Enhanced Aluminum Properties by Means of Precise Droplet Deposition

Melissa Orme and Robert F. Smith

Department of Mechanical and Aerospace Engineering University of California, Irvine CA 92697-3975

Accepted for Publication in

ASME Journal of Manufacturing Science and Engineering

Abstract

The use of molten aluminum droplets is investigated for potential application to Precision Droplet-Based

Net-Form Manufacturing (PDM). In the proposed application, final structural components are made

from the raw stock in one integrated operation by depositing molten metal droplets, layer after layer, via

computer information. This work investigates the feasibility of the proposed technology by investigating

the issues associated with generating molten aluminum droplets from capillary stream break-up, and

examining the mechanical characteristics of the fabricated aluminum components. New results are

presented which illustrate the generation of stable streams of molten aluminum droplets at rates of

24,000 droplets/second for a droplet stream speed of 10.9 m/s, corresponding to throughput rates of

2.3×10-4 kg/s (1.85 lb./hour). The droplets travel 2,500 droplet diameters in an inert environment

before impingement with the substrate. Microstructural images are completely devoid of splat

boundaries, which have been removed by remelting, and the grain size is approximately uniform

throughout the field of view of the image that, in most cases presented, contains easily upwards of 30

splats. Also, it has been found that the presence of aluminum oxide in the melt does not influence the

2

average grain size of the component. An oxide barrier however will encapsulate each grain if the oxides

are not removed by filtration in the pre-jetting stage. The presence of aluminum oxide in the melt does

not prohibit the removal of the splat boundaries. Mechanical analysis shows that fabrication with molten

aluminum droplet deposition results in a 30% increase in ultimate tensile strength compared to the raw

ingot stock.

1. INTRODUCTION

Rapid prototyping is now an established Time Compression Technology which leads to significant

competitive advantages in the manufacturing sector (Atkinson, 1997). Typical rapid prototyping

technologies such as Stereolithography (Jacobs, 1992) create prototypes necessarily made of

photopolymer resins or thermoplastic materials limiting the prototype to applications which do not

require the strength a fully functional part (e.g., used for applications of geometry and configuration

validation). Structural ceramic components are fabricated by the process of Fused Deposition Ceramics

(FDC) (Argarwala et al., 1995) which is based on the technology of Fused Deposition Modeling

(FDM) to make ceramic components from powder/binder filaments. Selective Laser Sintering (SLS) is

the Rapid Prototyping technology in which prototypes are constructed layer by layer from CAD

information where each layer of thermoplastic, metallic, or ceramic powder is locally heated with a laser

to fuse particles (Klock et al., 1995; Griffin et al., 1995; Pintat et al., 1995). Until recently, all SLS

green prototypes required post-treatment to reduce the large degree of porosity which would otherwise

severely limit their mechanical usefulness. SLS/HIP is a new process under development in which the

SLS processed part is post-processed by containerless Hot Isostatic Pressing to full density (Das et al.

1997).

3

Net-form manufacturing technologies are commonly viewed as a subset of Rapid Prototyping

technologies since the final component is manufactured directly from computer information in one

integrated operation. However, in net-form manufacturing, structural metallic components can be

made without any post-treatment.

This paper describes the generation and deposition of molten aluminum droplets from capillary stream

break-up for application to the emerging technology of Precision Droplet-Based Net-Form

Manufacturing (PDM) (Orme and Muntz, 1992; Orme, 1993), and examines the microstructure and

mechanical properties of the preliminary fabricated components. The technique of PDM could

theoretically employ any metal that can be contained in a crucible in its molten state. However, the work

presented in this paper, which examines the feasibility of the technology, is restricted to the generation

and deposition of pure molten aluminum droplets.

The technique of PDM, which is still under development, uses precisely controlled streams of molten

metal droplets as the deposition element that are characterized by speed dispersions ≤1×10-6 times the

average droplet speed and angular dispersions on the order of one micro-radian (Orme, 1991).

Droplets will be electrostaticly charged and deflected onto a substrate. The 3-D part is fabricated layer

by layer with conditions carefully controlled such that the thermal energy of the incoming droplets is

sufficient to locally remelt the previously delivered and solidified material (substrate), thereby maintaining

a high strength component by removing via phase change the boundaries between neighboring splats

(Orme et al., 1996, 1997; Amon et al. 1996, 1998). Microstructures of the fabricated components are

refined over their conventionally fabricated components (e.g., cast) due to rapid solidification of the

4

droplets. It is well known that rapid solidification leads to refined microstructures, which in turn leads to

enhanced mechanical properties.

Microstructural refinement from droplet deposition has been the subject of interest since at least the

early 1980’s. The literature is abundant with studies focused on the microstructural refinement of

components due to rapid solidification resulting from droplet deposition. A sample of these studies

include that due to Singer (1983), who evaluated a generation of new material characteristics including

refined microstructures with the technology of Spray Forming; Fielder et al. (1987), who examined the

microstructural refinement due to spray forming René 95 components; Lavernia et al. (1992), who

studied the microstructural refinement due to rapid solidification in aluminum alloys fabricated with

droplet spray deposition; Orme (1993), who studied the microstructural refinement of Rose’s metal

components that were fabricated with controlled droplet deposition; Orme et al., (1996), who studied

the microstructural characteristics of aluminum components fabricated with rapidly solidified molten

aluminum droplets; and Amon et al. (1996) who investigated the microstructural characteristics of

stainless steel and carbon steel droplets in their study that was primarily focused on substrate remelting.

Figure 1 is a conceptual schematic of the Precision Droplet-Based Net-Form Manufacturing (PDM)

technique. In this work, molten metal droplets are formed from capillary stream break-up at rates of

24,000 drops per second for a 50 micron radius orifice and for droplets travelling at 10.9 m/s

corresponding to a driving pressure of 20 PSI. Droplet deposition frequency is inversely proportional to

orifice radius so that droplets emanating from a 25 micron radius orifice would be generated at a rate of

48,000 droplets/second. The droplets are ejected into an inert environment and are targeted by means

of electrostatic deflection onto a substrate. Upon arrival at the substrate, they undergo a “splatting”

5

action that, under carefully controlled thermal conditions, entails simultaneous spreading, remelting of the

previously deposited and solidified material, followed by solidification. Movement of the x-y table in

accordance to CAD information coupled with electrostatic charging and deflection allows subsequent

droplet deliveries to build the 3-D part micro-layer by micro-layer without any mold. Since the droplets

rapidly solidify, detailed structures can be fabricated in the absence of macroscopic fluid flow.

PDM bears similarities with other emerging technologies of net-form manufacturing such as Shape

Deposition Manufacturing, SDM, (Prinz et al., 1995; Amon et al., 1996; Amon et al., 1998, Chin et al.,

1995) and 3-D printing (Sachs et al., 1992). In SDM, a feedstock wire located directly over the

substrate is melted using a plasma-welding torch. A discrete droplet of typical dimension 1-10 mm

(depending on plasma conditions) falls off the wire and onto the substrate thereby building the material

component. In 3-D Printing, a green part is created by printing a stream of binder droplets onto a metal

powder bed according to CAD information. However in that process, several post-treatment steps are

necessary to insure a mechanically sound component.

The aim of this paper is not to demonstrate PDM, which is still an immature technology. Rather, this

paper is focused on substantiating the feasibility of the technology by providing new engineering details

of molten aluminum droplet formation and the requisite apparatus, as well as microstructural and

metallurgical analysis of preliminary structures fabricated with the technology.

2. BACKGROUND

The crux of the PDM technique lies in the ability to generate continuous streams of precisely controlled

droplets. Previous research has illustrated the intrinsic lateral stability on the order of a micro-radian for

6

droplet streams composed of oil or molten solder (Orme and Muntz, 1990). It has been shown that the

vertical stability, which is characterized by the speed dispersions of the droplets, can be controlled to a

high degree of accuracy (e.g., ≤ 1 part in 106) when imposing carefully controlled frequency

perturbations on the stream which initiates droplet formation (Orme, 1991). In order to achieve

accurate electrostatic charging and subsequent deflection, the charging waveform must be perfectly

synchronized with the frequency disturbance that drives droplet formation. A droplet stream

characterized by large speed dispersions would result in gross targeting errors since speed dispersions

directly translate to errors in droplet charge acquired. Hence, the droplet generator must be one that

does not introduce additional unwanted harmonics on the capillary stream that would otherwise cause

velocity dispersions or jitter.

Figure 2 illustrates the general phenomenon of droplet generation from capillary stream break-up. An

axisymmetric jet of radius ro issues from an orifice as shown. A periodic disturbance of wavelength λ

resulting from motion of a vibrating rod initiates a radial disturbance on the circumference of the jet. The

motion of the rod is due to that of a piezoelectric crystal that is bonded to the upper end of the rod. If

the disturbance is chosen such that ko*<1.0, where ko

*=2πro/λ, then the disturbance will grow until

droplets of radius rd are pinched from the stream. The droplet radius is related to the undisturbed stream

radius, ro, by conservation of mass:

31

20

31

*0 4

323

=

=

fVr

krr od

π

7

In the above, V is the stream speed and f is the disturbance frequency. Previous work (Orme, 1991)

has shown that a droplet stream is most uniform (i.e., is characterized by the most uniformly separated

and sized droplets) when the capillary stream is perturbed with the ko* which corresponds to the

maximum disturbance growth rate, β, on the capillary stream. For fluids that are nearly inviscid such as

water and most molten metals, this ko*=0.697. Hence, for the conditions employed in this work (initial

stream radius, or , of 50 µm, and a stream speed, V, of 10.9 m/s), the driving frequency used is that

which generates a highly uniform droplet stream which is given by the relation

)2/()( *oo rVkf π= =24,000 Hz, where ko

*=0.692 is used.

The droplets are charged by surrounding the capillary stream with a charge electrode at the location that

includes the droplet formation point. By carefully synchronizing the periodic disturbance initiating

droplet formation with the charging waveform, drops can be charged on a drop-to-drop basis. This

paper is focused on details of molten aluminum droplet formation, and not on droplet charging and

deflection. Previous studies have detailed results from droplet charging and deflection of molten metal

(Orme et al., 1996).

3. EXPERIMENTAL APPARATUS

Previous work has illustrated the stable generation of oil or molten solder droplets from capillary stream

break-up (Orme, 1993; Orme and Muntz, 1990; Orme, 1991; Orme et al., 1996). Although it is a

straightforward extension in droplet formation theory, the generation of molten aluminum droplets from a

100 µm diameter orifice is not a straightforward extension in technology due to its higher melting point

and its corrosive nature in the molten state. These two issues are addressed in the design and

8

fabrication of a state-of-the-art apparatus for molten aluminum droplet generation that is described in

this section.

Figure 3 is a schematic of the existing facility. The main component is an environmental chamber that

sits upon a vibrationally isolated optical bench. An observation cylinder equipped with eight optical

quality glass windows couples the environmental chamber to the droplet generator. Two CCD array

cameras and two light sources aid in monitoring the droplet stream’s position as it flies through the

observation cylinder. The droplet generator is housed in a socket assembly that allows angular

positioning of the droplet stream. The distance between the exit face of the orifice and the copper

substrate is 50 cm. In operation, ambient air is evacuated from each component of the facility, as

oxygen has been found to have a detrimental effect on both droplet formation and the integrity of the

resulting deposit, and an inert gas such as argon or nitrogen is bled into the system. The purity of the

bottled gas is specified at 99.998%, and the oxygen level of gas in the environmental chamber is

monitored to insure that the O2 level does not exceed 25 PPM. A background pressure of 14.7 psi is

used in this work.

Aluminum is severely corrosive in the molten state, and was found to dissolve high grade stainless steel,

steel alloys such as Inconel, and some refractory materials such as molybdenum. It was found that

titanium corroded the least when in contact with the melt. Hence, the main component of the droplet

generator, which consists of a cylindrical cartridge fitted with an orifice on the lower end and a cap on

the upper end, was machined from titanium and coated with boron nitride. Figure 4 illustrates a

conceptual schematic of the titanium cartridge, resistance heater, and plunger rod. Not shown is the

restraining hardware of the piezoelectric crystal. While ceramics such as alumina would also be resistant

9

to corrosion and suitable for use in certain design configurations, the hardware design employed in this

work requires a material to be used which can bear a load in compression, deeming titanium a more

suitable choice.

The vibration that initiates the stream’s radial instability and subsequently produces droplets is

introduced with motion from a piezoelectric crystal (PZT5-H). When jetting low temperature fluids

such as oil, water, or molten solder, the PZT can be immersed directly in the fluid thereby reducing

structural vibrations. However the PZT must be shielded from extreme temperatures such as those

associated with the jetting of molten aluminum. We have therefore designed a vibrating rod system that

is machined from titanium and coated with boron nitride that restricts the transfer of thermal energy to

the PZT (by employing water cooling lines near the PZT) but provides the necessary pressure

fluctuation to the molten metal directly above the orifice. Care has been taken to decouple the motion

of the PZT from the rest of the apparatus thereby reducing the application of unwanted disturbances

such as apparatus resonances.

Natural diamond was chosen as the orifice material since it was assumed to be most resilient to

corrosion. Usage of the diamond orifice results in a stable and uniform droplet stream. However,

examination of the orifice with the use of a Scanning Electron Microscope illustrates that the diamond

was eroded by the molten aluminum. Further energy dispersed X-ray analysis (EDXA) of the diamond

orifice indicated peaks at carbon, aluminum, and oxygen. It is emphasized that the aluminum eroded

away the diamond as aluminum was found within the diamond and not on the outer surface. No traces

of AlC were found. Figure 5a is a SEM photograph illustrating the surface of the diamond orifice that

had been jetted for a two-hour period. The originally smooth circular entrance had become notched at

10

several locations. A different view with increased magnification is illustrated in figure 5b. The region on

the right half of the figure has been exposed to molten aluminum for a two-hour period, while the region

to the left has been exposed for only one hour with the use of a gasket. Clearly, the region that was

exposed for two hours has been significantly eroded on the surface. This is an interesting finding in its

own right since the diamond is contained in an inert environment and maintained at temperatures much

less than that which is known to cause diamond erosion.

4. RESULTS

Droplet streams were generated with pure aluminum, which has a melting temperature of 933K, a liquid

density of 2380 kg/m3, a solid density of 2707 kg/m3, and a latent heat of 402 J/kg. The jetting

temperature is held at 30C above its melting temperature, and the background environment gas is argon

held at a constant 14.7 psi.

4.1 Droplet Generation

In order to achieve a stable droplet stream (characterized by minimal angular and speed dispersions), it

was found necessary to filter the molten aluminum. While it is true that molten aluminum is a highly

corrosive fluid, and therefore dissolves many of its own contaminants thereby acting as a self-filtering

fluid, it was found that aluminum oxides were immune to its self-filtering nature. Hence, even though the

raw stock was advertised to be of highly pure quality (99.99% from Accumet Materials Co of Ossining,

NY), it was found that there was inevitably a significant amount of aluminum oxides in the melt which

11

have proven to have a deleterious effect on stream stability. Our approach was to filter the melt with

custom fabricated graphite filter plates.

Figure 6 illustrates a molten aluminum droplet stream after having traveled a distance of approximately

0.4 meters in an inert environment, which corresponds to approximately 2,500 droplet diameters. The

droplets were driven with a stagnation pressure of 20-psi (137.8 kPa) above ambient which

corresponds to a stream speed of 10.9 m/s. It is the author’s understanding that this is the first report of

the generation of molten aluminum droplet streams from capillary stream break-up and the subsequent

stable propagation to significant distances in an inert environment. Hence, with proper care to insure

that the obstacles of oxidation and materials corrosion are overcome, molten aluminum droplet

generation from capillary stream break-up is achievable and therefore appears to be feasible for the

emerging PDM technology.

4.2 Droplet Deposition

Droplet deposits were made by allowing the droplets to impinge upon a substrate where they solidify.

In this phase of PDM development, droplet charging is not employed so that consecutive droplet

depositions to one location on the substrate build a solid circular cylinder. Cylinders as high as 0.15

meters have been deposited and analyzed as discussed below. Since the droplets are deposited onto

the same location of the substrate every 42 microseconds (1/24,000 Hz), the previously deposited

material has had insufficient time to cool to the original substrate temperature (initially 293K), causing

the temperature of the deposit to increase with time (i.e., with successive splat depositions). This

warming trend of the previously deposited and solidified material causes the solidification time to

12

increase, which has two effects. First, the grain size is expected to be larger than that which would

occur in a rapid solidification process (e.g., the final realization of the PDM process that includes droplet

charging and deflection), and hence the metallurgical properties that vary with grain size will be of a

lower quality. Second, during the initial droplet depositions, the diameter of the circular cylinder

increases as the number of splats increases to a size which is larger than a single splat diameter. Hence,

the diameters of the circular cylinders deposited were found typically to increase from a few millimeters

at the lower end (i.e., at the interface of the first splat with the initial substrate) to approximately 1.0 cm

at a distance from the initial substrate which varies with jetting conditions (such as deposition frequency

and impact temperature), but is typically on the order of 1.0 cm. At heights above 1.0 cm the diameter

is relatively constant. The reason why the cylinder diameter becomes wider than the splat diameter is

due to the fact that as the droplets are deposited onto the same location at high deposition rates, the top

surface of the cylinder has insufficient time to cool before the next droplet is delivered. Hence, the

temperature of the top surface of the cylinder increases with distance away from the substrate (or

number of splat delivered), causing the remelt zone to increase with increasing distance from the

substrate. The relatively thick molten cylinder tip causes the fluid to flow prior to solidification, which in

turn causes cylinder widening. The issue of cylinder widening has no practical consequence for the

development of the PDM technology since in the final realization, the droplets would be electrostaticly

deflected onto the substrate in order to create complicated 3-D geometries, thereby permitting the

droplets to rapidly solidify prior to the arrival of a new splat.

4.3 Deposit Characterization

13

Microstructures

Micrographs of sections of the deposited cylinders are shown in Figures 7-9. Figure 7 is included to

illustrate the variation in grain size at different locations in the cylinder, and figures 8 and 9 are included

to illustrate the effect of melt filtration. All micrographs are taken from vertical sections of various

deposited cylinders so that they are cut through many splat layers. The samples in Figure 7 were etched

at UCI with a modified version of Keller’s Reagent, which consists of 31% H20, 31% HCl, 31% HNO3

and 7% HF. Before applying the etchant, the surfaces were manually polished to a reflecting shine. The

samples were initially polished with a synthetic sapphire polish that had a grain size of 1 µm, and

subsequently polished with Buehler micropolish compound of 0.5 and 0.3µm grain respectively.

Figure 7 illustrates the microstructure of the same component at three locations from the substrate. The

component was fabricated with coarsely filtered molten aluminum. The left image illustrates the

microstructure 1.0 mm from the substrate surface. It can be seen that the grains are on the order of 50

µm. The center photograph is a section of the microstructure at a distance 5.0mm from the substrate,

and the photograph on the right illustrates the microstructure at a distance of 25mm from the substrate.

In these optical photographs, as well as in all observations, no splat boundaries were found at any

location. The retention of individual splat boundaries would lead to a component with poor mechanical

qualities. In the photographs illustrated in Figure 7, the average grain size appears to be on the order of

50 µm for the samples evaluated at 1.0 and 5.0 mm from the substrate. It is seen that the grain size is

somewhat larger at a distance of 25.0 mm from the substrate, though not appreciably. It is worth noting

that the grain size of the cast ingot was on the order of millimeters.

14

The micrographs in Figure 8 provide a dramatic presentation of the effect of filtering the jetting melt.

Both images are at the same magnification, and the total field of view is 160x190µm. The image on the

left is taken from a vertical section of a cylinder that was fabricated with unfiltered molten aluminum, and

that on the right corresponds to a deposited cylinder that was fabricated with filtered molten aluminum.

The structure that was fabricated with unfiltered molten aluminum has identifiable pores and oxide grain

boundary encapsulation. The microstructure of the component that was fabricated with filtered molten

metal is free of visible oxide grain boundaries and pores. The observable speckle on both figures

(which is more prominent in the figure on the right) is believed to be an artifact of etching and are not

pores (as attested by density measurements discussed later).

The images in Figure 9 illustrate expanded views of the images in Figure 8, each with a total view of

800x970µm. Again, individual splat boundaries are not present, and the grain size is approximately

uniform throughout the entire field of view. Pores and oxide grain boundaries are visible in the image

that was fabricated with unfiltered molten aluminum, and are invisible in the image corresponding to the

component fabricated with filtered molten aluminum.

Mechanical Characteristics

Three test coupons were machined from three separate droplet deposited circular cylinders fabricated

with filtered molten aluminum (specimen 1, 2, and 3). An additional coupon was machined from a

circular cylinder that was fabricated with unfiltered molten aluminum (specimen 4). The asterisk (*)

indicates that the measurement was made six months after fabrication. Specimens 1 and 2 were sent to

the Boeing Commercial Airplane Group for evaluation, and specimens 3 and 4 evaluated by Anaheim

15

Test Laboratory. Table 1 below provides a summary of the results of the three specimens in addition to

the results from the raw ingot that was used as the jetting material for comparison purposes.

The average ultimate tensile strength of the droplet-deposited specimens 1, 2 and 3 (all filtered) is 8,900

where the corresponding measured value for the raw stock is 6,816 PSI. Hence, even in the current

immature stage of the PDM technology, a gain of 30% in ultimate strength is achieved. Measurements

of the 0.2% offset yield strength for the same three specimens showed a modest increase over the raw

stock with values of 4,500, 4,900, and 6,415 PSI, compared to a value of 4,010 PSI for the raw

stock. No appreciable change in elongation was observed in the specimens fabricated with filtered

aluminum (specimens 1, 2 and 3), and the raw stock as the average elongation for specimens 1, 2 and 3

is 38.3%, and that for the raw stock is 37%. The density of the droplet deposited specimens (fabricated

from filtered melt, i.e., specimens 1, 2, and 3) and the raw stock had no measurable difference,

indicating very low porosity as attested visually by the photographs in Figures 7, 8, and 9 of the

specimens fabricated with filtered molten aluminum.

Additionally, a sample that was fabricated with unfiltered molten aluminum was evaluated. This sample

was fabricated early in this research endeavor before the importance of filtering with respect to stream

stability was recognized. Using unfiltered molten metal results in orifice plugging in approximately 80%

of the experimental attempts. Then, when a stream is ultimately established, its trajectory is often erratic

with time due to the passage of micro-contamination (which is believed to be aluminum oxide) through

the orifice, thereby reducing the probability of fabricating a constant diameter cylinder suitable for tensile

strength evaluation. Hence, investigation of the effects of filtration on mechanical properties is not a

straightforward task. We have, however, fabricated one sample that was made with unfiltered molten

16

aluminum from which a test coupon was machined for tensile strength evaluation. This sample is listed

as specimen 4 in the table below. The asterisk indicates that the mechanical testing was performed 6

months after the sample was fabricated, whereas the other specimens without asterisks were analyzed

within one week of fabrication, hence a comparison may not be entirely valid since the mechanical

properties may have varied over time. With this in mind, it can be seen that the ultimate tensile strength is

9,461 PSI, the 2% offset yield strength is a comparatively high value of 8,734 PSI, the elongation and

density are measured to be 41% and 2,670 kg/m3 respectively indicating the presence porosity as

attested visually by the photographs in Figures 8 and 9. In order to ascertain if the unexpected increase

in strength for the unfiltered specimen is due to age hardening, we reevaluated one of the filtered

specimens (specimen 2) for strength after a six month waiting period. The results of this test are

included in the table with as entry 2*. It can be seen that there is a modest increase in tensile strength,

and significant increases in 2% offset yield strength and elongation. The density is unchanged as

expected. Hence, one can only conclude that the values of the aforementioned properties tend to

increase with time, and therefore, the cause of the increase in the said mechanical properties in the

sample fabricated with unfiltered metal is due to age hardening rather than the fact that it was made with

unfiltered molten aluminum. A direct comparison between un-aged filtered and unfiltered components is

not possible due to the difficulty in acquiring unfiltered cylinders as discussed above.

Sample Ultimate Tensile Strength

0.2% Offset Yield Strength

Elongation Density

1: filtered 8,800 PSI 4,900 PSI 48% Not available

2: filtered 9,000 PSI 4,500 PSI 34% not available

17

2: filtered* 9,030 PSI 6,260 PSI 42% 2,700 kg/m3

3: filtered 8,901 PSI 6,415 PSI 33% 2,700 kg/m3

4: unfiltered* 9,461 PSI 8,734 PSI 41% 2,670 kg/m3

Ingot Stock 6,816 PSI 4,010 PSI 37% 2,700 kg/m3

Table 1: Summary of Mechanical Properties for molten aluminum droplet deposited samples

(*properties of sample were measured 6 months after fabrication)

It should be emphasized that all of the results provided here are expected to be lower than those of the

final PDM technology in which the droplets are deflected onto different locations of the substrate (not

onto one location, which results in the formation of a cylinder). When the droplets are deflected onto

different locations, they will solidify more rapidly, thereby reducing their grain sizes and providing even

greater improvements in mechanical characteristics provided that the splat boundaries are completely

removed by the process of remelting.

5. DISCUSSION

It is of interest to understand the heat transfer phenomena associated with the droplet deposition

parameters encountered in this work, and specifically, to estimate the substrate heating due to

successive droplet depositions at high frequency (where substrate refers to the original copper substrate

and the previously deposited and solidified aluminum splats). To this end, a 1-D heat transfer model

used to study phase change and heat transfer of splat impingement was employed (Orme and Huang,

18

1997). For the purpose of gaining a phenomenological understanding, it is assumed that semi-infinite 10

µm thick splats are delivered in the molten state at a rate of 1 every 42 µseconds (corresponding to the

inverse of the deposition frequency, 24,000 Hz) onto a previously solidified substrate. In this first

approximation, the splats are delivered as 10 µm thick slabs and no spreading takes place, i.e., this is a

pure heat conduction problem. The dynamics of droplet flight, which would include convective cooling,

and the dynamics of droplet impact are ignored. Perfect thermal contact is also assumed. In reality,

however, the droplets travel 50 cm in an inert environment where they will undergo convective heat

transfer. The degree of cooling due to convection will depend on the amount of deflection the droplets

experience out of the main droplet stream. If no deflection occurs, as in the case examined in this work,

the droplet cooling will be minimized since the droplets will be thermally protected by the wake of the

neighboring droplet.

The model used is similar to the one developed by Amon et al. (1996, 1998) where they studied the 1-

D heat transfer of a droplet decoupled from the spreading process. They assumed that the solid-liquid

interface was governed by the Stefan solution (i.e., the location is proportional to time). In contrast, the

model employed in this study uses a coordinate transformation to transform the one-dimensional moving

boundary phase change problem into a fixed boundary problem. The model enables efficient tracking of

the solid/liquid interface through the sequential solidification and remelting regimes, and most

importantly, through the turning point defining the transition between solidification and remelting (i.e., the

solid liquid interface can change directions).

Figure 10 illustrates the predicted substrate temperature increase as a function of the number of splats

delivered. As can be seen in the figure, the temperature of the top surface of the substrate increases

19

with the number of splat delivered due to the rapid rate of droplet deposition which provides insufficient

time for complete splat cooling to occur between successive droplet deliveries.

Figure 11 illustrates the numerically determined criterion used to estimate the substrate temperature,

Tsub, and the incoming droplet temperature, Ti, required to initiate the remelting of the previously

deposited and solidified splats for the purpose of removing splat boundaries. Shown is the solid/liquid

interface through a two-splat system, where each splat is 10 µm thick. It can be seen from the figure

that the solid liquid interface penetrates down into the previously solidified material for temperature

difference ratios, (Ti-Tm)/(Tm-Tsub), equal or greater than 1.4, where Tm is the melting temperature, Tsub

is the substrate temperature, and Ti is the incoming droplet temperature. For increasing values of the

ratio of temperature differences, the remelt depth through the previously deposited and solidified splat

increases. Values of the ratio much less than 1.4 prohibits the downward motion of the solid/liquid

interface. In these cases, the material impinges on the previously delivered and solidified splat and

solidifies without remelting the previous splat. Hence the slope of the solid liquid interface is always

positive for these cases. The model provides a first estimate of the substrate temperature necessary to

initiate the onset of remelting and the remelt depth. As the temperature ratio increases (e.g., as the

substrate temperature increases), the depth of the remelt zone increases as can seen in Figure 11 by the

downward extent of the solid-liquid interface into the previously deposited material.

For the conditions employed in this work, in which the incoming droplet temperature is held constant at

30C above the melting point, it is estimated that a substrate temperature of 911K is required to achieve

substrate remelting. Examination of Figure 10 indicates that this temperature is achieved after a

deposition of approximately 40 splats, which corresponds to a vertical distance of approximately 400

20

µm from the initial substrate surface (within the approximations of this model). Hence, at deposit heights

greater than 400 µm, it is expected that substrate remelting will occur and therefore all splat boundaries

at will be obliterated by the mode of phase change. This is attested by our experimental observations

shown in figures 7 (frames center and right), and figures 8 and 9, the later two were taken at distances

greater than 1.0cm from the substrate. The splats in the component at vertical positions less than 400

µm can also be removed by remelting due to increased remelt depths caused by increased substrate

heating resulting from droplets delivered at higher distances. Hence the criterion for the onset of

remelting is estimated to be when the temperature ratio (Ti-Tm)/(Tm-Tsub) = 1.4, and the remelt depth is

minimal. However for higher substrate temperatures, the temperature ratio will have higher values

thereby causing the solid-liquid interface to penetrate deeper into the previously solidified material,

remelting larger vertical zones of material. This phenomenon is observed in experiment, where thick red

zones (on the order of 10.0 mm) can be seen visually during the droplet deposition process. Hence it is

felt that the splat boundary removal at the 1.0mm distance is due to the deposition of droplets far from

the 1.0mm distance which, at the time of their delivery, have caused considerable substrate heating

which subsequently causes the solid-liquid interface to penetrate deep into the previously solidified

material removing the splat boundaries.

In order to facilitate splat boundary remelting throughout the entire component, either the substrate

could be heated, or the incoming droplet temperature can be increased to compensate for a cooler

substrate as illustrated in Figure 11. Either solution will necessarily be employed in the final realization of

the PDM technology.

21

6. SUMMARY

This work describes the early phase of development of the PDM net-form manufacturing technology.

Though the technology is still immature, several new and significant findings have been presented in this

work and are summarized here.

It has been shown that streams of stable, uniformly sized molten aluminum droplets can be formed from

capillary stream break-up and travel distances up to 2,500 droplet diameters in an inert environment.

The main issues that have been found to have a profound affect on stream stability are aluminum oxides

in the pre-jetted melt, and corrosion of the nozzle due to the solvent nature of the molten aluminum. The

latter issue is a problem not only due to degradation of the nozzle, but also because of the introduction

of contaminants into the pre-jetted melt.

It was found that molten aluminum corroded a natural diamond orifice, even though it was in an inert

environment and held at a temperature lower that that known to cause diamond corrosion. No

formation of AlC was found from the analysis of the nozzle. It is believed that this is the first time such

an aluminum/diamond reaction has been reported.

Circular cylinders (towers) were fabricated up to 0.15m high by depositing the molten aluminum

droplets onto the same location on the substrate. Microstructures of vertical sections revealed no splat

boundaries in the component, and approximately uniform microstructures throughout the field of view of

the images. A small increase in grain size was observed in a sample that was 25mm from the substrate

compared to samples at 1.0 mm and 5.0 mm from the substrate.

The presence of oxide grain boundaries and pores in cylinders fabricated with unfiltered molten

aluminum was reported. Cylinders fabricated with filtered molten aluminum were devoid of visible

22

oxide grain boundaries and voids. The grain size of the components fabricated with unfiltered and

filtered molten aluminum was approximately equal.

Mechanical testing revealed that the droplet deposited components had a 30% increase in ultimate

tensile strength (filtered or unfiltered). The components made from filtered molten aluminum showed no

decrease in density over the raw stock, indicating minimal, if any, porosity. The samples fabricated with

unfiltered molten aluminum showed a 1.1% decrease in density.

Hence, this work illustrates that fabrication of components from molten aluminum droplet deposition

leads to significant microstructural refinement and mechanical property enhancements when care has

been taken to filter the molten metal prior to jetting. It is believed that the results presented here are an

underestimate of the microstructural refinement and mechanical quality enhancements of the final PDM

technology. This is because the droplets did not solidify as rapidly as they will in the final PDM

technology since electrostatic charging and deflection was not employed. In the final realization of

PDM, the droplets will be deflected to different locations on the substrate, allowing the splats to cool

prior to the next splats arrival. In this manner, the temperature of the previously deposited and solidified

material will not increase with time, allowing the droplets to rapidly solidify which leads to refined

microstructures and enhanced mechanical properties.

ACKNOWLEDGMENTS

This work was made possible with the generous grants from Boeing Commercial Airplane Group

(BCA-23483), Lawrence Livermore National Laboratories (B345710), and the National Science

23

Foundation (DMI-9457205, DMI-9622400). Additional thanks is extended to the Boeing Commercial

Airplane Group for taking the micrographs shown in Figures 8 and 9, to Mr. Charles Huang for

preparing the sample shown in Figure 7, and to Mr. Changzheng Huang for providing the simulations

shown in Figures 10 and 11.

24

REFERENCES

Atkinson, D. 1997, Rapid Prototyping and Tooling, A practical Guide, Strategy Publications, Weltech

Centre, Ridgeway, Welwyn Garden City, Herts AAL72AA, UK

Amon, C.H., Schmaltz, K.S., Merz, R., Prinz, F.B., 1996 “Numerical and Experimental Investigation

of Interface Bonding Via Substrate Remelting of an Impinging Molten Metal Droplet” ASME J. of Heat

Transfer, 118, 164-172

Amon, C.H., Beuth, J.L., Merz, R., Prinz, F.B., and Weiss. L.E., 1998, “Shape Deposition

Manufacturing with Microcasting: Processing, Thermal and Mechanical Issues, ASME J. Manufacturing

Science and Engineering, Vol. 120, pp 656-667

Argarwala M.K., Van Werren R, Jamalabad V., Langrana N., Whalen P., Danforth S.C., and Ballard

C. 1995, “Quality of Parts Processed by Fused Deposition” Proceedings to the Solid Freeform

Fabrication Symposium, University of Texas at Austin

Chin, R.K., Beuth, J.L., and Amon, C.H., 1996 “Thermomechanical Modeling of Molten Metal

Droplet Solidification Applied to Layered Manufacturing,” Mechanics of Materials, Vol. 24, pp 257-

271

Das, S., Wohlert, M., Beaman, J., Bourell, D., 1997 “Direct Selective Laser Sintering and

Containerless Hot Isostatic Pressing for High Performance Metal Components” Proceedings to the

Solid Freeform Fabrication Symposium, University of Texas at Austin

25

Fielder, H.C., Sawyer, T.F., Kopp, R.W., and Leatham, A.G. (1987), “The Spray Forming of

Superalloys,” Journal of Metals, 28-33

Griffin E.A., McMillan S., Griffin C. 1995, Proceedings to the Solid Freeform Fabrication Symposium,

Univ. of Texas at Austin

Jacobs, P. 1992,“Rapid Prototyping and Manufacturing – Fundamentals of StereoLithography”

Published by the Society of Manufacturing Engineers, Dearborn, MI

Klocke F., Celiker T., Song Y.A. 1995, Proc of the 4th European Conference on Rapid Prototyping

and Manufacturing, Belgirate, Italy

Lavernia, E.J., Ayers, J.D., and Srivatsan, T.S. 1992, “Rapid Solidification Processing With Specific

Application To Aluminum Alloys,” International Materials Reviews, vol. 37

Orme M & Muntz E.P., 1990, “The Manipulation of Capillary Stream Breakup Using Amplitude

Modulated Disturbances: A Pictorial and Quantitative Representation,” Phys. Of Fluids, 2, (7)

Orme M, 1991, “On the Genesis of Droplet Stream Microspeed Dispersions,” Physics of Fluids, 3,

(12)

Orme M., Muntz E.P. 1992, United States patent Number 5,171,360, December 15

Orme M., 1993, “A Novel Technique of Rapid Solidification Net-Form Materials Synthesis,” Journal

of Materials Engineering and Performance, 2, (3)

26

Orme M, Huang C and Courter J, 1996, “Precision Droplet Based Manufacturing and Material

Synthesis: Fluid Dynamic and Thermal Control Issues”, ILASS Journal of Atomization and Sprays

vol. 6,

Orme, M.E., Huang, C. 1997, “Phase Change Manipulation for Droplet-Based Solid Freeform

Fabrication”, ASME Journal of Heat Transfer, 119, 818 – 823,

Pintat T., Greul M., Greulich M., Wilkening C. 1995, Proceedings to the Solid Freeform Fabrication

Symposium, University of Texas at Austin

Prinz, F.B., Weiss, L.E., Amon, C.H. and Beuth, J.L., 1995, “Processing, Thermal and Mechanical

Issues in Shape Deposition Manufacturing” Solid Freeform Fabrication Symposium, Austin, Texas,

118-129

Singer, A.R.E. (1983) “A New Generation of Engineering Materials Produced by Spray Forming”

Materials & Design, Vol. 4, 1983

Sachs E., Cima M., Williams P., Brancazio D., Cornie J., 1992, J. of Eng. For Ind., 114, 4, 481-488

27

1. FIGURES

RapidPrototypedartifact

ball placementaccuracy to12.5 microns

computer control ofdroplet charge &substrate motion

x-y table

environmentalchamber

droplet generator

dropletstream

deflectionelectrodes

Figure 1: Conceptual schematic Precision Droplet-Based Net-Form Manufacturing

Figure 2: Illustration of droplet formation from capillary streams

λ λ

2rd

Charge electrode

2ro

orifice Vibrating

rod

Molten metal

28

Figure 3: Schematic of Aluminum Droplet Generation and Deposition Facility

optical bench

environmental chamber

deflection plates

CCD array camera

vibrating rod molten metal

cartridgeresistance heater

observation windows

substrate

nozzle

29

Figure 4: Schematic of cartridge, heater, and plunger assembly

Figure 5: (a) SEM photograph of a 100 µµm diameter natural diamond orifice which was jetted for two hours; (b)

enlarged view which illustrates the rough region on the right which was exposed for two hours, and the smoother

region on the left which was exposed for one hour.

(a)

(a) (b)

30

Figure 6: Molten aluminum droplets generated from a 100 µµm diameter orifice

after traveling approximately 0.4meters in an inert environment

Figure 7: Three optical microscope images of the microstructure of a droplet deposited aluminum component at

various distances from the substrate. Left: Image is a vertical cut 1.0 mm from the substrate illustrating

approximate grain sizes of 50 microns. Center: microstructure through the specimen at a distance of 5.0 mm from

the substrate. Right: microstructure of section 25mm from substrate (please not the different scales).

31

Figure 8: Micrographs of droplet deposited aluminum cylinders. Field of view in both images is 160x190µµm. Splat

boundaries are not evident in either image. Left: melt was unfiltered. Oxidation and pores are evident around the

grains; Right: melt was filtered. Grain boundaries are not encapsulated with an oxide film.

Figure 9: Zoomed out micrographs of droplet deposited aluminum structures. Field of view in both images is

800x970µµm. Splat boundaries do not exist in either image. Left: melt was unfiltered. Oxidation and pores are

evident around the grains; Right: melt was filtered. Oxide grain boundaries are not present.

32

Figure 10: Plot of temperature increase of the top surface of the deposited component as a function of number of

splat delivered for a splat delivery rate of 24,000Hz and a splat thickness of 10µµm.

33

Figure 11: Plot of the solid/liquid interface through a two-splat system for various values of the temperature

difference ratio.

Sol

id/li

quid

inte

rfac

e lo

catio

n (m

icro

ns)