The grain size number, explained

Every steel test certificate seems to carry it: "Grain size: G 7.5, ASTM E112." It is one of the most-quoted numbers in metallurgy — and one of the most misunderstood. Here is what G actually says, where it comes from, and what quietly corrupts it in practice.

Bigger number, finer grain

The grain size number is logarithmic and inverted: every step up in G means roughly twice as many grains in the same area — so grains half the average sectional area. G 5 is coarse enough to resolve comfortably at low magnification; G 10 is a fine-grained structure with over thirty times as many grains per unit area.

The classical ASTM definition ties G to the number of grains seen at 100× magnification:

In metric terms, working from the number of grains per square millimetre at 1× (N_A), E112 gives the equivalent relation:

ISO 643 defines its grain size index the same exponential way, from the grain count per square millimetre. In routine work the two scales agree closely, which is why certificates often cite both.

Why anyone cares

Grain size is one of the few microstructural parameters that moves strength, toughness and formability at the same time. Finer grains mean more grain-boundary area obstructing dislocation motion — the familiar Hall–Petch trend of higher yield strength — and generally better toughness at low temperature. That is why so many material specifications quote a minimum (or a range of) grain size number, and why auditors look hard at how the number was produced.

Two roads to G

The planimetric (Jeffries) procedure

Count grains inside a known area: grains fully inside count as one, grains cut by the boundary of the measurement region count as half. Divide by the true area — in calibrated square millimetres — to get N_A, and convert to G with the formula above. Simple, visual, and sensitive mainly to whether every boundary is actually revealed.

The intercept (Heyn) procedure

Lay test lines of known total length over the structure and count how many times grain boundaries intersect them. The mean lineal intercept ℓ̄ — line length divided by intersection count — characterizes the grain size directly, and E112 relates it to G:

The intercept method is generally quicker to apply consistently, and with directed test lines it handles elongated, worked structures better than area counting.

What corrupts the measurement

• Wrong calibration. The classic failure: the image was captured at one magnification, the software assumed another. Every downstream number is wrong by that ratio. This is why calibration should follow the objective automatically, not depend on operator memory.
• Under-etching. Boundaries that are not revealed are not counted; the structure reads coarser than it is. Automatic detection helps consistency, but the etch has to show the boundaries first.
• Twin boundaries. In austenitic steels and brasses, annealing twins are not grain boundaries. Counting them inflates G. They must be excluded — by eye or by an editable overlay.
• Duplex structures. A mix of coarse and fine populations does not average into one honest G. Standards treat duplex distributions separately; a single number hides what matters.
• Too few fields. One field is an anecdote. Accumulate several representative fields before quoting a value to a customer.

Reading G like a metallurgist

A few anchors help make the scale intuitive. Around G 3–5, individual grains are visible to a trained eye at modest magnification — typical of castings and hot-worked structures. G 7–8 is the everyday territory of normalized and quenched-and-tempered structural steels. Above G 10 you are in fine-grained, grain-refined practice, where the structure is measured, never guessed. When a certificate says "G 7.5", it is claiming roughly 90 grains per square millimetre of section — a checkable, physical claim. That is the point of the number: it turns "looks fine-grained to me" into something two labs can agree on.