How to clean a sapphire inspection window without scratching it

A sapphire inspection window needs one sequence, run in this order: confirm the coating, blow grit off with a squeeze bulb from 10-15 cm, drag reagent-grade acetone then 99% IPA across the glass in single straight passes with a fresh wipe surface each time, and prove the result with a flat-field image whose dark-speck count drops by half or more. The window is usually the last optical surface your light crosses before the part, and the first one the camera looks through, so a film of oil, dust or process splash shows up as contrast loss and stray glare long before you spot it by eye. It is also the surface most often ruined in cleaning: the crystal shrugs off abuse, but its anti-reflective coating does not, and a wipe dragged over grit laps the surface like sandpaper. A squeeze bulb, two solvent bottles, lint-free wipes and ten minutes cover the whole job.

What actually scratches a sapphire inspection window?

Bare sapphire is one of the hardest optical materials made — about 2,000 kg/mm² on the Knoop scale, Mohs 9 — so shop dust cannot cut it; the scratch you are avoiding comes from dragging grit under a wipe, or from marking the soft anti-reflective coating on top of the window.

Sapphire is single-crystal aluminum oxide, not a glass, and it is chosen for inspection ports precisely because it resists scratching and most chemicals. On the Mohs scratch scale, quartz and sand sit at 7, sapphire and alumina at 9, diamond at 10. Ordinary dust is mostly silica, so it cannot cut bare sapphire — but alumina or carbide grit from a nearby grinding or lapping operation can, and any particle dragged under a wipe will mark an AR coating, which is a thin film with a fraction of the crystal's hardness.

Two different problems come off a dirty window. A removable film of oil and dust scatters light and costs contrast; cleaning removes it. A scratch or an etched coating is permanent; only careful technique prevents it. Note that a camera's own cover window or an acrylic guard is a softer material with a gentler routine — the camera window cleaning guide covers that case, while sapphire lets you use real solvents.

Is your window coated, and does it change the solvent?

Tilt the window toward a lamp: a colored interference reflection — a blue, green or purple sheen — means an anti-reflective coating, while uncoated sapphire throws a bright, neutral image of the lamp at about 7.7% reflected light per surface, versus under 0.5% per surface when coated.

Each air-to-sapphire boundary reflects light by the normal-incidence Fresnel reflection:

Sapphire's refractive index is about 1.77 in the visible, so R = (0.77 / 2.77)² ≈ 0.077, that is 7.7% per surface. Both surfaces together leave (1 − 0.077)² ≈ 85% transmitted, which is why an uncoated window is visibly reflective. A good AR coating drops below 0.5% per surface and gives the difference back as that colored tint — the quick visual test.

Confirm the coating from the machine drawing or the window spec before the first clean; if you cannot confirm it, treat the window as coated. The decision that follows: dielectric AR stacks tolerate acetone and IPA chemically, but the coating dies mechanically — dragged grit, hard rubbing, abrasive cleaners. A coated window gets neutral reagent solvents and light, fast drags, nothing else. An uncoated window is nearly indifferent to solvents, and the material around it becomes the limit instead, which is the next question.

What should you blow the window off with first?

Blow the window off with a rubber squeeze bulb, filtered canned air held upright, or clean dry nitrogen — never raw factory compressed air, which carries oil aerosol, water and pipe scale — held 10-15 cm from the glass at a shallow angle, so grit leaves the surface at its edge.

Blowing off is the only zero-risk step, so it is also the one to repeat freely: before every wipe pass, before every inspection, and per shift in dusty plants. Compressor air re-dirties the window with an oil film and can embed fine rust particles; a tilted aerosol can sprays cold liquid propellant that chills the crystal and leaves residue, which is why the can stays upright with short bursts. If fine dust settles straight back, the surface is charged — short bursts plus a grounded fixture move it better than more pressure does.

Which solvents are safe on sapphire, and which are not?

Use reagent- or spectroscopic-grade acetone first, then reagent-grade isopropyl alcohol at 99% or stronger; bare sapphire tolerates both, and the solvent choice is decided by the coating and the seal around the window, not by the crystal itself.

Acetone dissolves fingerprint oils and most process grease and evaporates in seconds. Isopropyl alcohol follows to pick up acetone residue, but it dries slowly, so each stroke must finish in one motion or it leaves drag marks. Grade matters more than brand: 500 mL of 70% pharmacy IPA carries 150 mL of water (0.30 × 500 mL = 150 mL), and that water dries into mineral spots across the window. Buy reagent grade, keep the bottles capped and dedicated to the window.

One trap: if the part is an aluminum front-surface mirror rather than a transmissive window, skip the IPA — alcohol attacks aluminum coatings. Household cleaner belongs to ordinary guard glass, not coated optics; if the pane you are actually looking through is a machine door window rather than an inspection port, the machine door window cleaning guide covers that case.

How do you drag a wipe without dragging grit across the glass?

Lay a solvent-wetted wipe flat on the window and drag it off in one straight pass, so the pad lifts grit and film away instead of grinding them in — one pass per wipe surface, never a circular scrub, never a dry wipe.

1. Put on clean nitrile gloves and work under a bright lamp; fingerprints are the main contaminant, and you cannot fight what you cannot see.
2. Blow the window off first (previous section). A drag wipe only ever starts on a grit-free surface.
3. Fold a lint-free optical wipe into a pad about 30 mm across — thick enough that your fingers never come within a wipe thickness of the glass.
4. Wet the pad with reagent-grade acetone until it is damp, not dripping.
5. Lay the pad flat on the window at one edge and pull it straight across in one slow pass, holding the trailing edge. The wipe contacts on the lay-down and the drag only; it lifts residue instead of rubbing it.
6. Fold to a fresh surface, or take a new pad, for every pass. Never retrace a stroke.
7. Follow with one or two IPA passes the same way, then finish with a dry pad section dragged once to stop drying marks. Inspect at a glancing angle and repeat only where residue is still visible.

How do you prove the window is clean again?

Verify with a flat field: keep the light, exposure and gain fixed, image a uniform target before and after cleaning, and count dark specks — a successful clean shows the speck count dropping by more than half and the mean gray level recovering to its reference value.

A bright flashlight at a glancing angle finds films and scratches that the camera view hides. The flat field turns the same judgement into a number, and a printed machine vision reference target makes the comparison repeatable from week to week.

Set the trigger where your reject rate starts moving. A practical starting rule: clean when specks pass 0.1% of the frame or the mean level drops 3% below the recorded clean reference, and re-clean once if a pass does not move the number. Specks that survive two passes unchanged are not dirt — they are a scratch or an etched coating, and no wipe will move them. If the contamination is not on the window at all, it may be inside the camera: the sensor cleaning guide covers that job and its own precautions.

How often should you clean it, and when does it get replaced?

Clean on evidence, not on a calendar: start with a weekly flat-field check, clean when specks pass 0.1% of pixels or the mean level drops 3% below the clean reference, and replace the window once a scratch or chip is visible in the image, because sapphire cannot be re-polished in place.

The asymmetry is the point. Blow-offs carry no risk, so run them daily or per shift in dusty plants. Solvent passes carry a small coating risk every time, so they happen when the flat field asks for one, not on a schedule. Environments with washdown chemistry or oily mist film windows in days rather than months — the washdown lighting selection checks cover those environments and how to spec the station for them.

Replace, don't restore. A scratch appears in images as a fixed dark line or a glint that survives cleaning; an edge chip spreads under thermal cycling; a coating etched by process chemistry never comes back. Keep one spare window on the shelf in its original packaging. Mounting hardware, spare fasteners and the light cables around the port sit in the lighting accessories category — the accessories finder filters them by size — and for the wider camera-side care list, the cameras topic page collects the guides.

Baseline the window this week: shoot one flat field, save the speck count and mean level, and the next cleaning decision becomes arithmetic instead of judgement. If your window, seal or mount is non-standard, the contact page connects you to the applications team.