1. Let’s Get One Thing Straight: This Isn’t About Making Pretty Glass
If you’ve never had a quartz viewport fail inside a live etch chamber, consider yourself lucky. I’ve seen it happen three times in my career. The first time, we lost an entire batch of 300mm wafers – about $450,000 in scrap – before someone even noticed the chamber pressure had drifted. The crack was barely visible to the naked eye, maybe 12 microns wide at the surface. But under vacuum and thermal cycling, it propagated straight through the full thickness of the window.
That’s the reality of machining quartz glass for semiconductor tools. You’re not making a decorative window. You’re making a pressure boundary that sits between a plasma environment and a cleanroom. If it fails, the tool goes down, the wafers get contaminated, and your fab manager starts asking very uncomfortable questions.
So when we talk about “preventing micro-cracking,” we’re not talking about cosmetic perfection. We’re talking about preventing catastrophic failure.
This guide is written for CNC programmers, process engineers, and procurement leads who are tired of inconsistent results with quartz. I’m going to walk through exactly what works in practice – not just what the textbook says – based on running thousands of quartz viewports across multiple 5-axis machining centers over the past eight years.
2. Why Quartz Breaks Differently Than Metals
Before we get into the machine settings, you need to understand the physics of what’s happening at the cutting edge. I’m not going to bury you in equations, but there’s one concept you absolutely have to internalize.
Metals – aluminum, steel, titanium – deform plastically. When your endmill engages a chunk of 6061, the material shears, the chip curls, and the energy dissipates as heat. You can push hard, back off the feed, and the material forgives you.
Quartz does not forgive you.
Quartz (SiO2SiO2) is a covalent glass with no slip systems. It stores elastic energy until the stress exceeds its fracture toughness (KIcKIc), and then it lets go all at once. That’s called brittle fracture.
When a diamond grit or cutting edge plows into quartz, two types of cracks form:
·Lateral cracks run parallel to the surface, roughly 10–30 µm below it. These are the ones that actually remove material – they pop out chips and create the surface you see.
·Median (or radial) cracks run perpendicular to the surface, driving straight down into the bulk material. These are the killers. They don’t show up on a visual inspection. They don’t affect surface roughness. But they’re sitting there, waiting for thermal stress or vacuum pressure to pull them open.
Here’s the dirty secret that many machining guides don’t tell you: you can achieve a beautiful, glossy surface finish and still have median cracks 50–80 µm deep underneath it. I’ve seen parts pass a visual inspection, pass a profilometer check, and then fail in thermal cycling after two weeks in a CVD tool. The surface looked like a mirror. The subsurface was a minefield.
The goal of semiconductor-grade quartz machining isn’t just to make the surface smooth. It’s to eliminate the subsurface damage entirely – or at least push it down to a depth that’s no longer structurally significant.
3. The Ductile Regime: Where Quartz Actually Flows
The only reliable way to prevent median cracks is to operate in what’s called the ductile-regime machining zone. This isn’t a marketing term – it’s a well-documented physical phenomenon that’s been studied in ultra-precision grinding since the late 1980s.
At very small depths of cut, brittle materials don’t fracture. They flow plastically, just like metal. The chips look like continuous ribbons instead of shattered dust. The surface ends up with a mirror finish, and more importantly, the subsurface is free of median cracks.
The threshold depth is called the critical chip thickness (dcdc). For fused quartz, that threshold is typically between 0.1 µm and 1.0 µm, depending on the specific grade and the tool geometry.
That number – less than one micron – is why standard CNC milling doesn’t work for quartz finishing. If your depth of cut per pass is 0.05 mm (50 µm), you’re fifty times above the ductile threshold. You’re forcing the material into brittle fracture from the first engagement.
So the recipe for crack-free machining is straightforward in theory, but brutal in practice:
·Roughing (remove bulk material): Operate in brittle mode. Deep cuts, high material removal rates. Accept that you’ll get subsurface cracks. That’s fine – you’ll remove the damaged layer later.
·Finishing (final surface): Operate in ductile mode. Depth of cut under 1 µm. Feed rates slow enough to keep chip thickness below the threshold. This pass removes the cracked layer left by roughing and leaves a pristine subsurface.
The challenge is that ductile-mode finishing takes forever. A single viewport might take four hours of machine time for the finishing pass alone. That’s why many shops skip it and try to polish their way out of the problem. That’s also why those shops have high field-failure rates.
4. Rotary Ultrasonic Machining (RUM): The Game-Changer
This is where the real industrial practice diverges from the academic literature. In a research lab, you can use a single-point diamond turning tool with a 0.5 µm depth of cut and get perfect ductile-mode results on a small sample. In a production shop, that approach is too slow to be economical.
The production-proven solution is Rotary Ultrasonic Machining (RUM) .
Here’s how it works: a standard CNC spindle is fitted with an ultrasonic actuator that superimposes a high-frequency axial vibration – typically 20 to 40 kHz – on top of the normal rotation. The diamond tool vibrates in and out of the workpiece thousands of times per second, with an amplitude of about 5–15 µm.
This vibration does two things:
- It reduces the average cutting force by 40–50%. The tool contacts the quartz, removes a tiny chip, and then retracts before the force builds up to the fracture threshold. The material never sees the peak load that would create a median crack.
- It improves coolant delivery. The pumping action created by the vibration forces coolant into the cutting zone more effectively than flood cooling alone. This is critical because heat buildup is a major contributor to thermal micro-cracking.
We’ve run side-by-side tests on the same quartz grade, same tool, same coolant, same feed rate. Conventional milling produced visible edge chipping and subsurface cracks averaging 35 µm deep. RUM, with the same parameters, produced a crack-free surface with subsurface damage under 5 µm – and that remaining 5 µm was easily removed in the finishing pass.If you’re machining quartz for semiconductor viewports and you’re not using RUM, you’re making your life unnecessarily difficult.
5. Tooling Selection: The Details That Actually Matter
Every shop knows you need diamond tools for quartz. But not every shop understands which diamond tools, and that’s where many projects go off the rails.
Forget electroplated diamond tools for finishing. I’ve used them. They’re cheap, they hold diamonds well, and they’re great for roughing fiberglass or ceramics. But for quartz finishing, they’re a liability.
The problem is the bond matrix. Electroplated tools have a thin layer of nickel holding the diamond grits. The bond is hard and stiff. When a grit hits a quartz surface, the force transfers directly through the nickel into the tool body. There’s no give. If one grit digs in too deep, it creates a median crack before the adjacent grits can share the load.
Resin-bonded diamond tools are different. The diamonds are suspended in a phenolic or polyimide resin matrix that’s about 30–40% softer than nickel. This soft matrix does two things:
·It allows the tool to “dress itself” during cutting. Worn diamonds pull out, fresh diamonds expose themselves. The cutting edge stays sharp without manual intervention.
·It absorbs shock. When a grit hits a hard spot, the resin deflects slightly, distributing the load across multiple adjacent grits. That load-sharing effect is the difference between plastic flow and brittle fracture.
One more detail that’s rarely mentioned: the grit shape matters as much as the grit size. For roughing, sharp, blocky grits are fine – you’re removing material fast, and you don’t care about subsurface damage yet. For finishing, you want rounded, “truncated” grits. This is achieved through a truing process where the wheel is run against a SiC or alumina brake, flattening the sharp peaks. A truncated grit produces a shallower scratch and a lower stress concentration than a sharp grit.For a typical 200mm semiconductor viewport, our standard finishing tool is a resin-bonded diamond wheel, 1200 mesh (grit size ~5–8 µm), with a concentration of 75–100%, trued to a radius of about 0.5 mm at the edge.
6. Cutting Parameters: The Real Window That Works
I can give you a generic parameter table, but that’s not very useful because every machine – even two machines of the same model – behaves slightly differently. Instead, I’ll give you the starting point we use, and then explain how to adjust it based on what you see and hear.
Our standard starting parameters for RUM finishing of fused quartz:
| Parameter | Roughing | Semi-Finishing | Finishing (Ductile) |
| Spindle Speed (RPM) | 6,000 – 8,000 | 10,000 – 12,000 | 14,000 – 18,000 |
| Feed Rate (mm/min) | 200 – 400 | 80 – 150 | 20 – 50 |
| Depth of Cut (mm) | 0.10 – 0.30 | 0.02 – 0.05 | 0.002 – 0.005 |
| Ultrasonic Frequency (kHz) | 20 | 25 | 30 |
| Ultrasonic Amplitude (µm) | 10 | 8 | 5 |
You’ll notice the finishing depth of cut is 2 to 5 microns. That’s not a typo. It’s painfully slow, but it’s the only way to stay in the ductile regime when using a multi-grit wheel.
How do you know if your parameters are correct? Listen to the machine. A stable ductile-mode cut produces a high-pitched, continuous hiss – almost like fine sandpaper on glass. If you hear chattering, squealing, or intermittent banging, you’re in brittle mode, and you’re creating cracks.
Also, check the chips. In ductile mode, the swarf from quartz looks like fine white flour – powdery, but with individual particles that are smooth and rounded under magnification. In brittle mode, the chips are jagged shards with sharp edges. If you have access to a simple optical microscope at 200x, you can tell the difference in thirty seconds.
7. Workholding: The Overlooked Failure Point
I’ve seen more quartz parts scrapped by bad fixturing than by bad toolpaths. Here’s why: quartz is stiff but not ductile. When you clamp it, it deflects elastically. After machining, when you release the clamp, the quartz springs back to its original shape – and the machined surface cracks at the points of highest stress concentration.
Rule one: never use mechanical clamps directly on quartz. Point loads are a guaranteed path to edge chipping. Use vacuum chucks that distribute the holding force evenly across the entire back surface. For a 200mm viewport, you want a minimum of six vacuum zones so you can stage the hold-down force progressively.
Rule two: always use a sacrificial backing plate. This is non-negotiable for through-hole viewports. When your tool exits the bottom of a hole, the unsupported quartz edge tends to break out – that’s “exit chipping,” and it’s the most common location for cracks in finished parts.
Mount the quartz workpiece onto a backing plate made of the same material (or a compatible glass/ceramic) using UV-curable adhesive or wax. The backing material supports the edge at the exact moment of tool exit. The tool cuts through the quartz and then continues slightly into the backing plate. The quartz edge comes out clean; the backing plate takes the exit damage.We replace the backing plate after every 10–15 parts. It’s a consumable, and it’s cheap compared to scrapping a $2,000 quartz blank.
8. Coolant: It’s Not Just About Lubrication
With quartz, coolant serves three roles: lubrication, chip evacuation, and temperature control. If you neglect any one of these, you’ll get cracks.
Lubrication: The tool-chip interface generates enormous friction at low feed rates. If you’re not reducing that friction, you’re generating heat – and heat differentials across the quartz surface create tensile stress, which triggers cracks even without mechanical overload. Use a synthetic coolant with high EP (extreme pressure) additives. Standard water-soluble oils are fine for aluminum, but they don’t have the film strength for quartz.
Chip evacuation: Quartz chips are abrasive. If they stay in the cut zone, they get trapped between the tool and the surface, and they act like rolling abrasives that scratch the finish and can initiate micro-cracks. Flood coolant at 300+ PSI, directed through the spindle, is the minimum. For RUM, the ultrasonic vibration helps pump coolant into the gap, but you still need high pressure to flush the chips out.
Temperature control: Quartz has a low thermal conductivity – about 1.4 W/m·K, which is lower than most ceramics. That means heat stays where it’s generated. A localized hot spot of 50°C above ambient creates enough differential expansion to cause thermal cracking. Use a coolant temperature controller to keep the fluid at a stable 20–22°C, and consider using through-spindle coolant with a jet directed specifically at the tool engagement zone.
What not to use: chlorinated or sulfurized coolants that leave residues. Semiconductor cleanrooms have strict ionic contamination limits. If your coolant leaves sodium, potassium, or chlorine residues on the viewport, that part will fail the outgoing cleanliness inspection, regardless of how good the surface finish is.
9. Post-Processing: Two Steps That Catch the Cracks You Missed
Even with all the precautions above, some subsurface damage may remain. It might be only 2–3 µm deep, but for a vacuum viewport, 2–3 µm is enough to be a failure initiation site under cyclic loading.
We use a two-step post-processing sequence for every semiconductor quartz viewport that leaves our facility.
Step 1: Chemical etching in dilute HF (hydrofluoric acid). This is a controlled process – 5–10% HF solution, 2–5 minutes at room temperature, with continuous agitation. What this does is blunt the sharp tips of any micro-cracks that remain below the surface. A sharp crack tip creates a high stress concentration factor (KtKt). By rounding that tip through chemical dissolution, you effectively arrest the crack. It can’t propagate because the stress required to extend it has gone up by a factor of 2–3.
Step 2: Chemical-mechanical polishing (CMP) with a cerium oxide slurry. This is the step that takes the surface from “good” to “optical-grade.” CMP removes the last 0.5–1.0 µm of material, eliminating the chemically etched surface and leaving a finished surface with Ra<0.2 nmRa<0.2nm. At that level, the surface is flat to within a few atomic layers, and the subsurface is completely free of any detectable damage.
Does every quartz viewport need CMP? No. For many applications – thick windows, low-pressure chambers – a good ductile-machined finish with HF etching is sufficient. But for plasma etch chambers with optical emission spectroscopy (OES) windows, CMP is required because the surface roughness affects the optical signal.
10. Putting It All Together: What Actually Works in Production
Here’s the reality check. You can read all the theory in the world, but if your shop floor can’t execute consistently, you’re going to scrap parts.
Here’s the minimum checklist for successful quartz viewport machining:
·Machine: Rigid 5-axis or 3+2 axis CNC with hydrostatic spindles. Minimum spindle speed 15,000 RPM for finishing. Loop stiffness above 200 N/μm200N/μm.
·Tooling: Resin-bonded diamond wheels, 1200–2000 mesh for finishing. Truncated grits. Replace or re-true after every 20 parts.
·Parameters: Depth of cut below 5 µm for finishing. Feed rate 20–50 mm/min. Use RUM if available – it’s not a luxury, it’s a productivity enabler.
·Workholding: Vacuum chuck with at least six zones. Always use a sacrificial backing plate for through-holes.
·Coolant: High-pressure (300+ PSI) synthetic fluid at controlled temperature. No chlorinated additives.
·Post-processing: HF etch (dilute) + CMP (cerium oxide) for optical-grade surfaces. At minimum, do the HF etch to blunt any crack tips.
·Inspection: Don’t rely on visual inspection alone. Use a profilometer for surface roughness and a transmission microscope for subsurface crack detection. If you can’t inspect it, you can’t guarantee it.
11. Final Word: The Real Cost of Cutting Corners
I’ll be honest with you. Machining quartz correctly is slow. It’s expensive. The tooling wears fast, the cycle times are long, and you’ll scrap some parts while dialing in the process.
But I’ve also seen the alternative. I’ve seen fabs shut down for three days because a viewport cracked during a PM and dumped atmosphere into a $15 million etch tool. I’ve seen a single cracked window cause more damage than the entire cost of setting up a proper quartz machining cell.
At Runsom Precision, we’ve built our process around what we’ve learned from those failures. We don’t cut corners on tooling, we don’t rush the finishing passes, and we inspect every part like it’s going into a tool that’s running 24/7 for the next three years – because it is.
If you’re sourcing quartz viewports for semiconductor tools, the lowest bidder is rarely the cheapest. The cheapest is the one that works reliably, every time, without fail.
That’s what we deliver.
Got a quartz part that’s giving you trouble? Send us the drawing. We’ll look at the geometry, the tolerance callouts, and the surface finish requirements. We’ll tell you if it’s feasible, and we’ll recommend the process that gets you there without the surprises.
Contact us: sales17@runsom.com | Response within 24 hours