High Strength and Ductility of Additively

Manufactured 316L Stainless Steel Explained

MD. SHAMSUJJOHA, SEAN R. AGNEW, JAMES M. FITZ-GERALD,
WILLIAM R. MOORE, and TABITHA A. NEWMAN

Structure–property relationships of an additively manufactured 316L stainless steel were

explored. A scanning electron microscope and electron backscattered diffraction (EBSD)
analysis revealed a fine cellular-dendritic (0.5 to 2 lm) substructure inside large irregularly
shaped grains (~ 100 lm). The cellular structure grows along the h100i crystallographic
directions. However, texture analysis revealed that the main h100i texture component is inclined
by ~15 deg from the building direction. X-ray diffraction line profile analysis indicated a high
dislocation density of ~1 9 1015 m2 in the as-built material, which correlates well with the
observed EBSD microstructure and high-yield strength, via the traditional Taylor hardening
equation. Significant variations in strain hardening behavior and ductility were observed for the
horizontal (HB) and vertical (VB) built samples. Ductility of HB and VB samples measured 49
and 77 pct, respectively. The initial growth texture and subsequent texture evolution during
tensile deformation are held responsible for the observed anisotropy. Notably, EBSD analysis of
deformed samples showed deformation twins, which predominately form in the grains with
h111i aligned parallel to the loading direction. The VB samples showed higher twinning activity,
higher strain hardening rates at high strain, and therefore, higher ductility. Analysis of annealed
samples revealed that the observed microstructures and properties are thermally stable, with
only a moderate decrease in strength and very similar levels of ductility and anisotropy,
compared with the as-built condition.

https://doi.org/10.1007/s11661-018-4607-2
 The Minerals, Metals & Materials Society and ASM International 2018

I. INTRODUCTION with proper selection of laser processing parameters and

scanning strategies, it is possible to obtain near fully
LASER powder bed fusion is a type of laser-based dense parts.[2–4] The microstructural development in
additive manufacturing (AM) process during which AM materials is complex because of the complicated
metallic parts are fabricated from a 3D computer-aided thermal distribution. Each location in the component
drawing (CAD) file by selective melting of successive first undergoes rapid solidification, and then undergoes
layers of powder using a sharply focused laser beam, repeated reheating and cooling with an additional
e.g., selective laser melting (SLM), direct metal laser deposited layer. This complicated thermal history can
melting (DMLM). This ‘‘freeform’’ manufacturing result in microstructural heterogeneity in the as-built
method can produce extremely complex shapes relative part.[5]
to a traditional subtractive manufacturing method 316L stainless steel is one of the most widely
(machining).[1] The density, microstructures, and, con- investigated materials for AM. This is because 316L
sequently, the mechanical properties of the additively has a wide range of applications in marine, biomedical,
manufactured material are sensitive to laser processing nuclear reactor, chemical, and petrochemical industries
parameters such as laser power, scanning speed, and owing to its high strength, good ductility, and high
scanning strategies. Indeed, it has been reported that corrosion resistance.[6–8] Several studies have been
performed to optimize the laser processing parameters
to fabricate near fully dense 316L SS parts.[4,9–11] These
studies reported that the average strengths of the
MD. SHAMSUJJOHA, SEAN R. AGNEW, and JAMES M. optimized parts are higher than the counterpart as-cast,
FITZ-GERALD are with Materials Science and Engineering, hot-pressed (HP), or hot isostatic-pressed (HIP) 316L
University of Virginia, Charlottesville, VA 22904. Contact e-mail: stainless steel parts, while the ductility has been reported
[email protected] WILLIAM R. MOORE and TABITHA A.
NEWMAN are with the Naval Surface Warfare Center Dahlgren to be lower.[7,8,12] The microstructure in the as-built
Division, Dahlgren, VA 22448. 316L stainless steel materials is reported to be a
Manuscript submitted September 3, 2017. submicron cellular-dendritic solidification structure,
Article published online April 11, 2018

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3011

resulting from the high cooling rate inherent to AM plate and a layer thickness of 20 lm were maintained
process.[8,13,14] Many researchers have attributed the throughout the process. Figure 1 shows the component
higher strength of the as-built 316L stainless steel to the built on a base plate for this study: (a) illustrates how
refined microstructure and high dislocation density. the tensile samples, in accordance with ASTM E8,[25]
However, a quantitative connection between the dislo- were directly fabricated; (b) and (c) show the X axis of
cation density and the observed strength is yet to be the specimen is defined as the long transverse direction
made. (TD) and the Y axis as parallel to the building direction
Anisotropy of the mechanical properties, especially (BD). Selected AM samples were annealed in a vacuum
the ductility, with respect to as-built directions is still a furnace at 1100 C for 1 hour (HT1) and 13 hours
matter of controversy. For example, some research- (HT13), and cooled at a rate of 100 C/min.
ers[15–18] reported a higher ductility of the built material
when the sample is tested perpendicular to the building
direction relative to the samples tested along the B. Microstructural Characterization and Chemical
building direction. They attributed this low ductility to Analysis
the presence of defects along the melt pool boundary. Samples were observed using a Leica light optical
Carlton et al.[19] reported that even a very low level of microscope and an FEI Quanta 200 SEM equipped
porosity can significantly reduce the ductility of the with an AZtec energy dispersive spectrometer (EDS).
as-built parts. On the other hand, other authors[20–22] For microstructural analysis, the samples were sectioned
showed a higher ductility of samples tested along the using conventional band saw and slow-speed diamond
building direction. Tomus et al.[21] reasoned that the saw methods, mounted in Konductomet and mechan-
higher number of grain boundaries results in lower ically ground up to 1200 grit silicon carbide paper,
ductility of the samples deformed perpendicular to the polished with a diamond pastes of 3 and 1 lm, and
building direction. This apparent controversy will be finally repolished using 0.05 lm colloidal silica. Finally,
examined by taking into account of the specific pre- the samples were electro-etched at the room temperature
ferred crystallographic orientation (texture) that results in a solution of 10 g oxalic acid and 90 mL distilled
from AM processing.[23] The effects of the initial texture water at 6 VDC for 60 seconds. EDS analysis measured
and its evolution during deformation need to be the chemical composition of the material; however, EDS
considered. is only a semiquantitative measure of the chemical
The main objective of this study is to examine the composition and cannot yield compositional data for
microstructure and crystallographic texture that devel- light elements with high precision. Therefore, the
ops during AM and correlate the observed microstruc- compositions of light elements, such as carbon and
ture with mechanical properties determined by tensile nitrogen, were measured using EXTR Leco TCH600
testing perpendicular and parallel to the building according to ASTM E1019-11.[26]
direction. Electron-backscattered diffraction (EBSD) was per-
formed using an FEI Company Quanta 650 scan-
ning microscope equipped with an Oxford Instruments’
II. EXPERIMENTAL AND METHODS AztecHKL system, NordlysNano EBSD detector, and
HKL Channel 5.0 EBSD postprocessing software. The
A. Laser Powder Bed Fusion (LPF) and Heat microscopes were operated at 20 kV, at a working
Treatment distance ranging from 15 to 20 mm. The EBSD scans
The nitrogen-atomized spherical prealloyed 316L were performed at a step size ranging from 0.1 to
stainless steel powders acquired from the EOS product 0.7 lm. The fraction of points successfully indexed was
specification[24] were used as a starting material. Table I always greater than 95 pct. The Kernelaveraged misori-
lists the chemical composition of the EOS powders.[24] entation (KAM) obtained from the EBSD analysis was
Bulk material was fabricated using an EOS EOSINT used to estimate the geometrically necessary dislocation
M280 AM system at the Naval Surface Warfare Center, (GND) calculation. In this analysis, KAM was deter-
Dahlgren Division, Virginia, equipped with a 400-W mined up to the second-nearest neighbor. The misori-
continuous wave Nd:YAG fiber laser using the EOS entation values exceeding a threshold value of 2 deg
MS-1 speed parameter set, which includes specifications were excluded from the KAM calculation, since they
of a bidirectional scanning strategy wherein the scanning were most likely from adjacent grains. The KAM
vector is rotated by 70 deg between successive layers. To magnitude is affected by the step size of an EBSD scan,
minimize oxidation, laser-melting operations were con- and, therefore, only the scan performed at a step size of
ducted in a nitrogen environment. A stainless steel base 0.1 lm was used in the final analysis.

Table I. Chemical Compositions of the As-Built Parts Using the EDS and Leco Analyses for the Chemical Compositions of Heavy
and Light Elements (C, N, and S), Respectively

Cr Ni Mo C Mn S Si N Fe
As-built 20 16 2-3 0.006 1.00 0.005 0.4 0.049 bal.

3012—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

 Vercal build (VB)
(a) Transverse direcon (TD)
(c)

Horizontal build (HB)

Transverse direcon (TD)
(b)

Fig. 1—(a) The as-built components. The vertical and horizontal built samples are shown in (b) and (c), respectively.

C. Hardness and Tensile Tests conventional PANalytical X’Pert Pro MPD [multipur-
Vickers microhardness measurements were made pose diffractometer] X-ray diffractometer with a Cu-Ka
using a diamond pyramid indenter with an apical angle sealed-tube source operated at 45 kV and 40 mA.
of 136 deg, an indentation load of 0.5 kg, and an Similar sample surface preparation methods were used
indentation duration of 15 seconds. The measurement for the X-ray diffraction analysis as for microstructural
was performed on a face perpendicular to the building characterization.
direction, and the reported results are an average of at
least 10 indentation measurements per sample. Uniaxial 1. Phase identification and dislocation density
tensile tests were performed using an MTS 318.1 system The Bragg–Brentano focusing geometry was
with a nominal strain rate of 1 9 103 s1 at ambient employed to collect diffraction pattern over the 2h
temperature. A contactless MTS LX 500 laser exten- range from 30 deg–110 deg with a step size of 0.004 deg
someter was used to measure axial strain in the sample and counting time of 50 seconds per step for phase
gage upon loading. As mentioned above, these flat quantification and dislocation density measurement.
dog-bone-shaped samples were directly fabricated by The PANalytical X’Celerator linear position sensitive
AM, and both sides of the horizontal built samples are detector was used to collect the diffracted beam during
machined to remove the support structure to be consis- this h:2h scans. A programmable divergence slit of
tent (even though the support structure was only on one 0.5 deg was used for both the incident and diffracted
side). Both the machined and the as-received vertical beams. A sample spinner programmed to a revolution
built sample types were tested. Both showed identical time of 2 seconds was used to obtain better grain-sam-
mechanical properties, to within the uncertainty limits pling statistics. Instrumental broadening effects were
of the measurement, and therefore, no distinction is evaluated and corrected using a silicon standard.[27]
made between the two sample types. Both the modified Williamson–Hall (mWH)[28] and the
more sophisticated Convolutional Multiple Whole Pro-
file (CMWP)[29,30] methods were employed to investigate
the dislocation density present in the as-built and
D. X-ray Diffraction
annealed samples.
X-ray diffraction (XRD) was used to characterize the In the modified Williamson–Hall (mWH) method, the
phase content, dislocation density, and crystallographic full-width half-maximum (FWHM) obtained from sin-
texture of the as-built, annealed, and deformed samples. gle peak fitting of the diffraction data can be represented
All XRD measurements were performed using a by the following equation:

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3013

  2   average contrast factor value was used in both the
2 0:9 M 2 b2
ðDKÞ ¼ þ p qðK2 CÞ; ½1 CMWP and mWH methods.
D 2
where K is the diffraction vector defined by K ¼ 2 sin h 2. Texture measurement
k ,
D is the coherent domain size, b is the burgers vector Texture measurements were performed with the X-ray
(0.255 nm for FCC [face-centered cubic] iron), C is the sealed tube source oriented in point focus mode. A
average contrast factor, and M is a dimensionless 2 mm 9 1 mm cross slit was used to minimize defocus-
parameter related to the dislocation density and outer ing at higher v tilts. A graphite monochromator was
pffiffiffi placed in front of the point detector in order to block
cutoff radius Re by M ¼ Re q. The parameter M is
sensitive to the dislocations arrangement in the mate- fluorescence and Kb radiation. The (1 1 1), (2 0 0), and (2
rial. M > 1 indicates that dislocations are randomly 2 0) incomplete (v = 0 deg to 80 deg) pole figures were
distributed to form a screening strain field, while measured on a 5 deg 9 5 deg grid of azimuth (B) and
M < 1 indicates that dislocations are arranged into tilt (v) angles. An experimental defocusing curve was
low-energy dislocation structures (LEDSs). The values obtained from a randomly textured Ni powder sample.
of the average contrast factor for cubic crystals can be The background subtraction, defocusing correction,
defined as calculation of the complete orientation distribution
!! function (ODF), and recalculation of complete pole
h2 k 2 þ h2 l 2 þ k 2 l 2 figures were performed using the MATLAB MTEX
C ¼ Ch00 1  q 2
; ½2 library.
ðh2 þ k2 þ l2 Þ

where Ch00 is the average contrast factor of h00 reflection.

Ch00 can be calculated from the elastic constant of the III. RESULTS
crystal. The anisotropic elastic constants of the face-cen- A. Chemical Composition and Phase Identification
tered cubic (FCC) austenite are C11 = 210 GPa,
C12 = 130 GPa and C44 = 120 GPa.[31,32] These values The chemical compositions are listed in Table I,
were used to calculate the value of Ch00 using ANIZC which suggest that the composition of the as-built parts
computer program[33] for both edge (0.281) and screw is consistent with that of the powder supplier’s specifi-
dislocations (0.289). However, the average value of 0.285 cation.[24] (It is noted that the chemical composition of
for the mixed (i.e., equal proportion of edge and screw the starting powder was not measured during this
types) dislocations were used, since variations in contrast study.) A specific point to note here is that the N
factor are small and the consequences of this simplifying content of the as-built parts is within the powder
approximation are small. The q parameter was calculated supplier’s specification, which is contrary to several
by minimizing the difference of DK values between the studies[34–36] that reported the possibility of increased
measured and calculated ones from the linear fitting of nitrogen content for parts additively manufactured in a
the Eq. [2]. Additional detail describing the method used nitrogen environment.
can be found in Reference 28. The collected diffraction The X-ray diffraction analysis revealed that both the
patterns were analyzed for determining the lattice con- as-built and annealed parts were fully austenitic. No
stant, a, and FWHM using X’Pert High Score Plus traces of second phases (d-ferrite, a-ferrite or martensite)
software, assuming pseudo-Voigt peak shapes. It should were observed within the detection limit of X-ray
be noted here that the slope of the mWH plot is related to diffraction (See Figure 2). Saeidi et al.[37] also reported
both the M parameter and the square root of dislocation the formation of single-phase austenite during selective
density, and therefore, the dislocation density cannot be laser melting of 316L stainless steel. However, they
directly calculated from this method. reported the formation of ferrite after d-annealing heat
For the quantitative analysis of the dislocation treatment at 1100 C for one hour followed by furnace
density, the X-ray diffraction patterns were evaluated cooling and cited solute partitioning as responsible.
by the CMWP fitting procedure. In CMWP, the peak Similar to the current results, Sistiaga et al.[38] reported a
profile functions are calculated as the convolution of the single-phase austenite after a similar annealing heat
size, strain, and instrumental broadening. The measured treatment at 1095 C for 2 hours followed by water
diffraction pattern is then fitted with the calculated cooling. Finally, it is noted that due to a high initial
profile function using a nonlinear least squares dislocation density, the as-built samples have signifi-
method.[29,30] Although CMWP is one of the most cantly broader diffraction peaks, compared with the
sophisticated methods available for line broadening annealed samples (as shown in Figure 2).
analysis, it does involve a large number of fitting
parameters, and different combinations of initial values B. Microstructure and Crystallographic Texture
can yield different solutions associated with local min-
ima in the sum of the squared residuals. Therefore, the Figure 3(a) shows an optical micrograph of the
parameters obtained from mWH were used as initial as-built 316L stainless steel, constructed out of sections
guess values for the parameters in the CMWP fitting to create a 3D view. The as-built samples exhibit a
procedure. This procedure enabled the CMWP to reach typical layered microstructure, characterized by melt
the global minimum more efficiently. Note that the same pool boundaries generated along the laser scanning

3014—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

paths. It should be noted that the differently oriented in the orientation image maps, which confirms the
melt pool boundaries are the result of bidirectional epitaxial nature of resolidification. The dark black line
scanning strategy, i.e., 70 deg rotation between succes- in Figure 3(b) outlines the grain boundary (misorienta-
sive layers. However, in general, an arc-shaped melt tion angle is greater than 15 deg). The gradient in color
pool morphology was observed. Optical images show inside these coarse grains suggests the presence of
that coarse grains are growing inside these melt pool significant dislocation substructure. Closer SEM inspec-
boundaries. Since the previous layers are partially tion of the microstructure reveals the formation of
re-melted, heterogeneous nucleation on the metallurgi- cellular-dendritic solidification microstructure inside the
cal solid/liquid interface occurs easily, and new grains grain, as shown in Figure 4.
grow epitaxially in the direction of the maximum Figure 4(a) shows the cellular-dendrite structure with
temperature gradient. varying morphologies and growth directions inside the
EBSD analysis confirms the presence of coarse grains laser track (marked by red outlines) due to the variation
of up to 200 lm along the building direction, as shown in local solidification condition. AM is a dynamic
in Figure 3(b). No melt pool boundaries are noticeable process because the heat source is continuously moving.
As the laser beam moves away, the maximum temper-
ature gradients change direction. The growing columnar
crystals seek to follow the maximum temperature
gradients while maintaining their preferred growth
direction, h100i. The morphological difference in cell
structures depends on their growth direction with
respect to metallographic section. A higher magnifica-
tion image reveals that the cross-sectional dimensions of
the elongated intergranular cells are ranging from about
0.5 to 1 lm, as shown in Figure 4(b). Colonies of these
cells have the same crystallographic orientation and
belong to the same coarse grain, as revealed by the
EBSD analysis. Prior studies[8,13] have shown Mo
enrichment at the cell boundaries using transmission
electron microscopy (TEM).
Due to the preferred growth direction, the AM-built
part usually exhibit the so-called growth texture, i.e.,
h100i direction is parallel to building direction.[39]
Fig. 2—XRD patterns of the as-built and annealed samples Figure 5(a) shows the (100) pole figure of the as-built
confirming single-phase austenite.
316L stainless steel, where BD is at the center of the pole

Fig. 3—(a) The 3D optical micrograph of the as-built 316L stainless steel showing the laser tracks along the building direction and (b) the
cellular microstructure within the irregularly shaped grains, as indicated by the faint gray low-angle boundaries.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3015

 Build direcon

(a) (b)
Build direcon

(c) (d)

Fig. 4—SEM micrographs of the (a, b) front views and (c, d) top views of the as-built samples. The red solid lines outline the melt pool
boundaries.

figure and TD of the sample is vertical. The observed processes have occurred. Furthermore, a few misorien-
texture is distinct from that of published or other AM tation-free small grains were also observed, indicating
processes. There is a h100i fiber tilted ~ 15 deg away partial recrystallization having occurred. The macrotex-
from the BD, with a peak intensity of 2.8 times random, tures of the one-hour annealed (HT1) samples show no
and there is another h100i || TD texture component. The significant change. It still shows the near h100i fiber
inverse pole figures (IPFs) show that the h100i intensity texture, where (100) pole is about ~ 15 deg tilted with
is stronger along the TD than the BD, as shown in respect to the building direction (see Figure 7), which is
Figures 5(b) and (c), respectively. The type and strength another indication that little recrystallization has
of the texture developed during AM largely depend on occurred. This is consistent with the observation made
the scanning strategy.[23,39] The bidirectional scanning by Sistiaga et al.[38] Evidence of recrystallization and
strategy used in this study changes the heat flow grain growth was observed after annealing for 13 hours,
direction between layers, which inhibits the growth of as shown in Figure 6(c, d). Note the presence of larger,
very large columnar grains. This is why the observed more equiaxed grains and twin boundaries, R3 (60 deg
texture is not very strong. about h111i), indicated in red, which highlight the
Figure 6 shows the microstructures of the samples formation of annealing twins after the 13-hour anneal.
annealed at 1100 C for (a, b) 1 hour (denoted HT1),
and (c, d) 13 hours. After annealing for one hour, the
laser tracks and cellular-dendritic structures are no C. Dislocation Density
longer apparent (see Figure 6(b)). The EBSD analysis
also reveals that there is still misorientation present Figure 8 shows the mWH plot of the as-built and
inside these large grains, which suggests that recrystal- HT1 samples, in which the FWHM of each peak is peak
lization is not complete. However, the average misori- is plotted as a function of K2C, where K ¼ 2sinh
k and C is
entation inside the grains is smaller relative to the the average contrast factor. Accordingly, the FWHM is
as-built samples, an indication that dislocation recovery plotted as DK ¼ Dh 2cosh
k , where Dh is the FWHM of the

3016—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 5—The macrotexture (a) (100) pole, and (b, c) IPFs parallel to TD and BD obtained from the XRD measurement of the as-built sample.

diffraction peak. As seen from the Eq. [1], the mWH plot parameter M. Both the as-built and annealed materials
gives two important
 microstructural parameters. The show M < 1, i.e., M values are 0.17 ± 0.03 and
2 2
slope p M2b q is related to the microstrain and 0.56 ± 0.12, respectively, indicating that the disloca-
tions are arranged into low energy dislocation structures
dislocation
 density, whereas the intercept of the line (LEDS). The high dislocation density of the as-built
0:92 
D provides the size of the coherent scattering materials correlates well with microstructural observa-
domain (D). The larger slope and intercept of the tions of other researchers, who reported that a high
as-built samples indicates that these samples have higher concentration of dislocations exists at the cellular-den-
microstrain owing to higher dislocation density and dritic boundaries.[14,37,41,42] The short-term annealing
smaller subgrain (coherent scattering domain) size that results in a reduction in dislocation density without
relative to the annealed samples, as listed in 2. It should changing the subgrain structure at the micron scale was
be noted here that the dislocation density present in the also reported by Saeidi et al.[37] As usual, the use of the
samples annealed at 1100 C for 13 hours is below the coherent scattering domain-size determined using line
detection limit of the XRD-based line broadening broadening techniques as a proxy for the average
analysis. subgrain size yields a somewhat lower value than the
The line broadening analysis performed using the subgrain size observed in the SEM and EBSD images
CMWP procedure is shown in Figure 9 for HT1 sample. (see Figure 4). (In short, although the two quantities are
The green and red lines show the measured and fitted related, they are not of the same value.)
data, respectively, with the residual shown at the bottom In order to complement the XRD-based dislocation
of the figure. density calculation, the geometrically necessary disloca-
As seen from the inset (magnified (111) peaks), a good tion (GNDs) density also studied from the EBSD
fit is obtained. The area-weighted mean subgrain size analysis using the KAM method. By this approach,
the dislocation density can be estimated from the
(hXiarea ¼ mexpð2:5r2 Þ, where m is the median and r2 is
following formula[43]:
the log-normal variance of the crystalline size and
dislocation density (q), which are the two microstruc- 2hKAM
tural parameters obtained from the CMWP procedures, qGND ffi ; ½3
bd
and the values are listed in Table II.
Similar to the mWH method, a higher dislocation where b is the Burgers vector magnitude (0.255 nm for
density and smaller subgrain size were observed for the FCC iron) and d is the step size. Figure 10 shows the
as-built samples relative to the HT1 sample. As Table II KAM maps for the as-built and HT1 samples. A
indicates, the dislocation density of (1.18 ± 0.11) 9 significant increase in subgrain size (or decrease in the
1015 m2 of the as-built materials decreased to density of subgrain boundaries) can be visually
(2.4 ± 0.2) 9 1014 m2, and the subgrain size of observed, in the KAM maps obtained from the HT1,
190 ± 28 nm increased to 296 ± 80 nm after an anneal- which lends support to the notion that smaller subgrains
ing heat treatment for 1 hour. The uncertainties of the merge into larger subgrains within the bigger grain.[44]
dislocation density and subgrain size reported here were Table II presents the GND density of the as-built and
calculated using error propagation of the uncertainties HT1 samples calculated from the Eq. [3] revealing a
in the parameters of the CMWP fitting routine.[40] The 58 pct reduction in the GND density. Comparing the
dislocation arrangement can be inferred from the estimated GND values with total dislocation density

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3017

Fig. 6—Microstructures of the heat-treated AM parts annealed at 1100 C for (a, b) 1 h and (c, d) 13 h.

calculated from XRD line profile analysis, which properties of the as-cast 316L stainless, obtained from
accounts for both GNDs and statistically stored dislo- Reference 7. Each as-built sample condition exhibited
cations, it appears that GNDs account for about 50 and low scatter in the mechanical properties, as shown by
90 pct of the total dislocation density of the as-built and the small standard deviation listed in Table III. This
HT1 samples, respectively. indicates that the processing conditions adopted in this
study are capable of producing 316L stainless with
highly reproducible properties.
D. Mechanical Properties and Deformation Behavior The as-built samples have higher hardness, strength,
Figure 11 shows representative (a) engineering and ductility relative to the as-cast 316L stainless steel.
stress–strain and (b) true stress–strain of both the Both the horizontal (HB) and vertical (VB) as-built
as-built and HT1 samples tested along the building samples show a similar yield strength. However, the HB
direction (vertical built) and the transverse direction samples show a higher tensile strength and lower
(horizontal built), and the respective work hardening ductility (49 pct) relative to VB samples. The high
  hardness and strength values of the as-built materials
dr
rates h ¼ as a function of true strain are shown are in good agreement with the results reported in
de
Figure 11. Eight vertical and four horizontal as-built References 7, 8, 15, 17 and 20. However, the ductility of
samples were tested during this study, and the single AM 316L observed in the present study is significantly
curve presented should be taken as representative of higher than the Reference 7 as-cast material as well as
those multiple tests. However, only one vertical and one that reported in other studies of AM material, with the
horizontal sample from the HT1 material was tested. exception of Reference 20, in which they also reported
Average tensile properties such as yield strength, ulti- 70 pct elongation at failure for the vertical built samples.
mate tensile strength, uniform strain, and strain to However, no explanation for this high ductility was
failure were calculated from these plots are listed in provided. Annealing heat treatment at 1100 C for one
Table III, along with the Vickers microhardness and the hour reduced the hardness value of the as-built samples

3018—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 7—The macrotexture (a) (100) pole and IPFs parallel to TD (b) and BD (c) obtained from the XRD measurement of the samples annealed
at 1100 C for 7 h.

Fig. 8—Modified Williamson–Hall plot of an as-built sample as well as a sample annealed at 1100 C for 1 h. The FWHM of each peak plotted
as a function of K2C shows a higher dislocation density in the as-built material.

by only about 8 pct. A more significant reduction a sharp decrease in h value was observed, typical of
(23 pct) in hardness value was observed for samples materials in which the plastic deformation proceeds by
annealed for 13 hours (HT13) in which recrystallization slip. With an increasing strain, a nearly constant h value
is observed to have made significantly greater progress. (linear hardening) was observed up to a strain level of
HT1 samples show significant decreases in yield and 0.17 (Stage-II) followed by a final decrease in hardening
tensile strengths, but the strain hardening behavior and rate (Stage-III).
ductility remain largely unchanged. This multistage behavior is typical of materials that
In addition to the differences in tensile strength and exhibit deformation twinning-induced plasticity (i.e., the
ductility, the vertical and horizontal samples show a TWIP effect). Indeed, h vs e plot of the vertical samples
distinct work hardening behavior ðh vs eÞ (see show the typical hardening behavior observed in the
Figure 11(c)). The h vs e plot of the HB samples can be twin-induced plasticity (TWIP) steel[45–47] and can also be
characterized by three distinct stages. Stage-I ðe<0:09Þ, characterized by three distinct stages. Stage-I of VB

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3019

Fig. 9—The measured and fitted diffraction patterns obtained using CMWP of the as-built samples. The magnified view of the (111) peak shows
a good matching between the measured and simulated profile.

Table II. Subgrain Size and Dislocation Density Estimated from the Modified Williamson–Hall (mWH) and CMWP Methods,
Along with the Geometrically Necessary Dislocation (GND) Density Estimated Using the Kernel-Averaged Misorientation (KAM)
Data Obtained Using EBSD

Subgrain Size Microstrain Subgrain Size Dislocation Density Parameter (M), GND Density
Specimen (nm) (mWH) (mWH) (nm) (CMWP) (1014 m2) (CMWP) (CMWP) (1014 m2)
As-built 180 90 190 ± 28 11.8 ± 1.1 0.17 ± 0.03 5.3
HT1 500 50 296 ± 80 2.40 ± 0.19 0.56 ± 0.12 2.2

Fig. 10—Kernel-averaged misorientation map of (a) as-built and (b) HT1 samples.

samples again shows a rapid decrease in the hardening stage). With progress in deformation (Stage-III), h
rate up to a strain value of 0.09. Stage-II ðe<0:23Þ shows a remains essentially constant in the strain ranging from
sharp abrupt increase of the h value (i.e., primary twining 0.23 to 0.41. Above this strain level, the sample has

3020—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 11—Representative (a) engineering stress–strain and (b) true stress–strain plots of the as-built and one-hour annealed samples; (c, d) show
the work hardening rate (h)–true strain plot of the as-built and one-hour annealed samples, respectively.

Table III. Mechanical Properties, e.g., Yield Strength, Ultimate Tensile Strength, Total Elongation, Uniform Elongation, and
Microhardness, of the Investigated As-built and One-Hour Annealed Materials. For Comparison, Mechanical Properties of an
As-cast Reference Material Are also Listed

Yield
Strength Ultimate Tensile Engineering Strain Uniform Uniform True Strain Vickers
Samples (MPa) Strength (MPa) to Failure True Strain (Considère Analysis) Microhardness
As-built (horizontal) 584 ± 11 667 ± 15 0.49 ± 0.03 0.23 ± .03 0.22 199 ± 6.5
As-built (vertical) 588 ± 20 622 ± 27 0.77 ± 0.03 0.37 ± 0.02 0.41
Annealed one-hour 339 568 0.50 0.29 0.28 184 ± 3.02
(horizontal)
Annealed one-hour 322 495 0.81 0.43 0.42
(vertical)
As-cast[7] 200 450 0.45 165 ± 7.15

undergone plastic instability. Therefore, the stress and the dynamic refinement of the microstructure by fine
strain levels reported at higher strains are only approx- twins, which continually introduce new barriers to
imate, and the intrinsic strain hardening response cannot dislocation motion, as explained by Bouaziz et al.[47]
be readily ascertained. Note that the onset of necking is Figure 11(d) suggests that the work hardening behavior
well predicted by the conventional plastic dr instability
 (h vs e) of the annealed VB and HB samples are similar to
analysis based on Considère’s criterion de ¼ r . Mech- the hardening behavior of the as-built VB and HB
anistically, stages II and III are understood to result from samples, respectively.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3021

Fig. 12—(a, b) LD-IPF maps and (c, d) calculated Taylor factor of the horizontal and vertical built samples strained to 0.18 true strain. Loading
direction is horizontal.

EBSD analysis of the sample deformed to a true strain with other orientations contain twins. This is true for
value of 0.18 revealed deformation twins ({111} h112i) both VB and HB samples. Several authors employed
in both the HB and VB samples (dark black features Schmid’s law to explain the dependence of twinning on
interior to grains in Figure 12). At this strain level, twins crystal orientation.[48,49] In a similar way, consideration
usually appear as bundles, and therefore, it was possible of the Taylor factor values of the investigated grains of
to index these nanoscale twins even with a relatively the HB and VB can be insightful (Figures 12(c) and (d)).
large step size of 0.4 lm. Figure 12(a, b) shows the IPF The grains with favorable orientation for twining
for the crystal direction along the tensile loading (h111i||LD) have the highest Taylor factor (M) value
direction (LD) for VB and HB samples, respectively. of ~ 3.65. However, twinning is also observed in the
The deformation texture is characterized by a weak grains having a value of M as low as 3.06. On the other
h111i||LD and a strong h100i||LD fiber. This texture hand, no twin activity was observed in the grains with a
evolution is atypical of TWIP steel, which usually shows value of M < 2.6 (h100i||LD).
a strong h111i||LD and a weak h100i||LD.[47] The macrotexture obtained from X-ray diffraction
Initial grain orientation has a strong effect on the measurement is consistent with the EBSD analysis.
activity of twinning. Deformation twinning mainly Figure 13 shows the inverse pole figures (IPFs) of HB
occurs in the grains that are closely orientated to the and VB samples strained to different levels. After
h111i||LD direction and only a small fraction of grains straining to e ¼ 0:18, HB samples show a relatively

3022—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 13—The IPFs parallel to tensile loading direction (LD) of initial (a), after being deformed at e ¼ 0:18 (b), and e ¼ 0:40 (failed samples) (c)
for horizontal built samples. IPF of Initial (d), after being deformed at e ¼ 0:04 (e), e ¼ 0:18 (f), and e ¼ 0:57 (failed samples) (g) of VB samples
are also shown.

Table IV. Comparison Between Yield Strength Calculated Using Taylor’s Equation (Complete with Uncertainty Propagation) and
Experimentally Measured Values

Dislocation
Taylor Density Yield Strength, Yield Strength
Samples a Factor (M) (q (m2)) (MPa (Predicted)) (MPa (Experimental))
As-built (horizontal) 0.26 ± 0.015 2.6 11.8 ± 1.1 9 1014 573 ± 35 584 ± 11
As-built (vertical) 588 ± 20
Annealed one-hour (horizontal) (2.40 ± 0.19) 9 1014 313 ± 15 339
Annealed one-hour (vertical) 322

stronger h100i||LD fiber and weaker h111i||LD fiber. No only 2.5 times random), which is also typical for TWIP
significant strengthening in texture was observed beyond steel.[46,47] Based on these observations, it appears that
this strain level. Texture measurement of the HB there was more twinning in the case of the VB sample
samples strained at e ¼ 0:40 (true strain at failure) than the HB sample.
exhibits a weaker h111i||LD and stronger h100i||LD
fiber texture. On the other hand, the texture evolution of
VB samples shows a distinct behavior. The initial
texture does not show any preferred crystal orientation IV. DISCUSSION
along the sample loading direction. The texture evolu-
tion up to e ¼ 0:18 is modest, only about two times A. Microstructure, Dislocation Density, and Thermal
random. However, samples deformed to e ¼ 0:57 (true Stability
strain at failure) show a stronger h111i||LD relative to The observed complex microstructure is rather typical
h100i||LD fiber, which is typical for TWIP steel. Still, of additive manufacturing. This is because the inherent
the texture evolution is rather weak (peak intensity of high cooling rate of the AM process, where the material

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3023

in the melt pool undergoes rapid solidification, at typical B. Higher Yield Strength
cooling rates in the range of 103 to 108 K/s.[7] Evolution It is already well accepted that the higher yield
of cellular solidification morphology can be understood strength of the AM-built material relative to the
using classical solidification theory. According to this conventional as-cast materials is the consequence of
theory, the solute will pile up ahead of the interface the refined microstructure and high density of disloca-
owing to the smaller solubility in the solid when the tions resulting from the rapid solidification. Several
equilibrium partition coefficient (kE) is less than one. researchers have invoked the Hall–Petch relation, i.e.,
Under these so-called constitutional undercooling con- the macroscopic yield strength is inversely proportional
ditions, the stability of the planar interface will break to square root of grain size (D1/2), to account for the
down, and a cellular structure will form. The interface cell size effects on the yield strength.[13,41,58] However,
stability theory proposed by Mullins and Sekerka[50] for materials with well-defined dislocation cell structures
additionally takes capillarity effects into consideration as observed in AM materials, it has long been known
and can explain the evolution of cellular substructure that the macroscopic yield strength depends on the
observed in AM parts. The perturbation tip radius (R) inverse of the dislocation cell size (d1).[59–61] Further-
decreases with increasing solidification velocity. At very more, the cell size (d) can be related to dislocation
high growth velocity, as observed in AM process, the density (q) by the following equation[60]:
capillary force becomes important because of the 1
smaller value of tip radius. The magnitude of the solute d ¼ Kq2 ½4
diffusion length in the interdendritic region and capil-
larity effects govern the formation of cellular structures where K is an arbitrary constant. Therefore, the higher
during rapid solidification.[51] yield strength of AM-built materials should be corre-
Rapid solidification also results in high dislocation lated with the dislocation density. According to Tay-
density substructures.[52] However, the exact mechanism lor’s equation, the yield strength can be expressed as:
of dislocations generation during solidification is yet to pffiffiffi
ry ¼ r0 þ MaGb q; ½5
be understood, although several mechanisms of dislo-
cations formation have been proposed including the where r0 is the friction stress, M is the Taylor factor, a is
following[53]: the proportionality constant, G is the shear modulus (77
(a) Steep solute concentration gradients at the cell GPa for 316 stainless steel), b is the Burger’s vector of
boundary will result in coherency strains. In order dislocations, and q is the dislocation density. The value
to dissipate the associated strain energy, disloca- of constanta depends on the type, density, and interac-
tions could be formed. tion of dislocations at different strain levels.[62,63] Here, a
(b) Very high thermal stresses could nucleate and value of 0.26 ± 0.015 for a, obtained from References
drive dislocation motion, again, in order to 64–66, was used. These authors empirically assessed the
dissipate strain energy. value of a for high manganese and 304L stainless steels,
(c) Quenched-in clusters of vacancies could form which show similar TWIP behavior as the material
prismatic dislocation loops. under this study. Several authors have investigated the
(d) The misorientation of the neighboring cells could grain size effects on the yield strength, i.e., the
result in interfacial dislocations upon collision of Hall–Petch relationship, of 316L stainless and reported
the cells. friction stress to be 100 ± 20 MPa.[63,67] This high value
of friction stress corresponds to the high concentration
Saeidi et al.[37] qualitatively revealed a high density of of solute in the iron matrix.
dislocations at the cell boundaries in SLM 316L stainless On the other hand, the Taylor factor value of
steel. Similar qualitative observations also reported for polycrystalline material depends on texture. Here, the
other AM-built materials.[8,14,42,54,55] The present quantita- viscoplastic self-consistent (VPSC) model[68] was used to
tive investigation using XRD-based line profile analysis calculate the an effective value of M for both the HB and
revealed a dislocation density on the order of ~ 1015m2. VB samples in the as-built and annealed conditions. The
This dislocation density is typical of 20 pct cold work material was idealized as 1000 discrete ‘‘grain’’ features
austenitic stainless steel.[56] However, SLM-built materials with orientations and volume fractions selected to match
are in a lower energy state compared with cold work the texture. Details of the Taylor factor calculation from
material due to the low energy configuration of the the initial texture can be found in Reference 69. The
dislocations in the former, as indicated by a low value of initial value of M for the HB and VB samples was
dislocations arrangement parameter (M < 1) in line broad- calculated to be 2.6 for both the as-built and annealed
ening analysis. LEDS based on TEM analysis have also conditions, which is consistent with the observed
been reported.[8,14] Owing to this relatively low stored isotropy in yield strength.
energy in the material, the present AM-built materials are Table IV shows the comparison between calculated
thermally stable up to a very high temperature. Annealing of and experimentally measured yield strength. The calcu-
the AM samples at a very high temperature of 1100 C for lated yield strength of the as-built and annealed samples
one hour was insufficient to drive complete recrystallization shows a good agreement with the experimentally mea-
or grain coarsening (see Figure 6). For comparison, note sured yield strength. It should be noted here that the
that Herrera et al.[57] reported recrystallization in cold-rolled yield strength uncertainties in each sample were calcu-
316L stainless steel T ~ 600 C after one hour. lated by considering the error propagation from

3024—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

uncertainties in the Taylor’s equation resulting from direction (Figures 7 and 13). At low strain level of
uncertainties in r0, a and dislocation density (q). e ¼ 0:04, a h100i||LD fiber is beginning to be developed.
No significant intensification of this fiber was observed
at the strain level of e ¼ 0:18, i.e., no h111i||LD was
C. Deformation Behavior and Higher Ductility
observed even at this intermediate level of strain. This
Austenitic stainless steels deform either by dislocation indicates a higher activity of twinning for VB material.
slip, mechanical twining and/or martensitic transforma- The reinforcement of h111i||LD fiber only observed for
tion. Often, these deformation mechanisms coexist.[70] samples strained at e ¼ 0:57. Depending on initial
The stacking fault energy (SFE) of the material, which texture, deformation might activate different deforma-
depends on the chemical composition, is a controlling tion mechanism at the same strain level with respect to
factor in determining which mechanism(s) will be the loading direction. Future work will seek to quantify
operative. Frommeyer et al.[71] suggested that at the the relative contributions of the slip and twinning
SFE lower than the 16 mJ/m2 martensitic transforma- modes.
tion (c to eÞ dominates, while twinning occurs for SFE
higher than 25 mJ/m2. Based on the empirical relation-
ship proposed by Pickering et al.,[72] and the alloy
chemistry listed in Table I, the SFE value of 316L V. CONCLUSIONS
stainless is estimated to be 32 mJ/m2, suggesting that
The microstructure and mechanical properties of an
twinning will be favored over martensitic transforma-
additively manufactured (AM) 316L stainless steel were
tion. The analysis of the deformed samples at different
investigated. The following conclusions can be drawn
strain level using EBSD at e ¼ 0:18 (see Figure 12) and
from this study:
XRD at e ¼ 0:05; 0:18; ef (not shown in the interest of
space, although the data appears so in Figure 2) 1. AM of 316L stainless steel results in a single-phase,
confirms the absence of martensitic phase. austenitic microstructure that remains stable during
Twinning activity is one of the main factors contribut- tensile deformation, i.e., no stress-induced marten-
ing to the high ductility of AM built materials, partic- sitic phase transformation was observed.
ularly for the VB samples. Twin boundaries, similar to 2. An elongated, cellular microstructure resulting from
the grain boundaries, act as obstacles to dislocation the AM process is characterized by walls of high
movement. As twins are introduced, they decrease the dislocation density 1:2  1015 m2 measured by
mean free path of the dislocations on the slip planes XRD line profile, which is consistent with the
intersecting the twinning planes. This dynamic refine- EBSD-based estimates of the geometrically neces-
ment of the microstructure enhances the work hardening sary dislocation density.
behavior of the material during plastic deformation, 3. The observed high dislocation density is shown to
which in turn delays the onset of plastic instability. be responsible for the relatively high-yield strength
(~588 MPa) of the as-built AM material.
4. Preferred growth during the solidification process and
D. Texture Evolution and Anisotropic Behavior bidirectional laser scan pattern result in a relatively
Considering that the samples exhibit negligible poros- weak texture with h100i||transverse direction and
ity, or lack of fusion defects in the material being studied h010i inclined ~ 15 deg from the building direction.
(density > 99 pct), it is suggested that variation in 5. Although the yield strength is relatively isotropic,
texture is responsible for the observed anisotropic the initial texture contributes to anisotropic strain
mechanical behavior. FCC polycrystals with high stack- hardening and ductility (49 and 77 pct elongations
ing fault energy develop a pronounced h111i||LD fiber observed for horizontal (HB) and vertical built (VB)
texture during straining.[73] However, low SFE FCC samples, respectively).
materials deformed by mechanical twinning result in a 6. EBSD analysis confirms the formation of twins
h100i||LD fiber. The initial grain size and grain orien- during tensile deformation. Previous research has
tation have strong effects on twin formation and twin shown that twinning results in a dynamic Hall–
activity. In general, h111i||LD oriented grains favor Petch relationship, which contributes significantly
twinning due to their high Taylor factor. Formation of to strain hardening. The texture evolutions for the
twins within h111i||LD grains reinforce the h100i||LD tensile-deformed HB and VB samples indicate a
fiber. A limited number of twins form within the higher degree of twinning activity in the VB
h100i||LD orientation reinforce the h111i||LD fiber. samples, consistent with the observed higher resis-
Crystallographic texture analysis of the as-built samples tance to plastic instability (and hence enhanced
(see Figure 13) shows a preferred orientation of (100) ductility) in those samples. Annealing at 1100 C
grains along TD (see Figure 5). The texture evolution reveals that AM-316L stainless steel is thermally
observed during loading along the TD direction, i.e., stable, shows only moderate decrease in dislocation
horizontal built samples is indicative of primarily density and strength, and displays very similar
slip-based strain accommodation. On the other hand, strain hardening response, level of ductility, and
no preferred orientation was observed along the built anisotropy compared with the as-built condition.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3025

 ACKNOWLEDGMENTS 27. D. Black, D. Windover, A. Henins, D. Gil, J. Filliben, and J.P.
Cline: Adv. X-ray Anal., 2009, vol. 53, pp. 172–79.
The authors thank the Naval Sea Systems Com- 28. T. Ungár and A. Borbély: Appl. Phys. Lett., 1996, vol. 69,
mand and the Office of Naval Research for sponsoring pp. 3173–75.
29. G. Ribárik, J. Gubicza, and T. Ungár: Mater. Sci. Eng. A, 2004,
the research discussed in this paper. The work was vol. 387, pp. 343–47.
mentored by Naval Surface Warfare Center, Dahlgren 30. G. Ribárik, T. Ungár, and J. Gubicza: J. Appl. Crystallogr., 2001,
Division, as a project through the Naval Engineering vol. 34, pp. 669–76.
Education Consortium Program. 31. M. El-Tahawy, Y. Huang, T. Um, H. Choe, J. Lábár, T.G.
Langdon and J. Gubicza: J. Mater. Res. Technol., 2017 (in press).
32. M. Mangalick and N. Fiore: Trans Metall. Soc. AIME, 1968,
vol. 242, p. 2363.
33. A. Borbély, J. Dragomir-Cernatescu, G. Ribárik, and T. Ungár: J.
REFERENCES Appl. Crystallogr., 2003, vol. 36, pp. 160–62.
1. N. Shamsaei, A. Yadollahi, L. Bian, and S. Thompson: Addit. 34. T. LeBrun, T. Nakamoto, K. Horikawa, and H. Kobayashi:
Manuf., 2015, vol. 8, pp. 12–35. Mater. Des., 2015, vol. 81, pp. 44–53.
2. J.P. Kruth, P. Mercelis, J. Van Vaerenbergh, L. Froyen, and M. 35. L.E. Murr, E. Martinez, J. Hernandez, S. Collins, K.N. Amato,
Rombouts: Rap. Prototyp. J., 2005, vol. 11, pp. 26–36. S.M. Gaytan, and P.W. Shindo: J. Mater. Res. Technol., 2012,
3. Q. Jia and D. Gu: J. Alloys Compd., 2014, vol. 585, pp. 713–21. vol. 1, pp. 167–77.
4. J.A. Cherry, H.M. Davies, S. Mehmood, N.P. Lavery, S.G.R. 36. L. Facchini, N. Vicente, Jr, I. Lonardelli, E. Magalini, P. Robotti,
Brown, and J. Sienz: Int. J. Adv. Manuf. Technol., 2015, vol. 76, and A. Molinari: Adv. Eng. Mater., 2010, vol. 12, pp. 184–88.
pp. 869–79. 37. K. Saeidi, X. Gao, F. Lofaj, L. Kvetková, and Z.J. Shen: Alloys
5. L. Thijs, K. Kempen, J. Kruth, and J. Van Humbeeck: Acta Compd., 2015, vol. 633, pp. 463–69.
Mater., 2013, vol. 61, pp. 1809–19. 38. M. L. Sistiaga, S. Nardone, C. Hautfenne, and J. Van Humbeeck:
6. Harry Bhadeshia and Robert Honeycombe: Steels: Microstructure Annual International Solid Freeform Fabrication Symposium, 2016,
and Properties, 4th ed., Butterworth-Heinemann, and imprint of pp. 558–565, Austin.
Elsevier, Cambridge, MA, 2017. 39. F. Geiger, K. Kunze, and T. Etter: Mater. Sci. Eng. A, 2016,
7. F. Bartolomeu, M. Buciumeanu, E. Pinto, N. Alves, O. Carvalho, vol. 661, pp. 240–46.
F.S. Silva, and G. Miranda: Addit. Manuf., 2017, vol. 16, 40. J.R. Taylor: An Introduction to Error Analysis: The Study of
pp. 81–89. Uncertainties in Physical Measurements, 2nd ed., University Sci-
8. K. Saeidi: Stainless steel Fabricated by Laser Melting: Scaled- ence Books, Sausalito, 1997.
Down Structural Hierarchies and Microstructural Heterogeneities, 41. D. Zhang, W. Niu, X. Cao, and Z. Liu: Mater. Sci. Eng. A, 2015,
Stockholm University, Stockholm, 2016. vol. 644, pp. 32–40.
9. R. Li, Y. Shi, Z. Wang, L. Wang, J. Liu, and W. Jiang: Appl. Surf. 42. W.M. Tucho, P. Cuvillier, A. Sjolyst-Kverneland, and V. Hansen:
Sci., 2010, vol. 256, pp. 4350–56. Mater. Sci. Eng. A, 2017, vol. 689, pp. 220–32.
10. S. Dadbakhsh, L. Hao, and N. Sewell: Rapid Prototyp. J., 2012, 43. M. Calcagnotto, D. Ponge, E. Demir, and D. Raabe: Mater. Sci.
vol. 18 (3), pp. 241–49. Eng. A, 2010, vol. 527, pp. 2738–46.
11. B. Zhang, L. Dembinski, and C. Coddet: Mater. Sci. Eng. A, 2013, 44. J.C. Li: J. Appl. Phys., 1962, vol. 33, pp. 2958–65.
vol. 584, pp. 21–31. 45. I. Gutierrez-Urrutia and D. Raabe: Acta Mater., 2011, vol. 59,
12. D. Wang, C. Song, Y. Yang, and Y. Bai: Mater. Des., 2016, pp. 6449–62.
vol. 100, pp. 291–99. 46. B.C. De Cooman, Y. Estrin, and S.K. Kim: Acta Mater., 2017 (in
13. Y. Zhong, L. Liu, S. Wikman, D. Cui, and Z. Shen: J. Nucl. press).
Mater., 2016, vol. 470, pp. 170–78. 47. O. Bouaziz, S. Allain, and C. Scott: Scr. Mater., 2008, vol. 58,
14. Z. Sun, X. Tan, S. Tor, and W. Yeong: Mater. Des., 2016, vol. 104, pp. 484–87.
pp. 197–204. 48. I. Gutierrez-Urrutia, S. Zaefferer, and D. Raabe: Mater. Sci. Eng.
15. J. Suryawanshi, K. Prashanth, and U. Ramamurty: Mater. Sci. A, 2010, vol. 527, pp. 3552–60.
Eng. A, 2017, vol. 696, pp. 113–21. 49. H. Beladi, I.B. Timokhina, Y. Estrin, J. Kim, B.C. De Cooman,
16. C. Haase, J. Bültmann, J. Hof, S. Ziegler, S. Bremen, C. Hinke, A. and S.K. Kim: Acta Mater., 2011, vol. 59, pp. 7787–99.
Schwedt, U. Prahl, and W. Bleck: Materials, 2017, vol. 10, p. 56. 50. W. Mullins and R. Sekerka: J. Appl. Phys., 1964, vol. 35,
17. A. Röttger, K. Geenen, M. Windmann, F. Binner, and W. pp. 444–51.
Theisen: Mater. Sci. Eng. A, 2016, vol. 678, pp. 365–76. 51. R. Trivedi and W. Kurz: Int. Mater. Rev., 1994, vol. 39, pp. 49–74.
18. W. Shifeng, L. Shuai, W. Qingsong, C. Yan, Z. Sheng, and S. 52. A. Glezer, and I. Permyakova: Melt-Quenched Nanocrystals. Boca
Yusheng: J. Mater. Process. Technol., 2014, vol. 214, pp. 2660–67. Raton : CRC Press (Taylor & Francis Group imprint), 2013.
19. H.D. Carlton, A. Haboub, G.F. Gallegos, D.Y. Parkinson, and 53. V. Zolotorevsky, N. Belov, and M. Glazoff: Casting Aluminum
A.A. MacDowell: Mater. Sci. Eng. A, 2016, vol. 651, pp. 406–14. Alloys, Elsevier Science, Amsterdam, 2007.
20. E. Liverani, S. Toschi, L. Ceschini, and A. Fortunato: Mater. 54. D. Tomus, Y. Tian, P.A. Rometsch, M. Heilmaier, and X. Wu:
Process. Tech., 2017, vol. 249, pp. 255–63. Mater. Sci. Eng. A, 2016, vol. 667, pp. 42–53.
21. D. Tomus, Y. Tian, P. Rometsch, M. Heilmaier, M. Heilmaier, 55. J. Walker, K. Berggreen, A. Jones, and C. Sutcliffe: Adv. Eng.
and X. Wu: Mater. Sci. Eng. A, 2016, vol. 667, pp. 42–53. Mater., 2009, vol. 11, pp. 541–46.
22. I. Tolosa, F. Garciandı́a, F. Zubiri, and F. Zapirain: Int. J. Adv. 56. S. Murugesan, P. Kuppusami, E. Mohandas, and M. Vijayalakshmi:
Manu. Tech., 2010, vol. 51, pp. 639–47. Mater. Lett., 2012, vol. 67, pp. 173–76.
23. X. Zhou, K. Li, D. Zhang, X. Liu, J. Ma, W. Liu, and Z. Shen: J. 57. C. Herrera, R.L. Plaut, and A.F. Padilha: Mater. Sci. Forum,
Alloys Compd., 2015, vol. 631, pp. 153–64. 2007, vol. 550, pp. 423–28.
24. EOS GmbH. EOS Stainless Steel 316L material data sheet. [On- 58. K. Saeidi, X. Gao, Y. Zhong, and Z.J. Shen: Mater. Sci. Eng. A,
line] https://cdn1.scrvt.com/eos/77d285f20ed6ae89/dd6850c010d3/ 2015, vol. 625, pp. 221–29.
EOSStainlessSteel316L.pdf. Accessed 21 August 2017. 59. G. Langford and M. Cohen: Metall. Mater. Trans, 1970, vol. 1,
25. ASTM E8/E8M-16a: Standard Test Methods for Tension Testing of pp. 1478–80.
Metallic Materials, American Society for Testing and Materials, 60. M. Staker and D. Holt: Acta Metall., 1972, vol. 20, pp. 569–79.
West Conshohocken, 2016. 61. D. Kuhlmann-Wilsdorf: Metall. Trans., 1970, vol. 1, pp. 3173–79.
26. ASTM E1019-11: Standard Test Methods for Determination of 62. F. Lavrentev: Mater. Sci. Eng., 1980, vol. 46, pp. 191–208.
Carbon, Sulfur, Nitrogen, and Oxygen in Steel, Iron, Nickel, 63. X. Feaugas and H. Haddou: Metall. Mater. Trans. A, 2003,
and Cobalt Alloys by Various Combustion and Fusion Tech- vol. 34, pp. 2329–40.
niques. West Conshohocken: American Society for Testing and 64. G. Dini, R. Ueji, A. Najafizadeh, and S.M. Monir-Vaghefi: Mater.
Materi. Sci. Eng. A, 2010, vol. 527, pp. 2759–63.

3026—VOLUME 49A, JULY 2018 METALLURGICAL AND MATERIALS TRANSACTIONS A

65. B. Hutchinson and N. Ridley: Scr. Mater., 2006, vol. 55, 70. Y.F. Shen, X.X. Li, X. Sun, Y.D. Wang, and L. Zuo: Mater. Sci.
pp. 299–302. Eng. A, 2012, vol. 552, pp. 514–22.
66. M. Kassner: Acta Mater., 2004, vol. 52, pp. 1–9. 71. G. Frommeyer, U. Brüx, and P. Neumann: ISIJ Int., 2003, vol. 43,
67. B. Kashyap and K. Tangri: Acta Metall. Mater., 1995, vol. 43, pp. 438–46.
pp. 3971–81. 72. F. Pickering: Physical Metallurgy and the Design of Steels, Applied
68. J.W. Hutchinson: Proc. R. Soc. London A Math. Phys. Eng. Sci., Science Publishing Ltd, Essex, 1978.
1976, vol. 348, pp. 101–27. 73. C. Haase, C. Zehnder, T. Ingendahl, A. Bikar, F. Tang, B.
69. C.N. Tome, C.R. Canova, and U.F. Kocks: Acta Mater., 1984, Hallstedt, W. Hu, W. Bleck, and D.A. Molodov: Acta Mater.,
vol. 32, pp. 1637–53. 2017, vol. 122, pp. 332–43.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 49A, JULY 2018—3027