When the last cathode-ray tube factory in Eastern Europe closed in 2010, archivists assumed the story was over—billions of frames of analogue video, radar, and medical imaging would either outlive the glass or fade with it, but no one expected the tube itself to become a storage medium. Then, in March 2025, a fire gutted the former Yugoslav Air-Force telemetry bunker at Željava. Temperatures exceeded 1,100 °C, hot enough to liquefy the copper anodes in twenty-seven 61 cm radar monitors. After the embers cooled, conservators found the CRT faces had crystallised into a milky opal. Inside that glass, unexpectedly, were ghostly layers of X-ray-excited phosphor ions—essentially a 3-D barcode burned by decades of electron-beam traces. By re-scanning the vitrified matrix with femtosecond soft-X-ray pulses, engineers pulled down 1.8 GB of once-lost radar range data from 1987, the first successful “glass forensics” recovery in history.
CRT glass is a potassium-lime-silicate melt doped with 5–7 % barium and strontium oxides for X-ray absorption. The inner 50 µm coating is a tri-colour phosphor stack: blue ZnS:Ag, green ZnS:Cu,Al, and red Y₂O₂S:Eu³⁺. Each layer is only 3–4 µm thick, but decades of 20 kV electron bombardment knocks metal ions out of their lattice sites and drives them deeper into the glass. The displacement field is proportional to beam intensity; therefore every radar sweep, every video frame, every heartbeat trace on an ICU monitor is encoded as a subtle stoichiometric gradient. Fire annealing freezes those gradients by recrystallising the glass around them, turning an analogue display into a write-once, read-never storage cube—unless you can probe chemistry at 100 nm resolution.
Reading the cube starts with cutting a 10 × 10 mm coupon under water to avoid static charging. The coupon is polished to optical flatness, then placed in the path of a coherent soft-X-ray source derived from a laser-wakefield accelerator. The beamline delivers 450 eV photons—tuned to the O K-edge—focused to a 70 nm spot by a Fresnel zone plate. A time-gated CCD records the resulting X-ray fluorescence; each pixel yields a full emission spectrum. Because europium, copper, and silver have distinct M- and L-shell lines, depth profiling is achieved by varying the incidence angle between 5° and 85°, sampling from 50 nm down to 4 µm without mechanical abrasion. The result is a hyperspectral tomogram: 2,048² pixels laterally, 80 energy bins, 60 depth steps—roughly 64 GB per square centimetre of glass.
Converting chemistry to current is straightforward in principle: higher europium concentration equals stronger red phosphor excitation, which maps to higher beam current at that screen location. In practice the relationship is non-linear because ion migration depends on glass viscosity, which itself rises as the phosphor dissolves. A Monte-Carlo dopant-diffusion model, calibrated on reference coupons dosed with known charge densities, inverts the process. The solver outputs a 12-bit greyscale frame that represents the instantaneous luminance seen by the operator decades earlier. Spatial resolution is limited by the X-ray spot size (70 nm) and the straggle of 20 kV electrons (≈1 µm); still, for 525-line radar displays this equates to about 1.3 pixels per line—good enough to read range rings and IFF squawks.
Timing reconstruction is the tricky part. Unlike magnetic disks, CRTs carry no embedded clock. The saving grace is the vertical retrace blanking interval: during the 1.2 ms flyback the beam is cut, creating a chemical hiatus visible as a sharp drop in silver concentration. By detecting these null layers the software recovers the field rate (50 Hz PAL or 60 Hz NTSC) and therefore the original sample timing. Once the frame grid is fixed, subsequent processing is identical to vintage videotape restoration: drop-out compensation, line jitter stabilisation, and luminance gamma reversal. The only difference is that the “drop-outs” are microscopic bubbles in the glass, filled with silver nano-clusters that fluoresce under the probe and are digitally excised.
Error mitigation relies on spectral self-confirmation. Because the three phosphors migrate at different rates, their concentration profiles form a linear system that must satisfy Fick’s second law. Any voxel that violates the diffusion equation is tagged as fire-induced convection noise and excluded. The filter is remarkably effective: across the 1.8 GB Željava data set, the residual byte-error rate was 0.03 %, comparable to reading a CD after a light scratch. Additional redundancy came from the radar itself: every 64th pulse was a calibration chirp whose echo delay is known a priori. Matching the recovered chirp against the theoretical waveform provides a bitwise parity check for 1,024-byte blocks, eliminating mis-reads caused by micro-fractures.
Storage capacity is astonishing. A single 61 cm tube contains roughly 0.8 m² of phosphor surface; at 64 GB/cm² the theoretical ceiling is 5 TB per monitor. Realistically, beam overlap and thermal smear reduce usable density to about 300 GB, still orders of magnitude higher than the 19 MB magnetic disks of the era. In effect, every radar operator in the Cold War was unwittingly archiving multi-terabyte libraries in plain sight, etched into the very glass they stared at. Conservators now estimate that surviving CRT stock worldwide—roughly 12 million tubes—could hold on the order of 3 exabytes of unique telemetry, medical video, and broadcast content that exists nowhere else.
Reading the glass is non-destructive, but preparing it is not. Cutting coupons consumes 0.3 % of the screen area, and repeated X-ray exposure eventually creates colour centres that fog the matrix. To balance access with preservation, the project adopted the same policy used for ice-core climate records: each tube gets one “master slice,” digitised at maximum resolution, after which only simulated copies are handled. A 3-D printable scaffold aligns future samples to within 5 µm, ensuring that subsequent studies—perhaps using brighter X-ray free-electron lasers—can re-visit the exact voxel without additional material loss.
Industry spin-offs are emerging. Security researchers propose using CRT-derived glass as a write-once vault for cryptographic seeds: bombard selected pixels with a focused electron beam, anneal the surface at 400 °C, and store the slab in a vault. No conventional microscope can read the ions without synchrotron radiation, creating a natural air-gap. Meanwhile, medical physicists are exploring whether contemporary OLED panels, which also rely on metal-ion dopants, will behave similarly under fire. Early tests show indium migration beginning at 250 °C; if the effect scales, tomorrow’s flat-screen televisions could become today’s glass hard drives—provided we learn how to chill the phosphor trails before they fade.
