How to Choose Carbide End Mills by Workpiece Material

A 10 mm, 4-flute carbide end mill running 80 m/min in 304 stainless fails in a predictable order: the edge rubs before it shears, the surface it just cut work hardens from roughly 180 HV toward 300 HV or more, built-up edge grows on the rake face and pulls coating and edge material off with it, flank wear triples, and the pocket held to 0.10 mm goes out of flat. The same cutter, in the same machine, on a tougher fine-grain grade at full engagement and 46-107 m/min per published data, finishes the pocket with a straight edge. That is a selection failure, and selection starts at the workpiece, because the material dictates the grade group, coating chemistry, flute count, helix, edge preparation, and coolant decision that follow in every section below.

Six workpiece families, one decision table, and the vendor source row behind every speed band: public standard values or published cutting data cited as recorded, nothing interpolated, with the spindle calculation and the five selection mistakes that show up most in the quote fixes we are asked to review.

The Snapshot

• Material comes first, geometry second: carbide milling grades are grouped against ISO 513 application families (P steel, M stainless, K cast iron, N non-ferrous, S superalloys and titanium, H hardened, same letters our speed calculator sorts every row by), and a grade chosen for the wrong group fails at any speed.
• Steel wants hardness plus a tough finish: fine-grain AlTiN/TiAlN-class grades in the published 110-290 m/min band for alloy steel at 800-1100 N/mm2, flood coolant on interrupted cuts.
• Austenitic stainless (304/316, roughly 180 HB) runs 46-122 m/min in published end-mill data; the constraint is work hardening plus adhesion, not abrasion, so toughness and a smooth coated edge beat wear resistance.
• Cast iron runs 150-250 m/min (indexable published range), dry or air blast: coolant cycles the edge thermally and turns graphite dust into grinding paste.
• Aluminum (wrought, 60 HB) publishes 375-430 m/min on solid carbide to 1500 m/min indexable; polished non-stick flutes, 2-3 flutes at 45-55° helix, sharp edge, uncoated, or diamond where high-Si cast alloys abrade plain carbide.
• Titanium and superalloys are the low-speed regime: 46-107 m/min published on Ti-6Al-4V end mills, 60-125 on HRSA, generous feed at small radial engagement, always wet, never dry.
• Hardened steel 45-55 HRC runs 30-107 m/min published on short, rigid carbide, climb direction only; 55 HRC is about 600 HB on the ASTM E140 conversion printed in any vendor grade selector.
• The machine is the hidden fourth axis: every published range assumes a rigid, short, true-running setup, so start at the bottom of the band beyond about 3x stickout or measurable backlash.

The Four Knobs Every End Mill Ships With

An end mill is a substrate, a coating, a geometry, and an edge preparation. The workpiece sets all four, which is why "which end mill for steel" has no single answer: a continuous finishing cut in 1045 and an interrupted roughing cut in 4140 both read "steel" and want different settings on each knob.

Substrate. Tungsten carbide grains in a cobalt binder. Fine grain sizes (roughly 0.5-1.0 µm) trade hardness for transverse rupture strength, what interrupted cuts and hard machining need; coarse, low-cobalt grades resist abrasion, what cast iron and high-silicon aluminum need. The ISO 513 P-M-K-N-S-H groups exist because grade recipes are tuned per group, not per brand.

Coating. This is where the whole coating story lives in this guide; the material sections below just apply it. TiN tops out around 500-600 °C; TiAlN/AlTiN PVD coatings run into the 800-900 °C band per PVD supplier data sheets, which is why steel, stainless and hardened work are AlTiN/TiAlN territory: in those families the edge contact temperature is the limiting factor, and the coating's job is hot hardness, not lubricity. Two chemistry exceptions matter. Diamond and DLC coatings have no affinity for aluminum, so built-up edge cannot anchor on them, but diamond is carbon and reacts chemically with ferrous materials at cutting temperature, so a diamond-coated cutter never goes in steel. CVD alumina-rich coatings carry thicker, tougher layers for the turning-insert and cast-iron domain, and their process heat rounds sharp edges, which is why sub-12 mm solid end mills are almost all PVD.

Geometry. Flutes carry chips: fewer flutes means fewer, larger gullets. Helix pulls the cut in smoothly and shears: 30-35° general purpose, 45-55° the aluminum finishing band, variable helix breaks chatter harmonics in stainless and titanium. The core adds or removes stiffness: a ball nose buys its form by sacrificing core, which is why the same diameter ball nose cannot rough like a square end mill.

Edge preparation. The knob most buyers never see listed: a honed (rounded) edge land of 0.01-0.10 mm, sometimes a TCKK or negative chipbreaker land. A sharp, positive edge cuts soft and sticky materials; a supported edge survives the hammering of cast iron scale and interrupted steel cuts. Choosing sharp versus supported is choosing between aluminum and steel: opposite answers to opposite problems.

How to Read a Material Before You Read a Catalog

Before touching a catalog, answer four questions about the workpiece, in this order. These four collapse the catalog from hundreds of SKUs to a handful.

1. What group is it, by hardness and chemistry? Plain carbon and alloy steel is P. Austenitic stainless is M. Grey, ductile and compacted graphite iron is K. Aluminum, copper, brass, magnesium is N. Titanium, Inconel, HRSA is S. Anything hardened above 45 HRC is H. A material can sit in two rows (a precipitation-hardened stainless looks like M at solution age and like S at peak age); the group decides the grade family, not the marketing name on the box.

2. How hard is it, in one number you can verify? Brinell (HB), Rockwell C (HRC), or Vickers (HV). If the mill certificate gives HRC and the vendor table gives HB, convert with the ASTM E140 table rather than eyeballing it: 55 HRC is about 600 HB, 45 HRC about 448 HB, and published grade boundaries are drawn in those numbers. The hardness conversion chart runs the same table the other way, from HV readings on a finished surface back to the HRC the tool was chosen for.

3. Continuous or interrupted? A continuous finishing cut allows the high end of every speed band and a harder, more abrasive grade. An interrupted cut (keyways, holes, scale pockets, cast skin) is a hammering test: drop to the low end of the band, move to a tougher finer grain, thicken the edge land. This is the axis most selection errors hide on, because two buyers with the same part name often disagree about it.

4. What can the machine actually hold? Published ranges assume rigidity: stickout under about 3-4x diameter, a true-running holder (ER collet under 0.01 mm TIR, hydraulic or shrink-fit preferred), no table lurch. Manual or older CNC machines with backlash, or deep pockets needing long reach, take the low end of every range in this article, climb direction where the axis is tight, conventional on scaly stock. A 250 m/min published figure on a flexing machine shows up as chatter at 250 and a broken cutter at 280, and neither is the tool's fault.

Write the four answers as a one-line spec ("M, 304, ~180 HB, continuous pocket with 2 cross-over slots, rigid machine, hydraulic holder") and the rest of this article becomes a lookup.

Steel: The P Group, Where Most Selection Error Starts

Plain carbon and alloy steels are the largest family in the end mill catalog and the group most often mis-cut by over-soft or over-hard choices. Free-machining steels with sulfur or lead additions (12L14 class) break chips easily and run at the top of the steel bands, so treat them as the easy case. The selection work is in low-carbon through quenched-and-tempered alloy at 800-1100 N/mm2.

The published numbers: indexable turning data records 210-395 m/min for low-carbon and free-machining steel, and the solid-end-mill side records 110-180 m/min for alloy steel at 1100 N/mm2 and 160-250 m/min for the 800 N/mm2 class (vendor steel-milling tables, SFM columns converted as recorded). To start a roughing program in 4140 pre-hard (roughly 28-34 HRC): 160-200 m/min with 0.06-0.10 mm per tooth on a 4-flute carbide sits inside every published band rather than on its edge.

Grade choice within P is a toughness-versus-hardness dial. Continuous light finishing: a harder, more abrasive-resistant grade, high end of the band. Heavy interrupted roughing: tougher fine grain, edge land 0.10 mm plus, 20-40% slower. The mistake that eats budget is running it backwards: a hard finishing grade into an interrupted cut micro-chips within the first pocket, and slowing the feed to "fix" it makes chips too thick for the gullets, which accelerates the chipping. Coating follows the rule from the knob section above: AlTiN/TiAlN-class PVD for anything you would call steel cutting, because coating buys hot hardness, not a higher feed; chip load still comes from substrate toughness and rigidity, which is why two identical-looking TiAlN 10 mm cutters from different grade lines can carry a 2x difference in feed per tooth.

Direction is the remaining P-group decision, and it is made per operation, not as a blanket setting: climb is the default on rigid CNCs, and the scale, backlash and interrupted-edge exceptions with their force-mechanics numbers are worked out in the climb versus conventional guide.

Stainless: Work Hardening Is the Constraint

Stainless is not "hard steel." Austenitic 304/316 arrives at roughly 180 HB, softer than many P-group steels. The problem is its behavior under strain: the surface hardens toward 300-350 HV where it has been deformed but not cut, and the nickel-chromium matrix welds itself to any hot, clean surface, including your cutter. So the end mill for stainless is selected to fight adhesion and rubbing, not abrasion.

Three properties, in order: a smooth, well-prepared edge (any rubbing phase work-hardens the skin the next tooth must cut, a self-feeding loop), a tough fine-grain substrate that resists micro-deformation at the edge under drag, and the smooth AlTiN/TiAlN-class PVD coating from the knob section. Published solid-carbide end-mill data: 304-class austenitic on variable-geometry cutters records 46-107 m/min (the 150-350 SFM column, column mapping verified by x-coordinate in the source), 400-series ferritic/martensitic 91-122 m/min (300-400 SFM). The pattern that surprises people: the "easier" 400 series takes the higher number because it does not gall and weld the way 304 does.

Feed per tooth in stainless cannot be cheated: under about 0.04 mm per tooth on a 10 mm 4-flute, the edge rides the work-hardened skin instead of cutting below it, and the tool rubs, glazes, and dies. If the machine cannot carry the load, take smaller radial engagement and more depth, not a smaller feed. Radial engagement on 304 stays at or below about 8-12% of diameter at full axial immersion: the tool engages, shears, and gets out. Coolant is a working condition, not a luxury: flood or through-spindle aimed at the engagement, because its job is cooling the weld zone as much as rinsing chips, and the galling tendency rises with temperature. The one-line rule for visible built-up edge on stainless: it is almost always too low a speed, an edge honed rounder than it should be, or dry cutting where wet belongs.

Cast Iron: Abrasive Dust, Thermal Cracking, No Coolant

Grey, ductile and compacted-graphite iron are abrasive in a way steel is not: graphite flakes and hard carbide phases lap the flank, and the machining is dry and hot. The K-group wear mechanism is flank abrasion, so the grade wants the opposite of the stainless prescription: hard, cobalt-poor, wear-resistant, often CVD-coated for the thick alumina layer (per the coating chemistry above). Published indexable milling data records grey iron at 150-250 m/min and ductile slightly under that.

Coolant on cast iron is the trap. Hot cutting zone then cold jet micro-cracks the carbide surface; flakes and dust slur into a grinding paste that strips the coating faster than dry air ever would. The correct practice is dry, or air blast for chip evacuation in deep pockets. If the shop insists on wet for part cleanliness, run the jet wide-open and continuous, never spray-pause-spray: it is the cycling that kills the edge, not the water.

Geometry: the supported edge wins here, a wiper flat (TCKK) or robust positive geometry on wear-protective grades; for solid end mills, 2-3 flutes at 30-45° helix clear powdery chips without packing gullets. High-silicon aluminum castings sit on the same abrasion mechanism and pull the grade choice toward the K side, the crossover visible row by row in our speed table. One boundary check before you buy: a scale-covered steel casting labeled as iron is a surface condition, not a material, and the first pass over scale runs conventional for the impact-protection reasons in the direction guide.

Aluminum: Welded Built-Up Edge Is the Enemy

Aluminum cuts at 10-20x steel speed because it shears easily and carries heat away with the chip. Soft 6061-class (about 60 HB) machines like plastic; 7075-T6 and high-silicon casting alloys are the abrasive end of the family. The default failure mode in the soft end is welded built-up edge: aluminum smears onto the rake face, grows past the chip thickness, tears off with coating or edge material attached, and the next half-second of the cut happens with that rubble riding the surface, leaving the finish you are rejecting the part for.

Published solid-carbide data (vendor aluminum-mill tables, recorded as parsed): wrought non-hardened 60 HB at 375-430 m/min slotting/roughing columns (1230-1411 SFM), casting alloys up to 12% Si at 200-230 m/min, above 12% Si at 160 m/min; indexable carbide dry to 1500 m/min on the pure-aluminum side. Feed per tooth wants 0.05-0.15 mm, more than most operators feed in steel, because the chip must be thick enough to carry the heat out and leave the gullet before it welds.

The tool is built for evacuation and non-stick, and this is where the coating rule reverses: uncoated or polished bright flutes (TiAlN wants to bond to aluminum, the exact problem), sharp near-zero hone (a rounded edge squeezes and smears), 2-3 flutes, 45-55° helix, generous primary clearance. Diamond or DLC coating where free silicon in high-Si cast alloys acts as a grinding wheel against plain carbide; PCD tooling for those alloys holds order-of-magnitude life gains and sits on the superhard tooling line rather than in the 375 m/min question. For the specific searches: 5-flute aluminum cutters are for deep-slot and long-reach work where gullet volume trades away and core stiffness beats chatter; 3-flute at 45° is the general pocket-and-profile answer; 4-flute is a small-diameter exception (under about 3 mm), not a productivity upgrade. The full chip-load math behind 2 versus 4 flutes gets its own decision article; inside aluminum, count gullet volume, not flutes.

One arithmetic trap: 400 m/min on a 10 mm cutter is 12,732 rpm, which a 3,000 rpm head cannot reach; on small machines the answer is a smaller diameter tool at the same surface speed, the spindle-RPM rearrangement built into the calculator, not running aluminum at half the published band with a 10 mm tool.

Titanium and Superalloys: The Low-Speed Regime

Titanium and nickel superalloys are S-group by physics: low thermal conductivity means the heat stays in the cut instead of leaving with the chip, and the chemistry has affinity for tool materials. The whole regime lives between about 25 and 125 m/min. Published data: Ti-6Al-4V on variable-geometry solid carbide records 46-107 m/min (150-350 SFM column), indexable on titanium alloys 30-50 m/min, generic titanium from heavy roughing to light finishing 25-120 m/min, HRSA on indexable carbide 60-125 m/min. At 10 mm and 80 m/min the spindle answer is 2,546 rpm, and if that feels alarmingly slow after stainless work, the surface speed is the story, not the spindle counter.

Grade follows the coating chemistry of the knobs section with one titanium-specific twist: many vendor tables recommend uncoated tough fine grain for Ti precisely because AlTiN-family coatings have metallurgical affinity for titanium at cutting temperature and can weld to the chip and tear off with it; AlCrN-class or uncoated are the coated-safe answers, and superalloys take tough fine grain uncoated or AlCrN. Every S-group row sits at a toughness cliff, so the tool must be the stiffest thing in the system: shortest possible stickout, rigid holder, no tapered contact anywhere.

Strategy: keep a constant, generous chip load (0.03-0.08 mm per tooth published starting field on carbide end mills, more on roughers at small engagement), wide axial and 10-15% maximum radial engagement in finishing, climb direction, and never stop feeding inside the cut. Coolant at high pressure, through-spindle if available, always. Dry cutting at reduced speed is not a safer titanium process: the tool stays in contact longer per unit removed and oxidation runs regardless. The three visible failure signatures are diagnosis shortcuts: rubbing means chip load below the band; built-up-then-pulled means the speed is wrong for the grade-coating pair; notch wear at the depth line means toolpath and coolant aim, not grade.

Hardened Steel Above 45 HRC

Hard machining is a real process with published bands of its own: fine, tough carbide with a prepared edge covers roughly 45-60 HRC, CBN takes over above that. The published end-mill numbers: hardened steel 45-50 HRC on solid carbide records 53-69 m/min (the 175-225 SFM column), a 35 HRC tool-steel class at 46-107 m/min, hardened under 54 HRC on indexable carbide 30-40 m/min. Use the conversion chart to put the mill certificate and the table in the same units: 45 HRC is about 448 HB, 55 HRC about 600 HB, and the P/M-to-H boundary on grade selectors sits near 45 HRC.

The knob settings: micro-grain substrate for edge strength under the hammering of a hardened cut, a hot-hard AlTiN-class coating (the 800-900 °C band from the knobs section is exactly where hardened turning and milling contact temperatures live), edge land 0.05-0.10 mm. Geometry: 2-4 flutes, short engaged length, small radial stepover (0.1-0.2 mm finishing on ball nose and tapered radius tools), high axial with small radial, stickout under 3x diameter. Direction is fixed: climb only, because the conventional entry-rub phase smears and work-hardens an already-hardened surface, which is the mechanism laid out in the direction guide. Where the drawing demands ground surface integrity above 60 HRC, the end mill becomes a support tool and the process conversation belongs to CBN turning or grinding.

The Decision Table: Material Family at a Glance

One row per family, every setting the four knobs take when the material is the only thing you know. Starting speeds are the published bands, sources in the next section's table: the middle of a band on a rigid machine, the bottom third when machine or holder is in question. The grade/coating column names families, not part numbers, because the point is to make the catalog readable, not to replace it.

← swipe to scroll →

Published Speed Starting Points, With Sources

The audit trail for every range used above: each row is taken as recorded from vendor cutting-data publications parsed into the site's data foundation, converted to m/min only where the source printed SFM, nothing interpolated. Where a source gives one column for slotting and another for roughing, the row says so. Use any number from this page with its row attached.

← swipe to scroll →

To turn any band into the two numbers a control asks for: spindle rpm = 1000 × vc / (π × D), feed = rpm × flutes × feed per tooth. A 10 mm end mill in 304 at 80 m/min is 2,546 rpm; at 0.05 mm chip load on 4 flutes, about 509 mm/min. The cutting speed calculator carries these fields precomputed for every published row, with its own source note per row.

Five Selection Mistakes That Cost More Than the Cutter

1. Choosing the grade for the machine instead of the material. The machine sets the ceiling (rigidity, rpm, coolant); the material sets the grade, coating, edge and band. A rigid machine does not make a steel grade cut aluminum cleanly. If tool problems keep getting solved by changing machine settings first, the selection is usually the real cause.

2. Copying a parameter from a friend who "runs it fine." His band transfers only if his material spec, rigidity, holder, stickout and coolant aim match yours, across all four questions above, which they rarely do. Ask him for the source row, not the tuned number: the recorded band is transferable, the tuned S-parameter is not.

3. Too many flutes for the chip you are making. Aluminum and deep-pocket steel both want gullet volume; 4 and 6-flute cutters pack, rub and burn. The right count is the one whose gullets stay clear at the feed you can actually run.

4. Running stainless or titanium dry at reduced speed. Slowing down does not remove the adhesion loop in the M-group or the heat-concentration problem in the S-group; it just extends contact time per unit removed. These families are wet processes with generous chip loads.

5. Ignoring stickout and holder as selection variables. Published bands are measured in short, stiff test setups. A long-reach cutter in a worn ER collet at 5x stickout is not the machine the table describes: halve the engagement before touching the band. The holder belongs in the selection, not the accessory list; holder types are cataloged on their own page.

A sixth if you want the meta-mistake: treating selection as one-time. The tool that works on the prototype aluminum bracket is wrong for the high-Si production casting; re-ask the four questions whenever the material spec line moves, even when the drawing number has not.

Frequently Asked Questions

Q1. Which end mill for aluminum?

Uncoated or polished-flute fine-grain carbide, 2-3 flutes, 45-55° helix, sharp edge, starting 375-430 m/min on wrought 60 HB stock per published solid-carbide data, with a generous 0.05-0.15 mm feed per tooth so chips carry the heat out. On high-silicon casting alloy, move to diamond/DLC coated or an abrasion grade at 160-230 m/min.

Q2. What coating is best for stainless steel?

A smooth, tough PVD TiAlN/AlTiN-class coating over a fine-grain substrate, run inside the published 46-107 m/min band so the edge shears instead of rubbing. The coating is the smaller half of the answer; speed, toughness and edge prep do the work.

Q3. Square end mill or bull-nose for a 3-axis pocket?

Square (flat, 90° corner) where floors and walls must be square and true; bull-nose (corner radius) for contoured finishing, because its contact arc allows larger stepovers at equal cusp height and its corner survives deep-pocket engagement better than a square corner cutting full depth.

Q4. How do I turn a speed into spindle rpm and feed?

rpm = 1000 × vc / (π × D), feed = rpm × flutes × fz. Example: 10 mm tool, 80 m/min, 4 flutes, 0.05 mm gives 2,546 rpm and about 509 mm/min. The calculator page carries this math with every published row precomputed.

Q5. Can one grade cover both steel and aluminum in a mixed shop?

General-purpose PVD fine grains do both at reduced parameters, but nothing does both at published speeds: the sharp edge aluminum wants chips in steel, and the hard steel coating welds to aluminum. Two short lists per group beat one long list of compromises.

Q6. Can hard milling really replace grinding under 55 HRC?

To about 55 HRC, yes, on a rigid machine: short micro-grain coated carbide, climb only, small radial stepover, 30-107 m/min published bands, finishing allowances of 0.1-0.3 mm and Ra in the 0.4-0.8 µm range are routine. Above 60 HRC, or where the drawing demands ground surface integrity, CBN turning or grinding owns the process.

The Bottom Line

Read the workpiece, not the catalog. The material family sets the grade group, coating chemistry, flute count, helix and edge land all at once, and each of those is a published number in a supplier's cutting-data table before it is a sales page. Run the four questions (group, hardness, continuity, machine capability), take the row from the decision table and the source row behind its band, and calculate the spindle numbers once instead of tuning them forever.

A wrong end mill is a one-tool problem; a wrong material spec is a whole-lot problem, and that is the line the quote answers we send back are built on. InsertCore sources carbide end mills by material group against these published bands and replies with the grade family, coating and the source reasoning behind the choice, saying so where the data stops being honest.