How to test mineral hardness properly
Test on the base, wipe the mark, and look again. Kyanite is hardness 5.5 along the blade and 7 across it, so direction matters as much as the mineral does.
Test on the base or on matrix, never on a display face, and wipe the mark before deciding. Kyanite is hardness 5.5 along the crystal and 7 across it, so direction changes the answer. Hardness narrows a field of mineral species; it rarely closes a determination, and it will not separate pyrite from marcasite at all.
In short
- Half of all apparent scratches are metal, not damage. A steel point drawn across a harder mineral leaves a grey trail of steel. Wipe it with a damp finger: if it comes off, the mineral is harder than the tester.
- Hardness can vary with direction on the same crystal. The Handbook of Mineralogy gives kyanite as 5.5 parallel to [001] and 7 parallel to [100] — a whole point and a half, on one blade, testable with a steel point.
- Test on the base, in one short stroke, under magnification. A hardness test is a permanent alteration to the specimen, so it goes where nobody will look and it happens once.
- It will not separate the pairs you most want separated. Pyrite and marcasite are both 6 to 6.5. Brochantite and malachite are both 3.5 to 4. Cerussite and anglesite are within half a point. In each case the answer lies elsewhere.
- Know your testers’ real values. A fingernail is about 2.5, a copper coin about 3.5, a steel point or knife about 5.5, window glass about 5.5, a streak plate about 6.5, and a steel file about 6.5.
| Tester | Approximate hardness | Marks these | Does not mark these | Watch out for |
|---|---|---|---|---|
| Fingernail | ~2.5 | Gypsum, talc, halite, torbernite, molybdenite | Calcite and everything above it | Testing on a soft display face. Use the base |
| Copper coin | ~3.5 | Calcite, bornite, cerussite, crocoite, barite | Fluorite and above | Modern UK coins are plated steel, not copper. Use a pre-1992 penny |
| Steel point or knife blade | ~5.5 | Fluorite, apatite, brochantite, goethite at the low end | Feldspar, quartz and above | The grey metal trail. Wipe before judging |
| Window glass | ~5.5 | The same range, from the other direction | — | Scratching glass proves hardness above 5.5, not the mineral's identity |
| Streak plate | ~6.5 | Everything below quartz | Quartz, topaz, corundum | This is a streak test that also gives hardness information |
| Steel file | ~6.5 | Pyrite, marcasite, feldspar | Quartz and above | Files vary. Calibrate yours against a known quartz |
The method, and why the wipe matters more than the scratch
1. Choose the spot before you choose the tester. The base, the back, an area of matrix, or an already-damaged edge. Never a termination, never a display face, never the part of the specimen anyone will photograph. A hardness test is permanent.
2. One short, firm stroke. Not a sawing motion and not a long drag. Two millimetres is plenty.
3. Wipe the mark with a damp finger or a soft cloth, then look under a lens. This is the step that is skipped and it is the step that decides the result. If the mark wipes away, it was metal from the tester and the mineral is harder. If a groove remains with displaced material at its edges, the mineral is softer.
4. Bracket rather than guess. Find the softest tester that marks it and the hardest that does not. “Marked by a steel point, not by a copper coin” is a real result — somewhere between about 3.5 and 5.5 — and it is more honest than a single number.
5. Record the result, including the direction. On a catalogue card: not marked by steel point on the base, 12 September 2026. That is a determination someone can check. The certainty conventions are on labelling and cataloguing a collection.
Direction changes the answer, and kyanite proves it
Mohs hardness is treated as a property of a mineral. On some species it is a property of a mineral in a direction, and the difference is large enough to test at a table.
The Handbook of Mineralogy gives kyanite’s hardness as 5.5 parallel to [001] and 7 parallel to [100]. That is a point and a half of difference on a single blade. A steel point at about 5.5 will mark a kyanite blade along its length and fail across it, and the same blade will therefore appear to be two different minerals depending on which way you drew the point. Our page on kyanite is built on exactly this.
Kyanite is the extreme case but it is not the only one. Phosgenite’s hardness is recorded as 2 to 3, varying with direction. Micaceous and platy species behave differently across a cleavage plane from along it. And any species with a perfect cleavage will appear softer when the test happens to run along the cleavage direction, because you are not measuring hardness at all — you are opening a cleavage plane.
The practical rule: if a hardness result surprises you, repeat it in a different direction before you believe it. And if the two directions disagree, that disagreement is itself diagnostic information worth recording. Whether you have opened a cleavage or made a scratch is covered on cleavage, parting or fracture.
Where hardness does not help, with three worked examples
Hardness is the most over-used test in mineral identification, and knowing where it fails is worth more than knowing how to do it. Three cases from our own species pages, all with the figures from the Handbook of Mineralogy.
Pyrite against marcasite: both 6 to 6.5. Same formula, FeS2, same hardness, streaks too close to call, densities of 5.018 and 4.887 — a 2.6% difference. Hardness contributes nothing. The separation is habit: pyrite is cubic and makes cubes, pyritohedra and octahedra; marcasite is orthorhombic and makes tabular crystals, cockscomb ridges and spearhead twins. See marcasite.
Brochantite against malachite: both 3.5 to 4. Densities 3.97 and 4.05. Both give a pale green streak. Three standard tests, three identical answers. The separation is chemical: malachite is a carbonate and effervesces in cold dilute acid, brochantite is a sulphate and does not. See brochantite.
Cerussite against anglesite: 3 to 3.5 against 2.5 to 3. Overlapping. Densities 6.55 and 6.38, both adamantine, both pale, both in the same oxidised lead zone. Again the acid test settles it, because cerussite is a carbonate and anglesite a sulphate. See cerussite.
The general lesson is that hardness is a filter, not a determination. It reduces a field of hundreds to a field of dozens in a few seconds and at almost no cost, which is genuinely valuable. It closes a case only when combined with habit, streak, density, cleavage, associated species and locality — the sequence set out on identifying a mineral specimen.
The tester problems nobody mentions
Your copper coin is probably not copper. UK pennies and twopences struck from 1992 onwards are copper-plated steel. Draw one across a specimen and you are testing at steel hardness with a thin copper skin, which gives an unpredictable and usually too-hard result. Use a pre-1992 coin, or a piece of copper wire, and check by trying it on a known calcite.
Steel varies. Knife blades, needles and files run from about 5 to well over 6.5 depending on alloy and heat treatment. Calibrate your own testers once, against known specimens, and write the results down. A file that will not mark a known quartz is above 7 and is a different tool from one that does.
Glass is not a fixed number either. Ordinary window glass sits near 5.5; borosilicate and toughened glass are harder. And scratching glass tells you the mineral is harder than that piece of glass, which is a lower bound and nothing more.
Weathered surfaces read soft. An oxidised, altered or clay-coated surface can be a point or more below the fresh mineral underneath, and testing it gives the hardness of the alteration product. Find a fresh area, or clean one gently first — see removing clay from a specimen.
And some specimens should not be hardness-tested at all. Anything soft and fine-crystallised, anything you intend to sell, anything with intact terminations and no hidden surface, and anything toxic where the test would generate dust. On those, the answer is to use every other test first and accept an undetermined identification if they do not settle it. A correct “undetermined” is worth more than a confident guess with a scratch across the front of the specimen. The rest of the collecting guides cover the alternatives, and the wanted list is where to tell us what you are hunting for.