Metallography, Microstructure, and Analysis

, Volume 7, Issue 6, pp 650–660 | Cite as

A Metallographic Study on the Diffusion Behavior and Microstructural Transformations in Silicon-Containing Powder Metallurgy Steels

  • Thomas F. MurphyEmail author
  • Christopher T. Schade
  • Alan Lawley
  • Roger Doherty
Technical Article


The choice of alloying method for ferrous powder metallurgy alloys is often dictated by the oxidation potential of the alloying elements used in powder and compact manufacture. Silicon is effective in improving properties of ferrous PM steels; however, if added to the melt prior to atomization, the likelihood of oxidation is high. Alternatively, Si-rich and more complex particulates can be combined with the base powder, i.e., iron or low-alloy steel, with alloying occurring by solid-state diffusion during sintering. The effectiveness of these additives to improve material properties is dependent on their distribution throughout the material volume, as determined by diffusion of each added element. In this study, the distribution of silicon was quantified using energy-dispersive spectroscopy on an iron-based alloy composition. These data were compared with the volume fraction of the various transformation products in cross sections of hardenability samples using both light and electron microscopy. Predictions on the effects of local chemical composition, cooling rate, and microstructure are made.


Powder metallurgy Microstructure Phase transformations Quantitative metallography 


The beneficial effects of alloying ferrous powder metallurgy (PM) compacts with silicon, both alone and in combination with other elements, have been detailed by Schade et al. [1]. In that study, silicon was used in multiple concentrations as both the sole alloying additive in an unalloyed iron powder and with several levels of other additives, including chromium, manganese, in an iron powder and an iron-molybdenum prealloy.

The study showed the improvements on mechanical properties realized with the silicon additions resulted from increases in ferrite hardness and alloy hardenability, and a decrease in pearlite spacing [2]. These led to greatly improved tensile strength and higher as-sintered part hardness. While the improvements described in the original study were mostly property-related, the project, as described here, was expanded to study the effects of local chemical composition on the hardenability of these silicon-containing alloys.

Silicon is a common alloying element in wrought and cast ferrous alloys, but not used as frequently in PM. In the manufacture of wrought and cast steels, Si can be introduced into the molten metal bath prior to pouring but must be protected from exposure to atmospheric oxygen. In a comparison with the manufacture of ferrous metal powders and compacts, individual alloying additives, including Si, can be added into the powder using several techniques. The choice of alloying method is usually determined by the susceptibility of the specific element to oxidization during water atomization of the molten metal stream. For instance, when alloying with molybdenum, it is usually added to the molten metal bath and prealloyed with the iron because any molybdenum oxides created during atomization are easily reduced during the sintering process [3]. In comparison, chromium, manganese, and silicon are often mixed into a base powder in the form of iron-containing master-alloy particles. The different alloying methods are used because of the high propensity for the latter elements to form stable oxides during water atomization and the resulting difficulty to reduce these oxides during the sintering [4]. Comparing the alloy distributions of each alloy type, the prealloyed powder is uniform in particle-to-particle alloy content because the elemental additions are made to the molten metal bath, while alloys created using mixtures of powders are usually heterogeneous in alloy distribution. As a mixture of individual particles, there are variations in alloy concentration due to the physical distribution of the added particles throughout the mix; with the locations containing concentrations of additive particles having the highest elemental content. These highly concentrated areas originate along the iron or low-alloy steel particle boundaries in the as-pressed compact and are the starting locations for solid-state diffusion of the additive into the base powder particles.

If the element intended for alloying is oxidized during atomization, it makes no contribution to the intended sintered property improvements and may lead to degradation of some mechanical properties or material behaviors. Therefore, if oxidation has occurred, reducing the oxides during sintering is essential to utilizing the element for alloying. Since these complex oxides (Cr, Si and/or Mn) are relatively stable at a conventional sintering temperature (1120 °C–2050 °F), the use of higher temperature sintering (e.g., 1260 °C–2300 °F) can be useful for oxide reduction. Figure 1 shows the conditions needed for the oxidation and reduction of numerous elements under an atmosphere of hydrogen. Although pure hydrogen is not normally used as the atmosphere in the sintering furnace, the graph gives a good indication of the temperatures and dew points needed to reduce an oxide and make the particle available for alloying. Once the oxides are reduced, the high temperature provides the added benefits of increasing diffusion of the additive, developing a more uniform alloy distribution and subsequently, enhancing hardenability over a larger volume of the compact. In some cases, the additive may reach areas near the center of larger base particles that are usually unalloyed under sintering conditions using lower temperatures.
Fig. 1

The temperature and dew point conditions required to reduce various oxides. The line showing what is required for oxidation/reduction of SiO2 is indicated using the arrows [4]

The distribution of the alloying particles added to the iron or low-alloy volume determines the starting points for alloy diffusion, and these locations are between the base powder particles and along the pore edges in the as-pressed compact. As sintering starts, the elements diffuse into the base iron or low-alloy steel particles and the local chemical composition changes, creating steep concentration gradients throughout the compact volume. At increased temperatures and/or for longer times at temperature, the depth of diffusion increases, and the alloyed gradients become shallower. The final local alloy hardenability is created at the end of residence time in the hot zone of the sintering furnace. Upon entering the furnace cooling zone, the cooling rate determines the transformation product that will form based on the local chemical composition. With the combination of high alloy concentrations and a rapid cooling rate, martensite forms. In areas containing an insufficient number of additive particles or with a slower cooling rate, lower hardness transformation products form. Consequently, the amount and distribution of the additives, the effectiveness of the additives to improve hardenability, diffusion rates, the sintering time and temperature, and cooling rate all affect the combination of transformation products that form and the resulting mechanical and physical properties of the sintered part. It was shown by Schade et al. [1] that the diffusion of silicon from a master-alloy additive resulted in local variability in both silicon content and hardenability. This was confirmed with microindentation hardness measurements in a Jominy-like hardenability test and with the use of light optical and electron microscopy to characterize the correlation of local chemical composition with the transformation products.

To gain a better understanding of the silicon-containing alloys used in this study, experiments were undertaken to quantify the distribution of the alloy elements within several alloying systems after sintering. This information is vitally important because the type and amount of transformation products are directly related to the distribution of the alloying additives in the sintered compact. Techniques developed by Gungor [5] and then DeHoff [6] were used to describe the variability in chemical composition and the corresponding differences in microstructure. These techniques use stereological methods to estimate uniformity in some aspect of the structure and help develop a correlation with the local chemical composition and the resulting transformation products. The test methods use techniques similar to point counting, where uniformly spaced points are overlaid on cross-sectional areas of metallographically prepared specimens. Samples from four composition groups containing silicon, either alone or in conjunction with another alloying element, were analyzed. However, only the Fe-Si alloy will be discussed here considering the Si addition to the Mo, Cr, and Mn alloys increased the hardenability sufficiently to virtually through-harden the test samples and the relevance of the Si distribution on the transformation products would be less apparent.

Experimental Procedures

Mixtures of a base powder and a master-alloy containing silicon were used to prepare the compacts for hardenability testing and subsequently, chemical analysis using scanning electron microscopy (SEM) and examination using light optical microscopy (LOM). The mean particle size (d50) of the additives was 8–15 μm. Ancorsteel® 1000B (a water-atomized iron powder typically containing 0.10 Mn, 0.03 Cr, 0.05 Cu, 0.05 Ni, and < 0.01 C all wt/o) was used as the base for the iron-silicon alloy system. The d50 of the ferrous base powder was approximately 75–85 μm, with a small fraction as large as 200 μm in diameter.

The powders were mixed with Acrawax C (ethylene bis stearamide) lubricant and fine graphite powder. Graphite additions were 0.70 wt/o, resulting in a carbon content after sintering of approximately 0.65 wt/o. The test pieces were sintered in a high temperature Abbott continuous-belt furnace at 1260 °C (2300 °F) for 30 min in an atmosphere of 90 vol/o nitrogen / 10 vol/o hydrogen.

The effect of silicon on quenched hardenability was studied using cylindrical compacts that were 25 mm in diameter × 25 mm high. These were pressed to a green density of 6.95 g/cm3 (approximately 88% dense) at a nominal pressure of 690 MPa (50 tsi). After sintering using the conditions described above, the density of the samples was 7.10 g/cm3 (approximately 90% dense). They were then reheated to 900 °C (1650 °F) for 60 min at temperature and oil-quenched. The quenched samples were sectioned across the diameter, with the exposed rectangular cross section mounted for hardness testing, microstructure evaluation, and chemical analysis. Microindentation hardness traverses were made at the middle of the sample height, from the edge to the center of the compacts (approximately 12 mm). In this way, the hardenability could be measured in a similar fashion to that in the Jominy test.

Upon completion of microindentation hardness testing, the rectangular cross sections were reprepared for metallographic analysis using standard grinding and polishing procedures. LOM was performed after etching the samples with a combination of 2 vol/o nital and 4 wt/o picral. Furthermore, both etched and unetched surfaces were used for SEM examination utilizing secondary and backscattered electron imaging (SEI and BEI) in addition to energy-dispersive spectroscopy (EDS) analysis.

LOM coupled with an image analysis (IA) system was used to estimate the proportions of the specific transformation products in the material volume at 1 mm increments from the quenched surface to the sample core. A manual point count employing an x/y grid with 88 equally spaced points was used to generate these estimates. A 20× objective lens defined the magnification used on the IA system [7, 8]. Multiple fields totaling an area of approximately 0.7 mm2 were analyzed at each depth, starting at a point near the sample surface. The initial location for the counting was approximately 0.5 mm inside the surface edge to eliminate any effects of decarburization from heat treating. The individual categories used for the estimates were martensite, ferrite, non-martensite (fine pearlite or bainite), and porosity.

Chemical composition estimates were performed using several EDS methods. These included: examination of areas exhibiting differences in BEI contrast (either atomic number or microstructure), where locations appearing as varying shades of gray were analyzed individually, linescans consisting of multiple analysis points through specific transformation products, elemental maps displaying visual variations in alloy content, and point count analyses on systematically chosen areas. In the point count analysis, a 6 × 5 equally spaced x/y grid was superimposed on fields chosen at predetermined distances from the sample surface and chemical analysis performed at each x/y intersection. The magnification used was 400×, giving a point spacing of 50 μm in both the x and y directions. All samples were analyzed in the unetched condition. Considering the heterogeneous alloy distribution is similar throughout the compact volume and the sample volume excited by the electron beam was the same for each analyzed point, the chemical composition of the compact was estimated on a volume fraction basis.

Estimation of the volume-related chemical compositions was sometimes complicated by the position of the analysis points and locations of the pores. An analysis spot falling in a pore sometimes resulted in an unusable EDS spectrum. In some locations, the information generated by the electron beam was partially trapped within the pore and unable to reach the detector. When this was encountered, the point in question was considered “unfit for use” and eliminated from the data. Several observational techniques were used to make the decision to keep or discard the information from a specific test point. Those used most often were the appearance of the spectrum, what elements and peaks were identified, and the total number of counts within the generated peaks. Sometimes it was simply whether the Fe Lα line was present. A typical spectrum contained this iron peak, so if it was not included, the point was omitted due to insufficient counts and the analysis location being hidden from the detector. In some cases, the decision was a combination of peak identification and a visual inspection of the location of the information. It was also recognized that some pores contained an excessive amount of silicon, probably as an oxide from the location of an alloying additive particle. These data did not reflect diffusion of silicon into the base particles and had no effect on hardenability, and therefore they were also omitted. Regardless of the location of the features in the microstructure, no sampling points were moved to generate an acceptable spectrum. The number of usable points included in this volume fraction chemical composition estimate exceeded 225 in each sample.

The relative accuracy of the silicon EDS analysis was improved through the use of standards made in the Hoeganaes Corporation (HC) Research & Development Laboratory. Compositions containing 0 to 3.2 wt/o silicon in an iron matrix were melted, cast, and homogenized after casting to minimize alloy segregation. The new standards were then metallographically prepared using the same techniques required for the silicon-containing hardenability samples, then the silicon content measured using calibrated x-ray analysis. Once the Si content was measured using x-ray analysis, EDS analysis was performed on multiple areas of each new standard and a linear regression analysis of the x-ray and EDS results was performed. The R2 value from this correlation was 0.99. Consequently, all silicon concentrations reported here are corrected values as determined using the linear regression expression.

In this alloying/hardenability program, four alloy composition groups were evaluated with their chemical compositions shown in Table 1. A water-atomized iron was used as the base powder for all mixes except those containing molybdenum, which was a water-atomized prealloy of iron with approximately 0.30 wt/o molybdenum. Both base powders were carbon free, which necessitated a graphite addition of 0.70 wt/o in all mixes. After sintering, the sintered carbon content of each mix was approximately 0.65 wt/o. The silicon, chromium, and manganese additions were made with concentrated master-alloy powders. As mentioned previously, selected results from the LOM and SEM testing of only the Fe and Fe-Si mixtures are discussed here because the other Si-containing compositions were basically through-hardened. Consequently, no correlation of Si content with transformation products could be observed.
Table 1

Nominal bulk chemical compositions of test specimens

Sample ID


Graphite (wt/o)

Silicon (wt/o)

Molybdenum (wt/o)

Chromium (wt/o)

Manganese (wt/o)




































Results and Discussion

Microindentation Hardness Profiles

Figure 2 shows the microindentation hardness traverses from the sample edge to the cross-sectional core for the Fe and Fe-Si compositions. It should be emphasized that the locations for all hardness impressions were selected randomly on unetched surfaces. In choosing these areas for testing, the distance from the quenched surface was determined using the automated stage of the tester and the surrounding area at that position was visually examined to find suitable pore-free locations capable of supporting the hardness indentations. The readings show a clear separation in the hardness values of the martensite and non-martensite regions. In all cases, the hardness of the martensite ranges from approximately 700 to 800 HV 0.05 and the non-martensite is between 300 to 400 HV 0.05. It should be recognized the surface in each sample contained a thin layer of decarburization from heat treatment. The soft decarburized regions are seen as the first data points near the sample edge in Fig. 2.
Fig. 2

Microindentation hardness traverse from the edge of the sample to the core of the Fe and Fe-Si sintered and oil quenched compacts

In Fig. 2, there is a clear separation of the martensitic and non-martensitic areas in the Si-containing sample. This line oscillates between basically two hardness levels, with the higher being typical of a martensitic microstructure and the lower indicating the presence of pearlite or upper bainite. The sample without silicon is lower in hardenability and, other than near the surface, is pearlitic. In addition to the martensite in the Fe-Si sample, there is a slight increase in hardness of the lower hardness transformation products indicating a small hardening of the ferrite in the pearlite/bainite with the reported decrease in the pearlite spacing from the increased silicon content [1, 2].

Alloy Distribution

Figures 3 and 4 use two EDS methods to describe the variable silicon content in a completely martensitic area near the sample surface of the Fe-Si sample. The more concentrated dot population in local areas of the elemental Si dot map in Fig. 3b indicates a somewhat higher silicon content. This is quantified in Fig. 4 where the relative positions along the dashed line are the corresponding positions along the graph. Although a large portion of the composition in this area contains < 0.5 wt/o Si, the cooling rate was sufficient to transform the entire region to martensite. In these images, the pores are the irregularly shaped gray or black features within the metallic areas.
Fig. 3

Martensitic area from the Fe-Si sample near the surface of the cross section. (a) SEI and (b) an elemental silicon map (Si Kα). The higher concentration of yellow dots indicates higher silicon content. Total image width is approximately 300 μm. The sample has been lightly etched with 2 vol/o nital + 4 wt/o picral

Fig. 4

The same area as Fig. 2 with the silicon content measured along a 50 point linescan located along the dashed line SEI, lightly etched with 2 vol/o nital + 4 wt/o picral

Figures 5 and 6 illustrate similar analyses in a region midway to the cross-sectional core. In contrast to Fig. 3, this microstructure contains multiple transformation products, where the lighter gray areas in the SEI are pearlite/bainite and the darker gray is martensite. Each transformation product has variable silicon contents as can be seen in the linescan graph. The combination of the alloy composition and sample cooling rate in this area of the compact was not sufficient to produce a fully martensitic microstructure. Again, the pores are the irregularly shaped gray to black features within the matrix of the compact.
Fig. 5

Area containing a combination of transformation products half way to the core of the Fe-Si cross section. (a) SEI and (b) Si Kα image showing higher amounts of silicon as denser concentrations of yellow dots. Total image width is approximately 300 μm. lightly etched with 2 vol/o nital + 4 wt/o picral

Fig. 6

Same area as Fig. 4 with the silicon content measured along a 50 point linescan located at the dashed line. SEI, lightly etched with 2 vol/o nital + 4 wt/o picral

As can be seen in Figs. 3, 4, 5 and 6, a single transformation product can be produced from multiple chemical compositions. An example of the variety of microstructures is seen in Fig. 7, a BEI which shows a typical sample area from the Fe-Si sample at a slower cooling rate towards the center of the sample. The presence of two transformation products is seen as two distinct shades of gray in the backscattered electron image. The chemical compositions from the two microstructures are shown in Table II. These are not necessarily taken from this exact area, but represent analyses made on several locations in the same vicinity. It is important to note the differences in Figs. 3, 4, 5, 6, 7, where Figs. 3, 4, 5, and 6 are SEI and Fig. 7 is BEI. The SEI grayscale images in Figs. 3, 4, 5 and 6 show the martensite as a darker gray compared with the lighter gray pearlite/bainite. In the BEI grayscale image in Fig. 7, the appearance is the opposite, with martensite as the lighter gray and pearlite/bainite as the darker gray. The BEI was chosen for the more accurate EDS analysis since BEI contrast was obtained with the unetched surface, while the SEI was lightly etched, resulting in a slightly roughened surface texture from the etching reaction.
Fig. 7

Typical area in the Fe-Si sample approximately half the distance to the center of the cross-section. The variety of silicon levels shown in Table 2 are examples of what is found in the two transformation products (martensite—light gray, bainite/pearlite—darker gray). BEI, unetched

The above examples (Figs. 3, 4, 5, 6, 7 and Table 2) illustrate the effect of incomplete solid-state diffusion, resulting in a variation of chemical composition that controls the individual transformation products seen throughout the compact volume. The combination of martensite and bainite/pearlite is a product of both the local chemical composition and the cooling rate experienced by the compact. At the surface, the cooling rate is sufficiently high to transform the austenite to martensite, regardless of Si content. As the location of examination is moved toward the center of the compact, the cooling rate is slower and more non-martensitic transformation products are formed. The non-distribution of the alloying elements determined by the high temperature sintering process is, of course, independent of the subsequent cooling rate from the 900 °C austenitizing heat treatment.
Table 2

Silicon content (wt/o) in specific regions such as those shown in the BEI to the left in Fig. 7

Light areas

Dark areas





















Figure 8 shows the difference in microstructure with LOM images taken at distance intervals starting at the cross-sectional surface. It is clear the change in cooling rate has a pronounced effect on the combination of transformation products that are formed at these distances. The first images are located at or near the surface, with the remaining at distances progressing farther toward the core.
Fig. 8

Change in transformation products with movement toward the center of the compact. The light and medium tan phase is martensite and the darkest brown transformation product is bainite/pearlite. Movement from the surface progresses from (a) (edge), to (b), to (c), through to (d) (core). LOM, 2 vol/o nital + 4 wt/o picral

The change in the microstructure appears as a progression from all martensite (Fig. 8a) to a small amount of bainite/pearlite (Fig. 8b) with movement several millimeters from the edge. As the distance from the edge becomes greater, increasing amounts of the lower hardness transformation product are observed until approximately half of the microstructure is martensite near the compact center (Fig. 8d). In these photomicrographs, the martensite is the lightest and intermediate tan colors with the bainite/pearlite the darker brown color. The two shades of tan in the martensitic areas are an etching effect caused by variations in the Si content. The pores are the darkest, virtually black features.

Correlation of the chemical composition with the transformation products was accomplished using both SEM and LOM. Light microscopy was used to estimate the proportion of the transformation products at various distances from the sample surface, and SEM/EDS analysis was used to estimate the distribution of the alloying additives through the compact volume. Each method gives volume-related information from the compacts and provides data to help predict what transformation product will form with specific chemical composition-cooling rate combinations.

Figure 9 is used to estimate the distribution of the Si diffused into the compact volume. From the measured Si values at the systematically selected points, the values were arranged in increasing order and cumulative number with silicon content equal to or less than each value plotted. Since each sampling point has the same measured volume, this is a cumulative volume estimate expressed as a percentage of the total number of positions counted. These PM materials alloyed with particulate additives are confirmed to be heterogeneous in alloy distribution and consequently, vary in the regions with sufficient hardenability to transform to martensite upon cooling.
Fig. 9

The variability of the silicon content in the sample volume is estimated by the location of the curve in the space defined by the silicon content (abscissa)–cumulative volume (ordinate) relationship. The curve separates the sample volume containing silicon contents less than or equal to the amount described by the curve as the cumulative percent below the curve and the sample volume containing a higher silicon content than the specific amount as that above the curve

The proportion of transformation products is shown in Fig. 10 for the Fe-Si alloy. These estimates were made using the manual point count described earlier at 1 mm intervals near the center line of the sample cross section. The graph shows a clear progressive change in the amounts of the two transformation product classifications. The proportions appear to stabilize at approximately 10 mm. This indicates, as might be qualitatively expected from the 2-dimensional heat flux in a cooling cylinder, that the central core of a radius of about 2.5 mm had an essentially constant cooling rate.
Fig. 10

Quantitative measurements of the percent martensite and non-martensite (bainite/pearlite) in the Fe-Si alloy. The estimate is made on the same cross section used for the quantitative chemical analysis

By quantifying area percent of the transformation products at specific distances from the sample surface and estimating the alloy distribution throughout the compact volume, an approximate alloy content can be determined that will transform to martensite with a given cooling rate. This may apply in a through-hardening application or hardening to a particular depth below the surface.

By combining the Fe-Si data in Figs. 9 and 10, the minimum silicon contents required to transform the compact to martensite at increments of 1 mm can be predicted. As an example, the transformation product curves in Fig. 10 show that the proportion of martensite to non-martensite at the 6 mm distance is approximately 70:30 vol/o. By examining the Fe-Si cumulative curve in Fig. 9 and finding the 30 cum vol/o on the y-axis (the non-martensite percentage), the silicon content needed to give the 70:30 ratio is shown to be approximately 0.37 wt/o. These data combinations are shown in Fig. 11, which is a predictor of the approximate silicon content needed to give a specific combination of transformation products. Local silicon contents greater than these levels (any content above the curve) increase the hardenability sufficiently to transform to martensite and conversely, any Si content below the curve will not.
Fig. 11

Predicted silicon content in the Fe-Si alloy required to raise the hardenability to the level at which martensite transforms. It should be noted, as seen in Fig. 10, that the fraction of martensite given in this plot falls from 100 vol/o near the surface to approximately 50 vol/o at the core of the cylinder

The information gained from these volume fraction estimates can also be used in alloy and process development. Quantities such as chemical analysis distribution, local hardness, microstructure, and density that show variability in the compact volume can be evaluated using these techniques. In alloy development, the effectiveness of complex or multiple additives with different base alloys can be investigated and correlated with both the combination of transformation products in the microstructure and the resulting material properties. Part processing variables of sintering time, temperature, and cooling rate can be evaluated by looking at alloy distribution and/or microstructure. This information may also be relevant in part design, where part size and shape are strong considerations in alloy selection and how the microstructure is transformed.


The following conclusions can be drawn from the alloying evaluation:
  • Silicon has been shown to be effective in enhancing the hardenability of steels as the sole alloying additive in ferrous metal powders.

  • Ferroalloys or more chemically complex particulate additives can be added to a base iron powder and alloy during sintering by solid-state diffusion.

  • Metallographic techniques can be used to estimate the volume fractions of both the transformation products and the chemical content distribution in a PM compact.

  • These estimates can be used to predict alloy compositions needed to harden compacts of specific sizes.

  • It may be possible to use these stereological metallographic techniques in the evaluation of alloying methods such as additive research, diffusion alloying, and combinations of solid and liquid phase diffusion to assess the distribution and the effectiveness of the added elements.

  • PM materials, especially those alloyed with particulate additives, are often heterogeneous in chemical composition and consequently, microstructure. Metallographic techniques used to quantify alloy distribution clearly help increase our understanding of these materials.



The authors would like to thank Wing-Hong Chen for his assistance in making and testing the samples used in this study and Jerry Golin, Barry Diamond, and Eric Alesczyk for their assistance with the metallographic sample preparation and testing.


  1. 1.
    C. Schade, T. Murphy, A. Lawley, R. Doherty, The influence of silicon on the mechanical properties and hardenability of PM steels. Int. J. Powder Metall. 49(4), 9–20 (2013)Google Scholar
  2. 2.
    T. Murphy, B. Lindsley, C. Schade, A metallographic examination into fatigue-crack initiation and growth in ferrous PM materials. Int. J. Powder Metall. 49, 23–34 (2013)Google Scholar
  3. 3.
    B. Lindsley, H. Rutz, Effect of molybdenum content in PM steels, in Advances in Powder Metallurgy and Particulate Materials, vol. 7, ed. by A. Lawley, P. McGeehan, R. Lawcock (MPIF, Princeton, NJ, 2008), pp. 1–10Google Scholar
  4. 4.
    N. Bredzs, C.C. Tennenhouse, Metal-metal oxide-hydrogen atmosphere chart for brazing or bright metal processing. in Welding Research. Supplement to the Welding Journal, May 1970, (Wall Colmonoy Corporation, Detroit, MI, 1970), pp. 189s–193sGoogle Scholar
  5. 5.
    M.N. Gungor, A statistically significant experimental technique for investigating microsegregation in cast alloys. Met. Trans. A 20A, 2529–2533 (1989)CrossRefGoogle Scholar
  6. 6.
    R.T. DeHoff, Application of stereology to the statistical description of the spatial distribution of composition and other scalar properties in three dimensional space. Acta Mater. 51, 2259–2269 (2003)CrossRefGoogle Scholar
  7. 7.
    E.E. Underwood, Quantitative Stereology, 1st edn. (Addison-Wesley Publishing Co. Inc., Reading, MA, 1970), pp. 29–30Google Scholar
  8. 8.
    J.E. Hilliard, Measurement of volume in volume, in Quantitative Microscopy, ed. by R.T. DeHoff, F.N. Rhines (McGraw-Hill Book Co., New York, NY, 1968), pp. 45–76Google Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature and ASM International 2018

Authors and Affiliations

  • Thomas F. Murphy
    • 1
    Email author
  • Christopher T. Schade
    • 1
  • Alan Lawley
    • 2
  • Roger Doherty
    • 2
  1. 1.Hoeganaes CorporationCinnaminsonUSA
  2. 2.Drexel UniversityPhiladelphiaUSA

Personalised recommendations