You are on page 1of 9

Applied Materials Today 9 (2017) 341–349

Contents lists available at ScienceDirect

Applied Materials Today


journal homepage: www.elsevier.com/locate/apmt

In situ absorptivity measurements of metallic powders during laser powder-bed


fusion additive manufacturing
Johannes Trapp a , Alexander M. Rubenchik b , Gabe Guss b , Manyalibo J. Matthews b,∗
a
Technische Universität Dresden, 01062 Dresden, Germany
b
Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, CA 94550, USA

a r t i c l e i n f o a b s t r a c t

Article history: The effective absorptivity of continuous wave 1070 nm laser light has been studied for bare and metal
Received 29 June 2017 powder-coated discs of 316L stainless steel as well as for aluminum alloy 1100 and tungsten by use of
Received in revised form 16 August 2017 direct calorimetric measurements. After carefully validating the applicability of the method, the effective
Accepted 18 August 2017
absorptivity is plotted as a function of incident laser power from 30 up to ≈540 W for scanning speeds
of 100, 500 and 1500 mm s−1 . The effective absorptivity versus power curves of the bulk materials typi-
Keywords:
cally show a slight change in effective absorptivity from 30 W until the onset of the formation of a recoil
Additive manufacturing
pressure-induced surface depression. As observed using high-speed video, this change in surface mor-
3D printing
Selective laser melting
phology leads to an increase in absorption of the laser light. At the higher powers beyond the keyhole
Powder bed fusion transition, a saturation value is reached for both bare discs and powder-coated disks. For ≈100 ␮m thick
Optical absorption powder layers, the measured absorptivity was found to be two times that of the bare polished discs for
Laser keyhole low-laser power. There is a sharp decrease when full melting of the powder tracks is achieved, followed
Metal powder by a keyhole-driven increase at higher powers, similar to the bare disc case. It is shown that, under con-
Metal liquid absorptivity ditions associated with laser powder-bed fusion additive manufacturing, absorptivity values can vary
greatly, and differ from both powder-layer measurements and liquid metal estimates from the literature.
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction and materials. High-fidelity models describing heat transport and


hydrodynamics include the laser melting of the metal powder, the
Laser powder-bed fusion additive manufacturing (PBFAM), also melt pool structure, and hydrodynamics under the effect of capil-
known as Selective Laser Melting (SLM), or Laser Beam Melting lary forces, recoil pressure, and evaporation (see e.g., review [8] or
(LBM) produces metal parts through layer by layer deposition and [9,10]). The thermal histories predicted by these models will ulti-
selective fusion of the powder layers using a laser beam [1]. PBFAM mately inform thermodynamic and microstructural models, which
is not limited to metals and has been applied to polymers [2] and can be related to final part properties and performance. However,
ceramics [3] as well. The unique design freedom offered by this in practically all models describing the interaction of the laser beam
so-called 3D printing of metals offers industries such as aerospace, with the powder-layer surface, the absorption of laser light is typ-
medical and automotive [4] or power [5] new tools to improve part ically treated as constant, based on either calculated or measured
production. Some of the positive aspects of this process are off- values for polished clean metal plates [11,12] as well as rough
set by its complexity. It has been indicated previously [6,7] that at [13,14] and oxidized surfaces [11,14]. Recently, the absorptivity
least 157 parameters could affect the final manufacturing process, for several metal powders was measured at low homologous tem-
thus establishing a challenging phase space over which process peratures [15] and calculated using a ray tracing approach [16].
optimization is to be explored. Only the optimal selection of these Traditional direct measurements of the absorptivity at higher tem-
parameters results in manufactured parts that are free of defects, peratures use the collection of reflected light via an integrating
and minimizes residual stresses and distortions. sphere [17–19] and are difficult and expensive. In practice, once the
Numerical modeling provides a powerful tool for estimating temperature of the metallic surface heated by the laser reaches the
the optimal set of system parameters for specific part geometries melting point, the laser absorption is a complex process including
a variety of physical effects [20]. The melt surface interaction with
the laser is non-stationary and modulated due to the melt motion,
∗ Corresponding author. which can effect the absorptivity [21]. At higher intensities, when
E-mail address: ibo@llnl.gov (M.J. Matthews). the vapor recoil pressure is strong enough to create a deep surface

http://dx.doi.org/10.1016/j.apmt.2017.08.006
2352-9407/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
342 J. Trapp et al. / Applied Materials Today 9 (2017) 341–349

Fig. 1. View of the micro-calorimetry apparatus inside of the process chamber (a); a drawing of the sample holder made from porous Al2 O3 with a mounted cup (b) as well
as a magnified cross-section of the cup filled with powder with given dimensions (c).

depression in the melt pool – the keyhole – light interacts with the For measurements using bare discs, metal sheets were
keyhole walls at steep angles and with the ejected vapors. A frac- laser cut to discs of 10-mm diameter. The absorptivity of
tion of the net absorbed energy is lost to evaporative cooling and a powders was measured using cups as depicted in Fig. 1(c)
part of the laser light can be directly absorbed in the vapor plume that were precision machined from the discs with a rim
or in the plasma through inverse Bremsstrahlung absorption [22]. height of 100 ␮m, which was used to define the thickness
An adequate and self-consistent model formulation of these com- of the powder layer such that the influence of the powder-
plex effects is extremely difficult and remains a challenge to current layer thickness [25,26] is not explicitly considered in this
modeling efforts. It would therefore be highly beneficial to directly study.
measure net absorptivity for typical PBFAM materials to allow accu-
rate modeling of the laser energy coupling and subsequent physical
processes. 2.2. Experimental method
In this study, direct calorimetric measurements of the effective
laser absorptivity are presented in the temperature range from A key design component of the calorimetric setup is careful
below the melting point to the boiling point of metal materials thermal isolation of the sample during measurements. A sample
by performing single- and multi-track experiments and measuring holder custom made from porous alumina (0.14 W(m K)−1 ther-
the resulting temperature in the specimen for different scanning mal conductivity at 25 ◦ C) by Foundry Service & Supplies, Inc.
velocities, including conditions of the typical PBFAM process. After (Ontario, CA) as shown in Fig. 1(a) and (b) was designed to minimize
presenting the experimental methods used and the steps taken the conductive cooling of the sample well below the tempera-
to ensure reliability of the results, changes in absorptivity as a ture where convective heat loss dominates. Type K thermocouples
function of laser power for a fixed scanning speed are presented with bare wires of 76-␮m diameter each were spot welded to the
and evaluated for stainless steel 316L surfaces. After taking into rear side of the discs/cups: one thermocouple centered and one
account the impact of the scanning speed, the effects of the powder about three millimeter away from the center. Thin gold foils of
layer on the effective absorption of the laser light are demon- 20-␮m thickness were used as an interlayer to weld the thermo-
strated. High-speed videography and optical metallography are couples to the aluminum alloy discs. The wires were then insulated
used to further support the identification of keyhole formation. using alumina tubes. For the experiments with powder, a pow-
Finally, first results for aluminum alloy 1100 and tungsten are pre- der layer with a thickness of about 100 ␮m defined by the rim
sented and the expected effective absorptivity for the laser and height of the cups was spread on the cups using a steel razor
scan parameters typically used in commercial PBFAM systems is blade.
discussed. The experiments were carried out in a custom-built chamber
[21] with a suit of diagnostic tools with access to the melt pool as
well as the ability to apply a controlled atmosphere and adjust the
2. Methods and materials flow rate of the protective argon gas used. The flow rate of argon
was set to a fixed value of 2150 and 715 SCCM for the bare discs and
2.1. Materials the powder samples, respectively, both resulting in a gas pressure
slightly above atmospheric pressure due to the pressure limiting
Rolled metal sheets of 316L stainless steel, tungsten (99.95% valve used.
purity) (both 0.5-mm thick) and aluminum alloy 1100 (0.6-mm A 600 W Yb-fiber laser was focused to a 60 ± 5 ␮m 1/e2
thick) were obtained from Goodfellow Ltd. (Huntington, United diameter Gaussian spot and scanned using x–y galvanometer scan-
Kingdom). The 316L stainless steel powder used was the com- ning mirrors. Focusing of the laser beam was achieved using
mercially available alloy 20ES [23] from Concept Laser GmbH a convex–concave lens pair integrated into the scanner sys-
(Lichtenfels, Germany). The mean particle size of the spherically tem (Nutfield, model 3XB). This optical system was designed
shaped powder particles was 29.9±9 ␮m. Detailed characterization to closely approximate that of a typical commercial PBFAM
of the powder can be found in [24]. system.
J. Trapp et al. / Applied Materials Today 9 (2017) 341–349 343

compensate for temperature losses before temperature equilibra-


tion over the disc. These losses are primarily caused by convection
as the heat lost through the thin thermocouple wires and the low
conductivity sample holder is negligible. The total time of energy
deposition was in the millisecond range, much shorter then the
temperature equilibration time.
With known disc temperature increase, the effective absorptiv-
ity can be calculated by dividing the energy necessary to uniformly
heat the specimen from starting temperature T0 to end temperature
T1 (EH ) by the energy input due to the laser irradiation (EL ):
 T1
EH T0
mCp (T ) dT
Aeff = = (1)
EL l/vP
whereby m is the mass of the disc including the powder layer (if
applicable), P is the nominal laser power, v the scanning speed
and l the total length of the laser track (6 mm for a single track
line). The temperature-dependent heat capacity over the temper-
Fig. 2. Typical temperature measurement curves used to evaluate net energy
absorption during laser scanning experiments. The data shown here correspond to ature range measured (25 < T < 150 ◦ C) is given by the function
Cp (T) = Cp,0 (1 + ˛T) with Cp,0 and ˛ as shown in Table 1.
 disc.
temperature measurements of two thermocouples for a 316L steel  The dashed
line represents a least-squares fit to the function T (t) = T1 exp −t
t1/2
+ T0 with t1/2 The calculation of the absorptivity from the temperature differ-
being the time constant for cooling (the two fit curves exactly overlap). The insert ence is only valid, if the extrapolated cooling curve is not affected
shows practically identical effective absorptivity values for one to five tracks as by the position of the tracks on the disc or the total power input,
validation of the method. Multiple measurements for the same number of tracks i.e., the maximum temperature reached. This was tested by repeat-
correspond to different track patterns or positions, respectively. A trigger signal
ing experiments with the same laser power and scanning speed for
indicates the time t1 .
one to five tracks and for different positions of single tracks on the
disc (front, center, rear) or different patterns of multiple tracks (all
High-speed videos of the tracks were captured to better under- tracks in the center, two tracks at the front of the discs plus two
stand the absorption results. To magnify the track enough to see tracks at the rear side). The absorptivity calculated for the different
details in the melt pool, a 10× Mitutoyo long working distance experiments was found to be practically independent on the track
microscope objective was used. The objective could not be incor- pattern, position and number of tracks as shown in the insert in
porated into the chamber, so the videos were not captured during Fig. 2.
the absorption measurements. Instead, the chamber was removed Typically each set of parameters is measured twice, right after
and the experiments were repeated in a sample holder that allowed each other on the same specimen. The reported absorptivity is then
placing a Shimadzu HPV-X high-speed camera, Navitar microscope the mean value and the error bars indicate the higher and lower
body and objective at a 30◦ angle to the laser beam. Argon flow over values of the two measurements. For the bare disc experiments, 4
the sample was maintained for comparable conditions. The hori- stainless steel specimens have been used with a total of 149 tracks,
zontal field of view imaged onto the cameras sensor was 550 ␮m for aluminum 2 specimens (62 tracks) and for tungsten 2 specimens
wide, and due to the 30◦ viewing angle, a roughly 100-␮m region (85 tracks). The influence of the powder layer was studied using 14
was in focus, centered in the image vertically. A Cavilux Smart HF specimens with a total of 76 tracks.
light source was injected coaxially through the microscope body. After the experiments, top-view optical microscopy images
Movies were captured at 20 kHz, with the light source pulsed at and (for the samples with a powder layer) height profiles of the
100 ns to freeze the motion of the melt pool. A narrow band-pass fil- tracks were obtained using a Keyence model VX100 Laser scanning
ter centered on the 808 nm wavelength light source eliminated the microscope. The specimens were then cross-sectioned through the
influence of incandescent light, so that all variation in (reflected) tracks perpendicular to the scanning direction and electrochemi-
light originating from the melt pool can be attributed to shape cally etched with oxalic acid after embedding in epoxy, grinding,
changes. and polishing, so that the former melt pool could be clearly distin-
The laser scans produced single or multiple tracks on the disc guished.
or powder surface. The temperature time curve was measured for
a total time of 60 s with a measurement frequency of 14 Hz using 3. Results
thermocouples. A typical result of the measurements is presented
in Fig. 2. 3.1. Absorptivity of bare stainless steel discs
The time t1 − t0 for thermal equilibration via thermal diffusion
across the thin discs used was up to 12 s for the steel disc, the mate- At first, the laser absorption measurements with flat steel discs
rial with lowest thermal conductivity used. An exponential decay in the absence of metal powder are presented. The absorptivity
function was fitted to temperature data of each thermocouple for dependence on power for a scanning speed of 500 mm s−1 is shown
t > t1 and extrapolated the temperature to the time t0 corresponding in Fig. 3(a). The curve can be divided into three regimes. At low
to the time the tracks were created. This procedure was chosen to laser power, the absorptivity slightly increases with power from

Table 1
Physical properties of the materials used. Cp,0 denominates the specific heat capacity at 0 ◦ C, ˛ the temperature coefficient of Cp , and k and  are the thermal conductivity
and the mass density at room temperature, respectively. Tm is the melting point and Tb is the boiling point.

Material Cp,0 , J(g K)−1 ˛, 10−5 K−1 k, W(m K)−1 , g cm−3 Tm , ◦ C Tb , ◦ C Tb , eV

316L steel [31] 0.48 37.7 15 7.95 1410 ≈2800 0.27


Al 1100 [30] 0.89 51.6 209 2.7 640 2519 0.24
W [30] 0.13 15.9 159 19.3 3422 5555 0.5
344 J. Trapp et al. / Applied Materials Today 9 (2017) 341–349

Fig. 4. Flat surface absorptivity of 316L stainless steel as a function of laser power
for scanning velocities of 100, 500 and 1500 mm s−1 .

to an increasing absorptivity for 500 mm s−1 and even steeper


increase for 1500 mm s−1 . Increasing scanning speed also has the
effect of stretching the entire curve along the power-axis such that
the onset of rapidly increasing absorptivity (regime II) is shifted
to a higher power as well. For the fastest scanning speed, the
saturation value of the absorptivity is apparently not reached.
The main reason is that for 1500 mm s−1 an incident laser power
of about 540 W is only sufficient to reach a keyhole depth of
207 ␮m, well below that which would be attained at the satu-
Fig. 3. (a) Absorptivity versus laser power for 316L stainless steel discs measured
ration absorptivity level. Additionally, due to optical component
with a scanning speed of 500 mm s−1 . (b) and (c) show the optical metallography inertial effects, the scanning mirrors start to move the laser beam
cross-section and a top view of the tracks, both with the laser power correspond- about half a millisecond after the laser is turned on. During
ing to each track indicated. The sharp increase in absorptivity corresponds to the this time and for high laser power, the keyhole depth already
beginning of the keyhole formation starting at a laser power of about 70 W. The
exceeds the thickness of the disc, i.e., the laser initially penetrates
absorptivity saturates when the keyhole reaches a depth of about 300 ␮m.
through the disc for a short time. Thus, data points correspond-
ing to v = 1500 mm s and P > 300 W are expected to contain some
0.29 to 0.31 (regime I). At an apparent critical power threshold, a error.
sharp increase in absorptivity is observed to values over two times
that at low power (regime II), and finally it saturates at about 0.78
(regime III).
3.2. High-speed video
To understand the physics behind the absorptivity changes, an
optical microscopy image of the etched tracks cross-sectioned per-
Cross-sections of the tracks show a sharp increase in melt depth
pendicular to the scanning direction is presented in Fig. 3(b). The
when the absorptivity rises from 0.3 to 0.78. The molten region in
evolution of the melt pool shape as a function of laser power can be
the cross-sections is the final frozen melt pool shape, but it does
compared to the absorptivity curve in Fig. 3(a). Two different melt
not reveal how such a deep region of melting occurred. In Fig. 5(a)
pool geometries can be seen. For low power, a hemi-spherical shape
(see video 1 in supplementary material online), taken at a laser
can be observed that is a result of thermal conduction-controlled
power of 34 W and scanning speed of 500 mm s−1 , the track is dark,
melting. The depth of the melt pool increases with increasing laser
indicating a flat surface profile. This corresponds to the flat sur-
power in regime I. Coinciding with the onset of rapidly increas-
face absorptivity. In Fig. 5(b) (video 2), taken at a laser power of
ing absorptivity, the melt pool depth increases and a keyhole is
68 W and scanning speed of 500 mm s−1 , the track is bright on
formed. The growth in depth of the keyhole correlates with fur-
one half, indicating that it is raised above the plate surface. Fur-
ther increases in absorptivity (regime II, deep laser penetration
ther, three times during the video a bright ring is observed in
regime). After reaching a depth of about 300 ␮m, the absorptivity
the advancing region of the track, shown in Fig. 5(c) (video 2 as
saturates (regime III). These findings generally agree with the evo-
well). This indicates a depression in the melt pool – the keyhole
lution of absorption mechanisms with increasing power density as
– that traps light and leads to the increase in laser absorption.
discussed in Grigoryants et al. [20].
These parameters are at the transition of two regimes, where the
keyhole is unstable. Note how quickly it appears at its full size of
3.1.1. Influence of the scanning speed 30 ␮m and then disappears with the transitions occurring over one
Fig. 4 shows the power-dependent absorptivity for a few differ- movie frame of 50 ␮s. Fig. 5(d) (video 3), taken at a laser power
ent scanning speeds. Qualitatively, the power dependence of the of 270 W and scan speed of 500 mm s−1 , represents the saturated
absorptivity is almost the same as that presented in Fig. 3(a). With absorptivity of 0.78. Now the depression is three times as large,
increasing scanning speed, the behavior in the low-power regime and rippling and modulations of the track behind the melt pool are
I changes from a slight reduction of absorptivity for 100 mm s−1 visible.
J. Trapp et al. / Applied Materials Today 9 (2017) 341–349 345

Fig. 5. Examples of tracks indicating (a) a flat surface profile for P = 34 W and v = 500 mm s−1 , (b) slightly raised melt surface on the front edge of the molten track, (c) small
depression in the melt pool (both for P = 68 W and v = 500 mm s−1 ) and (d) well-developed keyhole for P = 270 W and v = 500 mm s−1 .

Fig. 6. Absorptivity of 316L steel with a 100-␮m powder layer compared to the absorptivity of a bare disc at v = 100 mm s−1 (a) and 1500 mm s−1 (b).

3.3. Absorptivity of metal powder-coated substrate aluminum show a lower effective absorptivity at low power than
steel and the onset of increasing absorptivity is shifted to higher
The absorptivity for the powder-coated substrate is presented power.
in Fig. 6 for v = 100 mm s−1 and 1500 mm s−1 . At low power,
the absorptivity of the powder is about two times higher then
the flat surface absorptivity. With increasing laser power, the 4. Discussion
absorptivity dramatically decreases. As soon as the recoil pres-
sure starts to form a keyhole, the absorptivity again increases To discuss the different regimes and the influence of scanning
quickly. The power dependence of the absorptivity for the powder- speed and material parameters on the absorptivity, it is crucial
coated case at high intensities is qualitatively similar to the flat to first know the peak temperature reached on the track sur-
2
plate. Numerically, the results can be different depending on face. Considering a Gaussian beam I = I0 exp(− 2r2 ) moving with

laser beam parameters. This will be discussed in the following speed v along a flat metal surface, the peak temperature increase
sections. (in the center of the laser beam and on the surface of metal,
thus r = 0 and z = 0) caused by a single laser pulse of duration t
3.4. Aluminum alloy and tungsten can be calculated after Bäuerle [27], modified for the 1/e2 beam
diameter:

In Fig. 7, the results for aluminum and tungsten discs are IA 8Dt
present in comparison with the steel data for a scanning speed TAl,W (r = 0, z = 0, t) = √ tan−1 . (2)
k  2
of 1500 mm s−1 . The results are qualitatively similar with dif-
ferences in the absolute value of the effective absorptivity as Using the dwell time /v for the time t and I = 2P2 for the intensity

well the onset of increasing absorptivity. Both tungsten and I of the laser, the peak temperature increase is:
346 J. Trapp et al. / Applied Materials Today 9 (2017) 341–349

Fig. 8. Top view of the laser tracks produced on an aluminum sample for a scanning
speed of 1500 mm s−1 with increasing laser power from the bottom to the top. At
Fig. 7. Absorptivity of aluminum alloy 1100, tungsten and 316L stainless steel discs low power (<100 W), no melt pool is generated. The transition to the keyhole regime
as a function of laser power for a scanning speed of 1500 mm s−1 . is observed between 250 and 400 W.

Table 2
Laser power for the surface temperature to reach the melting or boiling point of corresponding to a 1-␮m thick (and 100-␮m wide times 6-mm
the material, calculated after Eq. (3) for scanning velocities of 100 (steel only) and long) slab of steel. Another possible reason for a decreasing absorp-
1500 mm s−1 .
tivity is the absorption or reflection of laser light in the vapor plume
Steel 100 mm s−1 Steel 1500 mm s−1 Aluminum Tungsten generated before the light hits the metallic surface. As the plume
Melting power 12 21 125 286
is more and more tilted when the scanning speed is increased [33]
analytic and interaction between nominal incidenting laser light and vapors
Boiling power 26/24 42/41 473/107 446/185 is therefore reduced, this could explain the different behavior for
analytic/FEM different scanning speeds observed for the steel samples.
Melting <30 <30 100 ± 15 120 ± 10
As for tungsten and aluminum, the lowest power scans do not
observed
Boiling/keyhole 48 ± 10 108 ± 20 265 ± 15 275 ± 25 lead to melting (see Fig. 8 and Table 2), the difference in the absorp-
observed tivity of (a) solid and liquid metal and (b) an oxide-coated surface
compared to a pure metallic surface can add more to the changes
 in absorptivity. This is especially relevant for aluminum with the
 
 2PA 8D high absorptivity of the oxide compared to a bare metallic surface
TAl,W r = 0, z = 0, t = = √ tan−1 . (3) [14], even though this effect is considered to be small as the discs
u k 3 u
were ground before the experiments.
As the calculation after Eq. (3) does not take into account the tem- It is also apparent that the absolute value of the effective absorp-
perature dependence of the material parameters nor the limited tivity extrapolated to zero power matches the literature value,
thickness of the discs, these values were also derived numerically indicated by the circles on the y-axis, only for steel. The extrap-
by the finite element method using COMSOL (see Guss et al. [28] olated zero power absorptivity for aluminum is higher, and that
and Matthews et al. [29] for a more detailed description; material for tungsten lower than expected. While a higher value can be
parameters used were taken from [30,31]). The power to reach the explained by enhanced light absorption caused by the lower refrac-
melting and boiling points of the materials used calculated after Eq. tive index of the oxide layer [11,14], the effects causing the lower
(3) and using the FE-method are given in Table 2 for the following than expected absorptivity for tungsten still have to be investi-
discussion. gated. A possible factor might be the volatility of the tungsten oxide
Bare disc curves: regime I. The changes in absorptivity are small WO3 leading to a fast evaporation at a comparably low homologous
in this regime and vary between growing by 15% for the steel sam- temperature.
ples and scanning speeds of 500 and 1500 mm s−1 , decreasing by a Bare disc curves: regime II. According to the optical microscopy
comparable value for the steel samples with low-scanning velocity cross-sections, the onset of increased absorptivity correlates to the
as well as the tungsten samples and a constant value for aluminum. beginning deep penetration regime (the keyhole regime). There-
According to Eq. (3), the temperature in the melt pool grows fore, the onset of regime II can be compared with the power
linearly with the laser power. For an absorptivity of A = 0.3 for required to reach the boiling point, which is a necessary prerequi-
steel, the bulk metal surface starts to melt at a power of ≈10 W for site for the keyhole formation [34], given in Table 2 for the materials
100 mm s−1 and P ≈ 28 W for 1500 mm s−1 , so that the tracks melt, used.
as observed, under all conditions investigated. Literature data typ- After the peak temperature exceeds the boiling temperature for
ically show a higher absorptivity for the flat liquid metal surface the metal, the recoil pressure of the ejected vapors starts to form
than for the solid one [32]. Even though for the steel samples, a a keyhole depression. The multiple laser light scattering on key-
melt pool exists even for the lowest power and highest scanning hole walls results in enhanced light absorption [35–37]. This is
speed, a larger melt pool for increasing laser power (see also the consistent with the observed steep increase in absorptivity. The
melt pool width given in Fig. 3) should lead to an increased absorp- power when the temperature reaches the boiling point as common
tion. On the other hand, heat loss due to evaporation of material threshold for keyholing is depicted together with the power of the
reduces the absorbed energy. With the enthalpy of evaporation experimentally observed onset of increased absorptivity (that cor-
(starting from 25 ◦ C using a value for pure iron of 399 kJ mol−1 responds well to the onset of keyhole formation seen in the metal-
[30]), the reduced absorptivity for the 100 mm s−1 curve could be lographic images) in Fig. 9. Eq. (3) correctly predicts the threshold
explained by evaporation of about 4 ␮g of steel from each track power increasing with scanning speed, but numerically, the
J. Trapp et al. / Applied Materials Today 9 (2017) 341–349 347

Fig. 10. 3D height profile measured using a laser scanning confocal microscope
showing a growing fraction of the track molten with increasing power from 30 to
150 W.

Fig. 9. Calculated and experimentally observed threshold for keyhole formation for
the steel samples as a function of the square root of the scanning speed. The neces- the boiling point and experimentally observed keyhole threshold
sary power calculated is always lower than that of the experimentally observed.
are almost identically. An important feature is nevertheless that

the cooling by evaporation does not affect the v scaling of peak
predicted threshold is about 2 ± 0.2 times lower than observed. temperature and laser power. It is interesting to note that this
One reason is that the recoil pressure must first overcome the sur- additional laser power due to capillarity translates to an additional
face tension of the molten track. This additional power needed normalized enthalpy as previously discussed by King et al. [34] and
to create the keyhole is much more pronounced than in welding can explain at least some of the differences in the measured versus
experiments, as the laser beam size used in this experiments (and predicted conduction-to-keyhole transition point for steel 316L.
for PBFAM in general) is smaller than that used in laser welding Bare disc curves: regime III. For laser powers above the keyhole
and therefore the melt pool curvatures are smaller and internal threshold, the fast absorptivity growth eventually saturates below
pressures are larger, scaling as /a (where  is the specific surface 1, as part of the radiation is absorbed by the vapors and, due to
energy and a the curvature of the melt pool, which scales with the the Gaussian distribution of the laser light intensity, only a part of
beam size , so that a =  can be used). This effect can be estimated the laser intensity hits the keyhole. The value of the absorptivity at
using the recoil pressure given by the expression [27] saturation of about A ≈ 0.7–0.8 for steel is consistent with results
p
 1 1
 from welding studies [36,39]. For all materials the tracks made with
= K exp  − (4) a scanning speed of 1500 mm s−1 , the saturation level seems not
p0 Tb Ts
yet to be reached. Despite this, as an increasing scanning speed
where p0 is the atmospheric pressure, K a constant of about 0.55 leads to a more shallow keyhole [33], it is reasonable to expect the
taking into account the momentum carried away by the vapor,  the absorptivity for the highest scanning speed to saturate at a lower
evaporation enthalpy per atom (for steel the value for Fe of 4.3 eV is level.
used), Tb the boiling point and Ts the surface temperature, whereby Influence of the powder layer. The presence of a powder layer on
both temperatures are given in eV. To overcome the surface tension, top of the substrate increases the absorptivity due to the multiple
the surface temperature must (theoretically) exceed Tb by T: reflections and scattering, as has recently been studied numerically
T

T
 2  −1 [16] and measured experimentally [15] for low laser power and
= 1 − b ln −1 (5) corresponding track temperatures below the melting point of the
Tb  Kp0
materials used. The results at low power presented in this paper
Taking the surface tension for steel as ≈ 1.4 Nm−1 [38] leads to are consistent with those results.
T/Tb ≈ 0.05. The linear dependence of the peak temperature on Increasing the laser power leads to melting of the powder layer,
the laser power (Eq. (3)) leads to the estimate that the increase in which is associated with a sharp decrease of the absortpivity to
required power is 5% only. Therefore, the very efficient local cooling levels similar to that of the bare discs. Fig. 10 elucidates this by
via evaporation must be taken into account additionally. If all the showing the height profile of the tracks remaining on the cups after
energy exceeding that to reach the boiling point would be used for removing the loose powder. The increase in the fraction of powder
evaporation, the recoil pressure would scale linear with the laser being molten with increasing power is consistent with the decrease
power. Consequently, the threshold for keyhole power PK can be in absorptivity with increasing power measured calorimetrically.
calculated in terms of the power to reach the boiling point Pb , using Taking into account the lower thermal conductivity of the pow-
that the pressure at the boiling point is p = Kp0 . der layer, its reduced heat capacity due to the estimated packing
density of about 60%, and the initially higher absorptivity, the
PK 2
= (6) threshold for the keyhole formation of the powder track is reduced
Pb Kp0
compared to the value for the bulk material as can be seen in Fig. 6.
Using the parameters for steel as given above leads to PK /Pb ≈ 2, However, as a result of the particles’ minimal heat conduction in
which is consistent with the experimental data. These simple esti- the powder layer in the scanning direction, powder melts on the
mates neglect a lot of physics behind the process, especially the heat front edge of the beam. This means that the keyholing process is
conduction from the molten track that, other than for steel, plays an similar for a flat surface and a powder-coated one, which explains
important role for tungsten and aluminum. Taking into account the the similar behavior of absorptivity in this range of powers. The
very low surface tension of aluminum at the boiling point (about numerical value of absorptivity for a given power is yet differing for
0.28 Nm−1 [31]), the additional pressure to overcome surface ten- flat and powder-coated surfaces in the regime of a developed key-
sion is minimal, so that the numerically determined power to reach hole. The interesting fact for the powder-coated samples is that the
348 J. Trapp et al. / Applied Materials Today 9 (2017) 341–349

Fig. 11. Still frames showing the growing keyhole with increasing power for different scanning speeds. (a) 500 mm s−1 and 300 W, (b) 1500 mm s−1 and 500 W, (c) 500 mm s−1
and 150 W as well as (d) 1500 mm s−1 and 200 W. The measured absorptivity for (b) and (d) is almost identical.

saturation value of the effective absorptivity is lower than that of set of data for other materials than stainless steel and varying
the bare disc value for 100 mm s−1 but higher for 1500 mm s−1 . A powder-layer thickness will be discussed in a later paper. Still there
closer look at the results shows that the saturation value is almost are other parameter such as beam size and shape [40] or powder
constant for the powder-coated samples but differs for the bare characteristics [41] that will influence the process as well. Both the
dics. This is a result of the contradicting effects of (a) the powder modeling and experimental communities should benefit from the
entrainment in the laser beam [33] resulting in beam scattering methods presented here to gain further insights to the process of
by the powder particles and (b) a still increased absorptivity of the PBFAM.
laser light in the powder layer alongside the molten track as a result From the viewpoint of additive manufacturing practice, it can
of the Gaussian laser intensity distribution. Also, while for a bare be seen that operating with higher laser powers can improve the
disc a more perfect keyhole is formed, particles colliding with the process efficiency by increasing the effective absorptivity. Even
melt deform the melt pool, reducing the change in absorptivity seen more, working with a higher scanning speed reduces the sensi-
for the flat discs, which is caused by the varying keyhole depth. The tivity of the process to deviations in the process parameters by
differences in melt pool morphology for flat and powder-coated eliminating the sudden change of the aborptivity during transi-
samples for 500 and 1500 mm s−1 are shown in Fig. 11 and can tion to the keyhole regime. The latter has been shown earlier
better be seen in the supplementary videos 4–7. by plotting the density of a built part against the power for
different scanning speeds [42], where for a scanning speed of
5. Conclusions 1500 mm s−1 , a density over 99% is reached in the power range
between 250 and 400 W, as lower power leads to incomplete
The implications of our work are primarily to show the very melting and higher power to keyhole porosity. From Fig. 6(b),
strong absorptivity change as a function of laser and material one can see that this range corresponds to an absorptivity
parameters, which up to now have not been assessed quanti- value of about 0.6, almost twice the flat surface absorptivity
tatively. Direct calorimetric measurements are presented as a value. A similar result was also observed for lower scanning
powerful tool to gather pressingly necessary accurate data for the speeds.
effective absorptivity of laser light in PBFAM as well as to further As a complete melting of the powder-layer track and the sub-
improve the understanding of the complex physics of the process. strate underneath is necessary to achieve full bonding in AM, and
On the one hand, the results match expectations and data published the low density and thermo-isolation of the single particles ease the
earlier, especially melting of the powder layer, powder absorptivity plays a minor role
for the absorption properties close to or in the keyhole regime. Our
(a) the high absorptivity of the powder layer as long as it is not results even show the attenuating effects of the powder layer on
molten the influence of the scanning speed on absorptivity compared to
(b) the absorptivity of powder-coated and flat surface being quali- the bare discs.
tatively comparable once the powder on the track is completely The experimental data on effective absorptivity will help to
molten and improve the fidelity of the process modeling and can be used
(c) the increasing absorptivity once a keyhole if formed. to validate energy coupling models. The direct calorimetric mea-
surements can be used to assess conduction versus keyhole mode
On the other hand, it becomes obvious that the process differs regimes when establishing a new process (beam size, scanning
for different scanning speeds and can change substantially when speed, laser power, environment, etc.) or feedstock (material, pow-
other materials are used. This becomes apparent for the transition der size distribution), that can even be used when fast evaluation
point from conduction to keyhole mode, that cannot be estimated of absorption properties of less known materials are needed, e.g.
by peak temperature calculations without taking into account for additive manufacturing in space missions [5]. The measure-
local evaporation and liquid metal surface tension. An extended ments for solid substrate have an additional value in that they are
J. Trapp et al. / Applied Materials Today 9 (2017) 341–349 349

relevant to micro-welding conditions and should be of interest to [16] C. Boley, S. Mitchell, A. Rubenchik, S. Wu, Metal powder absorptivity:
the welding community in general. modeling and experiment, Appl. Opt. 55 (2016) 6496–6500.
[17] N.K. Tolochko, T. Laoui, Y.V. Khlopkov, S.E. Mozzharov, V.I. Titov, M.B. Ignatiev,
Absorptance of powder materials suitable for laser sintering, Rapid Prototyp.
Acknowledgments J. 6 (3) (2000) 155–160.
[18] R.P. Martukanitz, R.M. Melnychuk, S.M. Copley, Dynamic absorption of a
powder layer, in: Proceedings of 23rd International Congress on Applications
The authors thank Wayne King, Sheldon Wu and Nicholas Calta of Lasers and Electro-optics, vol. 97, 2004, pp. 1404–1409.
for helpful discussion as well as Sharon Torres for the prepara- [19] R.W. McVey, R.M. Melnychuk, J.A. Todd, R.P. Martukanitz, Absorption of laser
tion of the metallographic samples. This work was funded through irradiation in a porous powder layer, J. Laser Appl. 19 (4) (2007) 214–224.
[20] A.G. Grigoryants, Basics of Laser Material Processing, Mir Publishers/CRC
a Laboratory Directed Research and Development grant 15-ERD- Press, Moscow/Boca Raton, 1994.
037 and performed under the auspices of the U.S. Department of [21] M.J. Matthews, G. Guss, S.A. Khairallah, A.M. Rubenchik, P.J. Depond, W.E.
Energy by Lawrence Livermore National Laboratory under contract King, Denudation of metal powder layers in laser powder bed fusion
processes, Acta Mater. 114 (2016) 33–42.
DE-AC52-07NA27344. Johannes Trapp received funding by a grant
[22] M. von Allmen, A. Blatter, Laser-Beam Interactions with Materials: Physical
of the Graduate Acadamy of the Technische Universität Dresden Principles and Applications, vol. 2, Springer Science & Business Media, 2013.
covered by the German Excellence Initiative. [23] http://www.conceptlaserinc.com/wp-content/uploads/2014/10/111123 CL-
20ES.pdf.
[24] Knowledge Based Process Planning and Design for Additive Layer
Appendix A. Supplementary data Manufacturing (KARMA), Materials Datasheets (Ti6Al4V, ASTM F75 Cobalt
Chrome, CL20, CL50, DSM Somos 18420, DSM Somos Next, PA 2200 and
PA3200), Technical Report, Karma, Valencia, 2010.
Supplementary data associated with this article can be found,
[25] M.M. Savalani, J.M. Pizarro, Effect of preheat and layer thickness on selective
in the online version, at http://dx.doi.org/10.1016/j.apmt.2017.08. laser melting (SLM) of magnesium, Rapid Prototyp. J. 22 (1) (2016) 115–122.
006. [26] S.L. Sing, F.E. Wiria, W.Y. Yeong, Selective laser melting of lattice structures: a
statistical approach to manufacturability and mechanical behavior, Robot.
Comput. Integr. Manuf. 49 (2018) 170–180.
References [27] D.W. Bäuerle, Laser Processing and Chemistry, Springer, Berlin, Heidelberg,
2013.
[1] ASTM International, Standard Terminology for Additive Manufacturing [28] G.M. Guss, A.K. Sridharan, S. Elhadj, M.A. Johnson, M.J. Matthews, Nanoscale
Technologies, ASTM International, West Conshohocken, PA, 2010. surface tracking of laser material processing using phase shifting diffraction
[2] W.S. Tan, C.K. Chua, T.H. Chong, A.G. Fane, A. Jia, 3D printing by selective laser interferometry, Opt. Express 22 (June (12)) (2014) 14493–14504.
sintering of polypropylene feed channel spacers for spiral wound membrane [29] M.J. Matthews, S.T. Yang, N. Shen, S. Elhadj, R.N. Raman, G. Guss, I.L. Bass, M.C.
modules for the water industry, Virtual Phys. Prototyp. 11 (3) (2016) 151–158. Nostrand, P.J. Wegner, Micro-shaping, polishing, and damage repair of fused
[3] S.L. Sing, W.Y. Yeong, F.E. Wiria, B.Y. Tay, Z. Zhao, L. Zhao, Z. Tian, S. Yang, silica surfaces using focused infrared laser beams, Adv. Eng. Mater. 3 (2014).
Direct selective laser sintering and melting of ceramics: a review, Rapid [30] W.F. Gale, T.C. Totemeier (Eds.), Smithells Metals Reference Book, Elsevier
Prototyp. J. 23 (3) (2017) 611–623. Butterworth-Heinemann, Oxford, Burlington, 1992.
[4] T.T. Wohlers, Wohlers Report 2016 3D Printing and Additive Manufacturing [31] K.C. Mills, Recommended Values of Thermophysical Properties for Selected
State of the Industry Annual Worldwide Progress Report, Wohlers Associates, Commercial Alloys, Woodhead Publishing, 2002.
Inc., Fort Collins, CO, 2016. [32] N.R. Comins, The optical properties of liquid metals, Philos. Mag. 25 (4) (1972)
[5] A. Goulas, J.G.P. Binner, R.A. Harris, R.J. Friel, Assessing extraterrestrial regolith 817–831.
material simulants for in-situ resource utilisation based 3D printing, Appl. [33] S. Ly, A.M. Rubenchik, S.A. Khairallah, G. Guss, M.J. Matthews, Metal vapor
Mater. Today 6 (2017) 54–61. micro-jet controls material redistribution in laser powder bed fusion additive
[6] I. Yadroitsev, Selective Laser Melting, LAP Lambert, Saarbrücken, 2009. manufacturing, Nat. Sci. Rep. 7 (1) (2017) 4085.
[7] O. Rehme, Cellular Design for Laser Freeform Fabrication, Cuvillier Verlag, [34] W.E. King, H.D. Barth, V.M. Castillo, G.F. Gallegos, J.W. Gibbs, D.E. Hahn, C.
2010. Kamath, A.M. Rubenchik, Observation of keyhole-mode laser melting in laser
[8] W.E. King, A.T. Anderson, R.M. Ferencz, N.E. Hodge, C. Kamath, S.A. Khairallah, powder-bed fusion additive manufacturing, J. Mater. Process. Technol. 214
A.M. Rubenchik, Laser powder bed fusion additive manufacturing of metals; (12) (2014) 2915–2925.
physics, computational, and materials challenges, Appl. Phys. Rev. 2 (4) [35] P. Solana, G. Negro, A study of the effect of multiple reflections on the shape of
(2015) 041304. the keyhole in the laser processing of materials, J. Phys. D: Appl. Phys. 30 (23)
[9] S.A. Khairallah, A.T. Anderson, A. Rubenchik, W.E. King, Laser powder-bed (1997) 3216.
fusion additive manufacturing: physics of complex melt flow and formation [36] R. Fabbro, Melt pool and keyhole behaviour analysis for deep penetration
mechanisms of pores, spatter, and denudation zones, Acta Mater. 108 (2016) laser welding, J. Phys. D: Appl. Phys. 43 (44) (2010) 445501.
36–45. [37] G.G. Gladush, I. Smurov, Physics of Laser Materials Processing, vol. 146,
[10] C. Panwisawas, C. Qiu, M.J. Anderson, Y. Sovani, R.P. Turner, M.M. Attallah, Springer, Berlin, Heidelberg, 2011.
J.W. Brooks, H.C. Basoalto, Mesoscale modelling of selective laser melting: [38] T. Hibiya, K. Morohoshi, S. Ozawa, Oxygen partial pressure dependence of
thermal fluid dynamics and microstructural evolution, Comput. Mater. Sci. surface tension and its temperature coefficient for metallic melts: a
126 (2017) 479–490. discussion from the viewpoint of solubility and adsorption of oxygen, J.
[11] B. Karlsson, C.G. Ribbing, Optical constants and spectral selectivity of stainless Mater. Sci. 45 (8) (2010) 1986–1992.
steel and its oxides, J. Appl. Phys. 53 (9) (1982) 6340–6346. [39] U.D. Bello, C. Rivela, M. Cantello, M. Penasa, Energy Balance in High-Power
[12] F. Dausinger, J. Shen, Energy coupling efficiency in laser surface treatment, ISIJ CO2 Laser Welding, vol. 1502, 1991, pp. 104–116.
Int. 33 (9) (1993) 925–933. [40] T.T. Roehling, S.S.Q. Wu, S.A. Khairallah, J.D. Roehling, S.S. Soezeri, M.F. Crumb,
[13] D. Bergström, J. Powell, A.F.H. Kaplan, The absorption of light by rough metal M.J. Matthews, Modulating laser intensity profile ellipticity for
surfaces – a three-dimensional ray-tracing analysis, J. Appl. Phys. 103 (10) microstructural control during metal additive manufacturing, Acta Mater. 128
(2008) 1–12. (2017) 197–206.
[14] A.M. Rubenchik, S.S.Q. Wu, V.K. Kanz, M.M. LeBlanc, W.H. Lowdermilk, M.D. [41] A.T. Sutton, C.S. Kriewall, M.C. Leu, J.W. Newkirk, Powder characterisation
Rotter, J.R. Stanley, Temperature-dependent 780-nm laser absorption by techniques and effects of powder characteristics on part properties in
engineering grade aluminum, titanium, and steel alloy surfaces, Opt. Eng. 53 powder-bed fusion processes, Virtual Phys. Prototyp. 12 (1) (2017) 3–29.
(12) (2014) 122506. [42] C. Kamath, B. El-Dasher, G.F. Gallegos, W.E. King, A. Sisto, Density of
[15] A. Rubenchik, S. Wu, S. Mitchell, I. Golosker, M. LeBlanc, N. Peterson, Direct additively-manufactured, 316L SS parts using laser powder-bed fusion at
measurements of temperature-dependent laser absorptivity of metal powers up to 400 W, Int. J. Adv. Manuf. Technol. 74 (1–4) (2014)
powders, Appl. Opt. 54 (24) (2015) 7230. 65–78.

You might also like