Resource estimation for heterogeneous computing

Materials Science and Engineering A336 (2002) 274– 319 www.elsevier.com/locate/msea

Review

Nanostructured coatings Jianhong He, Julie M. Schoenung * School of Engineering, Uni6ersity of California Ir6ine, Ir6ine, CA 92697 -2175, USA Received 20 June 2001; received in revised form 11 December 2001

Abstract This review is based essentially on the results in the field of synthesis and characterization of nanostructured coatings obtained by the authors themselves or their colleagues. Characteristics of feedstock powders for synthesizing nanostructured coatings such as particle size and morphology, changes in chemical composition and grain size are summarized. The evolution of microstructure caused by mechanical milling in two typical powder systems, Cr3C2 – NiCr and Inconel 625, and mechanisms governing the development of the nanostructure are discussed. Mechanical properties and microstructure of several nanostructured coatings are evaluated by using microhardness testing, scanning electron microscopy, transmission electron microscopy, and X-ray diffraction. As background information, a review of the agglomeration process for milled powders and of thermal spraying technologies to synthesize nanostructured coatings are also included. In addition, the methodologies used to characterize the performance of the milled powders and nanostructured coatings, as well as the practice techniques for sample preparation, are described in detail. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Nanostructured; Coatings; Mechanical milling; High velocity oxygen fuel (HVOF)

1. Introduction In thermal spraying technology, molten or semimolten powders are deposited onto a substrate to produce a two-dimensional coating or in some cases, a three-dimensional self-standing material. The microstructure and properties of the material depend on the thermal and momentum characteristics of the impinging particulate [1], which are determined by both the spraying methodology and the type of feedstock materials employed. Powders, rods and wires can be used as feedstock materials. Various coatings are deposited on the surface of a substrate to either provide or improve the performance of materials in industrial applications. Nanostructured (or nanocrystalline) materials are characterized by a microstructural length scale in the 1 – 200 nm regime [2]. More than 50 vol.% of atoms are associated with grain boundaries or interfacial boundaries when the grain is small enough. Thus, a significant amount of interfacial component between * Corresponding author. Tel.: +1-949-824-5620; fax: + 1-949-8243672 E-mail address: [email protected] (J.M. Schoenung).

neighboring atoms associated with grain boundaries contributes to the physical properties of nanostructured materials [3]. Using nanostructured feedstock powders, thermal spraying has allowed researchers to generate coatings having higher hardness, strength and corrosion resistance than the conventional counterparts [4–6]. The purpose of this paper is to review: (a) the synthesis and characterization of nanostructured feedstock powders; (b) the agglomeration of these powders for use in coatings; and (c) the processing and characterization of nanostructured thermal spray coatings.

2. Synthesis of nanostructured feedstock powders Preparation of nanostructured feedstock powders is the first step for synthesis of nanostructured coatings. A number of techniques that are capable of producing nanostructured materials include gas condensation, mechanical alloying/milling, crystallization of amorphous alloys, thermochemical method, spray conversion processing, vapor deposition, sputtering, electro-deposition, and sol –gel processing techniques [7]. Of these techniques, only mechanical alloying/milling and ther-

0921-5093/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 0 9 3 ( 0 1 ) 0 1 9 8 6 - 4

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

mochemical techniques have been used to produce large quantities of nanostructured materials for possible commercial use [7]. Mechanical milling is reportedly employed to synthesize nanostructured powders of varying compositions [6 –19]. Fig. 1 shows a mechanical milling facility for industrial use. A single milling run can produce 20 kg of nanostructured powder. The thermochemical approach is primarily used for the synthesis of tungsten carbide (WC)– cobalt (Co) nanostructured powders [20,21]. Therefore, in this paper, mechanical milling has been chosen as the most viable technique for producing nanostructured feedstock powders. Historically, the mechanical alloying/milling process was originally developed by the International Nickel Company in 1966 for the production of oxide disper-

Fig. 1. Attritor developed by Rocketdyne Power and Propulsion, which is able to cryomill up to 20 kg powder in a single batch.

275

sion strengthened superalloys [22]. More recently, the mechanical milling process has attracted considerable interest, primarily as a result of its potential to generate nanocrystalline and other non-equilibrium structures in large quantities [8–19]. Vibratory mill, planetary mill, uniball mill and attritor mill are used to carry out mechanical milling [19]. Of the four types of mills, only the attritor mill has the highest capacity of powder charge. Accordingly, the attritor milling is employed to synthesize feedstock powder used for the fabrication of nanostructured coatings primarily because a large quantity of powder (e.g. 500 g) is needed for the synthesis of a coating. In this paper, unless specified otherwise, the milling has been accomplished using an attritor mill at a rotating rate of 180 rpm in a stainless steel tank. The ball-to-powder mass ratio used is 20:1. Fig. 2 shows a schematic diagram of an attritor mill. In an attritor mill, the powder and balls are charged into a vertical cylindrical tank. Several horizontal impellers, joined to a rotating vertical shaft driven by a motor, generate movement of the balls and the powders. The moving balls, which are energized by rotating impellers, cause grain size refinement of the powder particles by impact. Concurrently, a controlled milling environment is introduced into the tank for the following reasons: (1) to decrease contamination of powder caused by air; (2) to control powder temperature because milling causes an increase in temperature of the powders; (3) to introduce a reactive solvent to form the expected phase; and (4) to control powder particle size and morphology. The milling environment used for these purposes was: (a) gaseous environment (i.e. argon, nitrogen and hydrogen); (b) liquid environment (such as liquid nitrogen, methanol, acetone and hexane); and (c) solid environment (i.e. stearic acid). Argon, dry nitrogen, methanol, acetone and hexane were chosen and used for reducing the contamination of powder from the atmosphere and controlling the temperature. Hydrogen is used to form hydrides. Liquid nitrogen reduces contamination of the powders and provides control of the temperature, but also works as a reactive solvent to form nitrides that facilitate in retarding grain growth of the synthesized powder during thermal exposure [23]. The oxygen-rich stearic acid is used to reduce powder particle size through a reduction in the extent of welding between powder particles.

3. Characterization of nanocrystalline powder synthesized by mechanical milling

Fig. 2. Schematic diagram of attritor mill.

It is important to be able to identify powder characteristics, such as particle size, powder morphology, grain size, phase constituents and deformation faults as a function of milling parameters. Unfortunately, a realtime monitoring system for quantifying these character-

276

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 3. Dependence of the average particle size of attritor milled powders on milling time [25,32].

istics is not available. The construction of the tank permits the removal of powder samples from the tank during milling to conduct analysis of powder characteristics without disturbing the milling. Therefore, the dependence of powder characteristics on milling time can be established for an individual powder system. To determine whether a nanocrystalline structure has been achieved or not, and to also understand the milling mechanism, powder samples are unloaded from the tank at pre-determined time intervals. The samples were analyzed using various instruments: (a) X-ray diffraction (XRD) for structure and strain analysis; (b) scanning electron microscopy (SEM) for morphological analysis; (c) transmission electron microscopy (TEM) for structure and morphology characterization; (d) particle size analyzer for quantifying particle size; (e) nanoidentator for determining mechanical characteristics, and (f) thermal analysis instruments for phase transformation analysis. In this paper, several powder characteristics are described and the effects of milling conditions are elaborated.

3.1. Particle size To ascertain the continuous flow of feedstock powder in the powder feed system, the thermal spray process typically requires a powder particle size within the range of 10–50 mm. In addition, the particle size significantly affects particle temperature and speed during flight which subsequently influences the properties of the coating. Therefore, the particle size is an important parameter for thermal spraying. The size distribution of the particles can be determined by a particle size analyzer or using statistical measurements made in scanning electron micrographs. It should be noted here that

the nanocrystalline nature of the feedstock powder and the subsequent coating is not determined by the particle size, which for nanocrystalline materials is actually agglomerate size. Instead, it is determined by composition of the powder and the milling parameters, as is discussed later. Fig. 3 reveals the dependence of the average particle size of the powders on milling time [25]. In Fig. 3, Cr3C2-25 (Ni20Cr) powder is milled in hexane [H3C(CH2)4CH3], and Inconel 625 superalloy powder is milled in liquid nitrogen. As milling time increases, the average particle size of Cr3C2-25 (Ni20Cr) powder decreases and approaches a relatively constant value of 5 mm, while that of Inconel 625 increases and approaches a constant value of 84 mm. The commonality of the average particle size of the two powders is that a drastic change occurs within the first 4 h of milling, and subsequently stabilizes as milling time increases. The fact that the average particle size changes as milling time increases and approaches a constant value is an indication of fracture and cold welding occurring during milling. The smaller particles grow while the larger ones fracture. Benjamin [24], the developer of mechanical alloying/milling, proposed a mechanism to explain the stabilization of the average particle size. The fracture strain of the particle decreases with increasing particle size, while the strain value at which cold welding occurs remains constant. Hence, average particle size tends to stabilize as milling time increases. The fracture and welding strains depend strongly on powder characteristics and milling conditions. In related studies, it is reported that the average powder size of Ni-base superalloys with a small amount of Y2O3 increases and approaches a constant value, between 110 and 130 mm as milling progresses [24]. Brittle powder systems, such as the Cr3C2-25 (Ni20Cr) system, possess a lower fracture strain; and the average particle size approaches a constant value on the order of 10 mm [25]. By contrast, a continuous decrease in powder particle size for both ductile–ductile (Ti–Al) and ductile–brittle (Ti –Si) systems occurs with increasing milling time [26]. The milling environment also affects the relationship between average particle size and milling times of milled Ni powder [27]. As milling time increases, the Ni powder particle size decreases when milled in methanol but increases when milled in liquid nitrogen. On the basis of the data in the published literature, the steadystate particle sizes of some milled powders are listed in Table 1. Generally, brittle material systems lead to smaller steady-state particle sizes, while ductile material systems yield larger steady-state particle sizes.

3.2. Powder morphology Regardless of the initial morphology of as-received powders, milling causes a drastic change in morphology

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

of the powders that undergo severe plastic deformation during milling. Although a light optical microscope is sometimes used to examine the morphology of milled powders, SEM is more commonly used because of its long focal-depth. Fig. 4(a) [25] shows the morphology of as-received Cr3C2-25 (Ni20Cr) powder, which consists of a Cr3C2 carbide phase and a Ni20Cr phase (a solution of 20 wt.% Cr in Ni). A small amount of the powder exhibits a spherical morphology while the re-

277

mainder shows an irregular morphology having sharp facets. Fig. 4(b) shows the result of X-ray energy dispersive spectrum (EDS) mapping of Ni. There is a high Ni content in the regions where spherical particles appear, while almost no Ni appears at the sites where irregular-shaped particles are present. This indicates that the Ni20Cr phase shows a spherical morphology and Cr3C2 carbides are present in the form of irregularshaped particles. Fig. 4(c) and (d) shows the changes in

Table 1 Steady-state particle size of milled powder Material

Ball to powder ratio

Environment

Particle size (mm)

References

BaFe12O19 BaFe12O19 Cr3C2-25 (Ni20Cr) Ti–Si Ti–Al Ni–20Cr–5Al–1Ti–1Y2O3 67%Ni–33%W Ni-base alloy with small amount of Y2O3 Co–29Cr–6Mo Inconel 625 Pure Al

60:1 60:1 20:1 40:1 40:1 25:1 10:1 20:1 20:1 20:1 30:1

Air Low pressure air (10−110−2 Pa) Hexane Argon Argon Argon Argon

0.1–0.5 1 5 B10 B10 2030 75 100130 106 84 150

[28] [28] [25] [26] [26] [29] [30] [24] [31] [32] [32]

Methanol Liquid nitrogen Liquid nitrogen

Fig. 4. SEM micrographs showing the morphology of the Cr3C2 – NiCr powder for different milling times: (a) As-received powder; (b) Ni-map of (a) using EDS; (c) 8 h milled powder; and (d) 16 h milled powder [25].

278

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 5. SEM images showing the morphology of the binder-induced, self-agglomerated composite powders. (a) 4 h milled powder; (b) higher magnification of (a); (c) 20 h milled powder; and (d) higher magnification of (c) [33].

the powder after 8 and 16 h milling, respectively. After 8 h of milling, some of the brittle (hard) carbide phase is unaffected by the milling, as evidenced by the remaining carbide particle, seen at the arrow. X-ray mapping reveals the particle, having sharp facets, is a carbide particle. After 16 h of milling, the pure carbide particles were not observed. In the milled powder, the ductile (soft) Ni-rich phase is no longer seen as spherical particles, but is distributed among the homogeneous, finer particles. The surface of the powder becomes rough with an increase in milling time. In milled composite powders consisting of a brittle phase constituent and a ductile binder constituent, such as Cr3C2 –NiCr and WC – Co systems, it is difficult to clearly observe individual particles without using a high resolution SEM. Furthermore, during mechanical milling, powder self-agglomeration is often observed. Two types of self-agglomeration behavior are observed in a composite powder system: (a) binder-induced agglomeration, and (b) metallurgical agglomeration [33]. Fig. 5 shows SEM images illustrating binder-induced agglomeration. After 4 h of milling, see Fig. 5(a), a large proportion of the small particles are self-agglomerated. However, the larger carbide particles, with sharp facets, remain non-agglomerated. A detailed view of the agglomerated powder, indicated by the arrow in Fig. 5(a), is shown in Fig. 5(b). The regular rectangular parallelepiped particles present, as indicated by the arrows, are thought to be Cr3C2 carbides having an orthorhombic crystal structure (JCPDS—International Center for Diffraction Data, file number: 35-084) that fracture along low Miller index planes, possibly the {100} planes, during milling. X-ray dot mapping

confirms these particles to be carbides. Fig. 5(c) shows almost complete self-agglomeration in the powders that have been milled for 20 h. A higher magnification of the agglomerated powder indicated by the arrow in Fig. 5(c) is shown in Fig. 5(d). X-ray mapping results reveal the particles in Fig. 5(d) to be neither pure carbide nor pure NiCr phase. The self-agglomerated powders, shown in Fig. 5, are not very dense, and serve as an example of binder-induced agglomeration. Binder-induced agglomerates are primarily bonded by milling environment, and can be decomposed back to their original powder form. For example, when the powder agglomerates are immersed in methanol and vibrated for a few minutes, the agglomerates tend to break down. Fig. 6 shows TEM images illustrating the process of metallurgical agglomeration. A metallurgical agglomerate is made of particles A–D. The selected area diffraction (SAD) pattern on the interface of powders A and B is shown on the bright field image. This is a complex pattern of polycrystal diffraction rings and single crystal diffraction spots. On the side of powder A, the pattern comes from a polycrystal. In fact, a number of sharp carbide fragments can be observed in powder A. While, on the side of powder B, the pattern is essentially one of an fcc single crystal. Similar complex SAD patterns are also diffracted on the interfaces of powders A with C, and A with D. Therefore, powder A is a polycrystal composite, and powders B–D are essentially single crystal NiCr solid solutions that are embedded with a few fragments of carbide particles. The interface of powders A and B is completely continuous and without microcracks or microvoids. The continuous and smooth interfaces between the powder particles in the agglomerates provide direct evidence that mechanical milling promotes metallurgical bonds between ductile NiCr solid solution materials and polycrystal composite powders. The self-agglomerated powder in Fig. 6 is characterized by the metallurgical bond between the powders. The self-agglomeration process is primarily controlled by cold welding and fracturing that occurs during the milling process, and cannot easily be decomposed because of the presence of metallurgical bonds between powders. The measurement on micrographs of high resolution SEM indicates that the average particle size of Cr3C2-25 (Ni20Cr) powder, shown in Fig. 3, is actually the average size of self-agglomerates. The real particle size is plotted in Fig. 7 [33]. For comparison, the average size of self-agglomerates is also re-plotted in Fig. 7. In the powder milled for 2 h there is no evidence of self-agglomeration. However, in the powder milled for 4 h a large portion of the smaller particles self-agglomerate; while the larger carbide particles remain segregated. In an individual powder system, the size of the binder-induced agglomeration powder is thought to

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

279

Fig. 6. TEM micrograph of metallurgical agglomerated Cr3C2 – NiCr powders: (a) bright field image; and (b) dark field image [25].

depend on the dynamic factors in the milling process, such as rotation, ball-to-powder mass ratio, tank dimensions, and the behavior of the milling environment used. Dry inert gas, air, liquid nitrogen, liquid chemicals (such as methanol, acetone and hexane) are often used as the milling environment. The behavior of the milling environment affects the size of the agglomerates. In some individual cases, binder chemicals are often added to the powder to change behavior of the binder-induced agglomeration. However, a systematic study on the binder-induced agglomeration is currently unavailable. Under the conditions used in Ref. [24], the size of the binder-induced agglomeration particles approaches a constant value of 5 mm after 8 h of milling. Interestingly, the particle size also approaches a constant value of 0.5 mm. Fig. 8 reveals the morphology of Inconel 625 powder milled in liquid nitrogen [32], a process referred to as cryomilled hereafter. In contrast to the milled composite powder particles, the interfaces of the milled Inconel 625 powder particles are well defined. Moreover, the density of the agglomerates is high. In this particular case, the agglomerated particle size of the powder is equal to the powder particle size. High resolution SEM micrographs on cross-sections of the powder indicate that the agglomeration is a metallurgical agglomeration, which is formed by cold welding. Ductile metals (such as Al, Co and Ni) and pre-alloyed (i.e. stainless steels and superalloys) cryomilled powders usually show a similar morphology. Fig. 9(a) and (b) show the influence of milling environment on the powder morphology [27]. Ni powder milled in liquid nitrogen shows well-defined powder particles. The size distribution of the powder particles is relatively narrow. In contrast, milling in methanol re-

sults in powders having a flake-like morphology in which a large quantity of fine fragments are present.

3.3. Changes in chemical composition due to milling Contamination is unavoidable during the mechanical milling process although its extent can be limited under some conditions. Two issues associated with contamination are often critical: (a) what is the extent of contamination? (b) how does the contamination affect the properties of the resultant materials? To profile the extent of contamination, conventional chemical analysis of as-milled powder is almost always conducted. Tables 2 and 3 reveal the changes in chemical composition of Cr3C2-25 (Ni20Cr) powder milled in hexane [25] and Inconel 625 powder milled in liquid nitrogen, respectively [32].

Fig. 7. Variation of average sizes of agglomerate and particles in Cr3C2 – NiCr powders [33].

280

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 8. Morphology of Inconel 625 powders [32]: (a) as-received powder; (b) powder cryomilled for 8 h; (c) powder cryomilled for 16 h; and (d) powder cryomilled for 20 h.

Fig. 9. Influence of milling media on the morphology of commercially pure Ni powder [27]: (a) milled in methanol for 10 h; and (b) milled in liquid nitrogen for 10 h.

On the basis of the results in Tables 2 and 3, it can be seen that, as milling time increases, the percentage of carbon, nitrogen, oxygen and iron increases. This indicates that contamination occurs during milling, regardless of the milling environment and original chemical composition of as-received powders. The carbon, nitrogen and oxygen are likely to originate by diffusion from the milling environment, whereas the presence of iron is attributed to the stainless steel ball media and chamber components. However, no new peaks are evident in the XRD spectra of the powder milled, suggesting cases the amount of contamination to be less than the minimum limit (approximately 2%), which XRD can detect. Suryanarayana [34] summarized the origin of contamination of milled powders. The magnitude of con-

tamination is related to the milling time, the milling intensity, the milling environment, and difference in hardness of powder and the milling medium. 1–4 wt.% Fe was present in most of the powders milled with the steel milling medium; however, amounts as large as 60 Table 2 Dependence of chemical composition of Cr3C2–NiCr on milling time (wt.%) [25] Milling time (h)

Cr

Ni

C

N

O

0 8 20

70.0 66.2 64.3

19.2 19.7 18.8

9.83 9.71 9.36

0.20 0.38 0.51

0.21 1.11 1.93

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

281

Table 3 Dependence of chemical composition of Inconel 625 on milling time (wt.%) [32] Time (h)

Ni

Cr

Mo

Nb

Fe

Ti

Al

C

N

O

0 4 8 20

66.5 66.0 64.3 59.3

20.6 20.9 20.5 20.7

8.99 8.96 8.81 8.10

3.55 3.53 3.87 3.32

0.045 0.14 1.87 5.81

0.012 0.014 0.015 0.014

0.24 0.25 0.21 0.29

0.006 0.017 0.082 0.089

0.075 0.10 0.27 0.41

0.015 0.062 0.087 1.86

at.% Fe were reported during milling of W– 5Ni alloy for 60 h in a Spex mill [35]. Large amounts of oxygen (up to 44.8 at.%) have been reported to be present in Al –6Ti powders milled for 1300 h in a low energy ball mill [36]. A simple approach to minimizing the contamination from the balls and the container is to use the same material for the container and balls as the powder being milled. An alternative way is to choose a material for the container and balls that is harder/stronger than the powder being milled. A significant influence of milling environment on oxygen content was found. Oxygen content of Ni powder milled in liquid nitrogen was less than that for Ni milled in methanol [27]. It is worth noting that contamination during milling does not necessarily lead to a degradation of the properties of nanocrystalline materials. In related studies, it was reported that the nitride and oxide phases formed during cryomilling played a critical role in retarding growth of the nanocrystalline materials [23].

3.4. Grain size On the basis of an original study on microstructures of ball-milled Ru and AlRu, the formation of a nanocrystalline structure is thought to evolve from the development of dislocation cell structures within shear bands [10]. Plastic deformation leads to: (a) the formation of dislocation cells within shear bands; (b) the dislocation cells transform into low-angle grain boundaries; and (c) finally form nanocrystalline grains surrounded by high-angle grain boundaries via grain rotation [10–12]. In a dislocation cell mechanism, the contribution of fracture and welding processes in the powder particles, as described by Benjamin and coworkers [22,24,37], to the formation of a nanocrystalline structure has never been studied. It is thought that the welded fragments of the original coarse grains should form new grains, although Benjamin and coworkers did not explicitly demonstrate the existence of a relationship between the repeated fracturing– welding process and the refinement of grains. During mechanical milling, the grain size decreases with milling time down to a minimum value that reportedly depends on the magnitude of the microstrains rather than milling energy or ball– powder – ball collision frequency [14,15]. Fig. 10 shows the relationship

between microstrain and grain size for milled Al, Ni, Fe, W, Ru and Ti [14]. An accurate determination of grain size and microstrain is an essential requirement for quantitative studies of nanostructured materials. Transmission electron microscopy and XRD are the two analysis techniques most commonly employed for describing quantitatively the evolution of grain size in nanostructured materials. TEM provides a direct image of grains, thus the grain size distribution and the morphology of grains can be characterized. However, the information deduced from TEM observations is often limited to a small local region, whose size depends on intricate sample preparation techniques. XRD analysis can provide information on the structural characteristics of materials, for example, mean grain size, microstrain, as well as crystal imperfections that may be present in the sample as a whole. In addition, the simplicity associated with sample preparation for XRD analysis renders it an ideal technique for quantitative study of a large number of samples. A number of theoretical studies have facilitated simple approaches, such as the Scherrer equation [38,39], the linear fitting [40,41], the single-line approximation [42], that can be used to obtain structural information through mathematical analysis of the XRD profile.

Fig. 10. Dependence of reciprocal grain size on microstrain [14].

282

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

3.4.1. Sample preparation for TEM analysis For TEM studies, sample preparation is both difficult and time consuming. A number of techniques for TEM sample preparation have been successfully developed [43]. Twin-jet polishing and ion-milling are two of the most useful approaches for sample preparation of conventional bulk materials. However, a TEM powder sample cannot be directly prepared by the twin-jet polishing or ion-milling. Individual powder samples for TEM analysis can be prepared via one of three commonly used methodologies (although novel approaches are continuously being developed). 3.4.1.1. Carbon grid method. The carbon grid method is a simple and widely used approach. In this technique, powders are first immersed in methanol in a small container and agitated using a supersonic instrument for a few minutes. Then, a standard copper grid with a carbon film is used to collect some small fragments of powder particles from the methanol suspension. Finally, after drying at room temperature, the sample is directly observed using TEM. The advantage of this method is its simplicity. However, only edge regions of the small fragment of powder particles can be observed. It is therefore difficult to image a micrograph with high quality.

late grain size and microstrain in powders on the basis of the measured XRD spectra. Among these, the Scherrer equation, the linear fitting and the single-line approximation are the three most frequently used; therefore, these three methods are discussed next.

3.4.2.1. Scherrer equation. If the physical origin of broadening of XRD reflections is associated with the small grain size alone, the relationship between grain size (D) and the full-width at half-maximum (FWHM, referred as D(2q)) of XRD reflections is expressed by [38,39], D=0.9u/D(2q) cos q

(1)

where u is the wavelength, (in the case of Cu target, u= 0.15406 nm), and q is the diffraction angle. The true peak broadening, D(2q), can be obtained using the following equation: D(2q)=[(D(2q)h)2 − (D(2q)g)2]1/2

(2)

where D(2q)h is the FWHM of the measured profile, and D(2q)g is the FWHM of the profile from the standard sample for the same reflection. The grain size, obtained using the above described Scherrer equation, is hence volume-averaged in a direction perpendicular to the diffraction plane.

3.4.1.2. Cold compact method. Powders are compacted using high pressure (i.e. 3.5 GPa [44]) into 3 mm pellets at room temperature, followed by the standard twin-jet polishing method. Using this method, the microstructure in a powder sample can be clearly observed, similar to the case of a bulk sample; however, the parameters for twin-jet polishing need to be modified because of the presence of porosity in the compacted pellets.

3.4.2.2. Linear fitting. Assuming that the overall broadening of XRD reflections is comprised of two components, grain size and microstrain, the identification of the individual contributions from grain size and microstrain must be separated. The peak broadening component resulting from a microstrain effect, D(2q)s, can be expressed as [40,41],

3.4.1.3. Metal nut method. Powders are mixed with epoxy to create a slurry, which is then mounted into a metal nut with an outside diameter of 3 mm to form a 3-mm diameter disk, followed by the normal ionmilling process. A TEM powder sample with a wide region of uniform thickness can be prepared by the metal nut method without difficulty.

where m is the microstrain. The broadening component originating from the small crystallites, D(2q)c can be expressed by the Scherrer equation, that is

3.4.2. Analysis of XRD profile XRD is more conducive to analysis of powder samples than is TEM. It is worth noting, however, that the accuracy of the XRD measurement is crucial to the quantitative analysis for grain size and microstrain, which is determined on the basis of the change in width of XRD profiles. For an individual XRD reflection, the change in value of the width of the spectra is extremely low, usually less than 0.5°. Therefore, it is advisable to use a low scanning rate, i.e. approximately 0.1° min − 1, to guarantee the accuracy of the measurements. A number of formulations have been developed to calcu-

D(2q)s = 2m tan q

D(2q)c = 0.9u/D cos q

(3)

(4)

Therefore, the overall broadening of the XRD reflection is given by D(2q)= D(2q)s + D(2q)c = 2m tan q+0.9u/D cos q

(5)

thus, D(2q)cos q=2m sin q+ 0.9u/D

(6)

If D(2q) cos q is plotted against sin q for a number of XRD reflections at different angles, the data should fall on a straight line, with a slope of 2m and an intercept of 0.9u/D. It is worth noting that there is an inherent isotropic assumption in the linear fitting that implies the presence of identical microstrains for different crystallographic orientations, because the same slope (2m) is applied to different orientations. This is not true for

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319 Table 4 Stable values of grain size for several milled engineering materials Materials

Grain size (nm)

Milling media

References

Cr3C2-25 (Ni20Cr) Inconel 625 49Fe–49Co–2V M50 steel Inconel 718 Ni Ni

15

Hexane

[25]

31 35 10 25 44 19

Liquid nitrogen Methanol Liquid nitrogen Liquid nitrogen Methanol Liquid nitrogen

[32] [46] [47] [48] [27] [27]

283

where subscripts C and G denote the Caucy and Gaussian components of the respective Voigt profiles. From Eq. (7) it follows that the Caucy and Gaussian components, i Cf and i Gf , of the integrated widths for the f profile, are given by i Cf = i hC − i gC

and

(i Gf )2 = (i hG)2 − (i gG)2

(8)

The grain size (D) and microstrain (m) can be calculated by D= 0.9u/i Cf cos q m= i /4 tan q f G

(9) (10)

Thus, the objective of the single-line approximation is to establish the Caucy and Gaussian components, iC and iG, of the h and the g profiles. The constituent iC and iG of the h and the g profiles can be obtained using graphical methods or interpolation from tables [42]. In addition, the following empirical formulae have been derived by de Keijser et al. [39] to calculate iC and iG iC = i{2.0207− 0.4803[D(2q)/i]− 1.7756[D(2q)/i]2} (11) iG = i{0.6420+ 1.4187[D(2q)/i − 2/y}1/2 − 2.2043D(2q)/i + 1.8706[D(2q)/i]2}

(12)

where i is the overall integral widths, and D(2q) is the FWHM of the profile.

Fig. 11. Dependence of grain sizes of milled powder on milling time (grain size of as-received Inconel 625 powder is 2.71 mm) [25,32].

many cases. For instance, in the calculations of grain size relevant to nanostructured Fe using the linear fitting method [44], the data deduced from {100}, {211} and {220} reflections are only used. While those resulting from {110} and {310} reflections are excluded because the elastic modulus on {110} and {310} planes differ significantly from those on the former three crystallographic planes. This leads to different microstrains in Ž110 and Ž310 orientations relative to those in the Ž100, Ž211 and Ž220 orientations, and hence all the five data points, if used, will not fall on an identical straight line.

3.4.2.3. Single-line approximation. An approach for the simultaneous determination of grain size and microstrain using a single XRD reflection, was developed by de Keijser et al. [42]. Assuming that the XRD profile can be matched by a Voigt function [45], the experimentally measured line profile h is actually the convolution of the structurally broadened profile f and the standard profile g, that is, hC =g*f C C and hG =g* G fG

(7)

3.4.3. Value of grain size Fig. 10 shows that reciprocal grain size, in case of pure metals, depends on microstrain [14]. The lattice strain at the atomic level increased with decreasing grain size, and then decreased with further decreases in grain size. It is still an open question as to why the maximum in strain with respect to a change in grain size was present [14]. Typically, the grain size of nanocrystalline materials achieved via attritor milling ranges from 10 to 45 nm, as shown in Table 4. A smaller grain size was achieved in Ni with cryomilling, compared to methanol milling [27]. This is thought to be due to the presence of high oxygen content in methanol, which decreases the interaction between powder particles during milling. Fig. 11 reveals the variation in grain size, as determined by XRD, with milling time for Cr3C2-25 (Ni20Cr) and Inconel 625. With an increase in milling time, grain size decreases and approaches a relatively constant value. A sharp decrease in grain size often occurs within the first hours of milling. TEM observations confirm the achievement of nanoscale grain size. Fig. 12(a) and (b) display a bright field and the corresponding dark field image of Cr3C2-25 (Ni20Cr) powder after 20 h of milling. Carbide particles, with an average particle size of approximately 15 nm, are embedded into the matrix. Fig. 13(a)–(d) illustrate TEM bright field images, the corresponding dark field image, and the SAD pattern

284

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

of Inconel 625 powder cryomilled for 8 h. Fig. 14 indicates a narrow distribution of grains, which are measured from the TEM bright images of Inconel 625 powder cryomilled for 8 h, with an average dimension of 31 nm. The distribution of grain size is relatively uniform. It can be seen that only a few grains are larger than 65 nm in Fig. 13(a), in which approximately 2000 grains are illustrated.

3.5. Structural e6olution during milling To illustrate the evolution of microstructure during milling, two systems are discussed in detail, Cr3C2 – Ni20Cr and Inconel 625. The selection of these two systems was prompted by their distinct physical properties: (a) Cr3C2 –Ni20Cr system is a mixture of a brittle phase (Cr3C2) and a ductile matrix (Ni20Cr); whereas (b) Inconel 625 is an alloy system containing multiple elements. Moreover, the former system was milled in a hexane environment whereas the latter was milled in liquid nitrogen. These are discussed in detail in the sections that follow.

3.5.1. Cr3C2 –Ni20Cr powder The Cr3C2 –Ni20Cr system is commonly used for wear applications, especially in corrosive environments at elevated temperature [6]. Fig. 15(a) shows a TEM bright field image consisting primarily of a NiCr solid solution in the case of Cr3C2 – Ni20Cr powder following milling in hexane for 2 h, and Fig. 15(b) and (c) are the corresponding SAD patterns of powders A and B in Fig. 15(a) [25]. Based on the SAD patterns, powder A is essentially a single crystal of NiCr solid solution, although a few fragments from other powders are also seen. The carbide particles were identified to be embedded in powder B from the incomplete diffraction rings

that overlap the SAD pattern of single crystal NiCr solid solution. With increasing milling time, more carbide particles are embedded in the NiCr solid solution. Fig. 16(a) displays a TEM bright field image of the milled powder for 4 h, and Fig. 16(b) shows the corresponding dark field image, using a carbide diffraction spot, in which a number of carbide particles can be seen. In this figure, clearer carbide diffraction rings, which overlap the SAD pattern of the single crystal NiCr solid solution, are observed. Fig. 17 shows a carbide particle, in the 8-h milled powder, fractures into a number of fragments that are embedded in the NiCr solid solution. It is verified that the carbide and binder metal combine into a polycrystal nanocomposite powder because the SAD pattern is comprised of diffraction rings. The carbide fragments in the polycrystal nanocomposite reveal a sharp morphology. Fig. 12(a) and (b) in Section 3.4.3 reveal a TEM bright field image and corresponding dark field image of a polycrystal nanocomposite powder after 20 h of milling. Large proportions of carbide fragments have transformed into round carbide particles. Evidently, the micrographs and SAD patterns shown in Fig. 15 through 17 exhibit transition microstructures of the powders during milling. The milled powders are found to continually overlap, cold weld, fracture, and gradually transform into a polycrstal nanocomposite, in which round nanostructured carbide particles are uniformly distributed. Fig. 18 shows XRD spectra between the diffraction angle (2q) range of 15–25° (Mo target, u= 0.070923 nm) for the as-received powder and the corresponding powder milled for 16 h. The diffraction peaks from the milled powder broaden noticeably, and their intensities decrease drastically. This indicates a significant change in the structure of the powder as a result of milling.

Fig. 12. Cr3C2 – NiCr powder milled for 20 h [25]: (a) Bright field image; and (b) dark field image. Particles (white) are Cr3C2, and matrix (dark) is NiCr.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

285

Fig. 13. Nanocrystalline Inconel 625 structure with a relatively uniform distribution of grain sizes [32]: (a) bright field image; (b) dark field image taken using {111} + {200} reflection rings; (c) SAD; and (d) magnification of (a).

3.5.2. Inconel 625 Inconel 625, a nickel-based superalloy with a fcc crystal structure, is widely used in the aerospace and nuclear industries because of its high strength, excellent fabricability and corrosion resistance [49,50]. Fig. 19(a) reveals a cross-section SEM image of the as-received Inconel 625 powder. The equiaxed grains possess an average grain size of 2.71 mm. Fig. 19(b) shows a TEM bright field image of the as-received powder. A high dislocation density is observed in the powder. Fig. 20 shows a typical TEM image of the Inconel 625 powder cryomilled for 4 h. Grains are not clearly visible in the powder; however, the presence of significant difference in contrast is observable. Fig. 21(a)– (c) shows a region with the darkest contrast, which usually reveals an elongated morphology with an average dimension of 0.15 (width)×0.7 (length) mm, containing a high den-

sity of deformation faults. The corresponding dark field image indicates the darkest region to be one single elongated grain, in the case of Fig. 21(c), with a size of 0.13× 0.58 mm. The elongated grains containing a high density of deformation faults are similar to microstructures formed by strain-induced martensitic transformation on thinning [43,51,52]. However, there is no evidence to indicate the presence of a martensitic phase in the diffraction pattern and XRD of the powder. The density of faults is too dense, as well as their small dimensions, to identify the nature of the faults. Fig. 21(d) shows a magnification of the region between the elongated grains. The dimension of the Moire´ fringes indicates grain size of approximately 30 nm although the grain boundaries are not visible. Therefore, the powder cryomilled for 4 h exhibits a mixed configuration of elongated coarse grains and nanoscale grains.

286

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 14. Distribution of grain sizes in Inconel 625 powder, cryomilled for 8 h [32].

As the cryomilling time increases, the dimension of the elongated grains decreases. Fig. 22 shows the elongated grains in the powder cryomilled for 6 h. The average size of the elongated grains decreases to approximately 30 nm (width)×120 nm (length). In addition, fractured fragments, with equiaxed morphology, were observed in the powder following cryomilling for 6 h. Fig. 23 shows the elongated grains that fracture into segments. As the cryomilling time is increased further, the elongated grains disappear. Fig. 13, shown in Section 3.4.3, shows the equiaxed grain structure in the powder cryomilled for 8 h. During further increase in milling time up to 20 h, the average grain size remains a constant value of around 30 nm, and a near invariant morphology and size distribution of grains are observed. XRD spectra of Inconel powders for different milling

times are shown in Fig. 24. The as-received Inconel 625 powder is a single-phase Ni-base solution with an fcc crystal structure. To illustrate the change in the sharpness/width of XRD reflections visually, Ka2 peaks are included in the spectra shown in Fig. 24, although the effects of Ka2 are corrected in the calculation of grain size. With increasing milling time, XRD peaks broaden and decrease drastically. Peaks originated from Ka2 are clearly visible in the spectrum of the as-received powder, but not in the cryomilled powders because of the broadening of the reflections caused by cryomilling. It is evident from the XRD spectra shown in Fig. 24 that the cryomilling for 4 h leads to a sharp decrease in the intensity and a sharp increase in the width of the reflections. As cryomilling time increases further, the intensity and the width of XRD reflections approach a relatively constant value. This indicates that the primary structural changes are likely to have occurred within the first 4 h of milling time.

3.6. Milling mechanisms On the basis of observations on structural evolution of powders during milling, the development of nanocrystalline structure in a composite powder, consisting of a hard particle constituent and a tough binder constituent, differs from an alloyed powder containing a unitary constituent. Three mechanisms are discussed to explain the formation of nanocrystalline structure during milling.

3.6.1. Embedding On the basis of SEM, TEM and XRD results, the development of a nanocrystalline structure in a com-

Fig. 15. TEM image of NiCr solid solution following 2 h of milling [25]: (a) bright field micrograph; (b) SAD pattern of powder A; and (c) SAD pattern of powder B.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

287

Fig. 16. TEM image of Cr3C2 –NiCr powder following 4 h of milling [25]: (a) bright field; and (b) dark field image.

Fig. 17. Cr3C2 carbide fragments in the NiCr matrix of Cr3C2 – NiCr powder milled for 8 h [25]: (a) bright field image; and (b) SAD pattern.

posite powder is schematically summarized in Fig. 25 [6]. In a composite powder, such as Cr3C2 – NiCr and WC – Co, there are a hard and brittle carbide particle constituent and a tough metal binder constituent. Hard and brittle carbide particles are fractured into sharp fragments and embedded into the metal binder. The metal binder, with lower hardness, is subjected to enhanced milling from both the balls and hard carbide particles. As milling time increases, carbide fragments are continually embedded into the metal binder. The metal binder and the polycrystal composite experience continuous overlapping, cold welding, and fracturing. With time, the sharp carbide fragments in the polycrystal composite are shaped into round particles. Finally, a polycrystal nanocomposite powder system, in which round nanoscale carbide particles are uniformly distributed in a metal binder, is formed. As an example, Fig. 12 shows such a Cr3C2-25 (Ni20Cr) polycrystal nanocomposite powder. Clearly shown are large pro-

portions of carbides, in the form of round particles, uniformly distributing themselves in the NiCr solid solution. In addition to the Cr3C2 –NiCr and WC –Co systems, it is possible to use mechanical milling to synthesize other nanocomposite powder systems with a hard particle and tough binder duplex structure; examples of such systems are WC–NiCr, TiC–NiCr, TiC– Ti, and SiC–Al.

3.6.2. E6olution of dislocation cells On the basis of a study on microstructures of ballmilled Ru and AlRu, the formation of a nanocrystalline structure is thought to evolve from the development of dislocation cell structures within shear bands [10]. Plastic deformation leads to the formation of dislocation cells within shear bands, then dislocation cells transform into low-angle grain boundaries, and finally form nanocrystalline grains surrounded by high-angle grain boundaries via grain rotation [10–12]. During mechani-

288

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

cal milling, powder particles are repeatedly stressed by impact loading in a direction that changes randomly with increasing milling time. Plastic deformation significantly occurs in powder particles as evident by the powder morphologies shown in Figs. 8 and 9. Therefore, the random cyclic loading and resultant strain, with an amplitude in the plastic deformation regime, is thought to promote random strain fatigue [53] (or a random low-cycle fatigue) of the powder particles. In related studies, Plumtree and Pawlus [54] carefully examined the changes in microstructures in Al during strain fatigue. In their work the development of dislocation cells is stated as the primary characteristic of the deformed microstructure during strain fatigue. Dislocation cells form at the onset of steady-state (corresponding to 30 cycles) with thick and uncondensed walls. There is a high density of dislocations in the interior of the cells. As the number of cycles increases, the cells become more distinct and dislocations in the interior of the cells drastically decrease. At failure (11 300 cycles),

a well-defined cell structure is formed, the cell walls are narrow, and the interior of the cells contains very few dislocations. However, quantitative analysis of the dislocation cells, shown in Fig. 26, indicates that the cell size and misorientation between neighboring cells remain unchanged from the onset of steady-state to failure. The increase in strain amplitude from 1.0 to 2.0% does not influence the misorientation value. These results imply that the development of dislocation cells during strain fatigue does not promote grain size refinement. It is possible that there is a significant difference in strain amplitude and number of cycles between conventional strain fatigue and mechanical milling; however, Koch [14,15] stated that the total strain, rather than milling energy or ball-powder-ball collision frequency, was responsible for determining the nanocrystalline grain size. Presumably, the strain amplitude and number of cycles are not crucial factors for the achievement of nanocrystalline structure under typical mechanical milling conditions. The evolution of dislocation

Fig. 18. XRD spectra for (a) Cr3C2-25 (Ni20Cr) as-received and (b) milled for 16 h [25].

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

289

fracture and welding processes in the powder particles, as described by Benjamin and co-workers [22,24,37], to the formation of a nanocrystalline structure has not yet been studied. It is anticipated that welded fragments of the initial coarse grains should form new grains, although Benjamin and co-workers do not explicitly demonstrate the relationship between the repeated fracturing– welding process and the refinement of grains. It is well known that severe plastic deformation can lead to the formation of elongated grains along the tensile stress direction of a material. Under cryomilling conditions, the powder particles of the Inconel 625 first transformed into disks from an initially spherical geometry [32]. The radius of the ball was significantly larger than that of the powder particle (6.35 mm to 84 mm); therefore, the load applied to powder particles can be considered as simple impact loads normal to the large faces of the disk powder particle, as shown in Fig. 27. A compressive stress acts normal to the large faces of the disk, whereas a tensile stress is parallel to the directions of the disk radius. Hence, the grains would be elongated along the direction of the disk radius. Shingu et al. [55,56] simulated the mechanical milling process and the results are shown in Fig. 28. Mechanical milling brings about a random mixing of initially separated constituents of a sample to form a lamellar structure that is in agreement with the idea proposed by Gilman and Benjamin [37]. However, in the model of Benjamin, a layer of lamellar structure resulted from a particle. A fine lamellar structure (elongated grains) with a nanoscale dimension was commonly observed during intermediate stage cryomilling of Inconel 625, see Fig. 22 and Ref. [32]. This indicates that the above

Fig. 19. Microstructure of the as-received Inconel 625 powder [32]: (a) SEM micrograph indicating grains; and (b) dislocation configuration.

cells in deformation bands is not likely to explain the formation of elongated grains, shown in Figs. 21 and 22. These discussions indicate that further work is needed to provide an insight into the dislocation cell mechanisms, as related to the formation of a nanocrystalline structure during milling.

3.6.3. Fracture and cold welding In reference to the mechanism responsible for the evolution of dislocation cells, the contribution of the

Fig. 20. A typical TEM image of Inconel 625 powder cryomilled for 4 h [32].

290

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 21. A mixed microstructure consisting of the elongated coarse grains and nanoscale grains in Inconel 625 powder cryomilled for 4 h [32]: (a) bright field image; (b) dark field image; (c) detailed view of an elongated coarse grain; and (d) magnification of the region containing nanoscale grains.

random process occurred not only for the interaction between the particles but also that between the grains. From the standpoint of strain fatigue, the relationship between plastic strain and fatigue life is expressed by the Manson–Coffin equation [53], Dmp/2=m%f(2Nf)c

(13)

where Dmp/2 is the plastic strain amplitude, m%f is the fatigue ductility coefficient, and c is the fatigue ductility exponent. The fatigue crack growth rate is a function of the plastic strain range Dmp, that is, da/dN =f(Dmp)

(14)

Eqs. (13) and (14) indicate the fact that strain fatigue leads to fracture. As long as the fracturing occurs transgranularly (strain fatigue at room temperature or lower temperature always led to transgranular fracture [57]), the fracturing and cold welding are likely to refine grains. This observation that suggests the refinement model of fracturing and cold welding is consistent with the co-existence of elongated coarse grains and nanoscale grains during the intermediate stage of cryomilling Inconel 625 powder. The larger fragments appear as elongated coarse grains and the finer ones are present as nanoscale grains.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

In summary, the evolution of the elongated grains from development to disappearance suggests that repeatedly strain fatigue fracture, caused by cyclic impact type loads in random directions, and cold welding are responsible for the formation of a nanocrystalline structure in an alloyed powder with uniform constituents.

3.7. Nanoscale mechanical twins Although dislocations were observed in milled powder, such as in milled Ru and AlRu [10], dislocation activity was not evident in the cryomilled powders. However, mechanical twins were present in Inconel 625 powders cryomilled for different cryomilling times [32]. Fig. 29 shows a TEM bright field image, the corresponding dark field image, and the SAD pattern of mechanical twins in Inconel 625 powder cryomilled for 4 h. The streaks along the [1( 11] orientation of the matrix and the extra reflections are observed on the SAD pattern. The thickness of the twins is much smaller than their length and width. Consequently, the two-dimensional characteristic of these twins leads to two-dimensional diffraction effects to form the streaks. The projection of (1( 11) plane on (110) plane is a line along [11( 2] orientation. Thus the observed twins lie along [11( 2] orientation of the matrix. The above analysis indicates that these twins form on (1( 11) plane of the matrix. Twins are generally classified into annealing twins that grow during heating, and mechanical twins formed

Fig. 22. Elongated grains, with smaller dimension, in Inconel 625 powder following cryomilling of 6 h [32].

291

during plastic deformation. No twins were observed in the as-received Inconel 625 powder, see Fig. 19. The milling was conducted in liquid nitrogen, the temperature was maintained at a relatively constant value of 100 K. Severe plastic deformation occurred in the cryomilled powders. A related study on an fcc Ni-base superalloy deformed at room temperature revealed the occurrence of mechanical twinning in the {111}Ž110 system [51]. Accordingly, the observed twins are considered as mechanical twins. As cryomilling time increased to 8 h, a uniform nanoscale microstructure formed, see Fig. 13 [32]. A high mechanical twin density was observed in such a uniform nanoscale microstructure. Fig. 30 shows detailed views of the nano-twins in Inconel powder cryomilled for 8 h. Usually, plastic deformation at low temperatures is conducive to the occurrence of mechanical twinning [58]. However, a normal tensile deformation at room temperature can result in the formation of mechanical twins in Inconel 718, a Ni-base superalloy having an fcc crystal structure [51]. Furthermore, low temperature, individual milling environment and chemical composition are not necessary conditions for nano-twinning during mechanical milling. For comparison, nano-twins were observed in other mechanically milled powders. Nano-twins have been reported in Inconel 625 powder milled at room temperature (in methanol) for 8 h, Ni20Cr powder at room temperature (in hexane) for 8 h, and pure Al powder at 100 K (in liquid nitrogen) for 16 h [32]. The contrast that originated from nano-twins is strongly dependent on the orientation of the sample. It is therefore difficult to clearly image a number of individual nano-twins in a micrograph at high magnification. Fig. 31 shows the dependence of the contrast of nano-twins on the orientation of a sample in powders cryomilled for 20 h. Arrows indicate the changes in the image of nano-twins with tilting operation. A slight tilting operation may hide the image of nano-twins. Therefore, to observe the nano-twins, particular attention is needed while operating the TEM because of their small dimension and sensitivity to orientation of the sample. The fact that the contrast of nano-twins is dependent on orientation of the sample brings about difficulty in carrying out quantitative analysis of nanotwins using TEM. Twin faults in an fcc structure lead to an asymmetry in the corresponding XRD profile [59], and the methodology of XRD analysis, used to quantitatively characterize twin faults in fcc materials, has been established [59–62]. On the basis of a comprehensive mathematical formulation, the probability i, of finding a twin between neighboring (111) plane of fcc structure, is expressed as [60]:

292

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 23. Elongated grains fracture into a few segments during cryomilling [32]: (a) bright field image; (b) detailed view of (a); and (c) dark field image of (b), arrows indicate a fractured grain.

i=

DCG(2q)111 − DCG(2q)200 11 tan q111 +14.6 tan q200

(15)

where q is reflection angle, in degrees, and DCG is the displacement of center of gravity of a reflection peak from the peak maximum. On the basis of Eq. (15), the dependence of probability of twin formation on cryomilling time in Inconel 625 is calculated and the results are plotted in Fig. 32. The value of i increases to 0.025 with an increase in cryomilling time. The spacing of {111} crystallographic plane, d111, in fcc Inconel 625 alloy is equal to 0.207 nm, and grain size in the powders cryomilled for over 4 h is equal to approximately 30 nm. Therefore, i =0.025 means the presence of 3.6 twin boundaries in each grain. Thus, a relatively high density of nano-twins are present in cryomilled fcc powders. In related studies [63– 68], the microstructure and mechanical response of nanocrystalline Cu and Pd, synthesized by inert gas condensation followed by hot

pressing at 1.4 GPa pressures, were investigated. High resolution TEM and XRD studies, as well as conventional TEM observations, indicate the presence of a large number of highly twinned grains (i=0.01 0.05), whose origin is possibly related to the inert gas condensation process [68] or formed by shear stresses during sintering of the nanocrystalline powder [66]. The twinned structure was thought to be responsible for the higher creep resistance in the nanocrystalline Cu and Pd, because twin boundaries are poor paths for vacancy diffusion and resist grain boundary sliding [67].

4. Agglomeration Spherical powder particles, with dimensions ranging from 10 to 50 mm, are typically required for most available thermal spray systems. As discussed in Section 3.1, as-synthesized nanocrystalline powders do not sat-

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

293

isfy these requirements for the size and morphology of the particle. Accordingly, an agglomeration procedure is often necessary, even for conventional powders that do not meet these requirements. Spray drying is a popular agglomeration technique [72– 74]. In this technique, a slurry of liquids and solids or a solution is atomized into droplets in a chamber through which heated gases, usually air, are passed. The liquids are evaporated from the droplets and the solids are collected continuously from the chamber. Spray drying is also widely used in the pharmaceutical, chemical and food processing industries. The size distribution of particles is a function of atomization conditions and properties of the slurry.

Fig. 26. Size of dislocation cells and misorientation between neighboring cells during strain fatigue of Al [54]: (a) variation of cell size with cycles for DmT =1.0%; and (b) misorientation between cells for DmT =1.0, and 2.0%.

Fig. 24. X-ray diffraction spectra of Inconel 625 powders following different cryomilling times [25].

Fig. 27. Schematic diagram illustrating stresses applied to a powder particle during cryomilling [32].

4.1. Binders

Fig. 25. Schematic diagram of milling mechanism for duplex structure powder: (a) initial stage; (b) NiCr matrix overlaps and deforms, Cr3C2 fractures and embed into NiCr; (c) binders deform, fracture, and weld, carbide fracture further; and (d) nanocomposite powder.

Suitable binder materials must be homogeneously dispersible in the liquid used to form a slurry. When dried, binders can form a coating and/or adhere to the materials being agglomerated. Also strength is required for binders to resist crushing of the agglomerates caused by the subsequent handling of powder. Three groups of binders are listed in Table 5.

294

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

4.2. Agglomeration of cermet carbides

Fig. 28. Simulated image of mechanical alloying/milling. Lamellar structure is formed during mechanical milling [55].

Nanostructured WC–Co coatings are of interest in applications that require abrasion and erosion resistance. A typical agglomeration procedure for WC– Co powder is described as follows [72]. Closed-cycle spray drying is required for most cermet carbide powders because the binders that are used are soluble only in volatile organic fluids. The nitrogen drying gas that is used in the spray drying of cermet carbides is heated from 75 to 100 °C. The content of solid in the slurry varies from 75 to 80 wt.%. Pressures for nozzle atomization range from 0.59 to 1.47 MPa. A SEM micrograph indicating spray-dried, spherical-shaped, near nanostructured WC–18% Co powder is shown in Fig. 33.

Fig. 29. TEM images of mechanical twins in Inconel 625 powder cryomilled for 4 h, B =[110] of the matrix [32]: (a) bright field image; (b) dark field image; and (c) SAD and index.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

295

Fig. 30. Detailed views of nano-twins in Inconel 625 powder cryomilled for 8 h [32]: (a) nano-twins; (b) microstructure containing nano-twins A and B; (c) detailed view of nano-twin A in (b); and (d) detailed view of nano-twin B in (b).

5. Thermal spraying of nanostructured coatings Today, a number of thermal spraying techniques are available. Flame spraying (FS), arc spraying (AS), detonation gun spraying (DGS), continuous detonation spraying (CDS), atmospheric plasma spraying (APS), twin wire arc spraying (TWAS), low pressure plasma spraying (LPPS) or vacuum plasma spraying (VPS), controlled atmosphere plasma spraying (CAPS), high velocity flame spraying (HVFS) and high velocity oxygen fuel spraying (HVOF) are widely used to produce coatings for different industrial applications. The process and parameters of the spraying techniques are

described in Refs. [1,4]. Basically, thermal spray facilities are used for synthesizing conventional coatings, and can be used to spray the corresponding nanostructured coatings.

5.1. High 6elocity oxygen fuel HVOF spraying is the most significant development in thermal spraying industry since the development of plasma spraying [1] and has been the topic of many excellent investigations in the recent decade [75–77]. HVOF is characterized by high particle velocity and low thermal energy when compared to plasma spraying.

296

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

The applications of HVOF have expanded from the initial use of tungsten carbide coatings to include coatings that provide improved wear or erosion/corrosion resistance [78]. Fig. 34(a) illustrates a typical Sulzer Metco Diamond Jet HVOF thermal spray facility that is available at University of California Irvine [6]. The main constituents of the facility are a DJC control unit, a powder hopper feeder, a ‘Parker’ X– Y automated traverse unit, ‘In-Flight’ diagnostic equipment, and the Diamond Jet spray gun. The DJC control unit monitors and controls the gas flow into the gun and allows the proper stoichiometric ratios to be set for optimum spray performance. The powder feeder is a fluidized bed powder feed unit that allows proper control of powder flow into the gun. To obtain repeatable results,

a ‘Parker’ automated X–Y system was installed to aid in producing a predictable, uniform coating thickness because the substrate is uniformly scanned by the traverse controlled gun. With the ‘In-Flight’ diagnostic equipment, accurate average particle temperatures can be measured, as well as trajectory and particle flow characteristics. The ‘In-Flight’ particle pyrometer detects changes in temperature based on emissivity of light expelled from the particles as it traverses from the barrel of the gun to the substrate. The pyrometer interfaces with the torch diagnostic system directly. A light emitting diode is involved to guarantee alignment of the pyrometer to the center of the flame. Infrared rays radiated from the particles, not flame, are transmitted by a fused fiber coupler to two infrared detectors

Fig. 31. Dependence of the contrast of nano-twins on the orientation of the sample [32]: (a) a typical TEM image; (b) tilting 1° relative to (a); (c) tilting 2° relative to (a); and (d) tilting-1° relative to (a).

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

297

Fig. 32. Dependence of probability of twin (i) in Inconel 625 powder on cryomilling time [32].

of different spectral sensitivity. A value for the particle temperature is obtained by calculating the ratio of the output voltage from these two detectors. The Diamond Jet spray gun is a hybrid water-cooled gun that allows easy transitions between two fuel gases, hydrogen and propylene. As shown in Fig. 34(b), the Diamond Jet brings in oxygen, air, and fuel, from the DJC into the rear of the gun in the proper stoichiometric ratios. This gaseous mixture is ignited by an arc current creating a supersonic, low temperature flame with gas velocities of 1830 m s − 1 and temperatures around 2700 K. From the hopper powder feed unit, nitrogen carrier gas brings the feedstock powder into the rear of the gun and then

Fig. 34. HVOF thermal spray facility and process [6]: (a) Sulzer Metco Diamond Jet HVOF thermal spray facility; and (b) Schmatic diagram of HVOF process.

Table 5 Typical binders used for spray drying [72] Organic binders Inorganic binders Plasticizers

Polyvinyl alcohol, natural gums, carboxyl-methyl cellulose salts, polyvinyl acetate, methyl cellulose, ethyl cellulose, polyvinyl butyral dispersions, protein colloids, acrylic resin emulsions, ethylene oxide polymers, water-soluble phenolics, lignin sulfonates, propylene glycol alginates, flour, starches Sodium silicate, boric acid, borax, carbonates, nitrates, oxylates, oxychlorides Glycerine, ethylene glycol, triethylene glycol, dibutyl phthalate, diglycerol, ethanolamines, propylene glycol, glycerol monochlohydrin, polyoxyethylene aryl ether

Fig. 33. Agglomerated near nanostructured WC –18% Co powder [116]: (a) spherical agglomerated powder; and (b) detailed view of (a).

298

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

axially into the flame. The powder particles are injected axially into the jet gas, heated, and propelled toward the substrate. The velocity and temperature of the HVOF sprayed particle powders can be controlled independently [79]. With the relatively low temperatures of the flame gas associated with the HVOF systems, superheating or vaporization of individual particles are often prevented [80]. Furthermore, the lower particle temperatures lead to carbide coatings that exhibit less carbide loss than that of the plasma-sprayed coatings. In essence, the advantages of the HVOF process over conventional plasma spraying are higher coating bond strength, lower oxide content, and improved wear resistance due to a homogeneous distribution of carbides [4,81]. Therefore, the coatings discussed in this review paper have been synthesized using HVOF thermal spray technique.

5.2. Microstructure of as-sprayed coatings During thermal spraying, individual droplets are estimated to cool at rates of around 107 K s − 1 [82]. However, in order to build up a thick coating the spray gun traverses over previously deposited material. This leads to localized reheating both from the direct thermal energy of the gas jet and also from latent heat evolution as successive layers of molten splats solidify. This is a non-equilibrium process that results in a refined grain size, even in coatings sprayed with conventional feedstock powders [83,84]. Fig. 35 shows TEM micrographs of coatings sprayed with conventional and nanocrystalline Ni powders [27]. A decreased grain size is observed in the coating sprayed with nanocrystalline Ni powder, as compared to that sprayed with the conventional Ni powder.

The microstructures of conventional and nanostructured Cr3C2-25 (Ni20Cr) coatings, examined using SEM, are shown in Fig. 36 [6]. A uniform and dense microstructure is observed in the nanostructured coatings, compared to the conventional Cr3C2-25 (Ni20Cr) coating that is observed to have an inhomogeneous microstructure. Five carbide (dark phase) and matrix areas (bright phase) were randomly chosen for SEM EDS analysis of the chemical composition. The results of this analysis are listed in Table 6 [6]. The distribution of Cr and Ni in the carbide particle and matrix phase are obtained from the average value of five readings. The row labeled ‘average’ in Table 6 is obtained from a low magnification analysis of a region containing carbide and matrix phases. The results shown correspond to five readings. In the matrix phase of the conventional coating, the contents of Cr and Ni are close to the nominal chemical composition of a pure NiCr solid solution (80% Ni and 20% Cr), thus the matrix phase was considered to be the NiCr solid solution phase [6]. A minor amount of Ni exists (0.85%) in the carbide phase. For the nanostructured coating, a higher Cr content in the matrix phase and a higher Ni content in the carbide phase are observed. This is thought to result from the high degree of mixing and small scale associated with the nanostructured constituents, in combination with the beam size limitations of the EDS. To compare the average chemical composition (labeled ‘average’ in Table 6) of the coatings with those of the feedstock powders listed in Table 6, a simple correction for data in Table 6 is made because the light elements, such as C, N, and O, are not included in the EDS analysis [6]. In Table 6, the sum of Cr and Ni content is 100%. Actually, the sum of Cr and Ni

Fig. 35. TEM images of Ni coatings [27]: (a) conventional Ni coating; and (b) nanostructured Ni coating.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

299

Fig. 36. Microstructure of Cr3C2-25 (Ni20Cr): (a) conventional coating; (b) magnification of (a); (c) nanostructure coating; and (d) magnification of (c) [6].

content is 89.2% in the conventional feedstock powder, and 83.1% in the nanocomposite feedstock powder, see Table 2. Therefore, the content of Cr and Ni is multiplied by factors of 0.892 in the conventional coating, and 0.831 in the nanostructured coating. This correction represents a simple approximation because it implies there are no changes in the contents of all elements during thermal spraying, whereas in practice, this may not be the case. After making this simple correction, the average contents of Cr and Ni are 41.3 and 47.9% in the conventional coating, and 63.4 and 19.8% in the nanostructured coating, respectively. It is worth noting that there is a large average chemical composition difference between the conventional and nanostructured coatings. The average contents of Cr and Ni in the nanostructured coatings are close to those of feedstock nanocomposite powder. However, those same values found for the conventional coating differ markedly from the conventional feedstock powder. Because the contents of Ni and Cr in both the matrix and the carbide phase are close to their nominal contents, the chemical composition difference between the conventional coating and feedstock powder is attributed to a decrease in the volume fraction of the carbide phase in the conventional coating. In related studies, the decarburization of tungsten carbide has been widely

reported [85,86], whereas chromium carbide is quite stable during thermal spraying [87]. Therefore, the measured decrease in the volume fraction of the carbide phase in the conventional coating may not be attributed to decarburization or oxidation of chromium carbide. Individual, large-sized regularly shaped carbide particles and spherically shaped matrix phase particles are present in blended Cr3C2 –NiCr powder. During spraying, the large-sized carbide particles, with a high melting point of 2200 K, may remain solid or semi-molten in the HVOF system (short dwell time and low temperature flame), yielding low adherence with the substrate surface. Conversely, melt matrix phase droplets (melting point is 1690 K) have a more fluid characteristic than the carbide particles. The greater fluidity can result in effective contact of the matrix phase with the Table 6 Chemical composition of Cr3C2-25 (Ni20Cr) coatings (wt.%) [6]

Average Matrix Carbide

Conventional coating

Nanostructured coating

Cr

Ni

Cr

Ni

46.3 16.21 99.15

53.7 83.79 0.85

76.25 31.12 94.34

23.75 68.88 5.66

300

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 37. TEM observation of nanostructured Cr3C2-25 (Ni20Cr) coating: (a) bright field; (b) dark field; and (c) diffraction pattern [6].

substrate surface. Thus, the volume fraction of the carbide phase decreases in the conventional coatings as a fraction of the carbide particles fail to adhere, and simply bounce off from the substrate. In the case of the nanocomposite feedstock powder, there is high probability that the nanoscale carbide particles will be completely surrounded by the matrix phase, as a result of their extremely small size (15 nm). The nanoscale particles are therefore co-coated with the matrix phase and effectively adhere onto the substrate surface. Thus the improved fluidity leads to the nanostructured coating having a close composition to the feedstock powder. Commercially available blended Cr3C2 – NiCr powders are mixtures of Cr3C2 and NiCr solid solution, and these mixtures are inherently difficult to handle because of segregation during storage, transportation and spraying. Consequently, these mixtures usually produce an inhomogeneous microstructure characteristic [88,89]. A few pre-treatment methods have been developed to overcome these types of microstructural variations since coating performance is very susceptible to the non-uniformity of microstructure. Two such methods, referred to as ‘pre-alloying’ [89] and ‘cladding’ [88], are widely used. Using the ‘pre-alloying’ method, the powders are first agglomerated using an organic polymeric binder and then heated and pre-sintered in hydrogen. The powders are then densified using a plasma flame in an inert atmosphere, and are finally milled, screened and classified to yield the desired particle size. In a related study [89], the hardness of a plasma-sprayed Cr3C2 – NiCr coating, using the ‘pre-alloyed’ powder, increased from 594 to 796 HV300. In the ‘cladding’ method, each Cr3C2 carbide particle is clad with an essentially continuous layer of NiCr solid solution; therefore, the Cr3C2 carbide is present as a discrete second phase particle randomly embedded in a NiCr solid solution. The hardness of a plasma-sprayed Cr3C2 – NiCr coating increased from 620 to 860 HV300, and the wear resistance was also improved as compared to the standard blend coating method, presumably as a result of the uniform ‘clad’ powder [88]. Using a Metco Diamond Jet system, Sasaki et al. [90] compared the behavior of HVOF thermally sprayed coatings made using four different

types of Cr3C2 –NiCr feedstock powders (blend, agglomerated/sintered, sintered/crushed and sintered/ crushed/clad). They found that the coating made of the sintered/crushed/clad powder showed the best characteristics as compared with those coatings sprayed by the three other types of powders. Therefore, regardless of the spraying method employed, the uniformity of microstructure in a coating has a significant positive influence on its performance. The Cr3C2-25 (Ni20Cr) nanocomposite powder used by He et al. [6] was synthesized using mechanical milling and agglomeration following the milling process [25]. In this approach, the carbides in the nanocomposite powder are uniformly distributed in the NiCr solid solution. In essence, mechanical milling yields a ‘clad’ powder and hence a uniform microstructure is obtained in the nanostructured coating. The TEM bright field image of the nanostructured Cr3C2-25 (Ni20Cr) coating, the corresponding dark field image and diffraction pattern are shown in Fig. 37(a)–(c), respectively [6]. The average carbide particle size is approximately 24 nm. This indicates that the coating has a nanoscale microstructure. In the nanostructured WC– 12% Co coating [18], TEM examination revealed a microstructure consisting of nano sized WC carbide particles in an amorphous matrix phase. While in the nanostructured Cr3C2-25 (Ni20Cr) coatings, the diffraction pattern does not clearly exhibit the presence of an amorphous matrix phase [6]. Instead of an amorphous matrix phase, a few discontinuous elongated amorphous phases were observed in the nanostructured Cr3C2-25 (Ni20Cr) coating, shown in Fig. 38(a) and (b). The inserted diffraction patterns, which were from the elongated phases marked A, show diffuse rings. Many fine diffraction spots are sharply imaged in the diffraction patterns, indicating the diffraction patterns to be well focused. The diffuse rings are thus indicative of an amorphous phase rather than a false appearance caused by under/over focusing. The elongated amorphous phases, which have dimensions of around 100 nm wide and 1 mm long, are discontinuously distributed in the coating. Guilemany and Calero [87] also observed amorphous matrix phases in a conventional HVOF thermally sprayed Cr3C2 –NiCr coating.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

5.3. Mechanical properties of as-sprayed coatings Microhardness is a very important performance characteristic for coatings. The average microhardness of the nanostructured Cr3C2-25 (Ni20Cr) coating, taken on the cross-section, increases from a value of 846 for the conventional Cr3C2-25 (Ni20Cr) coating to 1020 HV300 for the nanostructured coating [6]. Hence the nanostructured coating exhibits a 20.5% increase in microhardness as compared with the corresponding conventional coating. Several published hardness values of Cr3C2-25 (Ni20Cr) coatings are listed in Table 7. The spraying methodology and the type of feedstock powder used (blend, agglomeration or clad) have a significant influence on the hardness of coatings on the basis of the data in Table 7. Effects of spraying method [91,92] and type of feedstock powder [86,88– 90] on hardness have been extensively investigated. HVOF is

301

characterized by high particle velocity and low thermal energy. This combination of high kinetic energy and low thermal energy leads to a high hardness [85]. The clad feedstock powders produce uniform Cr3C2 –NiCr coatings and hence high hardness [88–90]. It has been reported that the hardness of nanostructured materials often exhibits a two- to five-fold increase in hardness compared with that of the conventional materials although it is lower than that predicted using the classical Hall–Petch equation [69–71]. In a related study, Kear and McCandlish [21] also indicated the nanostructured WC –23% Co coatings with a higher hardness than a conventional coating of the same composition. The high hardness of the nanostructured Cr3C2 –NiCr coatings results primarily from two aspects: (1) uniformity of microstructure, caused by the process of synthesis of nanocomposite feedstock powder; and (2) the intrinsically high hardness of nanostructured phases.

Fig. 38. Elongated amorphous phase in nanostructured Cr3C2-25 (Ni20Cr) coating: (a) bright field; and (b) dark field [6]. Table 7 Published hardness data of Cr3C2-25 (Ni20Cr) coatings Spraying method

Powder

Hardness

Source

High velocity oxygen fuel High velocity oxygen fuel High velocity oxygen fuel High velocity oxygen fuel High velocity oxygen fuel Detonation gun spraying Atmospheric plasma spraying, Ar/H2 Atmospheric plasma spraying, Ar/He Continuous detonation spraying Atmospheric plasma spraying, Ar/H2 Atmospheric plasma spraying, Ar/He Continuous detonation spraying Detonation gun spraying High velocity oxygen fuel High velocity oxygen fuel

Commercial Commercial (sintered/crushed/clad) Commercial (blend) Commercial (agglomeration) Commercial (blend) Commercial (powder size: 10–44 mm) Commercial Commercial Commercial Commercial Commercial Commercial Commercial Commercial (blend) Nanostructured

855 (HV300) 950 (HV300) 700 (HV300) 914 (HV300) 697(HV300) 800 (HV300) 830 (HV300) 871 (HV300) 828 (HV300) 837 (HV300) 943 (HV300) 889 (HV300) 945 (HV300) 846 (HV300) 1020 (HV300)

[93] [90] [90] [86] [86] [94] [91] [91] [91] [92] [92] [87] [92] [6] [6]

302

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

decreased friction coefficient was also observed in a nanostructured WC–23% Co coating [21]. The abrasion resistance of thermal sprayed coatings is related to the relative fracture toughness [96]. The indentation fracture method is often employed to characterize the relative fracture toughness of coatings [96]. An indentation fracture examination was performed on the conventional and nanostructured Cr3C2-25 (Ni20Cr) coatings and the results are shown in Fig. 40(a)–(d) [6]. Under the same load, indentation marks in the nanostructured coating are smaller than those in the conventional coating because the nanostructured coating has a higher hardness. Under a load of 1000 g, many cracks caused by an indentation along the phase interface of carbide phase with metal matrix phase were observed in the conventional coating, and some cracks were also found in the nanostructured coating. When the load was decreased to 500 g, a few cracks around the indentation were still observed in the conventional coating, whereas none were present in the nanostructured coating [6]. These results suggest that the nanostructured Cr3C2 – NiCr coating possess a higher apparent fracture toughness relative to that of the conventional material. Fig. 39. Scratch-resistance of Cr3C2-25 (Ni20Cr) coatings [95].

5.4. Stability of nanostructured powder during thermal exposure

In related studies, increased microhardness values have been reported for nanocrystalline Ni [27] and Inconel 718 coatings [48]. In related studies on conventional and nanostructured Cr3C2-25 (Ni20Cr) coatings, the friction coefficient and scratch-resistance of coatings were measured using a CETR Micro-Tribometer [95]. The scratch head was a sapphire ball with a radius of 0.75 mm, and the scratch tests were performed using a scratch rate of 22.7 rpm (535 mm min − 1) and a scratch normal load of 5 N. Fig. 39 compares scratch depths in the as-sprayed conventional coating and in the as-sprayed nanostructured coating. A reading of scratch depth was taken every 0.01 s during the tests that were automatically controlled by a computer. Thus a few tens of thousands of data illustrating the relation between scratch depth and time were recorded during a single scratch run. The as-sprayed conventional coating produced an average scratch depth of approximately 100 mm, whereas a depth of around 50 mm was found in the as-sprayed nanostructured coating, see Fig. 39. Thus, the nanostructured coating exhibits a scratch resistance that is twice that of the conventional one. Coefficient of friction, under the mode of the ‘ball-ondisk’ friction, can also be obtained from the scratch tests, and are 0.495 and 0.216 for the conventional and nanostructured coatings, respectively [95]. Compared with the coefficient of friction in the as-sprayed conventional coating, a reduced coefficient of friction was observed in the as-sprayed nanostructured coating. A

The microstructural stability of many engineering materials, often involves, not only grain growth, but also phase transformations. For example, in the case of WC –Co coatings, which are widely used because of their high wear resistance, the formation of hard and brittle W2C phase originating from the decomposition of the WC phase during fabrication [97–108], drastically degrades performance. In a related study, it has been reported that a decrease in WC particle size in the feedstock powder led to an increase in the extent of decomposition of the WC phase [100], and thus significant amounts of W2C phase have been reported in nanostructured WC– Co coatings [18,103,104,107,108]. The chemical composition and structure of the matrix phase has a significant effect on the mechanical properties of the material [109,110]. Therefore, both phase transformations and grain growth in nanostructured WC –Co during exposure to elevated temperatures are of interest. To illustrate the issues involved, the phase, phase transformations and grain growth in WC–18% Co nanostructured powder, synthesized using cryomilling, are presented and discussed. In order to evaluate the thermal stability of cryomilled powders, He and Lavernia [111] isothermally treated samples at pre-determined temperatures ranging from 473 to 1273 K for 4 h, followed by quenching. The powders were sealed in stainless steel cans with nitrogen and isothermally treated. In the lower temperature range of 473–673 K, the powders were directly heated in laboratory air.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

303

Fig. 40. Indentation fracture on the Cr3C2-25 (Ni20Cr) coatings [6]: (a) conventional coating using 1000 g load; (b) nanostructured coating using 1000 g load; (c) conventional coating using 500 g load; and (d) nanostructured coating using 500 g load.

5.4.1. Phase transformations in the cryomilled WC – 18% Co powder The influence of heating in air on the XRD profiles of the cryomilled WC– 18% Co powder is shown in Fig. 41 [111]. Compared to the XRD of the as-received WC –18% Co powder, shown in Fig. 42, the XRD reflections of the WC phase in the cryomilled powders are noticeably broadened, and the peaks of Co disappear. However, no new phases were observed in the cryomilled powder despite the fact that nitrogen and oxygen contents were significantly increased. According to the XRD profiles in Fig. 41, there is no evidence indicating that significant oxidation of the cryomilled powder occurred in air below 623 K. This is favorable for storage, transportation, and agglomeration of cryomilled powders because the powders may be exposed to air during spray drying which is normally conducted at temperatures between 373 and 473 K. At 648 K, oxidation started to occur in the cryomilled powder exposed to air. The WO2 and WO3 oxide phases were detected by XRD. However, phases that form through the decarburization of WC, such as W2C and metallic W phases, were not detected. The

Fig. 41. XRD of the cryomilled WC – 18% Co powder exposed to air [111].

304

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 42. XRD of the as-received near nanostructured WC – 18% Co powder [116].

Fig. 43. XRD of the cryomilled WC – 18% Co powder exposed to a nitrogen atmosphere [111].

XRD spectra of the cryomilled WC–18% Co powders heated in nitrogen are shown in Fig. 43. When the temperature reached 873 K, the XRD reflections of fcc Co reappeared. In the specimen treated at 1273 K, the Co6W6C phase was detected by XRD. Once again, W2C metallic W phases, were not detected as a result of the low partial pressure of oxygen in the environment despite the fact that the powder contained 1.49 wt.% oxygen. The chemical constituents and structural characteristics of the Co matrix in WC–Co system materials have been widely studied [97,103,104,108,112–115]. Amorphous/Cox Wy Cz compounds have been documented experimentally in WC–Co powders and coatings [97,113–115]. In comparison to conventional WC–Co coatings, a pronounced amorphous ‘hump’ has been reported in XRD spectra of nanostructured WC coatings [103,104,108]. For example, in a WC–Co nanostructured bulk material sintered at 1623 K for 1 h, high resolution electron microscopy revealed [112] the Corich matrix phase to be composed of small (less than 8 nm) crystalline regions and interdispersed amorphous ones. The spacing values of the lattice fringes in the small crystalline regions matched those of the fcc crystal structure of Co. However, the chemical composition analysis, on the basis of EDS measurements, indicated the absence of pure Co [112]. Approximately 20 wt.% W and 80 wt.% Co in the small crystalline regions, and 43 wt.% W and 57 wt.% Co in the amorphous regions were measured. A larger Co lattice constant in the nanostructured samples as compared to those in the conventional ones was argued to indicate higher W dissolution in the matrix phase of nanostructured materials [109]. The chemical constituents and structural characteristics in the matrix phase are also functions of the temperature the material is exposed to. The XRD of the as-received WC–18% Co powder, shown in Fig. 42, indicates a peak at 2q = 44.24°, which corresponds to {111} reflection of the fcc Co crystal with a lattice constant of 0.3544 nm. In a related study, as-received WC –18% Co powder was separated into different powder particle size ranges [116]. With these powders of varying powder particle size, WC–18% Co near-nanostructured coatings were synthesized using HVOF thermal spraying [116]. The XRD of these coatings, synthesized at different powder particle temperatures, are shown in Figs. 44, 48 and 49. A peak at near 2q=44.24°, indicated by the arrows in Fig. 44, was also observed. This peak is noticeably broadened and became broader with increasing particle temperature. Concurrently, the peak moved towards a low diffraction angle, showing an increased crystallographic plane spacing and lattice constant, and finally formed a broad peak (‘hump’) consistent with the presence of an amorphous phase. The W and C elements, which originate

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

from the decomposition of WC particles, dissolve in the Co lattice and lead to an increase in crystallographic plane spacings and lattice constants. These results clearly indicate the formation of an amorphous broad peak is attributable to the dissolution of large amounts of W and C elements in the Co crystal. The formation of non-WC phases depends closely on the composition and structure of the matrix phase. The decarburization transformation from WC to W2C and W was thought to occur as a result of the direct oxidation on the surface of solid WC, that is, non-WC phases formed on the surface of WC [117– 120]. However, on the basis of TEM examination, Stewart et al. [108] proposed that, during thermal spraying, the Co matrix melts and WC particles are dissolved. In this case, the periphery of the semi-molten particle is decarburized by oxidation, promoting further WC dissolution in this region. In related work, particle quenching during impingement with a substrate resulted in precipitation of W2C in the matrix phase [108]. On the basis of the two mechanisms described above, a schematic diagram indicating the formation of non-WC phases is shown in Fig. 45. The argument that melting of matrix phase leads to formation of non-WC phases may explain an absence of non-WC phases in cryomilled powders, exposed to elevated temperature, because powders did not melt during thermal exposure in this case.

5.4.2. Influence of heat treatment on grain size On the basis of the XRD of the samples treated at different temperatures, He and Lavernia [111] investi-

Fig. 44. Influence of particle size on XRD of near nanostructured WC–18% Co coatings sprayed using hydrogen as fuel with the fuel/oxygen ratio of 2.80. SH — 20.33 mm; MH— 32.04 mm; and LH— 38.55 mm [116].

305

gated the dependence of grain size on temperature. XRD profiles shown in Fig. 41 indicate that milling leads to a disappearance of XRD reflections of the Co phase and the pronounced broadening of those of the WC phase. The reflections located at diffraction angles (2q) larger than 60° were so close that it was difficult to accurately measure the full-width at half-maximum (FWHM — D(2q)) and integral width (IntW—i) of these XRD reflections [111]. Therefore, the calculations of grain size and microstrain were performed using the data corresponding to the three strongest peaks of the WC phase, that is, {0001}, {101( 0} and {101( 1} reflections occurred at near 2q= 31.7, 35.9 and 48.5°, respectively. The FWHM and IntW of these three reflections are measured using the corresponding software package in the Siemens D5000 diffractometer. The variation of grain sizes (WC particle size) with temperature was determined by the Scherrer equation, described in Section 3.4, and the results are plotted in Fig. 46 [111]. The results show that, in the Ž0001 orientations, particle size had its maximum value and minimum value in Ž101( 1 orientations. The average size of the WC particles was approximately 25 nm in the as-cryomilled powder. Below 873 K, the size of the WC particles remained invariant with increasing temperature. Growth of WC particles started above 873 K. Therefore, there are two stages for WC particle growth, a near constant value below 873 K and a growth over 873 K. In a related study [44], two different mechanisms of grain growth were suggested to be responsible for the difference in grain growth of lower and higher temperature ranges. In the lower temperature range, nanostructured Fe exhibited the same general trend as zone-refined polycrystalline Fe, while a pinning effect on the grain boundary, in the higher temperature range, played a primary role. In calculations using the Scherrer equation, microstrain is not invoked. The linear fitting method [40,41] is invalid to calculate the microstrains because an inherent assumption made in the linear fitting method is that the material is an isotropic one in which the microstrains in different crystallographic orientations are identical (an unique slope). The elastic modulus values of WC single crystal on {0001}, {101( 0} and {101( 1} planes are not available. However, the microhardness values of WC single crystals exhibited a significant difference between {0001} and {101( 0} planes [121]. The particle sizes of the WC phase calculated using the Scherrer equation, shown in Fig. 46, also showed a clear difference in particle size for Ž0001, Ž101( 0 and Ž101( 1 crystallographic orientations. This indicates non-isotropic behavior of the WC phase. Accordingly, it is unlikely that the data corresponding to {0001}, {101( 0} and {101( 1} reflections will fit a single straight line. Therefore, the linear fitting method is not suitable for calculations of particle sizes and microstrains in the cryomilled WC– 18% Co powder [111].

306

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 45. Schematic diagram indicating the two mechanisms responsible for the formation of non-WC phases. Light areas denote WC particles or Co matrix; gray areas denote matrix containing W and C, and dark areas denote non-WC phases [111]. (a) Non-WC phases form by direct oxidation on the surface of WC phase: (i) initial stage; and (ii) non-WC phases formed on the surface of the WC phase. (b) Melting of the binder phase leads to the formation of non-WC phases: (i) initial stage; (ii) Co matrix melted and WC particles dissolved; (iii) WC further dissolution on periphery; and (iv) after impingement with substrate, precipitation of non-WC phases in binder.

The single-line approximation developed by de Keijser et al. [42], described in Section 3.4 in detail, was employed to calculate particle size and microstrain in an individual crystallographic orientation [111]. In this case, the isotropic assumption was not required for the single-line approximation calculation. Using a Microsoft Excel calculating table, WC microstrains and particle size were calculated and the results are plotted in Fig. 47. The calculated microstrains ranged from 0.12 to 0.36%. The microstrains decreased with increasing temperature. Below 623 K, as temperature increased, the microstrain values decreased and then increased. Total microstrains are presumably composed of the microstrains that originated from cryomilling and that from quenching (rapid cooling). As temperature increases, the microstrains generated by cryomilling decrease and those formed by quenching increase. In contrast to quenching, an annealing heat treatment

Fig. 46. Particle sizes of WC calculated using the Scherrer equation [111].

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

307

1023 K [118]. However, Jia et al. [112] reported that in WC –Co bulk nanostructured materials, sintered at 1623 K for 1 h, there was no evidence indicating significant particle growth of the WC phase. The sintering of pure nanostructured WC powder involved three stages [126]. In the first stage, WC particles were rearranged at low temperatures (less than 1023 K) without particle growth. In the second stage (1273–1523 K), a neck developed between powder particles and initial particle growth was observed. In the final stage, the pores were eliminated, accompanying massive particle growth of the WC phase at 1573–1723 K. Particle growth of WC particles occurred on preferred crystallographic planes, i.e. grain growth on (101( 0), (11( 03) and (011( 1) planes was reportedly observed [126]. In a recent study [111], the powder was not compacted into green compacts prior to thermal exposure. This is thought to influence the temperature at which WC particles begin to grow.

Fig. 47. Particle sizes and microstrains of WC calculated using the single-line approximation [111]: (a) particle size; and (b) microstrain.

(slow cooling) decreased the microstrains caused by the cooling process. However, an annealing process leads to a variable heating time for the powders treated at the different temperatures. The entire line broadening of XRD in the Scherrer equation is treated as a contribution of small crystallites, while the partial line broadening in the single-line approximation is treated as the contribution of the microstrain. Thus, particle sizes calculated using singleline approximation were larger than those calculated using the Scherrer equation. However, both calculation methods led to self-consistent particle size values. TEM observations of WC particles in the as-cryomilled powder showed an average value of 18.9 nm, close to the result from calculation of the Scherrer equation [111]. The growth of nanostructured WC particles during sintering has been widely investigated [20,21,112,119,122– 126]. Porat et al. indicated that significant particle growth of the WC phase occurred at

5.4.3. Influence of spraying parameters In addition to characteristics of feedstock powder, the spraying parameters, such as fuel-chemistry and fuel/oxygen ratio also affect the behavior of nanostructured coatings. In the case of WC–Co coatings, it has been reported that the decomposition of WC into W2C, W3C significantly decreased performance, and the decomposition of WC increased as grain size decreased [100]. However, experimental results indicated that the decomposition of WC was closely related to the spraying parameters chosen [116]. This section illustrates how spraying parameters can influence the behavior of near-nanostructured WC–18% Co coatings. In a study conducted by He et al. [116], near-nanostructured WC– 18% Co feedstock powder shown in Fig. 33, which consisted of 200–500 nm WC grains in a Co matrix, was agglomerated into particles with a size ranging from 10 to 56 mm. The as-synthesized powder was separated into different particle size ranges using a sieve with mesh number of 500. There was no evidence indicating the presence of W2C, W3C in the as-synthesized powder, see Fig. 42. In these experiments, either propylene or hydrogen was used as fuel. The combustion reaction of propylene (C3H6) with oxygen is shown in Eq. (16): C3H6 + 4.5O2 = 3H2O+ 3CO2

(16)

Therefore, the stoichiometric fuel/oxygen ratio is 1/ 4.5= 0.222. If the value of the fuel/oxygen ratio is increased, the gaseous mixture is fuel rich; and if fuel/ oxygen ratio is less than 0.222, the gaseous mixture is oxygen rich. The combustion of hydrogen with oxygen is indicated in Eq. (17): H2 + 0.5O2 = H2O

(17)

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

308

Table 8 Influence of fuel-chemistry, fuel/oxygen ratios, and particle size on near nanostructured WC–18% Co coatings [116] Code

Size (mm)

Fuel

Fuel/O2

Temperature (K)

W2C (vol.%)

Porosity (vol.%)

MH45 MH62 MH75 SH62 LH62 MP32 MP40 MP48

32.04 32.04 32.04 20.33 38.55 32.04 32.04 32.04

H2 H2 H2 H2 H2 C3H6 C3H6 C3H6

1.98 2.80 3.46 2.80 2.86 0.242 0.304 0.367

1519 1807 1840 1866 1687 2130 2291 2223

1.57 3.80 4.68 8.65 2.89 3.79 6.71 5.69

25.3 11.98 9.50 10.83 12.31 2.08 1.99 1.98

Thus, the stoichiometric fuel/oxygen ratio for the combustion reaction of hydrogen is 1/0.5 = 2. Correspondingly, larger than 2 is fuel rich, and less than 2 is oxygen rich. Because high-pressure air, which also contains oxygen, was introduced into the mixture during spraying, the real stoichiometric fuel/oxygen ratio for propylene and hydrogen must be smaller than 0.222 and 2, respectively. A quantitative investigation of the particle temperature while impinging on the substrate was conducted, using In-flight diagnostic equipment described in Section 5.1, to establish a relationship between the particle size, fuel-chemistry and the fuel/ oxygen ratio, and the resultant particle temperature. The results of these studies are summarized in Table 8 [116]. As the hydrogen/oxygen ratio increased, the particle temperature increased. The maximum particle temperature measured in the M powders was 1870 K, which was generated by hydrogen with a hydrogen/oxygen ratio of 3.46. Particle temperature significantly increased when the hydrogen/oxygen ratio increased from 1.98 to 2.80, but slightly increased when the hydrogen/oxygen ratio was increased from 2.80 to 3.46. This indicates that increases in hydrogen/oxygen ratio may not increase the particle temperature further. Particle temperature was also seen to increase as particle size decreased. An increase in average particle size from 20.33 to 38.55 mm led to a decrease in the maximum particle temperature of 350 K. The highest particle temperature observed was 2291 K, which was generated with a propylene/oxygen ratio of 0.304 and was 421 K higher than that of hydrogen. Compared to changes in particle temperature caused by different fuel/oxygen ratios, the variation of fuel-chemistry and particle size have a much stronger influence on particle temperature. XRD of the near nanostructured WC– 18% Co coatings are shown in Figs. 44, 48 and 49. In all coatings, the diffraction peaks from W phase were not detected. However, a W2C phase peak was observed at near 2q= 40°. On the basis of ratio of intensity of W2C phase peak to that of WC phase peak on XRD spectra, the volume fraction of W2C phase in coatings was determined and the results are also listed in Table 8.

Fig. 48 shows the influence of the hydrogen/oxygen ratio on the phase constituents. An increase in the hydrogen/oxygen ratio, which led to an increased particle temperature, caused an increase in the intensity of the W2C phase peaks. The amount of the W2C phase increased from 1.57 to 4.68% when the hydrogen/oxygen ratio increased from 1.98 to 3.46. Fig. 44 shows the influence of particle size on the XRD of the coatings. The intensity of the W2C phase peak increased as particle size decreased because a decrease in particle size led to an increase in particle temperature during spraying. In the coating sprayed using the L (large) powder, 2.89% W2C phase was observed, while 8.05% W2C phase was detected in the coating sprayed using the S (small) powder. Fig. 49 indicates the effects of the propylene/oxygen ratio on phase constituents. The coating sprayed using a propylene/oxygen ratio of 0.304 had a maximum intensity for

Fig. 48. Influence of hydrogen/oxygen ratio on XRD of near nanostructured WC – 18% Co coatings sprayed with hydrogen. MH45 — H2/O2 =1.98; MH62 —H2/O2 =2.80; and MH75 —H2/O2 = 3.46 [116].

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 49. Influence of propylene/oxygen ratio on XRD of near nanostructured WC – 18% Co coatings sprayed with propylene. MP32 — C3H6/O2 = 0.242; MP40 —C3H6/O2 = 0.304; and MP48— C3H6/O2 = 0.367 [116].

the W2C phase peak because this propylene/oxygen ratio produced highest particle temperature. However, the degree of decomposition of WC phase was on the same level as that in the coatings sprayed using hydrogen, although propylene generated much higher particle temperature than hydrogen. This result implies that both fuel-chemistry and particle temperature play an important role in the decomposition of the WC phase. The results suggest that propylene (C3H6) provides more benefits than hydrogen (H2) for spraying of WC– Co system coatings. The flame enthalpy is the primary driving force for the decomposition of the WC phase. However, the gaseous mixture chemistry also plays an important role in the decomposition of the WC phase. A related study shows that Ar or ArHe plasma spraying usually results in less decomposition of the WC phase than in the case of ArH2 plasma spraying [117]. Similarly, during spraying with hydrogen, the WC phase experienced a combination of decomposition by an oxidation reaction with oxygen and a reduction reaction with the H2, i.e. 2WC + 2H2 =W2C+ CH4. Conversely, during spraying with propylene, the WC phase experienced decomposition by an oxidation reaction with oxygen, and the free carbon liberated from the WC phase to form CO and CO2 gases. The product of the combustion reaction of propylene with oxygen, as indicated in Eq. (16), contained CO2. Therefore, the decomposition reaction with oxygen, i.e. WC+ O2 =W2C +CO, and 2CO + O2 = 2CO2, was significantly reduced by addition of free carbon to the starting powder, although the pres-

309

ence of residual free carbon in the coating led to an increase in wear rates [127]. An effort was proposed to introduce a suitable amount of free carbon into the powder to prevent decomposition of the WC phase. But this did not leave free carbon in the coating [127], because this approach is equivalent to introducing CO and CO2 into the gaseous mixture. Therefore, although propylene generated a much higher flame temperature, the extent of decomposition of the WC phase in the coatings sprayed using both propylene and hydrogen was comparable. As a result of the analysis, it is concluded that the amount of non-WC phases in the coatings depends both on the particle, i.e. size and morphology, and flame characteristics, such as temperature and chemistry. The cross-sectional microstructures of the nearnanostructured WC–18% Co coatings sprayed using hydrogen are shown in Fig. 50 [116]. A large proportion of carbide particles exhibited rectangular boundaries, and their size ranged from 200 to 500 nm, identical to that of the feedstock powder. The porosity measured as a fraction of area, was statistically measured and the results are listed in Table 8 [116]. The porosity in the coatings depended primarily on the temperature the powder particle experienced during spraying; low temperature resulting in a higher porosity. Very high porosity was observed in coatings sprayed using hydrogen, regardless of hydrogen/oxygen ratio and powder particle size. Thus, hydrogen is not a suitable fuel for spraying near-nanostructured WC– 18% Co coatings. However, dense near nanostructured WC –18% Co coatings with low non-WC phase were successfully deposited using propylene. Mathematical simulations of the HVOF process have been performed [80,128,129]. Figs. 51 and 52 show the results of the simulations, indicating the influence of particle size and morphology on particle temperatures. These results were similar to those obtained experimentally. Sliding wear and abrasion tests were conducted for near-nanostructured WC–18% Co coatings [130]. The sliding wear tests were on a ball-on-disk tribometer at room temperature, where a WC–Co coated steel plate slid against a commercial Si3N4 ball without lubrication. The load was 9.8 N and the sliding velocity was 30 mm s − 1. The diameter of the traveling circle of the pin on disk was 8–13 mm. The sliding distance was 10 000–12 000 m corresponding to 250 000–480 000 rotations of the disk. The wear volume was calculated according to the profile of the wear scar measured by means of a profilometer. Generally, the amount of material removed on sliding wear is proportional to the normal force on the ball in Newtons and the distance slid in meters. The wear rate can be defined in mm3 N − 1 m − 1. The inverse of the wear rate in N m mm − 3 is defined as the wear resistance. Abrasion

310

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 50. Cross-sectional microstructures of the near nanostructured WC – 18% Co coating (sample MH62). (a) – (d): different magnifications as shown [116].

tests were performed on a ML-100 pin-on-disk tribometer. A WC –Co coating was used as pin against disk with bonded silicon carbide abrasive. The silicon carbide plain-back abrasive paper was 120 grit (106 mm average particle size). The pin was displaced along the radius direction by 1 mm per turn while the disk moved at 54 rpm. This ensured that fresh abrasive was constantly used during the test. The load was 7 N, and the sliding distance was 38 m according to the equation: S = y(r 22 −r 21)/L, where L was the radial displacement of the pin in one spin of the disk, and r1 and r2 were the distances from the center of the disk to the pin at the beginning and at the end of a test run, respectively. In this study, r1 = 5 mm, r2 =110 mm, and L = 1 mm. The wear volume was calculated from the weight loss [130]. The average sliding wear rate of the coatings, determined from at least ten profilometer traces for each sample, are listed in Table 9 [130]. The wear resistance, the inverse of wear rate, is plotted against hardness in Fig. 53. The sliding wear resistance of WC–Co coatings increased roughly linearly with the hardness [130]. However, samples MP48, MH75, and SH62 exhibited much lower wear resistance than this trend. This higher wear rate is attributed to the brittleness of the material, which is the result of extensive decarburization. In these samples, decarburization did not seriously occur and another cause of low wear resistance of the samples should be present.

As expected from their higher hardness, the coatings deposited with propylene fuel in a near-stoichiometric ratio had higher wear resistance than those sprayed by hydrogen. Sample MP48, which was sprayed with a high propylene/oxygen ratio, showed low wear resistance. In order to obtain more information on wear behavior, the worn surfaces were examined with normal and 45° incidence SEM, as shown in Fig. 54 [130]. Large pits in the wear scar of WC–Co materials were ob-

Fig. 51. Simulated results indicating influence of diameter of spherical particle on particle temperature [128].

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

311

Fig. 52. Simulated results showing influence of particle morphology on particle temperature during HVOF thermal spraying. E is aspect ratio of spherical particle [128].

served, which suggest that wear occurred predominately by the removal of entire splats [131]. These pits were more numerous in sample MP48, shown in Fig. 54(b), than in MP32, shown in Fig. 54(a). Sample MP48 showed large cracks in the wall of the pit, which were not present in sample MP32. A similar crack can be seen on the worn area of sample MH75 (upper right corner of Fig. 54(c)). This crack encircled an area about 2 mm in diameter, which would probably fall out in further sliding. The micrographs at 45° incidence on the right side of Fig. 54, show clearly that the cracks on samples MP48 and MH75 were roughly parallel to the surface, which reinforce the view that material was removed by fracture along such cracks. No cracks were visible on sample MP32. Because the whole surface of the sample was abraded in the abrasive tests, a correlation of local abrasion resistance and local hardness could not be established [130]. It is noted that there is no simple relation between hardness and abrasion resistance of these samples, although, as a group, the harder samples deposited with propylene fuel had a higher abrasion resistance than the coatings sprayed with hydrogen. A decreased abrasion resistance is attributed to the pres-

Fig. 53. Wear-resistance of WC – 18% Co coatings as a function of their hardness [130]: (a) sliding wear resistance of WC –18% Co coatings; and (b) abrasion wear resistance of WC – 18% Co coatings.

ence of cracks in the boundaries between splats, just as it did in sliding wear [130].

Table 9 Hardness, sliding wear and abrasive wear rates of near-nanostructured WC–18% Co coatings [130] Sample

Particle size (mm)

Fuel/oxygen ratio

HV1000 (Kg mm−2)

Sliding wear rate (×106 mm3 N−1 m−1)

Abrasive wear rate (×10−2 mm3 N−1 m−1)

SH62 MH62 LH62 MH75 MP32 MP40 MP48

20.33 32.04 38.55 32.04 32.04 32.04 32.04

H2/O2:2.80 H2/O2:2.80 H2/O2:2.80 H2/O2:3.46 C3H6/O2:0.242 C3H6/O2:0.242 C3H6/O2:0.242

1004 825 735 1018 1240 1156 1122

0.28 0.32 0.30 0.38 0.19 0.20 0.36

1.50 1.20 1.00 1.10 0.78 0.90 1.10

312

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

5.5. Post-treatment of nanostructured coatings At the beginning of this century, Wilm [132] discovered the phenomenon of precipitation hardening in Al alloys. It was postulated that the increase in hardness of

aluminum alloys with time was due to precipitation of a new phase in an initially supersaturated solid solution [133]. Subsequent studies using TEM confirmed the precipitation of very fine particles in aged alloys [134]. Mott and Nabarro [135] and Orowan [136] put forth

Fig. 54. Worn Surface of WC –Co coatings deposited by: (a) sample MP32, C3H6/O2 =0.242; (b) sample MP48, C3H6/O2 =0.367; and (c) sample MH75, H2/O2 =3.46 viewed at normal (left side) and 45° incidence (right side) [130].

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

313

Fig. 55. Precipitates in the nanostructured Cr3C2-25 (Ni20Cr) coating. Arrows O indicate the initial carbides, and arrows P indicate the precipitates [95]: (a) bright field image; and (b) dark field image.

dislocation models of precipitation/dispersion hardening. A systematic description of mechanics models relating to the precipitation strengthening is available in a review article by Ardell [137]. The phenomenon of supersaturation, which results in subsequent precipitation, is commonly observed in materials manufactured by non-equilibrium processes, i.e. rapid-quenching, mechanical alloying/milling, and thermal spraying. In related studies, a significant amount of W and C were dissolved into the matrix phase of WC– Co coatings [115], and a non-equilibrium microstructure was detected in the as-sprayed nanostructured Cr3C2-25 (Ni20Cr) coatings synthesized by mechanical milling and HVOF thermal spraying [6]. Therefore, it is anticipated that precipitation will occur in these coatings under certain conditions. This section discusses the precipitation in nanostructured Cr3C2-25 (Ni20Cr) coatings and the influence of precipitation on coating properties. The as-sprayed coatings were thermally exposed to air at 473, 673, 873 and 1073 K for 8 h [95]. Fig. 55(a) and (b) reveals spherically shaped precipitates in the nanostructured coating treated at 1073 K that were likely to have formed by nucleation and growth in the matrix. In addition to original carbide particles, some very fine precipitates were observed. The average size of the original carbide particles increased from 24 nm in the as-sprayed nanostructured coating to 39 nm in the nanostructured coating exposed at 1073 K, whereas the precipitates had an average size of 8.3 nm. In addition to precipitation, structural changes in the elongated amorphous phases were also observed [95]. In the assprayed nanostructured coatings, elongated amorphous phases were often observed, see Fig. 38 and Ref. [6]. The morphology and dimensions of the elongated amorphous phases did not change during heat treat-

ment. However, a few changes can be observed in Fig. 56 [95]. Fig. 56(a) is a TEM bright field image. Arrow M indicates the matrix containing a high density of precipitates and A illustrates an originally elongated amorphous phase. The SAD pattern of area A, shown in Fig. 56(b), indicates the amorphous phase has crystallized. Very thin and long streaks in the SAD pattern indicate the presence of thin twins. Using a diffraction streak, the dark field image of the original elongated phase was taken and is shown in Fig. 56(c), in which very thin thick twins can be seen. Precipitates were not found in the elongated structure. Therefore, the changes in the elongated amorphous phases during annealing can be summarized as follows: (1) the morphology and dimensions have not been altered; (2) the amorphous structure has crystallized; (3) deformation twins were formed inside the phase; and (4) no precipitation occurred inside. Fig. 56(d) is a SAD pattern of the matrix M consisting of two sets of diffraction patterns from the fcc matrix and the Cr2O3 particles [95]. Arrows P indicate the diffraction spots of Cr2O3 and arrow M indicates diffraction spots of fcc matrix. The dark field image of the precipitates, taken using a P spot and is shown in Fig. 56(e), reveals a distribution of spherically shaped precipitates in the matrix. The dark field image of the matrix, taken using a M spot is shown in Fig. 56(f), shows an average grain size of the matrix to be approximately 150 nm. Phase transformations in conventional Cr3C2 –NiCr coatings at high temperatures have been reported in the literature [90,94,138–140]. Lai [139,140] showed that there were structural changes in carbides from Cr3C2 to Cr7C3 to Cr23C6 during exposure to high temperature. The relative thermodynamic stability of carbides was determined by their standard free energies for formation. The Cr2O3 phase was detected by XRD in the

314

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Cr3C2 –NiCr coatings exposed to air [94,138], high-pressure helium [139,140], and N2-3%H2 gas [90] at high temperatures. Oxygen was not observed by X-ray mapping in the NiCr matrix phases. Thus, the precipitate

was not found in the original elongated amorphous phases [95]. In the as-sprayed coatings, a high oxygen content was observed in the nanostructured coating, while there was a low oxygen content in the conven-

Fig. 56. TEM analysis of precipitate, matrix and changes in the original amorphous phase [95]: (a) Bright field image containing matrix, precipitates and the original amorphous phase; (b) SAD pattern of the original amorphous phase, very thin and long streaks illustrating thin twins are observed; (c) Dark field image of the original amorphous phase; (d) SAD pattern containing diffraction spots from both matrix and precipitates; (e) Dark field image of the precipitates; and (f) Dark field image of the matrix, illustrating grains.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

Fig. 57. Variation of microhardness of Cr3C2-25 (Ni20Cr) coatings with heat treatment temperature [95].

Fig. 58. Scratch-resistance of nanostructured Cr3C2-25 (Ni20Cr) coatings as a function of temperature used [95]. Table 10 Influence of heat treatment on coefficients of friction of nanostructutred Cr3C2-25 (Ni20Cr) coatings [95] Temperature (K)

As-sprayed

673 K

873 K

Coefficient of friction

0.216

0.205

0.183

tional coating. In the coatings exposed at 873 K, high oxygen contents were found in both conventional and nanostructured coatings [95]. Therefore, oxidization occurred in the carbide phases of Cr3C2-25 (Ni20Cr)

315

coatings exposed to air at high temperatures regardless of original oxygen content in the coatings. The average microhardness, measured on the crosssection, of both conventional and nanostructured Cr3C2-25 (Ni20Cr) coatings is shown in Fig. 57 [95]. Microhardness of the conventional coating only increased slightly with increased exposure to all temperature ranges, while that of the nanostructured coating drastically increased from 1020 to 1240 HV300 in the temperature range 700–900 K, and then approached a constant value. The scratch tests, described in Section 5.3, were employed to investigate the influence of heat treatment on scratch depth of the nanostructured coatings [95]. The scratch depth decreased with increasing heat treatment temperature, see Fig. 58. With a load of 5 N, no scratch was found in the nanostructured coating treated at 873 K [95]. The scratch depth results were supported by the measured microhardness values. Coefficient of friction, under the mode of the ‘ballon-disk’ friction, was also obtained from the scratch tests and are listed in Table 10 [95]. The coefficient of friction decreased as heat treatment temperature increased. A number of publications report the application of Cr3C2 –NiCr coatings at elevated temperatures [86,87,90,94,138–140]. In a related study [90], the Cr3C2 –NiCr was heated at 1123 K for 2 h and the results showed that the hardness on cross-sections did not decrease even though Cr2O3 was detected by XRD. The results published by Fukuda and Kumon [94] showed that heat treatment at 923 K for 1000 h yielded a maximum hardness value of 1100 for D-gun sprayed Cr3C2-25 (Ni20Cr) coatings. In their experiments, 900 HV for the as-sprayed coatings, 1100 for the coatings heated at 923 K for 1000 h, and 1000 for the coatings heated at 1123 K for 1000 h were reported. Cr3C2 – NiCr coatings, with the highest hardness, offered the best abrasive and erosive wear resistance [94]. Therefore, it is concluded that a suitable post-sprayed heat treatment is beneficial for properties of both conventional and nanostructured Cr3C2 –NiCr coatings. Microhardness of a material is strongly dependent on its microstructure. Precipitation, depending on size, distribution and behavior of the precipitates formed, usually caused hardening [135,136]. High density of fine, hard and dispersed precipitates in the matrix significantly increased microhardness [137]. The size, distribution and density of precipitates resulted from the combined effect of nucleation and growth of the precipitates, which are determined by aging processing, mainly aging temperature and time [141–147]. For most types of precipitates, the relationship between the nucleation of precipitates and aging temperature can be represented by a C-shaped curve [148]. In other words,

316

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

an incubation is needed for the nucleation of precipitates and the shortest incubation appears in a specified temperature range. The temperature corresponding to the shortest incubation is usually chosen as the aging temperature. A typical hardness versus aging time curve in precipitation hardening alloys showed three stages, under-aged, aged and over-aged stages [148]. The hardness of the alloy increases sharply in the under-aged stage and reaches its maximum value in the peak aged stage, then decreases with increasing aging time in the over-aged stage because of coarsening of precipitates. On the basis of the distribution of oxygen in the coatings [95], it is proposed that the observed Cr2O3 particles in the nanostructured coatings evolved as follows. First, there was a high oxygen potential in the as-sprayed nanostructured coatings, due to the presence of oxygen in the starting powder (1.93 wt.% , see Table 2), as well as oxygen entrapment during thermal spraying. Second, oxygen reacted with chromium to form CrO, CrO2, CrO3, Cr2O3, and Cr3O4. However, CrO, CrO2, CrO3 and Cr3O4 were metastable and/or present only under high pressure [149]. Therefore during thermal exposure, Cr2O3 was directly formed from the chemical reaction: 2Cr3C2 +17/2O2 =3Cr2O3 +4CO2

(18)

The formation of the Cr2O3 phase is supported by TEM analysis that revealed the presence of the diffraction pattern from the hexagonal (a = 0.4954 nm, c= 13.584 nm) Cr2O3 phase, see Fig. 54 [95]. The Cr2O3 phase, with the Gibbs free energy of formation of DG = − 251.70 kcal mol − 1 [150] was relatively stable. Third, the nucleation and growth of Cr2O3 phase uniformly occurred throughout the coating because most of the oxygen was primarily observed in nano-Cr3C2 particles [95] that were uniformly distributed in the coating. Therefore, this internal oxidation process led to the formation of fine (8.3 nm) and dispersed Cr2O3 precipitates in the nanostructured Cr3C2 coatings [95]. During thermal exposure, the crystallization of the amorphous phase in the coatings occurred, see Fig. 54 and Ref. [95]. Crystallization of the amorphous phase usually caused a decrease in hardness [151]. Therefore, this change in the amorphous phase has a negative influence on the hardness of the coatings. However, the decrease in hardness caused by crystallization of the amorphous phase was compensated for by the increase created from the precipitation of oxide particles. Accordingly, on the basis of the findings, the observed increase in microhardness of Cr3C2 – NiCr coatings exposed to high temperatures is attributed to the precipitation of oxides. Particularly in the nanostructured coatings, the high density, nano-sized oxide particles lead to significant increases in microhardness and an increase in scratch-resistance.

6. Summary This article reviews the synthesis and characterization of nanostructured feedstock powders, the agglomeration of these powders used in coatings, and processing and characterization of nanostructured thermal sprayed coatings. The published results show that mechanical milling can be effectively used to synthesize nanostructured powders. Whether a composite or a single-phase starting powder is involved, mechanical milling leads to the formation of nanocrystalline structure under certain milling conditions, although mechanisms governing the formation of a nanocrystalline structure depend strongly on the characteristics of the starting constituents. Using nanostructured powders as feedstock powders, various nanostructured coatings with improved performance have been successfully synthesized, while process parameters and post-sprayed treatment can significantly influence mechanical and physical properties of nanostructured coatings. There are still a number of scientific and technological issues that have not been solved. Two of them are of imperative interest. First, the origin of a nanocrystalline structure in the nanostructured coatings remains an open question. It is not clear whether the nanocrystalline structure results from a retention of a nanocrystalline structure in feedstock powders or from the formation of nanocrystalline structure during spraying of nanocrystalline powders. Second, the fundamental factors that govern observed thermal stability of nanostructured powders have not been established. Despite these limitations, available experimental evidence suggests that nanostructured coatings have strong technological potential, and rapid industrial growth is expected over the next decade.

Acknowledgements The authors gratefully acknowledge financial support provided by the Office of Naval Research under grants N00014-94-1-0017, N00014-98-1-0569 and N00014-001-0109, and N00014-01-C-0384 as well as many useful discussions with Professor Enrique J. Lavernia at University of California, Irvine.

References [1] L. Pawlowski, The Science and Engineering of Thermal Spray Coatings, Wiley, England, 1995. [2] C.M. Hackett, G.S. Settles, in: C.C. Berndt (Ed.), Thermal Spray: Practical Solutions for Engineering Problems, Materials Park, Ohio, 1996, p. 665. [3] C. Suryanarayana, C.C. Koch, in: C. Suryanarayana (Ed.), Non-equilibrium Processing of Materials, Pergamon, New York, 1999, pp. 313 –344.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319 [4] E.J. Lavernia, M.L. Lau, H.G. Jiang, Thermal spray processing of nanostructured materials, in: C.M. Chow, N.I. Noskova (Eds.), Nanostructured Materials, Kluwer, Dordrecht, The Netherlands, 1998, pp. 283 –302. [5] R. Birringer, in: G.C. Hadjipanayis, R.W. Siegel (Eds.), Nanophase Materials: Synthesis – Properties –Applications, Kluwer, Dordrecht, The Netherlands, 1994, p. 157. [6] J. He, M. Ice, E.J. Lavernia, Metall. Mater. Trans. A31 (2000) 555 – 564. [7] C. Suryanarayana, Int. Mater. Rev. 40 (1995) 41 –64. [8] E. Hellstern, H.J. Fecht, Z. Fu, W.L. Johnson, J. Mater. Res. 4 (1989) 1292 – 1295. [9] H.J. Fecht, E. Hellstern, Z. Fu, W.L. Johnson, Metall. Trans. A21 (1990) 2333 – 2337. [10] E. Hellstern, H.J. Fecht, C. Garland, W.L. Johnson, J. Appl. Phys. 65 (1989) 305 –310. [11] H.J. Fecht, Nanostruct. Mater. 6 (1995) 33 – 42. [12] J. Eckert, J.C. Holzer, C.E. Kill III, W.L. Johnson, J. Mater. Res. 7 (1992) 1751 –1761. [13] J.S.C. Jang, C.C. Koch, J. Mater. Res. 5 (1990) 489 –510. [14] C.C. Koch, Nanostruct. Mater. 2 (1993) 109 –129. [15] C.C. Koch, Nanostruct. Mater. 9 (1997) 13 –22. [16] D.G. Morris, Mechanical Behavior of Nanostructured Materials, Trans. Tech. Publ., Switzerland, 1998, p. 85. [17] H.J. Fecht, G. Han, Z. Fu, W.L. Johnson, J. Appl. Phys. 67 (1990) 1744 – 1748. [18] J. He, M. Ice, S. Dallek, E.J. Lavernia, Metall. Trans. A31 (2000) 541 – 553. [19] B.S. Murty, S. Ranganathan, Int. Mater. Rev. 43 (1998) 101 – 141. [20] L.E. McCandlish, B.H. Kear, B.K. Kim, Nanostruct. Mater. 1 (1992) 119 – 124. [21] B.H. Kear, L.E. McCandlish, Nanostruct. Mat. 3 (1993) 19 – 30. [22] J.S. Benjamin, Met. Trans. 1 (1970) 2943 –2951. [23] H.G. Jiang, M. Ruhle, E.J. Lavernia, J. Mater. Res. 14 (1999) 549 – 559. [24] J.S. Benjamin, Mater. Sci. Forum 88 –90 (1992) 1 –18. [25] J. He, M. Ice, E.J. Lavernia, Nanostruct. Mater. 10 (1998) 1271 – 1283. [26] J.H. Ahn, H.S. Chung, R. Watanabe, Y.H. Park, Mater. Sci. Forum 88– 90 (1992) 347 –354. [27] M.L. Lau, H.G. Jiang, W. Nuchter, E.J. Lavernia, Phys. Status Solidi (A) 166 (1998) 257 –268. [28] W.A. Kaczmarek, Mater. Sci. Forum 179 –181 (1995) 313 – 320. [29] Y.S. Song, W.C. Kim, T.H. Kim, Mater. Sci. Forum 88 – 90 (1992) 191 – 196. [30] J. Zbiral, G. Jangg, Mater. Sci. Forum 88 –90 (1992) 19 – 26. [31] M.L. Lau, E. Strock, A. Fabel, C.J. Lavernia, E.J. Lavernia, Nanostruct. Mater. 10 (1998) 723 –730. [32] J. He, E.J. Lavernia, J. Mater. Res. 16 (2001) 2724 –2732. [33] J. He, M. Ice, E.J. Lavernia, in: C.M. Chow (Ed.), Nanostructured Films and Coatings, Kluwer, The Netherlands, 2000, pp. 131 – 148. [34] C. Suryanarayana, Prog. Mater. Sci. 46 (2001) 1 –84. [35] T.H. Courtney, Z. Wang, Scr. Metall. Mater. 27 (1992) 777 – 782. [36] S. Saji, S. Abe, K. Matsumoto, Mater. Sci. Form. 88 –90 (1992) 367 – 374. [37] P.S. Gilman, J.S. Benjamin, Annu. Rev. Mater. Sci. 13 (1983) 279 – 300. [38] P. Scherrer, Nachr. Ges. Wiss. Goettingen, Math.-Phys. K1.2 (1918) 98. [39] H.P. Klug, L.E. Alexander, X-ray Diffraction Procedures, Wiley, New York, 1974, 643pp. [40] J.W. Jeffery, Methods in X-ray Crystallography, Academic Press, London and New York, 1971, 83pp.

317

[41] C. Suryanarayana, M.G. Norton, X-ray Diffraction —A Practical Approach, Plenum Press, New York, 1998. pp. 207 –218. [42] Th.H. de Keijser, J.I. Langford, E.J. Mittemeijer, A.B.P. Vogels, J. Appl. Crystallogr. 15 (1982) 308 – 314. [43] Sir P. Hirsch, A. Howie, R.B. Nicholson, D.W. Pashley, M.J. Whelan, Electron Microscopy of Thin Crystals, Robert E. Krieger Publishing Co. Inc., New York, 1977, 24pp. [44] T.R. Malow, C.C. Koch, Acta Mater. 45 (1997) 2177 –2186. [45] J.I. Langford, J. Appl. Crystallogr. 11 (1978) 10. [46] J. He, F. Zhou, G. Chang, E.J. Lavernia, J. Mater. Sci. 36 (2001) 2955 – 2964. [47] M.L. Lau, H.G. Jiang, R.J. Perez, J. Juarez-Islas, E.J. Lavernia, Nanostruct. Mater. 7 (1996) 847 – 856. [48] V.L. Tellkamp, M.L. Lau, A. Fabel, E.J. Lavernia, Nanostruct. Mater. 9 (1997) 489 – 492. [49] H.K. Kohl, K. Peng, J. Nucl. Mater. 101 (1981) 243 –250. [50] H.K.D.H. Bhadeshia, Mater. Sci. Eng. A223 (1997) 64 –77. [51] J. He, S. Fukuyama, K. Yokogawa, Mater. Sci. Technol. 11 (1995) 914 – 919. [52] G. Han, J. He, S. Fukuyama, K. Yokogawa, Acta Mater. 46 (1998) 4559 – 4570. [53] K. Sobczyk, B.F. Spencer Jr., Random Fatigue from Data to Theory, Academic Press, Inc., London and New York, 1992, 32pp. [54] A. Plumtree, L.D. Pawlus, Basic Questions in Fatigue, in: J.T. Fong, R.J. Fields (Eds.), vol. 1, ASTM STP 924, American Society for Testing and Materials, Philadelphia, 1988, pp. 81 – 97. [55] P.H. Shingu, K.N. Ishihara, A. Otsuki, Mater. Sci. Forum 179 – 181 (1995) 5 – 10. [56] R.S. Mackay, J.D. Meiss, I.C. Percival, Physica D13 (1984) 55. [57] T. Tanaka, M. Kosugi, Basic Questions in Fatigue, in: J.T. Fong, R.J. Fields (Eds.), vol. 1, ASTM STP 924, American Society for Testing and Materials, Philadelphia, 1988, pp. 98 – 119. [58] J. He, G. Han, S. Fukuyama, K. Yokogawa, Mater. Sci. Technol. 15 (1999) 909 – 920. [59] C.N.J. Wagner, Acta Mater. 5 (1957) 427. [60] J.B. Cohen, C.N.J. Wagner, J Appl. Phys. 33 (1962) 2073 – 2077. [61] B.E. Warren, X-ray Diffraction, Addison-Wesley, London, 1969, 275pp. [62] L.H. Schwartz, J.B. Cohen, Diffraction from Materials, Academic Press, New York, 1977, 362pp. [63] G.W. Nieman, J.R. Weertman, R.W. Siegel, Scr. Met. Mater. 24 (1990) 145 – 150. [64] G.W. Nieman, J.R. Weertman, R.W. Siegel, J. Mater. Res. 6 (1991) 1012 – 1027. [65] G.J. Thomas, R.W. Siegel, J.A. Eastman, Scr. Metall. Mater. 24 (1990) 201 – 206. [66] W. Wunderlich, I. Ishida, R. Maurer, Scr. Metall. Mater. 24 (1990) 403 – 408. [67] P.G. Sanders, M. Rittner, E. Kiedaisch, J.R. Weertman, H. Kung, Y.C. Lu, Nanostruct. Mater. 9 (1997) 433 – 440. [68] P.G. Sanders, A.B. Witney, J.R. Weertman, R.Z. Valiev, R.W. Siegel, Mater. Sci. Eng. A204 (1995) 7 – 11. [69] X.D. Liu, M. Nagumo, M. Umemoto, Mater. Trans. JIM 38 (1997) 1033 – 1039. [70] C. Suryanarayana, D. Mukhopadhyay, S.N. Patankar, F.H. Froes, J. Mater. Res. 7 (1992) 2114 – 2117. [71] D.A. Konstantinidis, E.C. Aifantis, Nanostruct. Mater. 10 (1998) 1111 – 1118. [72] D.L. Houck, in: Metals Handbook, nineth ed., ASM, Metals Park, Ohio, 1985, pp. 73 – 78. [73] D.L. Houck, Metal Powder Report, Dec. (1985) 707 –709. [74] E. Lugscheider, M. Loch, H.G. Suk, Proceedings of the international Thermal Spray Conference and Exposition, Orlando, Florida, May 28 – June 5, 1992, pp. 555 – 559.

318

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319

[75] S. Eidelman, X. Yang, Nanostruct. Mater. 79 (1997) 9. [76] V.V. Sobolev, M. Guilemany, Int. Mat. Rev. 41 (1969) 13. [77] O. Knotek, U. Schnaut, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 811 – 816. [78] D.W. Parker, G.L. Kutner, Adv. Mat. Process 139 (1991) 68. [79] C.M. Hackett, G.S. Settles, Thermal Spray: Practical Solutions for Engineering Problems, in: C.C. Berndt (Ed.), ASM International, Materials Park, Ohio, 1996, 665pp. [80] D.J. Varacalle, M.G. Ortiz, C.S. Miller, I.J. Steeper, A.J. Rotolico, J. Nerz, W.L. Riggs, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 181 – 187. [81] D. Apelian, D. Wei, B. Farouk, Metall. Trans. 20B (1989) 251. [82] C. Moreau, P. Cielo, M. Lamontagne, S. Dallaire, J.C. Krapez, M. Vardelle, Surf. Coat. Technol. 46 (1991) 173. [83] H. Edris, D.G. McCartney, A.J. Strgeon, J. Mater. Sci. 32 (1997) 863 – 872. [84] A.H. Dent, A.J. Horlock, D.G. McCartney, S.J. Harris, Surf. Coat. Technol. (2002) in press. [85] P. Vuoristo, K. Niemi, T. Mantyla, L.M. Berger, M. Nebelung, in: C.C. Berndt, S. Sampath (Eds.), Thermal Spray Science and Technology, ASM International, Materials Park, Ohio, 1995, pp. 681 – 686. [86] L. Russo, M. Dorfmann, in: A. Ohmori (Ed.), Thermal Spraying: Current Status and Future Trends, High Temperature Society of Japan, Osaka, Japan, 1995, pp. 681 –686. [87] J.M. Guilemany, J.A. Calero, in: C.C. Berndt (Ed.), Thermal Spray: A United Forum for Scientific and Technological Advances, ASM International, Materials Park, Ohio, 1997, pp. 717 –721. [88] J.D. Reardon, R. Mignogna, F.N. Longo, Thin Solid Films 83 (1981) 345 – 351. [89] D.L. Houck, R.F. Chency, Thin Solid Films 118 (1984) 507 – 513. [90] M. Sasaki, F. Kawakami, C. Komaki, M. Ishida, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 165 – 170. [91] P. Vuoristo, K. Niemi, A. Makela, T. Mantyla, in: C.C. Berndt, S. Sampath (Eds.), Thermal Spray Industrial Applications, ASM International, Materials Park, Ohio, 1994, pp. 121 – 126. [92] K. Niemi, P. Vuoristo, T. Mantyla, G. Barbezat, A.R. Nicoll, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 685 –689. [93] D.C. Crawmer, J.D. Krebsbach, W.L. Riggs, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 127 – 136. [94] Y. Fukuda, M. Kuman, in: A. Ohmori (Ed.), Thermal Spraying: Current Status and Future Trends, High Temperature Society of Japan, Osaka, Japan, 1995, pp. 107 –111. [95] J. He, E.J. Lavernia, Mater. Sci. Eng. A301 (2001) 69 –79. [96] Karimi C. Verdon, G. Barbezat, Surf. Coat. Technol. 57 (1993) 13 – 23. [97] V. Ramnath, N. Jayaraman, Mater. Sci. Technol. 5 (1989) 382 –388. [98] H.J. Kim, Y.G. Kweon, R.W. Chang, J. Therm. Spray. Technol. 3 (1994) 169 – 178. [99] J.E. Nerz, B.A. Kushner, A.J. Rotolico, Protective Coatings: Processing and Characterization, in: R.M. Yazici (Ed.), The Minerals, Metals and Materials Society, Warrendale, PA, 1969, pp. 133 – 143. [100] T.A. Mantyla, K.J. Niemi, P.M.J. Vuoristo, G. Barbezat, A.R. Nicoll, Proceedings of Second Plasma-Technik Symposium, in: S. Blum-Sandmeier, H. Eschnauer, P. Huber, A. Nicoll, (Eds.),

[101] [102]

[103]

[104] [105]

[106] [107]

[108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118]

[119] [120] [121] [122]

[123]

[124]

[125] [126] [127]

vol. 1, Sulzer Metco AG, Wohlen, Switzerland, 1991, pp. 287 – 297. G. Barbezat, A.R. Nicoll, A. Sickinger, Wear 162 – 164 (1993) 529 – 537. T.P. Slavin, J. Nerz, in: T.F. Bernecki (Ed.), Thermal Spray Research and Applications, ASM International, Materials Park, Ohio, 1991, pp. 159 – 165. S. Usmani, S. Sampath, H. Herman, Thermal Spray Processing of Nanoscale Materials —A Conference Report with Extended Abstracts, in: C.C. Berndt, E.J. Lavernia, (Eds.), J. Thermal Spray Technol.7 (1998) 429 – 431. D.A. Stewart, P.H. Shipway, D.G. McCartney, Wear 225 –229 (1999) 789 – 798. P. Vuoristo, K. Niemi, A Makela, T. Mantyla, in: C.C. Berndt, T.F. Bernecki (Eds.), Thermal Spray: Research, Design and Applications, ASM International, Materials Park, Ohio, 1993, pp. 173 – 178. S.F. Wayne, S. Sampath, J. Therm. Spray Technol. 1 (1992) 307 – 315. P.R. Strutt, Thermal Spray Processing of Nanoscale Materials —A Conference Report with Extended Abstracts, in: C.C. Berndt, E.J. Lavernia (Eds.), J. Thermal Spray Technol. 7 (1998) 413 – 415. D.A. Stewart, P.H. Shipway, D.G. McCartney, Acta Mater. 48 (2000) 1593 – 1604. J.H. Yang, Cemented Carbide 2 (1984) 17. V.I. Kudryavteva, I.N. Chaporova, Sov. Powder Metall. Metal Ceramics 26 (1987) 396. J. He, E.J. Lavernia, J. Mater. Res. 16 (2001) 478 – 488. K. Jia, T.E. Fischer, B. Gallois, Nanostruct. Mater. 10 (1998) 875 – 891. J.E. Nerz, B.A. Kushner, A.J. Rotolico, J. Therm. Spray. Technol. 1 (1992) 147 – 152. A. Karimi, Ch. Verdon, G. Barbezat, Surf. Coat. Technol. 57 (1993) 81 – 89. H.L. de Villiers Lovelock, J. Therm. Spray. Technol. 7 (1998) 357 – 373. J. He, Y. Liu, Y. Qiao, T.E. Fischer, E.J. Lavernia, Metall. Mater. Trans. A 33A (2002) 145 – 157. C.J. Li, A. Ohmori, Y. Harada, J. Mater. Sci. 31 (1996) 785 – 794. M.S.A. Khan, T.W. Clyne, in: C.C. Berndt (Ed.), Thermal Spray: Practical Solutions for Engineering Problems, ASM International, Materials Park, Ohio, 1996, pp. 113 – 122. P. Aroto, L. Bartha, R. Porat, S. Berger, A. Rosen, Nanostruct. Mater. 10 (1998) 245 – 255. J. Subrahmanyan, M.P. Srivastava, R. Sivakumar, Mat. Sci. Eng. A84 (1986) 209. I. Jenkins, J.V. Wood, Powder Metallurgy: An Overview, The Institute of Metals, London, 1991, 320pp. Z. Fang, J.W. Eason, Proceedings of the 13th International Plansee Seminar, Study of Nanostructured WC – Co Composites, Reutte, Austria, 1993, 625pp. R. Porat, S. Berger, A. Rosen, in: D.L. Bourell (Ed.), Synthesis and Processing of Nanocrystalline Powder, TMS, Warrendale, PA, 1992, p. 217. R. Porat, S. Berger, A. Rosen, in: G. Kneringer, P. Rodhammer, P. Wilhartitz (Eds.), Proceedings of the 14th International Plansee Seminar, Reutte, Austria, 1997, 582pp. R. Porat, S. Berger, A. Rosen, Nanostruct. Mater. 7 (1996) 429 – 436. G.R. Goren-Muginstein, S. Berger, A. Rosen, Nanostruct. Mater. 10 (1998) 795 – 804. L.M. Berger, in: C.C. Berndt (Ed.), Thermal Spray: International Advances in Coatings Technology, ASM International, Materials Park, Ohio, 1992, pp. 253 – 258.

J. He, J.M. Schoenung / Materials Science and Engineering A336 (2002) 274–319 [128] D. Cheng, Q. Xu, G. Trapaga, E.J. Lavernia, Metor. Trans. B 32 (2001) 525 – 535. [129] S. Eidelman, X. Yang, J. Therm. Spray Technol. 5 (1996) 175 – 184. [130] Y. Liu, Y. Qiao, J. He, E.J. Lavernia, T.E. Fischer, Metall. Mater. Trans. A 33 (2002) 159 –164. [131] K. Jia, T.E. Fischer, Wear 200 (1996) 206 –214. [132] A. Wilm, Metallurgie 8 (1911) 225. [133] P.D. Merica, R.G. Waltenberg, H. Scott, Trans. AIME 64 (1920) 41. [134] A. Kelly, R.B. Nicholson, Prog. Mater. Sci. 10 (1963) 149. [135] N.F. Mott, F.R.N. Nabarro, Proc. Phys. Soc. 52 (1940) 86. [136] E. Orowan, Symposium on Internal Stress in Metals and Alloys, Session III Discussion, Institute of Metals, London, England, 1948, 451pp. [137] A.J. Ardell, Metall. Trans. A16 (1985) 2131. [138] S.Y. Hwang, B.G. Seong, in: C.C. Berndt, T.F. Bernecki (Eds.), Thermal Spray Coatings: Research, Design and Applications, ASM International, Materials Park, Ohio, 1993, p. 587. [139] G.Y. Lai, Thin Solid Films 53 (1978) 343 –351. [140] G.Y. Lai, Thin Solid Films 64 (1979) 271 –280.

319

[141] D.F. Pauloins, J.M. Oblak, D.S. Duvall, Trans. Am. Soc. Metals 62 (1969) 611 – 622. [142] I. Kirman, D.H. Warrington, Metall. Trans. 1 (1970) 2667 – 2675. [143] R. Cozar, A. Pineau, Metall. Trans. 4 (1973) 47. [144] Y. Han, P. Deb, M.C. Chaturvedi, Mater. Sci. 16 (1982) 555 – 561. [145] M. Sundararaman, P. Mukhopadhyay, S. Banerjee, Acta Metall. Mater. 36 (1988) 847 – 864. [146] J. He, X.Y. Tang, S. Fukuyama, K. Yokogawa, Acta Metall. Mater. 43 (1995) 4403 – 4409. [147] J. He, G. Han, S. Fukuyama, K. Yokogawa, Acta Metall. Mater. 46 (1998) 215 – 223. [148] P. Haasen, Physical Metallurgy, third ed., Cambridge University Press, 1996, 370pp. [149] G. Bank, T. Schmit, P. Ettmayer, B. Lux, in: T.B. Massalski (Ed.), Binary Alloy Phase Diagrams, second ed., ASM International, Materials Park, OH, 1990, pp. 1304 – 1305. [150] J.A. Dean (Ed.), Lange’s Handbook of Chemistry, thirteenth ed., McGraw-Hill, New York, 1985, pp. 9 – 19. [151] A. Inoue, Acta Mater. 48 (2000) 279 – 306.