ASTM Committee E-4 and Grain Size Measurements

The following article describes the evolution of grain size measurement and ASTM stardard E 112.

Summary

ASTM Committee E-4 has been a world leader in the standardization of grain size measurement methods. Initially, Methods E 2 recommended the ]effries planimetric method as the preferred measurement method. This method is more difficult to apply on a production basis than the intercept method due to the need to mark off the grains as you count them to minimize counting errors. This is unnecessary with the intercept method.

With the 1974 revision of Test Methods E 112, the intercept method, as modified by Halle Abrams, became the preferred analysis technique. The three-circle intercept method, as described in Test Methods E 112 since 1974, provides a more precise estimate of the grain size in much less time than required by the planimetric method. In manual methods, it is essential to recommend the most efficient method for any measurement.

Test Methods E 112 is designed for rating the grain size of equiaxed grain structures with a normal size distribution; the standard is presently being revised to provide better instructions for rating the grain size of deformed grains. Other standards have been introduced by E-4 to handle the measurement of occasional, very large grains present in an otherwise uniform, fine grain size dispersion (E 930, Methods of Estimating the Largest Grain Observed in a Metallographic Section (ALA Grain Size)) or for rating the grain size when the size distribution is non-normal, for example, bi-modal or "duplex" (E 1181, Methods of Characterizing Duplex Grain Sizes). Committee E-4 has recently developed a grain size standard for ratings made using semiautomatic or automatic image analyzers (E 1382, Test Methods for Determining the Average Grain Size Using Semi-Automatic and Automatic lmage Analysis). No other standards writing organization has developed standards similar to Methods E 930, Methods E 1181 or Test Methods E1382.

Metals, except in a few instances, are crystalline in nature and, except for single crystals, they contain internal boundaries known as grain boundaries. When a new grain is nucleated during processing (as in solidification or annealing after cold working), the atoms within each growing grain are lined up in a specific pattern that depends upon the crystal structure of the metal or alloy. With growth, each grain will eventually impinge on others and form an interface where the atomic orientations are different.

As early as the year 1900, it was well known that most mechanical properties were improved as the size of the grains decreased. A few notable exceptions exist where a coarse grain structure is desired. Alloy composition and processing must be controlled to achieve the desired grain size. Metallographers examine polished cross sections of specimens from appropriate locations to determine the grain size.

Complications--Grain Characteristics

Grain size measurement is complicated by a number of factors. First, the three-dimensional size of the grains is not constant and the sectioning plane will cut through the grains at random. Thus, on a cross-section we will observe a range of sizes, none larger than the cross section of the largest grain sampled. Grain shape also varies, particularly as a function of grain size. One of the earliest studies of grain shape was made by Lord Kelvin in 1887. He showed that the optimum space-filling grain shape, with a minimum surface area and surface tension, is a polyhedron known as a tetrakaidecahedron, which has 14 faces, 24 corners, and 36 edges. While this shape meets most grain criteria, it does not satisfy the required 120 degree dihedral angles between grains where three adjacent grains meet at an edge, unless the faces exhibit a minor amount of curvature. Another ideal grain shape, the pentagonal dodecahedron, agrees well with observations of grains, but is not a space filling shape. It has twelve five-sided faces. However, it must be recognized that we are sampling grains with a range of sizes and shapes. In most cases, the grains observed on a polished cross-sectional plane exhibit a range of sizes around a central mean and individual measurements of grain areas, diameters, or intercept lengths exhibit a normal distribution. In the vast majority of cases, we merely determine the mean value of the planar grain size, rather than the distribution. There are cases where the grain size distribution is not normal but bimodal, or "duplex." Also, our grain shapes can be distorted by processing procedures so that they are flattened and/or elongated. Different product shapes, and different processing procedures, can produce a variety of non-equiaxed grain shapes. This, of course, does influence our ability to measure the grain size.

Grain size measurement is also complicated by the different types of grains that can be present in metals, although their fundamental shapes are the same. For example, in body-centered cubic metals, such as Fe, Mo, and Cr, we have ferrite grains; in face-centered cubic metals, such as Al, Ni, Cu, and certain stainless steels, we have austenite grains. The grains exhibit the same shapes and are measured in the same way, but we must be careful in describing what kind of grains we are measuring. In the face-centered cubic metals, we may observe so-called twin boundaries within the grains. Aluminum alloys, however, rarely exhibit twins. When twins are present, they are ignored if we are trying to define the grain size. However, if we are trying to establish a relationship between microstructure and properties, for example, strength, we must consider twin boundaries as they influence dislocation movement, just as grain boundaries do. Hence, we must recognize the intent of the work being performed.

In heat-treated steels, it is recognized that the grain size of the product of the heat treatment, usually martensite, is not measured or cannot be measured. For low-carbon steel, the martensite forms in packets within the parent austenite grains. In high-carbon martensites, we do not observe any convenient structural shape that can be measured. In most cases, we try to measure the size of the parent austenite grains that were formed during the high temperature hold during the heat treatment. This is usually referred to as the "prior-austenite grain size" and it has been widely correlated to the properties of heat treated steels. The most difficult process here is the etching procedure needed to reveal these prior boundaries. Sometimes they cannot be revealed, particularly in low-carbon steels. In this case, it may be possible to measure the low-carbon lath martensite packet size, which is a function of the prior-austenite grain size.

Complications---Different Measures of Size

Another complicating factor is the different measures of grain size. The planimetric method, described below, yields the number of grains per square millimeter area, N_A, from which we can calculate the average grain area, A. It is common practice to take the square root of A and call this the grain diameter, d, although this assumes that the cross sectional shape of the grains is a square, which it is not. The intercept method yields a mean intercept length, L₃ ; its relationship to N_A, A, or d is not exceptionally well defined. A variety of planar grain size distribution methods have also been developed to estimate the number of grains per unit volume, N_v, from which the average grain volume, V, can be calculated. The relationship between these spatial measures of grain size and the above planar measures is also ill-defined.

It is now common to express grain sizes in terms of a simple exponential equation: (Equation 1)

n = 2 ^{G - 1}

where:
n = the number of grains per square inch at 100X magnification, and
G = the ASTM grain size number.

This approach was developed and introduced in 1951 with the premiere of ASTM standard E 91, Methods for Estimating the Average Grain Size of Non-Ferrous Metals, Other Than Copper and Their Alloys. Although the N_A, d, or L₃, values had been used for many years as measures of grain size, the G values were adopted readily due to their simplicity. As shown in Eq. 1, we can directly relate the number of grains per unit area to G, but the relationship between L₃, and G, or N_V and G are not as clearly defined. This problem is one of many being addressed by ASTM Committee E4 on Metallography.

Measurement Methods

Although Committee E-4 was formed in 1916 for the express purpose of establishing standard magnifications for micrographs, its first standard, E 2-17T, Methods of Preparation of Micrographs of Metals and Alloys, was partly devoted to grain size measurement. Two basic approaches to measure grain size were being developed at that time. In the United States in 1894, Albert Sauveur published a "planimetric" approach, which was further developed by Zay Jeffries with two 1916 publications. This approach measured grain size in terms of the number of grains visible on a cross section within a fixed area, the number per square inch at 100X, or the number per square millimetre at 1X, N_A. From this value, the average cross-sectional area of the bisected grains can be computed. This is not an average of the maximum cross-sectional area of each grain because the sectioning plane does not intersect each grain at its maximum width.

In Germany in 1904, Emil Heyn published an intercept approach for measuring grain size. In this method, one or more lines are superimposed over the structure at a known magnification. The true line length is divided by the number of grains intercepted by the line. This gives the average length of the line within the intercepted grains. This average intercept length will be less than the average grain diameter but the two are interrelated.

Many grain size raters expressed the need for simpler ways to estimate the grain size. In some cases, such as heat clearance, grain size measurement is required. In many cases, it is required that G be 5 or greater (i.e., "fine- grained"). Hence, if the grain size is substantially finer than this, a quick method, which may not be as precise as an actual measurement, is adequate. A comparison chart method with examples of grain sizes meets this need adequately, as long as the grain size distribution is normal. Additionally, the specimens should be etched in the same manner as depicted on the chart. If the grain size is near the specification limit, an actual measurement is preferred due to the improved precision. The first grain size comparison chart was introduced in Methods E 2 in its 1930 revision; this chart was for copper.

Note that these methods are applied on the polished surface of the specimen, that is, on a plane that cuts through the three-dimensional grains. Thus, these are planar rather than spatial measures of the grain size. The planimetric, or Jeffries method, defines the grain size in terms of the number of grains per unit area, the average grain area, or the average grain diameter, while the Heyn intercept method defines it in terms of the average intercept length. The comparison chart method expresses the grain size only in terms of G, except for the copper charts, which use d.

Evolution of Test Methods E 112

Methods E 2-17T was only slightly more than three pages long and had three sections: standard magnifications, lenses, and grain size. The grain size section did not actually detail the measurement method, it merely suggested the method to apply depending on whether the grains were equiaxed (Jeffries planimetric method) or elongated (Heyn intercept method). The 1920 revision of Methods E 2 added details on performing the Jeffries planimetric measurement method. The 1930 revision of Methods E 2 witnessed the addition of Committee E-4's first standard chart, a grain size chart (ten pictures) for brass, i.e., a twinned austenitic structure with a grain contrast etch at 75X magnification. The chart was developed by a special committee formed on June 28, 1928, which consisted of: C.H. Davis, chairman (American Brass Co.); Henry S. Rawdon (U.S. Bureau of Standards); Edgar H. Dix, Jr. (Aluminum Co. of America); and Francis F. Lucas (Bell Telephone Laboratories). Types of grain structures are shown in the.

A special subcommittee to study grain characteristics of steels was formed in 1931 with Clarence J. Tobin (General Motors Research Laboratory) as chairman. They decided to adopt the McQuaid-Ehn carburizing test for evaluating the grain growth characteristics of steel, again with the aid of a comparison chart. The proposed chart method was approved as E 19-33T, Classification of Austenite Grain Size in Steels. At that time, grain size was defined in terms of the number of grains per square inch at 100X; ASTM grain size numbers were not introduced until much later. However, this chart was criticized for being inaccurate and it was eventually dropped when E 112, Test Methods for Determining the Average Grain Size, was introduced.

Oscar E. Harder took over this special subcommittee in 1936, with the idea of revising Classification E 19 and adding a non-carburizing method. The next year, Dr. Marcus A. Grossman (Carnegie-lllinois Steel Co.) took over control of this group, which became Subcommittee Vlll (Arabic numerals are now used) on Grain Size in 1938. Grossman ---famous for his work on hardenability---was chairman of Subcommittee Vlll until his death in 1952. Subcommittee Vlll formed three sections (the term task group was not used at that time), referred to as A, B, and C. Section A was chaired by Grossman and was concerned with improving Classification E 19 on austenite grain size of steels. Section B was chaired by R. Earl Penrod (Bethlehem Steel-Johnstown Plant) and was to develop a ferrite grain size rating method and chart. Section C was chaired by Carl Samans (American Optical Co., later with Standard Oil Co. of Indiana) and was to develop charts for nonferrous metals and alloys that could not be rated by the brass chart in Methods E 2. The brass grain size chart and grain size measurement information was deleted from Methods E 2 in the 1949 revision and this information was incorporated into a new standard, E 79-49T, Methods for Estimating the Average Grain Size of Wrought Copper and Copper-Base Alloys. Two pictures were added to the chart; later when it was transferred to Test Methods E 112, two more pictures were added (14 in all). Methods E 2 was discontinued in 1984 when E 883, Guide for Metallographic Photomicrography, was introduced.

Section B produced E 89-50T, Methods for Estimating the Average Ferrite Grain Size of Low-Carbon Steels, with a chart depicting a ferritic grain structure as revealed by nital etching. This was the first chart (eight pictures) to define grain size in terms of the now familiar ASTM grain size numbers (1 to 8 in this chart). Methods E 89 also marked the first detailed description of the Heyn intercept method with equations and a conversion approach to yield ASTM grain size numbers. Earlier Methods E 2 versions only gave a general description of how to do the intercept test with no interrelationship to the results from the planimetric method. Methods E 89, however, had a short life, being discontinued when Test Methods E112 was adopted.

Section C produced E 91-51T, Methods for Estimating the Average Grain Size of Non-Ferrous Metals, Other Than Copper, and Their Alloys. This consisted of two charts, one for twinned alloys, the other for non-twinned alloys; both charts had 17 pictures with grain sizes from 2 to 10. Methods E 91 also had a short life, also being discontinued when Test Methods E 112 was adopted. Neither of the charts of Methods E 91 were incorporated in Test Methods E 112.

The net result was four standards (Methods E 19, E 79, E 89, and E 91) dealing with various aspects of grain size measurement. It was recognized that all four shared many common points and it was believed that they could be combined into one overall grain size standard, hence the birth of Test Methods E 112. However, the story of ASTM and grain size measurement does not end with the adoption of Test Methods E 112 in 1961. Since then, the standard has been revised nine times and is presently now under intense scrutiny for further refinement. (webmasters note: E112 has been updated and reissued as E112-96)

Beyond Test Methods E 112

Test Methods E 112, one of the most widely cited ASTM standards, is chiefly concerned with the measurement of grain size when the grains are equiaxed in shape, that is, non-deformed, although it does contain some information about measurement of grain size when the grains have been elongated by processing. There are other situations where Test Methods E 112 is not helpful and other standards have been developed. For example, certain alloys may not exhibit a uniform distribution of grain sizes. Instead, a bimodal distribution may exist; several types have been observed. Two ASTM standard test methods deal with such structures. Standard E 930, Test Methods for Estimating the Largest Grain Observed in a Metallographic Section (ALA Grain Size), is used to measure the size of an unusually large grain in an otherwise uniformly fine grain size distribution, while standard E 1181, Test Methods for Characterizing Duplex Grain Sizes, is used to measure the grain size when the distribution is non-normal. With the growth of image analysis, test methods for performing measurements must be established and a new standard, E 1382, Test Methods for Determining the Average Grain Size Using Semiautomatic and Automatic Image Analysis, completed the balloting process in 1990. This standard describes a number of equivalent approaches for measuring grain size using both tablet digitizer systems and fully automatic systems.

The lnternational Scene

Committee E-4's work on grain size has been followed closely by other industrial countries and the lnternational Organization for Standardization (ISO). Many countries have adopted one or more of the grain size charts of ASTM Test Methods E 112. Some countries have also developed very useful charts. For example, for rating McQuaid-Ehn carburized specimens, most U.S. raters etch the pearlitic matrix dark as depicted in Plate IV of Test Methods E 112. As the sidebar on grain structures demonstrates, it is easier to see the intergranular carbide phase if we use an etchant that darkens the grain boundary cementite. The French grain size standard, NF A04-102, contains a rating chart where the grain boundary cementite was darkened with alkaline sodium picrate. The German SEP 1510 grain size standard also contains a very useful chart. It illustrates non-twinned grains (such as ferrite grains) that are equiaxed or deformed (elongated 2 to 1 and 4 to 1) by cold working. Eq. 1 described the approach used to compute ASTM grain size numbers which, de- veloped in the United States in the late 1940s, was based on English units rather than metric units. Countries that used the metric system at that time developed an alternate equation that produces nearly identical grain size numbers: (Equation 2)

m = 8(2^Gm)

where:
m = the number of grains per mm² at 1 X, and
G_m = the metric grain size number.
G_m is slightly greater than G but the difference is negligible. Eq. 2 is used in the Swedish (SIS 11 11 01), ltalian (UNI 3245), Russian (GOST 5639), French (NF A04-102), and ISO (ISO 643) standards.

The German standard (SEP 15l0)also uses the metric system, but a different equation is employed : (Equation 3)

K= 3.7 + 3.33Log(Z)

where:
K = the photomicrograph serial number (same as G), and
Z = the number of grains per cm² at 1OOX.

In this case, K equals G. Japanese standards JIS G 0551 and G 0552 also use the metric system, with a slightly different equation than Eq. 2 (but equivalent) that produces the same values as Eq. 2: (Equation 4)

m = 2^(Gm+3⁾

where m and G_m are defined as before.

GRAlN STRUCTURE TYPES The measurement of grain size, whether by the chart comparison method or by manual or automated measurement methods, is complicated by the different types of grain structures encountered and by the etched appearance of the grains. For example, as shown in Figure A, we may have ferrite grains in a non- heat treated or non-hardenable body-centered cubic (bcc) metal or alloy. These do not contain annealing twins, but could contain deformation twins, and second-phase constituents may be present. The example shown is ferrite in a low-carbon sheet steel; carbides are present. This specimen was etched with nital and not all of the grain boundaries are visible; those that are visible are variable in darkness and width. These factors are a minor nuisance for manual rating and a significant problem for automatic rating. Figure B depicts a single phase austenitic alloy that contains annealing twins. Like the previous micrograph, it shows the boundaries as dark lines, a so-called "flat etch." The austenitic alloy shown, L605, illustrates a common problem with such alloys, they are very difficult to etch so that all of the grain boundaries are visible. This makes it very difficult to measure the grain size with a high degree of precision. Also, when rating grain size the twin boundaries must be ignored, which is not easy, especially by image analysis. Not all austenitic alloys will exhibit annealing twins, aluminum alloys rarely are twinned. Austenitic alloys may also be etched with reagents that produce grain contrast or color variations as a function of their crystallographic orientation. Figure C shows the twinned austenitic grain structure of cartridge brass that was etched producing grains with different contrast in black and white. Note that unlike the flat etched L605 specimen, all of the grains are revealed. This structure is easy to rate by the comparison method if the grain size chart depicts grains etched in the same manner. This condition is virtually impossible to measure by automatic image analysis, however. Again, twins are present but the coloration or contrast varies within the grains. To measure twinned austenitic grain structures by image analysis, we need to either suppress the etching of twins or be able to identify and ignore them. At the same time, all of the grain boundaries must be revealed and be identifiable. The best solution is to use an etchant that reveals only the grain boundaries. To illustrate this, the next micrograph, Figure D, shows AlSl 316L stainless steel electrolytically etched with 60% nitric acid in water (Pt cathode, 0.8 V dc, 45 s). The grain boundaries are almost completely revealed but no twins are visible. The accompanying micrograph, Figure E, shows a tint etched view of this specimen at the same magnification where the twins are visible. In dealing with carbon and alloy steels, the steelmaker generally performs a test known as the McQuaid- Ehn test, to determine if the steel is inherently fine grained. A specimen is carburized at 1700"F for 8 h and furnace cooled. The excess carbon in the carburized case precipitates during cooling as cementite in the austenite grain boundaries present at the end of the carburizing cycle. The specimen is cut and polished so that the case structure is revealed. Generally, nital is used as the etchant and a comparison chart rating is made where the Test Methods E 112 chart exhibits the same contrast. Such a structure, Figure F, is not very good if actual measurements are made, especially if image analysis is employed. The alternative is to darken the grain boundary cementite films. A number of etchants will darken cementite, the one used here, Figure G, was Beraha's sodium molybdate tint etch but the familiar alkaline sodium picrate etch works well also. Etched in this way, the grain structure shows up much more clearly and image analysis could be used. Once an alloy steel part is heat treated, only etching can be used to try to reveal the prior-austenite grain boundaries, that is, the austenite boundaries present when the part was soaked at the austenitizing temperature. While many etchants have been developed for this purpose, such work is fraught with difficulty. One of the most successful prior-austenite grain boundary etchants is a saturated aqueous solution of picric acid containing a wetting agent, several of which have been used. This etch is sensitive to phosphorus segregated to the prior-austenite grain boundaries and will not work otherwise. Figure H illustrates a fairly successful effort with a quenched and tempered experimental alloy steel. This type of etch rarely, if ever, yields an etch quality adequate for image analysis and is usually accompanied by substantial pitting.

GRAlN SlZE DlSTRlBUTlONS The previous examples of grain structures were of equiaxed grains with a uniform size distribution, that is, if a frequency histogram is plotted of the grain diameters, intercept lengths, or areas, a single peak is observed which will be symmetrical in shape if the dimension scale is logarithmic. However, there are occasions where more complex distributions are observed, often in specimens where recrystallization is incomplete, or at the onset of rapid grain growth. Test Methods E 112 is not well suited for such specimens--Methods E 930 or E 1181 must be used, usually the latter. As an illustration, Figure A (the figures for this sidebar are in the margins) shows the microstructure of an experimental hot work tool steel that was austenitized at 1975 degrees F ; grain growth was very rapid at slightly higher temperatures. Etching revealed the prior-austenite grain boundaries which, except for one very large grain, are from a single distribution. In measuring the grain size of this specimen, we could determine the grain size of this one very large "rogue" grain by Methods E 930 and the rest oi the grains using Test Methods E 112. A somewhat similar example is given in Figure B, which shows the onset of grain growth in A286. These are twinned austenite grains revealed by a grain contrast etch, rather than a flat etch as in Figure A. We could measure the largest of these grains by Methods E 930 and the fine region by Test Methods E 112, or we could use the approaches in Methods E 1181 for the entire structure. Sometimes the duplex nature of the grain size distribution is highly segregated, in other cases the two distritbutions are intermixed. Figure C shows a relatively well mixed bimodal distribution of grain sizes. This is a superalloy specimen with a flat etch containing annnealing twins. Figure D, on the other hand, shows an extremely segregated example of a duplex condition. Note that grain growth has occurred at two locations along the surface of this low-carbon steel specimen. Another segregated form of a duplex condition is the so-called "necklace" type, as shown in Figures E and F. Both are highly alloyed stainless steels that have not been fully recrystallized. The fine recrystallized grains surround the large non-recrystallized grains. The main difference between the two examples is the pronounced elongation of the non-recrystallized grains in Figure F due to the difference in sectioning plane orientation, transverse vs. longitudinal, for the two. The amounts of the fine and coarse grains can vary considerably. Obviously, the percentage of the number of grains of each type will be much different than the area or volume percentage of each type. Figure G shows a more equal area percentage of fine and coarse grains in a nickel-base superalloy. In some cases, the different grain size distributions may be in a layered or banded manner due to the influence of segregation. Figure H shows such a pattern in ferritic stainless steel plate specimen. Note the very long, thin non-recrystallized grain. The grain size of such specimens should not be described by a single average value, as it is quite possible that there will not be any actual grains of that size in the specimen; or, the average chosen may be in one of the tails of the two grain size distributions. The best approach is to determine the area percentage of each grain size distribution and the average grain size of each distribution in specimens with a duplex or bimodal grain size distribution, as described in Methods E 1181 .