Twenty Grooves Where Two Thousand Should Be

11:23 PM — Commercial diffraction gratings cost $200 minimum. Edmund Optics lists a 25×25 mm reflection grating, 600 lines/mm, blazed for 500 nm, at $347 USD plus shipping. For measuring emission spectra that’s probably worth it—way better resolution than the prism, linear dispersion, proper blaze angle to maximize light at the wavelengths you care about.

But you’re not learning anything by buying one. Someone else ruled those grooves on a machine they spent five years calibrating. You get data, not understanding.

Started reading about ruling engines last night. Turns out the first diffraction grating was made in 1785 by stringing hairs between threaded screws—literally a harp tuned for light instead of sound. By 1900, Henry Joseph Grayson’s engine was ruling 120,000 lines per inch. That’s 4,724 grooves per millimetre. At visible wavelengths around 550 nm, groove spacing becomes 211 nanometres, which is less than half the wavelength you’re trying to diffract. Absurd precision.

I’m not attempting 4,700 lines/mm. Targeting 300 lines/mm, which is 3.33 micrometres groove spacing. Still ridiculous but theoretically achievable with a decent micrometer stage and a diamond stylus that doesn’t wander.

Saturday 8:47 AM — Micrometer stage borrowed from the university surplus shop, $40 because the Vernier scale is slightly bent. Digital readout still works: 0.001 mm resolution, 25 mm travel. Clamped to bench. Aluminum blank cut from scrap sheet metal, 50×25 mm, polished with 2000-grit wet/dry sandpaper then buffed with jeweller’s rouge until it reflected the ceiling lights clearly. Any scratch, any fingerprint, any dust particle becomes a permanent defect in the grating.

Stylus is the hard part. Diamond scribe from the engraving kit—meant for marking glass and carbide tooling, not precision optical ruling. The tip radius is maybe 50 micrometres under the loupe, which is huge compared to professional ruling diamonds (5–10 µm radius). Larger radius means shallower groove angle, which means less diffraction efficiency. But I’m testing proof of concept, not manufacturing astronomy gratings.

Mounted the stylus in a makeshift tool holder: brass rod drilled and tapped for a set screw, stylus clamped at 90° to the aluminum blank. Spring tension provided by a bent paperclip pressing down on the back of the brass rod. Target pressure: 10 grams, measured by pressing the stylus against a postal scale until it reads 10 g, then locking the set screw at that height. Wildly imprecise but better than guessing.

9:14 AM — First ruling pass. Micrometer advanced 3.33 µm (one full revolution = 500 µm, divided by 150 clicks). Drag the stylus across the aluminum blank by hand, 50 mm stroke. Takes about eight seconds. Check under loupe: faint line visible but inconsistent depth. Right side deeper than left—stylus pressure wasn’t uniform, probably because I pulled faster at the end of the stroke.

Ruled five more lines. Each one looks different. Some barely visible, some gouged deep enough to tear aluminum. The problem is hand pressure variation and surface irregularities in the blank. Even 2000-grit polish leaves microscopic ridges that push the stylus around.

10:03 AM — Motorized traverse would solve the hand-drag issue. Don’t have one. Considered adapting the small lathe but the carriage runs perpendicular to what I need and the gear train has too much backlash. This is why ruling engines took decades to develop—Jesse Ramsden’s design in the 1770s required a custom screw-cutting lathe that was “particularly advanced” according to the Wikipedia article. You need the precision tooling before you can build the precision optical instrument.

Trying a different approach: clamp the aluminum blank to the micrometer stage, fix the stylus position, advance the blank under the stationary diamond. Same principle, reversed geometry. Gives better control over traverse speed because I’m turning the micrometer knob instead of dragging a tool by hand.

10:41 AM — Twenty grooves ruled at 300 lines/mm spacing. Took 47 minutes. At this rate, a 25 mm grating (7,500 grooves) would take 293 hours. Professional ruling engines ran overnight, sometimes for weeks, in temperature-controlled rooms with vibration isolation. Rowland’s engines had massive granite bases to dampen building vibrations. Mine is clamped to a wooden workbench. Every footstep shows up as a depth variation.

Testing: pointed the ruled section at the sodium street lamp through the window. Held it at ~30° incidence, looked for first-order diffraction. Got a faint yellow streak offset from the zero-order reflection. Not two separate lines—my 300 lines/mm grating doesn’t have enough resolving power to split the 589.0/589.6 nm doublet, or the grooves are too irregular to produce clean diffraction. Probably both.

Commercial 600 lines/mm grating would double the angular dispersion and might resolve the doublet. But 600 lines/mm means 1.67 µm spacing, which is beyond what my micrometer stage can reliably reproduce. The Vernier scale clicks are 3.33 µm—I’d need to advance half a click per groove, and there’s no detent at half positions.

11:58 AM — Temperature is the other nightmare. Coefficient of thermal expansion for aluminum: 23 µm/m/°C. For a 50 mm blank, that’s 1.15 µm per degree. At 300 lines/mm (3.33 µm spacing), one degree of temperature change shifts groove spacing by 34%. The grating stops working. Professional shops rule at night when HVAC is off, building temperature is stable, and no sunlight is heating the room through windows.

Workshop temperature right now: 19.2°C according to the thermostat. Varies ±2°C during the day when the furnace cycles. Completely unworkable for serious ruling. Would need to wait until 2 AM, close all doors, turn off heat, rule in the dark by flashlight.

Not doing that today.

1:32 PM — Tried a reflection test with white LED light instead of sodium. Pointed the ruled grating at a 5000K LED work lamp, looked at the first-order diffracted beam. Saw a faint rainbow spectrum, red on one side grading to blue-violet on the other. Proves the grooves are diffracting light, at least. Can’t measure wavelengths without a proper mounting and detector, but qualitatively it works.

The rainbow is backwards from a prism. Prisms refract blue more than red (higher refractive index at shorter wavelengths), so blue bends farther. Diffraction gratings spread angles by wavelength directly from the grating equation: d sin θ = mλ. Longer wavelength (red) diffracts to larger angle. Opposite physics, opposite result.

2:19 PM — Blaze angle is the next problem I’m not solving today. “Blazed” gratings have asymmetric groove profiles—triangular teeth like a saw blade—angled to reflect light preferentially into a specific diffraction order at a target wavelength. You’re not scribing a vertical groove; you’re carving a tiny angled facet with the diamond held at maybe 2–4° tilt. Get it wrong and your grating is brightest at the wrong wavelength, or all the light stays in zero-order reflection (mirror mode) and nothing diffracts.

My grooves are probably symmetric V-shapes because the diamond stylus is perpendicular to the blank. No blaze angle, so light spreads equally into all diffraction orders. Inefficient but simpler.

Would need a tiltable tool mount to introduce blaze. That’s a machining project on top of an optics project. Not today.

3:46 PM — Ghost lines are the error signature I can’t see yet. Periodic defects in the lead screw—wear patterns, thread pitch errors, thermal expansion cycles—create spurious spectral lines called “ghosts” that appear at predictable offsets from real emission lines. Every ruling engine has unique ghosts. Rowland’s engines, MIT’s engines, they all had published ghost maps. You didn’t just buy a grating; you bought one from a specific serial-numbered machine with known artifacts.

My micrometer stage definitely has ghosts. Can feel a slight catch every 18–20 rotations, probably a worn section in the lead screw threads. That’ll show up as a periodic groove spacing error every 60 micrometres, which at 300 lines/mm means every 18 grooves. Won’t know what the ghost spectrum looks like until I rule enough grooves to test it properly and compare against known emission lines.

Twenty grooves isn’t enough. Need at least 500 to see diffraction quality, probably 2,000 to measure ghosts. That’s six hours at current ruling speed, assuming no mistakes, no interruptions, no temperature drift.

4:02 PM — Stopping here. The 20-groove test section diffracts light, which proves the concept works. Scaling up to a usable grating requires better vibration isolation, temperature control, motorized traverse, and probably a week of overnight ruling sessions when the building is quiet and thermally stable.

Still cheaper than $347.