How to measure specific gravity at home
Weigh the specimen in air, then in water, and divide. The specimen must outweigh your scale's resolution by 150 times for quartz and 750 times for galena.
Weigh the specimen dry in air, weigh it again suspended in water, and divide the first figure by the difference. A specimen of 12.40 g in air and 9.28 g in water gives 12.40 divided by 3.12, or 3.97 — libethenite. The method is only as good as the scale: the specimen must outweigh your scale's resolution by about 150 times for a light mineral and 750 times for a heavy one. We quote a measured density for all 108 species this site covers.
In short
- The whole method is one division. SG equals the weight of the crystal in air divided by the weight in air minus the weight in water. No calibration, no reference liquid, no correction — water at room temperature is close enough to 1.000 for this purpose.
- The limit is your scale, and the rule is exact. To land within 0.05 of the true value, the specimen must weigh at least ten times s(2s−1) times the scale's resolution, where s is the density you expect. That is 150 times for quartz and 750 times for cobaltite.
- A 1 g kitchen scale is useless for this. It would need 153 g of a mineral at density 3.0 and 745 g of one at 6.33. A scale reading to 0.01 g brings those down to 1.6 g and 7.5 g.
- Porous and soluble species are excluded outright. Turquoise, chrysocolla, howlite and any zeolite absorb water during the measurement; halite and chalcanthite dissolve in it, and hardness is no guide to which. For those the answer is not a worse measurement, it is no measurement.
- The test needs a clean single-species fragment, which is the constraint that actually defeats most specimens. A druse of small crystals on matrix returns the density of the matrix, and breaking it off costs more than the answer is worth.
| Expected density | Example | 1 g scale | 0.1 g scale | 0.01 g scale | 0.001 g scale |
|---|---|---|---|---|---|
| 2.65 | Quartz | 117 g | 11.7 g | 1.2 g | 0.2 g |
| 3.00 | Magnesite | 153 g | 15.4 g | 1.6 g | 0.2 g |
| 4.00 | Libethenite, olivenite | 284 g | 28.5 g | 2.9 g | 0.3 g |
| 6.33 | Cobaltite | 745 g | 74.5 g | 7.5 g | 0.8 g |
The method, start to finish
You need a scale reading to 0.01 g, a glass of room-temperature water, a length of fine thread or fishing line, and something to suspend the thread from above the scale pan. Nothing else.
1. Weigh the specimen dry, in air. Call this A. Let it settle and take the reading twice.
2. Tie the thread and weigh the specimen suspended in the water, hanging clear of the sides and bottom of the glass, with the glass standing on something other than the scale pan and the thread hanging from a support above it. Call this W. Add one drop of washing-up liquid to the water first: it breaks the surface tension so bubbles do not cling to the specimen, and bubbles are the single commonest source of a wrong answer.
3. Divide. SG = A / (A − W).
A worked example, with real numbers from a real question. A green prismatic crystal weighs 12.40 g in air and 9.28 g suspended in water. The difference is 3.12 g. 12.40 divided by 3.12 is 3.97. That is libethenite, not the olivenite at 4.46 the label said, and the two are indistinguishable by eye.
The physics behind step 2 is Archimedes and nothing more: the apparent loss of weight in water equals the weight of the water displaced, so A − W is the mass of a volume of water equal to the specimen's volume. Dividing gives the ratio of the two densities, which is what specific gravity is.
How good is the answer? The rule, derived
Every guide to this method tells you to use a good scale. None of them says how good, so here is the arithmetic.
Write s for the specific gravity, A for the weight in air, and e for the largest error in a single reading — half the scale's resolution, since that is the worst a correct rounding can be. Propagating both readings through s = A/(A−W) gives a worst-case error of:
error = e × s(2s − 1) / A
Rearranged for a target accuracy of 0.05, and with e as half the resolution δ, that becomes Amin = 10 × s(2s − 1) × δ.
The factor s(2s−1) is the part worth remembering, because it grows fast. At s = 2.65 it is 11.7; at s = 3 it is 15; at s = 4 it is 28; at s = 6.33 it is 74.5. So a heavy mineral needs a much larger specimen than a light one for the same accuracy — roughly six times as much between quartz and cobaltite — which is the exact opposite of most people's intuition, since heavy minerals feel like they should be easier to weigh.
As a rule of thumb: your specimen must outweigh your scale's resolution by at least 150 times for a light mineral and 750 times for a heavy one. The table above gives the figures; the formula gives them for anything not in it.
When not to do this at all
The method has three failure modes, and two of them destroy the specimen rather than merely giving a wrong number.
Soluble species. Halite dissolves. Chalcanthite dissolves. Anything the Handbook records as soluble in water is excluded outright, and the damage is immediate and total. Check before the specimen goes near the glass, not after.
Porous species. Turquoise, chrysocolla, howlite, variscite and the fibrous zeolites all take water into themselves during the measurement. The reading drifts upward as they soak, the specimen may stain or lose cohesion, and the answer is wrong in a direction you cannot correct for. This is why the density test does not settle a turquoise question, which is a real limitation of our own note on telling real turquoise from fake and is stated there.
Mixed specimens. A crust on matrix, a druse of small crystals, an intergrowth of two arsenides — all of these return a weighted average of whatever is in the piece, which corresponds to no species at all. The measurement needs a clean fragment of one mineral, of the minimum mass above, and getting one usually means breaking something.
That last constraint is the honest cost of this method and it defeats more specimens than the scale does. A 0.01 g scale is inexpensive and the rest of the equipment is a glass and some thread. What you cannot buy is a loose 5 g fragment of the species you are asking about, and on a good specimen it is not worth making one. For those, record the uncertainty on the label instead: our note on labelling and cataloguing sets out the certainty states worth using.
Where this test earns its place, and where it does not
Density is decisive on some questions and useless on others, and knowing which in advance saves an afternoon.
It is decisive where two look-alikes differ by more than about 0.3. Molybdenite at 4.62 to 4.73 against graphite at 2.09 to 2.23 — a factor of more than two, resolvable on a kitchen scale. Libethenite 3.97 against olivenite 4.46. Cobaltite 6.33 against pyrite 5.018. The whole silver-white arsenide group, which nothing else separates.
It is useless where the ranges overlap. Scolecite 2.25 to 2.29, mesolite 2.26 and natrolite 2.20 to 2.26 — our scolecite note works through why twinning has to do that job instead. Stilbite against heulandite, for the same reason. Epidote against clinozoisite, except at the ends of the series.
Read the two densities before you weigh anything. If the gap is under 0.1 the test cannot settle it at home, and knowing that in advance is worth more than the measurement would have been. Our collecting notes quote densities on every species page for exactly this purpose.
We are not a laboratory and offer no analytical service. Where a determination genuinely matters and the home test cannot reach it, the answer is a university department or a commercial analyst, not a better kitchen scale. Our wanted list sets out the specimens and documentation we are looking for.