Measuring Inclusions and Grain Size in Steel

The old rating and the new measurement

Microstructural ratings for quality control and specification compliance were historically done by comparing a field with a standard chart. Grain size was read against the chart in ASTM E112 and inclusion content against the chart in ASTM E45, and both are still in use because both are fast and both are agreed between buyer and seller. What has changed since is the arrival of stereological measurement, which describes a structure with numbers such as area fraction, count per unit area, mean lineal intercept and interfacial area per unit volume, and automated image analysis, which collects those numbers far faster than an operator can.

The two approaches answer slightly different questions, and the second one has a property the first does not. Chart comparison returns a value from a fixed scale, so two careful operators tend to converge on the same rating. A stereological measurement returns a real number, and the value of that number depends on how the instrument was set up. Automation removes operator fatigue, which is a genuine source of error in manual work, and replaces it with a set of setup decisions that have to be controlled deliberately. The rest of this page is about those decisions.

Setting the threshold is the largest error in the room

An automatic analyser detects a feature by its grey level. The operator sets a threshold, and everything darker or lighter than that threshold is counted as the phase of interest. Detection by threshold is described in the source material as the most significant instrument error in the whole procedure, and reproducibility between operators, or between two days for the same operator, suffers mainly because the procedure used to establish the threshold changes.

The published example is a low-carbon steel etched in 4 % picral, so the phases present are ferrite and pearlite, and the same field was detected three times by moving only the threshold. The result is worth carrying around, because it is the reason a single figure should never be taken as the property of the steel.

One field of a low-carbon steel etched in 4 % picral, detected three times by changing only the threshold. Source, Figure 4 of Image Analysis, in ASM Handbook, Volume 10.

Two details in that example are practical. Picral is the more accurate etchant for this kind of measurement because it does not attack the ferrite grain boundaries, which would otherwise be detected along with the pearlite, and because it darkens pearlite more uniformly than nital does. And the grey level range of a typical image is divided into something between 64 and 100 increments, so the operator is choosing one step out of a hundred with a live image as the only feedback. The standard trick is to alternate between the live image and the detected image while watching how well the sizes correspond, which is the only way to see an error that is invisible in the final number.

Magnification changes the answer

The second variable is the objective. Measuring the same specimen at a higher magnification is normally assumed to give a better answer, and on grain size the opposite is true when the frame is not handled correctly. In the published example a specimen of about ASTM 5.2 was measured at three magnifications without deleting the grains that intersected the edge of the measuring frame. The intercept length fell and the reported grain size number rose, because a larger share of the grains in each field were cut by the frame and therefore measured short.

Two specimens measured by automatic image analysis at several magnifications. Source, the grain size examples in Image Analysis, in ASM Handbook, Volume 10.

The effect is not limited to edge errors. In a medium-carbon hot-rolled steel the measured volume fraction of proeutectoid ferrite rose as the magnification was raised, because at higher resolution the ferrite lamellae inside the pearlite were resolved and counted as well. The best determination in that experiment came from the 16 times objective, which is not the finest one available and is not obvious from the numbers alone. There is also a plain sampling argument that runs the other way. The field area at 500 times is one twenty-fifth of the area at 100 times, so the same area of specimen demands twenty-five times as many fields, and a specimen with a wide or bimodal particle size range cannot be measured properly at a single magnification at all.

Etch time changes the answer too

Etching is where a specification usually stops at naming the reagent, and the published example shows why the time has to be part of it. A low-carbon sheet steel was etched in 2 % nital, an orientation sensitive etchant, and measured after different etch times. After a few seconds the mean lineal intercept came out at about 66 μm, which is ASTM grain size 4.5. After about twenty seconds of etching the same specimen gave about 17 μm, which is ASTM 8.5. Both figures were produced by the same operator on the same instrument from the same piece of steel, and they differ by four grain size numbers.

Past roughly twenty seconds the result settled down and became reproducible, so the rule is to etch to a controlled end point and then measure, not to etch until the structure looks right and then stop. Heavy etching with nital is specifically to be avoided, because the etch relief it produces makes the surface reflect light unevenly and corrupts the measurement rather than improving the contrast. For a tool steel the same logic applies to any measurement where the structure is fine, and it applies with more force in a hardened structure where the etch window is narrow.

How many fields are enough

Inclusion measurement has a statistical problem that grain size does not, because the quantity being measured is small and its distribution is not uniform. In a resulfurized steel used as the example, the inclusion volume fraction measured over fifty or more fields settled between about 0.5 and 0.7 %. Over ten fields or fewer the scatter was substantial. When the same data were replotted against the total area examined rather than the number of fields, the relationship became usable, and in that example about 10 mm² of polished surface had to be measured to reach 10 % relative accuracy.

That last figure is the one to keep. Ten percent relative accuracy is described as the working minimum for this kind of work, and the area needed to reach it depends on how much of the constituent is present and how evenly it is spread, so a cleaner steel needs a larger area for the same confidence. Measurement below one percent by volume is specifically called out as the acute case. A cleanliness figure quoted without the area examined, or the number of fields, is a figure that cannot be checked or compared, and this is the single most common defect in an inclusion report.

Preparation decides whether the number is real

Sample preparation is described in both source articles as the factor most often decisive and most often neglected. Inclusions are measured on an unetched surface, because detection is more reliable there, and the grinding damage left by the cut off wheel extends to roughly 1 mm below the surface on a typical specimen and has to be ground and polished away before anything is counted. Polishing damage has its own signature, and the comet tails thrown out from a poorly polished hard particle are the classic artefact on a tool steel. Where automatic polishing is used it is because the sample volume is too large to do by hand at a consistent quality, not because hand polishing is inaccurate.

Two further effects are worth knowing. Features that cross the edge of the measuring frame are measured short, so they either get deleted, with the measuring area corrected afterwards, or measured only when they intersect two of the four borders. And the environment matters more than it sounds, because dust settling on the specimen during an inclusion count or humidity staining it both introduce errors that look like real features. On a hardened tool steel at high magnification the same caution applies to the depth of field, because autofocus at high magnification is where an out of focus field slips through unnoticed.

What the instruments in a metallography laboratory each deliver. Compiled from the two source articles in ASM Handbook, Volume 10.

What this means for a tool steel order

When a purchase order carries an inclusion requirement or a grain size requirement, the useful part of the specification is the method rather than the number. The standard governs the rating, so ASTM E45 or ASTM E112 in the clause matters, and the report should carry the magnification, the number of fields or the area examined, the etchant and the etch time, and whether foreign material was measured on a polished or an etched surface. Given those, two laboratories can be compared. Without them, a report that says the steel is clean is a statement of opinion.

Inclusion content also carries directly into performance, which is why a buyer asks about it at all. Inclusions sit behind reduced ductility, they are named as sites where fatigue cracks and fracture begin, and their size and distribution are what limits how far a tool steel surface can be polished before the microstructure starts to show. That is why a cleaner steel is bought through a cleaner melting route, and the remelting side of that question is set out in ESR tool steel. Our own practice around incoming verification is set out in how we control tool steel quality, the surface finish consequence is covered in polishing and texturing of plastic mold steel, the carbide side of the same subject is in carbide distribution in high speed steel, and the defects that rolling can leave in a bar in the first place are collected in steel bar defects. The grain size number and the diameter it corresponds to are tabulated in the ASTM grain size chart for tool steel, and the mold grades where polishability is the buying reason are gathered under plastic mold steels. Where the requirement still has to be matched to a grade, the tool steel material finder and the contact page are the two places to start.

Source: ASM Handbook, Volume 10, Materials Characterization, ASM International, 1986.