Engineering Materials: A Practical Reference

Carbon steel is iron with 0.05–1.0% carbon and very little else; low-alloy steel adds a few percent of chromium, molybdenum, or nickel so that the hardening reaches the middle of a thick section. Two facts set everything else on this sheet. Young's modulus is 200–210 GPa regardless of grade, alloy content, or heat treatment, so no steel is stiffer than any other steel and a stiffness problem can only be solved with a bigger section. And carbon plus a quench-and-temper cycle moves yield strength from 250 MPa in mild A36 to roughly 1,600 MPa in 4340, inside one alloy family, using furnaces every heat treater already owns. The AISI-SAE numbering encodes the recipe: 10xx is plain carbon with the last two digits giving hundredths of a percent carbon, 41xx adds chromium and molybdenum, 43xx adds nickel on top. At roughly $1/kg for hot-rolled product it is the cheapest structural metal there is, which makes it the baseline every other material here has to beat.

Steel is cheap, stiff, tough, weldable, available in every form from 0.2 mm strip to 300 mm plate, and recycled at full value. It also has a fatigue endurance limit at roughly half its tensile strength, so a steel part can be designed for infinite life instead of for a finite life plus an inspection interval, and aluminum, magnesium, and polymers cannot do that. The weaknesses are density and rust. At 7.85 g/cm³ its specific strength is about 50 kN·m/kg, the lowest of any structural metal on the sheet, and bare steel corrodes at a rate you have to design around, so paint, zinc, or oil is not optional outdoors. The failure mode that catches people out is brittle fracture: carbon steel is body-centered cubic and has a ductile-to-brittle transition, typically somewhere between -50 °C and +20 °C depending on grade, section thickness, and steelmaking cleanliness, so a plate that is ductile in a room-temperature tensile test can fracture with no warning on a cold morning. Weldability also falls as strength rises. Above a carbon equivalent of roughly 0.45 you need preheat and controlled cooling or the heat-affected zone cracks.

Start here and make everything else justify a premium. Pick carbon steel whenever the part is limited by strength, stiffness, or cost and the environment is dry, painted, or oiled: frames, brackets, shafts, gears, fasteners, pressure vessels, buildings, and anything welded. If it needs to be harder or stronger, change the heat treatment before you change the material, because 4140 hardened to 45 HRC costs nearly the same per kilogram as 1018. Leave the family only when one specific constraint binds. If that constraint is corrosion, price paint or hot-dip galvanizing first and move to stainless (004) only when a coating cannot be inspected or maintained, since stainless costs three to six times as much per kilogram. If it is weight, aluminum (006) has the same stiffness per kilogram at a third the density, so it helps only where you can afford a thicker section. If the part is a stamped body panel, the answer is advanced high-strength sheet (002) rather than mild steel. And if service runs below about -20 °C, specify Charpy impact energy at the service temperature rather than assuming the room-temperature datasheet applies.

Young's modulus 200–210 GPa across every grade and heat treatment · density 7.85 g/cm³, so specific strength about 50 kN·m/kg for mild steel · yield strength 250 MPa for A36 up to roughly 1,600 MPa for quenched-and-tempered 4340 · fatigue endurance limit near half the tensile strength · ductile-to-brittle transition typically between -50 °C and +20 °C · carbon equivalent above roughly 0.45 needs preheat to weld · US hot-rolled coil around $1,150 per short ton in early August 2026, about $1.25/kg, against a longer-run $0.7–1.2/kg range.

Three mechanisms account for most steel failures. General corrosion thins the section from the outside, which is easy to see and easy to prevent, so it rarely causes surprises; pitting and crevice attack under a failing coating or inside a lap joint do, because the lost section is hidden until something leaks or buckles. Brittle fracture is the one that takes down structures. Below the transition temperature a crack runs the length of a plate in a fraction of a second with essentially no plastic deformation and no prior warning, which is what split the Liberty ships and why pressure-vessel and arctic codes are written around notch toughness rather than tensile strength. The third is hydrogen embrittlement in high-strength grades. Above roughly 1,200 MPa, hydrogen absorbed during electroplating, pickling, or cathodic protection migrates to the most highly stressed region and cracks the part hours to days after assembly under a load it has already carried, which is why plated high-strength fasteners are baked within hours of plating. The warning signs worth looking for are rust staining bleeding out of faying surfaces, cracks starting at weld toes and at any abrupt change of section, and beach marks on a fracture face, which point straight back to where the fatigue crack started.

A36 and A992 hot-rolled sections in buildings and bridges; API 5L line pipe; 1018 and 1045 bar for shafting; 4140 and 4340 in axles, crankshafts, and landing-gear forgings; 8620 case-carburized gears; A516 pressure-vessel plate; and rebar, which is the largest single tonnage of all. The producers are among the largest industrial companies on earth: China Baowu, ArcelorMittal, Nippon Steel (which completed its acquisition of U.S. Steel in 2025), HBIS, POSCO, Nucor, and Steel Dynamics.

Steel is made two ways and the split drives the economics. Integrated mills reduce iron ore with coking coal in a blast furnace and refine it in a basic oxygen furnace, which is capital-heavy, has to run continuously, and emits roughly 1.8–2.3 tonnes of CO2 per tonne of steel. Electric arc furnaces melt scrap, cost far less to build, can follow the power price, and emit a fraction of that; they are about 70% of US production and a much smaller share of China's. Price is set regionally by scrap and iron-ore cost plus trade policy, and policy currently dominates: US Section 232 duties went to 50% in 2025 and were extended to a long list of derivative products, which is why US hot-rolled coil sat near $1,150 per short ton in mid-2026 while world prices ran well below that. China produces about 52% of world crude steel, and its export volumes set the floor everywhere else. Three practical consequences for anyone building on steel: the material is a small share of a finished part's cost, so machining and coating decisions matter more; scrap carries real resale value at end of life, which composites do not; and the green-steel projects (HYBRIT, Stegra, and several hydrogen direct-reduction plants behind them) are selling a carbon attribute at a premium rather than a better material, so their customers are the ones with scope-3 targets rather than the ones with engineering problems.

25 minUnderstanding Steels and Heat TreatmentThe Efficient Engineer · 500k+ views

8 minSteel Numbering SystemJason Lonon Toolmaker · 100k+ views

6 minHow carbon content affects mechanical properties in steel microstructuresTaylor Sparks · 10k+ views

Steel material properties for structural design and performance (SteelConstruction.info) · Ferrous Alloys, Part IB lecture notes (University of Cambridge)

Advanced high-strength steel is automotive sheet whose strength comes from a controlled mixture of phases rather than from carbon content, which is what lets it stay formable and weldable at strengths that plain high-carbon steel cannot reach without becoming glass. Dual-phase grades put 10–20% hard martensite islands in a soft ferrite matrix, so the sheet yields low, work-hardens fast, and ends up strong in the finished stamping. TRIP grades hold retained austenite that transforms to martensite as the part deforms, buying elongation at a given strength. Press-hardened steel takes a different route: a boron-alloyed 22MnB5 blank is austenitized near 900 °C, formed in a water-cooled die, and quenched in the die in a few seconds, coming out fully martensitic at about 1,500 MPa where the incoming coil was around 500. This is the reason the volume car body is still steel. Mild sheet is roughly 270 MPa; the same body in AHSS runs 600–1,500 MPa for a premium of a few tenths of a dollar per kilogram, and it stamps on presses the plant already owns. The forming operations themselves are on the manufacturing-processes sheet.

The case for AHSS is entirely economic. It removes mass from a vehicle at roughly $2–4 per kilogram saved against $8–15 for aluminum and well above $30 for carbon fiber, it drops into an existing stamping and spot-welding plant, and it recycles into the same scrap stream as everything else. The physical cost is ductility. Total elongation falls from around 40% for mild sheet to 5–8% at 1,500 MPa, so the strength-versus-elongation "banana" curve is the real design chart: first-generation grades (DP, TRIP, CP, martensitic) sit along the lower band, second-generation TWIP sits high but needs 15–25% manganese and has never gone volume, and third-generation grades built on quenching-and-partitioning or medium manganese are filling the middle at 980–1,180 MPa with enough elongation to cold-form. Higher yield strength also means more elastic springback, so dies get compensated by simulation and recut, and abrasive high-strength sheet wears tooling far faster than mild steel does. Above roughly 1,200 MPa hydrogen embrittlement becomes a live design constraint rather than a footnote.

Use AHSS when the part is a formed sheet-metal structure carrying crash or fatigue load and mass matters: rockers, B-pillars, roof rails, bumper beams, door intrusion beams, seat structures, and battery-enclosure frames. The decision inside the family is mostly about how the part fails. If it has to absorb energy by crushing, pick a dual-phase or TRIP grade in the 600–980 MPa range, because the structure needs to fold rather than fracture. If it has to not deform at all, which is the passenger cell, pick press-hardened 22MnB5 and accept that the part is formed hot in a dedicated line. If you need 980 MPa or more but cannot justify a hot-stamping line, look at third-generation cold-formable grades first. Skip AHSS for outer skin panels, where dent resistance and surface finish, not strength, decide, and for anything where the tooling is amortized over a few thousand parts and the die wear will not pay back. And if the mass target survives all of that, the next step up is an aluminum body (006), which is what Ford did on the F-150, at roughly four times the cost per kilogram saved.

Mild sheet around 270 MPa against press-hardened 22MnB5 at about 1,500 MPa · premium of roughly $0.1–0.4/kg over mild sheet · about $2–4 per kilogram of mass removed, against $8–15 for aluminum · total elongation falls from roughly 40% in mild sheet to 5–8% at 1,500 MPa · specific strength 75–125 kN·m/kg for common dual-phase grades and about 190 for press-hardened · hydrogen embrittlement becomes a design constraint above roughly 1,200 MPa · press-hardening blanks austenitized near 900 °C and die-quenched in a few seconds.

The failures that matter are all delayed or invisible, which is what makes them expensive. Hydrogen-assisted delayed cracking is the headline one: in grades above roughly 1,200 MPa, hydrogen picked up in pickling, coating, or service migrates to the highest-stress location and opens a crack hours or days after the part was made, with no change in load, usually starting at a sheared edge or a hole where residual stress is highest. Liquid metal embrittlement is the second: when zinc-coated high-strength steel is resistance spot welded, molten zinc penetrates the austenite grain boundaries in the heat-affected zone and leaves cracks that are hard to see from outside the joint, which is why weld schedules and coating choices get requalified for every new grade. Third is edge cracking. A punched or sheared edge carries a damaged, heavily work-hardened layer, so a flange splits during forming even though the coil passed its tensile test; the number that predicts it is hole expansion ratio, not elongation, and dual-phase grades score badly on it because cracks run along the hard-soft phase boundaries. And the most common failure of all is dimensional rather than structural: the stamping springs back out of tolerance and will not assemble.

ArcelorMittal's Usibor 1500 press-hardened steel and Ductibor for tailored blanks, SSAB's Docol range, thyssenkrupp, POSCO, and Nippon Steel supply most of the coil. Gestamp and Benteler are the large independent hot-stamping tier-ones, and TWB Company and ArcelorMittal Tailored Blanks laser-weld blanks of different grade and thickness into one stamping so the soft zone lands where the part is meant to fold. Volvo's boron-steel safety cage and Honda's ACE body structure are the well-documented early production applications, and WorldAutoSteel's FutureSteelVehicle and Steel E-Motive programs are the industry's public design studies for how far the material can go.

AHSS is the steel industry's answer to losing body-in-white volume to aluminum, and it has largely worked: a mill sells a commodity product at a premium of a few tenths of a dollar per kilogram, using the same casters and rolling mills, by controlling cooling on the run-out table and the annealing line. The capital moves downstream instead. A hot-stamping line needs a roller-hearth furnace, a transfer system, and water-cooled dies, so press-hardened parts come from tier-one suppliers with dedicated cells rather than from the OEM's own press shop, and that is why third-generation cold-formable grades are commercially interesting even where they are slightly heavier: they avoid the line. Trade policy sits on top of all of it, with US Section 232 duties at 50% since 2025 covering steel and a long list of derivative parts. If you are evaluating anyone selling into this market, the durable position is the qualification package and the forming know-how, not the composition, since grade chemistries are published and the hard part is proving a die and a weld schedule that make a million parts.

26 minThe mad metallurgy of advanced high strength steel in modern cars | Auto Expert John cadoganAuto Expert John Cadogan · 50k+ views

1 minHot Stamping in eight secondsANDRITZ Schuler · 50k+ views

7 minDual Phase SteelMetallurgical Engineering · 1k+ views

AHSS Application Guidelines (WorldAutoSteel) · Multi-Scale Microstructural Tailoring and Associated Properties of Press-Hardened Steels: A Review (Materials, via PMC)

Tool steel is high-carbon, high-alloy steel bought for what it does after heat treatment: 58–65 HRC, held against an abrasive workpiece for millions of cycles. It is the material behind almost every mold, die, punch, and cutting tool on the manufacturing-processes sheet. Hardness comes from martensite, and wear resistance comes from hard carbide particles suspended in it, which is why the grade letters are worth knowing as a trade-off map rather than as a catalog: W is plain water-hardening carbon, O oil-hardening and A air-hardening for general cold work, D high-carbon high-chromium for abrasive cold work, S shock-resisting for punches and chisels, H hot work for anything that touches hot metal, and M and T the molybdenum and tungsten high-speed steels for cutting tools. More carbide means more wear resistance and less toughness, in a straight line, and that single trade is most of what grade selection is. Hot-work grades add a second requirement, hot hardness: H13 is tempered near 550 °C so it does not soften when the die face sees 500–600 °C every shot.

Nothing else gives you that combination of hardness, bulk toughness, machinability in the annealed state, and price. You cut the tool at about 20 HRC with ordinary carbide inserts, then harden it, which is the whole reason tool steel beat cemented carbide for most tooling: carbide is harder and lasts longer but has to be ground or EDM'd to shape, costs several times as much per kilogram, and chips instead of deforming. The weaknesses show up in heat treatment and at the edges. Hardening changes size by a few tenths of a percent and distorts the part, so tools are rough-machined, stress-relieved, hardened, and then finish-ground, and you leave roughly 0.2–0.5 mm of grind stock per surface for that. At working hardness the material has almost no ductility, so it fails by chipping and cracking rather than by bending, and every sharp internal corner, EDM recast layer, and grinding burn is a crack starter. Powder-metallurgy grades such as CPM 10V fix the wear-versus-toughness trade partway by dispersing very fine vanadium carbides instead of coarse segregated ones, at $30–50/kg against $3–10 for conventional bar.

Pick the grade from what will destroy the tool, not from a hardness target. If the tool wears out against an abrasive workpiece (glass-filled plastic, high-strength sheet, ceramics), go up in carbide content: D2, then a PM grade like CPM 10V or Vanadis 4 Extra, and accept the price and the grinding cost. If the tool chips or cracks first, go down in carbide and up in toughness: S7 or A2, or drop the working hardness a few points, which usually buys more life than a harder grade would. If the tool touches hot metal, you are in hot-work grades and H13 is the default for die casting and aluminum extrusion. For plastic molds, start at P20 pre-hard around 30 HRC if the resin is unfilled and the run is short, and move to a hardened and polished grade if the resin is filled or the surface has to stay optical. For tooling that will make a few thousand parts, 4140 pre-hard is often enough and skips heat treatment entirely. And if wear resistance still is not enough after all of that, the next steps are a coating (PVD titanium aluminum nitride, on the surface-modification part of the manufacturing sheet) or cemented carbide, not another steel.

58–65 HRC after hardening, roughly 20 HRC annealed for machining · H13 tempered near 550 °C with die surfaces at 500–600 °C · around 2,000 MPa tensile at working hardness, about 250 kN·m/kg, though tool steels are chosen on hardness and used mostly in compression · vanadium carbide around 2,800 HV against roughly 1,800 HV for chromium carbide, which is why PM grades outwear D2 · conventional bar $3–10/kg, PM grades such as CPM 10V $30–50/kg · leave 0.2–0.5 mm of grind stock per surface for heat-treat distortion.

Hot-work dies fail by heat checking. The die surface heats and quenches on every shot, so it cycles in tension against a cold interior, and after tens of thousands of cycles it develops a craze network of fine thermal-fatigue cracks that then prints onto every part. You can see it coming: the pattern starts at the hottest, sharpest features and coarsens, and once it does the die gets welded and re-cut or scrapped. Die-casting dies add soldering, where molten aluminum chemically attacks the steel and welds the casting into the cavity, which is why die-casting alloys carry deliberately high iron. Cold-work tools fail the other way, by chipping and gross fracture, usually because the grade or the hardness was pushed past what the toughness supported, and almost always starting at a sharp corner, an EDM recast layer, or a grinding burn rather than in the middle of a face. The third mechanism is heat treatment itself: quench cracks in sections that change abruptly, and slow dimensional drift weeks later as retained austenite transforms, which is what deep-freeze and double-tempering cycles exist to prevent. If a tool that measured right at inspection is out of size a month later, that is the mechanism.

P20 and 1.2738 in injection mold bases, H13 in aluminum die-casting and extrusion dies, D2 and A2 in blanking and forming dies, S7 in punches and shear blades, M2 and M42 in drills, taps, and broaches, and CPM 10V or Uddeholm Vanadis 4 Extra in high-volume blanking of advanced high-strength sheet. Printed tooling is a real niche now, mostly H13 and 1.2709 maraging powder for mold inserts with conformal cooling channels that cannot be drilled. The specialty mills are voestalpine BÖHLER Uddeholm, Daido Steel, Proterial, Erasteel, Crucible Industries, and Swiss Steel.

Tool steel is a small-tonnage, high-value corner of the steel industry, made by specialty melt shops using electroslag or vacuum-arc remelting for cleanliness and gas atomization for the powder grades, which is why it prices like an alloy rather than like a commodity. The number that decides purchases is not $/kg, because the steel is usually well under a fifth of what the finished tool costs; machining, EDM, grinding, polishing, heat treatment, and try-out dominate. That arithmetic is why a grade at five times the price of D2 sells: if it doubles the interval between regrinds on a tool that costs $80,000 to build and stops a production line when it is out, the material premium disappears into the rounding. It also means the durable businesses here are heat treaters, toolmakers, and coaters rather than the mills, and that a new tool material has to prove tool life in a real press before anyone will switch, which takes a year of running.

9 min7 TYPES OF TOOL STEEL AND THEIR APPLICATIONSJames Sword Engineering · 10k+ views

4 minThe Four Types of Steel (Part 5: Tool Steel) | Metal SupermarketsMetal Supermarkets · 100k+ views

4 minH13 vs P20 Tool Steel Comparison | Die & Mould Material Selection ExplainedCNC CAD CAM ACADEMY of SIGMA YOUTH ENGINEERS · 1k+ views

Die Science: Tool steel made simple (The Fabricator) · Effect of Different Surface Conditions on Toughness of Vanadis 6 Cold Work Die Steel—A Review (Materials, via PMC)

Stainless steel is any steel with at least 10.5% chromium, which is the threshold where the surface forms a chromium-rich oxide film a few nanometres thick that reforms within seconds of being scratched. That self-healing film is the entire product; everything else about the family follows from keeping it intact and from what the alloy has to give up to do so. Nickel is the second lever. Add enough of it and the structure stays face-centered-cubic austenite at room temperature, which is non-magnetic, extremely ductile, has no ductile-to-brittle transition, and cannot be hardened by heat treatment. Leave it out and you get ferritic stainless, cheap and magnetic; add carbon instead and you get martensitic stainless that hardens like tool steel; split the difference and you get duplex, roughly half austenite and half ferrite. The chromium costs about three to six times carbon steel's price per kilogram, and the honest framing of this entry is what that premium buys and where it does not help, because stainless is not corrosion-proof and the ways it fails are specific and predictable.

The 300-series austenitics are among the most useful engineering materials in existence: they resist most of what a food plant, a chemical plant, or a building facade throws at them, they weld and form well, they stay tough to cryogenic temperatures where carbon steel shatters, they hold their oxidation resistance in air to about 870 °C, and they are hygienic and cleanable. The trade-offs are all on the shop floor and in the specifics of the chemistry. Annealed 304 yields at only about 215 MPa, so a stainless part is often heavier than the carbon-steel one it replaces. It work-hardens under a dull tool, so machining rates are roughly half of carbon steel's and interrupted cuts glaze the surface. Its thermal conductivity is about 16 W/m·K against roughly 50 for carbon steel and its expansion is higher, so welds distort more and heat concentrates locally. Austenitic threads gall and seize against each other. And the failure that ends careers is chloride stress-corrosion cracking, which attacks 304 and 316 specifically, in hot chloride environments, with no warning and no metal loss to inspect.

Default to 304 when the part will get wet, handled, or cleaned and a coating is impractical or cannot be maintained: food and pharmaceutical equipment, sinks, fasteners, architectural trim, tanks, and general hardware. Move to 316 when chlorides are present at all (marine air, de-icing salt, chlorinated water, brine). If chlorides, heat, and tensile stress are all present at once, do not go to 316 and hope; go to duplex 2205, a ferritic grade, or a nickel alloy, because that combination is exactly what cracks austenitics. If you need real strength out of stainless, use 17-4 PH, which machines soft and ages to about 1,310 MPa with very little distortion. If you need hardness and edge retention, use martensitic 420 or 440C and accept that it is the least corrosion-resistant of the family. If cost and nickel-price volatility matter more than weldability in thick section, ferritic 439 or 444 does most of what 304 does without the nickel. And if the environment is seawater at temperature, or a chloride process stream where even duplex is marginal, the answer is titanium (009), not another stainless.

10.5% chromium minimum to form the passive film · 304 annealed around 505 MPa tensile and 215 MPa yield, 17-4 PH in the H900 condition about 1,310 MPa · specific strength roughly 65 kN·m/kg for annealed 304 up to about 170 for 17-4 PH · duplex 2205 at 450 MPa yield with a pitting resistance number near 35, against about 24 for 316 and 18 for 304 · thermal conductivity around 16 W/m·K, roughly a third of carbon steel's · $3–6/kg against about $1.25/kg for carbon steel · usable in air to about 870 °C for 304.

Chloride stress-corrosion cracking is the signature failure and it needs three things at once: an austenitic grade, chlorides, and tensile stress, usually above roughly 50–60 °C. The cracks branch through the grains, carry almost no corrosion product, and remove no measurable metal, so a vessel or a hanger that passes every visual and thickness inspection splits without warning. It has killed people: collapsing 316 stainless suspension rods brought down swimming-pool ceilings in Uster in 1985 and Steenwijk in 2001, in warm chlorinated air that would look harmless on a corrosion chart. The second mechanism is crevice and pitting corrosion, which happens wherever oxygen cannot reach the surface to rebuild the film, so it starts under gaskets, deposits, and lapped joints rather than on open surfaces, and pitting resistance number ranks grades against it. The third is sensitization: hold an unstabilized grade between about 425 °C and 815 °C, which any weld does somewhere in its heat-affected zone, and chromium carbides precipitate on the grain boundaries and strip the adjacent metal of chromium, so corrosion later runs in a neat band a few millimeters from the weld. Low-carbon L grades and titanium- or niobium-stabilized 321 and 347 exist for that reason. The everyday warning signs are rust bleeding from crevices and fasteners rather than from open faces, and brown heat tint left beside a weld, which is a chromium-depleted layer and needs pickling or passivation rather than a wire brush.

304 and 316L piping, tanks, and vessels across food, dairy, brewing, and pharmaceutical plants; 316L in surgical implants, marine fittings, and architectural exteriors; 17-4 PH in aerospace fittings and as a standard powder for laser powder-bed fusion; duplex 2205 in chemical tankers, desalination plants, and flue-gas desulfurization; 409 in automotive exhaust systems and 430 in appliance trim; 440C in bearings and knife blades. The mills are Outokumpu, Acerinox (which owns North American Stainless), Aperam, POSCO, Nippon Steel Stainless, and, at much larger volume, China's Tsingshan, TISCO, and Baosteel.

Stainless is sold as a base price plus an alloy surcharge that tracks the exchange prices of nickel, chromium, and molybdenum, which is why a 316 quote moves month to month while the mill's margin does not. Nickel is the swing factor: LME nickel settled around $16,750 per tonne in early August 2026, and the difference between a 300-series and a ferritic grade is mostly whether you are exposed to that number. The structural story of the past decade is Indonesian and Chinese capacity. Tsingshan's integrated nickel-pig-iron-to-stainless operations in Indonesia changed the industry's cost floor, China now makes well over half the world's stainless, and Western mills have consolidated or closed capacity in response, with trade cases and safeguard measures on top. Two practical implications: if you are designing to a budget, specifying a ferritic grade removes most of the price volatility rather than just a few dollars a kilogram, and if you are sourcing, the scrap-based supply chain means stainless carries 60% or more recycled content and retains real value at end of life, unlike a coated carbon-steel assembly.

21 minThe Problem with Stainless SteelThe Efficient Engineer · 1m+ views

8 minStainless Steel Grades ExplainedAalcoMetals · 100k+ views

9 minStainless Steel Types - What is the diffrence between Austenitic, Martensitic, Ferritic, & DuplexMachining Doctor · 10k+ views

Introduction to stainless steels (worldstainless) · High-Performance Stainless Steels (Nickel Institute)

Cast iron is iron with 2–4% carbon and 1–3% silicon, which is more carbon than the metal can hold in solution, so the excess comes out as free graphite while it solidifies. The shape that graphite takes decides everything. Flakes give gray iron, nodules give ductile iron, short thick worms give compacted graphite iron, and suppressing graphite entirely gives white iron, which is used only where abrasion matters. Ductile iron is not a different alloy so much as a different treatment: about 0.03–0.05% residual magnesium added to the ladle changes the graphite from flakes to spheres and turns a brittle material into one with real elongation. The composition sits near the iron-carbon eutectic, so it melts at 1,150–1,200 °C rather than the roughly 1,500 °C steel needs, which means cheaper furnaces, thinner walls, and better fill. Casting is the only way it is made, and the manufacturing-processes sheet covers those processes; here the point is that no other material gets you a complex, stiff, damped, machinable shape for this little money.

Cast iron is the cheapest route to a complicated part that has to hold its shape. Gray iron's graphite flakes damp vibration roughly an order of magnitude better than steel, which is why machine-tool bases have been gray iron for 150 years and still are, and they break the chip and lubricate the cut, so gray iron machines faster than almost anything else. Compressive strength runs three to four times tensile, which suits the frames, housings, and bases it is used for. The weaknesses follow from the same graphite. Gray iron has essentially no tensile ductility, so it does not bend before it breaks. Its modulus is 80–145 GPa depending on class, well below steel's 200–210, so the stiffness advantage comes from being able to afford a deep ribbed section rather than from the alloy. It is section-sensitive, meaning a thick wall cools slowly, grows coarser graphite, and ends up weaker than the separately cast test bar that qualified the heat. And it is not practically weldable, since the heat-affected zone forms martensite and cracks, so a repair is a specialty operation rather than a shop-floor fix.

Choose cast iron when the part is geometrically complex, needs stiffness and damping more than tensile strength, and the volume will pay for a pattern: machine-tool bases and columns, engine blocks and heads, housings, pump and valve bodies, brake rotors, counterweights, and pipe. Pick gray iron when damping, machinability, and thermal conductivity matter and the loading is compressive: bases, blocks, rotors, and manifolds. Pick ductile iron whenever the part sees tension, impact, or a bolted joint that someone will over-torque: crankshafts, suspension knuckles, pipe, wind-turbine hubs. Pick compacted graphite iron only if you are chasing higher cylinder pressure in a diesel block and can pay for the process control, because it is markedly harder to machine than gray iron. Do not use cast iron for a one-off or a low-volume part, where a steel weldment gives you more strength with no tooling; do not use it where the part will be welded into an assembly; and if the driver is weight rather than cost, look at cast aluminum (007), which will be roughly a third the density and several times the price per kilogram.

Gray iron 150–400 MPa tensile with essentially no elongation · ductile iron 400–900 MPa, with elongation of about 18% at the low-strength end falling to roughly 2% at the top · gray iron modulus 80–145 GPa and ductile iron about 170 GPa, against 200–210 GPa for steel · melts at 1,150–1,200 °C against roughly 1,500 °C for steel · compressive strength three to four times tensile in gray iron · liquid metal under $1/kg, with tooling and machining setting what a casting actually costs.

Gray iron fails by brittle fracture, and the mechanism is built into the microstructure: every graphite flake is effectively a pre-existing crack, so the material has no yield plateau and no plastic reserve. Over-torque a bolt into a gray-iron boss, drop a casting on a corner, or load a base in tension and it breaks cleanly with no bending to warn you. In hot service the failure is growth and heat crazing. Repeated heating lets oxygen run in along the graphite flakes, the casting permanently swells by a percent or more, and the surface develops a network of thermal-fatigue cracks, which is why brake rotors and exhaust manifolds are consumable parts. Ductile iron adds a failure that is specific to how it is made: magnesium fades from the treated melt over roughly ten to twenty minutes, so castings poured late from a ladle can have degenerate graphite and a fraction of the expected elongation while looking identical from outside, which is why foundries run nodularity checks per pour rather than per heat. And section sensitivity means the part often breaks at the place the drawing called thickest, because that section cooled slowest and has the coarsest graphite. Warning signs are a dull gray fracture face with no shear lip, and casting defects, shrinkage porosity, cold shuts, and sand inclusions, which turn up on a machined face after most of the value has been added.

Machine-tool bases, columns, and lathe beds from Haas, DMG Mori, and Mazak; engine blocks, heads, and crankshafts; brake rotors and drums; ductile iron water and sewer pipe from McWane, U.S. Pipe, and Saint-Gobain PAM; manhole covers and hydrants; wind-turbine hubs and main frames, which are among the largest ductile iron castings made at 10–40 tonnes apiece; high-chromium white iron mill liners and slurry-pump parts. Compacted graphite iron is the specialty case, with SinterCast process control used in the Ford 6.7L Power Stroke and Audi V6 TDI blocks, where the higher strength lets the block take more cylinder pressure at the same size. The common specifications to know are ASTM A48 for gray iron and ASTM A536 for ductile.

The metal itself is close to free by engineering-material standards: scrap and pig iron melted in a cupola or an induction furnace, with in-house returns remelted, so the foundry's real input cost is the delta between what it buys and what it ships. Everything else is the foundry. Pattern and core tooling, molding line time, cleaning, and machining set the price, which is why cast iron only makes sense above a few hundred parts and why quoted casting prices are several times the metal price. The structural trend is consolidation: gray iron foundries in the US and Europe have been closing for two decades on energy, labor, and air-permitting costs, and China and India have taken the volume, so lead times and minimum order quantities have stretched even though the material is cheap. Two things follow for anyone building on it. Casting yield, meaning shipped weight divided by poured weight, and machining stock are where a foundry's margin actually lives, so design changes that improve either are worth more than a price negotiation. And the competitive threat to iron castings is not another iron; it is aluminum castings taking weight-driven parts and fabricated steel taking low-volume ones.

8 minTypes Of Cast Iron And Their Differences | An Overview.James Sword Engineering · 50k+ views

7 minHow to Identify Cast Iron, Ductile or Grey IronFireball Tool · 100k+ views

4 minWhat is Ductile Iron? || THORS Ductile Iron Fundamentals Course PreviewTHORS eLearning Solutions · 10k+ views

Ductile Iron Data (Ductile Iron Society) · Cast Irons (University of Cambridge)

Wrought aluminum is the default lightweight structural metal: 2.70 g/cm³, a modulus of 69 GPa, no ductile-to-brittle transition, and an oxide film that protects it in ordinary atmospheres without any coating. The alloy series tells you the strengthening mechanism and therefore what the metal can do. 1xxx is nearly pure and gets used for conductors; 3xxx and 5xxx are strengthened by solid solution and cold work, so they are formable and weldable but capped around 200–300 MPa; 2xxx, 6xxx, and 7xxx are precipitation hardened, which is where the strength lives, and the T-temper suffix records the heat treatment that produced it. The misconception worth correcting up front is stiffness. Aluminum's modulus per unit density is about 26 GPa per g/cm³, and steel's is also about 26, as are titanium's and magnesium's. Swapping steel for aluminum buys nothing on deflection unless you can make the section thicker, which is exactly why aluminum wins in beams and panels, where a thicker section is worth more than the density penalty, and does not win in a tension member.

The real advantages are density, extrudability, and corrosion behavior. Aluminum extrudes into complex hollow profiles that would be a welded assembly in steel, which routinely saves more cost than the material premium costs, and it conducts heat at 237 W/m·K for pure metal and around 167 for 6061-T6, so it doubles as a heat sink. Recycling takes roughly 5% of the energy of primary smelting, so scrap has real value. The weaknesses are all temperature, joining, and fatigue. Aluminum loses about half its strength by 200 °C and creeps well below that, so it is out of the running for anything hot. Welding softens 6xxx back toward its T4 condition, and 2xxx and 7xxx are not reliably fusion weldable at all, so high-strength aluminum structures are riveted, bolted, bonded, or friction stir welded. It expands at 23 ppm/K, about twice steel, which matters at every mixed-material joint. And there is no fatigue endurance limit: the S-N curve keeps falling, so an aluminum structure is designed to a finite life with an inspection interval rather than to infinite life.

Reach for aluminum when mass matters and the part runs below roughly 150 °C: chassis and frames, housings, heat sinks, marine hulls and superstructures, aircraft skins and machined structure, bicycle and vehicle frames, and anything extruded. Pick the alloy from the process, not the datasheet. If it will be extruded or welded, use 6061 or 6063; if it will be formed and welded and see salt water, use 5052 or 5083; if it will be machined from solid and strength governs, use 7075 or 7050; if it is aircraft skin loaded in fatigue, 2024-T3 has better damage tolerance than 7075. Avoid aluminum if the part is stiffness-limited and cannot get thicker, since you gain nothing per kilogram; if it runs hot, where titanium (009) or steel take over; or if it will be fastened directly to steel or carbon fiber in a wet environment without isolation, because aluminum will be the anode. And check whether the mass target justifies the price: aluminum takes weight out of a vehicle body at roughly $8–15 per kilogram saved against $2–4 for advanced high-strength steel (002).

2.70 g/cm³ and 69 GPa modulus, so about 26 GPa per g/cm³, the same as steel, titanium, and magnesium · 6061-T6 at 310 MPa tensile and 276 MPa yield, about 115 kN·m/kg · 7075-T6 at 572 MPa and 2.81 g/cm³, about 205 kN·m/kg · thermal conductivity 237 W/m·K for pure aluminum and roughly 167 for 6061-T6 · thermal expansion 23 ppm/K, about twice steel's · LME aluminum near $3,280 per tonne in early August 2026, with common mill product roughly $4–8/kg · no fatigue endurance limit, so structures get a finite life and an inspection interval.

Fatigue is the one that defines how aluminum structures are managed. Because there is no endurance limit, every cyclically loaded part is accumulating damage at any stress, so the design deliverable is a life and an inspection program rather than a safe stress. The Aloha Airlines 737 that lost most of its upper fuselage in 1988 is the reference case: fatigue cracks at rows of lap-joint rivet holes linked up into one long crack, and the industry's aging-aircraft inspection rules were rewritten around it. The second mechanism is stress-corrosion cracking in the 7xxx series, which needs sustained tension in the short-transverse direction, meaning through the thickness of a plate or across the grain flow of a forging, plus ordinary humid air. It shows up in thick machined parts and in interference-fit bushings and bolts, and it is the reason T73 and T7351 overaged tempers exist: they trade roughly 10–15% of strength for immunity. Related is exfoliation corrosion, where attack runs along the elongated grain boundaries of rolled plate and lifts the metal apart in leaves, jacking a joint open from inside. Finally, aluminum sits low in the galvanic series, so at any wet joint with steel or carbon fiber the aluminum dissolves; the warning sign is white powdery product bleeding out of a fastener line. In a welded 6xxx structure, expect the failure at the weld toe, because the heat-affected zone is the softest metal in the part.

6061-T6 in machined structure, frames, and fixtures; 6063 in architectural and heat-sink extrusion; 5052 and 5083 in boat hulls, tanks, and cryogenic vessels; 2024-T3 in aircraft skins; 7075 and 7050 in wing spars, ribs, and machined aerospace structure; 3104-H19 can bodies with 5182 ends, which is one of the highest-volume single products in metals. The Ford F-150 aluminum body from the 2015 model year is the best-documented mass-market switch, taking about 700 lb out of the truck and forcing self-piercing rivets, structural adhesive, and a rewritten collision-repair network with it. Primary producers are Alcoa, Rio Tinto, Norsk Hydro, Chalco, and Hongqiao; rolled and extruded product comes from Novelis, Constellium, Kaiser Aluminum, and a long tail of regional extruders.

Aluminum is mostly electricity: smelting takes roughly 13–15 kWh per kilogram, which is why smelters sit next to hydro power in Norway, Iceland, Quebec, and the Pacific Northwest, and why a smelter's viability tracks the power contract more than the metal price. Price is the LME figure plus a regional physical premium plus a fabrication charge, and the premium is where trade policy lands: US Section 232 duties went to 50% in 2025, which raised landed cost for US buyers without changing anything about the metal. China produces roughly 60% of the world's primary aluminum against a self-imposed capacity cap near 45 million tonnes, so incremental global supply increasingly has to come from recycling or from new smelters in places with cheap power. Recycling is the structurally important fact for anyone building a business here: secondary aluminum uses about 5% of the primary energy, scrap trades at a large fraction of primary value, and the constraint on using more of it is sorting and alloy contamination rather than availability. That is also why extrusion and rolling capacity is regional: profiles and coil are expensive to ship relative to their value, so the local converter, not the smelter, is usually who you are actually buying from.

5 minIntro to Wrought Aluminum Alloys - 7075 - 7050 - 6061 - 2024 - Guide to AluminumMichlin Metals Inc · 10k+ views

6 minIntroduction to Aluminum The Heat Treated Tempers - 6061-T6 - 7075-T7351- 2024-T4Michlin Metals Inc · 10k+ views

5 minWhich Aluminum Grade Should I Use | Metal SupermarketsMetal Supermarkets · 100k+ views

Teal Sheets: alloy designations and composition limits (The Aluminum Association) · Aluminium Automotive Manual: Materials – Microstructure and properties (European Aluminium)

Cast aluminum is a different alloy world from wrought aluminum, and most of the aluminum in a car is here rather than in rolled or extruded form. The controlling addition is silicon, at 7–12%, which puts the alloy near the aluminum-silicon eutectic, drops the melting range, and makes the metal fluid enough to fill a thin cavity. Two alloys cover most of the field. A356 is 7% silicon with a little magnesium, cast in sand or permanent molds, heat treatable to about 262 MPa with roughly 5% elongation, and weldable, which makes it the structural casting alloy. A380 and its Japanese equivalent ADC12 are the high-pressure die-casting alloys: 8.5% silicon plus 3.5% copper and, deliberately, up to 1.3% iron, because iron stops the casting from soldering itself to the steel die. That iron is why A380 reaches about 325 MPa but only around 3% elongation. The casting processes themselves are on the manufacturing-processes sheet; what matters here is that the process and the alloy are chosen together, and the process decides which properties you can actually have.

Casting buys geometry. A die casting comes out of the machine as a near-net shape with walls down to about 1.5–2 mm, bosses, ribs, and cored holes already in place, in a cycle measured in tens of seconds, and no other route gets that combination at that cost. The alloys keep aluminum's density and most of its thermal conductivity, and because they tolerate scrap chemistry they are made largely from recycled metal. The weaknesses come from how the metal solidifies. Aluminum shrinks roughly 5–7% by volume on freezing, so sand and permanent-mold castings need risers and a feeding path or they pull voids. High-pressure die casting fills so fast that it entrains air, and that trapped gas means a conventional die casting cannot be solution heat treated (it blisters) and cannot be reliably welded, which is why die-cast parts are used as-cast and joined mechanically. Porosity also blocks pressure-tight applications and ruins a decorative anodized finish. High-vacuum die casting and the structural alloys built for it (Silafont-36, Castasil-37, Aural) are the fix, and they cost more.

Cast aluminum when the part is complex, moderately loaded, and made in volume: housings, transmission cases, engine blocks and heads, brackets, wheels, heat sinks, and electronics enclosures. Choose the process first. If the part must be heat treated, welded, pressure-tight, or fatigue-critical, use A356 in a sand or permanent mold and accept the longer cycle and more machining. If it is a high-volume housing where as-cast strength is enough, use A380 or ADC12 in a die and let the die do the work. If it needs die-casting economics and structural properties both, you are into high-vacuum die casting with a structural alloy, and the price goes up accordingly. Zinc is the contrast worth checking before you commit: Zamak 3 casts at about 6.6 g/cm³ against aluminum's 2.7, so parts are heavier, but it pours at around 420 °C instead of 660–700 °C, which is gentle enough that a zinc die outlives an aluminum die by a large multiple, it holds walls down to 0.3–0.6 mm, and it plates directly to a bright finish. For a small, thin, decorative or precise part in high volume, zinc usually beats aluminum on total cost even though it loses on weight. If the part needs full wrought properties, machine it from bar (006) and pay for the chips.

A356-T6 at about 262 MPa and roughly 5% elongation in permanent mold · A380 and ADC12 die cast at about 325 MPa but only around 3% elongation, giving roughly 118 kN·m/kg · iron held at 0.7–1.3% in die-casting alloys for die release, at a direct cost in ductility · solidification shrinkage roughly 5–7% by volume · die-cast aluminum walls typically 1.5–2 mm against 0.3–0.6 mm for zinc · zinc at 6.6 g/cm³ poured near 420 °C, against 660–700 °C for aluminum · giga-press machines at 6,000–9,000 tonnes of clamping force.

Castings fail from their defects, not from their alloys, and porosity is the defect. Gas porosity comes from air entrained during a turbulent fill and from hydrogen dissolved in the melt; shrinkage porosity comes from a section that froze without a feed path. Either one intersecting a machined sealing face makes the part leak, which is usually discovered on a pressure test at the end of the line after all the machining value has been added, and either one below the surface is a fatigue crack starter, so a casting's fatigue strength is set by its largest defect rather than by its composition. That is why cast fatigue data scatters so widely and why radiographic or CT acceptance classes are written into casting drawings. Two more mechanisms are specific to the process. Trapped gas expands during a solution treatment and blisters the skin, which is the direct reason conventional die castings are not heat treated. And hot tearing opens a ragged crack at a restrained corner or boss while the metal is still semi-solid, so it looks like a casting flaw rather than a fracture. In service, the iron-bearing β-phase platelets in high-iron die-casting alloys behave like internal cracks and are most of the reason A380 breaks at 3% elongation, so a die-cast bracket that is overloaded snaps rather than bends.

A356-T6 low-pressure cast road wheels; 319 and A356 engine blocks and cylinder heads; A380 and ADC12 transmission cases, gearbox housings, brackets, and consumer-electronics frames; high-vacuum structural castings in shock towers and body nodes at Audi, Ford, and most European OEMs. The Tesla Model Y underbody giga-castings are the landmark case, replacing on the order of 70 stamped and welded parts with one casting off an IDRA press. Zinc's territory is a good contrast set: locks, hinges, connector shells, zippers, and plated hardware, mostly in Zamak 3 and 5. Casting suppliers include Nemak, Ryobi, and Georg Fischer, with machine builders Bühler and IDRA.

Cast alloys are the aluminum industry's recycling sink. Silicon-bearing casting compositions tolerate the mixed chemistry of post-consumer scrap that wrought alloys cannot use, so cast alloy prices sit below wrought and carry a much smaller energy and carbon footprint than primary metal. The cost of a casting is tooling plus cycle time plus scrap rate, in that order: a die represents a real capital commitment amortized over the program, cycle time sets the machine rate, and a few percent of scrap is the difference between a profitable job and a loss. Giga-casting pushed all three numbers up together. A 6,000–9,000 tonne press with its cell, robots, and trim tooling is a major installation, and the payoff is deleting assembly line and labor rather than saving metal, which also concentrates risk: one bad casting scraps far more value than one bad stamping, and a damaged structural casting can total a vehicle. The most useful spillover is alloy development. Because a 1.5 meter casting distorts in a solution furnace, giga-casting forced self-hardening alloys that skip heat treatment and tolerate higher recycled-scrap iron, and those alloys make ordinary castings cheaper and greener too.

3 minWhich Aluminum Alloy Should You Use? | 319 vs A356 vs A380 ExplainedBatesville Products, Inc. · 10k+ views

7 minHow to prevent/ manage POROSITY in PRESSURE DIE CASTING | Serious Engineering: Ep 29Star Rapid · 10k+ views

Aluminium Automotive Manual: Manufacturing – Casting methods (European Aluminium) · Microstructural Characteristics of High-Pressure Die Casting with High Strength-Ductility Synergy Properties: A Review (Materials, via PMC)

Magnesium is the lightest structural metal at 1.74 g/cm³, about two thirds of aluminum and a quarter of steel, and that is the whole reason anyone uses it. Its modulus is 45 GPa, which works out to about 26 GPa per g/cm³, the same specific stiffness as aluminum, steel, and titanium, so once again the gain comes from being allowed a thicker section in bending rather than from the metal being stiffer per kilogram. The crystal structure is hexagonal close-packed, which leaves too few slip systems for good room-temperature ductility, so magnesium is mostly die cast rather than formed, and sheet has to be worked at 200–300 °C. Alloy names read directly: the ASTM code gives the two main additions by letter and their rounded percentages, so AZ91D is 9% aluminum and 1% zinc in the high-purity D grade, and AM60B trades some strength for elongation. The aluminum in those alloys is also what limits them, because the magnesium-aluminum phase softens and lets the metal creep above roughly 120 °C.

Magnesium die casts better than aluminum in almost every respect: it holds walls of 1–1.5 mm routinely and down to about 0.6 mm, it can run on a hot-chamber machine, its lower heat content gives faster cycles and much longer die life, and it machines faster than any other structural metal. It damps vibration well and shields electromagnetic interference, which is why it ends up in laptop and camera bodies. The weaknesses are why it stays a niche despite being the lightest option. Creep limits standard alloys to about 120 °C, so anything near a powertrain needs a rare-earth or strontium alloy at a price premium. Magnesium is the most anodic structural metal, so every joint with a steel or aluminum fastener is a galvanic cell with the magnesium as the anode. Machining chips and grinding dust ignite, and water-based coolant makes it worse by generating hydrogen, so shops run dry or on mineral oil and keep Class D extinguishers. There is no fatigue endurance limit. And roughly 85–90% of primary magnesium comes from one country.

Use magnesium when mass is the binding constraint on a die-cast part running below about 120 °C, and when the part is large and thin enough that a thicker wall in a lighter metal actually wins: instrument-panel cross-car beams, steering-wheel armatures, seat frames, housings, and consumer-electronics enclosures. Pick AM60B when the part must absorb impact or deform, AZ91D when it is a rigid housing, and an AE, AJ, or MRI creep-resistant alloy when a bolted joint will run hot. Do not use it for a bolted joint on a hot component in a standard alloy, because you will lose clamp load rather than break anything. Do not use it in an unprotected joint with steel unless the fastener is isolated or coated. And check two alternatives before committing: aluminum die casting (007) is roughly 50% heavier for the same wall but has none of these problems and a supply chain that will not surprise you, and a 30% glass-filled engineering thermoplastic (015) is lighter still, cheaper, and immune to corrosion, though its modulus is roughly a fifth of magnesium's.

1.74 g/cm³, about two thirds of aluminum and a quarter of steel · modulus 45 GPa, so about 26 GPa per g/cm³, the same as aluminum, steel, and titanium · AZ91D die cast at roughly 230 MPa with about 3% elongation, around 130 kN·m/kg · standard aluminum-bearing alloys creep above roughly 120 °C · die-cast walls routinely 1–1.5 mm and down to about 0.6 mm · roughly 85–90% of primary magnesium comes from China.

The characteristic magnesium failure has no crack in it. A bolted magnesium housing running above about 120 °C creeps under the clamp load, the bolt tension relaxes over weeks to months, and the joint leaks, buzzes, or loosens while every part in it still measures within drawing. That is the reason a magnesium transmission case needs a creep-resistant alloy rather than a thicker flange, and the reason bolt-load-retention testing, not tensile testing, is what qualifies a magnesium part in a powertrain. The second mechanism is galvanic corrosion, and it is fast. Magnesium sits at the anodic end of the galvanic series, so a steel bolt through a magnesium boss in road salt dissolves a crater around the hole rather than corroding the bolt, and the fix is an isolating washer, an aluminum-coated fastener, or a coating that has to survive assembly. High-purity D-grade alloys, which cap iron, nickel, and copper at very low levels, improved general salt-spray corrosion by orders of magnitude and did nothing about the galvanic problem. The third is not a service failure but a shop one: chips and grinding fines burn, they burn hotter when someone puts water on them, and a magnesium fire in a chip conveyor is a plant-level event. Warning signs to look for are weeping or loosening fasteners on hot assemblies and mounded white corrosion product around fastener holes.

AM60B steering-wheel armatures and instrument-panel cross-car beams are the highest-volume automotive uses, since both are large, thin, and lightly loaded. AZ91D shows up in gearbox and accessory housings. The creep-resistant alloys have narrower but well-documented uses: AE44 in the engine cradle of the 2006 Corvette Z06, and AJ62 in BMW's magnesium-aluminum composite inline-six crankcase. Outside vehicles, magnesium bodies are standard on premium laptops, mirrorless camera bodies, and drone frames, where stiffness per unit weight and electromagnetic shielding both matter. Primary production is concentrated in China's Shaanxi cluster; US Magnesium in Utah is the only US primary producer, and Magontec is the main Western alloying and recycling specialist.

Almost all primary magnesium is made by the Pidgeon process, which reduces calcined dolomite with ferrosilicon in retorts, and it is both the most concentrated and the most carbon-intensive supply chain of any common structural metal. What that concentration means in practice was demonstrated in the autumn of 2021, when power curtailments in Shaanxi cut Chinese output and European spot prices more than quadrupled within weeks, and the shortage hit the aluminum industry harder than it hit die casters, because 5xxx and 6xxx aluminum alloys need magnesium as an alloying addition. That is the key structural fact for anyone specifying it: magnesium parts compete for supply with the entire aluminum industry, so the price is set by a market you are not really in. On top of that sit EU antidumping duties on Chinese magnesium and a US primary supply base of one plant that has already declared force majeure once. The engineering case for magnesium is genuine and narrow; the reason it has not grown is that the supply risk and the corrosion engineering both have to be paid for, and aluminum usually gets close enough.

15 minThe Dark Side of the World’s Lightest Structural MetalReal Engineering · 500k+ views

6 minMagnesium - lightweight materials of the futureSciencemovies · 10k+ views

Magnesium Alloys Overview (International Magnesium Association) · Chromate-Free Corrosion Protection Strategies for Magnesium Alloys—A Review: Part I (Materials, via PMC)

Titanium buys two things at once that no other metal offers together: about 215 kN·m/kg of specific strength and near-immunity to seawater, chlorides, and body fluids. Ti-6Al-4V, also called Grade 5, is roughly half of all titanium used, and it runs about 950 MPa at 4.43 g/cm³ with a modulus of 114 GPa, which is again about 26 GPa per g/cm³, so it is no stiffer per kilogram than aluminum or steel. The commercially pure grades 1 through 4 give up strength for formability and the best corrosion resistance in the family; the near-alpha engine alloys such as Ti-6242 push continuous service to 500–600 °C; the beta alloys such as Ti-5553 harden through thick sections for landing gear. Corrosion resistance comes from a titanium dioxide film that is stable in oxidizing conditions, which is why titanium shrugs off hot seawater and wet chlorine but is attacked by hydrofluoric acid and hot concentrated reducing acids. Everything about its cost traces back to the Kroll process, an unchanged 1940s batch reduction that turns titanium tetrachloride into sponge with magnesium over days per cycle.

On a properties basis titanium is close to ideal for structure that gets wet or hot: strong for its weight, non-magnetic, tough, fatigue-tolerant, biocompatible enough to bond to bone, and usable to about 400 °C in the workhorse alloy. The problem is that everything downstream of the sponge is expensive. Sponge runs $8–12/kg, mill bar $30–70/kg, and a finished aerospace part far above that, because titanium conducts heat at only about 7 W/m·K, roughly a seventh of carbon steel, so the heat from cutting stays in the tool rather than leaving in the chip and cutting speeds run about a fifth of steel's. Buy-to-fly ratios of 5–20:1 on machined structure mean most of the billet becomes chips. It also reacts with oxygen and nitrogen when hot, so melting is done under vacuum, welding needs full inert shielding front and back, and heating in air leaves a brittle oxygen-enriched surface layer. It galls against itself, it burns in oxygen-enriched atmospheres, and its low modulus means more springback in forming and more deflection under load than a steel part of the same section.

Specify titanium when a specific constraint makes cheaper metals fail, not because it is a premium material. The three that justify it are seawater or chloride service where stainless pits, temperatures between roughly 200 °C and 500 °C where aluminum is finished and steel is too heavy, and implants where nothing else has the biological record. If corrosion is the driver and strength is not, use commercially pure Grade 2 rather than Grade 5, because it is cheaper, more formable, and more corrosion-resistant. If strength and weight are the driver, use Ti-6Al-4V and design for the machining cost from the start, since a shape that would be trivial in aluminum can double the part price here. Do not use titanium in high-pressure oxygen service or where a rub can happen against a rotating part, because it burns. And run the honest comparison first: stainless is a quarter the price and heavier, aluminum is cheaper and cooler, and a carbon-fiber laminate (026) beats titanium on specific strength by a factor of nearly two if the load path is directional and you can qualify the part.

Ti-6Al-4V at about 950 MPa and 4.43 g/cm³, roughly 215 kN·m/kg · modulus 114 GPa, about 26 GPa per g/cm³, the same as steel and aluminum · thermal conductivity around 7 W/m·K, roughly a seventh of carbon steel's, so cutting speeds run about a fifth · Kroll sponge $8–12/kg and mill bar $30–70/kg · buy-to-fly ratios of 5–20:1 on machined aerospace structure, against roughly 1.5:1 for additive · continuous service to about 400 °C for Ti-6Al-4V and 500–600 °C for the near-alpha engine alloys · the Boeing 787 is roughly 15% titanium by weight.

The everyday failure is galling. Titanium slides against titanium by cold-welding rather than by rubbing, so threads seize and tear out and unlubricated sliding surfaces destroy themselves, which is why titanium fasteners are supplied with a solid-film lubricant or a dissimilar-metal nut. The failure that changed an industry is melt cleanliness. United 232 lost its tail engine and hydraulics in 1989 because a hard-alpha inclusion, a nitrogen-rich brittle particle left from melting, had seeded a fatigue crack in a Ti-6Al-4V fan disk, and modern triple-vacuum-arc-remelt practice, ultrasonic billet inspection, and full melt traceability exist because of that class of defect. Two surface mechanisms account for most of the rest: heating titanium in air above roughly 600 °C grows an oxygen-enriched alpha case that is brittle, cracks under load, and has to be chemically milled off after forging or heat treatment, and hydrogen picked up from pickling or cathodic protection precipitates hydrides that turn a ductile part brittle with no dimensional change. Above roughly 250–300 °C, chloride residue from a fingerprint or a cleaning agent can drive hot-salt stress-corrosion cracking. And titanium burns: a compressor rub or a chip pile can ignite, and it keeps burning in its own oxide, which is why engine designers restrict where titanium is allowed and why high-pressure oxygen systems exclude it entirely.

The Boeing 787 is about 15% titanium by weight, mostly in fittings that had to be compatible with the carbon-fiber structure around them, since carbon and aluminum corrode each other and titanium and carbon do not. Jet engine fan blades, compressor discs, and blades are titanium up to the temperature where nickel takes over. Ti-6Al-4V ELI is the standard for hip stems, knee components, spinal cages, and dental implants, and electron-beam and laser powder-bed printing of acetabular cups was the first serious production use of metal additive manufacturing. Titanium heat exchangers and piping run desalination plants and chemical service that would destroy stainless. Consumer use is small but visible, including Apple's Grade 5 iPhone frames. The supply base is short: VSMPO-AVISMA in Russia, TIMET and ATI in the US, Osaka Titanium and Toho Titanium in Japan, and Baoti in China, with IperionX among the startups trying to build a scrap-fed alternative to Kroll.

Titanium's price is a process problem, not a scarcity problem. The element is the ninth most abundant in the crust, but the Kroll process is a batch reduction that takes days, consumes magnesium and chlorine in a closed loop, and produces a sponge that then needs double or triple vacuum-arc remelting before it becomes a billet. Nobody has commercialized a continuous replacement despite decades of attempts, so the cost floor has barely moved. Two things have changed the economics at the margin. Additive manufacturing matters more for titanium than for any other metal, because when buy-to-fly goes from 10:1 to near 1.5:1 the material cost of a part falls by most of an order of magnitude, and titanium is expensive enough that the slow, costly printing process still wins. And sourcing has become a geopolitical question since 2022, because VSMPO-AVISMA was the single largest supplier of aerospace-grade product to Western airframers, which has pushed both requalification of alternative mills and investment in scrap-based routes. If you are assessing a titanium business, ask whether the value is in metal production, in the qualification package for a specific part family, or in near-net-shape processing, because the last two are where the margin has actually been.

9 minHow Is Titanium Made?History of Simple Things · 1m+ views

5 minTitanium: Kroll MethodInnovations in Manufacturing at ORNL - Archived · 100k+ views

5 minThe Problem With TitaniumSimple Things - Surprising Histories · 50k+ views

Titanium and its Alloys, Part II Course C9 notes (University of Cambridge) · Bandwidth Study on Energy Use and Potential Energy Savings Opportunities in U.S. Titanium Manufacturing (US Department of Energy)

Copper is bought for conductivity and then alloyed away from it. Pure C11000 sets the electrical benchmark at 100% IACS and carries about 400 W/m·K of heat at 8.96 g/cm³, and every addition that makes it stronger, more machinable, or more corrosion-resistant scatters electrons and takes conductivity with it. Free-cutting brass lands near 26% IACS, beryllium copper near 22%. That trade is the organizing idea of the whole family. Brass is copper and zinc, cheap and easy to machine and form. Bronze is copper and tin or aluminum, used where a bearing surface or seawater is involved. Cupronickel is copper and nickel, which resists seawater and biofouling well enough to run condenser tubing for decades. Beryllium copper is the outlier, precipitation hardening to about 1,300 MPa, which makes it the only copper alloy with real spring properties and the standard for non-sparking tools. Copper alloys are also unusually pleasant to make parts from, since they form, machine, cast, braze, and solder easily.

Nothing practical beats copper on conductivity except silver, which costs far too much, so for wire, busbar, windings, and heat spreaders it is effectively the reference material. It is ductile, joins by soldering or brazing without special equipment, resists corrosion in fresh water and most atmospheres, and is antimicrobial enough that copper touch surfaces are EPA-registered. The weaknesses are weight, price, and heat. At 8.96 g/cm³ copper is heavier than steel, and its specific strength in the annealed condition is around 25 kN·m/kg, so it is never a structural choice. It anneals and creeps at modest temperatures, which is why the alloys are used hot as electrodes and heat sinks but not as load-carrying parts above roughly 200–300 °C. Price has become a live design constraint rather than a background number, with LME copper settling near $14,200 per tonne in early August 2026. And two specific chemistries bite: brass cracks in ammonia and dezincifies in aggressive water, and beryllium copper carries an occupational exposure limit strict enough to change who is willing to machine it.

Use copper where the specification is conductance, and choose the alloy from the second requirement. If you need maximum conductivity, use C11000 and accept that it is soft. If the part will be machined in volume and conductivity is secondary, use free-cutting brass, or one of the silicon-brass replacements if it touches drinking water. If it is a spring, a connector contact, or a tool that must not spark, use beryllium copper and plan for the machining controls. If it runs in seawater, use 90/10 or 70/30 cupronickel, or aluminum bronze for propellers and bushings. If it is a plain bearing, leaded or tin bronze is still the default. The comparison that decides the big-volume cases is aluminum: at 61% IACS and 2.70 g/cm³ an aluminum conductor carries roughly twice the current per kilogram and about 60% per unit volume, so aluminum wins wherever there is room and the terminations are engineered for it (overhead lines, medium-voltage cable, busbar, some motor rotors), and copper wins wherever space is tight. And do not use copper alloys as structure; they are not chosen for strength and they will not surprise you by having any.

Pure copper at 100% IACS, about 400 W/m·K, and 8.96 g/cm³, with annealed specific strength around 25 kN·m/kg · free-cutting brass about 26% IACS and beryllium copper about 22% · beryllium copper aged to roughly 1,300 MPa · aluminum conductor at 61% IACS and 2.70 g/cm³, so roughly twice the conductance per kilogram and about 60% per unit volume · LME copper cash settlement near $14,200 per tonne in early August 2026, about $14/kg, with mill product carrying a fabrication premium on top · lead capped at a 0.25% weighted average of wetted surfaces in US drinking-water hardware.

Brass has two failure modes that both look like manufacturing defects and are not. Season cracking is stress-corrosion cracking in ammonia or amine environments, and it needs only the residual tension left by drawing or bending, so a formed brass part cracks along its grain boundaries weeks or months after it was made, under no service load at all; the name comes from brass cartridge cases cracking in storage near stables, and the cure is a stress-relief anneal rather than a better alloy. Dezincification is the other: in stagnant or aggressive water, zinc leaches out of high-zinc brass and leaves a porous copper skeleton that holds the part's exact shape while losing nearly all of its strength, so a fitting passes inspection and then splits under pressure. Dezincification-resistant brass with an arsenic inhibitor exists for exactly this. In copper tube, the mechanism is erosion-corrosion: above roughly 1.5 m/s in water the flow strips the protective film and cuts horseshoe-shaped pits immediately downstream of elbows and burred cut ends, which is why plumbing codes cap velocity. And in electrical service the failure is thermal rather than chemical. A terminal loosens as the copper creeps under the screw, contact resistance rises, the joint heats, oxidation raises resistance further, and it runs away, which is why infrared surveys of switchgear look for warm connections rather than for corroded ones.

C11000 in building wire, busbar, and motor windings; C36000 free-cutting brass in machined fittings and valves, now displaced from potable plumbing by C69300 and similar silicon brasses; C46400 naval brass in marine hardware; C95400 aluminum bronze in bushings, valve seats, and ship propellers; C93200 leaded tin bronze in plain bearings; C70600 and C71500 cupronickel in seawater piping and heat-exchanger tubing; C18150 chromium-zirconium copper in resistance-welding electrodes; C17200 beryllium copper in connector contacts, springs, plastic-mold inserts, and non-sparking tools, with Materion as the dominant producer.

Copper is priced on the exchange, so a fabricator sells metal plus a conversion charge and hedges the metal, which means the interesting question about any copper business is the conversion margin rather than the copper price. The demand side is the strongest structural story in base metals: grid buildout, renewables, and data centers all need conductor, and a battery-electric vehicle uses roughly three to four times the copper of a combustion car. Supply responds slowly, because a new mine takes ten to twenty years from discovery to production and ore grades at existing mines keep falling, which is why the price has run well above its historical band. Scrap covers roughly a third of consumption and copper recycles without any loss of properties, so the secondary market is a genuine buffer rather than a downcycle. Two regulatory shifts matter to anyone specifying copper parts. US drinking-water rules capping lead at a 0.25% weighted average of wetted surfaces removed leaded free-cutting brass from plumbing and pushed the industry onto silicon brasses that machine noticeably worse, so machining cost went up across a whole product category. And the beryllium exposure limit of 0.2 µg/m³ has concentrated beryllium-copper machining into shops set up for it, which shows up as lead time rather than as price.

5 minCopper, Brass & Bronze alloys explainedAalcoMetals · 100k+ views

3 minThe Difference Between Copper, Brass and BronzeMetal Supermarkets · 1m+ views

Copper 101 (Copper Development Association) · Copper for Busbars: Guidance for Design and Installation (Copper Development Association)

Nickel superalloys are what lets a turbine run with gas hotter than the melting point of the metal in the flow path. Their strength comes from gamma prime, an ordered Ni3(Al,Ti) precipitate that forms coherently inside the nickel matrix at 60–70% by volume in blade alloys and, unusually, gets stronger as it gets hotter up to about 800 °C rather than softening. The family has two poles. Wrought alloys, above all Inconel 718, are forged and machined into discs, casings, shafts, and fasteners; 718 ages to roughly 1,240 MPa, holds useful properties to about 650 °C, costs $35–60/kg, and is somewhere near a third of all superalloy tonnage. Cast alloys go the other way: directional solidification removes the grain boundaries running across the blade, single-crystal casting removes them entirely, and second-generation alloys such as CMSX-4 and René N5 hold 1,050–1,100 °C of metal temperature while the gas around them runs 1,500–1,700 °C, using film cooling through drilled holes and a 100–300 µm yttria-stabilized zirconia thermal barrier coating to bridge the gap.

No other class of metal keeps useful strength at 0.8 of its absolute melting temperature, and the corrosion-oriented members of the family (Inconel 625, Hastelloy C-276) survive chemistry that destroys stainless. The costs are spread across the whole value chain rather than concentrated in the alloy price. Superalloys machine at 5–15 m/min against 150 m/min or more for steel, and they work-harden ahead of the tool so a dwell puts a hard layer where the next pass has to cut, which means tool life is measured in minutes and finished parts cost far more than their material. Density is 8.2–8.9 g/cm³, so specific strength is around 150 kN·m/kg, no better than press-hardened steel. Welding the gamma-prime-rich alloys causes strain-age cracking during post-weld heat treatment, so blade repair is a specialty trade rather than a shop operation. And the manufacturing itself is the moat: single-crystal casting yields are low, closely held, and concentrated in a handful of foundries worldwide.

Superalloys are a temperature answer, so start by checking whether you actually need one. Below about 650 °C, stainless or a ferritic-martensitic steel is far cheaper and easier to make parts from, and moving to a superalloy for margin rather than need is the most common way to overspend. Between roughly 650 °C and 900 °C in a rotating or pressure-bearing part, IN718 or Waspaloy is the default and there is no serious alternative. Above 900 °C in a gas path, you are into cast blade alloys with cooling and coatings, and the design question becomes cooling-air budget rather than metal choice. If the requirement is corrosion rather than heat, pick from the solid-solution alloys (625, C-276, alloy 825) instead of the gamma-prime ones, since you are paying for nickel and molybdenum, not for creep strength you will not use. And above about 1,200 °C the metal has run out: a SiC/SiC ceramic matrix composite (028) does that temperature at roughly a third the density, which is why CMC shrouds and liners have started replacing superalloy ones.

IN718 aged to roughly 1,240 MPa at 8.19 g/cm³, about 150 kN·m/kg, useful to around 650 °C, at $35–60/kg · single-crystal alloys such as CMSX-4 holding 1,050–1,100 °C metal temperature · gas path temperatures of 1,500–1,700 °C, above the alloy's own melting range · gamma prime at 60–70% by volume in blade alloys, and strength rising with temperature to about 800 °C · thermal barrier coatings 100–300 µm thick · machining at 5–15 m/min, roughly a tenth of steel's or less · LME nickel near $16,750 per tonne in early August 2026.

Creep sets the design life, and it is a budget rather than a limit. At blade temperature the metal deforms slowly under centrifugal load until the tip clearance closes and the blade rubs the shroud, so parts are rated in hours at temperature and the industry's rule of thumb is that roughly 25 °C of extra metal temperature halves creep life. That single sensitivity is why cooling-hole blockage from dust or a coating repair that changes hole geometry is treated as a serious defect. The second mechanism is thermal-mechanical fatigue: every start and shutdown cycles the surface against the interior, and cracks initiate at cooling holes and fillets where the constraint is highest, which is why an engine's life is quoted in cycles as well as in hours. Third is hot corrosion, which is faster than oxidation and works differently: sodium sulfate deposits from sea salt or sulfur-bearing fuel flux away the protective alumina scale, in a low-temperature form around 650–800 °C and a high-temperature form around 800–950 °C, and a marine gas turbine can lose a blade set in a fraction of its rated life to it. Finally, the coating usually fails before the metal. A thermal barrier spalls once the thermally grown oxide under it reaches a critical thickness, and the moment the ceramic comes off, the metal underneath is exposed to the full gas temperature and the creep clock runs many times faster.

Inconel 718 in compressor and turbine discs, casings, shafts, and fasteners across essentially every gas turbine; CMSX-4 and René N5 single-crystal high-pressure turbine blades; Mar-M247 in directionally solidified blades; Waspaloy and Udimet 720 in discs; Inconel 625 and Hastelloy C-276 in chemical, marine, and flue-gas service where corrosion rather than heat is the problem; Haynes 282 in advanced ultra-supercritical steam plant. Mill products come from Special Metals, ATI, Carpenter Technology, Haynes International, VDM Metals, and Aubert & Duval; investment castings from Precision Castparts, Howmet Aerospace, and Doncasters. In additive manufacturing, 718, 625, and Hastelloy X are the printable ones, and the gamma-prime-rich blade alloys largely are not, because they crack during printing and post-weld heat treatment.

The metal price is a minority of the cost. LME nickel near $16,750 per tonne would put raw nickel around $17/kg, and IN718 mill product sells at $35–60/kg because of niobium content, double or triple vacuum melting for cleanliness, small lot sizes, and rigorous certification. The finished part then costs several times the billet again, because of machining rates, buy-to-fly, and inspection. Two inputs carry real supply risk: cobalt, most of which comes from the Democratic Republic of the Congo, and rhenium, which appears at 3% in second-generation single-crystal alloys, is produced only as a molybdenum by-product, and costs orders of magnitude more per kilogram than nickel, which is why later alloy generations have worked to use less of it. The durable business position here is not the composition, which is published in the open literature, but the process yield: single-crystal casting, hot isostatic pressing, coating lines, and the certification packages that go with them. That is also why engine aftermarket repair of hot-section parts is a better business than making them, and why any new entrant claiming a better superalloy should be asked who can cast it and who has qualified it.

13 minThe Insane Properties of SuperalloysThe Efficient Engineer · 100k+ views

22 minGas Turbine Blades and their Heat-Defying Single-Crystal SuperalloysAsianometry · 100k+ views

13 minThe Story of Nickel Superalloys: Saving the World in a Different WayThe Mat Sci Guy · 50k+ views

Superalloys: A Primer and History (TMS) · An Overview of Thermal Exposure on Microstructural Degradation and Mechanical Properties in Ni-Based Single Crystal Superalloys (Materials, via PMC)

The refractory metals are tungsten, tantalum, molybdenum, and niobium, and they exist on this sheet because they are the only metals that work above about 1,200 °C. Their melting points are the highest in the periodic table after carbon: tungsten 3,422 °C, tantalum 3,017 °C, molybdenum 2,623 °C, niobium 2,477 °C. That same property is why they are not melted and cast like other metals. Tungsten and molybdenum are made by powder metallurgy, pressed from reduced oxide powder and sintered near 2,000 °C, then swaged or rolled to close the remaining porosity; tantalum and niobium are electron-beam melted under vacuum. The catch that shapes every application is oxidation. Above roughly 500–600 °C in air, tungsten and molybdenum form oxides that volatilize rather than protect, so the metal keeps disappearing instead of passivating, and any real high-temperature use means vacuum, inert gas, or a coating. Sintering and the powder routes themselves are on the manufacturing-processes sheet.

These metals hold real strength and stiffness at temperatures where nickel superalloys are liquid, and tungsten's 411 GPa modulus is roughly twice steel's, the highest of any metal in common use. Tantalum has the additional distinction of resisting nearly every acid short of hydrofluoric. The counterintuitive weakness is that specific strength is bad. Tungsten runs about 50 kN·m/kg, the same as mild steel, because its 19.25 g/cm³ density cancels its strength, and its specific stiffness at roughly 21 GPa per g/cm³ is actually below steel's 26. So nobody chooses tungsten to make something light or stiff for its weight; they choose it to make something heavy in a small volume, or stiff in an absolute sense, or hot. On top of that, recrystallized tungsten is brittle at room temperature, molybdenum and tungsten need protective atmospheres, the whole class machines slowly and abrasively, and mill forms run $50–200/kg with tantalum at $250–400.

Ask what you actually want from the metal, because the four uses barely overlap. If you need mass in a small envelope, use tungsten or a tungsten heavy alloy at 17–18.5 g/cm³: aircraft and motorsport ballast, vibration-damping tool holders, radiation shielding, and collimators, where density is the specification and heat never enters into it. If you need hot structure in a furnace or a vacuum chamber, use molybdenum or TZM for heating elements, hot zones, hearths, and extrusion dies, and design the atmosphere as carefully as the part. If you need chemical resistance beyond what nickel alloys give, use tantalum in liners, heat exchangers, rupture discs, and implants. If you need superconductivity, niobium is the only option, in Nb-Ti and Nb3Sn wire and in accelerator cavities. Do not use these metals as a general high-temperature upgrade over superalloys (011), because the oxidation problem usually makes them impractical in an open gas path, and for hot structure exposed to air a ceramic, a ceramic matrix composite (028), or carbon-carbon (025) is more likely to be the workable answer.

Melting points of 3,422 °C for tungsten, 3,017 °C for tantalum, 2,623 °C for molybdenum, and 2,477 °C for niobium · tungsten at 19.25 g/cm³ and 411 GPa modulus, about twice steel's, but only around 50 kN·m/kg of specific strength, the same as mild steel, and about 21 GPa per g/cm³ of specific stiffness, below steel's 26 · tungsten and molybdenum oxidize away in air above roughly 500–600 °C · tungsten heavy alloys at 17–18.5 g/cm³ · mill forms typically $50–200/kg, with tantalum at $250–400 · roughly 80% of mined tungsten comes from China.

The oxidation failure has no plateau, which is what makes it dangerous. A protective oxide works by forming once and then stopping the reaction; molybdenum and tungsten oxides volatilize instead, so a molybdenum element running in a furnace with a small air leak loses section continuously at a steady rate until it burns through, and the size of the leak sets the schedule rather than whether the part survives. The yellowish oxide smoke is the warning sign, and by the time it is visible the part is being consumed. The second mechanism is embrittlement by recrystallization. Worked tungsten has some room-temperature ductility, but one thermal cycle hot enough to recrystallize the grain structure moves the ductile-to-brittle transition well above room temperature, so a part that was tough when installed cracks during the next cold handling, thermal shock, or bolt-up, with no change in composition and nothing to see. Third, all of these metals absorb interstitials. Oxygen, nitrogen, carbon, and hydrogen dissolve into the lattice at temperature and embrittle it, which is why welding is done in vacuum or high-purity argon and why a tantalum liner electrically coupled to a steel flange in acid service fails by hydrogen embrittlement rather than by corrosion: the tantalum is the cathode, it takes up the hydrogen, and it cracks.

Tungsten and tungsten heavy alloys such as Plansee's Densimet in radiation shielding, medical collimators, aircraft and Formula 1 ballast, boring bars, and kinetic penetrators, plus tungsten X-ray anode targets and electrodes. Molybdenum and TZM in furnace hot zones and hearths, glass-melting electrodes, hot extrusion dies, sputtering targets, and flat-panel display backplanes. Tantalum in electrolytic capacitors, which is still where most tantalum goes, and in chemical process equipment and implants. Niobium is the odd one out, since roughly 90% of it is consumed as ferroniobium in high-strength low-alloy steel and only a small fraction reaches metal form, in Nb-Ti and Nb3Sn superconducting wire for MRI magnets and fusion machines and in superconducting radiofrequency accelerator cavities. The fabricators to know are Plansee, H.C. Starck, Global Advanced Metals, and Elmet, with CBMM supplying the large majority of world niobium.

These are small markets with concentrated supply, which makes them policy-sensitive out of proportion to their tonnage. China accounts for roughly 80% of mined tungsten and added tungsten, molybdenum, tellurium, bismuth, and indium-related items to its export control list in February 2025, which turned a catalog purchase into a licensing question for Western buyers more or less overnight. Tantalum comes substantially from Central Africa and carries conflict-mineral reporting obligations under US and EU rules, so the compliance overhead is real even at small volumes. Niobium is more concentrated still, with CBMM in Brazil supplying most of the world from a single deposit, though the steel industry that consumes it has never treated that as a crisis. Prices in the $50–400/kg range mean the material is a meaningful share of part cost, unusually for a metal, and the processing routes are capital-intensive and slow, with sintering furnaces and vacuum equipment that nobody builds speculatively. For anyone evaluating a business here, the questions are whether the position is upstream of an export control, whether the part can be qualified in a lower-cost material, and whether the customer is buying density, temperature, or chemistry, because those three markets behave nothing alike.

6 minTungsten - The MOST REFRACTORY Metal ON EARTH!Thoisoi2 - Chemical Experiments! · 1m+ views

11 minHow Is Molybdenum Made? THE METAL THAT KEEPS JET ENGINES ALIVESecrets of Everyday Things · 10k+ views

Properties & Intermediates (International Tungsten Industry Association) · Molybdenum metal & alloys (International Molybdenum Association)

Polyethylene and polypropylene are the cheapest plastics that do real work, and together they are close to half of all plastic produced. Both are plain hydrocarbon chains with no polar groups on them, and most of their behavior follows from that one fact. There is nothing for acids, bases, or salt solutions to attack, so chemical resistance is excellent and water uptake is essentially zero. There is also nothing for an adhesive or a paint to bond to: surface energy sits near 30 mN/m, and wetting usually needs about 38, which is why polyolefin parts get flame- or plasma-treated before anyone prints on them or glues them. Both are semicrystalline with glass transitions well below room temperature (about -120 °C for PE, near 0 °C for PP), so they stay ductile down to the cold and they creep under sustained load at 20 °C. Density runs 0.90–0.96 g/cm³, which makes them the only common plastics that float.

The case for polyolefins is price at $1–1.6/kg, near-total chemical resistance, no moisture uptake at all (so a molded part is the same size in a desert and a swamp), broad food-contact clearance, fast molding with no resin drying step, and a genuine recycling stream for HDPE and PP. Against that, stiffness is poor: modulus is 1.0–1.5 GPa, roughly one one-hundred-fiftieth of steel's, and continuous service tops out near 90–100 °C for PP and 60–80 °C for HDPE. Nothing sticks to them, so assembly means welding, snap fits, or a surface treatment step. The failure mode that catches people is environmental stress cracking: HDPE held under tensile stress in contact with a detergent or surfactant cracks in weeks at a stress well below yield, and the part passes every static strength test first.

Default to a polyolefin for anything that holds a fluid, gets thrown away, or ships in millions of units, as long as stiffness and temperature are not the binding constraints. Use PP when you need the higher service temperature, a living hinge, or a clean autoclave cycle, and HDPE when you need toughness, impact at low temperature, or pipe. Reach for UHMWPE when abrasion is the problem and you can live with a part that is machined or compression molded instead of injection molded. If the part has to be painted, plated, or solvent-bonded, go to ABS (014). If it has to hold a bolt preload or cut a gear tooth, go to nylon or acetal (015). And if it sits outdoors, specify carbon black or accept a two-season life.

Resin $1–1.6/kg · density 0.90–0.96 g/cm³, so they float · PP tensile about 35 MPa at 0.90 g/cm³, near 39 kN·m/kg · modulus 1.0–1.5 GPa · continuous service 90–100 °C for PP and 60–80 °C for HDPE · surface energy near 30 mN/m against the roughly 38 an adhesive needs · close to half of world plastic volume.

Environmental stress cracking is the mechanism that surprises people. A surfactant or detergent gets into the amorphous regions between crystallites, lets the tie molecules pull apart, and turns a ductile material into a brittle one at a stress well under yield, usually over weeks. Creep is the second: with the glass transition below room temperature, chains keep sliding under load, so a bolted flange loses preload and a loaded shelf sags rather than breaking. Third is photo-oxidation, which chalks and then embrittles an unstabilized part within a season or two outdoors, faster in PP than PE because PP's tertiary hydrogens are easier to abstract; roughly 2–2.5% carbon black is the standard fix and it is why outdoor pipe and geomembrane are black. Watch for a chalky surface, fine crazing at gates and radii, and a part that has visibly changed shape under load.

HDPE in milk jugs, jerry cans, PE100 gas and water pipe, and landfill geomembranes. LDPE and LLDPE in film, wire insulation, and rotomolded tanks, kayaks, and Yeti-style coolers. PP in car bumper fascia and battery cases, woven bags, autoclavable labware, spunbond nonwovens for masks and diapers, and living-hinge lids of the Tic Tac type. UHMWPE in hip and knee bearing surfaces, conveyor wear strips, and Dyneema and Spectra fiber for rope and ballistic panels. The producers are ExxonMobil, LyondellBasell, Dow, SABIC, Borealis, Braskem, and Sinopec.

Polyolefin prices track ethylene and propylene, which track naphtha in Asia and Europe and ethane in the US, so US Gulf Coast crackers running on shale ethane have held a structural cost advantage for a decade. A large capacity build in the US and China between roughly 2017 and 2023 left the industry long, polyethylene margins have been poor since, and several European crackers have closed. For anyone building a product, resin cost is a small share of a molded part's cost, since tooling and cycle time dominate at anything under high volume, so the reason to pick a polyolefin is usually that it molds fast, needs no drying, and survives whatever is in the bottle. The one place the material price does bite is recycled content: food-grade recycled PE and PP carry a premium over virgin resin because supply is short against EU packaging targets, which is the opposite of how recycled material is supposed to price.

4 minWhat is Polypropylene Plastic | Can it be recycled?Plascon Plastics · 50k+ views

11 minHow To Bond Polyethylene & Polypropylene Plastics (Phillips Vision: Episode - 123)Phillips Vision · 10k+ views

4 minWhat is a Plastic Living Hinge? Hinge Design Basics.OneMonroe · 10k+ views

Polypropylene (PP) (British Plastics Federation) · Handbook of Polyethylene Pipe (Plastics Pipe Institute)

PVC and the styrenics are the cheap plastics you reach for when polyolefins are too floppy or impossible to bond. PVC is 57% chlorine by weight, so it is less than half oil-derived, it is priced partly off salt and chlor-alkali economics rather than off crude, and it does not sustain a flame without help. It is also glassy at room temperature, which gives it a 3 GPa modulus, and it dissolves in the right solvent, which is why PVC pipe is joined with a cement that fuses the two parts rather than gluing them. The styrenics start from polystyrene, a clear glassy polymer that is stiff, cheap, and brittle, and get useful by blending in rubber: polybutadiene gives HIPS, and grafting styrene-acrylonitrile onto polybutadiene gives ABS. ABS is the default molded housing material because it is the one commodity plastic that paints, chrome-plates, and solvent-welds without surface treatment.

Rigid PVC gives you stiffness, self-extinguishing behavior, good acid and base resistance, and outdoor life measured in decades when it is loaded with titanium dioxide, all for around $1–1.5/kg. Plasticized PVC covers the whole range from that down to a soft tube by adding 20–50% plasticizer, which no other single polymer does. ABS at $2–3/kg gives a stiff, tough, dimensionally stable molding that finishes well. The weaknesses are temperature and chemistry: rigid PVC is done at about 60 °C, its melt decomposes above roughly 180–200 °C into hydrogen chloride that autocatalyzes further decomposition and corrodes tooling, and both PVC and ABS are attacked by ketones, esters, and chlorinated solvents. ABS's own failure mode is UV: the polybutadiene phase oxidizes, so an outdoor ABS part yellows, chalks, and loses most of its impact strength while its tensile number barely moves.

Pick rigid PVC when the part carries water, sits in a wall, or has to pass a flame test, and 60 °C is a ceiling you can live with; use CPVC when it is hot water and you need about 95 °C. Pick plasticized PVC for cheap flexible tube and wire jacket where the service life is years rather than decades, and silicone or a TPE when plasticizer migration or low-temperature flexibility matters. Pick ABS for a molded housing that will be painted, plated, or solvent-bonded, and ASA instead whenever the part lives outdoors, since ASA swaps the butadiene rubber for an acrylic one that does not care about UV. Go up to polycarbonate or a PC/ABS blend (015) if you need a drop test at -20 °C or service above 90 °C, and down to polypropylene (013) if the part only has to hold liquid and nobody looks at it.

PVC 57% chlorine by weight · rigid PVC to about 60 °C, CPVC to about 95 °C · PVC melt decomposes above roughly 180–200 °C, releasing HCl · plasticizer at 20–50% takes PVC from 3 GPa rigid to Shore A 60–90 · PVC roughly $1–1.5/kg, ABS $2–3/kg · polystyrene elongation 2–3% at break · ABS heat deflection near 90 °C.

PVC has two clocks running. In the barrel, thermal degradation strips HCl off the chain, and the HCl catalyzes more of the same, so a short shot left sitting turns brown and then black, and the acid pits the tool; heat stabilizers (calcium-zinc or organotin now, lead historically) exist only to intercept that. In service, the failure is plasticizer loss: plasticizer is dissolved in the PVC rather than bonded to it, so it migrates out into whatever the part touches and evaporates at temperature, and the tube or the wire jacket stiffens and then cracks after years. Styrenics fail by notch sensitivity and by UV. Polystyrene breaks from any sharp corner or molded-in stress with 2–3% elongation and no warning, and ABS left in sunlight yellows and chalks as its rubber phase oxidizes, which is a surface effect that shows up as an impact failure long before any strength test notices.

PVC is mostly pipe, plus window and door profiles, siding, resilient flooring, wire insulation, roofing membrane, and medical tubing and blood bags. Styrenics show up as LEGO bricks and automotive interior trim in ABS, refrigerator liners in HIPS, disposable cutlery and clamshells in GPPS, EPS packaging and cups, chrome-plated grilles and badges on etched ABS, and motorcycle fairings and RV siding in ASA. The main PVC producers are Shin-Etsu (through Shintech in the US), Formosa Plastics, Westlake, INEOS, and Orbia; styrenics come from INEOS Styrolution, Trinseo, Chi Mei, LG Chem, and SABIC.

PVC economics are unusual because chlorine is a co-product of caustic soda: a chlor-alkali plant makes both whether or not you want both, so PVC resin pricing partly depends on what caustic is doing, and PVC is the largest single use of chlorine in the world. Demand is construction demand, which means it follows housing starts and infrastructure spending rather than consumer goods. Styrene monomer tracks benzene and ethylene, so styrenics move with oil in a way PVC does not. The live risks are regulatory rather than economic: phthalate plasticizers keep getting restricted in medical and toy applications, lead stabilizers have been phased out in Europe, PVC is a known contaminant in PET recycling streams, and vinyl chloride monomer is a confirmed human carcinogen handled in closed systems since the 1970s. If you are building a product on PVC, assume the additive package will be reformulated at least once during the product's life.

11 minThe PVC production process explainedAntonio Hernandez · 50k+ views

5 minWhat is ABS Plastic? | How to use ABS Plastic. 3D Printing? Lego?Plascon Plastics · 10k+ views

3 minWhat is Polystyrene Plastic? | Why Styrofoam is TERRIBLE!Plascon Plastics · 10k+ views

Polyvinyl Chloride PVC (British Plastics Federation) · Acrylonitrile Butadiene Styrene (ABS) and Other Specialist Styrenics (British Plastics Federation)

Engineering thermoplastics are the mid-tier: molded parts that carry real mechanical load at $3–6/kg, with tensile strengths of 60–80 MPa and moduli of 2.4–3.3 GPa instead of the commodity plastics' 1.0–1.5. They differ from polyolefins by having polar groups on the chain, which is where the strength, the stiffness, and every one of their problems comes from. Polar groups mean the chains attract each other, so the material is strong; they also mean water, solvents, and acids have something to interact with. Nylon takes up water and changes size, acetal comes apart in acid and in chlorinated water, and polycarbonate cracks in the wrong cleaner. The differences inside this family are larger than the difference between the family and the one below it, so picking "an engineering plastic" is not a decision, and picking nylon over acetal usually is.

This family is where you get a part that replaces a small machined metal one: a gear, a bearing, a latch, a connector body, a housing that has to survive a drop. Strength is 60–80 MPa, they mold to tight tolerance, and 30% short glass fiber roughly triples modulus (nylon 6/6 goes from about 3 GPa to 9–10) and raises heat deflection by a hundred degrees or more, because the fiber carries load above the glass transition. The price of glass fill is elongation, which falls from tens of percent to 2–3%, so a filled part is strong and brittle and it warps, since the fibers align with flow and the shrinkage differs along and across it. The weakness that costs the most money is that the datasheet number is a dry, room-temperature, injection-molded test bar. Nylon 6/6 quotes 80 MPa dry, but at 50% relative humidity it holds about 2.5% water, loses roughly half its modulus, and grows 0.5–0.7% in every dimension, which is enough to close a bearing clearance.

Reach for this family when a polyolefin is not stiff or strong enough and PEEK is not worth $80/kg. Then choose inside it by the thing that will actually go wrong. If the part must hold a dimension, pick acetal, because it takes up almost no water. If it must absorb impact and take abuse, pick nylon, and specify PA6 for toughness or PA66 for heat. If it must be transparent or survive a hammer, pick polycarbonate, and then check every fluid it will ever touch against a stress-cracking chart before you commit. If it must be dimensionally stable and hold a connector pin, pick glass-filled PBT. If it must be optically clear outdoors, pick PMMA and accept brittleness. Go to PPS or PEEK (016) once continuous service passes about 120 °C, and back down to polypropylene (013) if you only wanted chemical resistance.

Resin $3–6/kg · tensile 60–80 MPa, modulus 2.4–3.3 GPa · nylon 6/6 at 80 MPa dry, holding 2.5% water at 50% RH, losing roughly half its modulus and growing 0.5–0.7% · 30% glass fiber roughly triples modulus and cuts elongation to 2–3% · polycarbonate glass transition 147 °C, about 90% light transmission · acetal depolymerizes above roughly 230–240 °C · continuous service 90–125 °C across the family.

Moisture, temperature, and chemistry, in that order, and none of them show up on the front page of a datasheet. Nylon reaches equilibrium with ambient humidity over weeks, which plasticizes it: strength and modulus drop by roughly half and the part grows, so a press fit that was correct on the bench is loose after a summer. Environmental stress cracking is the polycarbonate failure and it needs stress plus a specific fluid at the same time, so a part passes every mechanical test and then crazes when someone wipes it with the wrong cleaner. Hydrolysis is the slow one, and it applies to polycarbonate, PBT, and to a lesser extent nylon: hot water cleaves the chain, molecular weight falls, and a part that looks unchanged fractures with no yielding after a year or two of steam or coolant exposure. Check for surface crazing near stress concentrations, brittle fracture surfaces on a normally ductile resin, and dimensions that have drifted in one direction.

Nylon 6/6 in cable ties, automotive intake manifolds and radiator end tanks, and power-tool housings. Acetal in gears, seat-belt and buckle mechanisms, fuel-system components, and zipper teeth, sold as Delrin (a standalone company since DuPont divested it in 2023), Celcon, and Hostaform. Polycarbonate in machine guarding, headlamp lenses, safety glazing, and medical device housings. Glass-filled PBT in automotive and appliance connectors, sold as Valox and Crastin. PMMA in signage, tail lamps, and aquarium glazing, from Röhm and Mitsubishi Chemical. The nylon chain runs through Invista, Ascend, BASF, and Asahi Kasei; acetal and PBT through Celanese, Polyplastics, and BASF; polycarbonate through Covestro, SABIC, and Wanhua.

This family sits in the awkward middle: expensive enough that resin cost shows up in a bill of materials, cheap enough that nobody qualifies a second source for years. Nylon 6/6 is the volatile one, because its supply chain runs through adiponitrile, which only a handful of plants in the world make; an outage at one of them moves the price by 50% or more within a quarter, as happened after the 2021 Texas freeze. Chinese capacity has been added across polycarbonate, PBT, and nylon 6 since about 2020, which has compressed Western producer margins and made the commodity grades cheap while the specialty compounds (flame-retardant, hydrolysis-stabilized, laser-weldable, medical-grade) hold their pricing. The durable business here is in compounding rather than polymerization, because a validated grade with a regulatory clearance and a color match is expensive to displace, and switching it restarts an automotive part approval or a medical filing.

14 minABS, POM, Polycarbonate, Teflon and PEEK for CNC machiningProtolabs · 10k+ views

4 minWhat is Polycarbonate Plastic? | Polycarbonate vs Acrylic? What are the Differences?Plascon Plastics · 10k+ views

7 minBoedeker TECH Talk Episode 6 | Understanding the Moisture Absorption of Plastics ASTM D570Boedeker Plastics · 5k+ views

Nylons (Polyamide) (British Plastics Federation) · Characterisation and Modelling of Moisture Gradients in Polyamide 6 (Polymers, via PMC)

These are the plastics that still work above 200 °C, and they cost 20 to 50 times what a commodity resin does. The chemistry behind them is aromatic rings and stiff linkages in the backbone (ketone, sulfide, imide, sulfone) instead of the flexible aliphatic chains of nylon and polyethylene, which raises the glass transition and makes the polymer hard to burn and hard to dissolve. PEEK is the reference point: glass transition 143 °C, melting point 343 °C, roughly 250 °C continuous in air, tensile near 95–100 MPa at 1.30 g/cm³, and $70–110/kg. The two members people forget are the interesting ones commercially, because PPS at $12–20/kg and PEI at $25–40/kg solve most of the same problems for a quarter of the money. All of them need melt processing at 350–400 °C with tools held at 180–200 °C, which is the real barrier: the resin is expensive, but the capital equipment is what stops a shop from quoting the job.

You get continuous service at 200–250 °C, resistance to almost every solvent and to steam and radiation, inherent flame retardancy with low smoke, low outgassing for vacuum and semiconductor work, and wear behavior good enough to run unlubricated bearings when filled with carbon and PTFE. Specific strength is respectable rather than remarkable at about 75 kN·m/kg for unfilled PEEK, which is 50% better than mild steel and a fifth of a carbon laminate. The honest weakness is that the metal-replacement case is usually not about mass: you buy these to delete a corrosion problem, a lubricant, a coating, or a machining step, and mass savings are a bonus rather than the argument. Cost is the obvious limit, and processing is the quiet one, since 400 °C barrels, hot runners rated for it, and corrosion-resistant hardened tool steel against abrasive glass-filled melt all have to be in place before the first shot.

Go here when continuous service passes about 150 °C, or when the chemistry rules out everything cheaper, and then work down the price ladder rather than starting at PEEK. If the part is under the hood and needs 200 °C with chemical resistance and cheap glass-filled molding, use PPS. If it needs 170–180 °C, transparency, and an aircraft-interior flame, smoke, and toxicity rating, use PEI. If it needs steam sterilization cycle after cycle, use PPSU. Use PEEK when you need 250 °C, hydrolysis resistance, radiation tolerance, or implant-grade material, and use PAI only when nothing else reaches. Drop back to a glass-filled nylon or PBT (015) if the real temperature turns out to be 130 °C, and compare against aluminum honestly: an anodized aluminum bracket is usually cheaper unless the environment is eating it.

PEEK glass transition 143 °C, melting 343 °C, roughly 250 °C continuous, $70–110/kg · PPS $12–20/kg at 200–220 °C continuous · PEI glass transition 217 °C, $25–40/kg · PEEK tensile 95–100 MPa at 1.30 g/cm³, near 75 kN·m/kg · melt processing at 350–400 °C with tools at 180–200 °C · 20–50 times commodity resin price.

Almost never by running out of strength, and usually by one of three mechanisms. The first is a misread temperature rating: PEEK's 250 °C number is thermal-oxidative stability, and its glass transition is 143 °C, so modulus drops by roughly half between about 140 and 170 °C, and a bracket sized on the room-temperature datasheet sags at a temperature the datasheet appears to allow. The second is crystallinity, which the mold sets rather than the resin: a semicrystalline part shot into a cold tool comes out partly amorphous, then crystallizes slowly in service, shrinking and going brittle months later. The third is environmental stress cracking in the amorphous members, since PEI and the sulfones have no crystalline phase to resist solvent ingress, so a stressed part crazes in chlorinated solvents, some hydraulic fluids, and a surprising number of cleaning agents. Watch for parts that changed dimension after a heat cycle, and for brittle fracture in a resin that was ductile when it shipped.

PEEK in spinal fusion cages and trauma implants (Invibio's PEEK-OPTIMA), semiconductor wafer carriers and test sockets, downhole seals and back-up rings, aircraft brackets and clips, and unlubricated bushings in carbon and PTFE filled grades. PPS in automotive thermostat housings, water-pump impellers, EGR and fuel-system parts, and electrical connectors, sold as Ryton and Fortron. PEI as Ultem in aircraft interior panels, sterilizable medical trays, and high-temperature electrical insulation. PPSU as Radel in surgical instrument cases and baby bottles, PAI as Torlon where 275 °C is required, and LCP in fine-pitch connectors and antenna film. Suppliers are Victrex, Solvay's specialty-polymer business (now Syensqo), Evonik, SABIC, Celanese, Toray, DIC, and a growing set of Chinese PEEK producers.

The economics are set by qualification rather than by chemistry. PEEK was a near-monopoly for Victrex for decades, and while the base patents are long expired and Chinese producers have entered, the price has fallen much less than a commodity view would predict, because medical, aerospace, and semiconductor customers buy a qualified grade with a traceable history rather than a spec. That is why PEEK sits at $70–110/kg while PPS, which is a genuine automotive commodity bought on price, sits at $12–20. Growth has been strongest in the cheap end of the family, because electrification put a lot of PPS into inverters, pumps, and connectors, and in additive manufacturing, where high-temperature filaments and sintering powders opened a market for small, expensive parts that nobody would tool for. If you are pricing a business here, look at whether the value is in the polymer, the compound, or the qualification package, because the last one is where the margin holds.

4 minWhat is PEEK?Victrex · 50k+ views

7 minHow PEEK Changed Engineering ForeverVision Miner · 100k+ views

7 minPolyphenylene Sulfide (PPS) Plastic: Properties & Applications in 3D PrintingVision Miner · 10k+ views

Tracing the History of Polymeric Materials, Part 26: High-Performance Thermoplastics (Plastics Technology) · Polyaryletherketone Based Blends: A Review (Polymers)

Fluoropolymers are carbon chains wrapped in fluorine, and the carbon-fluorine bond is the strongest single bond in organic chemistry. PTFE is the extreme case, with the chain completely shielded: essentially nothing attacks it below 260 °C except molten alkali metals and elemental fluorine, its friction coefficient of 0.05–0.10 is the lowest of any solid, and its dielectric constant of 2.1 with very low loss makes it the standard insulation for RF cable. The same shielding causes both of PTFE's hard limits. It has no intermolecular attraction to speak of, so it cold-flows under sustained load at room temperature, and its melt does not flow at all above the 327 °C melting point, so it cannot be injection molded. PTFE parts get compression molded from powder and sintered at 360–380 °C, or paste-extruded and sintered, closer to how a ceramic is made than how a plastic is. The melt-processable members (PVDF, FEP, PFA, ETFE) give back some chemical resistance or temperature in exchange for being moldable.

Nothing else combines this chemical inertness, this friction coefficient, this dielectric performance, and 260 °C continuous service. The weaknesses are mechanical and they are severe. Tensile strength is 20–35 MPa at a density of 2.15–2.20 g/cm³, which works out near 12 kN·m/kg, a quarter of mild steel's, so PTFE carries essentially no structural load. It creeps under any sustained stress, which is why a PTFE gasket relaxes out of a bolted flange over months and a plain PTFE bearing wears fast; 15–25% glass, carbon, bronze, or graphite filler cuts deformation under load by several times, at the cost of some chemical resistance and higher friction. Nothing bonds to it without a sodium-etch or plasma treatment. And it is porous to small molecules over time, so a lined vessel can corrode from the steel side while the liner looks perfect.

Specify PTFE when chemical attack or friction is the whole problem and the part carries no load: chemical-plant gaskets and linings, valve seats, unlubricated slide bearings under low pressure, and cable insulation where dielectric loss matters. Use a filled grade any time the part sees a sustained squeeze. Choose PFA when you need PTFE's chemistry and temperature but the geometry has to be molded, which is most semiconductor high-purity fluid handling. Choose PVDF when 150 °C is enough and you want chemical piping you can weld and machine like a normal plastic, and remember it is attacked by strong bases. Choose FEP for clear tubing and ETFE for tough film and wire jacket. If the part has to carry structure as well as resist chemicals, go to PEEK (016) instead, and if it has to seal dynamically with spring-back, go to an elastomer (021) or a spring-energized PTFE seal.

PTFE 260 °C continuous, melting 327 °C, sintered at 360–380 °C · friction 0.05–0.10, the lowest of any solid · tensile 20–35 MPa at 2.15–2.20 g/cm³, near 12 kN·m/kg · PTFE stock $20–35/kg, PFA well above that · dielectric constant 2.1 · 15–25% filler to control cold flow.

Cold flow is the characteristic one. Unfilled PTFE has no yield point in the usual sense, so under a constant load at room temperature it keeps deforming, and a gasket thins and squeezes out of the joint until the bolts lose preload and the flange weeps. The second is permeation, which is not attack: small molecules diffuse through the fluoropolymer over months, so the steel behind a PTFE or PFA liner corrodes and blisters the liner off from underneath while the liner itself is untouched. The third is thermal decomposition, since above roughly 400 °C PTFE gives off hydrogen fluoride and perfluoroisobutylene, which is why overheated cookware kills birds and why sintering ovens are ventilated. PVDF has its own mode: strong bases strip HF off the chain and the material discolors to brown or black before it cracks, so a color change in caustic service is the warning to act on.

PTFE in chemical-plant gaskets and vessel linings, valve seats and stem packing, non-stick cookware coatings, and RF and aerospace cable insulation. Expanded PTFE as Gore-Tex membrane, sealant tape, and vascular grafts. PFA in semiconductor tubing, fittings, and wet-bench tanks where extractable metals have to be near zero. PVDF as Kynar in chemical piping and tank liners, plenum-rated cable jacket, architectural coatings, and as the standard binder in lithium-ion cathodes. ETFE in stadium and greenhouse film, including the Allianz Arena and the Eden Project, and in aircraft wire. The producers are Chemours, Daikin, AGC, Syensqo, Arkema, Gujarat Fluorochemicals, and Dongyue, with 3M's Dyneon business exiting.

Two things dominate the outlook. Upstream, everything starts from fluorspar, which is on the EU and US critical-materials lists and is heavily concentrated in China, so the feedstock has the same political exposure as a critical mineral rather than the exposure of a petrochemical. Downstream, there is a regulatory clock: the universal PFAS restriction proposed under EU REACH in 2023 by five member states covers thousands of substances and includes fluoropolymers within scope, and 3M announced it would exit all PFAS manufacturing by the end of 2025, which took real capacity out of the market. Industry argues that fluoropolymers are polymers of low concern and should be carved out, and the outcome is genuinely unsettled. The practical consequence for anyone designing now is that PTFE and PVDF supply is tighter and pricier than it was, and that customers in Europe increasingly ask for a substitution study even where no substitute performs, so a program that depends on a fluoropolymer should carry a named alternative and an estimate of what losing the fluoropolymer would cost.

3 minHow Does PTFE Make Things Slippery?Science Channel · 10k+ views

4 minForever ChemicalsReactions · 10k+ views

Guide for the Safe Handling of Fluoropolymer Resins (Plastics Europe)

Thermoset resins start as a liquid and end as one crosslinked molecule, which is what makes them useful and what makes them a one-way trip. Epoxy is the reference: a bisphenol-A-based resin plus an amine or anhydride hardener, curing with under 2% shrinkage, reaching a glass transition of 120–180 °C, and bonding to almost anything it wets. Unsaturated polyester is the cheap alternative at $2–3/kg against epoxy's $4–8, thinned with styrene that both crosslinks and evaporates, and it shrinks 6–8% on cure. Phenolic is the flame, smoke, and toxicity answer, because it chars instead of dripping. The property that separates all of them from a thermoplastic is that the network forms once: no remelting, no welding, no reshaping, and essentially no recycling except grinding it up as filler or burning it for the fiber. Cure schedule matters as much as chemistry, because the network vitrifies once the glass transition reaches the cure temperature and the reaction stops there.

Epoxy's strengths are adhesion, low cure shrinkage, dimensional stability, electrical insulation, and the fact that it can be processed as a liquid at room temperature, which is what makes composites, potting, and structural bonding possible at all. Tensile is 60–80 MPa with a modulus around 3 GPa, so as a bulk material it is unremarkable; the value is that it holds fibers, chips, and steel plates in position. The weaknesses: it is brittle in the unmodified state, it yellows and chalks in UV without a coating, amine hardeners are strong skin sensitizers and epoxy dermatitis is one of the more common occupational allergies in construction, and mixing has to be exact because an off-ratio batch cures into something that looks right and is not. Thick sections have their own problem, since the cure is exothermic and self-accelerating, so a kilogram of resin in a cup can run to 200 °C and smoke while the same resin in a thin film cures gently.

Choose epoxy when you need adhesion, low shrinkage, or a structural composite, and when you can pay for a controlled cure. Choose unsaturated polyester when the part is large, open-molded, and cost-driven (boats, tanks, tub and shower, sheet molding compound) and 6–8% shrinkage is tolerable. Choose vinyl ester as the middle ground when polyester's hydrolysis resistance is not enough, as in chemical tanks, pipe liners, and marine skin coats against osmotic blistering. Choose phenolic when a flame, smoke, and toxicity spec is driving, which in practice means aircraft interiors, plus friction materials and foundry binders. Go to a bismaleimide or cyanate ester above about 200 °C, and go to a thermoplastic matrix when you want to weld, reform, or recycle the part, or when out-time and freezer storage of prepreg are the schedule risk. The processes themselves (hand layup, resin transfer molding, prepreg and autoclave, pultrusion) are on the manufacturing-processes sheet.

Epoxy $4–8/kg, vinyl ester $3–5, unsaturated polyester $2–3 · cure shrinkage under 2% for epoxy against 6–8% for polyester · glass transition 120–180 °C for structural epoxies, and roughly the cure temperature if you never post-cure · epoxy takes up 1–2% water, which drops wet glass transition by 20–30 °C · tensile 60–80 MPa at about 3 GPa modulus.

Under-cure is the most common failure and the hardest to see. An off-ratio mix, a cold shop, or a schedule that skipped the post-cure leaves an incompletely crosslinked network whose glass transition sits 30–50 °C below spec, and the part looks and feels finished. Then it fails by exceeding that glass transition in service, because modulus falls by one to two orders of magnitude within about 20 °C of it, so a bonded joint that holds at 20 °C lets go at 90 °C without anything breaking. Moisture makes this worse rather than causing its own failure: epoxy absorbs 1–2% water over months, the water plasticizes the network, and the wet glass transition drops another 20–30 °C, which is exactly why aerospace structures are designed to hot-wet allowables instead of dry ones. The fourth mode is microcracking from thermal cycling, since the resin shrinks far more than the fiber or the chip it is bonded to. Warning signs are a tacky or rubbery surface, a part that softens in hot water, and fine cracks along fiber directions.

Epoxy in every fiber-reinforced structure from wind blades to aircraft primary structure, in FR-4 printed-circuit laminate, in the molding compound that encapsulates almost every plastic-packaged integrated circuit, in underfill and potting, in structural adhesives such as Loctite and Araldite, and in tank and pipe linings. Unsaturated polyester in boat hulls, bathware, cultured marble, gel coats, and sheet molding compound for automotive closures. Vinyl ester in chemical storage tanks, FRP pipe, and marine skin coats. Phenolic in aircraft interior sidewalls and overhead bins, brake pads, grinding wheel binders, and foundry sand binders, and originally as Bakelite. The resin producers are Olin, Westlake Epoxy, Kukdo, Aditya Birla, Huntsman, Hexion, and Swancor.

Epoxy is a bisphenol-A and epichlorohydrin business, so its cost floor moves with those two, and its demand is dominated by wind energy, coatings, electronics laminate, and construction, which cycle differently from each other. Chinese capacity expansion has made basic liquid epoxy resin a commodity and pushed Western producers toward formulated systems, where the margin sits in the hardener package, the cure kinetics, and the qualification rather than in the resin. Two regulatory threads matter. Bisphenol A is under pressure in food contact, which affects can linings more than structures but keeps reformulation work going, and styrene emissions from open-mold polyester are regulated tightly enough that many shops moved to closed molding for compliance rather than for quality. The end-of-life problem is the durable one: a cured thermoset composite has essentially no scrap value, disposal costs money instead of returning it, and the retiring wind-blade fleet has made that a visible line item rather than a footnote.

7 minThermosets vs. Thermoplastics | Polymeric Materials SeriesThe Madison Group · 10k+ views

2 minEpoxy 101: How to Avoid A Runaway Exothermic ReactionSystem Three · 10k+ views

3 minUnsaturated polyester resin curingNouryon · 10k+ views

Materials & Processes: Resin matrices for composites (CompositesWorld) · Occupational exposure to epoxy resins (EU-OSHA)

Polyurethane is one reaction, isocyanate plus polyol, that covers a mattress, a forklift wheel, and the insulation in a refrigerator wall. What changes is the polyol: a long flexible one gives a soft foam, a short stiff one with more crosslinks gives a rigid foam or a hard elastomer. Foam comes from adding water, which reacts with isocyanate to release carbon dioxide, or from a physical blowing agent that boils as the exotherm rises. Flexible foam runs from about 20 kg/m³ up, rigid insulation foam sits at 30–50 kg/m³ with a thermal conductivity of 0.020–0.024 W/m·K, and cast elastomer covers 60 Shore A to 75 Shore D with no gas at all. The rest of the foam world is separate chemistry doing a similar job: expanded polystyrene and polypropylene beads for packaging and impact, and PVC and PET foam boards at 60–200 kg/m³ as structural sandwich cores.

Rigid polyurethane and polyisocyanurate have the lowest thermal conductivity of any common building insulation, roughly 0.020–0.024 W/m·K against 0.033–0.038 for expanded polystyrene and 0.035–0.040 for mineral wool, which is why they win wherever wall thickness is constrained. Cast polyurethane elastomer beats rubber on abrasion resistance and load capacity by a wide margin, which is why heavy wheels, mining screens, and scraper blades are urethane rather than rubber. Flexible foam is simply the cheapest way to fill a volume with something soft. The weaknesses come in three parts. Polyester-based grades hydrolyze in warm humid service and turn sticky and then crumbly. Aromatic isocyanates yellow in sunlight. And the isocyanate itself is a respiratory sensitizer, which has made handling a regulated activity rather than a housekeeping matter, with EU REACH requiring documented training for anyone using diisocyanates since August 2023.

Use rigid polyurethane or polyisocyanurate when insulation performance per inch is the constraint, as in refrigeration, cold chain, and retrofit wall assemblies, and use mineral wool instead when fire performance or vapor openness matters more than R-value per inch. Use flexible foam wherever comfort or bulk cushioning is the job, and check compression set if the part has to hold its height for a decade. Use cast polyurethane elastomer when abrasion and load-carrying beat everything else, which is most industrial wheels and rollers, and use rubber (020) instead when you need low compression set, low heat build-up, or resistance to hot water. Use expanded polystyrene for single-impact energy absorption and packaging and expanded polypropylene when the part has to survive repeated impacts. Use PVC or PET structural core when you need shear stiffness in a sandwich panel, with PET where recyclability or higher process temperatures matter.

Flexible foam from about 20 kg/m³, rigid insulation 30–50 kg/m³ · rigid polyurethane 0.020–0.024 W/m·K against expanded polystyrene 0.033–0.038 and mineral wool 0.035–0.040 · cast elastomer 60 Shore A to 75 Shore D, with roller and wheel compounds usually near 95 Shore A · PVC and PET structural cores 60–200 kg/m³ · expanded polystyrene 15–30 kg/m³ · commodity systems roughly $2–5/kg.

Hydrolysis is the signature failure of polyurethane and it is specific to polyester-based grades: water cleaves the ester link in the soft segment, faster the warmer and wetter it gets, so a shoe midsole or a camera grip that sat in a closet for fifteen years goes tacky and then crumbles under a thumbnail. Polyether-based grades resist that and give up some abrasion and oil resistance for it, which is the trade you are making when you specify one. Flexible foam fails by compression set instead: the cell walls buckle permanently under sustained load, so a seat cushion or a gasket loses height and stops pushing back, and the loss shows up as a leak or a sag rather than a fracture. Rigid insulation foam has a slower one, thermal drift, where the low-conductivity blowing agent diffuses out and air diffuses in over years, which is why insulation is specified on an aged long-term thermal resistance rather than on the value measured fresh off the line. All of these polymer foams also burn fast and give off carbon monoxide and hydrogen cyanide, which is why building codes require a thermal barrier such as gypsum board over exposed foam plastic.

Rigid polyurethane in refrigerator and freezer cabinets, insulated metal panels, spray foam building envelopes, and pipe-in-pipe district heating. Flexible foam in mattresses, furniture, and automotive seating, which is where most polyurethane by volume goes. Cast elastomer in forklift and roller-coaster wheels, skateboard wheels, mining screen panels, print and conveyor rollers, and pipeline pigs. Expanded polystyrene in packaging, bicycle helmet liners, and insulated concrete forms; expanded polypropylene in bumper energy absorbers and reusable crates. Structural cores as Divinycell and Corecell in boat hulls and as PET foam in wind-turbine blade shells and shear webs. The isocyanate producers are Wanhua, BASF, Covestro, Dow, and Huntsman; the core suppliers are Diab, Gurit, and Armacell.

Polyurethane is an isocyanate business, and the isocyanate market has consolidated toward a small number of very large MDI and TDI plants, with Wanhua now the largest producer and BASF, Covestro, Dow, and Huntsman holding the rest. That means system houses, which blend polyols, catalysts, blowing agents, and flame retardants into a two-component kit, capture the formulation value while the base chemistry stays commodity. Blowing agents have been the recurring cost event: CFCs to HCFCs to HFCs to hydrocarbons and HFO blends, each transition forcing reformulation and requalification of thermal and fire performance, and each one driven by ozone or global-warming regulation rather than by performance. Two things are worth watching. Recycled and bio-based polyols are being pushed by furniture and automotive buyers with content targets, and chemical recycling of flexible foam back to polyol has moved past pilot scale without yet being cheap. And in wind, PET core has taken share from PVC core and balsa on cost and process temperature, which is a reminder that in foams the substitution usually comes from a different chemistry doing the same job cheaper rather than from a better version of the incumbent.

3 minBasics of PolyurethaneBASF · 50k+ views

4 minHannah Fry and Dr Anna Ploszajski make Polyurethane foam - BBCBBC · 10k+ views

11 minThe Magic Behind PU Foam Making: How Chemicals Turn into SpongeSkills Creation · 500k+ views

Measuring the R-value of Polyiso Roof Insulation (PIMA) · A Review of Polyurethane Foams for Multi-Functional and High-Performance Applications (Polymers)

Rubber is a polymer above its glass transition, crosslinked lightly enough that the chains can still uncoil and recoil. That crosslinking is vulcanization, and it is what turns a sticky gum into an engineering material and makes the result impossible to remelt. Almost all commodity rubber is one of four polymers picked for a specific reason. Natural rubber and styrene-butadiene rubber go into tires and mounts because they have the best tear strength and fatigue crack resistance available. EPDM goes into weatherseals, roofing, and coolant hose because its backbone has no double bonds left in it, so ozone and UV have nothing to attack. Butyl and halobutyl go into tire innerliners and pharmaceutical stoppers because their gas permeability is roughly a tenth of natural rubber's. The compound around the polymer matters as much as the polymer: 30–50 parts per hundred of carbon black is reinforcement rather than filler, and it multiplies tensile strength several times over while setting hardness and abrasion resistance.

Rubber does things no rigid material does. It recovers from 200% strain, it seals against an imperfect surface, it damps vibration, and natural rubber in particular crystallizes as it stretches, which gives it tear strength and fatigue life that no synthetic matches. It is cheap, at roughly $1.5–2/kg for natural rubber and styrene-butadiene, $3–5 for EPDM, and $3–4 for butyl. The weaknesses are mostly incompatibilities rather than deficiencies. Natural rubber and styrene-butadiene are destroyed by ozone, sunlight, and oil, and they are done by about 80–90 °C. EPDM is excellent in water, steam, and brake fluid and swells uselessly in any hydrocarbon oil. Vulcanization also means every part is a cure cycle in a press, with no regrind and no welding, and the ground-up scrap goes into asphalt and playground surfacing rather than back into parts.

Pick natural rubber or a natural-blend compound when the part flexes for a living: engine and bridge mounts, conveyor belts, truck tire carcasses, and anything where a small cut must not grow into a tear. Pick EPDM whenever the part sees weather, water, steam, or glycol and never sees oil, which covers door and glass seals, roofing membrane, radiator hose, and brake system rubber. Pick butyl or halobutyl when holding gas in is the job. Pick styrene-butadiene when you want most of natural rubber's behavior at a stable synthetic price with better abrasion. Go up to nitrile or FKM (021) the moment oil, fuel, or more than about 120 °C is involved, and go sideways to a thermoplastic elastomer when cycle time, two-shot overmolding, or recyclability matters more than compression set and heat resistance, which is why many glass-run channels and constant-velocity joint boots are now TPV rather than EPDM. Compounding, extrusion, and vulcanization are covered on the manufacturing-processes sheet.

Natural rubber and styrene-butadiene roughly $1.5–2/kg, EPDM $3–5, butyl $3–4 · Shore A 40–95 covers nearly all of it, with tire tread near 60–70 · natural rubber to about 80–90 °C, EPDM to about 150 °C · butyl gas permeability roughly a tenth of natural rubber's · carbon black 30–50 parts per hundred of rubber · compression set under about 25% after 22 hours at 100 °C for a good seal compound.

Ozone cracking is the classic and it needs only parts-per-billion ozone plus a little tensile strain: the ozone attacks the double bonds left in the backbone of natural rubber, styrene-butadiene, and nitrile, and cracks open perpendicular to the strain direction. Antiozonant waxes are compounded in to bloom to the surface and physically block it, which is why tire sidewalls develop a brown film. EPDM and butyl are immune because their main chains are saturated, and that immunity, not any strength number, is the reason they get specified. The second mode is chemical swell, and it is a selection error rather than a wear-out: EPDM in oil takes up so much fluid that it loses shape and strength within days. The third is compression set from heat aging, where sulfur crosslinks rearrange and the rubber takes a permanent shape, so the seal stops pushing back and leaks after a thermal cycle rather than at the moment of installation. Look for surface crazing perpendicular to strain, a part that has grown or gone soft, and a seal that has flattened into the shape of its groove.

Natural rubber and its blends in truck and aircraft tire carcasses, engine and suspension mounts, elastomeric bridge bearings, and heavy conveyor belting. Styrene-butadiene and polybutadiene in passenger tire tread compounds, with silica-reinforced tread the standard low-rolling-resistance formulation. EPDM in automotive door and window seals, single-ply roofing membrane, coolant hose, and high-voltage cable jacket. Halobutyl in tire innerliners and pharmaceutical vial stoppers. The tire makers (Michelin, Bridgestone, Goodyear, Continental) consume most of the world's natural rubber; the synthetic polymers come from ARLANXEO, Dow, ExxonMobil, Lion Elastomers, Mitsui, and SK, and Santoprene TPV is now a Celanese product.

Natural rubber is an agricultural commodity grown mostly by smallholders in Thailand, Indonesia, Vietnam, and Côte d'Ivoire, so its price responds to weather, disease, and replanting cycles with a multi-year lag that no synthetic has, and the EU deforestation regulation now covers it, adding traceability cost to a supply chain of millions of small farms. Synthetic rubber tracks butadiene and styrene, which are cracker co-products, so its availability depends on what the ethylene industry is doing rather than on rubber demand. In a molded part the polymer is usually a minority of the cost, since carbon black, curatives, mixing, and the press cycle dominate, which is why compounders rather than polymer producers hold the customer relationship. Two things are shifting the mix: electrification removes a lot of oil-resistant hose and adds a lot of EPDM for high-voltage cable and thermal management, and thermoplastic elastomers keep taking parts where a cure cycle is the bottleneck.

10 minThe Story of Vulcanized Rubber: Goodyear's Remarkable DiscoveryThe Mat Sci Guy · 50k+ views

3 minWhat Is Vulcanized Rubber - Monroe EngineeringOneMonroe · 10k+ views

9 minSulfur vs. Peroxide Curing in EPDM Rubber: Key Differences, Processing & ApplicationsMy Rubber Heart · 10k+ views

Elastomers: Styrene Butadiene Rubber (SBR), Poly Butadiene, Nitrile Rubber (NPTEL) · Rubber Fatigue Revisited: A State-of-the-Art Review Expanding on Prior Works by Tee, Mars and Fatemi (Polymers)

These are the elastomers you specify when the fluid or the temperature rules out cheap rubber, and the family is best read as a price ladder. Nitrile handles oil and fuel to about 100–120 °C at $3–5/kg and is the default O-ring compound. Hydrogenated nitrile removes the leftover double bonds and buys roughly 150 °C. Fluoroelastomer, usually called FKM or by the Viton trade name, reaches 200–230 °C with excellent fuel and oil resistance at $30–60/kg. Silicone spans -60 to 230 °C and is physiologically inert, but it tears easily and swells in oil. Perfluoroelastomer reaches 327 °C with near-universal chemical resistance at over $1,000/kg, which puts individual O-rings in the tens to hundreds of dollars. Every step up that ladder is bought for a specific incompatibility, and the seal works because the elastomer is squeezed and pushes back, so the property that actually matters is not strength but how much of that push-back survives the service condition.

The strength of the family is coverage: between nitrile, EPDM, silicone, FKM, and FFKM there is a compound for nearly every fluid and temperature combination a machine will see, in standard sizes off a shelf. The weakness is that compatibility is a pairing rather than a rating, and the good compounds have sharp holes in them. FKM is superb in fuel and oil and poor in hot water, steam, amines, and brake fluid; EPDM is the opposite; silicone is inert and clean and has tear strength around 10–30 kN/m, well below a reinforced natural rubber compound, so it fails on installation damage rather than on chemistry. Cost climbs faster than temperature does, roughly ten times from nitrile to FKM and another twenty from FKM to FFKM. And low-temperature performance moves the wrong way as heat resistance improves: a standard FKM stiffens near -20 °C, which is why cold-climate applications reach for a low-temperature grade or for fluorosilicone.

Start from the fluid, not the temperature. If it is petroleum oil or fuel below 120 °C, use nitrile and stop. If it is oil above that, or refrigerant, use hydrogenated nitrile. If it is fuel, engine oil, or aggressive chemistry to 200–230 °C, use FKM, and check separately that it never sees steam, hot water, amines, or glycol brake fluid. If it is water, steam, glycol, or brake fluid, use EPDM (020), which is cheaper and better than anything on this list for those. If it is food, medical, or purely thermal service from -60 °C upward with no oil, use silicone. Reserve FFKM for semiconductor plasma chambers and aggressive chemical process where a failure costs more than the seal by orders of magnitude. And if elasticity is not required, a PTFE or spring-energized seal (017) will beat all of them on chemistry.

Nitrile to 100–120 °C at $3–5/kg · hydrogenated nitrile to about 150 °C · FKM to 200–230 °C at $30–60/kg · silicone -60 to 230 °C with tear strength around 10–30 kN/m · FFKM to 327 °C at over $1,000/kg · squeeze typically 15–30% for a static seal, with gland fill around 75–85%.

Compression set is the mechanism behind most seal failures and it is why a seal that passed on the bench leaks in the field. The elastomer, held squeezed and hot, gradually rearranges its crosslink network into the deformed shape, so the stored elastic force decays; the seal still fills the groove but has stopped pushing on the sealing face, and the leak appears on the first cool-down when the metal contracts and the rubber does not follow. Set rises steeply with temperature, so a compound that shows 15% after 22 hours at 70 °C can show several times that at its rated maximum. The second mode is dimensional change from the fluid, and shrinkage is far more dangerous than swell, because a compound losing plasticizer or extractables to the fluid pulls away from the gland while a swollen one at least stays in contact. The third is explosive decompression in high-pressure gas: gas dissolves into the elastomer under pressure and nucleates into bubbles when the pressure drops quickly, splitting the seal from the inside, which is why oilfield, carbon dioxide, and hydrogen service call out resistant compounds and a maximum depressurization rate. And a large share of what gets logged as seal failure is really geometry, since an extrusion gap that is too wide at pressure, a groove that is over- or under-filled, or a sealing surface rougher than about Ra 0.8 µm will kill a correctly chosen compound.

Nitrile in hydraulic cylinder seals, fuel-system O-rings, and shaft seals across industrial equipment. Hydrogenated nitrile in air-conditioning systems, timing belts, and downhole packers. FKM as Viton, Tecnoflon, and Dai-El in engine and transmission seals, fuel injectors, turbocharger hoses, and chemical pumps. Silicone in food and pharmaceutical gaskets, medical tubing, implantable devices, and oven and appliance seals, with fluorosilicone in aerospace fuel systems for the low-temperature end. FFKM as Kalrez, Chemraz, Isolast, and Perlast in semiconductor etch and deposition chambers and in aggressive chemical process equipment. The polymer producers are Chemours, Daikin, Syensqo, AGC, and ARLANXEO, with DuPont, Greene Tweed, Trelleborg, Parker Hannifin, and Precision Polymer Engineering converting them into seals.

Seals are a small fraction of the cost of the machine and a large fraction of its unplanned downtime, which is why this market prices on consequence rather than on material. An FFKM O-ring at several hundred dollars is trivial against a semiconductor chamber that is down for a shift, and the same logic sets the willingness to pay in oilfield, aerospace, and pharmaceutical service. The value sits in the compound and the qualification, not in the base polymer: manufacturers keep hundreds of proprietary formulations, publish compatibility data, and get designed in by part number, so switching suppliers means requalifying the joint. The live risk is the same one that hangs over fluoropolymers (017), since FKM and FFKM are fluorinated and fall inside the scope of the EU PFAS restriction proposal, and 3M's exit from fluoropolymer and fluoroelastomer manufacturing has already tightened supply. There is no drop-in replacement for FFKM at 300 °C in an etch chamber, so the realistic outcome is a carve-out plus higher prices rather than substitution, but any program that depends on a fluoroelastomer should be carrying that as a named risk.

7 minO-Rings? O-Yeah! How to Select, Design, and Install O-Ring Sealstarkka · 1m+ views

2 minA Guide to Common Seal MaterialsTotally Seals · 10k+ views

7 minWhat is Compression Set?Precision Polymer Engineering Ltd · 5k+ views

Characteristics of Elastomer Seals Exposed to Space Environments (NASA) · Degradation Mechanism of Perfluoroelastomer (FFKM) in the Acidic SC2 Solution of Semiconductor Radio Corporation of America (RCA) Cleaning (Polymers)

Oxide ceramics are metal oxides pressed or molded from powder and fired to near full density, and alumina is most of what actually ships. Depending on purity, a 96–99.5% Al2O3 part runs about 1,500 HV hard and up to 370 GPa stiff, insulates at roughly 15 kV/mm while still passing 25–30 W/m·K of heat, and holds its shape to about 1,600 °C. Zirconia is the other one worth knowing. Yttria-stabilized 3Y-TZP gives up nearly all of that thermal conductivity, running 2–3 W/m·K, and buys 900–1,200 MPa flexural strength and 8–10 MPa·√m fracture toughness, the toughest monolithic ceramic in common use. It gets there by transformation toughening: the stress field ahead of a crack tip flips metastable tetragonal grains to the monoclinic phase, and the roughly 4% volume increase that comes with the flip clamps the crack shut. Both families are shaped green and then sintered, which the manufacturing-processes sheet covers as a process, and the part shrinks 15–20% linearly in the kiln, so any surface that has to hold a fit gets diamond ground afterward. Neither one yields, because the ionic and covalent bonding blocks dislocation motion at room temperature, so the material goes from elastic straight to fractured.

Alumina is the cheapest way to get hardness, electrical insulation, and dimensional stability at temperature in one part, and at $20–60/kg for a finished piece it is affordable enough to design in rather than to argue for. It is inert to almost everything below its use temperature, it does not creep where steel would, and its stiffness barely moves with heat. The weaknesses all come from having no ductility. Tensile strength is a statistic rather than a property, with a Weibull modulus typically 5–15, so nominally identical parts break over a wide spread and the design stress has to sit well under the mean measured strength. There is no crack-tolerant regime either, so a scratch from a steel tool, a chipped edge, or a bolt torqued past spec is a starter crack rather than a blemish. Zirconia adds one of its own: it degrades hydrothermally in 100–200 °C water or steam, where the surface transforms back to monoclinic over months and roughens and microcracks, which makes it a poor choice on the steam side of anything even though its room-temperature numbers look excellent.

Reach for alumina when a part has to be hard, electrically insulating, and stable hot, and you can load it in compression: seal faces, insulators, wear plates, circuit substrates, furnace fixtures, pump components. If the part sees bending or tension, size it against a design stress far below the average measured strength and expect to defend the safety factor, or move to zirconia and pay two to three times more per kilogram for the toughness. Pick zirconia when the duty cycle looks more like a metal's, such as a cutting edge, a small structural pin, or a dental crown, and skip it when the service environment is hot water or steam. If what you actually need is thermal conductivity or thermal shock resistance rather than hardness, this is the wrong family and the answer is silicon carbide or silicon nitride at two to four times the price. And if the part could be hardened tool steel at 60 HRC instead, use the steel: it costs roughly a fifth as much, it tolerates a crack, and it can be cut to size rather than ground.

96–99.5% alumina at roughly 1,500 HV, up to 370 GPa, and 300–400 MPa flexural strength · alumina thermal conductivity 25–30 W/m·K, dielectric strength around 15 kV/mm, useful to about 1,600 °C · 3Y-TZP zirconia at 900–1,200 MPa flexural and 8–10 MPa·√m fracture toughness, but only 2–3 W/m·K · 15–20% linear shrinkage during sintering · Weibull modulus typically 5–15 · finished alumina parts $20–60/kg.

Almost every failure is a brittle fracture from the largest flaw the part happens to carry, and that flaw is usually on a surface someone handled. Because strength is a distribution rather than a number, nominally identical parts break at stresses a factor of two or three apart, which is what a Weibull modulus of 5–15 means in practice and why the allowable sits so far under the average. The second mechanism is thermal shock: quench a hot part and the surface contracts against a still-hot interior, and for alumina the tension that generates exceeds the strength at a ΔT of roughly 200 °C, which is easy to reach by spraying coolant onto a hot fixture. The warning signs are edge chips, tool marks, and grinding scratches, all of which are starter cracks, plus one specific to zirconia: a surface that has gone rough and chalky after months in hot water is hydrothermal aging in progress.

Spark plug insulators from NGK and Bosch are the highest-volume alumina part in the world, at roughly 95% Al2O3. CoorsTek, Kyocera Fine Ceramics, CeramTec, Morgan Advanced Materials, and Maruwa supply most industrial alumina and zirconia components: mechanical seal faces, pump plungers, thread guides, circuit substrates, and semiconductor process-chamber parts. Zirconia ships as fiber-optic connector ferrules by the billion and as dental crowns, with Tosoh supplying much of the world's 3Y-TZP powder, and yttria-stabilized zirconia is the sensing element in every automotive oxygen sensor and the standard thermal barrier coating sprayed onto turbine blades.

Powder is a small part of the cost. Calcined alumina powder runs a few dollars a kilogram and zirconia powder many times that, but a finished alumina part sells for $20–60/kg because the money goes into tooling, firing, and grinding. Diamond grinding after firing is the item that surprises people; on a part with several tight fits it can be most of the piece price, so the useful design skill in this family is knowing which features can be left as-fired. Tooling amortization and kiln time make the cost curve steep in volume and flat over calendar time, since this is a mature industry with no learning curve left to ride. Supply is a handful of Western and Japanese specialists plus a large and improving Chinese base, and moving between them is a requalification rather than a purchase order, because each supplier's powder and firing schedule produce a different flaw population and therefore a different strength distribution. If you are building a business here, the durable asset is process control and the yield data behind it, not the composition.

2 minAlumina Ceramic Producing Process ｜How to Make High Density Alumina Ceramic Piecescsceramic co.,ltd · 10k+ views

9 mintransformation tougheningMSE Frary · 5k+ views

4 minDiamond Grinding & Green Machining of Advanced Ceramics | International Ceramic Engineering ICE | MAPart Gurus · 5k+ views

Brittle Fracture (DoITPoMS, University of Cambridge) · CARES/LIFE Ceramics Analysis and Reliability Evaluation of Structures Life Prediction Program (NASA)

Non-oxide ceramics are carbides and nitrides, and you buy them when alumina runs out of thermal conductivity, thermal shock resistance, or hardness. Silicon carbide carries 120–270 W/m·K depending on how it is densified, with a coefficient of thermal expansion near 4 ppm/K, and that pairing is what lets it shrug off thermal shock: heat leaves fast and the strain per degree is half alumina's, so a temperature gradient generates much less tension. It is also about 2,500 HV hard and 410–450 GPa stiff, which is why it owns mechanical seal faces, kiln furniture, and large lightweight mirrors. Silicon nitride goes the other way, trading some hardness for 800–1,000 MPa flexural strength and 6–7 MPa·√m toughness, which is enough to make it the only ceramic used routinely for rolling-element bearings. Boron carbide is the hardness extreme at roughly 3,000 HV and only 2.52 g/cm³, which is why it is the ceramic in most rifle-rated armor. All of these are strongly covalent and will not sinter on their own, so they are made by reaction bonding, pressureless sintering with additives, or hot pressing, and the route changes the properties as much as the chemistry does.

Nothing else combines this much hardness with this much thermal capability: a SiC seal face runs against another SiC face for years in abrasive slurry, and a silicon nitride ball rolls at 25,000 rpm without the lubrication a steel ball needs. Silicon nitride's low expansion and high strength give it a thermal shock ΔT of 500 °C or more, roughly two and a half times alumina's, and aluminum nitride puts 170–200 W/m·K underneath a power chip while still insulating it electrically. The weaknesses start with price: $50–200/kg for finished parts, two to four times alumina, before the diamond grinding that every tight feature needs. They are still brittle, so the Weibull statistics and the design-in-compression rule from oxide ceramics apply unchanged. And the manufacturing route carries its own limits that datasheets bury: reaction-bonded SiC contains 8–12% free silicon, which softens near 1,400 °C and dissolves in caustic, so a reaction-bonded part fails in service a sintered part would survive.

Pick silicon carbide when the requirement is thermal shock, thermal conductivity, abrasion, or dimensional stability at temperature, and you can load it in compression: seal faces, nozzles, kiln furniture, wafer-handling parts, mirror substrates. Pick silicon nitride when there is real contact stress or bending, which in practice means bearings, cutting inserts for nickel alloys, and glow plugs, and specify it in an EV traction motor's bearings when you need to break the shaft-current path electrically rather than add a grounding ring. Pick boron carbide only when mass matters more than anything else, since it is the lightest hard ceramic and also the most expensive and the most temperamental under high-velocity impact. If your temperature is under about 800 °C and the load is mostly compressive, drop back to alumina and save half to three quarters of the cost. If the part has to survive thermal cycling above 1,200 °C with any tensile load at all, a monolithic ceramic is the wrong answer and you want a ceramic matrix composite instead.

Silicon carbide at 120–270 W/m·K, CTE about 4 ppm/K, roughly 2,500 HV, and 410–450 GPa · silicon nitride at 800–1,000 MPa flexural, 6–7 MPa·√m fracture toughness, CTE about 3.2 ppm/K · thermal shock ΔT roughly 350–400 °C for SiC and 500 °C or more for silicon nitride, against about 200 °C for alumina · boron carbide at 2.52 g/cm³ and roughly 3,000 HV · aluminum nitride at 170–200 W/m·K while electrically insulating · finished parts $50–200/kg.

The brittle statistics are the same as any ceramic, but the practical triggers are different, and most of them are chemical or contact-driven rather than a single overload. A silicon carbide seal face fails when the lubricating film breaks down: friction heating puts a thermal gradient into the face, it heat-checks into a network of radial cracks, and one of them runs. Silicon nitride bearings fail by rolling contact fatigue, where a subsurface inclusion or a machining defect grows a spall over millions of cycles, which is the closest thing this family has to a warning. Two chemical ones catch people out: reaction-bonded SiC loses its free silicon to caustic and to temperatures near 1,400 °C, and aluminum nitride hydrolyzes in water to aluminum hydroxide and ammonia, so an AlN substrate stored or cleaned wet quietly loses the thermal conductivity it was bought for. Boron carbide has its own limit, since above roughly 20 GPa of impact pressure it amorphizes locally and loses strength, which is why the hardest armor ceramic is not automatically the best one against high-velocity threats.

Essentially every industrial pump seal has a silicon carbide face, supplied through John Crane, EagleBurgmann, and Flowserve, with the material from Saint-Gobain (Hexoloy), 3M, Schunk, or Morgan Advanced Materials. Mersen Boostec built the 3.5 m SiC primary mirror for ESA's Herschel telescope and Gaia's SiC optical bench, which is the clearest demonstration of the stiffness-per-kilogram case. Toshiba Materials, CoorsTek, and CeramTec supply silicon nitride bearing balls for machine-tool spindles and for the hybrid bearings now standard in EV traction motors. Boron carbide goes into NIJ Level IV armor plates and into reactor control rods as a neutron absorber. Kyocera, Denka, and Maruwa make the aluminum nitride substrates in power modules and RF packages. Single-crystal SiC wafers for power devices are a different product made a different way, and the semiconductor-manufacturing sheet covers those.

Cost here is dominated by densification and grinding, not by powder. Hot pressing and hot isostatic pressing tie up expensive equipment for hours per batch, pressureless sintering needs tightly controlled powder and additive chemistry, and everything with a tolerance gets diamond ground, so finished parts land at $50–200/kg against $20–60 for alumina. The one part of this family with a real cost curve is aluminum nitride, pulled down by power-electronics volume, and the one with a real growth story is silicon nitride, where EV traction bearings and power-module substrates have both moved from niche to platform in about a decade. Supply is concentrated in a small set of Western and Japanese specialists with Chinese producers moving up the quality ladder, and qualification is slow because the flaw population depends on the specific plant and firing schedule. If you are underwriting a business here, the question is whether the value sits in the powder, the densification know-how, or the finish grinding, because those are three different companies and only the middle one is hard to copy.

2 minAdvanced Engineering CeramicsInternational Syalons · 50k+ views

3 minThe Manufacturing of Silicon NitrideCTL Amedica · 10k+ views

12 minWhy Ceramic Armor? Hard but brittle!Military History Visualized · 100k+ views

Effect of Silicon Nitride Balls and Rollers on Rolling Bearing Life (NASA) · Silicon Carbide Ceramics for Armor Applications: A Review of Sintering Methods and Additive Systems (Molecules)

Technical glass is an amorphous silicate that is transparent, chemically inert, and dimensionally about as stable as any material available, and the property that separates the family is thermal expansion. Soda-lime glass, which is nearly all the glass ever made, expands at about 9 ppm/K. Borosilicate 3.3 drops that to 3.3, which is the whole reason for laboratory glassware and cookware, since a lower expansion coefficient means a smaller stress for the same temperature gradient. Fused silica reaches 0.55 ppm/K and holds continuous service near 1,000 °C, which puts it in furnace tubes, semiconductor process ware, and precision optics. Strength works differently from every other material on this sheet: the Si-O bond is good for several gigapascals, but real glass fails from surface flaws at a few tens of megapascals, so designers work to roughly 7 MPa for annealed glass under sustained load. Everything that makes glass stronger works by putting the surface into compression so those flaws cannot open, which is what thermal tempering and ion exchange do, and the manufacturing-processes sheet covers both as processes.

Glass is transparent from the ultraviolet into the near infrared depending on composition, it resists nearly every chemical except hydrofluoric acid and hot alkali, it does not creep, outgas, or age, and float glass costs $1–2/kg. Borosilicate takes thermal shock that would crack any ceramic in this part of the sheet, and fused silica is the reference material for anything that has to hold a dimension. The weakness is a single one with several faces. Glass has no ductility and a fracture toughness under 1 MPa·√m, roughly a hundredth of structural steel's, so its strength is set by whatever scratch, edge chip, or inclusion happens to be worst. That number is not stable either: under sustained tension in humid air, water attacks the crack tip and existing flaws grow slowly, so a glass part loaded for a year fails at a lower stress than the same part loaded for a minute. Tempering fixes the surface flaws and creates a new failure mode of its own, because once the compressive layer is breached the stored energy dices the whole pane at once.

Specify glass when you need optical transmission, chemical inertness, or dimensional stability, and the part can be designed so no surface sees sustained tension. Choose by expansion coefficient first: soda-lime where temperature is stable and cost matters, borosilicate where the part is cycled or heated, and fused silica where you need deep-UV transmission or an expansion coefficient near zero, at ten to a hundred times the price. If the part has to survive impact or handling, do not design against annealed strength; specify thermally tempered glass for roughly 100 MPa of surface compression, or ion exchange for 700–900 MPa in a 30–50 µm case, which is what a phone cover glass is. Remember that a tempered part cannot be cut, drilled, or ground afterwards, so every hole and edge has to exist before the furnace. If the requirement is impact resistance rather than scratch resistance or optical quality, use polycarbonate or PMMA instead and accept that they scratch, yellow, and expand eight times as fast.

Thermal expansion: soda-lime about 9 ppm/K, borosilicate 3.3, fused silica 0.55 · annealed glass designed to roughly 7 MPa under sustained load, against a theoretical bond strength in the gigapascals · thermal tempering adds roughly 100 MPa of surface compression, ion exchange 700–900 MPa in a 30–50 µm layer · fracture toughness under 1 MPa·√m, against 50–150 for structural steel · continuous service near 500 °C for borosilicate and near 1,000 °C for fused silica · float glass $1–2/kg.

Glass fails in tension from a surface flaw, and the flaw almost always arrives after manufacture: a handling scratch, a chipped edge from a bad cut, or a hard particle dragged across the surface. Because the flaw population sets the strength, measured strengths scatter widely and the design value has to be a low percentile rather than an average, the same Weibull problem the ceramics have with an even lower modulus. Two mechanisms are specific to glass and worth knowing. Static fatigue means a part under constant tension in humid air gets weaker with time, because water reacts at the crack tip and the crack extends subcritically, so a shelf that survived installation can break years later under the same load. And tempered glass carries a rare late failure of its own: a nickel sulfide inclusion left from melting slowly changes phase and expands, and if it sits inside the tensile core it detonates the pane spontaneously, sometimes years after installation, with no impact and nothing to inspect for beforehand.

Corning, Schott, AGC, and Nippon Electric Glass supply most of the technical glass in the world. Corning's Gorilla Glass is the volume ion-exchange product, and Schott's Borofloat and Duran are the reference borosilicates. Heraeus and Tosoh make fused silica for semiconductor process tubes, photomask substrates, and deep-UV optics. Type I borosilicate tubing drawn by Schott (Fiolax), Corning (Valor), and NEG becomes pharmaceutical vials and syringes, which is the segment that ran short during COVID vaccine manufacturing. Architectural tempered and laminated glass comes from Guardian, Vitro, Saint-Gobain, Xinyi, and Fuyao, and glass-ceramic cooktops and telescope blanks are a separate branch of the same family.

Float glass is one of the cheapest engineered materials made, at $1–2/kg, because the process is continuous, enormous, and about a century into its optimization. Cost rises steeply with purity and expansion control: borosilicate tubing runs a few dollars a kilogram, and optical or semiconductor-grade fused silica runs one to two orders of magnitude above that. Float lines and tube-drawing lines are billion-dollar capital assets that cannot be throttled, so the industry is cyclical, regionally traded on freight cost, and consolidated among a handful of firms. Energy is the largest variable cost in melting, which is why European glassmakers were squeezed hard by gas prices after 2022 and why electric and hybrid furnaces are being built now. Two segments have real pricing power: pharmaceutical tubing, where qualification is regulatory and switching suppliers restarts it, and specialty display and optical glass, where composition and forming tolerance are proprietary. Commodity flat glass has neither, and it competes on freight and furnace utilization.

12 minThe Story of Borosilicate Glass: Why Pyrex was SpecialThe Mat Sci Guy · 50k+ views

8 minHow to chemically strengthen glass (eg Gorilla Glass)Applied Science · 100k+ views

2 minHow Is Toughened Glass Different To Ordinary Glass? | BBC Earth ScienceBBC Earth Science · 10k+ views

Influence of Water on Crack Growth in Glass (NIST)

Engineered graphite is made by mixing petroleum coke with coal tar pitch, forming it, baking it near 1,000 °C, impregnating it with more pitch, and graphitizing it at 2,500–3,000 °C. What comes out is the only common material that gets stronger as it gets hotter, gaining roughly a factor of two in strength up toward 2,500 °C, because the internal stresses locked in during cooling relax as the part is reheated. It never melts at atmospheric pressure, it sublimes, so service above 2,500 °C is routine in vacuum or inert gas. The catch is oxidation: in air, measurable mass loss starts around 450 °C and accelerates from there, so the same part that holds 2,800 °C under argon burns away in a furnace with a leak. Isostatic graphite, the fine-grained isotropic grade pressed cold and used for tooling, runs $20–60/kg with a density of 1.75–1.90 g/cm³ and flexural strength of only 40–90 MPa. Carbon's other property worth designing around is extreme anisotropy, since a pyrolytic graphite sheet conducts about 1,500 W/m·K in plane and two orders of magnitude less through its thickness.

Graphite is the cheapest material that works above 1,500 °C, it machines fast with ordinary carbide tooling and no coolant, it conducts both heat and electricity well, it is chemically inert to most molten metals and salts, and it costs a small fraction of the refractory metals it competes with. It is also dimensionally stable, since its expansion coefficient is low and it does not creep the way metals do. The weaknesses are oxidation and mechanical fragility. Flexural strength of 40–90 MPa and a modulus of only 10–14 GPa mean a graphite part cracks if you clamp it wrong, and there is no ductility to warn you. Machining is easy but the dust is abrasive, electrically conductive, and gets everywhere, so shops running graphite need dedicated dust extraction and keep it away from electronics. And in air above about 450 °C the material is consumed rather than degraded, which makes every high-temperature application either inert-atmosphere or a consumable.

Use graphite whenever the job is above roughly 1,200 °C in vacuum or inert gas and the part is not highly loaded: crucibles, hot-zone furniture, sintering dies, susceptors for epitaxy and crystal growth, continuous-casting dies, and EDM electrodes, where it is the default because it erodes slowly and machines into fine detail. If the environment is air above about 450 °C, either accept the material as a consumable and budget for replacement, or specify a silicon carbide coating or an antioxidant treatment. If the part carries real tensile or bending load at temperature, go to carbon-carbon composite, which puts carbon fibers in a carbon matrix and reaches several hundred MPa, at ten to a hundred times the price. If you need the same temperature capability plus oxidation resistance, silicon carbide is the answer, and if you need conductivity near room temperature, copper is four times better and far tougher.

Service above 2,500 °C in vacuum or inert gas, against measurable oxidation in air from about 450 °C · strength rises with temperature, roughly doubling toward 2,500 °C · isostatic graphite at 1.75–1.90 g/cm³, 40–90 MPa flexural, and 10–14 GPa modulus · pyrolytic graphite sheet about 1,500 W/m·K in plane and two orders of magnitude less through thickness · isostatic graphite $20–60/kg · needle coke is roughly 60% of the cost of an electric-arc-furnace electrode.

In air, graphite fails by being consumed. Oxygen attacks at the exposed edges and pores first, so a hot-zone part thins, pits, and dusts before it breaks, and the warning sign is weight loss and dimensional drift rather than a crack. Mechanically it fails brittly and usually at a fixture, because the modulus is low but the toughness is lower, so a bolted or clamped joint that would be routine in metal splits the part along a machining feature. Carbon-carbon adds a third mode: it delaminates between plies, and oxidation attacks the fiber-matrix interface preferentially, so an aircraft brake disc loses strength internally while the surface still looks intact, which is why they are retired on measured thickness and mass rather than on appearance.

Toyo Tanso, IBIDEN, Tokai Carbon, SGL Carbon, Mersen, and Entegris (Poco) supply the isostatic graphite used for EDM electrodes, semiconductor susceptors, and hot-zone furniture. Graphite electrodes for electric arc furnaces are a separate and much larger business, made by GrafTech, Resonac, Tokai Carbon, HEG, Graphite India, and Fangda; GrafTech is the only major producer integrated back into needle coke through its Seadrift plant. Carbon-carbon shows up in aircraft brakes from Safran, Honeywell, and Collins Aerospace, in rocket nozzle throats and exit cones, and in hot-press dies. Pyrolytic graphite sheet from Panasonic and Minerals Technologies spreads heat inside phones and satellites. Battery anode graphite is a different product with a different supply chain, and the critical-minerals sheet covers natural and synthetic anode graphite separately.

The cost of engineered carbon is set upstream by needle coke, which makes up roughly 60% of an arc-furnace electrode's cost and comes from a handful of petroleum and coal-tar refiners. That concentration produces genuine chokepoint behavior: when Chinese capacity closures collided with rising electric-arc steelmaking in 2017 and 2018, electrode prices rose by roughly an order of magnitude within a year, and steelmakers who had treated electrodes as a consumable line item found them setting the cost of a heat. The same coke feeds synthetic battery anode material, so electrodes now compete for feedstock with a much larger and faster-growing buyer. Isostatic graphite is a smaller, higher-margin business dominated by Japanese producers and tied to semiconductor and solar capital cycles, and its capacity additions take three to five years because graphitization furnaces are slow and power-hungry. If you are underwriting anything in this family, check coke supply and electricity price before checking the product market, since those two inputs set most of the cost.

8 minHow Is High-Purity Graphite Made and Where Is It Used?History of Simple Things · 50k+ views

5 minThe 6th element. Isostatic graphite, a key product for the high tech industry.SGL Carbon · 1k+ views

4 minHow we’ve developed Long Life carbon brake 🇬🇧 | SafranSafran · 5k+ views

Graphite Design Handbook (DOE) · Are There Opportunities To Re-Think How We Manufacture Synthetic Graphite? (NETL)

A carbon-fiber composite is 55–65% by volume of carbon filament held in a thermoset resin, usually epoxy, and the fiber does essentially all the work. Standard-modulus fiber runs about 230 GPa and 4,000–4,900 MPa at a density of 1.8 g/cm³, and a laminate built from it lands near 1.55–1.6 g/cm³. A quasi-isotropic layup, meaning plies at 0, ±45, and 90 degrees so the panel behaves the same in every in-plane direction, reaches roughly 390 kN·m/kg of specific strength, about seven times mild steel and nearly twice titanium. The property that people get wrong is stiffness. Unidirectional carbon is genuinely in a class of its own at 85–125 GPa per g/cm³, but a quasi-isotropic laminate lands near 30, only slightly above the 26 that steel, aluminum, titanium, and magnesium all share, so the stiffness win arrives only when the load path is one direction and you can point the fibers along it. Everything gets shaped by layup, winding, or molding rather than by cutting, and the manufacturing-processes sheet covers those processes; here the point is that the process sets the fiber volume fraction, and the fiber volume fraction sets the properties.

Nothing else on this sheet gets close on strength per kilogram, and the axial thermal expansion is near zero, which is why satellite benches and metrology frames are carbon. It also does not fatigue the way aluminum does, so a well-designed carbon structure can run to its design life without the inspection interval an aluminum one needs. The weaknesses are cost, direction, and the failure mode. Standard-modulus 24k–50k tow costs $15–30/kg and aerospace-qualified intermediate-modulus prepreg $60–200/kg, before an autoclave cycle measured in hours, freezer storage at -18 °C, and an out-time clock that starts when the roll leaves the freezer. Properties in the through-thickness direction come from the resin alone, so a laminate is roughly a fiftieth as strong through its thickness as along the fibers, which is where damage lives. And there is no yielding at all: the laminate is linear to failure and then it is in pieces, so every design decision has to be made on a strain allowable rather than on a plastic reserve you can count on.

Use carbon when mass is the binding constraint and someone will pay for it, and when the load path is well enough understood to point fibers along it: aircraft primary structure, satellite and launch structure, pressure vessels, racing, prosthetics, high-end sporting goods. A useful test is cost per kilogram of mass removed. Advanced high-strength steel takes weight out of a vehicle at roughly $2–4/kg saved, aluminum at $8–15, and carbon well above $30, which is why volume cars are still steel and airliners are not. If you are chasing stiffness rather than strength, check whether your layup can actually be unidirectional; if it has to be quasi-isotropic, aluminum with a thicker section is usually cheaper and just as stiff per kilogram. If you need corrosion resistance, radio transparency, or impact tolerance rather than specific strength, glass fiber does the job for a tenth of the fiber cost. And if the part is going to be hit, inspected by non-specialists, or repaired in the field, think hard, because none of those are carbon's strengths.

Quasi-isotropic laminate around 390 kN·m/kg specific strength, roughly seven times mild steel · fiber volume fraction 30–40% for hand layup, 50–60% for infusion, 55–65% for autoclaved prepreg · standard-modulus 24k–50k tow $15–30/kg, aerospace intermediate-modulus prepreg $60–200/kg · autoclave cure measured in hours, prepreg stored at -18 °C with a shelf life and an out-time limit · axial thermal expansion near zero, against 23 ppm/K for aluminum · barely visible impact damage can remove 40–60% of compression strength.

Carbon laminates fail through the thickness, where there is no fiber and only resin. An impact from a dropped tool, hail, or a runway stone drives a cone of delaminations between plies while leaving a surface dent a few tenths of a millimeter deep, and that barely visible impact damage can take 40–60% off compression strength with almost nothing to see from outside. There is no yielding and no slowly growing crack, so the design method is a strain allowable low enough that undetected damage still carries limit load, backed by scheduled ultrasonic or thermographic inspection rather than visual checks. Two slower mechanisms matter as well: epoxy absorbs 1–2% water, which drops the hot-wet glass transition temperature by 20–30 °C and takes the compression allowable with it, and carbon behaves like a noble metal against aluminum, so a carbon panel bolted straight to an aluminum fitting corrodes the aluminum away at the joint unless the fasteners are titanium and a glass ply isolates the faying surface.

The Boeing 787 and Airbus A350 are roughly half composite by weight, which is the reference application. Toray, with Zoltek for industrial-grade large tow, is the largest producer, followed by Hexcel, Mitsubishi Chemical, Teijin (Toho Tenax), SGL Carbon, Syensqo, Hyosung, and a fast-growing Chinese base led by Zhongfu Shenying and Jilin Chemical Fiber. Type IV hydrogen pressure vessels wound from carbon are the fastest-growing non-aerospace use, and SpaceX, Rocket Lab, and most satellite bus builders use carbon primary structure. BMW's i3 passenger cell was the most serious attempt at a volume automotive carbon body, and the industry's retreat from it is the clearest evidence for the cost-per-kilogram-saved argument.

About half the cost of carbon fiber is the polyacrylonitrile precursor and the energy to oxidize and carbonize it, which is why the price has been sticky for two decades despite steady capacity growth: the process is slow, the lines are capital-intensive, and nobody has commercialized a cheaper precursor. Prices softened through 2025 and into 2026 as wind and industrial demand cooled against added capacity, with standard-modulus industrial tow trading in the $15–30/kg band, while aerospace-qualified grades are effectively a separate market priced on qualification rather than on cost. The part cost is where the real money goes: autoclave cycles, tooling, scrap, and inspection typically dwarf the fiber bill, and aerospace prepreg layup can turn a substantial fraction of the material into offcuts. End of life is a genuine liability, since thermoset composites cannot be remelted, pyrolysis recovers short unaligned fiber worth a few dollars a kilogram at best, and most scrap is landfilled or burned. If you are evaluating a carbon-fiber business, the durable positions are precursor and line uptime upstream, and qualified aerospace or pressure-vessel allowables downstream; the middle of the chain is the commodity.

8 minHow Carbon Fiber is Made in Factories | HOW IT'S MADEHow It's Made · 100k+ views

3 minCarbon Fiber: PAN Precursor Carbon FiberInnovations in Manufacturing at ORNL - Archived · 100k+ views

1 minBGA how to inspect a dent for delamination on a composite wing with a tap testgordon macdonald · 10k+ views

Materials & Processes: Fibers for composites (CompositesWorld) · Composite Aircraft Structure, AC 20-107B (FAA)

Glass-fiber composites are what people actually mean when they say "composite" in most industries: boats, wind blades, pipe and tanks, gratings and ladders, truck panels, and enclosures. E-glass roving costs $2–4/kg, about a tenth of carbon tow, and the fiber is strong (roughly 3,400 MPa) but not stiff, at 72–76 GPa against carbon's 230. That single ratio decides where GFRP belongs. A unidirectional glass-epoxy laminate reaches around 1,000 MPa at a density near 1.9–2.0 g/cm³, so specific strength is real, but a quasi-isotropic glass laminate has a specific stiffness below steel's, which makes GFRP a strength, corrosion, and electrical material rather than a stiffness material. Where stiffness is needed anyway, the standard answer is geometry: a sandwich panel with a 60–200 kg/m³ foam or balsa core moves the skins apart and buys bending stiffness that the laminate itself cannot supply. Glass also strains to about 4.5% before it breaks, roughly three times carbon, which changes the whole character of how these parts behave when overloaded.

GFRP is cheap enough to use in tonnage quantities, it does not corrode, it is transparent to radio frequencies, and it can be laid up over a plug in a shed or infused in a 100 m mold with the same basic chemistry. It is far more forgiving than carbon, since the higher failure strain means an overloaded part cracks and whitens visibly instead of shattering, and it does not set up a galvanic cell against aluminum or steel fasteners. The weaknesses are stiffness, temperature, and fatigue. Modulus is low enough that most GFRP designs are deflection-limited rather than strength-limited, which drives thickness and eats the weight advantage. Service temperature is set by the resin, so 100–150 °C for most polyester and epoxy systems. And glass laminates fatigue steadily under cyclic load with no endurance limit, which is the constraint that sizes a wind blade: the design stress for a 20-year, 10^8-cycle life lands at roughly 20–30% of the laminate's static strength.

Choose glass fiber when the job is corrosion resistance, electrical or radio transparency, or a large complex shape at low tooling cost, and stiffness is not the binding constraint: chemical tanks and pipe, gratings and handrails, radomes and antenna housings, boat hulls, pool and utility enclosures, truck and bus panels. If the part is deflection-limited, add a sandwich core before you add laminate thickness, and if that still does not close the gap, put carbon only in the spar caps or the flanges where the strain is highest, which is exactly what large wind blades do. Use vinyl ester or ECR glass when the service is acidic, because standard E-glass is leached by acids and cracks under combined stress and acid exposure. Move to S-glass or aramid when impact or ballistic performance matters. And if the environment is benign and the shape is simple, remember that coated or galvanized steel is usually cheaper per part than any composite, and the corrosion case has to carry the whole argument.

E-glass roving $2–4/kg, about a tenth of standard-modulus carbon tow · fiber strength roughly 3,400 MPa at 72–76 GPa modulus, against carbon's 230 GPa · unidirectional glass-epoxy near 1,000 MPa at 1.9–2.0 g/cm³ · failure strain about 4.5%, roughly three times carbon's · design stress for a 10^8-cycle blade life around 20–30% of static strength · sandwich cores 60–200 kg/m³ · resin-limited service temperature of 100–150 °C.

The impact and delamination story is the same as carbon's but far more visible, because glass strains three times as far before it breaks, so damage usually shows up as whitening and matrix cracking rather than as a hidden delamination. The mechanism that actually retires glass structures is cyclic fatigue at the fiber-matrix interface: each load cycle debonds a little more resin from the fiber surface, matrix cracks link up between plies, and the laminate loses stiffness gradually before it loses strength, which is why blade monitoring watches for a shift in natural frequency. Water is the other slow one. Moisture wicks along the fiber-matrix interface and, in polyester laminates, hydrolyzes the resin into acids that draw more water in, producing the osmotic blisters familiar on older boat hulls. In acid service the glass itself is attacked, with aluminum and calcium leaching out of E-glass and leaving a weakened fiber surface, so a tank that passed its pressure test cracks months later under a load it already carried.

Owens Corning, China Jushi, Taishan Fiberglass, Chongqing Polycomp, Nippon Electric Glass, and Johns Manville make most of the world's glass fiber, and China holds well over half the capacity. Wind blades are the largest single structural use, with Siemens Gamesa, Vestas, LM Wind Power, TPI Composites, and Mingyang building blades past 115 m for the largest offshore rotors. Pultruded profiles from Strongwell and Bedford go into gratings, ladders, and utility crossarms, and filament-wound GFRP pipe from Future Pipe and NOV runs corrosive duty that would eat carbon steel. Sheet molding compound presses truck body panels, electrical enclosures, and EV battery covers in one to three minute cycles. Radomes and antenna housings are almost universally glass or quartz composite, because carbon would short the antenna out.

Glass fiber is a commodity made in continuous furnaces, so the economics look like glass rather than like carbon: energy-intensive, freight-sensitive, cyclical, and priced on utilization. Volumes tell the story, with glass fiber shipping in millions of tonnes a year against carbon fiber on the order of 100,000, and the price gap has been stable for decades. Chinese capacity growth has kept prices flat to falling and squeezed Western producers, so the defensible positions are specialty sizings, ECR and boron-free chemistries for corrosive duty, and local supply for products too bulky to ship. The interesting cost problem is at end of life. Thermoset glass laminates have no resale value, blades have been landfilled or co-processed in cement kilns, and several European countries have banned blade landfilling, which turns disposal into a line item for operators. Vestas and Siemens Gamesa have both announced chemical or recyclable-resin routes, and whether those work at scale is the thing to watch, because the installed fleet coming off warranty over the next decade is very large.

3 minHow to build a wind turbine bladeLM Wind Power · 100k+ views

4 minSouthport Boats Hull Lamination and InfusionSouthport Boats · 50k+ views

7 minWhy Wind Turbine Blades Are So Hard to Recycle | World Wide WasteBusiness Insider · 1m+ views

Analysis of SNL/MSU/DOE Fatigue Database Trends for Wind Turbine Blade Materials, 2010-2015 (Sandia National Laboratories) · Manufacturing Technologies of Carbon/Glass Fiber-Reinforced Polymer Composites and Their Properties: A Review (Polymers)

These are composites with a metal or a ceramic holding the reinforcement instead of a resin, and they exist because epoxy stops working around 180 °C. The important one is the ceramic matrix composite, usually silicon carbide fiber in a silicon carbide matrix, which lands at 2.7–3.1 g/cm³ against 8.2–8.9 for a nickel superalloy and holds 1,200–1,400 °C with far less cooling air. What makes it a composite rather than a brittle ceramic is a thin boron nitride interphase coated onto every fiber: when a crack reaches a fiber it runs sideways along that weak layer instead of straight through, so the matrix cracks in many places while the fibers bridge across and keep carrying load. The result is a stress-strain curve that bends over instead of snapping, which is why a CMC tolerates a bolt hole and a monolithic ceramic does not. Metal matrix composites work on a plainer principle: mix 30–70% by volume of silicon carbide particles into aluminum and the mixture inherits the ceramic's low thermal expansion while keeping most of the metal's conductivity, which gives a power-module baseplate a CTE of 7–9 ppm/K against silicon's 2.6 and copper's 17. Both families are shaped by fiber layup, infiltration, and casting rather than by machining, and the manufacturing-processes sheet covers those; the thing to carry away here is that densification is what you are paying for.

A CMC does the job of a cooled superalloy part at a third the weight and 150–250 °C hotter, and the cooling air it frees up goes back through the core, which is where the fuel-burn gain comes from. It also degrades progressively rather than shattering, so a cracked liner keeps flying. The weaknesses are all cost and chemistry. SiC fiber sells for several thousand dollars a kilogram against $15–30 for carbon, and densification takes days to weeks in a furnace, since chemical vapor infiltration works by diffusing gas into a preform that is closing itself off as it fills. Silicon carbide also does not survive combustion gas on its own: the protective silica scale reacts with high-pressure steam to form volatile Si(OH)4 and the surface recedes, so every hot-section CMC needs an environmental barrier coating, typically an ytterbium silicate, and the coating becomes a second thing that can fail. The design allowable is the proportional limit where matrix cracking starts, roughly 100–150 MPa, not the 250–400 MPa ultimate, so you are buying a fraction of the strength on the datasheet. Aluminum MMCs have a duller problem: the same silicon carbide particles that fix the expansion also chew up carbide tooling, so parts get ground or cut with polycrystalline diamond, and elongation drops to a few percent.

Reach for a CMC when a static hot-section part is temperature-limited and the alternative is spending more compressor bleed air on cooling it: combustor liners, turbine shrouds, and nozzles. Rotating blades are still single-crystal superalloy, because the attachment stresses are tensile and concentrated and CMC design allowables are too low to carry them. If the temperature is under about 1,100 °C, a nickel superalloy is far cheaper and easier to attach, join, and repair. If the part sees high temperature but only compressive load and no thermal shock, a monolithic silicon carbide or silicon nitride does the same job for a tenth of the price. Oxide-oxide CMC, alumina fiber in an alumina matrix, is the sensible middle option below roughly 1,100 °C, because being an oxide already it needs no environmental barrier coating and costs substantially less than SiC/SiC. On the metal side, specify AlSiC when a power module's thermal cycling is cracking solder joints from expansion mismatch and copper is not an option; anywhere the requirement is just heat flow, plain copper is cheaper and conducts twice as well. For structural weight saving with no temperature requirement, use carbon fiber, which beats both of these on specific strength and costs far less.

SiC/SiC at 2.7–3.1 g/cm³, roughly a third of a nickel superalloy's 8.2–8.9 · 1,200–1,400 °C with an environmental barrier coating, against 1,050–1,100 °C metal temperature for a cooled single-crystal blade · proportional limit 100–150 MPa sets the allowable, against a 250–400 MPa ultimate · SiC fiber at several thousand dollars per kilogram, versus $15–30 for standard-modulus carbon · chemical vapor infiltration measured in days to weeks per part · AlSiC baseplates at 7–9 ppm/K and 170–200 W/m·K, against silicon at 2.6 ppm/K and copper at 17.

A CMC does not break in one event; it goes permeable first. Once stress passes the proportional limit the matrix develops distributed cracks that the fibers hold together, and the part still carries its load, but oxygen and steam now have a path to the boron nitride interphase. The interphase oxidizes, the fiber bonds to the matrix, crack deflection stops, and the composite turns brittle at a stress it passed on the test stand, which is why cyclic hot-time matters more than peak stress in CMC life prediction. The other mechanism is surface loss: uncoated SiC in a combustor at 1,300 °C and 10–15 atm recedes on the order of a micron an hour as silica volatilizes into steam, so a spalled environmental barrier coating starts a clock that runs whether or not anything is loaded. Aluminum MMCs fail differently, by fatigue nucleating at particle clusters and at particle-matrix decohesion, so their fatigue strength is a smaller fraction of tensile strength than the unreinforced alloy's, and an AlSiC baseplate usually dies at its solder or sinter joint rather than in the composite itself.

The GE9X carries five CMC parts in its hot section, including the inner and outer combustor liners and the first-stage shrouds and nozzles, and the CFM LEAP's first-stage turbine shroud has been the highest-volume CMC part in service since 2016. GE Aerospace makes finished parts in Asheville, North Carolina and built two adjacent plants in Huntsville, Alabama to make its own silicon carbide fiber and prepreg tape, which is the clearest signal available about where the bottleneck is. Fiber otherwise comes from NGS Advanced Fibers, a Nippon Carbon, GE, and Safran venture that makes Hi-Nicalon, and from UBE's Tyranno; Safran Ceramics, Rolls-Royce, and Pratt & Whitney all run their own CMC programs, and the XA100 and XA102 adaptive-cycle demonstrators use them heavily. Outside turbines, carbon-ceramic brake discs from Brembo and SGL are C/SiC made by the same melt-infiltration chemistry, General Atomics markets SiGA SiC/SiC as accident-tolerant nuclear fuel cladding, and 3M's Nextel oxide fiber goes into oxide-oxide exhaust structures through COI Ceramics. On the metal side, CPS Technologies, Denka, and Ferrotec supply AlSiC baseplates for IGBT and SiC power modules, aluminum-boron carbide sheet such as Metamic lines spent-fuel storage racks as a neutron absorber, and Duralcan Al-SiC brake rotors on the Lotus Elise and GM EV1 are the standing example of an MMC that worked technically and lost on cost.

Two things set CMC cost: fiber and furnace time. Polymer-derived SiC fiber is made in small volumes by a handful of Japanese and American producers at several thousand dollars a kilogram, and chemical vapor infiltration ties up a specialized furnace for days to weeks per load, so the plant's economics are furnace-hours rather than material throughput. GE's prepreg and melt-infiltration route exists mainly to shorten that step, and vertical integration back into fiber is the standard answer to the supply problem. Finished aerospace CMC parts run in the thousands of dollars per kilogram, an order of magnitude above technical ceramics, and the gap does not close with volume the way a molded part's would, because the slow step is chemistry rather than cycle time. That is why this entry carries the top cost band even though its metal-matrix half sits nearer $50–200/kg. Aluminum MMCs are the opposite case: they are cheap enough to be commodity-adjacent, and they stay niche because they compete against copper and direct-bonded aluminum nitride on a cost-per-watt basis rather than on any exotic property. If you are underwriting anything here, the durable positions are fiber production, interphase coating, and qualified allowables for a specific engine; a shop that only lays up and infiltrates someone else's fiber has no moat and no supply security.

9 minGE Aviation and the Ceramic Matrix Composite RevolutionGE Aerospace · 500k+ views

3 minGE's super material: How the CMC process worksGE Aerospace · 50k+ views

3 minLEAP engine - Ceramic Matrix CompositeSafran · 10k+ views

Aerospace Ceramic Materials: Thermal, Environmental Barrier Coatings and SiC/SiC Ceramic Matrix Composites for Turbine Engine Applications (NASA) · Advancements in SiC-Reinforced Metal Matrix Composites for High-Performance Electronic Packaging: A Review of Thermo-Mechanical Properties and Future Trends (Micromachines)

A permanent magnet is specified by two numbers, and everything else is detail. The first is maximum energy product, (BH)max in MGOe, which is how much field energy the magnet can push into an air gap per unit of its own volume: sintered neodymium-iron-boron runs 33–52, samarium-cobalt 20–32, AlNiCo 5–9, and ferrite 3.5–4.5. The second is the temperature it survives, and that one is set by coercivity rather than by melting or softening. Coercivity is the reverse field the magnet withstands before its magnetic domains flip, and in NdFeB it falls roughly 0.5–0.6% for every degree of heating, so a magnet with plenty of margin on the bench runs out of it inside a hot rotor. Substituting dysprosium or terbium for some of the neodymium raises the local anisotropy field and buys that margin back, which is why grade codes carry two parts: N42SH means 42 MGOe and a 150 °C limit, against about 80 °C for a plain N grade. Ferrite behaves the opposite way, since its coercivity rises with temperature, so a ferrite motor is at risk of demagnetizing on a cold start rather than a hot one. Sintered magnets are made by aligning powder in a magnetic field, pressing, and firing, then grinding and plating; the manufacturing-processes sheet covers sintering itself.

NdFeB puts about ten times the field energy of ferrite into the same volume, which is the entire reason a traction motor, a hard-drive actuator, or a cordless-tool motor is the size it is. It gets that field for free once magnetized, with no excitation current and no copper loss, so a permanent-magnet motor typically runs a few points more efficient than an induction motor of the same rating. The weaknesses are supply, temperature, and durability. The heavy rare earths that make a high-temperature grade possible are the most concentrated supply chain on this sheet, and they carry export-license risk described below. NdFeB is also a sintered intermetallic with roughly 75 MPa tensile strength, so it chips in handling and cracks under rotor hoop stress unless it is sleeved or potted. And it corrodes readily, because the neodymium-rich phase at the grain boundaries is anodic to the magnetic grains, so every magnet ships with nickel-copper-nickel plating, epoxy, or zinc, and a scratch in that coating is a real defect rather than a cosmetic one. Samarium-cobalt fixes the temperature and corrosion problems at 20–32 MGOe and a higher price, and ferrite fixes the supply problem at a tenth of the energy product.

Use sintered NdFeB when volume or mass is the binding constraint and the operating temperature stays under about 200 °C: traction motors, servos, actuators, wind generators, headphones, sensors. Pick the grade by the hottest spot the magnet actually sees under the worst duty cycle, not by average temperature, and remember that every step up the coercivity ladder from N to SH to UH to EH costs energy product and money. If the part runs above 200 °C, or sees radiation, or has to hold its flux to a fraction of a percent, use samarium-cobalt instead. If size does not matter, ferrite is still the right answer and is still most magnets by unit count, at $3–6/kg with no supply exposure. Use AlNiCo when you want flux that barely changes with temperature and can live with a magnet that demagnetizes if you look at it wrong, which is the guitar-pickup and instrument case. Use bonded NdFeB, powder in a nylon or epoxy binder at 5–10 MGOe, when the shape is complex or you need a net-shape multipole ring without grinding. If rare-earth supply is your real constraint, the answer is usually a different motor rather than a different magnet, and the robot-actuators sheet covers induction, wound-rotor, and switched-reluctance topologies from the motor-design side.

Sintered NdFeB 33–52 MGOe, SmCo 20–32, AlNiCo 5–9, ferrite 3.5–4.5 · standard N-grade NdFeB to about 80 °C, SH to 150 °C, EH to 200 °C, SmCo to 300–350 °C · NdFeB intrinsic coercivity falling roughly 0.5–0.6% per °C · sintered NdFeB roughly $60–120/kg for common grades in 2026, ferrite $3–6/kg, bonded NdFeB at 5–10 MGOe · NdPr oxide near $100–135/kg and terbium oxide near $800–970/kg through mid-2026 · grain-boundary diffusion cutting heavy rare-earth content by 50–80% for the same coercivity.

The failure that matters is irreversible demagnetization, and it leaves nothing to see. The magnet's operating point sits in the second quadrant of its B-H curve, at a position set by its own thickness and by the permeance of the circuit around it. Heat, an opposing field from stator current, or both together push that point below the knee of the intrinsic curve, at which point some fraction of the domains reverses and does not come back when the magnet cools or the current stops. A motor that has been through an inverter short-circuit fault or a locked-rotor event comes out physically intact with a permanently lower torque constant, and the usual way anyone notices is that the machine draws more current for the same torque. Thin magnets in large air gaps are the most exposed, because a low permeance coefficient puts the operating point close to the knee before any external field arrives. The second mechanism is corrosion: a chipped edge or a scratched plating lets moisture into the anodic grain-boundary phase, the magnet corrodes intergranularly, and it swells and crumbles rather than rusting evenly. Above the Curie temperature, 310–400 °C for NdFeB, the magnetism is gone entirely and only remagnetizing brings it back.

Almost every EV traction motor uses interior-permanent-magnet rotors with sintered NdFeB, and a direct-drive wind generator uses on the order of 600 kg of it per megawatt, which is why those two markets set rare-earth demand. Chinese producers dominate the sintered magnet business through JL MAG, Ningbo Yunsheng, and Zhenghai Magnetics; the non-Chinese incumbents are Shin-Etsu, Proterial, TDK, and Daido Steel in Japan and Vacuumschmelze in Germany. The Western build-out worth watching is MP Materials' Independence plant in Fort Worth, e-VAC Magnetics in Sumter, South Carolina, and Noveon Magnetics in San Marcos, Texas. Magnequench supplies most of the bonded-magnet powder in the world. Samarium-cobalt goes into downhole tools, aerospace actuators, and traveling-wave tubes, ferrite into loudspeakers, holding magnets, and low-cost motors, and AlNiCo into guitar pickups, meters, and sensors. Niron Magnetics is the most credible attempt at an iron-nitride magnet with no rare earths at all, aiming between ferrite and NdFeB on energy product.

Ferrite is most magnets by count and NdFeB is nearly all of them by value, and the NdFeB price is mostly a rare-earth price passed through: a sintered magnet is roughly 30% neodymium and praseodymium by weight, so it tracks NdPr oxide, which ran near $100–135/kg through mid-2026 after a volatile year. Heavy rare earths are the sharper problem, with dysprosium oxide around $210–220/kg and terbium oxide near $800–970/kg over the same period. China holds roughly 90% of sintered NdFeB output and close to 99% of heavy rare-earth separation capacity, and the critical-minerals sheet works through that chain properly. The policy state as of 9 August 2026: China's April 2025 licensing regime is in force and covers dysprosium- and terbium-bearing NdFeB, so shipments of high-temperature grades need an export license, while MOFCOM Announcements 61 and 62 from October 2025, which added a 0.1% de minimis re-export rule and a broader technology-transfer definition, were suspended on 7 November 2025 until 10 November 2026 unless extended. Two engineering responses matter commercially. Grain-boundary diffusion, which pushes dysprosium or terbium in only along the grain boundaries where it does the work, cuts heavy rare-earth content by 50–80% for the same coercivity and is now standard on automotive grades. Designing the cooling so an SH grade works where a UH grade was specified saves real money. If you are building a business on magnets, assume price volatility rather than a cost curve, and price the option of a magnet-free topology before you need it.

5 minRare Earth Magnets - HOW they're madeMP Materials · 50k+ views

5 minMAGNETS | How It's MadeDiscovery UK · 1m+ views

Rare Earth Permanent Magnets: Supply Chain Deep Dive Assessment (US Department of Energy) · Most frequently asked questions about the coercivity of Nd-Fe-B permanent magnets (Science and Technology of Advanced Materials)

Soft magnetic materials are the other half of every motor, transformer, inductor, and solenoid, and they are chosen for how little energy they waste rather than for what field they hold. The specification is core loss in watts per kilogram at a stated flux density and frequency, and it splits into two mechanisms. Hysteresis loss is the area of the B-H loop, paid once per cycle, so it scales with frequency. Eddy-current loss comes from the current the changing flux drives around inside the steel itself, and it scales with the square of both frequency and sheet thickness divided by resistivity, which is the whole reason cores are built from thin insulated laminations instead of solid blocks. A premium 0.35 mm non-oriented electrical steel runs roughly 2.3–3 W/kg at 1.5 T and 50 Hz, commodity grades closer to 4–5; grain-oriented steel, rolled and annealed so the easy magnetic axis lies along the strip, drops near 1 W/kg at a higher 1.7 T but only if the flux follows that one direction; and amorphous ribbon gets to about 0.2 W/kg. Adding 3–3.5% silicon roughly quintuples the steel's resistivity and cuts eddy loss, which is why electrical steel is silicon steel, and above about 3.5% the sheet gets too brittle to punch, which is why it stops there.

Electrical steel is remarkable value: a few dollars a kilogram for a material whose loss number directly sets a machine's efficiency, and it saturates at about 2.03 T, higher than almost anything else you could put in its place. The trade-offs are all about frequency. Eddy loss scales with thickness squared, so an EV traction motor running 800–1,200 Hz electrical frequency needs 0.20–0.25 mm laminations instead of 0.35, and the thinner material costs more per kilogram, punches worse, and stacks with more insulation and less iron in the same axial length. Amorphous ribbon cuts loss by about 70% against grain-oriented steel, but it is 25 µm thick, brittle, saturates at only 1.56 T, and cannot be punched, so it gets wound rather than stamped and the core has to be bigger. Ferrites solve the high-frequency problem by being electrical insulators, which makes eddy loss essentially disappear, and they pay for it with a saturation flux density of only 0.4–0.5 T and a Curie temperature of 150–250 °C. Cobalt-iron saturates highest of all at about 2.35 T and costs $100–300/kg, which confines it to aerospace generators and a few very-high-power-density motors.

At 50 or 60 Hz in a rotating machine, use non-oriented electrical steel and pick the grade by comparing the extra cost per kilogram against the loss you save over the machine's duty cycle; for a motor that runs continuously, a better grade usually pays back, and for one that runs a few hours a week it usually does not. Use grain-oriented steel only where flux stays in one direction, which means transformers and wound cores, since its advantage disappears in a stator where flux rotates. Above roughly 1 kHz, go thinner before you go exotic: 0.20–0.25 mm silicon steel handles most traction-motor work. Above about 20 kHz, switch material class entirely to MnZn ferrite for power conversion up to a few megahertz, NiZn ferrite above that, and nanocrystalline cores where you need very low loss and high permeability in the 20–100 kHz range that fast chargers and solid-state transformers live in. Use soft magnetic composite, insulated iron powder pressed to shape, when the flux path is genuinely three-dimensional, as in an axial-flux motor or a claw-pole alternator, and accept permeability of a few hundred instead of several thousand. And check the boring alternative first: in many machines you cut total loss more cheaply by adding copper and dropping current density than by upgrading the iron.

Premium 0.35 mm non-oriented electrical steel roughly 2.3–3 W/kg at 1.5 T and 50 Hz, commodity grades 4–5 · grain-oriented near 1 W/kg at 1.7 T, amorphous ribbon about 0.2 W/kg · eddy loss scaling with thickness squared, so 0.20–0.25 mm laminations for 800–1,200 Hz traction motors · saturation 2.03 T for silicon steel, 2.35 T for cobalt-iron, 1.56 T for amorphous, 0.4–0.5 T for MnZn ferrite · 3–3.5% silicon, above which the sheet is too brittle to punch · electrical steel a few dollars per kilogram, cobalt-iron $100–300/kg.

Soft magnetic parts rarely break; they quietly stop meeting their loss number. The most common mechanism is cut-edge damage: punching leaves plastic strain and residual stress in a band along every edge, and that stress pins domain walls and raises loss locally. On a wide transformer lamination it barely registers, but on a narrow stator tooth the damaged band is a large fraction of the tooth width, which is why a built motor's iron loss often measures 1.5–2 times the Epstein-strip datasheet value. Stress-relief annealing recovers most of it, and welding or bonding the stack afterwards puts some of it straight back. The second mechanism is an interlaminar short: a burr from a worn punch or a weld bead down the outside of the stack bridges laminations, so eddy currents circulate through the full stack thickness instead of one sheet, and the result is a local hot spot that cooks the insulation around it. Ferrites fail thermally, because permeability and saturation both fall as the core approaches its Curie temperature, so a converter that works cold can drive its core into saturation as it self-heats, at which point current spikes and the heating accelerates. Ferrite is also a brittle ceramic, so an overtightened clamp or a thermal shock cracks a core, and a crack acts as an unintended air gap that changes the inductance the circuit was designed around.

Nippon Steel, JFE, POSCO, Baosteel, ThyssenKrupp, voestalpine, Cleveland-Cliffs, and Tata Steel's Cogent unit supply most of the world's electrical steel, and thin high-grade non-oriented strip for EV traction motors has been the tight product in that market. JFE's Super Core reaches 6.5% silicon by chemical vapor deposition of silicon into the finished strip, which gets around the punchability limit at a price. Metglas, part of Proterial, and Advanced Technology and Materials in China make the amorphous ribbon behind low-loss distribution transformers, which have been deployed at scale in India and China where no-load loss is regulated tightly. Vacuumschmelze's Vitroperm and Proterial's Finemet are the standard nanocrystalline cores for common-mode chokes and high-frequency transformers. TDK, Ferroxcube, Fair-Rite, Magnetics, and DMEGC cover the ferrite market. Carpenter's Hiperco 50 and Vacuumschmelze's Vacoflux are the cobalt-iron grades used in aircraft generators and high-specific-power motors. Höganäs supplies most of the powder behind soft magnetic composite cores.

Electrical steel is a commodity steel product with a specialty tail: ordinary non-oriented grades trade at a modest premium over sheet steel, while thin high-silicon grades and grain-oriented strip command several times that, and the capacity for the thin grades is concentrated in a handful of mills because the rolling and annealing lines are specialized and slow to build. That concentration became the real constraint on EV motor production through the middle of the decade, and it is a capacity story rather than a raw-material story, which means it resolves with capital and time rather than with geology. Ferrite and amorphous are separate businesses with their own dynamics: ferrite is cheap and Chinese-dominated with margin in the specialty grades, and amorphous ribbon has a narrow producer base because casting a 25 µm metallic glass at production speed is genuinely hard. Cobalt-iron carries cobalt price exposure and stays a specialty item. The pattern worth noticing is that this whole family competes against copper and against switching frequency: a designer can spend money on better iron, on more copper, or on faster semiconductors that let the magnetics shrink, and wide-bandgap devices have been steadily pushing that choice toward the third option, which favors ferrite and nanocrystalline over steel.

4 minWhat is Eddy Current Loss & Hysteresis Loss | Simple TutorialElectrical lectures · 50k+ views

2 minStator laminations = more power! (but why?)Stephan_RC · 10k+ views

15 minHow to Identify an Unknown Ferrite CoreFair-Rite® · 50k+ views

Soft magnetic materials for a sustainable and electrified world (Sandia National Laboratories)

These are the materials specified to move heat from a die to the air, and the ladder of thermal conductivity runs further than most people expect: silicone gap pads at 1–6 W/m·K, thermal grease at 3–8, alumina at 25–30, aluminum nitride at 170–200, aluminum at 200, copper at 400, and pyrolytic graphite sheet at roughly 1,500 in plane. What the ladder hides is that the bottleneck is almost never the heat sink. Two machined surfaces pressed together touch on about 1–2% of their apparent area, and the rest is air at 0.026 W/m·K, so the interface material's job is displacing air rather than conducting well in absolute terms. The arithmetic is worth doing once: thermal resistance is thickness divided by conductivity and area, so a 50 µm bond line of 3 W/m·K grease over one square centimeter is 0.17 K/W, while three millimeters of copper over the same area is 0.075 K/W. The 50 µm of paste costs more than the 3 mm of metal. The second constraint is thermal expansion, because a power module stacks silicon at 2.6 ppm/K onto a ceramic at 4.5–7 onto a baseplate, and if that baseplate is copper at 17 ppm/K every power cycle works the solder joint between them.

The engineered members of this family solve the expansion problem without giving up much conductivity. Aluminum-silicon-carbide and copper-tungsten baseplates land at 7–9 ppm/K with 170–200 W/m·K, close enough to the ceramic above them that the joint survives the cycling that kills a copper baseplate. Aluminum nitride puts 170–200 W/m·K directly under a chip while staying an electrical insulator, which is a combination almost nothing else offers. Pyrolytic graphite sheet spreads heat sideways better than copper at a fifth the weight. The weaknesses are specific rather than general. Graphite sheet is roughly a hundred times worse through its thickness than in plane, so it spreads and does not conduct, and it is electrically conductive, so a flake in the wrong place is a short. Aluminum nitride is a ceramic and cracks; it also hydrolyzes if stored or cleaned wet. Beryllium oxide reaches 250–300 W/m·K and has been designed out almost everywhere, because machining or breaking it produces dust that causes chronic beryllium disease, and no thermal gain justifies that handling regime. And the interface materials themselves are the least stable thing in the stack: greases and pads degrade in service in ways their datasheet conductivity says nothing about.

Start by finding out whether your problem is spreading, conducting, or interfacing, because the answers are different materials. If a small die is dumping heat into a large sink, the problem is spreading, and copper, vapor chambers, or graphite sheet fix it. If the stack is already short and metallic, the problem is the interface, and the lever is a thinner bond line rather than a higher-conductivity paste. If the part fails after thousands of power cycles rather than at first power-up, the problem is expansion mismatch, and you want a CTE-matched baseplate and probably sintered silver instead of solder for die attach. Use a gap pad or a dispensed gap filler only where the gap actually varies, as it does between cells and a cooling plate in a battery pack, and accept that a 1–3 W/m·K pad over a millimeter is a large resistance you are choosing to pay for tolerance. Use phase-change pads in place of grease where power cycling is severe, since they stay solid until roughly 45–55 °C and so do not get squeezed out. If none of this is enough, the honest answer is usually active cooling or a bigger die rather than an exotic material, because you run out of conductivity ladder well before you run out of heat.

Gap pads 1–6 W/m·K, grease 3–8, alumina 25–30, aluminum nitride and AlSiC 170–200, aluminum 200, copper 400, pyrolytic graphite sheet about 1,500 in plane · air in an unfilled gap at 0.026 W/m·K, with bare surfaces touching on 1–2% of apparent area · a 50 µm grease line at 0.17 K/W per cm², against 0.075 K/W for 3 mm of copper · CTE 2.6 ppm/K for silicon, 4.5–7 for the ceramics, 7–9 for AlSiC and copper-tungsten, 17 for copper · phase-change pads softening at roughly 45–55 °C · sintered silver die attach around 200 W/m·K against 50–60 for solder.

Thermal materials fail by getting slowly worse, which is why the symptom is a machine that throttles or trips months after it passed its acceptance test. Grease fails two ways at once. Pump-out is mechanical: the die and the heat sink expand at different rates every power cycle, the bond line breathes, and the grease is progressively worked out to the edges until air fills the middle. Dry-out is chemical: the oil that carries the filler evaporates or bleeds into the surrounding material, and what remains is a dry powder with much worse contact. Either way junction temperature climbs a few degrees per thousand cycles until the protection circuit notices. The other mechanism lives one layer down, in the solder or sinter joint under the die. Expansion mismatch strains that joint on every cycle, cracks start at the corners where the strain is largest and grow inward, the conducting area shrinks, the local temperature rises, and the rise accelerates the cracking, which is why power modules are rated in cycles to failure at a stated junction-temperature swing rather than in hours. Ceramic substrates add a third: repeated cycling delaminates the copper from the ceramic at the edges of a direct-bonded copper pattern, and the ceramic itself can crack under the peeled copper.

Honeywell's PTM7950 phase-change film, Shin-Etsu and Dowsil greases, Henkel's Bergquist Gap Pad line, Parker Chomerics, and DuPont's Laird materials cover most of the interface market, with dispensed gap fillers for EV battery packs from Henkel, Dow, H.B. Fuller, and Elkem now the largest volume application by mass. Panasonic's PGS is the pyrolytic graphite sheet inside most phones and thin laptops. Kyocera, Denka, and Maruwa make aluminum nitride substrates, and Rogers' curamik and Ferrotec supply direct-bonded copper on alumina and AlN. CPS Technologies makes AlSiC baseplates, and Plansee and Mi-Tech supply copper-tungsten and copper-molybdenum. Heraeus and MacDermid Alpha sell the silver sintering pastes that have been replacing solder for silicon carbide power devices running 175–200 °C junctions. Element Six supplies CVD diamond heat spreaders for RF power amplifiers, which is the top of the ladder at roughly 1,500–2,000 W/m·K and a price that keeps it in defense and telecom.

Interface materials are a formulated-chemistry business rather than a materials business: the conductivity comes from alumina, boron nitride, or aluminum filler that costs very little, and the value is in loading that filler high enough to conduct while keeping the paste dispensable and stable for ten years. Prices run from a few tens of dollars a kilogram for commodity gap filler to several hundred for phase-change films and sintering pastes, and EV battery packs consume gap filler by the kilogram, which has commoditized the low end fast. The substrate and baseplate side behaves like technical ceramics, with aluminum nitride riding power-electronics volume down a real cost curve and copper-tungsten staying expensive because tungsten powder metallurgy is expensive. The structural shift worth tracking is the move from soldered silicon modules to sintered silver on silicon carbide, which raises junction temperature, makes the expansion mismatch worse, and pulls demand toward CTE-matched substrates and higher-grade interface materials at the same time. If you are selling into this market, the moat is qualification data: nobody designs in a new interface material on its conductivity number, they design it in after thousands of power cycles on their own stack, and that testing takes a year.

12 minThermal Resistance and Heat Transfer in PCB DesignAltium Academy · 10k+ views

4 minWhat happens if you use the WRONG amount of thermal paste?mryeester · 1m+ views

10 mindoes thermal paste ACTUALLY expire?mryeester · 100k+ views

A Survey of Thermal Interface Materials (Sandia National Laboratories) · Fillers and methods to improve the effective (out-plane) thermal conductivity of polymeric thermal interface materials - A review (Heliyon)

This is the one class on the sheet where nobody asks about strength. An optical material is specified by three things: refractive index, usually quoted as n_d at the 587.6 nm helium line; Abbe number, which measures how much that index changes across the spectrum, with high numbers meaning low dispersion; and the wavelength band it actually transmits. N-BK7, the default visible glass, sits at n_d 1.5168 and Abbe 64, and a dense flint like N-SF11 at 1.7847 and Abbe 26. Those two numbers are what makes an achromat work: the focal error a lens introduces from dispersion is proportional to its optical power divided by its Abbe number, so cementing a positive crown element to a negative flint element cancels the color error at two wavelengths while leaving net positive power. Outside the visible the material list changes completely, because ordinary glass stops transmitting past about 2.5 µm: fused silica covers 185 nm to 2.1 µm, sapphire 0.17–5.5 µm, zinc selenide 0.6–16 µm, and germanium 2–14 µm at a refractive index of 4.0, which reflects roughly 36% at every uncoated surface and makes anti-reflection coating mandatory rather than optional. Everything here gets ground and polished, or molded when the tolerance allows; the technical-glass entry covers the underlying glass families and why they behave as they do.

The value of this class is that its properties are specified to four or five decimal places and hold there. Optical glass is delivered to a melt-specific index tolerance, and the low-expansion glass-ceramics, Schott's Zerodur and Corning's ULE, hold near 0 ± 0.02 ppm/K over 0–50 °C, against about 7 for N-BK7, which is what makes a segmented telescope mirror hold its figure while the dome cools overnight. The costs are the obvious kind and one less obvious. Blanks run from roughly $200/kg for common optical glass to several thousand for infrared crystals, and germanium in particular is expensive and supply-exposed. Coating is a separate cost line and frequently exceeds the substrate, especially for ion-beam-sputtered laser mirrors and broadband infrared anti-reflection stacks. The less obvious cost is that many of these materials are physically delicate: calcium fluoride has an 18.9 ppm/K expansion coefficient and cracks from thermal shock, zinc selenide is soft enough to scratch during ordinary cleaning, and the fluorides and halides fog in humidity. Polymer optics invert the whole trade, moldable in a minute at pennies each, but limited to roughly 80–120 °C with eight to ten times the expansion of glass and a temperature coefficient of index near -100 ppm/K against about +3 for glass, so a plastic lens visibly defocuses as it warms.

If the application is visible-band and volume is low, start with N-BK7 or fused silica, and only reach for an exotic glass when the design needs a specific index-Abbe pair that nothing cheap provides. Use fused silica rather than N-BK7 when the wavelength goes below about 350 nm, when laser fluence is high, or when radiation is present. Go to the infrared crystals only when the band demands it, and pick by wavelength and thermal environment together: germanium for 8–14 µm thermal imaging if the optic stays cool, zinc selenide for CO2 laser work at 10.6 µm, sapphire when the window also has to survive sand, rain, or pressure. If the requirement is dimensional stability rather than transmission, as with a telescope mirror or a metrology frame, Zerodur or ULE is what you are buying, and the price is per piece rather than per kilogram. Use molded polymer optics whenever the volume is over roughly 100,000 units, the temperature stays moderate, and you can compensate the thermal defocus in the design, which is why every phone camera is plastic. And if the system can work in reflection, consider a diamond-turned aluminum or nickel-plated mirror instead: it has no chromatic aberration at all and is far cheaper to make in odd shapes, at the cost of more scattered light and more thermal sensitivity.

N-BK7 at n_d 1.5168 and Abbe 64, N-SF11 at 1.7847 and Abbe 26 · fused silica transmitting 185 nm to 2.1 µm, sapphire 0.17–5.5 µm, germanium 2–14 µm, zinc selenide 0.6–16 µm · germanium at n = 4.0, reflecting about 36% per uncoated surface · Zerodur and ULE near 0 ± 0.02 ppm/K over 0–50 °C, against about 7 for N-BK7 · polymer optics at 80–120 °C service, eight to ten times glass expansion, and dn/dT near -100 ppm/K against about +3 for glass · blanks from roughly $200/kg to several thousand.

Optical materials mostly fail by losing transmission rather than by breaking, and the failures are wavelength-specific. Germanium runs away thermally: its absorption rises steeply with temperature above roughly 50–70 °C as free carriers appear, the absorbed light heats it further, absorption rises again, and a germanium window in a hot housing or a strong beam goes hazy and then opaque without any mechanical damage. Coated optics in laser service fail at the coating rather than the substrate, where an absorbing defect or a fingerprint takes up energy locally and ablates the stack above a fluence threshold, which is why damage threshold is a coating and cleanliness specification and why it degrades in a dirty lab. The soft and hygroscopic materials fail environmentally: calcium fluoride and the halides fog or etch in humid air, and zinc selenide picks up scratches from routine cleaning that scatter light into the image. Fused silica in deep-ultraviolet service solarizes, meaning UV photons create color centers that raise absorption over hundreds of hours of exposure. Polymer optics drift instead, absorbing water and creeping until the focus has moved, which shows up as a slow loss of image quality rather than as a fault.

Schott, Ohara, Hoya, and CDGM supply most optical glass, with Heraeus and Corning covering synthetic fused silica and Corning's ULE and Schott's Zerodur splitting the low-expansion mirror-blank market. ESO's Extremely Large Telescope uses 798 hexagonal Zerodur segments about 1.4 m across for its primary mirror, polished by Safran Reosc, which is the clearest demonstration of what near-zero expansion is for. LIGO's test masses are ultra-low-loss fused silica. Umicore is the main supplier of germanium optics and blanks, and Coherent, formerly II-VI, grows most of the world's CVD zinc selenide and zinc sulfide. Sapphire windows and domes come from Kyocera, Rubicon, and Monocrystal. Zeon's Zeonex and Mitsubishi Gas Chemical's polymers are the cyclo-olefin optical plastics behind molded camera and sensor lenses. Thorlabs, Edmund Optics, and Newport are the catalog channel where most engineers actually buy this material, and Zygo makes the interferometers everyone measures it with.

Cost here splits into three independent lines, and people underestimate two of them. The blank is the smallest: common optical glass in the low hundreds of dollars per kilogram, infrared crystals in the thousands. Grinding and polishing is a labor-and-time business priced by surface figure and finish, and it scales badly, since a quarter-wave surface and a twentieth-wave surface differ by an order of magnitude in cost on the same substrate. Coating is the third, often exceeding the substrate for laser mirrors and broadband infrared stacks, and it also carries the lead time. The one genuine cost curve is molding: a precision glass-molded or injection-molded polymer asphere costs a few dollars or a few cents at volume against hundreds for a polished glass asphere, which is why consumer optics moved to molding and scientific optics did not. On supply, germanium is the pressure point, since China produces roughly 60% of it and has restricted exports since 2023, pushing prices up several times over and sending thermal-imaging designers toward chalcogenide glasses and silicon; the critical-minerals sheet covers that chain in detail. If you are building an optics business, the durable asset is polishing and metrology capability rather than material access, because glass is purchasable and figure is not.

13 minHow to make Very Flat Optical Surfaces on GlassHuygens Optics · 1m+ views

4 minThorlabs Optical Lens ProductionThorlabs · 50k+ views

14 minThe Importance of Abbe Value or Number in Lens MaterialsLaramy-K Optical · 10k+ views

TIE-29 Refractive Index and Dispersion (Schott) · A Review of the Precision Glass Molding of Chalcogenide Glass (ChG) for Infrared Optics (Micromachines)