EN

Calculate Rock Mineral Percentages (Point Counting)

On a thin section under a polarizing microscope, mineral volume is not measured directly. Points are counted on a regular grid, and the mineral under each station is recorded. The Delesse principle states that the fraction of points landing on a mineral is an unbiased estimate of its volume fraction in the whole rock — provided the sampling is uniform and dense enough.

The count for each mineral (up to six) is entered below. The calculator returns the volume percentage of each phase and flags whether the total is dense enough to be reliable.

The same arithmetic applies to measured areas (mm² per mineral) instead of points: switching the unit below leaves the percentage formula unchanged.

Count per mineral

Minerals counted (two minimum, blank rows are ignored)

Terms marked with an asterisk are defined in the glossary at the end of the page.

Mode versus CIPW norm: two mineral lists, two jobs

A mode* is the mineral list actually present in the rock, by volume: quartz 24%, alkali feldspar 31%, and so on, counted on a thin section* or a stained slab. A CIPW* norm is a different list, rebuilt from the whole-rock chemistry as if the magma had crystallised anhydrous, at equilibrium, into a short set of ideal end-members.

CIPW will report hypersthene and olivine where the slide shows hornblende and biotite, because the calculation does not spend water. Modes feed QAPF* names and the color index. Norms serve glassy volcanics, rocks whose primary crystals are gone, and suite comparison in geochemistry when a count is impossible. They are not two readings of the same quantity. Averaging them is not a method.

Why a point fraction estimates a volume fraction (Delesse)

Delesse showed in 1848 that, on a random section through a polyphase aggregate, the area fraction of a phase equals its volume fraction. Rosiwal turned the idea into linear intercepts. Glagolev and Chayes turned it into a grid of points. This calculator is the last of those steps: points (or mm²) over total, times 100.

%vol(i) ≈ points(i) / points(total) × 100

The estimate is unbiased if the section is a fair slice of the rock and the grid is uniform. A layered gneiss counted on one leucosome band is a count of the band, not of the unit. A single 30 µm slice through a 2 cm megacryst is a lottery ticket. For an ordinary isotropic granite the Delesse* principle holds, which is why petrographers still sit at the stage and click.

How many points to count: binomial error and the 300-point flag

For a phase at proportion p in N independent points, the binomial standard deviation is:

σ = √[p(1 − p) / N]

A mineral at 30 vol% : 100 points → σ ≈ 4.6%; 300 points → ≈ 2.6%; 1,000 points → ≈ 1.4%; 2,000 points → ≈ 1.0%. Accessories at 3% stay noisy until N is large. That is why Van der Plas & Tobi charts exist, and why this page flags totals under 300.

A trained counter finishes 1,000–2,000 points in well under an hour on a coarse granite. Fine-grained volcanics take longer because every station is a decision. The useful stop is when the rock name is stable, not when the third decimal looks scientific.

Under 300 points, a minority phase at 10% has a 2σ band that can swallow a QAPF boundary. The major mineral at 40% is already usable. A 200-point teaching count still separates granite from gabbro; it does not separate granodiorite from granite, and it does not fix biotite to a percent.

Zircon, or the 4% muscovite that decides two-mica versus biotite granite, needs a count past 1,000 — or an honest “present” instead of a fake mode. The form will still divide whatever integers are typed. The flag exists because the arithmetic never refuses a bad sample size.

Cutting the thin section and setting the grid

A standard rock thin section is a chip sawn, one face lapped, stuck to glass, ground to about 30 µm — quartz goes first-order grey-white under crossed polars, and that tint is the thickness gauge — then covered. The grinding recipe is not needed to count. A section of the right thickness, without pluck pits later called “pores” or “opaques”, is.

Grid spacing should be of the same order as the typical grain, or a little larger. Points that all land inside one crystal are correlated; they pretend to be a larger N than the rock has given. On a 2.5 × 4 cm granite section, 1 mm steps already give a thousand-odd possible stations. On a microlitic andesite, 0.3–0.5 mm is the range that still samples the groundmass. The whole slide is traversed, not the pretty corner.

Quartz, feldspar and mica under crossed polars

Plane-polarised light first, then crossed polars. The usual plutonic cast:

  • Quartz — colourless, low relief, no cleavage, first-order grey, undulose extinction, no twins.
  • Alkali feldspar — cloudy more often than quartz; tartan twinning if microcline, simple twins if orthoclase, perthite lamellae when they are there.
  • Plagioclase — polysynthetic albite twins; zoning if the magma did not sit still.
  • Biotite — brown/green pleochroism, bird’s-eye extinction, one perfect cleavage.
  • Hornblende — green to brown, 60°/120° cleavage, moderate relief.
  • Olivine — high relief, high birefringence, no cleavage, often a serpentine mesh.

A stained slab (HF etch, sodium cobaltinitrite for K-feldspar, a red stain for plagioclase) stops the pink-versus-white guess on the hand specimen. The microscope still owns quartz versus untwinned K-feldspar when both are clear.

From the mode to a rock name: the QAPF diagram

Mafics and opaques are dropped. Quartz (Q), alkali feldspar (A), plagioclase (P) and feldspathoid (F) are recast to 100%. If F = 0, the point sits on the QAP triangle. IUGS / Streckeisen fields, plutonic, quartz 20–60%:

  • P / (A+P) < 10% — alkali-feldspar granite
  • 10–35% — syenogranite
  • 35–65% — monzogranite (with syenogranite, “granite” in the broad sense)
  • 65–90% — granodiorite
  • > 90% — tonalite

Q between 0 and 20% is the quartz-syenite / quartz-monzonite / quartz-monzodiorite band. Q = 0 is syenite–diorite–gabbro depending on P/(A+P). F > 0 leaves the quartz triangle for the foid-bearing fields. Color index does not move the QAPF point; it says whether that granite is a leucogranite or a hornblende–biotite granite. The mafic prefix is added after the root name.

Heterogeneity, operator bias and altered grains

Three ways to lie with an honest grid.

  • Heterogeneity. Flow banding, schlieren, a pegmatite seam, a chilled margin: one section is not the rock. Two sections, or a note of which fabric was counted.
  • Operator bias. Distinctive minerals — biotite books, quartz with undulose extinction — are overcounted; cloudy feldspars drift into whichever bin was used last. Two operators on the same slide, or a recount a week later, is the calibration that matters. Distinctive is not abundant.
  • Alteration. Chlorite after biotite, sericite after plagioclase, serpentine after olivine: the secondary is counted as itself, or the primary is reconstructed from habit — not both conventions in one table. Grain-pluck holes are not opaques.

On a porphyritic rock, a grid that is fine for the groundmass will park a dozen consecutive points in one phenocryst and inflate that phase. Counting more on the same grid does not fix it.

Automated counting: image analysis and SEM maps

Pixel classification on stitched PPL/XPL mosaics is fast once trained, and it is repeatable in a way a tired Friday afternoon is not. It still inherits the stain, the thickness, and the training set. Quartz versus untwinned K-feldspar remains a fight in unstained light.

SEM*–EDS (or QEMSCAN-style) maps chemistry onto every grain. Fine-grained basalts, hornfelses, and anything with clay rims belong here. Delesse is then done with a mineral map, not a mechanical stage. The hitch is cost, and the temptation to publish three-decimal modes from a 4 mm² field. Area still has to represent the rock.

Worked example: 1,000 points on a granite

Holocrystalline plutonic section, 1,000 stations:

  • Quartz 240
  • Alkali feldspar 310
  • Plagioclase 280
  • Biotite 110
  • Hornblende 40
  • Opaques and accessories 20

Whole-rock modes: Q 24.0%, A 31.0%, P 28.0%, M (bt+hbl+opq) 17.0%. For QAPF, the 170 mafic-and-opaque points are dropped and 830 recast:

Q′ = 240/830 = 28.9%  ·  A′ = 310/830 = 37.3%  ·  P′ = 280/830 = 33.7%

P / (A+P) = 280/590 = 47.5%  ·  Q′ = 28.9% (in the 20–60% band)

That is monzogranite (granite sensu stricto). Dark minerals total 17% — leucocratic, a biotite–hornblende granite on CI. 1,000 points is enough to sit inside the granite field with room. It is not enough to fight over the syenogranite/monzogranite line if P versus A is 2% off.

Volume percent is not weight percent

Modes are volume. XRF majors and CIPW are weight. Magnetite at 5 vol% (ρ ≈ 5.2 g/cm³) in a granite framework (ρ ≈ 2.65 g/cm³) is nearer 9–10 wt%. Quartz at 30 vol% is a bit less than 30 wt% once biotite is in the mix. Convert before claiming that “the norm and the mode agree”.

wt%(i) = vol%(i) × ρi / Σ (vol%(j) × ρj) × 100

Grain density of the rock should match the mode-weighted mineral densities, plus a porosity correction if bulk rather than grain volume was measured.

Phenocrysts: counting a porphyritic rock without inflating the mode

On a slide with 8 mm feldspars in a 0.2 mm groundmass, a 0.5 mm grid is a phenocryst sampler with a groundmass hobby. Two honest methods: phenocryst area (or a coarse grid) over a large slab first, then the groundmass on a fine grid, combined with the phenocryst mode; or a very large N on a very large area, living with the variance. A 400-point traverse that happened to start on a megacryst is not a mode.

Volcanic rocks with a glassy mesostasis: glass is a phase. It is named, counted, and put in the table. Redistributing it into “normative feldspar” is a CIPW job, not a point count.

IUGS names on BGS and USGS maps

IUGS / Streckeisen is the international root-name system. British Geological Survey maps of the Scottish Caledonian granites, and USGS maps of the Sierra Nevada batholith, still carry older field language: “granite” for a wide QAP band, “adamellite” on some older BGS sheets for what is now monzogranite, “quartz monzonite” in much of the western United States for a rock a European petrographer would call granite. The mode does not change. The published name might.

A Cornish two-mica granite and a Sierra biotite granodiorite can share a field notebook heading and sit in different QAPF polygons. Recasting Q, A and P from this calculator, then reading the IUGS field, is how the two literatures are lined up. The petrography page is the right place to keep both the count and the map name, rather than forcing one onto the other.

Stained slabs for coarse granite

A polished slab, HF, sodium cobaltinitrite (K-feldspar goes yellow), a red stain for plagioclase, and a point count with a binocular or a transparent grid, is often better than one thin section for QAPF of a coarse granite. More crystals are seen. Quartz stays unstained and low-relief. Mafics stay dark. Undulose extinction and small accessories are lost; representativity is gained.

The slide still finishes the accessories, the myrmekite, the An-zoning. Both belong in the folder. The calculator does not care whether the “points” came from a mechanical stage or a slab grid. Delesse does not either.

Glossary

Mode
The mineral composition actually present in the rock, expressed as volume percent from a point count or an area measurement — as opposed to a norm calculated from chemistry.
Thin section
A rock slice ground to about 30 µm, mounted on glass, and viewed in polarised light. At that thickness quartz shows first-order grey, which doubles as a thickness check.
CIPW
A normative calculation (Cross, Iddings, Pirsson, Washington) that recasts a whole-rock analysis into an ideal anhydrous mineral list. Useful when crystals cannot be counted; not a substitute for a mode.
QAPF
The IUGS / Streckeisen double triangle used to name igneous rocks from quartz (Q), alkali feldspar (A), plagioclase (P) and feldspathoid (F), recast to 100% after mafics are removed.
Delesse principle
On a random section through a polyphase rock, the area fraction of a mineral equals its volume fraction. Point counting and Rosiwal intercepts are practical forms of the same statement.
SEM
Scanning electron microscopy. Paired with energy-dispersive X-ray spectrometry (EDS), it maps chemistry grain by grain and can yield a modal mineral map on fine-grained rocks.

Related calculators