Effect of Composition and Cooling Rate on Physical Properties and Microstructure of Prealloyed P/M Steels Bruce Lindsley, George Fillari and Thomas Murphy Hoeganaes Corporation Cinnaminson, NJ 08077, USA ABSTRACT The composition of P/M steels plays an important role on the microstructure and physical properties of sintered parts. The different levels of alloying elements prealloyed into base powders change the hardenability of the material. In addition, copper and graphite additives also play an important role. Further, different cooling rates will have a profound effect on the microstructure. Three base alloys with different levels of copper and graphite were sintered at 1120°C and cooled at either conventional rates or under sinter-hardening conditions. The effect of these variables on microstructure and mechanical properties will be discussed, with an emphasis on dimensional change. INTRODUCTION Dimensional change in ferrous sintered compacts is influenced by several factors, including particle size, density, composition, sintering time and temperature, cooling rate and microstructure. Reduced particle size and increased sintering time and temperature can be used to reduce dimensional change. Given a set base iron and sintering conditions, factors such as composition and cooling rate play an important role on microstructure, dimensional change and mechanical properties. The most widely used additives to P/M iron are graphite and copper. Separately, copper and graphite additions cause growth in the sintered compact. Dilatometric studies have shown that in Fe-Cu alloys, growth is observed at 1083 °C, which is the melting point of copper. The amount of growth is a direct function of Cu content up to levels of 10% (13). Detailed dilatometry work has also shown that, upon heating, graphite quickly goes into solution in iron after the alpha to gamma transformation (1, 3-5). During this solutioning, the sample exhibits considerable growth beyond that associated with the thermal expansion of iron. As the carbon remains in the steel on cooling in the form of carbide or in solution in martensite, the sample retains this growth. Again, the amount of growth increases with the amount of graphite added to the alloy. The dimensional change (shrinkage) that occurs at the sintering temperature is smaller than the carbon solutioning growth (5). When both copper and carbon are added to a base alloy, there is an interesting interaction between the two alloying elements. The carbon inhibits Cu diffusion in the compact. It has been shown that carbon content changes the wetting angle and dihedral angle of liquid copper on iron; increased carbon increases both angles (1,3). This has an important ramification on copper diffusion in the compact, as at higher carbon levels, liquid Cu does not flow along particle surfaces and into grain boundaries, thereby reducing Cu diffusion. The amount of growth observed in a dilatometer at 1083 °C is significantly reduced in high carbon alloys. Therefore, growth and dimensional change in a high carbon high copper mix can be less than if either additive were made alone. Base iron composition and cooling rate have a pronounced effect on the microstructure that forms in a sintered steel. Several phases can form upon cooling in the sintering furnace. The high temperature phase in steel, austenite, transforms to either ferrite + carbide (in the form of pearlite, divorced pearlite or bainite), martensite, or a combination thereof. In addition, austenite may remain in the microstructure in heavily alloyed regions (retained austenite). Austenite is the highest density phase in steel, followed by ferrite and carbide, and finally martensite, which is the lowest density phase. This decrease in density results in a corresponding increase in compact growth, and martensite formation leads to the largest growth as a result of microstructural changes. Length changes in fully dense materials can reach 1.4% upon transformation from austenite to martensite (6). Interestingly, tempering of martensite leads to a ferrite and carbide structure, which has a higher density. Tempering should therefore decrease the dimensional change of a martensitic sintered compact. The composition of the steel will determine its hardenability, or the ability to form martensite. Steel is primarily hardened by rapid cooling from an austentizing temperature, in order to form martensite. The cooling rate required is dependent on alloy content. Alloying elements such as C, Mo, Ni, Mn, Cr and Cu will increase the hardenability of steel. The amount of prealloyed elements in the base iron will therefore affect the transformation products found after sintering. The cooling rate, prealloyed elements and admixed additions will determine the final microstructure and mechanical properties in the sintered compact. The purpose of this paper is to investigate processing and composition effects on the dimensional change, microstructure and mechanical properties of three base alloy systems. EXPERIMENTAL PROCEDURE Three commercially available base irons were selected for study, Table I. The materials were chosen to represent a range in hardenability, from a commercial pure base iron to an alloy specifically designed for sinter-hardening. To each base iron, 3 levels of Cu (0%, 1%, 2%) and 5 levels of graphite (0.6%, 0.7%, 0.8%, 0.9%, 1.0%) were added for a total of 15 mixes per base material. 0.75% EBS wax was used as a lubricant for all mixes. Standard TRS bars (MPIF or ISO designation here) were pressed to a green density of 7.0 g/cm3 and sintered in 90% nitrogen / 10% hydrogen at 1120 °C (2050 °F). Samples were cooled at two rates using a convective cooling unit in Abbott belt furnace. The cooling rates between 650 °C (1200 °F) and 315 °C (600 °F) were 42 °C/min (1.3 °F/sec) and 94 °C/min (2.8 °F/min). For each cooling rate, samples were grouped into 4 conditions: as-sintered; sintered and tempered; sintered and liquid nitrogen (LN) quenched; and sintered, liquid nitrogen quenched and tempered. Tempering was conducted at 200 °C (400 °F) for 1 hour under a nitrogen atmosphere. Quenching was performed in liquid nitrogen in an effort to transform any retained austenite in the samples. Samples were covered with tape and placed in liquid nitrogen, followed by a hold in a dry atmosphere. The tape and dry atmosphere were used to avoid condensation on the cold samples after the quench. Samples were measured before sintering and for all 4 conditions and both cooling rates to determine density and dimensional change. Dimensional change is determined by the difference between the length of the die and the sintered length of the bar, divided by the length of the die. Hardness and transverse rupture strength were also measured for all conditions. Table I. Nominal composition of base irons tested (in wt%). Base Iron MPIF Designation Mn Mo ® F-0000 0.1 Ancorsteel 1000B Ancorsteel 4600V FL-4600 0.15 0.56 Ancorsteel 737SH FL-4800 0.42 1.25 Ni 1.83 1.4 Samples for metallographic examination were cross-sectioned, mounted, ground and polished using well established practices. The samples were etched in a 4 wt% picral – 2 vol% nital solution to reveal the microstructure. For retained austenite, samples were prepared by stain etching with an aqueous solution of 25 wt% sodium bisulphite, which stained the martensite and left the austenite unstained (white). Quantitative image analysis was used to quantify the amount of retained austenite. In addition, select as-sintered samples were polished and quenched in liquid nitrogen. Surface relief due to transformation of the retained austenite to martensite was detected by Nomarski differential interference constrast. Finally, as-polished samples were analyzed using electron dispersive spectroscopy (EDS) in the SEM to measure the copper content. Composition maps, line profiles and point spectra were used to measure the Cu distribution. RESULTS As-sintered The effect of Cu and C on the dimensional change (DC) of three different base alloys is shown in Figure 1. The overall shape of the curves is the same for all three alloys. With no added Cu, the 0.6% graphite addition had the smallest dimensional change of all compositions and the addition of further graphite to 1% increased the dimensional change (causes growth). At 1% Cu addition, the addition of graphite between 0.6% and 1.0% has little effect on dimensional ® Ancorsteel is a registered trademark of Hoeganaes Corporation change and at 2% Cu, the addition of C decreases the dimensional change. The difference in dimensional change is greatest at 0.6% graphite for the different levels of Cu, and further increases in C content reduces this difference as the curves tend to merge at high graphite levels. The interaction between the Cu and C has been shown by other authors in the literature, although these effect are generally not presented in this fashion. It can be seen that with the FL-4800 base alloy, the 2% Cu composition curve crosses the 0% Cu composition at 0.9% graphite. The graphite content at which these curves cross is much lower for the FL-4800 alloy than the other base alloys, and will be discussed later. Figure 1. Effect of C and Cu on dimensional change of 3 different base alloys. It is interesting to note that the two base materials that are prealloyed with Ni and Mo have a larger range in dimensional change as compared to the pure iron base material. At 0 and 1 % Cu, these prealloys have a smaller dimensional change from die size as compared to the pure iron base. However, at 2% Cu, the prealloys have considerably more growth than the pure iron base at 0.6% graphite, and the reduction in dimensional change with graphite is considerably steeper than the pure iron base. It can be seen that the 1% Cu mixes for all alloys are very dimensionally stable with respect to carbon content. However, typical sinter-hardening compositions consisting of Ni-Mo prealloys and 2% Cu admixes are not. Control of graphite content is critical in these compositions, both in the overall mix and locally within a compact. Bonding of the graphite to the base alloy is an effective way to minimize dimensional change variations in these compositions. The shape of the curves clearly displays the copper – carbon interaction. If no interaction existed, one would expect a series of parallel lines, each for a different Cu content, and all with positive slopes. It is thought that the effect of carbon on copper diffusion throughout the compact causes the different slopes. As carbon increases, copper diffusion decreases to the point where the 0% and 2% copper curves behave similarly (ie. at 1% graphite in Figure 1a). The liquid Cu diffusion is restricted to the point that with respect to dimensional change, it acts as if it had not been added. Literature data suggests that at high C levels, copper no longer diffuses down grain boundaries and is limited to the pore / prior particle boundary regions. To determine if the copper distribution changed with carbon content, line scans using electron dispersive spectroscopy (EDS) in an SEM were made across prior particles in sintered compacts of FL-4800 base alloy with 2% Cu and different carbon levels. A typical trace and copper profile is shown in Figure 2. The Cu levels are highest near the pores and decrease toward the center of the base iron particle. Elevated Cu levels were found to a depth of nominally 10µm from the edge of the particle. The results for several scans at the different graphite levels tested (0.6, 0.8, 1.0) are shown in Table II. The average values and ranges in Cu content for the edge and center locations are shown. It can be seen that the copper content near the pores is higher for the high carbon alloys, while the average value at the center of the particles is higher for the low carbon sample. The large range in Cu levels in the center of the particles at 0.6% graphite may be the result of Cu diffusion along grain boundaries, although no effort was made to determine if grain boundaries were present at these locations. The carbon content clearly affects the Cu distribution in the samples. No difference in C content was found along the same line scans. Figure 2. Line scan and Cu composition profile for FL-4800 with 2% Cu and 0.8% graphite. Table II. Effect of carbon on Cu distribution. Copper content measured by EDS analysis. Cu Content at Particle Edge (%) Cu Content in Particle Core (%) Graphite Content (%) Average Range Average Range 0.6 2.8 2.2 - 3.3 0.32 0.0 - 0.81 0.8 3.7 3.0 - 4.3 0.21 0.18 - 0.23 1.0 3.8 3.4 - 4.5 0.11 0.1 - 0.11 Another approached was used to show the effect of carbon on admixed copper diffusion and its subsequent effect on dimensional change. A 1% Cu prealloy was made and tested against the pure iron base (F-0000) at 0% and 1% admixed copper. Figure 3 shows the 1% Cu prealloy behaves like the pure iron base with no admixed copper. The prealloyed Cu is well distributed prior to carbon addition, and hence the effect of carbon on wetting and dihedral angle is irrelevant as no liquid copper is present. The dimensional change is clearly affected by the presence of admixed Cu by its penetration into grain and particle boundaries and corresponding spreading tendency in the liquid form, as other authors have shown previously (1-4). Figure 3. Dimensional change of the iron base with 0% and 1% admixed copper and an iron-1% Cu prealloy with different graphite contents. Effect of Tempering Tempering of steels is an important operation for sinterhardening grades. These grades form martensitic structures that are hard and brittle, and a low temperature heat treatment will greatly improve the tensile properties of the steel with a small reduction in hardness. Martensite is the lowest density phase in steels, and its formation causes growth. Tempering of martensite increases its density by changing the body center tetragonal (BCT) structure to ferrite (BCC) + carbide, thereby reducing the overall dimensional change of the compact. The effect of tempering on the dimensional change was studied for the three base materials. Little to no martensite formed with the pure iron base material at any of the conditions tested, as the cooling rates in the furnace were not high enough and the material has poor hardenability. Therefore, tempering had no effect on the dimensional change with this base material. Correspondingly, tempering had no effect on the hardness of these compacts either. With the addition of prealloyed Mo and Ni (and Mn), the hardenability of the FL-4600 and FL4800 is significantly higher than the pure iron base (F-0000), resulting in martensite formation during cooling after sintering (sinter-hardening) at high levels of C and Cu. The effect of tempering on dimensional change can be seen in Figure 4. Again, the data is plotted vs graphite content for the three different copper levels. The effect of tempering with 0%Cu is negligible, indicating little martensite formation with either base alloy. The 0%Cu curves were left out of Figure 4 for aesthetics. At 1% and 2%Cu, a difference in dimensional change with tempering was found for both materials. This difference increases with carbon content, with a change in DC of about 0.1% with 1% graphite. These graphs show that increasing martensite formation leads to greater differences in DC with tempering. The martensitic structure for FL-4800 with 2% Cu and graphite is shown in Figure 5. The lower carbon martensite has more of a lath structure, while the high carbon material is plate martensite. Prealloyed materials, such as FL-4800, lead to a uniform microstructure. Figure 4. The effect of tempering on the dimensional change of FL-4600 and FL-4800 sinterhardening base alloys. 0.6% 1.0% Figure 5. Microstructures of FL-4800 + 2% Cu at 0.6% and 1.0% graphite. As noted earlier, tempering is critical to mechanical properties other than hardness. The effect of tempering on the transverse rupture strength of the FL-4800 alloy with different copper and graphite levels is shown in Figure 6. A dramatic difference in strength is apparent with tempering of the martensitic compositions. The 1% and 2% Cu alloys with 0.9% graphite show strength increases from 1150 MPa to approximately 1600 MPa. Again, the 0% Cu alloy and the 1% Cu with 0.6% graphite composition are not martensitic under these conditions, so little change is evident with tempering. Similar effects of tempering were found with the FL-4600 material, and tempering of the F-0000 alloys had no effect on transverse rupture strength. As-Sintered 205 °C/1hr Temper Figure 6. Effect of tempering on the transverse rupture strength of FL-4800. Retained Austenite In martensitic steels with carbon contents higher than the eutectoid composition, plate martensite (also known as twinned martensite) becomes the predominate microstructure. As the hardenability of the steels increase due to carbon content and metallic alloying, the martensite finish temperature decreases below room temperature, at carbon contents above 0.6% (6). This results in some amount of retained austenite in the material. Below 0.6% C, the amount of retained austenite becomes negligible. An example of this can be found in steel carburization. In carburized steels, where carbon contents of 0.9%-1.0% are found in the case layer, the microstructure consists of plate martensite and retained austenite. The amount of retained austenite is often quoted as an important variable in carburized steels for gear applications. The same phenomenon occurs in sinter-hardening P/M steel grades, where carbon contents near 0.9% are often used. Between the martensite plates, some level of austenite is maintained from the sintering temperature. Evidence of retained austenite in a sinter-hardening composition has been shown previously (7,8). The presence of retained austenite affects the density of the metal, and therefore the dimensional change of the compact. A sample with retained austenite will have a lower absolute dimensional change than if the austenite all transformed to martensite. To test this supposition, samples were quenched in liquid nitrogen after sintering. The lower temperature will drive the austenite to martensite transformation. Those samples with significant levels of retained austenite should grow due to the formation of additional martensite. The pure iron base mixes, which did not form martensite and therefore contained no retained austenite, did not respond to the liquid nitrogen quench. No change in DC or hardness was observed. The sinter-hardenable alloys with 0%Cu also showed no effect of the liquid nitrogen quench, and have not been plotted. With increasing Cu and C contents, the effect of liquid nitrogen quenching on dimensional change was significant, Figure 7. Growth of the samples was observed with the quench, and at 1% graphite and 2% Cu, the effect of the liquid nitrogen quench on DC surpassed +0.1% for both base alloys. The increased alloy content of the FL-4800 base material increased hardenability over the FL-4600 base material. Along with the increase in hardenability also comes an increase in retained austenite. This is reflected in larger changes of DC with the liquid nitrogen quench. At 1%Cu, the hardenability of the FL-4800 base was great enough on the small test pieces to form martensite with retained austenite, whereas the FL-4600 base material did not show evidence of retained austenite. Figure 7. Effect of liquid nitrogen quench after sintering on the dimensional change of the Ni-Mo prealloyed base materials. Figure 8. Retained austenite (white) in FL-4800 with 2% Cu and 1.0% graphite. The samples were etched to reveal the retained austenite by light optical microscopy. Figure 8 shows the FL-4800 with 2% Cu and 1% graphite, and the retained austenite appears white in the microstructure. The retained austenite resides between the martensite plates and is found in the Cu rich regions of the P/M steel. The combination of high C and high Cu contents result in locally high retained austenite contents in the prior particle boundary regions. The retained austenite content was measured quantitatively by optical microscopy and plotted against the change in dimensional change due to the liquid nitrogen quench, Figure 9. A good correlation was found between retained austenite and the growth associated with the quench. Some amount of retained austenite was apparently not detected by optical microscopy, as evidenced by the two compositions with nominally 0% austenite and different growths. In wrought steels, up to 3% retained austenite cannot be measured by optical means (6). Figure 9. Correlation between retained austenite measured by optical microscopy and growth associated with the liquid nitrogen quench. FL-4800 with 2% Cu and 0.6 to 1.0% graphite. a b Figure 10. Surface relief in polished sample after liquid nitrogen quench. Sample imaged with differential interference contrast. Particle core (C) shows no surface relief in b. FL-4800 with 2% Cu and 0.9% graphite. Metallographic samples in the as-sintered state were polished and subsequently quenched in liquid nitrogen. Figure 10 shows the surface relief due to the martensite that formed as a result of the quench. Recall the austenite is present between the martensite plates and individual regions are quite small. The 0.9% graphite sample contained approximately 5% retained austenite, which matches well with the amount of surface relief found on the sample. No surface relief was present in the center of large particles, where Cu diffusion is limited. This is consistent with the etched microstructure in Figure 8. The presence of retained austenite is partially responsible for the slope of the 2% Cu curve in the FL-4800 material, Figure 1C. The increased density of the austenite found at high C levels increased the negative slope of the 2% Cu curve, so that the 2% Cu curve crosses the 0% Cu curve at lower C levels than the other base alloys. After quenching and transformation of retained austenite, the 2% Cu curve no longer crosses the 0% Cu curve at the carbon levels tested, and the FL-4800 base alloy behaves similarly to the other alloys. The amount of retained austenite can be used as a tool to control the dimensional change to a small degree in these compositions. Some of the samples quenched in liquid nitrogen were subsequently tempered at 200°C (400°F) for 1 hour. The tempering had a similar effect as the tempering of the as-sintered bars, where the density increased and the dimensional change decreased after tempering. The net effect of the quench and temper together was that the dimensional change of these samples was similar to the as-sintered samples. The additional growth due to quenching was balanced by the shrinkage due to tempering. Figure 11 shows the dimensional change of the 2 Ni-Mo prealloyed base materials with 2%Cu and 1.0% graphite and the 4 different processing conditions described above. Figure 11. Effect of tempering, liquid nitrogen (LN) quench, and LN quench and temper (Q & T) on the dimensional change of Ni-Mo prealloys with 2%Cu and 1.0% graphite. Effect of Cooling Rate The effect of cooling rate was studied using a convective cooling unit in the sintering furnace. The cooling rate in the range from 650 °C to 315 °C was increased from 0.7 to 1.6 °C/sec (1.2 to 2.8 °F/sec). Accelerated cooling is used to increase the amount of martensite formation and thereby the hardness of sintered compacts. It is expected that an increase in martensite formation should increase the dimensional change of the samples. The pure iron base samples showed no effect of accelerated cooling with respect to either dimensional change or hardness. The inherent hardenability of this material is sufficiently low that no martensite formed in any of the conditions tested. The cooling rate difference was not great enough to avoid the pearlite transformation. The two Ni-Mo prealloyed materials showed a significant effect of cooling rate on both dimensional change and mechanical properties. The increase in cooling rate increased the amount of martensite in many of the compositions. The effect was especially pronounced in the alloys with less copper and graphite. A significant increase in hardness was found for all combinations, except the FL-4600 + graphite (no copper) alloys. Figure 12 shows the effect of accelerated cooling on the hardness of FL-4800 for all Cu-C combinations. At the higher cooling rate, copper was not required to develop a fully hardened structure. The cooling rate of 1.6 °C/sec is well within the range of typical sinter-hardening cooling rates. The inherent hardenability of alloy FL-4800 allows for sinter-hardening without the addition of Cu. In parts with heavy gauge sections, the cooling rate may decrease, and additions of Cu can be made to ensure a fully hardened structure. The FL-4600 grade with 1% and 2% Cu and graphite levels ≥ 0.7% also showed an excellent sinter-hardening response at 1.6 °C/sec. Hardnesses of ≥ 70 HRA (40 HRC) were found under these conditions. 0.7 °C/sec 1.6 °C/sec Figure 12. Effect of cooling rate on the as-sintered hardness of FL-4800 alloys. It is interesting to note that not all increases in hardness were associated with a measurable increase in dimensional change. For example, both the FL-4800 + 1Cu + 0.8Gr and the FL-4600 + 2Cu + 0.8 Gr compositions showed hardness increases of 5 HRA (roughly 10 HRC) due to the increase in cooling rate, but did not result in a measurable increase in DC. When changes in DC are observed due to cooling rate, there was a large change in the material hardness. Figure 13 shows the effect of cooling rate on hardness and dimensional change on the two Ni-Mo prealloys with no copper added. The FL-4600 base does not have sufficient alloying to sinter-harden under these conditions, so little change in dimensions and hardness were found. However, the FL-4800 alloy does sinter-harden at the higher cooling rate, resulting in martensite formation with a corresponding increase in dimensional change and hardness. An increase in dimensional change of 0.1% corresponded to a 20HRA increase in hardness. (a) (b) Figure 13. Effect of cooling rate and base material on as-sintered hardness and dimensional change in alloys with no Cu. Increased alloy hardenability in (b) lead to martensite formation with accelerated cooling, resulting in increased DC and hardness. Solid symbols are dimensional change, open symbols are apparent hardness. CONCLUSIONS The composition of P/M steels and cooling rate in the sintering furnace play important roles on the microstructure and physical properties of sintered parts. The results from the three base alloys studied have shown that dimensional change is a function of many variables, including base alloy composition, Cu and C content, post sintering processing (cooling rate, tempering and liquid nitrogen quench) and microstructure. Evidence of the inhibiting effect of carbon content on copper distribution supports earlier work in the literature with respect to the interaction between Cu and C, and the resulting effect on dimensional change was shown. The composition of the alloys clearly affects the hardenability within the cooling rate range of sintering furnaces. Dimensional change control of those compositions that sinter-harden is more challenging than traditional materials, as martensite formation results in more growth than other phases, while tempering can be used to reduce this growth. The presence of retained austenite in heavily alloyed regions reduces the overall growth of sintered compacts. Mechanical properties, such as hardness and transverse rupture strength, were heavily dependent on the processing and the final structure of the material. ACKNOWLEDGEMENTS The authors would like to thank Ron Fitzpatrick, Eva Wagner and Dave Cochran for their assistance in collecting the data within this paper. REFERENCES 1. N. Dautzenberg and H. J. Dorweiler, “Dimensional behavior of copper-carbon sintered steels”, Powder Metallurgy International, Vol. 17, No. 6, 1985, p. 279. 2. Y. Trudel and R. Angers, “Properties of iron copper alloys made from elemental or prealloyed powders”, Int. Journal of Powder Metallurgy & Powder Technology, Vol. 11, No. 1, 1975, p. 5. 3. R. L. Lawcock and T. J. Davies, “Effect of carbon on dimensional and microstructural characteristics of Fe-Cu compacts during sintering”, Powder Metallurgy, Vol. 33, No. 2, 1990, p. 147. 4. F. Chagnon and M. Gagne, “Dimensional control of sinter hardened P/M components”, Advances in Powder Metallurgy & Particulate Materials, compiled by W. G. Eisen and S. Kassam, Metal Powder Industries Federation, Princeton, NJ, 2001, part 5, p. 5-31. 5. F. J. Semel, “Processes determining the dimensional change of P/M steels”, Advances in Powder Metallurgy & Particulate Materials, compiled by W. G. Eisen and S. Kassam, Metal Powder Industries Federation, Princeton, NJ, 2001, part 5, p. 5-113. 6. R. E. Reed-Hill, Physical Metallurgy Principles – 2nd Edition, PWS-Kent Publishing, Boston, MA, 1973, p. 701-748. 7. T. Murphy, “An investigation into the effect of copper and graphite additions to sinterhardening steels”, Advances in Powder Metallurgy & Particulate Materials, compiled by W. B. James and R. A. Chernenkoff, Metal Powder Industries Federation, Princeton, NJ, 2004, Part 10, p. 266. 8. M. C. Baran, A. H. Graham, A. B. Davala, R. J. Causton and C. Schade, “A superior sinterhardenable materials”, Advances in Powder Metallurgy & Particulate Materials, compiled by C. L. Rose and M. H. Thibodeau, Metal Powder Industries Federation, Princeton, NJ, 1999, Part 7, p. 185.