Flake Particle Synthesis from Ductile Metal Particles Using a Novel High-speed Vibrator y Mill

A novel particle processor was designed and built for the production of flake-shaped powders. 300 μ m magnesium and 140 μ m iron particles processed for 1 and 2 minutes were analyzed for dimensional, ductile, and morphological characteristics. Particle diameter distributions tended to broaden towards higher size ranges after 2 minutes of processing; the mean particle size was in the range of 400 μ m for magnesium and 300 μ m for iron. The flake thickness decreased over time, leading to a mean-thickness of 12 μ m for the magnesium particle sample processed with 2 x 6.0 mm milling ball media after 2 minutes. The effect of particle medium showed that the milling operation had greater influences on the more ductile material. Surface morphology also became smoother as the milling time increased. Larger ball media tended to produce samples with larger particle sizes with wider size distributions, while smaller ball media produced smaller particles with narrower size distributions. Loading weights also tended to have similar trends. The novel process was demonstrated as an effective and efficient method for the production of flake-shaped metal particles which greatly reduced the amount of milling time and energy required for flake particle production.


Introduction
A flake-shaped particle is a particle with a high diameter-to-thickness aspect ratio. Needs for flake particles var y across a variety of applications such as: pigments and inks, electrochemical electrodes, fuel cell hydrogen storage devices, explosives, lightweight concretes and obscurences [1][2][3][4][5][6] . Properties of flake-shaped particles have growing interests in the chemical and material industries, which continuously demand for increased efficiency in their processes.
Flake particles are desirable from several viewpoints. Due to its high aspect ratio, a flake particle has a larger specific surface area than a spherical particle of the same volume, which can enhance chemical reactivity 5,6) . High aspect ratio metal flakes, such as those made of aluminum, have been used to increase optical obscurence characteristics so as to mimic a metallic look when used as pigment in paint 7) . Additionally the use of flake powders for metal-hydride electrodes in modern batteries has shown significant improvement in electrical capacity 4,8) . Processes for the production of flake-shaped particles are relatively new and have yet to be extensively studied and mastered. Conventional methods for flake particle synthesis are attritor milling, vibratory ball milling, ball milling and wet milling 1,2,9) .
Attritor mills generally refer to the use of a stirring rod and pin to agitate a solution of grinding media and the target particles. During such agitation, the random collisions of balls statistically occur with the target particle in between, eventually flattening the particle after successive impactions 9) . This method, however, can take at least 5 hours to produce consistent flake particles, wasting a large amount of energy to fluid friction and fruitless collisions, and also requires milling to occur in a liquid solution, which can react with the target particle 9) . A vibratory mill operates under similar circumstances as that of the previously mentioned attritor mill; however, it produces a high frequency of randomly colliding ball media by the means of large vibrating forces, and it does not require a liquid milling solution 12) . Though this method allows for the highest kinetic transfer of those mentioned, it is also bounded by long operating times (up to 8 hours; e.g. SPEX model 8000 12) ). Ball milling methods typically require lubricants and anticoagulants such as stearic or oleic acid and/or mineral spirits to control the shape and quality of the product flake 1,2) . These reagents are incompatible when reactive metals such as Al, Fe, Mg, and Cu must remain in their elemental form and can possibly react with the reagents during milling, especially when high energy transfers are present. Wet milling methods such as basket mills are not compatible for similar reasons. Furthermore, these conventional methods are very time-consuming, ranging in processing time from 0.5 to 60 hours 1-3, 8, 9) . As an example, Hong and Kim 2) reported that processing scrap aluminum particles in a wet ball mill system with oleic acid requires 30 hours to produce micrometer thickness flake particles.
The objective of this study was to develop a novel high-speed vibratory mill process for the production of flake particles so as to significantly reduce the milling time required to achieve micrometer thickness without the need of subsidiary milling aides. During milling operations, a cycle of particle fragmentation and reformation was observed. The effects of plastic deformation, particle-to-particle welding and flake fragmentation of this cycle on particle size distribution were examined. Several operating parameters of the milling process were studied, including milling time, particle medium (i.e. Fe, Mg), initial particle size, loading weight and ball media size.

High Speed Orbiting of Ball Media
The milling operation occurs within a closed milling tube, where spherical par ticles are subjected to strong compression and shear forces produced by ball media that flatten the particles. The motion of ball media in the milling tube is illustrated in Fig. 1(a). Compression and shear forces are pro-duced by ball media that roll along the smooth inner wall of the milling tube at high speeds (several thousand rotations per minute) forming sustained orbiting regions. Fig. 1(b) is a photograph demonstrating the orbits created by the rolling ball media that are visible as long circular blurs functioning inside a glass test-tube.
Oscillations in the xz-plane of the milling tube produce the impetus for the media motion. Fig. 2 illustrates the frame-by-frame movements of the milling tube and the resulting movement of the containing media.
The motion of the milling tube is indicated as going in the direction from the gray position to the black position. The oscillations of the milling tube from position 1 2 3 4 repeated, cause an acceleration with a continuously changing direction away from the milling tube center -indicated as a gray arrow. In turn this causes the internal ball media to naturally find the opposite-most point from the direction of milling tube movement. This point is indicated by the position of the dark and filled ball media in Fig. 2. With a continuous change in the direction of the milling tube acceleration, the position of the ball media likewise changes continuously. Ultimately this motion causes the internal media to roll along the inside edges of the milling tube at the same frequency as the milling tube oscillation.

Plastic Deformation and Welding Effects
In the context of the presented novel milling system, plastic deformations are shape changes occurring to the initial particle sample during milling due to the high compression and shear forces from the orbiting ball media. Deformation occurs primarily in a way that results in very thin and wide flakes. Furthermore, progressive deformation can lead to fragmentation of flake particles, resulting in many smaller flakes. Evidence of this can be seen as an increase in the fraction of smaller particles.
Due to the high frequency of orbiting media and chaotic swirling motion of particles inside the milling tube, random flake overlaps readily occur. Composite welding occurs as various flakes are joined to one another by random overlappings resulting in a single compound flake. Instances of compound flake-to-flake welding have been reported for various types of milling devices 4,[7][8][9][10] . Since there are both large and very small particles accumulating, the resulting compound flake may be large or not so large, leading to a wider distribution of sizes for welding dominant systems. However, a compound flake must always be greater in size than its constituents, resulting in a shift of the particle size distribution towards larger particle size ranges.

Experimental System
For this research, a lab scale version of the novel vibrator y mill was built. A milling tube was constructed using 10 mm internal diameter aluminum tubing with a length of 50 mm. Milling tubes were mounted via plastic collars perpendicular to a highspeed vibrator, which provided 13,000 oscillations per minute (OPM) at 120 V. Chromium-Steel Cr-52100 balls (Norstone Inc.) with diameters of 2.0 4.0 and 6.0 mm were used as ball media in the milling tube. The area between the first orbiting ball and the last orbiting ball is defined as the milling zone. Whereas, the actual area which will receive the milling forces is defined as the milling area and is the sum of the ball diameter projections on the milling tube, disregarding the inter-orbit gaps, which are considered ineffectual. This method of classification is illustrated in Fig. 3. The milling area was fixed as being equal to a total of 12 mm for all the balls in the milling tubesuch that the number of balls in each tube would be 2, 3 and 6 for the 6.0 mm, 4.0 mm and 2.0 mm sizes, respectively. This arrangement was chosen with the aim in mind to make the effective milling area approximately equal for all samples so as to replicate the effective milling area for all ball sizes. Two types of particles were tested: 300 -m magnesium (Alfa Aesar) and 140 -m iron (Fisher Sci.). Each milling tube was sealed under ambient atmosphere with plastic end caps to contain the ball media and particles. Particles were weighed and loaded into each milling tube with the ball media in the tube. After processing, the products were separated from the ball media and then stored in labeled vials for analysis. Table 1 displays the parameters tested for each sample.

Product Characterization
Three basic characteristics of the resulting flake par ticles were analyzed, including mean par ticle diameter, mean particle thickness and morphology. Flake particle samples were evenly distributed on glass substrates and observed under an optical microscope (Olympus BX-60) using Spot Advanced (Diagnostic Instrument, INC) image capture software to acquire sample images. The sample images were then processed with Image Pro Plus (Media Cybernetic, L.P) to calculate particle number and dimen-sional statistics for each sample photo batch. Martin diameter, which is defined as the length of the line parallel to a given axis that divides the particle into two equal area regions, was measured. Approximately 4000 particles were analyzed per sample. To determine the thicknesses, flake particles were suspended and solidified in an epoxy resin. The dried epoxy was then cut transversally to obtain cross sectional slices of flakes suspended in the epoxy and subsequently examined under optical microscope. Optical microscopy was also used to study the surface morphology of bare flake particles for evidence of increased surface cracks and multiple-flake layering from flake-toflake welding.
Because of the nature of the milling process, and provided the understanding that the size distribution of particles is dependant on the fragmentation and welding qualities of the milling parameters, its distribution does not necessarily resemble a Gaussian curve. In this study, mean particle size and standard deviation were evaluated side-by-side, which can effectively signify in what regime (fragmentation/welding dominant) the sample distribution reflects.

300 m Mg Particle
The original par ticle size distributions for 300 -m Mg are displayed in Fig. 4(a). The ef fects of increased milling time on particle size distribution (Martin diameter) and thickness of 300 m Mg particles were studied. The results for samples 1 and 2, after milling for 1 and 2 minutes respectively, are displayed in Fig. 5(a) and (b).
The resulting mean diameters for samples 1 and 2 were 372 and 442 m and the standard distributions were 195 and 386 respectively. As shown in Figs. 5  (a) and (b), the majority of the sample remains near 400 m after both 1 and 2 minutes of milling. The peak fraction in the mean size range decreased from 1.2 in sample 1 to 0.8 in sample 2, while fractions of larger particles increased. This shows that a significant amount of the particles in the mean size range were redistributed into larger particle sizes and the size distribution became broader with increased milling time. This growth and broadening in particle size distribution can be explained by plastic deformation and particle welding as described earlier. The lack of change in the smaller particle size regime implies that the particle fragmentation was not as important as welding, and that there does not seem to be a change in the dominant mechanisms between 1 and 2 minutes. It can therefore be concluded that, since the mean diameter is large and increasing and that the standard deviation is also broadening, plastic deformation and composite particle welding are the dominant mechanisms over particle fragmentation in these samples.
The change in flake thickness was analyzed using optical microscope images of epoxy resin slices containing the flake particles. Fig. 6 (a) displays the change in thicknesses in samples 1 and 2. With increased milling time, it can be seen that the flake thickness distribution decreases, similar to the flake diameter trend. The mean particle thickness decreased from 300 m of unprocessed particles to approximately 35 m after only 1 minute. Within just 2 minutes of milling, there was an even greater decrease in flake thickness, with an average flake thickness of 12 m. Plastic deformation is responsible for this trend, since with the start of milling, flakes are continually molded thinner. It should be pointed out that due to the nature of the milling process, not all particles are milled simultaneously. During the early periods of milling, some are fully milled while others are yet to be milled. As milling time increases, the fraction of un-milled par ticles decreases, and the amount of milled particles arrives at the totaltherefore reducing the sample discrepancy caused by the remaining un-milled particles. This explains the decrease in the flake thickness distribution in Fig. 6 (a). In short, the fraction of the thicker and un-milled particles greatly decreases with processing time. Figs. 6 (b) and (c) are the microscope images of representative flake cross-sections from samples 1 and 2 respectively. The results of samples 1 and 2 demonstrate the novel high-speed vibratory milling as a very effective and efficient process for the production of flake-shaped particles.

140μm Fe Particle
Experiments were also carried out for Fe particles. The original particle size distributions for 140 -m Fe particles are displayed in Fig. 4 (b) and the results for samples 3 and 4 are displayed in Fig. 7. The resulting mean diameters for samples 3 and 4 were 238 and 241 m and their standard deviations 63.4 and 89.4 respectively.
It can be seen from Figs. 7 (a) and (b) that the majority of the particle size resides in the mean range of 240 m. Similar to Figs. 5 (a) and (b), the peak fraction decreased and larger particles developed, resulting in a wider particle size distribution. Similar to samples 1 and 2 of magnesium particles, the most possible explanation for the broadening particle size distribution is the effect of plastic deformation and particle welding -consequences of the high compression forces imposed on the flake particles inside the milling tube. In the smaller particle size range, there is little change in particle size distribution below the mean particle size. This further affirms that samples milled up to 2 minutes do not show significant signs of particle fragmentation dominance. Rather, because of the particle size distribution shift in the larger range, welding is the dominant mechanism that influences the particle size distribution in samples 3 and 4.
Comparing the results of samples 1, 2 and 3, 4 can reveal the effects of particle material. Ductility is the primar y factor that determines the sample's dimensional results. Ductile and malleable materials (those with a low Young's Modulus; Mg -45 MPa, Fe -211 MPa) tend to have larger particle size distribution changes over time, as they are more vulnerable to the compression and shear forces present during milling.
In addition to size characteristics, surface qualities of the flake particles were obser ved under optical microscope. Fig. 8 shows the acquired images of the Fe flakes. Two main characteristics of the flakes were examined: surface morphology and evidence of flake-to-flake welding events. Comparing Figs. 8 (a)  and (b), it can be seen that the surface of these iron particles became noticeably smoother with increased time.
Additionally, evidence of flake-to-flake welding can be seen (circled in both Figs. 8 (a) and (b)). The flake pieces circled are likely to be in an intermediate phase of the entire process; a loose flake is first taken up by another and then eventually milled until the two compress into each other to the point that they are indistinguishable.

Effect of Ball Size
To investigate the effect of ball size on the milling operation, in regards to mean particle size and particle size distribution, two additional samples were prepared with smaller ball sizes than that of sample 2. Samples 2, 5, and 6 can be compared for the effect of the milling ball size, with all other parameters kept constant. Their particle size distributions are graphed in Figs. 5 (c) and (d). It can be seen clearly that a simple trend exists between the resulting mean particle size and the ball media size: as the ball size is reduced, the mean particle size reduces (442, 172 and 84 m for 6, 4 and 2 mm ball media, respec-  tively). This trend can be explained by the larger ball sizes creating higher compression forces due to their greater mass, resulting in thinner and wider particles.
A more detailed inspection of the par ticle size distributions gives clues in regards to the welding and fragmentation characteristics of each sample. As mentioned earlier, fragmentation of initial flakes, by its own nature, will result in greater numbers of smaller flakes. Conversely, welding will result in generally larger flake particles, and depending on the extent and number of weldings, can have a broad range of particle sizes. It can be seen in Figs. 5 (c) and (d) that there is a much higher fraction of smaller particles present in sample 6 (2.0 mm balls) than in sample 5 (4.0 mm balls). In other words, there is a broader midrange distribution and a lesser fraction of small particles in sample 5 (4.0 mm balls) than in sample 6 (2.0 mm balls). This indicates that smaller ball media tend to show a dominance of fragmentation over welding, while larger ball sizes tend to have a more profound welding impact, resulting in a broader distribution of particles. This vision is also supported by the standard deviation values for samples 5 and 6. The low welding qualities of sample 6 can also be resorted to the low compression forces present in smaller media, which may not be sufficient to join two flake particles. Since fragmentation is readily present even in smaller ball sizes, all milling processes that use larger ball media have fragmentation occurrences. They differ only in their welding ability, which so far has been the most influential factor of the samples' particle size distribution.

Effect of Weight Loading
The effect of weight loading of Mg in the milling tube was studied by comparing three contrasting weights. Samples 7 and 8 were 15 mg and 35 mg respectively, and were compared to the similar sample 2 of 25 mg. Their particle size distributions are displayed in Figs. 5 (e) and (f). Comparing the trends of samples 7, 2, and 8, shows the effect of increased loading weight. As the loading weight increases, the size distribution can be seen to shift from smaller to larger particles. The mean diameter data for each sample also supports this observation (55, 442 and 997 m for 15, 25 and 35 mg, respectively). As mentioned earlier, smaller particle sizes, like those present in sample 7, imply a dominance of fragmentation and a deficit of welding instances. Since welding can only occur when particle overlapping and milling coincide, the frequency of welding is influenced by the probability of particle overlapping, which is directly related to the particle concentration. As the loading weight of the sample is increased, the concentration increases and the probability of overlapping increases as well, resulting in a greater amount of particle-toparticle welding and ultimately a shift in the particle size distribution towards a midrange size, with a typically larger standard deviation.

Conclusions
Several aspects of flake characteristics produced by a novel vibrator y mill process were studied, including: flake diameter, thickness, and morphology as a function of processing time, ball media count and weight loading. The results of these data showed similar particle size distribution trends in samples of magnesium and iron. With increased milling time, more particles in the larger size ranges developed. This expanding distribution can be explained largely because of the ef fects of plastic deformation and composite particle welding. The more ductile magnesium particles tended to show faster changes in their particle size distributions, since they were more subjective to the forces that caused the shifting distribution. Flake thickness of magnesium was found to decrease as a function of time, as did the particle thickness distribution. With compression and shear forces caused by the orbiting ball media, the flake particles were continuously milled thinner, eventually to a mean thickness of 12 m after 2 minutes of milling. Additionally, as milling proceeded, the influence of partially milled particles reduced and welding occurrences continued thus lowered the standard deviation of flake thickness with time.
The increasing weight of the milling ball media resulted in higher compression forces, and directly affected the mean particle size and its distribution based on their greater welding effects. Smaller ball media tended to show a dominance of fragmentation over welding, while larger ball sizes tended to have a more profound welding impact, resulting in a broader distribution of particles.
Increasing the concentration of flake par ticles (weight loading) increased the probability of overlapping and therefore resulted in a greater amount of particle-to-particle welding. This led to a shift in the particle size distribution towards a midrange size, with a larger standard deviation.
In summar y, a novel vibrator y mill consisting of ball media orbiting at high revolutions inside a milling tube produced high compression and shear

Dr. Chang-Yu Wu
Dr. Chang-Yu Wu is an Associate Professor in the Department of Environmental Engineering Sciences at University of Florida. He received his BS from Mechanical Engineering at National Taiwan University, both MS and PhD from Environmental Engineering at University of Cincinnati. His teaching and research interests include air pollution control, aerosol processes, nanoparticles and catalysis/photocatalysis. He has more than 40 refereed journal publications, 120 conference presentations, 16 invited lectures and 1 patent. He received 6 awards for his achievement in research and engineering education in the past 5 years.
forces capable of processing 25 mg of 300 m spherical magnesium particles into flakes of mean diameter 442μm and thickness of 12μm in only 2 minutes. As demonstrated, the novel high-speed vibrator y mill process is a very efficient and effective process for the production of flake-shaped particles from ductile metal particles.