Cobalt Glass Between Yellow and Everything Else

Sodium ruins everything. That’s the first real lesson in flame emission spectroscopy once you move past identifying clean salt solutions: every flame is contaminated with sodium, and sodium emits so intensely at 589 nm that it drowns out nearly every other element you’re trying to measure. Sweat from your hands. Dust on the bench. Microscopic salt crystals floating in the air. All of it vaporizes in the flame and screams yellow.

Historical workaround: cobalt blue glass. Hold a piece of dark blue glass between your eye and the flame, and the yellow sodium emission vanishes. The glass absorbs wavelengths from roughly 550–620 nm, filtering out the D-line doublet and letting you see the weaker emissions underneath—barium’s apple-green at 524 nm, strontium’s deep red at 606 nm, potassium’s faint violet at 404 nm. This was standard practice in 19th-century qualitative analysis, back when flame tests were done by eye and “greenish” had to be decoded into a specific element without instruments.

Cobalt glass still works. I tested it yesterday with the prism setup from yesterday—razor blade slit, estate-sale prism, dispersed spectrum on the far wall. Sprinkled table salt into a Bunsen burner flame, spectrum lit up with two bright yellow lines 0.6 nm apart. Held cobalt glass in front of the slit. The yellow lines disappeared entirely. Background blue glow from the flame remained, plus a faint green emission I hadn’t seen before. That’s copper contamination, likely from the burner hardware itself—copper chloride volatilizes at flame temperature and emits at 515 nm. Sodium was hiding it.

Tried this with forge flame samples. The coal fire at the makerspace has been banked since Monday’s session, but there’s a small propane forge for knife work that burns cleaner and hotter than coal. Lit it, brought the prism rig, pointed the slit at the flame from across the bench. Expected clean propane combustion—maybe C₂ Swan bands in the blue-green if combustion was incomplete, otherwise mostly thermal blackbody radiation from the refractory lining.

Got sodium instead. Huge yellow doublet, completely overwhelming. Propane burns around 1980°C with proper air mixing, well above sodium’s vaporization point (883°C). Any salt residue left in the forge chamber from previous flux-assisted welds or quenching operations is now atomized and emitting. Cobalt glass filtered it out. Underneath: strong orange-red line at 606 nm, which is strontium. Why strontium? Checked the forge lining—ceramic fibre blanket rated to 1260°C, manufacturer’s MSDS lists trace strontium aluminate as a binder. It’s outgassing at operating temperature, and the emission spectrum tells you what the lining’s made of before the label does.

Coal forge would be different. Coal combustion produces molecular emission bands—C₂ radicals emit in broad features from 470–565 nm (Swan bands), giving coal flames that characteristic blue-green tint you see in the hottest part of the fire. Not discrete atomic lines, but smeared bands from vibrational and rotational quantum states in diatomic carbon molecules. Atomic emissions are sharp: one wavelength, one electron transition, one photon energy. Molecular emissions are fuzzy: rotation adds extra energy levels, vibration adds more, and you get dozens of closely-spaced lines that blur into a band when your resolution is limited.

The prism can’t resolve Swan band structure. Needs roughly 0.05 nm resolution to separate individual rotational lines, and my setup gives maybe 2 nm at best—good enough for the sodium D-line doublet if I squint, useless for molecular fine structure. Diffraction grating would do better, but gratings are expensive and require precise collimation. For now: prism tells me “that’s molecular emission” by showing broad smeared bands instead of sharp lines. Good enough to distinguish chemistry.

Chloride trick works better than I expected. Made a 50/50 paste of copper sulfate (pool supply) and concentrated hydrochloric acid, dipped a nichrome wire loop into it, held it in the flame. Brilliant blue-green emission at 515 nm, way brighter than straight copper sulfate powder. Chlorides volatilize at lower temperature than sulfates or carbonates—copper(II) chloride boils at 993°C, copper(II) sulfate decomposes around 650°C but doesn’t fully vaporize until much higher. More atoms in the gas phase means more emission. This is why fireworks use chlorine donors like ammonium perchlorate to intensify metal colours—it’s not just oxidizer chemistry, it’s volatility.

Tested strontium chloride from a road flare (cut it open, scraped out the red pellet, dissolved it in water, evaporated to dryness, mixed with HCl). Deep orange-red emission at 606 nm, incredibly intense. Could see it clearly reflected off the ceiling. Tried strontium carbonate next—garden lime product, supposedly pure SrCO₃. Weaker emission, same colour, but took longer to appear and faded quickly. Carbonate thermally decomposes to oxide in the flame, oxide is stable and doesn’t vaporize easily. Chloride goes straight to gas phase.

Noble metals don’t emit. This surprised me more than it should have. Tried gold leaf (gilding supply), platinum wire (thermocouple scrap), silver powder (photographic chemistry leftovers). All three produced no characteristic flame colour and no detectable emission lines in the prism. Gold and platinum have filled d-orbitals and high ionization energies—thermal excitation in a Bunsen flame (~1400°C) isn’t enough to bump electrons into accessible excited states. Silver’s slightly more reactive but still stubbornly refuses to emit in the visible range. The technical jargon: “refractory” elements. They don’t play.

Calcium from limestone flux shows up everywhere. The forge uses crushed limestone (calcium carbonate) as a welding flux because it melts around 825°C and forms a protective slag layer that prevents oxidation during forge welding. Some of that flux vaporizes during the weld cycle. Pointed the prism at forge exhaust yesterday—strong orange-red line at 622 nm, which is calcium. Also a fainter green line at 527 nm, also calcium. The orange is more intense because that transition has a higher probability (stronger oscillator strength), but both are diagnostic.

This is forensic chemistry by firelight. You’re reading what’s contaminating the flame by looking at which electrons are dropping between which orbitals. Sodium? 589 nm, 3p→3s transition. Copper? 515 nm, multiple overlapping d-orbital transitions. Strontium? 606 nm, 5s5p→5s² transition. The wavelength is the fingerprint. The intensity tells you concentration, if you’re careful about calibration and measurement geometry.

Still haven’t pointed the prism directly at glowing steel in the forge. That’s pure thermal emission—blackbody radiation from hot metal, continuous spectrum, no discrete lines. Boring for spectroscopy because there’s no chemical information, just temperature. But the flame around the steel? That’s combustion products, metal vapor boiling off the workpiece, flux outgassing, coal ash contamination. Layered emission from multiple sources at different temperatures. Want to see if I can distinguish hot steel (continuous spectrum peaking in red-orange) from coal flame (C₂ Swan bands in blue-green) from sodium contamination (sharp yellow doublet) all in the same field of view.

Might need to build a proper spectroscope instead of squinting at a wall. Bunsen did it in 1855 by adding a prism and a telescope with a calibrated scale. I’ve got the prism, a DSLR tripod, and no excuse.