701 FTP

R
E
V
I
E
W
S
Physical Aspects of Process Control
in Selective Laser Sintering of Metals
By Suman Das*
1. Introduction
Selective Laser Sintering (SLS) is a laser based rapid manu-
facturing technology that builds objects layer by layer using
powder and a computer controlled laser. An object's comput-
er aided design (CAD) model is mathematically sliced into
thin layers. The object is then created by selectively fusing se-
quentially deposited layers of a powder with a scanning laser
beam. Each scanned layer represents a cross section of the
sliced CAD model. A schematic of the SLS process is shown
in Fig. 1.
Recent research efforts have demonstrated great potential
to manufacture metal components with microstructures and
mechanical properties equivalent to or superior than conven-
tionally processed materials. These efforts
[16]
have resulted
in the capability to successfully process high temperature,
high performance materials such as nickel base superalloys,
titanium alloys and superalloy cermets into functional com-
ponents including the titanium guidance housing for the
Sidewinder missile and cermet abrasive turbine blade tips for
a next generation fighter aircraft engine.
Although SLS holds great future promise as an advanced
manufacturing technology capable of directly manufacturing
functional metallic components from powder in a single step,
it is a complex process that requires significant further re-
search and understanding. A schematic of the complexity in
SLS is shown in Fig. 2. The density of the powder bed in SLS
is initially at approximately 55 % theoretical density. When
the laser beam strikes the powder surface, the powder first
undergoes densification by sintering. As additional laser en-
ergy is absorbed by the sintered powder, it eventually under-
goes melting and further volume reduction as additional
powder is dragged into the melt pool under the action of sur-
face tension. As the scanning laser beam proceeds to a neigh-
boring portion of the powder bed, strong convection currents
are formed due to surface tension gradients in the transient
melt pool formed under the laser beam. Eventually, the melt
pool undergoes solidification by nucleation of epitaxial grains
off the remelted surface of the previous layer. Additionally,
heat transfer by conduction and radiation takes place from
the laser processed zone into the surrounding powder, into
the substrate underneath the processed powder, and by radi-
ation and convection into space above the melt pool. These
complex phenomena are strongly material dependent and are
governed by a material's chemistry and its optical, thermal
and rheological properties.
As is evident from the foregoing description, Selective la-
ser sintering (SLS) of metals is a complex process exhibiting
multiple modes of heat and mass transfer, and in some in-
stances, chemical reactions. The strong interplay between
these physical mechanisms directly impacts the SLS process,
determining whether a given material can be processed suc-
cessfully and if so, how the microstructure and properties of
the processed material are influenced. In order to understand
the complex nature of these mechanisms, their interplay and
ADVANCED ENGINEERING MATERIALS 2003, 5, No. 10 DOI: 10.1002/adem.200310099 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 701
[*] Dr. Suman Das

Department of Mechanical Engineering
University of Michigan
Ann Arbor, MI 48109-2125 (USA)
E-mail: sumandas@umich.edu
Direct selective laser sintering (SLS) is a layered manufacturing technique that can produce fully
dense, functional components in high performance metals. In this review paper, a first step is taken
towards identifying and understanding some of the important physical mechanisms in direct SLS.
This study not only provides an insight into phenomena observed during direct SLS processing of a
variety of metallic materials but also helps in selecting those materials that are most amenable to direct
SLS processing. The physical mechanisms discussed include oxidation, non-equilibrium wetting,
epitaxial solidification, metal vaporization, and oxide purification. Understanding these mechanisms
is crucial for the design of direct SLS machines, process development, and process control.
Das/Physical Aspects of Process Control in Selective Laser Sintering of Metals
R
E
V
I
E
W
S
impact on process control in SLS, it is appropriate to first un-
derstand them at a fundamental level. Simple models that
can quantitatively reveal the influence of these mechanisms
as functions of processing conditions are needed first. Follow-
ing this step, increasingly detailed models that take into ac-
count the coupling between different mechanisms should be
investigated. Models of SLS have been developed previously
for polymers
[711]
and for multi-component metal powder
mixtures
[1216]
. These models focused on describing heat
transfer and sintering phenomena, attempting to predict the
densification and shape of the laser sintered powder bed as a
function of process parameters that include powder bed pre-
heat temperature, laser power, scan speed, scan spacing, par-
ticle size and packing density.
In this article, a first step is taken towards identifying and
understanding some of the important physical mechanisms
in direct SLS processing of metals and metal alloys. These
mechanisms are particularly relevant to SLS process control,
especially when the intent it to produce defect free, full den-
sity metal parts with mechanical properties suitable for ser-
vice conditions. This study not only provides an insight into
phenomena observed during direct SLS of a variety of metal-
lic materials, but also provides insights towards selecting
those materials that are most amenable to successful process-
ing by direct SLS. This article reviews the mechanisms of oxi-
dation, atmosphere control, wetting, epitaxial solidification,
metal vaporization under vacuum and oxide purification. An
understanding of these physical mechanisms is essential for
the design of direct SLS machines, their process control and
materials specific SLS process development.
2. Wetting
2.1 Oxidation, Wettability and Mechanical Properties
Direct SLS of metals is far more demanding than SLS of
polymers or indirect SLS of metals. In direct SLS, oxidation is
a serious problem at the higher processing temperatures nec-
essary for melting metals of interest (typically exceeding
1000 C). Since SLS is a layer-by-layer additive process, oxida-
tion of laser processed powder is a severe impediment to in-
terlayer bonding and causes defects such as balling. Balling
occurs when the laser melted metal powder layer does not
wet the underlying substrate due to a contamination layer of
oxide being present on the substrate and on the surface of the
melt. Oxidation also causes delamination induced by poor in-
terlayer bonding in combination with thermal stresses. These
phenomena severely degrade material properties and part ge-
ometry. Poor wetting of an oxide substrate by a liquid metal
is explained by the interfacial energies involved. The equilib-
rium of a liquid in contact with non-interacting solid and gas-
eous phases is shown in Fig. 3. The equilibrium of interfacial
free energies is governed by
c
SV
c
SL
= c
LV
cos(h) (1)
The liquid will wet the substrate as cos(h) 1 or equiva-
lently, if c
SV
c
SL
> c
LV
. This corresponds to the case on the
left. A large positive spreading coefficient S = c
SV
c
SL
c
LV
favours spreading. Surface free energies of metal oxides are
much lower than the corresponding liquid-vapor surface free
energy.
[12]
Thus, in the absence of a chemical reaction be-
tween the liquid and the substrate, that may aid wetting and
spreading, liquid metals do not wet surface oxide films, pre-
ferring to minimize surface area by balling instead.
In order to mitigate oxidation as well as to ensure good
wetting and successful layer-by-layer consolidation in direct
SLS of metals, processing must be conducted in a vacuum or
protective atmosphere using high purity inert gases. An El-
lingham diagram
[17]
of common oxides shows that at the
melting points of Nickel (1454.4 C), Chromium (1857 C) and
Titanium (1704.4 C), the equilibrium partial pressure of oxy-
gen must be less than 10
5
, 10
12
and 10
16
atm respectively to
prevent the oxidation reaction. Therefore, some degree of oxi-
702 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.aem-journal.de ADVANCED ENGINEERING MATERIALS 2003, 5, No. 10
Fig. 1. Selective Laser Sintering.
Previously fabricated layers
Remelted
substrate
Laser beam
Melt pool
Epitaxial grains
Powder layer
Motion
Fig. 2. Laser Material Interaction in SLS. Fig. 3. Three-phase equilibrium for wetting and non-wetting systems.
R
E
V
I
E
W
S
dation cannot be avoided under normal operating conditions.
In traditional furnace brazing operations for metals and al-
loys conducted under pure hydrogen atmosphere, it is impor-
tant to consider the dew point or moisture content of the at-
mosphere against the most stable oxide forming element. A
chart of metal-metal oxide equilibria under pure hydrogen at
1 atm is available in the literature.
[18]
This chart, calculated
for a variety of metal oxides shows the temperature and
water vapor partial pressure combinations under which the
reaction is reducing or oxidizing. For good wetting to take
place, reduction of surface oxides is necessary to provide a
clean metal-metal interface. Conventional practice involves
providing a continuous flow of hydrogen to maintain the
water vapor partial pressure levels favorably. It has been
noted that the same chart is equally applicable for continuous
flow inert gas atmospheres with similar oxygen content.
Some extremely reactive metals such as titanium cannot be
processed under hydrogen atmospheres for two reasons.
First, it is difficult to obtain hydrogen of the necessary purity
at the processing temperature of interest (the melt temperature).
Second, titanium and its alloys are extremely sensitive to
hydrogen embrittlement. Thus, traditional high temperature
sintering and vacuum brazing is conducted in the
10
4
10
6
Torr vacuum range. Although processing under
these vacuum levels yields satisfactory results with regards
to wetting, volatilization of alloying elements at high temper-
atures under reduced pressure can be a serious concern.
Oxidic cleanliness is also of paramount importance while
laser processing high performance metals, especially superal-
loys and titanium with extremely high strength, high creep
resistance properties at elevated service temperatures ap-
proaching the solidus temperature. The detrimental influence
of increasing oxygen content on the rupture life of superal-
loys is illustrated by Choudhry
[19]
. It is shown that increasing
the oxygen content from 25 ppm to 100 ppm in cast Udimet
500 at 670 C and 172 MPa decreases its rupture life by 46 %
while for IN 100 at 732 C and 689 MPa, rupture life is de-
creased by 40 %.
Commercially pure titanium (CP Ti) has a specification of
1400 ppm maximum acceptable level of oxygen content. The
titanium alloy Ti-6Al-4V has specification oxygen limits in
the final product ranging from 0.1 % to 0.4 % by weight de-
pending on the particular application. Molybdenum is classi-
fied as a superclean metal
[19]
and is typically purified by
electron beam melting at 10
5
10
6
Torr. When compared to
cast or wrought product, metal powders have higher degree
of contamination due to their much higher surface area per
unit volume. As-received powders are known to harbor mois-
ture, organics, adsorbed gases, oxide and nitride films on par-
ticle surfaces depending on the production method and shelf
life.
[20]
These contaminants not only inhibit successful wetting
and densification of the melted powder but also degrade me-
chanical properties of the consolidated product. Therefore,
processing under high vacuum (<10
5
Torr) or under ultra
high purity inert environments is necessary for successful di-
rect SLS of high performance metals.
2.2. Homologous Wetting
Formation of a metal object layer by layer in direct SLS re-
quires that the powder melted by the laser completely wets
the layer underneath to form a strong metallurgical bond.
Most of the work on wetting reported in the literature is
based on isothermal wetting of dissimilar substances, for ex-
ample wetting of a liquid on a solid substrate of chemical
composition different than the liquid. An excellent review of
the physical mechanisms governing this wetting regime is
given by De Gennes.
[21]
The wetting behavior of a molten liq-
uid on a substrate of its own kind was investigated only re-
cently by Schiaffino and Sonin.
[2224]
This particular situation
is known as homologous wetting and is inherently a non-
equilibrium phenomenon involving simultaneous heat trans-
fer, fluid flow and solidification.
In the case of a liquid droplet placed on a substrate of its
own kind, wetting involves simultaneous fluid flow by
spreading, heat transfer to the substrate and solidification at
the solid-liquid interface. A schematic of this situation is
shown in Fig. 4 where the liquid at temperature T
0
spreads at
a velocity U(t), the subcooled solid substrate is at temperature
T
t
and the fusion front is at temperature T
f
. As the liquid
droplet spreads on the substrate with a dynamic contact an-
gle h
a
, a solidification front grows into the liquid with a dy-
namic angle h
s
at the contact perimeter. The dynamic spread-
ing of the droplet is arrested when h
s
= h
a
. This particular
value of h
s
is denoted by h
*
.
In their experimental studies, Schiaffino and Sonin
[22]
in-
vestigated the spreading behavior of superheated droplets of
molten microcrystalline wax placed onto a subcooled sub-
strate of the same kind. They found that for a given material
with known thermophysical properties, h
*
= f(S), i.e. the angle
of molten contact line arrest is a function of the Stefan num-
ber. S is given by
S = c
T
f
T
t
L
(2)
where c is the specific heat and L is the latent heat of fusion.
In particular, their experimental data for Reed 6882 micro-
crystalline wax could be fitted with the power law h
*
= 85 S
0.5
.
In addition, they found that the spreading behavior of the liq-
ADVANCED ENGINEERING MATERIALS 2003, 5, No. 10 http://www.aem-journal.de 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 703
Fig. 4. Isothermal Wetting (left) and homologous wetting (right).
R
E
V
I
E
W
S
uid droplet before contact line arrest followed the
well known Tanner law
[21]
Ca =
lU
= 0:013h
3
a
(3)
where Ca is the Capillary number, l is the viscosity,
and r is the surface tension.
In direct SLS, melting of powder and subsequent
solidification onto a previously formed layer consti-
tutes homologous wetting. To observe homologous
wetting in direct SLS, experiments were conducted
by laser melting droplets onto prepared substrates
of the same material. Stainless steel 304L was cho-
sen as the material for investigation due to its ready
availability in both powder and plate form. Stain-
less steel plate 0.125 inches thick was prepared by
polishing with 120, 240, 400 and 600 abrasive grits,
followed by 3 lm diamond paste and a final polish
with 0.05 lm alumina slurry, followed by ultrasonic
cleaning in acetone. This preparation technique was
necessary to provide as clean and smooth a surface
as possible since the presence of surface oxides will
cause the material to spread poorly
[25]
and surface
roughness can drastically alter wetting behavior
[21]
.
Two types of experiments were conducted. In the first type
of experiment, stainless steel powder placed on the stainless
steel substrate was melted by a stationary Nd:YAG laser
beam incident on the powder. In the second type of experi-
ment, a strip of stainless steel, 1 mm wide and 4 mm long
was placed on the substrate and melted by the same laser
beam. Each type of experiment was conducted under vacuum
(<10
5
Torr) with no substrate preheat, 350 C preheat and
900 C preheat to document the influence of substrate pre-
heating on the wetting behavior (analogous to the Stefan
number in the studies of Schiaffino and Sonin
[22]
). The dy-
namics of melting, droplet formation, wetting and solidifica-
tion were captured by video microscopy. The objective in
these experiments was to visualize and qualitatively under-
stand the wetting behavior of a metal melt on a substrate of
its own kind. Therefore, no attempt was made to carefully
control the temperature of the superheated melt. Such an ar-
rangement could be implemented by controlling the laser
power via feedback from a pyrometer aimed at the melt pool.
All the experiments revealed that after melting and droplet
formation, the molten metal showed no tendency to spread
and wet the underlying substrate. Instead, it formed a nearly
spherical droplet having a point like contact with
the substrate. The influence of preheat temperature
on the wetting behavior was minimal and could not
be distinguished visually. Sequential frames from
video microscopy of one such experiment con-
ducted with stainless steel strip on stainless steel
plate are shown in Figs. 57. The image at 1 s shows
the point of incidence of the laser beam, approxi-
mately at midlength of the strip. At 33 s, the ends of
the strip begin to lift off the substrate and collapse rapidly to-
wards the center. The laser beam is turned off at 49 s, reveal-
ing a very nearly spherical molten droplet and by 54 s, the
droplet has completely solidified.
The solidified droplets (with largest dimension typically
2 mm) exhibited different shapes for powder and solid strips.
In the case of powder on plate type experiments, the solidi-
fied droplets, shown in Fig. 9 had an elongated peaked fea-
ture at the top and a finite contact area with the plate. On the
other hand, the solidified droplets of strip on plate type ex-
periments, shown on in Fig. 8 exhibited a much higher de-
gree of sphericity, and a very small area of contact with the
plate. The typical size of solidified droplets in the photos is
2 mm.
Poor metallurgical bonding was observed between the so-
lidified droplet and the substrate in each of the homologous
wetting experiments conducted. The solidified droplets were
easily detached from their respective substrates. In addition,
the substrates did not exhibit any evidence of melting at the
droplet-substrate interface. Despite excellent vacuum atmo-
sphere control (< 10
6
Torr), poor wetting observed in these
experiments resulted due to a lack of substrate remelting and
Fig. 5. Homologous wetting experiment at 1s, 4s and 7s
Fig. 6. Homologous wetting experiment at 15s, 33s and 35s.
Fig. 7. Homologous wetting experiment at 37s, 50s and 54s.
Fig. 8. Solidified droplets in strip on plate experiments.
R
E
V
I
E
W
S
ensuing epitaxial solidification. As shown in the following
section, remelting of the underlying substrate is necessary to
break down surface contamination such as oxide films in or-
der to provide a clean interface at the atomic level between
the solid and the liquid.
2.3 Epitaxial Solidification in SLS
Direct SLS of metals is very similar to fusion welding. In
fusion welding, the base metal and the weld metal may have
similar but not necessarily the same composition, whereas in
direct SLS of monolithic metallic materials, the base metal
and weld metal are necessarily the same. Some dilution of the
base metal is known to occur in all types of fusion welding.
Dilution refers to partial localized remelting of the base metal
during welding and is necessary to provide a continuous in-
terface free of oxide films or other contaminants between the
solidifying liquid weld metal and the base metal. Dilution is
deemed extremely important to ensure good weld integrity.
The most important implication of dilution or meltback is
that during solidification, grain growth will occur at the sol-
id-liquid interface between the base metal and the liquid in
contact with it. Grain growth is important because it has been
well established by metallographic inspection that initial so-
lidification in fusion welding occurs by epitaxial solidifica-
tion.
[26,27]
Due to the significant similarities between fusion
welding and direct SLS, it is instructive to understand the im-
plications of epitaxial solidification on SLS process control.
Epitaxial solidification is a process of heterogeneous nucle-
ation. In heterogeneous nucleation, a solid nucleus nucle-
ates from the liquid onto a substrate with which the liquid is
in contact. The shape of the nucleus is governed by the bal-
ance of interfacial free energies between the nucleus, solid
and the liquid. It can be shown that the total interfacial free
energy of this system is minimized when the nucleus as-
sumes the shape of a spherical cap as shown in Fig. 10.
At equilibrium, the balance of interfacial forces is given by
c
ML
= c
SM
c
SL
cos(j) (4)
where the interfacial free energies are c
ML
between the sub-
strate metal and the liquid, c
SM
between the solid nucleus
and the substrate metal, and c
SL
between the solid nucleus
and the liquid. The contact angle of the nucleus is w
Rearranging equation 4 gives
cos(j) = c
ML
c
SM
c
SL
(1)
In the context of direct SLS, the compositions of the solidi-
fied nucleus and the solid substrate are the same. Thus the in-
terfacial energy between the nucleus and the solid substrate
should be negligible while the interfacial energies between
the nucleus and the liquid, and the solid substrate and the liq-
uid should be about the same. This leads to the following ap-
proximations:
c
SM
~ 0
c
ML
~ c
SL
j 0 (2)
The change in free energy associated with the formation of
a solid spherical cap nucleus with radius of curvature r is
given by
DG
het
= V
S
DG
V
A
S
c
SL
DG
het
= S(j)
4
3
pr
3
DG
V
4pr
2
c
SL

(3)
where DG
v
is the free energy change per unit volume asso-
ciated with formation of the nucleus, V
s
is the volume of the
nucleus, A
s
is the surface area of the new interface created be-
tween the solid nucleus and the liquid, and S(j) is a shape
factor dependent on the contact angle given by
S(j) =
2cos(j) [ [ 1cos(j) [ [
2
4
(4)
By differentiating equation 7 with respect to r and setting
the result equal to zero, the critical radius r* to promote het-
erogeneous nucleation is given by
r
+
=
2c
SL
DG
V
(5)
By substituting the expression for r* into equation 7, the as-
sociated nucleation energy barrier for heterogeneous nuclea-
tion to proceed is
DG
+
het
=
16p
3
c
3
SL
DG
3
V
S(j) (6)
It is interesting to note the dependence of DG
het
on S(j).
From Eq. 6, it apparent that in epitaxial solidification, S(j)
Fig. 9. Solidified droplets in powder on plate experiments.
Fig. 10. Solid nucleus in contact with substrate metal and liquid.
R
E
V
I
E
W
S
0. This result implies that the energy barrier for heteroge-
neous nucleation to proceed during epitaxial solidification is
vanishingly small. This is in stark contrast to conventional
casting processes in which j is finite, and hence S(j) and
dG
het
are finite due to mold and cast materials being dissimi-
lar, as illustrated in Fig. 11 after Grong.
[27]
Nucleation proceeds immediately in epitaxial solidifica-
tion by virtue of the negligible energy barrier. For this mode
of solidification to occur, the substrate must be partially
melted back to promote growth from the prior grains of the
substrate. Epitaxial solidification by heterogeneous nuclea-
tion is indeed the mode of solidification in direct SLS process-
ing to full density. Evidence of epitaxial solidification was
confirmed by microstructural examination of sections of SLS
processed material encompassing several layer thicknesses.
A cross-section of specimen 102097, shown in Fig. 12a is
an example of incomplete homologous wetting and solidifica-
tion. In this case, the melt front from the laser-melted powder
layer above propagated down to barely contact the surface of
the previous layer. The uniformity of layer thickness is partic-
ularly notable. In contrast, layers in the cross-section of speci-
men 102497 produced at a higher laser power density, shown
in Fig. 12b, are completely indistinguishable indicating con-
tinuous grain growth across the layers. This result indicates
that sufficient remelt of the previous layer took place to in-
duce epitaxial solidification. It should be noted that these mi-
crographs traverse approximately three layer thicknesses.
Another benefit resulting from epitaxial solidification is
fine grain size resulting in improved mechanical properties.
Equation 9 shows that for heterogeneous nucleation to pro-
ceed, the critical radius of the nucleus r* depends on the sol-
id-liquid interfacial free energy c
SL
. In SLS, the solid and liq-
uid are of nearly the same composition (assuming negligible
contamination or segregation). Thus, c
SL
is very small imply-
ing that r* is very small. Since SLS is a localized solidification
process, this means that under identical processing condi-
tions (temperature gradients at the solid liquid interface, cool-
ing rates, etc), the SLS processed material will exhibit finer
and more uniform grain size compared to conventional cast-
ing. This hypothesis was also confirmed by examining the
etched microstruture of SLS processed Ti-6Al-4V which re-
vealed finer grain size than conventionally cast Ti-6Al-4V.
Thus, the mechanism for good wetting in direct SLS is
epitaxial solidification by heterogeneous nucleation. As an
analogue to fusion welding,
[26]
in SLS, partial remelting of the
previous layer is necessary to ensure full density. Remelting
the previous layer removes surface contaminants, breaks
down oxide films, provides a clean solid-liquid interface at
the atomic level and allows epitaxial grains to grow heteroge-
neously. Laser processing parameters in direct SLS should be
chosen so as to ensure that sufficient remelt of the previous
layer takes place and continuity of the solid-liquid interface
under the moving laser beam is maintained.
3. Volatilization
Volatilization of elements is a well known phenomenon in
vacuum metallurgy,
[2830]
electron beam processing
[31, 32]
and
thin film technology.
[33]
Evaporation and condensation of
metals has been noted in direct SLS as well. The evaporation
process from a metal melt is governed by four distinct re-
Fig. 11. Schematic of free energy change associated with heterogeneous nucleation in
casting and welding juxtaposed with free energy change in homogeneous nucleation
(after Grong, 1994).
a)
b)
Fig. 12. a) Etched microstructure of SLS processed Ti-6Al-4V exhibiting porosity and
insufficient substrate remelt and b) Etched microstructure of fully dense Ti-6Al-4V
specimen.
R
E
V
I
E
W
S
gimes: mass transport of atoms from the interior of the melt
to its surface, phase change to gaseous state at the surface,
mass transport in the gas phase above the melt and finally
condensation.
3.1. Evaporation of a Pure Metal
The molecular (or atomic) rate of evaporation per unit area
and time for a liquid in equilibrium with its vapor was de-
rived first by Hertz and later modified by Knudsen
[33]
to give
dN
e
A
e
dt
= a
1
2pmkT
r
(P
v
P) (7)
where m is mass of an individual molecule, k is Boltzmann's
constant, T is the absolute temperature, P
m
is the vapor pres-
sure at the liquid's surface temperature and P is the hydro-
static pressure acting on the liquid surface. The coefficient a
m
accounts for the fraction of vaporized molecules that are
ejected and not resorbed back into the evaporating surface,
hence, the maximum evaporation rate is attained when a
m
= 1.
Langmuir showed that the Hertz-Knudsen equation was
also applicable to vaporization from free solid surfaces. In ad-
dition, he demonstrated that for pressures below 0.1 Torr
(10
4
atm), the vaporization rate is independent of residual
gas pressure due to the negligible resistance offered. This re-
sulted in the Langmuir equation for the maximum vaporiza-
tion rate expressed in mass flux form as
[33]
C =
m
2pkT
r
P
v
(8)
The Clausius-Clapeyron equation describes the change of
equilibrium vapor pressure as a function of temperature and
is given by
dP
v
=
LdT
RT
2
(9)
where L is the molar latent heat of vaporization and R is the
molar gas constant. If the latent heat of vaporization is as-
sumed constant for the temperature range of interest, equa-
tion 13 can be integrated to give the relation
P
v
(T) = P
0
e
L
RT
(10)
For most materials, in the range of vapor pressures below
1 Torr (1.315 10
5
atm), it has been observed experimen-
tally
[34]
that the vapor pressure is indeed related to the tem-
perature in the form of Eq. 13. Substituting Eq. 14 into equa-
tion 12 for the evaporation rate yields the form
C = P
0
m
2pkT
r
e
L
RT
(11)
showing the near exponential dependence of the vaporization
rate on the vaporization temperature.
3.2. Alloy Melting and Vaporization Under Vacuum
In the melting of alloys under vacuum, volatile solute and
impurity atoms are vaporized in addition to base metal
atoms. Therefore, mass transport of the solute from the interi-
or to the surface must be considered in addition to evapora-
tion from the surface. Both processes contribute to the overall
transfer of metal atoms into the gas phase and it is therefore
important to determine the rate limiting step.
The average mass transfer coefficient in a stirred metal
melt is described by
[35]
k
m;l
= 2
D
ph
r
(12)
where D is the diffusion coefficient (m
2
/s) of the solute in the
liquid and h is the "residence time" (s) defined as the melt
pool characteristic length divided by the surface velocity of
the flowing melt. Thus, the flux to the surface from the bulk
liquid is
j
A
= k
m;l
(C
A;
C
s
A
) (13)
where j
A
is the flux (mol/m
2
s), C
A,
and C
S
A
are volumetric
concentrations of the solute in the bulk and at the surface re-
spectively. Evaporation from the surface of a pure metal was
described earlier. In the case of an alloy, the evaporation rate
or flux of an alloying component is given by
[35]
j
A
=
c
A
P
0
A
C
s
A
r
2pMRT
_ (14)
where c
A
is the activity coefficient of the component, P
0
A
is
the vapor pressure of the pure component at the temperature
under consideration, r is the molar density of the alloy and M
is the alloy's molecular weight. From equation 18, the mass
transfer coefficient for surface evaporation is
k
m;e
=
c
A
P
0
A
r
2pMRT
_ (15)
When the melting and vaporization takes place under vac-
uum better than 0.1 Torr, the resistance to mass transport in
the gas phase is negligible, as was indicated earlier. Under
this condition, the liquid phase mass transfer and evaporation
fluxes are in equilibrium. Therefore,
k
m;l
(C
A;
C
s
A
) = k
m;e
C
s
A
(16)
Subsituting for C
S
A
from Eq. 20 into Eq. 17, the expression
for the flux is given by
j
A
=
k
m;e
k
m:l
k
m:e
k
m;l
C
A;
= KC
A;
(17)
where K =
1
k
m;e
1
k
m;l
is the overall mass transfer rate constant.
R
E
V
I
E
W
S
By evaluating the magnitudes of the resistances 1/k
m,l
and 1/k
m,e
offered by each step, one can determine whether
they are the same order of magnitude or whether one is
clearly the rate limiting step. The higher the degree of stirring
or convection in a melt, the higher the value of k
m,l
and the
smaller its effect on the overall mass transfer rate. Conversely,
the lower the vapor pressure P
0
A
of the solute, the smaller the
value of k
m,e
and the greater the resistance offered by surface
evaporation to mass transport.
For example, it has been reported
[32]
that for electron beam
processing of a wide variety of liquid metal systems, k
m,l
falls
in the range 1 10
5
1 10
4
m/s. In the refining of silver
from liquid copper at 1920 K by electron beam melting in a
water cooled hearth, the rate constants reported were
k
m,e
= 6.1 10
4
m/s, K =1.4 10
4
m/s to yield k
m,l
= 1.9
10
4
m/s. Given that the hearth area was 0.16 m
2
and typical
surface velocity observed in this experiment was 7 10
2
m/
s, the residence time h is of order 3 s. This value is of the or-
der of the time expected in an inductively stirred melt.
[35]
If
this system is considered for full density SLS processing
(where a continuous melt pool would be maintained), as a
first approximation we can assume that the laser scan speed
is representative of the surface velocity in the melt pool and
the melt pool size is on the order of the beam diameter. For a
typical scan speed of 5.08 10
2
m/s (2 in/s) and a beam di-
ameter of 5 10
2
cm, the value of h is 0.01 s. Due to the high
surface velocity for the pool size under considera-
tion, it is expected that stirring is even more vigorous
in SLS and that k
m,l,SLS
= 10 k
m,l,EB
. This calculation is
based on a very conservative estimate of flow veloci-
ty at the pool surface. Surface velocities induced by
surface tension gradients that are 2 to 3 orders of
magnitude higher than the velocity of the moving
heat source have been observed and predicted
[30]
.
This computation indicates very good mixing in the
SLS melt pool and the rate limiting step for mass
transfer is volatilization at the pool surface.
Evaporation of alloying elements from the melt
pool also causes loss of input laser power. The
power loss due to evaporation of an alloying ele-
ment is given by
[29]
P
L;A
= C
A
(L
A
DH
A
) (18)
where P
L,A
is the power lost due to evaporation of
element A, C
A
is the evaporation rate of element A,
L
A
is the latent heat of vaporization of pure compo-
nent A and DH
A
is the partial molar heat of mixing
of A in the alloy. The evaporative power loss affects
the peak temperature attainable in the melt pool.
3.3. Vaporization Characteristics of Elements
Relevant to SLS
Empirical data obeying the form of Eqs. 14 and
15 have been used by Dushman
[34]
to tabulate data
for a wide variety of elements. The vapour pressure data for
selected elements of relevance to SLS are shown in Table 1.
Some important observations can be made in this table
with regard to SLS process control. For example, in the
10
5
10
4
Torr vacuum range, all elements in the table except
Al and Pb have vaporization temperature lower than their re-
spective melting points. Hence, all these elements will show a
very high tendency to vaporize when processed under this
vacuum range. The vaporization temperature approaches
and exceeds the melting point in the 10
3
10
2
Torr range. In
the 10
1
1 Torr range, most vaporization temperatures are in
excess of the melting points but not by a significant degree.
This range allows melt superheat by a few hundred degrees
Celcius before the vaporization temperature is attained. How-
ever, it should be noted that even at 1 Torr, Cr, Cu, Mg, Mn
will all vaporize. Therefore, in order to inhibit vaporization
during SLS, processing should be conducted under partial
pressure instead of high vacuum.
3.4. Reducing Mean Free Path by Partial Pressure
Processing
Volatilization of alloying elements has been observed dur-
ing SLS processing under high vacuum. The most deleterious
effect of this phenomenon is condensation of volatilized ele-
ments on the laser window, severely affecting consistency of
Table 1. Vaporization temperature of selected elements as a function of pressure.
Element T
m
( C) T
b
( C) Vaporization temperature at Various Pressures (Torr)
10
5
10
4
10
3
10
2
10
1
1
Al 659 2447 882 972 1082 1207 1347 1547
B 2027 3927 1687 1827 1977 2157 2377 2657
Co 1495 2877 1162 1262 1377 1517 1697 1907
Cr 1903 2665 1062 1162 1267 1392 1557 1737
Cu 1084 2578 942 1032 1142 1272 1427 1622
Fe 1539 2857 1107 1207 1322 1467 1637 1847
Mg 650 1104 287 330 382 442 517 612
Mn 1244 2051 697 767 852 947 1067 1227
Mo 2577 4827 1987 2167 2377 2627 2927 3297
Ni 1452 2839 1142 1247 1357 1497 1667 1877
Pb 328 1751 487 551 627 719 832 977
Si 1415 2787 1177 1282 1357 1547 1717 1927
Ti 1660 3287 1321 1431 1558 1703 1877 2083
V 1857 3377 1432 1551 1687 1847 2037 2287
W 3377 5527 2547 2757 3007 3297 3647 n/a
Zn 420 906 208 246 290 342 405 485
Zr 1852 4405 1837 2002 2187 2387 2647 2977
T
m
=melting point T
b
=boiling point
R
E
V
I
E
W
S
laser energy delivery. Condensation of volatilized elements
also takes place on chamber interior surfaces that can cause
contamination during subsequent processing of other materi-
als. Volatilization can also cause depletion of alloying ele-
ments resulting in alteration of alloy chemistry.
In the absence of laser spot temperature control, the melt
zone is likely to undergo some degree of superheat. There-
fore, it is necessary to select the operating pressure based on
the elemental constituents that have the highest vaporization
tendency. In order to maintain purity of the processing atmo-
sphere, this necessitates evacuating the SLS processing cham-
ber to high vacuum followed by purging the process chamber
with ultrahigh purity Argon or Helium (Nitrogen is avoided
due to the tendency for formation of undesirable nitrides dur-
ing processing). It is necessary to implement inert gas partial
pressure processing to decrease the mean free path of gaseous
element atoms so that condensation of volatile elements on
the laser window is inhibited. In addition, if a continuous
flow of inert gas is maintained to provide dynamic partial
pressure processing, volatilized material will tend to be
flushed away from the processing zone and out of the vacu-
um system.
The mean free path is defined as the average distance be-
tween successive collisions of gaseous molecules and is given
by
[33]
k =
kT
2
_
Pp
2
(19)
where k is Boltzmann's constant, T is the absolute tempera-
ture, P is the pressure and r is the molecular diameter. As-
suming r = 5 A, T = 298 K and P is expressed in Pa, Eq. 23
reduces to
k = 3:624 10
3
P (20)
Table 2 shows the mean free path as a function of pressure
over 10 decades. Upon inspection of Table 2, the reason for
substantial condensation of elemental volatiles on the laser
window and chamber walls is evident. The distance r from
the laser window to the scanned powder surface is typically
less than 0.5 m. At 10
7
Torr vacuum levels, when the mean
free path is about 36 m, volatilized metal atoms travel unim-
peded to the laser window and processing chamber walls. At
one order of magnitude lower vacuum, 10
6
Torr, the mean
free path of 3.59 m is still an order of magnitude larger than r.
It is only at 10
5
Torr that the mean free path is of same order
of magnitude as r. Thus, if the arrival rate of volatilized metal
atoms at their condensation on the laser window is to be sub-
stantially curtailed, it is necessary to operate at still higher
pressure levels. We assume that the arrival rate is substan-
tially reduced by decreasing the mean free path of volatilized
atoms to two orders of magnitude lower than r. This hypoth-
esis suggests that the desired operating vacuum (pressure)
level is in the neighborhood of 7.6 10
3
Torr. This hypothesis
was indeed confirmed by experiment. While substantial con-
densation was observed during SLS processing at 10
4
Torr
vacuum, condensation on the laser window was nearly com-
pletely eliminated at 10
3
Torr levels, allowing the fabrication
of large complex parts spanning several hundred layers and
several tens of hours uninterrupted SLS processing time.
3.5. Alloy Melt Refining by Evaporative Deoxidation
Melting of high performance alloys under high vacuum or
ultrahigh purity inert gas partial pressure enviroments can
provide the added benefit of alloy refinement by volatiliza-
tion of oxides, impurities and non-metallic inclusions. Refin-
ing of alloy melts during SLS processing under vacuum can
occur by mechanisms very similar to those in vacuum induc-
tion melting (VIM), vacuum arc remelting (VAR) and electron
beam cold hearth refining (EBCHR). The techniques VIM,
VAR and EBCHR are presently well established for produc-
tion of aerospace and specialty alloys including nickel base
superalloys and alloys of titanium
[36]
. From the perspective of
SLS process control, it is important to consider the refining
mechanisms and their impact on the chemistry of the pro-
cessed material.
If the stoichiometric form of a metal oxide is M
x
O
y
, then
the atomic ratio of oxygen to metal in the vapor is given by
O
M
=
P
yP
M
x
O
y
P
xP
M
x
O
y
(21)
where P
MxOy
is the partial pressure of the metal oxide M
x
O
y
gaseous species. The oxygen to metal ratio in the metallic
phase is defined as the ratio of mole fractions of oxygen and
metal. At low concentrations of oxygen, this ratio reduces to
the mole fraction X
0
of oxygen in the metal. The ratio R is de-
fined by Brewer and Rosenblatt
[37]
as
R =
O
M

vapor
O
M

metal
(22)
Table 2. Mean free path at various pressures.
Pressure (atm) Pressure (Torr) (m)
1 760 3.59 10
8
10
1
76 3.59 10
7
10
2
7.6 3.59 10
6
10
3
7.6 10
1
3.59 10
5
10
4
7.6 10
2
3.59 10
4
10
5
7.6 10
3
3.59 10
3
10
6
7.6 10
4
3.59 10
2
10
7
7.6 10
5
3.59 10
1
10
8
7.6 10
6
3.59
10
9
7.6 10
-7
35.9
R
E
V
I
E
W
S
If R > 1, vaporization of the metal phase will lower the
oxygen content of the base metal. Brewer and Rosenblatt state
that substantial purification (>50 %) will only occur if R > 10
since for R = 10, vaporization of 10 % of the metal results in
50 % reduction of oxygen impurity. Their data for various
metals is shown in Table 3.
Inspection of Table 3 reveals that the elements Si, Ge, Pb
and Sn will undergo high purification by evaporative loss of
oxide. However, the second group of metals which are of
most interest to SLS (consituents of steel, titanium and nickel
alloys) will not undergo substantial evaporative deoxidation.
Therefore, in order to meet specifications on the chemistry of
such alloys while processing by SLS under vacuum or ultra-
high purity inert gas partial pressure environments, it is im-
portant to monitor not only the initial chemisty of the feed-
stock material, but also to prevent further oxygen
contamination, the integrity and leak tightness of the process-
ing environment must be monitored carefully. On the other
hand, all the refractory metals except Zr have very high val-
ues of R and will thus undergo high purification.
4. Summary and Conclusions
This article presents some of the important physical mech-
anisms in SLS processing of metals. This understanding is
crucial to machine design and process control development.
The following conclusions can be made with respect to suc-
cessful direct SLS processing of metals.
First, due to their high reactivity at melting temperature,
most metals will readily form oxides even under very low
partial pressure of oxygen. The layered fabrication nature of
SLS is very exacting in that oxides on surfaces of previously
fabricated layers and in the laser processed melt strongly in-
hibit wetting, resulting in poor bonding between layers, and
balling or beading across a layer. Good wetting requires that
SLS processing should be undertaken in a high vacuum envi-
ronment or ultrahigh purity inert gas environments. How-
ever, these environments alone do not ensure complete inter-
layer bonding and good wetting. In SLS of monolithic metals
and alloys, homologous wetting is the principal mechanism
for wetting of liquid onto a substrate of its own kind. This is a
non-equilibrium situation involving simultaneous fluid flow,
heat transfer and solidification. To ensure complete bonding
(full density, continuous interface) between successively
fabricated layers, epitaxial solidification must take place by
heterogenous nucleation off the grains of the substrate (pre-
vious layer). This mode of solidification necessitates remelt-
ing a portion of the underlying substrate so as to break down
surface impurity films and to provide intimate contact at the
atomic level between the solidifying liquid and the partially
melted grains of the underlying substrate.
Although processing under high vacuum yields desirable
results with respect to wetting, most metals of interest vapor-
ize due to reduction in vaporization temperatures with de-
creasing vapor pressure. Vaporization of elements results in
condensates on the laser window that adversely affect laser
energy transmission and consistency of energy delivery to
the powder surface. Therefore, SLS pro-
cessing should be conducted under ultra-
high purity partial pressure inert gas en-
vironments with continuous flow.
Operating under such environments ad-
dresses the problem of elemental volatil-
ization in three ways. First, increasing
the residual gas pressure in the SLS pro-
cessing chamber raises the vaporization
temperature, allowing for some degree of
superheat above the melt temperature
before significant volatilization takes
place. Second, the increased residual gas
pressure decreases the mean free path of
volatilized atoms, inhibiting their unim-
peded travel from the melt pool to the la-
ser window condensing surface. Third,
the continuous flow of inert gas flushes
volatilized material away from the pro-
cessing zone.
If the starting metal powders are con-
taminated by oxides, laser processing
these metals under vacuum or ultra-high
purity inert gas environment can lead to
purification by evaporative deoxidation
mechanisms. However, it is noted that
the metals that are of most interest to SLS
Table 3. Vaporization behavior of selected elements, after Brewer and Rosenblatt [37] (1962).
Element Evaporating Oxide T(K) P
M
(atm) X
0
(sat) R (low X
0
)
Oxygen Alloys of 4
th
group metals
Si SiO 1685 10
6.3
10
3.7
10
8
Ge GeO 1210 10
9
10
2.5
10
9.5
Sn SnO 1000 10
10
10
2
10
6.3
Pb PbO 1000 10
4.8
10
3.1
10
2.6
Oxygen Alloys of Transition Metals
Ti TiO 1940 10
5.2
0.3 1
V VO 2190 10
4.4
0.3 10
Cr CrO 2171 10
2.2
10
2
10
0.5
Mn MnO 1516 10
2.8
10
2
10
4
Fe FeO 1810 10
4.5
10
2.2
10
0.4
Co CoO 1765 10
5.1
10
2.1
10
Ni NiO,O 1726 10
5.4
10
2.1
10
Oxygen alloys of refractory metals
Zr ZrO 2125 10
7.7
10
Nb NbO 2740 10
5.8
10
3
Mo Mo
3
O
9
, MO
3
, MO
2,
O 2880 10
4.3
10
6
Ta TaO 3250 10
5
10
5
W W
3
O
9
, WO
3
, WO
2
, O 3650 10
4
10
6
R
E
V
I
E
W
S
for engineering components will not undergo such purifica-
tion due to thermodynamic reasons. Therefore, the chemistry
of starting powder materials as well as the integrity of the
SLS processing environment are critical for achieving chemi-
cal purity of the finished product.
Received: May 25, 2003
[1] S. Das, M. Wohlert, J. J. Beaman, D. L. Bourell, in Ad-

vances in Powder Metallurgy & Particulate Materials, Metal
Powder Industries Federation, Princeton, NJ, 1997,
p. 2167.
[2] S. Das, Ph.D. Thesis, University of Texas at Austin, 1998.
[3] S. Das, M. Wohlert, J. J. Beaman, D. L. Bourell, JOM
1998, 50(12), 17.
[4] S. Das, J. J. Beaman, M. Wohlert, D. L. Bourell, Materials
& Design 1999, 20(2-3), 115.
[5] S. Das, T. Fuesting, G. Danyo, L. E. Brown, J. J. Beaman,
D. L. Bourell, Materials & Design 2000, 21(2), 63.
[6] D. L. Bourell, M. Wohlert, N. Harlan, S. Das, J. J. Bea-
man, Adv. Eng. Mater. 2002, 4, 663.
[7] J. C. Nelson, Ph.D. Thesis, University of Texas at Austin,
1993.
[8] M. M. Sun, Ph.D. Thesis, University of Texas at Austin,
1991.
[9] M. Kandis, T. L. Bergman, J. Heat Transfer 1997, 119,
824.
[10] R. M. Tarafdar, T. L. Bergman, J. Heat Transfer 2002,
124, 553.
[11] A. E. Tontowi, T. Childs, Rapid Prototyping J. 2001, 7,
180.
[12] D. E. Bunnell, Ph.D. Thesis, University of Texas at Aus-
tin, 1995.
[13] Y. Zhang, A. Faghri, J. Heat Transfer 1998, 120, 883.
[14] Y. Zhang, A. Faghri, Int. J. Heat Mass Transfer 1999, 42,
775.
[15] Y. Zhang, A. Faghri, C. W. Buckley, T. L. Bergman,
ASME J. Heat Transfer 2000, 122, 150.
[16] M. Shiomi, A. Yoshidome, F. Abe, K. Osakada, Int. J.
Machine Tools Manufacture 1999, 39, 237.
[17] D. R. Gaskell, Introduction to the Thermodynamics of Mate-
rials, Taylor, Francis, New York 1995, p. 370.
[18] N. Bredsz, C. C. Tenenhouse, in Source Book on Brazing
& Brazing Technology, ASM, 1980, p. 8690.
[19] A. Choudhry, Vacuum Metallurgy, ASM International,
Warrendale 1990, 101.
[20] B. Engel, D. L. Bourell, Rapid Prototyping J. 2000, 6, 97.
[21] P. G. De Gennes, Rev. Mod. Phys. 1985, 57(3), 827.
[22] S. Schiaffino, A. A. Sonin, Physics of Fluids 1997, 9, 2217.
[25] A. J. Wall, D. R. Milner, J. Inst. Metals 1961, 90, 394.
[26] K. Easterling, Introduction to the Physical Metallurgy of
Welding, Butterworth-Heinemann 1992, p. 66-69.
[27] . Grong, Metallurgical Modeling of Welding, The Insti-
tute of Materials 1994, p. 222227.
[28] O. Winkler, R. Bakish, Vacuum Metallurgy, Elsevier
1971, chapter III.
[29] A. Block-Bolten, T. W. Eagar, Metall. Trans. 1984, 15B.
[30] M. M. Collur, A. Paul, T. Debroy, Met. Trans. B, 18B
1987, 733.
[31] R. Bakish, Introduction to Electron Beam Technology, Wi-
ley, 1962, chapter 6.
[32] J. Herbertson, R. Harris, in Electron Beam Melting & Re-
fining-State of the Art 1985, Bakish Materials Corpora-
tion, 2, p. 2638.
[33] R. Glang, Handbook of Thin Film Technology, chapter 1,
McGraw-Hill, 1970.
[34] S. Dushman, Scientific Foundations of Vacuum Technique,
chapter 10, Wiley, 1962.
[35] D. R. Poirier, G. H. Geiger, Transport Phenomena in Ma-
terials Processing, chapter 15, TMS 1994, p. 560567.
[36] JOM 1998, TMS, 12-35.
[37] L. Brewer, G. M. Rosenblatt, Trans Metall. Soc. AIME
1962, 224, 1268.
______________________

701 FTP

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

701 FTP

Uploaded by

Copyright:

Available Formats

R

[*] Dr. Suman Das

[1] S. Das, M. Wohlert, J. J. Beaman, D. L. Bourell, in Ad-

You might also like