Choosing a special screw material: strength, heat, magnetism and chemicals — in numbers

Standard stainless is not strong. A4-70 (SUS316L) reaches 700 N/mm² tensile, and the A2-70 grade the same — which is why so many stainless designs are oversized in diameter to make up for it.

NBK’s high-strength stainless, SNSX, is made from SUS316L HiMo: reduced carbon, more chromium, nickel and molybdenum. It comes in class 8.8 (800 N/mm² tensile, 640 N/mm² proof) and class 10.9 (1,000 and 900). It stays non-magnetic, and it keeps a usable proof stress hot: 448 N/mm² at 600 °C for the 8.8, and 765 N/mm² for the 10.9.

The point is not the number, it is what the number buys. NBK gives the worked example: swapping to high-strength stainless let a flange go from M6 to M4 bolts at the same clamping force, and the flange itself shrank from ⌀80 mm to ⌀75 mm and from 10 mm to 7.5 mm thick. Strength per screw becomes space and weight saved in the machine.

Austenitic stainless is often called non-magnetic, and for a fridge door it is. For an electron beam it is not: cold working during thread rolling raises the permeability, and the A2 grade NBK measures comes out at µ 1.4 — enough to deflect the beam in an SEM or an ion implanter, or to become a projectile near an MRI magnet.

Titanium and phosphor bronze are in another class entirely: amplitude permeability 1.0001, with a magnetic flux density measured at 0 T where the A2 stainless reads 5 × 10⁻⁵ T. NBK measured both with an F.W. Bell 5080 gauss/tesla meter in DC mode, at 5 mm from the sample.

And remember the tool. NBK sells titanium hex keys (SKHTG) and a CFRTP adjustable wrench (SKMWCP-200) precisely because, near an MRI magnet, “tools are strongly attracted to the magnetic field, posing a danger”. A non-magnetic screw fitted with a steel Allen key in an MRI room is a solved problem replaced by a worse one.

NBK does not publish a universal corrosion table, and it is right not to: resistance depends on concentration, temperature and time. What it does publish is an ordinal ranking, which is exactly what you need to build a shortlist.

Metals, most resistant first: MAT21 → Hastelloy C-22 equiv. → Hastelloy C-276 equiv. → Monel 400 equiv. → titanium → Inconel 600 equiv. → duplex stainless → SUS316L HiMo → SUS310S.

Plastics, most resistant first: PTFE → PFA → PVDF → H-PVC → PP → PEEK.

Two data points are worth keeping. In sulfuric acid at 50 °C, SUS304 corrodes at 1.08 mm/year in 3 % concentration, where SUS316/316L reads 0 — and the SNSX HiMo grade holds at 0 in both 3 % and 10 %. And alumina ceramic (Al₂O₃ 99.5 %) shrugs off boiling 98 % sulfuric acid, boiling 70 % nitric acid and boiling 35 % hydrochloric acid — but is heavily corroded by potassium fluoride at 90 °C. There is no universally resistant material; there is only the right one for your chemistry.

A screw does not hold with torque. It holds with the axial force that torque produces — and how much of your torque becomes axial force depends almost entirely on friction in the thread.

NBK measured it on the same screw: an M5 socket head cap screw, class A2-70, tightened to 5 N·m. Untreated, it develops 2.3 kN of clamp load. The same screw with low-temperature black chrome develops 9.2 kN. Four times the preload from the same wrench setting, because the friction coefficient collapsed.

This cuts both ways, which is why it is a warning as much as a feature. If you specify a coated screw and keep the torque figure from the uncoated one, you may take the joint past yield. NBK gives the design rule: for class A2-70 the 0.2 % proof stress is 450 N/mm², and “it is recommended to use at an axial force at which stress at tightening is 70 % or less” of it. Change the surface, recalculate the torque.

The mechanism is worth knowing, too: “The main cause of seizing is frictional heat on the thread peak surface generated during tightening.” Lower the friction and you lower the heat — which is why the same treatments that raise preload also prevent galling.