. A true IRR camera is a five-figure euro capital item (Apollo listed near €49,000 in a 2026 retail catalogue). MA-XRF and hyperspectral RIS are museum or university equipment. Private owners work through an accredited conservator and a lab; E-RIHS competitive access serves research. Time on the object, licences and handling exceed the shutter cost.

Does infrared or X-ray imaging damage a painting?

IRR and infrared photography are non-contact optical methods. Heat from lamps is the practical risk; studios control distance, duration and, where possible, filter ultraviolet. X-radiography and MA-XRF use ionising radiation at doses conservation scientists consider acceptable for paintings under licensed protocols; they are still controlled procedures, not something to improvise. No imaging method replaces a conservator’s decision about whether the object can come off the wall.