A pilot plant reactor recently has been constructed for continuously coating fine powders using chemical vapor deposition (CVD) technology. The reactor is capable of coating particles as small as 0.5um in diameter with aspect ratios up to 100, using any of over 100 different elements and compounds. Initial product development efforts are focused on coating tungsten carbide powders with cobalt. This new facility – the largest operation of its kind – is capable of producing up to 100 tons per year of coated particulates. This capability is designed to bring engineered coated powder materials into a cost range acceptable for large-scale commercial applications.

Improved materials and processes can be obtained if fine powders are produced with controlled surface properties or if multiphase powders are produced with controlled phase distribution, such as through coating by CVD. Fluidized beds are an ideal tool for producing or modifying powder materials, allowing significant improvements in material properties and processing economies to be achieved. Unfortunately the most desirable powders for producing engineering materials are of such fine size (<30um) that they are very cohesive and very difficult to handle, fluidize, and process in a non-agglomerated form. Some fine powders can be fluidized by agglomerate formation, and several fine dielectric powders have been fluidized in this manner. However applying coatings or surface modifications at elevated temperatures can form hard agglomerates, an undesirable trait [1].

Ultramet discovered that operating a fluidized bed in the fast-transport regime (above transport velocity), or in the turbulent fluidization regime enables fine particles including fine powders and whiskers, chopped fibers, and such, to be fluidized with high product yields [2,3]. A recirculating fast-fluidized bed pilot plant reactor has been designed, constructed, and put into operation. Manufacturing processes are being developed to apply CVD coatings to fine powders using this system [4]. This technology will enable improved starting materials to be produced for plasma spray, powder metallurgy, ceramic, and composite materials, as well as for improved conductive fillers, reinforcements, and electronic materials.

Key improvements made possible through using coated powders include:

	Control over grain boundary composition
	Non-segregation and improved product uniformity
	100% dispersion and mixing
	Independent control over physical, mechanical, and electrical properties
	Ability to use high rate-forming techniques (i.e., powder injection molding, centrifugal slurry casting) with multiphase powders
	Reduced handling and safety concerns
	Reduced number of processing steps
	Consistent, uniform, reliable product

Designing and sizing the pilot plant reactor was based on requirements for obtaining market penetration with coated powders, which have been identified as a perceived high value added to the coated powder product, at an equivalent or nominally higher cost to the consumer. To meet these requirements, the following conditions were designed into the pilot plant reactor:

	The reactor is of sufficient capacity that only relatively low costs are added in the powder coating step.
	Reduced processing costs (attained through eliminating powder blending, sizing and mixing, using lower sintering temperature, and through more rapid consolidation) offset increased processing costs associated with powder coating.
	The reactor produces a consistent, uniform, reliable product with low levels of hard agglomerates and other inclusions or defects.
	The reactor provides a coated powder source with adequate capacity and rapid response and delivery, suitable for prototyping and producing coated powder-derived products.
	A high perceived value added is delivered to the customer (as in the case of cobalt coated tungsten carbide, with its increase strength and wear resistance and reduced handling and processing costs.)

Previously, coated powders were only available in sizes that could be coated in standard fluidized beds, and at add costs typically exceeding $500-$5,000 per kilogram of processed material. By using the recirculating fast-fluidized bed, fluidizing and coating powders as small as 0.5um with aspect ratios as high as 100 has been demonstrated [2,3]. Cost and yield estimates demonstrate a one hundredfold cost reduction at current capacity, and a five hundredfold cost reduction upon further scale-up to production capacity. The fast-fluidized bed pilot plant reactor will allow sizable quantities (25-250 kg per lot) of coated powders to be prepared at low cost (<$200 per kg), with rapid scale-up to production size lots envisioned at a concomitant cost reduction (to <$20 per kg).

The fluidized bed is a highly efficient solid-fluid contacting device well suited for coating particles, enabling numerous benefits to be achieved through more precisely controlling material microstructure and composition. The fluidized bed originally was adapted for coating large particles, a technology derived from the nuclear industry for producing uranium fuel kernels. However, most ceramic and metal powders and reinforcements of interest are in the 0.5-1.5um size range, and many have large aspect ratios. These particles fall into Geldart’s Class C, which are highly cohesive and difficult to fluidize [5]. Additionally, in standard fluidized bed approaches the fines concentration results in considerable elutriation and loss of a substantial percentage of material, due to entrainment and carryover. Previous studies attempting to fluidize these fine particulates typically have relied on first agglomerating the particles, which then can be fluidized but still with substantial elutriation [6,7].

Surface phenomena tend to dominate fine particle properties per their high surface-to-volume ratio, and electrostatic, van der Waals, and surface tension forces often have dominant effects on their properties and handling. Handling these fine particles is extremely difficult, much less modifying their surfaces through coating in a cost-effective manner without agglomeration. High aspect ratios such as those found in whiskers further aggravate the handling problem, and most fine powders are frequently handled in the agglomerated form. Coating or modifying these particles in the agglomerated form leads to hard agglomerates forming, which are virtually impossible to break up during processing.

The various operating regimes in a fluidized bed are a function of superficial gas velocity, and span superficial velocities from the minimum fluidization velocity to above transport velocity. Previously, state-of-the-art for Class C fine particle fluidization utilized a bubbling fluidized bed. These particles also can be fluidized in the turbulent and fast-transport regimes, where high gas shear forces and massive turbulence tend to cancel out the fine particles’ cohesive effects.

Another advantage of operating a fluidized bed in the turbulent regime (which extends from the bubbling regime up to the transport velocity) is that it allows high gas shear forces and particle collision forces to continually break up agglomerates as they form. However, the high superficial gas velocity leads to considerable entrainment and elutriation. Large particles and agglomerates are returned to the fluidized bed, as long as sufficient freeboard height (the column height above the bed surface) is allowed. The freeboard height required for stable operation is set by the equilibrium between agglomeration at the wall and agglomerate break up and fine particle re-entrainment or elutriation, and can be varied to evaluate a powder’s agglomeration characteristics.

As the superficial velocity in a fluidized bed increases from the bubbling regime, the heterogeneous two-phase flow characteristic of a bubbling fluidized bed first peaks and then gradually changes, giving way to a condition of increasing uniformity. This culminates in the turbulent state, in which large bubbles or voids are absent. In the turbulent fluidized bed, the bed’s upper surface is still present, although considerably more diffuse than in a bubbling fluidized bed. This is caused by greater freeboard activity per higher superficial gas velocities. The turbulent regime extends to the transport velocity. As transport velocity is approached, there is a sharp increase in the particle carryover rate; typical fluidized bed solid loading in this operating region is well below 5%. Recycling the entrained solids allows the solid loading to be increased to above the transport velocity, which moves reactor operation into the fast-transport regime.

In the absence of solid recycling, the bed would soon empty. Above the transport velocity, solids fed into the bottom of the column traverse it in fully entrained transport flow, and the resulting suspension’s concentration or density not only depends on the gas velocity but also on the solid flow rate. If the solid flow rate is small, dilute-phase flow results. If instead solids are fed to the column at a sufficiently high rate (such as by circulating solids carried over from the column via external cyclones, filters, and a standpipe), then it is possible to maintain the column at the relatively large solid loading fractions typical of the fast-transport regime. Turbulent and fast-fluidized beds are thus referred to as high-velocity fluidized beds. Adding a cyclone separator at the top of the column allows most of the particles to be returned to the bed (i.e., recirculated). Using cyclones, recirculation efficiencies (defined as mass of particles recirculated per pass divided by total powder mass) of >95% have been achieved, and adding a pulse-cleaned filter has resulted in >99% recirculation efficiencies.

The pilot-scale recirculating fast-fluidized bed is designed to operate continuously in the bubbling regime with larger particles, and in a constantly stirred mode when operated with fine Geldart Class C particles in the turbulent and fast-transport regimes. Good fluidization and stable operation can be achieved with cohesive powders. Excellent control over gas shear forces and volumetric loading provide the agglomeration control and mass contacting efficiency needed to prepare economic quantities of coated powders. This reactor allows coating and surface modification of Class C particles with minimal agglomeration.

The fast-fluidized bed operates by recirculating an inert fluidization gas. which passes through a countercurrent heat exchanger and compressor. The reactor is 20.3cm (8.0in) in diameter, with a total length (height) of just under 4.27m (14.0ft.).

The recirculation systems for both gas and powder flow streams are completely closed except for a balanced stream for each, allowing a small amount of product gases and/or powder product to be removed from the system, as desired. This set-up conserves fluidizing gases, which are circulated at flow velocities ranging from 0.1-2.2 m· s-1 and flow rates up to 0.118 m3· s-1 (250 ft3· min-1) at standard temperature and pressure. Reactant gas flow rates typically are in the 0.000017-0.00017 m3· s-1 (1-10 · min-1) range.

Total active reactor volume is 0.127 m3 (4.5ft. 3), with a standpipe volume of 0.074 m3 (2.6ft. 3). This volume allows for a maximum reactor charge of nearly 800kg (1750 lb.) of tungsten carbide powder when operated in batch mode. The reactor can produce 136-227kg per hour (300-500 pounds per hour) of cobalt coated tungsten powder when operated in continuous mode.

Initial fluidization trials and operational checkout were conducted using fine alumina (Al2O3) and copper powders. These powders (with average particle diameters of 22 and 17um respectively) are both cohesive near the "border" between cohesive and aeratable powders. Copper powder is slightly more cohesive than alumina powder per its finer size, which was evidenced by the higher shear forces (i.e., higher rpm’s) necessary to feed the powder into the reactor. The conductive copper powder’s tendency to agglomerate, however, was less than that of the dielectric alumina powder.

Agglomeration degree can be estimated by comparing the measured minimum fluidization velocity (Uminmeas.) and the bubbling-turbulent transition fluidization velocity (U1 meas.) with predictions from theory (Umin calc. and U1 calc., respectively) and back-calculating an approximate particle size (Dp calc.) from the measured fluidization velocities. Table 1 gives the measured and predicted fluidization velocities as well as calculated effective particle sizes, assuming 65% of theoretical density for agglomerates.

Table 1

Measured and Predicted Fluidization Velocities and Calculated Effective Particle Size for Alumina and Copper Powders Used in Fluidization Trials

Material        Uminmeas.     Umin calc.     U1 meas.     Umin calc.     Dp calc.                                                  (cm· s-1)          (cm· s-1)       (cm· s-1)       (cm· s-1)      (micron)

22 um Al2O3      0.26                  0.06               48                 5.75                46-79

17um Cu           0.21                  0.08               32                  7.80                25-35

Comparing the measured and calculated data, it is apparent that the powders act as agglomerates, with aluminum having a greater tendency to agglomerate than copper (which is to be expected based on their properties, as above noted). In addition, the dielectric alumina powders have a greater tendency to agglomerate at high superficial gas velocities, indicating that electrical charges created by gas shear dominate agglomeration tendencies. Conductive copper powders, however, show much lesser tendency to agglomerate with increasing gas velocity. A rough trend related to bed height or wall surface area for copper powders has been calculated, revealing a different agglomeration mechanism relative to wall effects, in addition to a much smaller agglomerate size.

Unfortunately, no agglomerate size measurement could be derived in the turbulent or fast-transport regimes to confirm reduced agglomeration, as hypothesized. However, photomicrographs revealed no evidence of agglomerates in powder fluidized at gas velocities above 55 cm· s-1 for copper and 85 cm· s-1 for alumina. The greatest degree of agglomeration most likely occurs at the bubbling-turbulent transition, after which the agglomeration state decreases per expanding bed volume and greater particle impact velocities. An equilibrium is implied by these observations, obtained at each velocity/particle loading and between tendencies for agglomeration formation and breakup, which shift toward agglomerate breakup at higher particle velocities and lower particle loadings.

Experiments also were conducted using 5um high aspect ratio silicon carbide whiskers. The minimum observed fluidization velocity was 4.8 cm· s-1, compared to a predicted value of 0.01 cm· s-1. This implies an effective particle size larger than 180um, a very large particle, demonstrating the high cohesivity these high aspect ration whiskers have. While no clear transition to the turbulent regime was observed for the whiskers, the best observation in a small bed was approximately 50 cm· s-1, indicating the effective particle size (agglomeration degree) did indeed decrease with increasing particle velocity and superficial gas velocity for highly cohesive powders.

Fluidized bed CVD powder coating has been viewed as a high cost process because of the R&D nature of coated powder development efforts to date. The operating characteristics and economics of the recirculating fast-fluidized bed pilot plant reactor disprove this commonly held misconception, and demonstrate that a viable production process is available for producing large quantities of coated powder products. At the pilot plant scale (approximately 10,000kg per month for coated powders) the dominant factors in process economics are material cost and overhead. Labor costs for repeat coated powder production are estimated to be $0.44-0.88 per kg ($0.20-0.40 per pound). Raw material costs are expected to range from $11-33 per kg ($5-15 per pound) on a non-production basis and $4.40-11.00 per kg ($2-5 per pound) on a pilot plant production basis. Estimated production costs for 5% cobalt coated 5um tungsten carbide powder are $6.60 per kg ($3 per pound) of added value, or $33 per kg ($15 per pound) including substrate powders cost. This compares to a material cost of $3.75 per kg ($1.70 per pound) for cobalt powder, or $30.20 per kg ($13.70 per pound) for cobalt-tungsten carbide powder produced by conventional blending techniques. At production quantities in excess of roughly 1,000,000 kg per year, estimated production costs are equivalent to raw material costs for blended powders.

1. D. Dunii and O. Levenspiel Fluidization Engineering, John Wiley & Sons, New York, 1966.

2. A.J. Sherman and V.M. Arrieta "Coated Micrograin Carbide Powders for Wear  Resistance," Final Report 9ULT/TR-94-6594) Grant DE-FG03-93ER81584, U.S. Department of Energy, Washington, DC, 1994.

3. A.J. Sherman and V.M. Arrieta "Fluidized Bed Modification of Fine Powders," Final Report (ULT/TR-94-6677), Grant III-9362136, National Science Foundation, Washington, DC, 1994.

4. A.J. Sherman and V.M. Arrieta "Coated Micrograin Carbide Powders for Wear Resistance, Phase II," work in progress under Grant DE-FG03-93ER81584, U.S. Department of Energy, Washington, DC, to be completed in 1996.

5. A.R. Abrahamson and D. Geldart "Behavior of Gas-Fluidized Beds of Fine Powders, Part I: Homogeneous Expansion," Powder Technology 26, 1980.

6. J. Yerushalmi et. al. "Flow Regimes in Vertical Gas-Solid Contact Systems," A.I.Ch.E. Symposium Series 74, 1978.

7. A.A. Avidan and J. Yerushalmi "Bed Expansion in High-Velocity Fluidization, "Powder Technology 32, 1982.