Anodizing: The 720 Rule, the 50/50 Split, and the 0.013 mm That Kills a 6g Thread
A 240-piece batch of 6061-T6 standoffs went out for 25 µm black Type II anodize. The drawing said "ANODIZE PER MIL-PRF-8625, TYPE II, CLASS 2, BLACK." Nothing about threads. The parts came back the colour of wet slate, cosmetically perfect, and every M6 screw galled a turn and a half in. Assembly stopped. The rework quote — strip, re-mask, re-anodize, re-inspect — was twice the original finishing cost, with a five-day slip.
The number that caused it is 0.013 mm. That is the radial clearance in an M6×1 6g/6H thread at maximum material. Anodizing a 25 µm coating adds about 12.5 µm per surface, and because a 60° thread flank turns radial growth into axial displacement, the pitch diameter moves by roughly four times the per-surface buildup. The coating asks for 0.050 mm of pitch-diameter room. The standard gives it 0.026 mm. The parts were never going to fit.
This is not a finishing defect. It is arithmetic that should have been done at the drawing stage.
1. The coating is not deposited. It is eaten.
Anodizing is a conversion process. The oxide is grown from the substrate, not laid on top of it, which is why it cannot peel or flake in the way a paint or a plating can. At the anode, water is oxidised and aluminium is consumed:
At the cathode, hydrogen evolves; the net cell reaction is simply 2Al + 3H_2O \rightarrow Al_2O_3 + 3H_2 plus heat. Every ampere that enters the tank leaves as heat, and roughly 0.34 g of aluminium per ampere-hour is converted.
The film is not a single uniform slab. It has two layers in series:
- A barrier layer at the metal interface, compact, insulating, and proportional to the forming voltage at 1.0–1.4 nm/V. At 18 V that is 18–25 nm; at 60 V, 60–84 nm. The electric field across it is on the order of 1 V/nm, which is high enough to move ions but not high enough to break down the film.
- A porous layer above it, a close-packed array of hexagonal cells with a central pore. In sulfuric acid at 20 °C, pore diameter is 10–20 nm, cell spacing is about 2.5 nm/V (50 nm at 20 V), and pore density lands around 10^{10}–10^{11} pores per cm². Porosity is roughly 10%.
That porous layer is the useful part, and it is also the part that makes anodizing a chemical balancing act. Oxide forms at the metal interface at a rate set by current; the acid simultaneously dissolves the oxide at the pore base at a rate set by temperature and field. At 20 °C, dissolution is fast enough to keep the pores open, and the practical ceiling for Type II is about 25 µm — push further and the film stops thickening because it is being eaten as fast as it grows. At 0–5 °C, dissolution nearly stops, the pores stay tighter, and 50–100 µm of Type III hardcoat becomes possible. The cold bath is not there to "freeze" the coating. It is there to suppress the dissolution that would otherwise cap the thickness.
2. Faraday, the 720 rule, and the 312 rule
The thickness is, to first order, a coulomb counter. Faraday's law gives the mass converted:
For aluminium, M = 26.98 g/mol and z = 3:
One gram of aluminium becomes 1.89 g of Al_2O_3, so the oxide mass rate is 0.635 g per ampere-hour. At a dense-alumina density of 3.1 g/cm³, that is a thickness rate of:
From there, the industry shortcut falls out. The 720 rule states that 720 ampere-minutes per square foot produce one mil (25.4 µm) of coating:
The metric restatement is the 312 rule. An exact conversion gives 3.05 rather than 3.12 — a 2% difference, well inside the ±10% that a real tank delivers.
Check the two against each other: 720 A·min/ft² is 12 A·h/ft², or 129 A·h/m² for 25.4 µm, which is 0.197 µm per A·h/m². Faraday with dense alumina says 0.205. The gap is porosity, sulfate incorporation, and hydration — the film is not fully dense Al_2O_3, and its effective density is closer to 2.5–2.9 g/cm³. The 720 rule is calibrated, not derived, and that is precisely why it survives on shop floors.
Current density (ASF) · Current density (A/dm²) · Minutes per 1 mil (25.4 µm)
12 · 1.29 · 60
18 · 1.94 · 40
24 · 2.58 · 30
30 · 3.23 · 24
40 · 4.31 · 18
Worked examples, which is where most quoting errors happen:
- 20 µm Type II at 12 ASF — 0.787 mil × 720 / 12 = 47 minutes of dwell.
- 12 µm black Type II at 18 ASF — 0.472 × 720 / 18 = 19 minutes.
- 50 µm Type III at 28 ASF — 1.969 × 720 / 28 = 51 minutes, which matches the NIST handbook's 25 min per 25 µm at 2.5–3.6 A/dm².
A ramped rectifier does part of the work before it reaches full current. A linear ramp from zero takes half the charge of an equivalent dwell, so:
A 10-minute ramp to 30 ASF is worth 5 minutes at full current. If the recipe says 48 minutes at 30 ASF, dwell 43 minutes after a 10-minute ramp. Ignore this and thick parts come out 5–10% heavy — which matters on a thread, and does not matter on a cosmetic cover.
3. Heat is the real constraint
The rectifier load is the product of current density, area, and voltage:
For Type II at 12 ASF, 129 A/m² at 18 V is 2.3 kW per square metre of work. For Type III at 28 ASF, 301 A/m² at an average 50 V during the ramp is 15 kW per square metre. All of that energy ends up in a bath that must be held at 20 °C, or at 0–5 °C. A chiller with a coefficient of performance of 2.5 needs about 6 kW of electrical input per square metre of hardcoat load, on top of pumping and agitation.
That single number explains most of the commercial difference between the two processes. Type III is not expensive because the chemicals are exotic — it is the same sulfuric acid — it is expensive because you are paying to remove 15 kW/m² from a tank held near freezing, for an hour, while the rectifier runs at 60–100 V.
Two failure modes live here. Burning is local current density above roughly 40 ASF, usually from a bad rack contact or a part shadowing another: the film grows powdery, white-to-brown, and non-coherent, and the area has to be stripped. Soft burn is the same mechanism at lower current, from insufficient agitation; the coating looks fine and fails abrasion. Cathode area should be at least equal to the work, and 1.5–2× is common, because a starved cathode simply raises the voltage and the heat.
4. The dimensional arithmetic
This is the part that scraps parts.
Anodizing consumes aluminium and grows oxide. The coating thickness t is split between penetration into the original surface and buildup above it. The shop rule is 50/50, and for Type III hardcoat that is close to correct: a 50 µm coating penetrates 25 µm and builds 25 µm. For Type II the split is closer to two-thirds in, one-third out. Either way, the number that matters for tolerances is the buildup per surface:
Because a dimension is bounded by two coated surfaces, the net change is one full coating thickness:
- An external dimension grows by about t: a Ø25.000 shaft becomes Ø25.025 after a 25 µm coating.
- An internal dimension shrinks by about t: a Ø20.000 bore becomes Ø19.975.
Put those against normal machining tolerances and the problem is immediate. A Ø25.000 ±0.010 shaft has a 0.020 mm total band; a 25 µm coating consumes 0.025 mm, more than the entire band, before the lathe has contributed any error at all. A Ø20 H7 bore (+0.021/0) closed by 0.025 mm lands at Ø19.975 — 0.025 mm under its lower limit, and the bearing will not start.
The thread case
Threads are worse than cylinders because the flank angle multiplies the error. For a 60° thread, a normal growth of \delta on each flank displaces the pitch line by 2\delta, and two flanks double it:
This is the same 4× rule used for electroplating: 0.001" of buildup moves the pitch diameter 0.004". It is not a rule of thumb pulled from air; it is the geometry of a 30° flank.
Now put ISO 965 numbers on it. For M6×1:
- External 6g pitch diameter: 5.324 / 5.212 mm
- Internal 6H pitch diameter: 5.350 / 5.500 mm
- Diametral clearance at maximum material: 5.350 − 5.324 = 0.026 mm
- Radial clearance per flank: 0.013 mm
That 0.026 mm is the entire budget ISO 965 leaves for coating. It was sized for the thin electroplating the standard calls "standard" — not for a conversion coating that grows into the part.
Coating t (µm) · Buildup \delta (µm) · \Delta d_2 = 4\delta (µm) · Share of 0.026 mm allowance
5 · 2.5 · 10 · 38%
10 · 5 · 20 · 77%
15 · 7.5 · 30 · 115%
20 · 10 · 40 · 154%
25 · 12.5 · 50 · 192%
50 · 25 · 100 · 385%
At 10 µm the thread is technically still inside the allowance, with 23% left for the actual machining error, the coating thickness spread (±10% is normal, and not less than ±2.5 µm), and the plating on the mating screw. At 15 µm it is already over. At 25 µm it is nearly double.
If both halves are anodized — an aluminium housing and an aluminium lid — the closures add: external +4\delta and internal -4\delta, so the pair loses 8\delta. A 25 µm coating on both parts closes 0.100 mm against a 0.026 mm allowance, and the assembly is 0.074 mm in interference before any torque is applied. This is the classic "it fit before finishing" failure.
The fixes, in order of preference:
- Mask the threads. Plugs and tapes are cheap; rework is not. Masking a thread costs seconds per part; stripping and re-anodizing costs 2–3× the original finish.
- Pre-machine to the finished condition. For external threads, cut undersize by 2\delta on the pitch diameter; for internal threads, cut oversize by 2\delta. Then anodize to the final size. This is what "machine to" dimensions on the drawing are for.
- Chase the thread after anodizing. A tap or die removes the brittle oxide from the flanks. It works, but it leaves bare aluminium in the thread and a cosmetic witness mark; it is a repair, not a process.
- Use a thinner coating. 8–10 µm of Type II will survive a 6g/6H fit if the machining is centred. It will not survive a 6H/6g pair that is already at maximum material, and it gives up most of the wear and salt-spray performance.
The same logic applies to any close fit: dowel holes, bearing bores, spigots, H7/h6 shafts, and press fits. Anodize is a dimensioning process, not a cosmetic one.
5. Design rules that survive the tank
The coating grows perpendicular to the surface, and current density is not uniform. Both facts produce geometry-specific rules:
- Edges and corners. Coating grows normal to the surface, so a sharp external corner receives a thin, rounded film and a sharp internal corner is current-shadowed. NASA's process specification calls for radii of 0.03", 0.06", and 0.09" for 25, 50, and 75 µm coatings respectively (0.76, 1.52, 2.29 mm). Most shops will accept less, but expect thin coating and exposed substrate at the edge.
- Blind holes and deep slots. Coating thickness falls with depth-to-diameter ratio because the field and the electrolyte exchange both degrade. A Ø4 × 20 mm hole will not see the same film as the outer face. If the hole is a bearing bore, mask it; if it is a clearance hole, add 2t to the drill.
- Rack marks. Every part needs an electrical contact, and that contact leaves a bare spot. Put it on a non-cosmetic face and say so on the drawing. Titanium racks are used for hardcoat because aluminium racks anodize and lose contact.
- Masking. Threads, bearing bores, ground strap pads, EMI gasket lands, dowel holes, and any surface that will be adhesively bonded or welded. Masking is labour, and on small parts it is the dominant cost.
- Laser marking. Mark after anodizing if you want a crisp, contrasting mark. Marking before anodizing means the film covers the mark; on 25 µm of dark dye it will disappear.
- Bending and welding. Bend and weld first, anodize last. Anodized film is brittle — a tight bend after coating will craze and crack along the outside radius. Welding over anodize contaminates the puddle with oxide and produces porosity.
- Electrical bonding. Anodize is an insulator. If the housing is a chassis ground, mask the pad or use a masked stainless insert; a star washer biting through 25 µm of oxide is not a bond.
6. Fatigue: the debit nobody puts on the drawing
Anodic coatings are brittle, microcracked, and firmly adhered. Under cyclic load, cracks initiate in the film and propagate into the substrate, and the fatigue limit drops. The magnitude is not small.
For 6082 aluminium tested in alternating tension–compression at R = −1, high-cycle fatigue strength fell from 130 MPa bare to:
Condition · Fatigue strength (MPa) · Change
Bare 6082 · 130 · —
Type II, 10 µm · 120 · −8%
Type II, 20 µm · 85 · −35%
Type II, 40 µm · 65 · −50%
Type III, 20 µm · 50 · −62%
Type III, 40 µm · 65 · −50%
Type III, 70 µm · 50 · −62%
The non-monotonic hardcoat row is not a typo; the 20 and 70 µm conditions both landed at the same low fatigue strength in that study, which is a reminder that coating morphology, not just thickness, drives crack initiation. Other data sets are consistent in direction: 7075-T6 loses about 40% of fatigue strength at 23 µm, and 2024-T351 loses 26–35% at conventional Type II thicknesses.
The design consequences are concrete:
- A rotating bending shaft, a connecting rod, or a pressure-cycle vessel should not carry 50 µm of hardcoat as a wear surface without a fatigue assessment.
- Chromic acid (Type I) and boric-sulfuric (Type IC, Boeing BAC 5632) anodize are less damaging than sulfuric; they are the aerospace choices for fatigue-critical parts, not because they are thinner by definition but because the film morphology is less crack-initiating.
- Shot peening before anodizing restores much of the loss. In one 7075 study, peening at 12A intensity regained the fatigue strength that anodizing had removed. If the part is fatigue-critical and must be hardcoated, peen first, then anodize — and test coupons, because the interaction is alloy-specific.
7. Sealing, dye, and what salt spray actually measures
The porous layer is why anodize takes dye and why it corrodes if left open. A dye molecule is 1–2 nm; the pore is 10–20 nm; the dye adsorbs on the pore walls during immersion. Sealing then closes the pore mouths, either by hydrothermal conversion of the oxide to boehmite in boiling deionised water, or by precipitation of nickel hydroxide from nickel acetate.
- Hot water seal: >95 °C, roughly 2 min per µm of coating — 40 minutes for 20 µm. Cheap, effective, and it slightly softens the film.
- Nickel acetate: 5 g/L at 85–90 °C, faster, colourless, and the standard for dyed work because it does not leach the dye. Nickel is a REACH concern in some markets; cobalt acetate is a substitute.
- Cold seal: nickel-fluoride chemistry at 25–30 °C, used where the part cannot take boiling water or where energy cost matters. Slower to reach full corrosion performance.
Sealing trades wear resistance for corrosion resistance. The Aluminum Anodizers Council is explicit: where extreme abrasion resistance is required, do not seal, because sealing softens the film. That is why a hardcoated wear surface is often left unsealed or impregnated with PTFE, while a decorative black cover is always sealed.
Corrosion performance is usually quoted as neutral salt spray hours per ASTM B117, and the numbers are thickness- and seal-dependent:
- Type II, 10 µm, sealed: about 1,000 hours to the AAMA 611 Class II requirement.
- Type II, 18 µm, sealed: 3,000 hours for AAMA 611 Class I, the architectural exterior benchmark.
- Type III, 50 µm, sealed: commonly 2,000+ hours, with 1,000 hours as the routine acceptance floor.
- Unsealed Type III: excellent wear, poor corrosion. Do not specify unsealed for a marine or wash-down environment.
Two tests keep the process honest: ASTM B244 eddy-current thickness on the finished part, and ASTM B136 dye-stain resistance for seal quality. A coating that is the right thickness but poorly sealed will pass a micrometer and fail a dye stain.
8. Alloy selection changes the answer
Anodizing is not alloy-neutral. The coating grows from the aluminium matrix; alloying elements that do not anodize (silicon, copper, iron) remain as second phases, and they change colour, coating ratio, and defect sensitivity.
Alloy · Anodize behaviour · Typical use
6061-T6 · Excellent. Clear to light grey, good coating ratio, takes dye well. Hardcoat 400–500 HV. · Machined brackets, housings, fixtures
6063 · Excellent, brightest clear finish, best for architectural extrusion · Frames, trim, heatsinks
6082 · Very good, slightly darker than 6061, common in Europe · Structural machined parts
2024-T351 · Moderate. Yellow-grey, lower coating ratio, copper-rich phases can burn at high current density · Aerospace structures, fatigue-critical
7075-T6 · Moderate. Bronze-grey to dark grey, lower coating ratio, sensitive to burning · High-strength airframe parts
5083 · Good corrosion resistance but dull grey; hard to get a bright decorative finish · Marine, tanks, welded fabrications
3003 · Good, light grey, soft substrate · Sheet metal covers, panels
A380 die casting · Poor cosmetic result. Silicon particles do not anodize; finish is dark grey to black and uneven · Cast housings — specify paint instead
The practical rule: if the part is cosmetic, use 6061 or 6063. If it is structural and fatigue-critical, use 2024/7075 with chromic or boric-sulfuric anodize, and peen before coating. If it is a high-silicon casting, do not expect a decorative anodized finish; the silicon phase will show through as a speckled dark surface.
9. Electrical and thermal, in numbers
Anodic film is a dielectric. Measured breakdown voltage is thickness-dependent and alloy-dependent, and it is lower than the handbook value for bulk sintered alumina because the film is porous and hydrated. The Aluminum Anodizers Council's technical data gives approximately:
- 8 µm: ~400 V breakdown
- 26 µm: ~800 V breakdown
That is 31–50 V/µm, not the 1.6–6.4 V/µm sometimes quoted from bulk-alumina handbooks — the handbook figure is for dense sintered ceramic, not a porous anodic film. Copper- and silicon-bearing alloys reduce breakdown further, because second-phase particles create local field enhancement.
For design, treat the measured value as a qualification number and derate it. A 25 µm coating on 6061 is a reasonable 500 V insulator; it is not a 2 kV insulator, and it is not a substitute for a proper creepage distance in a mains enclosure.
Thermally, the film is a poor conductor compared with the alloy. Bulk anodic alumina sits in the 0.2–1 W/m·K range, against 167 W/m·K for 6061-T6, so a coating adds an interfacial resistance:
At a heat flux of 5 W/cm² that is a 2 K penalty — acceptable on a heatsink, and worth checking on a 100 W IGBT module. The film is electrically insulating and thermally merely inconvenient, which is why anodized heatsinks work at all.
10. The cost model
Anodizing is priced by area, but the cost is driven by handling. A first-principles model for a job shop:
with c_A the area cost (energy, chemicals, DI water, effluent), n the parts per rack, and C_{min} the minimum batch charge. At ₹8/kWh and ₹400/h loaded labour:
Bucket · Type II, 20 µm · Type III, 50 µm
Rectifier energy · ₹13/m² · ₹90/m²
Chiller energy · ₹5/m² · ₹120/m²
Chemicals, dye, seal, DI · ₹40–80/m² · ₹50–90/m²
Labour, racking and jigging · ₹80–150/m² · ₹150–250/m²
Masking (per part, 1–3 min) · ₹7–20/part · ₹10–30/part
Estimated cost · ₹300–600/m² · ₹700–1,200/m²
Market rates for small machined parts in India run ₹150–350 per square foot (₹1,600–3,800/m²), with minimum charges of ₹200–450 per part for Type II and ₹450+ for Type III. Architectural extrusion is a different business entirely — ₹15–40/kg — because the area-to-mass ratio is enormous and the parts are racked by the hundred. The gap between the ₹300–600/m² cost and the ₹1,600–3,800/m² price is not margin alone: it is racking labour, QA, effluent treatment, scrap allowance, and the fact that a job shop quoting 20 brackets cannot amortise a 40-minute cycle the way an extruder can.
The line item that surprises engineers is masking. On a 30 g bracket with two M4 threads and a Ø8 H7 bore, masking can be 2–3 minutes of hand labour — ₹13–20 — on a part whose anodizing might otherwise cost ₹30. That is why "just mask it" is not free, and why the better answer is often to pre-machine the thread and skip the mask.
11. What to put on the purchase order
A finish specification that cannot be argued with:
ANODIZE PER MIL-PRF-8625, TYPE II, CLASS 2 (BLACK)
COATING THICKNESS 20 µm ± 3 µm, MEASURED PER ASTM B244
SEAL: HOT DI WATER, 40 min MIN AT 96 °C MIN
SEAL QUALITY: DYE STAIN PER ASTM B136, NO STAIN
MASK: ALL THREADS, Ø8 H7 BORES, GROUND STRAP PAD
RACK ON NON-COSMETIC FACE "B" ONLY
NO ANODIZE ON DATUM A MATING FACE
PRE-MACHINE THREADS TO FINISHED PITCH DIAMETER
Four lines do most of the work: type and class, thickness with tolerance, seal method, and mask list. Adding "pre-machine threads to finished pitch diameter" moves the dimensional responsibility to the machinist, where it belongs, instead of leaving it to a tank operator who never sees the assembly.
The arithmetic, once more
The 240 standoffs were not a bad batch. They were a correct batch of parts made to a drawing that did not account for 12.5 µm of growth per surface, a 4× pitch-diameter multiplier, and a 0.026 mm allowance that was already spoken for.
Anodizing rewards the same discipline as any other process: know the growth rate, know where the coating goes, and put the critical features either outside the tank or inside the tolerance. The 720 rule tells you the time. The 50/50 split tells you the size. The thread table tells you whether the assembly will go together. Do all three at the drawing stage, and the finish stops being a surprise.