The problem is not the software. It is what a photograph contains.
Mineral identification is a measurement discipline. The properties that settle a species are hardness, streak colour and specific gravity, with cleavage and reaction to acid close behind — and every one of those is a physical test rather than an appearance. A photograph contains none of them.
What a photograph does contain is colour, and colour is the least reliable diagnostic property in mineralogy. This is not a mild caveat; it is the first thing taught in the subject. Quartz is colourless when pure and occurs as purple amethyst, yellow citrine, pink rose quartz, brown smoky quartz, banded agate, opaque red jasper and golden tiger's eye — one species, seven names, identical hardness and fracture. Fluorite occurs in every colour there is.
Meanwhile the reverse also holds. Purple can mean amethyst or fluorite, which differ by three whole points on the Mohs scale and by a factor of many in value, and are separated in one second by a steel knife.
Colour is the only property a photograph reads well, and it is the one that proves least.
Where an identifier is genuinely useful
Habit and lustre are visual, and an app reads them well. A brassy cube is pyrite; a six-sided prism with a pointed termination and striated faces is quartz; a botryoidal green mass is likely malachite; radiating white spherulites in black glass is snowflake obsidian. Those are real identifications from real visual evidence.
The other genuine value is vocabulary. Mineralogy is a naming discipline and the hard part of researching an unknown stone is not the research — it is not knowing the words. An app that hands you 'botryoidal chalcedony' or 'bladed gypsum' has given you the search you could not otherwise have run.
So the correct use is as a hypothesis generator. The app proposes; the tile and the coin dispose.
The three tests, and the kit you already own
None of this requires buying anything. Four hardness references are already in your house, and the other two tests need a ceramic tile and a bottle of vinegar.
- Hardness: a fingernail is about 2.5, a copper coin 3.5, a steel knife or glass plate 5.5, a steel file 6.5. Work upwards and bracket it. Wipe away the powder before you decide — soft minerals leave dust that looks exactly like a scratch.
- Streak: drag the specimen hard across the unglazed back of a tile and read the powder colour, not the surface colour. Hematite streaks red-brown however silver it looks. Pyrite streaks greenish-black; gold streaks yellow.
- Specific gravity: weigh it dry, weigh it suspended in water, divide the dry weight by the difference. Amber is 1.05, opal 2.1, quartz 2.65, pyrite 5.0, gold 19.3.
- Cleavage: find a broken edge. Flat mirror faces meeting at consistent angles means cleavage; curved shell-like surfaces mean none. Quartz never cleaves; calcite, fluorite and feldspar always do.
- Acid: one drop of vinegar. Carbonates fizz, silicates do not. Watch for slow fine bubbles under a lens rather than expecting foam.
Have I found gold? Almost certainly not, and here is the proof
This is the most common question in the subject and the one where the gap between photograph and reality is widest, because pyrite and gold share the one property a camera sees — a brassy metallic colour — and differ on almost every property it cannot.
Streak the specimen on an unglazed tile. Gold leaves a yellow metallic smear; pyrite and chalcopyrite leave greenish-black powder. Then press a pin into an inconspicuous grain: gold is malleable and dents and smears, pyrite is brittle and crumbles into angular grit. Then weigh it — gold's specific gravity is 19.3 against pyrite's 5.0, which is a difference you can feel in your hand.
Pyrite also forms cubes, and gold does not form cubes in nature at any size you will find in the field. A perfect brassy cube is pyrite, and it is a nicer object to own than a flake of gold.
The materials where treatment matters more than species
There is a second category of limitation, separate from species identification, and it is where the money actually goes wrong. Six materials carry nearly all the fraud in this field, and in every case the fraud is the same: a cheap material dyed, stabilised, reconstituted or moulded to pass as an expensive one.
Turquoise is imitated with dyed howlite, which is hardness 3.5 and white beneath its blue skin. Malachite is imitated in pigmented resin, which weighs a third as much and does not fizz in acid. Lapis is imitated with dyed jasper, which has no pyrite. Amber is imitated with plastic, which sinks in saturated brine where amber floats. Jade is substituted with serpentine, which is much less dense. And 'magnetic hematite' is a manufactured ferrite, exposed by the fact that real hematite will not hold a paperclip.
Every one of those frauds is aimed at appearance, which means every one of them defeats a photograph and none of them defeats a knife, a tile and a cup of water.
How to photograph a stone so the reading is worth something
Given the limits, the captures that help most are the ones that show structure rather than colour.
- Clean it first, and photograph it dry — wetting a stone changes its apparent colour and lustre.
- Indirect daylight, no flash. Flash flattens lustre, which is one of the few useful visual properties.
- A plain mid-grey background, so the camera does not colour-balance against something bright.
- One frame of any crystal faces, square on, to show habit and striations.
- One frame of a broken edge, to show cleavage or conchoidal fracture.
- Something for scale in the frame — a coin works, and doubles as a hardness reference.