ArticlePDF Available

Non-oxide precipitates in additively manufactured austenitic stainless steel

  • CESI Engineering School

Abstract and Figures

Precipitates in an austenitic stainless steel fabricated via any Additive Manufacturing (AM), or 3D printing, technique have been widely reported to be only Mn-Si-rich oxides. However, via Transmission Electron Microscopy (TEM) studies on a 316L stainless steel, we show that non-oxide precipitates (intermetallics, sulfides, phosphides and carbides) can also form when the steel is fabricated via Laser Metal Deposition (LMD)—a directed energy deposition-type AM technique. An investigation into their origin is conducted with support from precipitation kinetics and finite element heat transfer simulations. It reveals that non-oxide precipitates form during solidification/cooling at temperatures ≥ 0.75T m (melting point) and temperature rates ≤ 10 ⁵ K/s, which is the upper end of the maximum rates encountered during LMD but lower than those encountered during Selective Laser Melting (SLM)—a powder-bed type AM technique. Consequently, non-oxide precipitates should form during LMD, as reported in this work, but not during SLM, in consistency with existing literature.
Content may be subject to copyright.
Scientic Reports | (2021) 11:10393 | 
Non‑oxide precipitates in additively
manufactured austenitic stainless
Manas Vijay Upadhyay1*, Meriem Ben Haj Slama1,2, Steve Gaudez1,3, Nikhil Mohanan1,
Lluis Yedra2,4,5, Simon Hallais1, Eva Héripré2 & Alexandre Tanguy1
Precipitates in an austenitic stainless steel fabricated via any Additive Manufacturing (AM), or 3D
printing, technique have been widely reported to be only Mn‑Si‑rich oxides. However, via Transmission
Electron Microscopy (TEM) studies on a 316L stainless steel, we show that non‑oxide precipitates
(intermetallics, suldes, phosphides and carbides) can also form when the steel is fabricated via
Laser Metal Deposition (LMD)—a directed energy deposition‑type AM technique. An investigation
into their origin is conducted with support from precipitation kinetics and nite element heat
transfer simulations. It reveals that non‑oxide precipitates form during solidication/cooling at
temperatures ≥ 0.75Tm (melting point) and temperature rates 105 K/s, which is the upper end of the
maximum rates encountered during LMD but lower than those encountered during Selective Laser
Melting (SLM)—a powder‑bed type AM technique. Consequently, non‑oxide precipitates should form
during LMD, as reported in this work, but not during SLM, in consistency with existing literature.
Additive Manufacturing (AM), or 3D printing, of alloys is a revolutionary technology that simultaneously manu-
factures a part with a desired geometry and creates the material microstructure without the need for tooling.
is unique ability has generated a widespread industrial interest to adopt AM not only to manufacture entire
alloy parts but also to repair broken/damaged parts. However, manufacturing entire parts or repairing portions
of a broken part via AM to obtain microstructures that exhibit desired material properties is a great challenge,
and it has been a subject of intensive research in recent years.
An AM process involves reading a 3D part geometry from a computer-aided design le and building it in a
layer-by-layer manner by locally melting a feedstock (powder/wire) using a moving heat-source (laser/electron
beam). Based on the manner in which heat-source and feedstock interactions occur, alloy AM processes can be
classied1 into (i) Directed Energy Deposition (DED), where the feedstock material is directly fed into the moving
heat-source, and (ii) Powder Bed Fusion (PBF), a two-step repetitive process involving powder deposition on a
powder-bed followed by scanning via a heat-source. AM of austenitic stainless steel parts is typically done via
Laser Metal Deposition (LMD: a laser-based DED approach) and Selective Laser Melting (SLM: a laser-based
PBF approach). In both approaches, the heat-matter interactions subject the material to a sequence of highly
non-equilibrium processes, viz. melt-pool dynamics, rapid solidication and solid-state cooling-heating cycles,
which result in the formation of hierarchical microstructures exhibiting physical and chemical heterogeneities
at multiple length scales1.
In the case of 316L Stainless Steel (316LSS) fabricated via SLM or LMD, the hierarchical microstructures are
composed18 of (i) precipitates, micro-segregations, porosities and dislocation structures at the intragranular level
and (ii) heterogeneous grain morphologies and sizes, texture and voids at the polycrystalline level. Amongst all
these features, precipitates are one of the smallest in size and they play a crucial role in determining the material
response3,4. During plastic deformation, precipitates can impede dislocation motion resulting in hardening via
mechanisms such as Orowan bypass, cross-slip, etc.9 Furthermore, the presence/absence of some precipitates
    
    Laboratoire de Mécanique des Sols, Structures et Matériaux (MSSMat), CNRS
           
     
         
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
can aect the corrosion, wear, fatigue and fracture resistance of a steel2,10. It is therefore important to understand
their origin in order to control their formation in the microstructure and optimize material properties.
Conventionally processed/post-processed 316LSS is known to contain a plethora of dierent kind of pre-
cipitates including oxides of transition metals (Fe, Cr, Ni, Mn, Mo) and silicon1113 as well as non-oxides such
as transition-metal carbides10, suldes1417, phosphides18 and intermetallics19,20. In contrast, however, a series of
recent Transmission Electron Microscopy (TEM) studies27 report the sole presence of Mn-Si–O precipitates in
SLM and LMD 316LSS. is signicant dierence in the type of precipitates between conventionally processed/
post-processed and additively manufactured (LMD or SLM) 316LSS can be attributed to the dierences in the
temperature rates encountered during these processes. In addition, oxidation of Mn and Si precipitates results
in the highest reduction in the Gibbs free energy in comparison to other transition-metal (Fe, Cr, Ni and Mo)
oxides21, which favors the formation of Mn-Si-O over other oxide precipitates.
It is well known that precipitate sizes in steels decrease with increasing solidication/cooling temperature
rates15,16. For conventionally processed/post-processed 316LSS, where temperature rates much less than 102K/s
are encountered, oxide and non-oxide precipitate sizes typically range from 1μm to several tens of microns15,16.
Meanwhile, during LMD and SLM processes, the maximum temperature rates encountered fall in the range
102–105K/s22 and 106–107K/s23, respectively. ese magnitudes are consistent with the average size of Mn-Si-O
precipitates reported in LMD 316LSS (> 100nm) and SLM 316LSS (< 100nm)3,7, respectively. However, none
of these works report the presence of non-oxide precipitates. Based on these studies, one could deduce that the
temperature rates at which non-oxide precipitates stop forming in 316LSS should be below 102K/s, which is the
lower end of the maximum rate encountered during any AM process.
Recent TEM studies on 316LSS powders fabricated via inert gas atomization1,7,24 have revealed the presence
of a signicant amount of non-oxide precipitates (rich in Mo, Cr, P and S) together with Mn-Si–O precipitates.
Interestingly, the maximum cooling rates encountered during inert gas atomization2527 are in the same range
as those occurring during LMD i.e., 102–105K/s, which suggests that there should be a signicant presence of
non-oxide precipitates in LMD 316LSS. However, none of the existing studies on precipitates in LMD 316LSS
conclusively report the presence of non-oxides. In a very recent study, Barkia etal.28 reported the presence of
an Mo-Cr-rich zone and an Mn–Mo–Cr-rich zone (around an Mn–Si–O precipitate) within the walls of an
intragranular cellular solidication structure in their LMD 316LSS. However, these cell walls are known to be
sites of preferential segregation of Mo and Cr1, and the reported presence of Mo–Cr and Mn–Mo–Cr zones can
be considered only as circumstantial evidence of the presence of non-oxide precipitates.
In light of the above, the main aim of this paper is to answer the following questions: Can non-oxide pre-
cipitates form during LMD of 316LSS? What is the threshold temperature and temperature rate above which
non-oxide precipitates stop forming in 316LSS? To answer the rst question, we have performed a series of
Scanning-TEM (STEM) studies together with Electron-Dispersive X-ray Spectroscopy (EDS). To answer the
second question, we have performed Finite Element (FE) heat transfer simulations to better understand the
temperature rates encountered during LMD and used this information together with the results of precipitation
kinetics simulations to understand the role of temperature and temperature rates on nucleation and growth of
non-oxide precipitates. is analysis is also applied to understand the absence of non-oxides in SLM 316LSS.
e 316LSS powder used in this work had been produced via the inert gas atomization process by Höganäs
AB. e wrought alloy used to manufacture this powder had the chemical composition in weight percent (wt.
%): Cr—16.9, Ni—12.7, Mo—2.5, Mn—1.5, Si—0.7, P—0.015, C—0.011, S—0.005 and Fe-balance. Presence of
trace amounts of oxygen in the inert gas atomization chamber can introduce oxygen into the 316LSS powder
obtained at the end of this process; according to the documentation of Höganäs AB29, a 316LSS powder particle
produced via their inert gas atomization process can contain up to 0.05 wt. % of O.
e 316LSS powder had been fed into the LMD machine to print a single-track (one linear laser pass per
layer) 3-layer wall on a hot-rolled 316LSS substrate via a bidirectional (forth–back–forth) scanning strategy
as illustrated in Fig.1a. Following the building of the 3-layer wall and aer the substrate had cooled down, a
60-layer wall was built on the same substrate as shown in Fig.1a. e Electron Back-Scattered Diraction (EBSD)
orientation map of a cross-section in the middle part of the 3-layer wall (Fig.1b) and the 60-layer wall (Fig.S1)
reveal a weakly-textured microstructure with columnar grains in the bulk surrounded by a thin layer of smaller
equiaxed grains, which is typical for an LMD process.
ree TEM lamellae (L1, L2 and L3), each thinner than 100nm, had been extracted from three dierent
columnar grains of cross-section A-A of the 3-layer wall (Fig.1b). Two lamellae (L4 and L5) were extracted from
two columnar grains, one near the top and one near the bottom of the 60-layer wall (Fig.S1). All lamellae were
studied via (i) High-Angle Annular Dark-Field (HAADF) STEM imaging to identify the location of precipitates
and (ii) EDS chemical mapping to understand their chemical composition (Fig.2 and Fig.S2). HAADF images
and EDS chemical maps help identify intragranular chemical heterogeneities; in LMD 316LSS, these heteroge-
neities are cellular solidication structures and precipitates.
Intragranular cellular solidication structures in any AM 316LSS typically have cell walls with lower concen-
tration of Fe and higher concentration of Cr and Mo than the surrounding matrix1,4; these cellular structures can
be discerned from the EDS maps in Fig.2 and Fig.S2. ese structures also have their long-axes aligned with the
crystal growth direction during solidication1. Since L1 and L2 had been extracted in directions nearly parallel
and perpendicular to the growth direction of their respective grains (Fig.1), the long-axis of cellular structures in
L1 and the grid-like cross-section of the cellular structures in L2 can be discerned. L3 forms a random angle with
the grain growth direction, which makes it dicult to discern the cellular structures. Nevertheless, a Y-shaped
band can be observed in its Fe, Cr and Mo maps. L4 also shows a complex intragranular cellular structure whose
precise orientation is hard to discern from the Fe, Cr and Mo maps (Fig.S2). L5 has the long-axes of its cellular
solidication structures aligned in-plane at a small angle with the horizontal direction. Similar to Barkia etal.28,
Mo-Cr and Mo-Cr-Mn rich zones are observed in the cellular solidication structures in L1 and L4, however,
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
at the magnication level shown in Fig.2 and Fig.S2, it is not possible to discern whether or not these zones
are forming precipitates.
e chemical composition of the lamellae L1–L5 obtained from the EDS maps shows that all lamellae have
a lower wt. % of Fe and Cr and higher wt. % of Mn, O, P, S and C than the wrought alloy (TableS1). Ni is higher
for L1, L2, L4 and L5 but lower for L3. Mo is higher for L3 and L5 but lower for the rest. Si is higher for L1, L2
and L3 but lower for the rest.
55 precipitates have been identied in these lamellae: 16 in L1, 5 in L2, 2 in L3, 17 in L4 and 15 in L5. Most
of them could be discerned from the HAADF images in Fig.2 and Fig.S2, however, some of them were discov-
ered aer performing high-resolution HAADF imaging (Fig.3 and Fig.S3). In order to facilitate identifying the
precipitates and dierentiating between them, a classication scheme and a nomenclature system is proposed
in Table1. Based on this classication scheme, out of the 55 precipitates, 16 are oxides (29.1%), 3 are non-oxides
(5.4%) and 36 (65.5%) are mixed.
Figure3 shows the HAADF images and EDS chemical maps for all non-oxide and mixed precipitates together
with at most one oxide precipitate (if any) in L1–L3; Fig.S3 shows the HAADF images and EDS maps for non-
oxide and mixed precipitates in L4 and L5. EDS line prole analysis (Methods) has been performed for each
precipitate; Fig.4 shows the EDS line proles for two of the most complex mixed precipitates. Following this
analysis, the chemical composition of all precipitates had been compared with the nominal composition of the
316LSS wrought alloy. Based on this comparison the elements in each oxide and non-oxide precipitate, and
all the inclusions in each mixed precipitate, have been classied into two categories: (i) higher and (ii) lower
composition than wrought alloy. is classication is used to tabulate each element in all precipitates of L1, L2
and L3 in TableS1.
All precipitates consistently show a decrease in Fe with respect to the wrought alloy and the lamella to which
they belong. Most precipitates show a decrease in Ni, however, these concentrations never become zero. TEM
lamella preparation can lead to partial slicing of precipitates and in most cases include some proportion of the
surrounding matrix along the thickness direction; this is also evidenced via the distortions caused by dislocation
lines in the HAADF images of oxides (Fig.3). Hence, non-zero compositions of Fe in all precipitates should
mainly arise from the matrix above or below the precipitate.
All oxides are amorphous in nature, which has been conrmed from high-resolution STEM imaging and
convergent beam diraction (not shown) and it is consistent with existing literature7. ey are richer in Mn–Si–O
with respect to the wrought alloy, however, in some cases, Si and O are lower in wt. % than their respective com-
positions in the associated lamellae (TableS1). Furthermore, proportions of Mn, Si and O vary (drastically in
some cases) from one oxide to another. ese results indicate the presence of dierent combinations of MnxSiyOz
(x,y 0, z > 0) in these precipitates. Along with Mn-Si-O, oxides have higher concentrations of P, S and C than
the wrought alloy but oen lower C concentration than their corresponding lamella (TableS1); furthermore, in
most cases, the P, S and C concentrations increase from the oxide center to the oxide-matrix interface.
Figure1. Design and crystallographic orientation map of the LMD 316LSS walls from which TEM lamellae
are extracted. (a) An illustration (not-to-scale) of the single-track bidirectionally-printed 3-layer and 60-layer
LMD 316LSS walls on a hot-rolled 316LSS substrate. Directions
respectively represent the printing,
thickness and building directions. e top view and the front view of a cross-section A–A are shown. e red
arrows in the front-view show the direction of printing of each layer. (b) An EBSD orientation map color-
coded according to IPF along building direction at a cross-section B–B approximately at the mid-section of the
3-layer wall. Dotted lines in (b) represent the approximate position of interlayer boundaries. e black solid
circles in (b) indicate the points X, Y and Z that correspond to the locations in the FE simulations from where
the temperature vs time data has been extracted and plotted in Fig.7. Subgures in (b) show zoomed-in EBSD
maps. Yellow-colored lines demark zones underneath which lamellae L1, L2 and L3 are extracted (in the out-of-
plane direction). ese lines correspond to the top side of L1, L2 and L3 in the HAADF images in Fig.2. Aztec
v4.2 (Oxford Instruments https:// nano. oxinst. com/ produ cts/ aztec/) and Microso PowerPoint have been used
to prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Amongst the 3 non-oxide precipitates, (i) P161 is richer in Cr–Mo–Mn–Si–O–P–S–C in comparison to the
wrought alloy and L1, (ii) P154 is richer in Ni–Mo–Mn–O–P–S–C in comparison to the wrought alloy but only
Mo-Mn-P-S are higher than those in L4, and (iii) P75 is richer in Ni–Mo–Mn–O–P–S in comparison to the
wrought alloy but only Mo-Mn-P are higher than those in L5. All non-oxide precipitates are embedded either
above, below or in between the matrix. Whenever a non-oxide precipitate contains a higher wt. % of Cr than its
surroundings, it results in a brighter contrast with respect to its surrounding in the HAADF image, as evidenced
from P161.
Similar to non-oxide precipitates, all mixed precipitates also exhibit a non-zero oxygen composition with
respect to the wrought alloy, and all non-oxide inclusions in mixed precipitates that have a higher wt. % of Cr
than the surrounding matrix show a brighter contrast in the HAADF images. Out of the 36 mixed precipitates,
34 (94.44%) contain both oxide and non-oxide inclusions; P151 and P144 contain only non-oxide inclusions.
Amongst these 34 mixed precipitates, 26 (72.22% out of 36) clearly show oxide inclusions at their core and non-
oxide inclusions around the oxides. However, it is harder to arrive at this conclusion for the remaining 8 (22.22%
out of 36): P11, P91, P34, P84, P124, P15, P25 and P45.
Figure2. STEM HAADF images and EDS chemical maps of L1, L2 and L3 extracted from the 3-layer wall of
Fig.1. EDS maps have been generated from within the black outlined region in each HAADF image. e dotted
green lines in each HAADF image represents the zone where the lamella composition has been computed
and presented in TableS1 (Supplementary Data). Precipitates that are easily visible in the HAADF images are
numbered in yellow font. Precipitates that have been detected aer high resolution imaging are numbered in red
font. Scale for EDS maps of each lamella are shown in their C map. HAADF images and EDS maps have been
acquired using the TIA v4.2 (FEI https:// www. fei. com) and the Esprit v1.9 (Brucker https:// www. bruker. com)
soware, respectively. Fiji v2.1.0/1.53c (https:// ji. sc/) and Microso Powerpoint have been used to prepare this
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Out of the 36 mixed precipitates, 30 are composed of two inclusions each that manifest either next to each
other or one surrounds the other; they are P121–P151, P12, P22, P42, P52, P24, P44–P64, P84–P104, P134, P144,
P164, P174, P15–P65, P95–P135. e remaining 6 i.e., P11, P91, P23, P34, P74 and P124 are composed of more than
2 inclusions.
Figure3. STEM HAADF images and EDS maps of all non-oxide and mixed precipitates, and at maximum
one oxide precipitate, in L1–L3. To facilitate visualization, picture corrections have been made for all the EDS
maps. ImageJ and Microso PowerPoint have been used to prepare this image. HAADF images and EDS maps
have been acquired using the TIA v4.2 (FEI https:// www. fei. com) and the Esprit v1.9 (Brucker https:// www.
bruker. com) soware, respectively. Fiji v2.1.0/1.53c (https:// ji. sc/) and Microso PowerPoint have been used to
prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
P11 is a complex combination of 3 zones and 5 small inclusions. Based on the morphology, brightness evolu-
tion and composition of these zones, we can deduce that (i) zone 1 was part of a larger oxide inclusion before the
lamella was prepared, (ii) zone 2 is made up of two inclusions: one is the same oxide as in zone 1 and the other
is a Cr–Mo–P–S non-oxide that overlaps the rst, (iii) the brightness changes across zone 2 as the amount of
the oxide inclusion along L1’s thickness direction decreases and that of the non-oxide inclusion increases. (iv)
e oxide terminates at the interface between zone 2 and zone 3 such that (v) zone 3 is rich only in Cr–Mo–P–S
(Figs.3 and 4). Due to the high O-content of the ve elliptical inclusions, they are classied as oxides. ey are
also rich in Cr–Mo–Mn–P–S–C with respect to the wrought alloy but these contributions could arise due the
overlap of these inclusions with zones 2 and 3 along L1’s thickness.
P91 also has three zones. Similar to P11, the morphology of these zones and of the surrounding matrix sug-
gests that prior to L1’s preparation, zone 1 was part of a larger oxide inclusion and zone 2 includes a portion from
the same oxide inclusion embedded underneath or above the surrounding matrix. e portion of this oxide in
L1 terminates at the outer boundary of zone 2. Zone 3 is a non-oxide inclusion richer in Cr–Mo–P–S–C with
respect to the wrought alloy and Cr–Mo–P–S with respect to L1.
e presence of dierent zones in P23 can be explained in a manner similar to P11 and P91. However, the mor-
phology of dierent inclusions, their composition and dierence between their contrast in the HAADF image,
clearly shows that the oxide inclusion in P23 is not circular and it is highly unlikely that it would have been spheri-
cal prior to the extraction of L3. Distorted oxide inclusions are also observed in other mixed precipitates such as
P121. Meanwhile, pure oxide precipitates always manifest as circles. erefore, these results strongly suggest that
the presence of non-oxides can deviate the morphology of oxide inclusions away from a perfect spherical shape.
Similar to P11, P34 and P124 also contain multiple small oxide inclusions embedded in non-oxide inclusions
(Fig.S3); their sizes and distributions make it hard to separate them into clear zones. e oxide inclusions in
P124 are rich only in Mn–O; the composition of Si is lower than that in the wrought alloy and the surround-
ing matrix. P74 contains a single large oxide inclusion at the core and it is surrounded by multiple non-oxide
inclusions richer in Mo–Cr–P and Mo–Cr–S than the surrounding matrix. Meanwhile, contrary to other mixed
precipitates containing oxide inclusions in the core, P84, P15, P25 and P45 contain a large non-oxide inclusion at
their core surrounded by a smaller oxide inclusion.
P151 and P144 are the only mixed precipitates that are composed of only non-oxide inclusions. P151 is made
up of two inclusions next to each other. e darker one is richer in Ni–Mo–Mn–Si–O–P–S–C than the wrought
alloy but only Mo–Mn–S–C are higher than in L1. e brighter one is also richer in Ni–Mo–Mn–Si–O–P–S–C
than the wrought alloy but only Mo–Mn–P–C are higher than in L1. P144 is made up of two nearly concentric
non-oxide inclusions (Fig.S3). e larger one is richer in Ni-Mo-Mn–O–P–S than the wrought alloy but only
Mo–P are higher than in L4. Interestingly, in comparison to the surrounding matrix, this inclusion is richer
in Cr–Mo–P; the presence of Cr results in a brighter contrast with respect to the surrounding matrix in the
HAADF image. e smaller darker one is richer in Mo–Mn–O–P–S than the wrought alloy but it is richer in
Cr–Mo–Mn–P than L4. Interestingly, P144 is the only precipitate that has negligibly small C composition with
respect to the wrought alloy.
e TEM analysis has clearly shown that a plethora of dierent kinds of non-oxide inclusions are present in
L1–L5. Furthermore, out of the 55 precipitates, 39 (70.9%) are either non-oxide precipitates or mixed precipitates
containing only non-oxide inclusions. Bearing this fact in mind together with the fact that the TEM lamellae
have been extracted from dierent locations of our LMD 316LSS walls and they occupy a very small volume in
these walls, we can claim with high condence that non-oxide inclusions are abundantly present in the 3-layer
and the 60-layer LMD 316LSS walls.
In order to check whether similar kind of precipitates occur or not in the 316LSS powder, a TEM lamella “LP”
had been prepared from a 316LSS powder particle (Fig.5). e composition of LP is similar to L1–L5 (TableS1).
It has lower Fe and Cr, slightly higher Ni, and the remaining alloying elements higher than the wrought alloy
along with a high proportion of O. HAADF image of LP reveals 5 mixed precipitates containing both oxide and
non-oxide inclusions with sizes similar to their counterparts in the 3-layer and 60-layer LMD 316LSS walls.
Table 1. Classication and nomenclature of precipitates in lamellae L1–L5. Precipitates are classied into
oxides, non-oxides and mixed based on the oxygen content (in EDS maps) and brightness (in HAADF image)
with respect to the surrounding (austenite) matrix. Each precipitate is numbered according the sequence
shown in the HAADF images in Fig.2 and the lamella number is added as a subscript.
Precipitate classication Description
L1 L2 L3 L4 L5
Single inclusion with nearly uniform
composition. Contains higher O than the
matrix and appears darker than matrix in
the HAADF image
P21–P81, P101, P111P32P13P14, P114P85, P145, P155
Single inclusion with nearly uniform
composition. Contains lesser oxygen than
the matrix and appears brighter than oxide
precipitates in HAADF images
P161 – P154P75
Mixed Two or more inclusions with dierent
elemental compositions P11, P91, P121–P151P12, P22, P42, P52P23P24–P104, P124–P144, P164, P174P15–P65, P95–P135
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Now, the only common factor between inert gas atomization and LMD is the maximum temperature rate
encountered during these processes, which is expected to be in the range 102–105K/s22,2527. Furthermore, it
is well-known that the precipitate sizes in steels decrease with increasing temperature rate1517. Based on this
knowledge and the TEM results of this work, we hypothesize that the temperature rate encountered during AM
of 316LSS is the main factor that determines the size of dierent kinds of precipitates. In order to verify this
hypothesis, we have performed thermodynamics computations to study the precipitation kinetics of a non-
oxide at dierent temperature rates. However, in order to perform these computations, we need (i) to assess the
oxide and non-oxide inclusion/precipitate size distribution in our LMD 316LSS walls, and (ii) to estimate the
temperature rates encountered during LMD of our 316LSS wall.
Figure6 shows the size distribution of oxide and non-oxide precipitates and inclusions in L1–L5. In general,
a wide scatter is observed in the size distribution of all precipitates and inclusions; the largest scatter occurs for
oxide inclusions in mixed precipitates. e size distributions could be either (i) an apparent scatter caused by
the 2D nature of a TEM analysis, or (ii) it could be due to dierences in inclusion nucleation and growth rates
arising from variations in local temperature rates. In order to diminish the inuence of these eects, only the
average sizes of the precipitates and inclusions are analyzed.
Figure4. HAADF images and EDS line proles for precipitates P11 and P91 along dierent paths. e points
in the curves have been tted using c-splines. e key for both HAADF images is presented below the HAADF
image of P11. HAADF images have been acquired using the TIA v4.2 (FEI https:// www. fei. com) soware. Fiji
v2.1.0/1.53c (https:// ji. sc), gnuplot v5.4 (http:// www. gnupl ot. info) and Microso PowerPoint have been used to
prepare this image.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
e average size of non-oxide precipitates/inclusions (72.69nm/85.67nm) is smaller than the average size of
oxide precipitates/inclusions (148.65nm/116.19nm). A similar trend in the size dierences between oxide and
non-oxide inclusions is also found in the 316LSS powder. Furthermore, a similar trend has also been reported
during solidication of a cast steel1517. From here one can deduce that for the same temperature rate and starting
temperature during processing of 316LSS, the average size of oxide inclusions/precipitates should be larger than
the average size of their non-oxide counterparts irrespective of the processing technique used.
In order to obtain an order of magnitude of the temperature rates that are encountered during the LMD
of our samples, solid-state heat transfer FE simulations (see Methods for simulation setup) were performed
for the 3-layer wall using the geometry shown in Fig.1a. e temperature v/s time history was extracted from
the mid-section of the 3-layer wall, approximately at the location of points X, Y and Z shown in Fig.1b. ese
temperature v/s time curves are plotted in Fig.7. Since points X and Y belong to the mid-layer of the 3-layer
wall, they experience a cooling-heating-cooling sequence. Since Z belongs to the top layer of the 3-layer wall, it
only experiences a cooling sequence. Just aer deposition, the three points experience a temperature above the
equilibrium solidus (1673K) of 316LSS. However, since the FE simulations do not account for uid dynamics
or solidication, we shall restrict the discussion to cooling rates at temperatures below the equilibrium solidus.
Between 1273K (typically the lowest annealing temperature for 316LSS) and equilibrium solidus, the three points
experience cooling rates between 1.15 × 104K/s and 2.91 × 104K/s; the latter rate is the highest rate encountered
by any point. e lowest cooling rate experienced by any point is 1K/s at 305K. Meanwhile, the highest heating
rate, 1.8 × 105K/s, is experienced by Y at 1074K. e lowest heating rate between 1273 K and 1673K experienced
by Y is 2.08 × 103K/s at 1517K.
e temperature v/s time curves shown in Fig.7 are relevant for lamellae L1, L2 and L3. Meanwhile, lamella
L4 should experience an initial cooling-heating-cooling similar to point X followed by 58 heating–cooling cycles
at decreasing temperature rates. L5 should experience a single cooling cycle at a temperature rate lower than
the maximum one shown in Fig.7. erefore, in the range of 300K to 1673K, any material point in the 3-layer
Figure5. SEM micrograph of (a) polished 316LSS powder particles and (b) a powder particle at higher
magnication. e TEM lamella LP in (c) is extracted from underneath the red zone in (b), which is embedded
within a grain whose outline is shown in blue. (c) STEM HAADF image of the lamella LP extracted from
the 316LSS powder particle in (b). (d) STEM HAADF images and EDS maps of all the precipitates of LP.
Precipitates are named by adding a prex “P” to their corresponding number in (c). SEM micrographs have
been captured using the XT v10 (ermoFisher https:// www. therm osh er. com/) soware. HAADF images and
EDS maps have been acquired using the TIA v4.2 (FEI http:// www. fei. com) and the Esprit v1.9 (Brucker https://
www. bruker. com) soware, respectively. Fiji v2.1.0/1.53c (https:// ji. sc/) and Microso PowerPoint were used to
prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
and the 60-layer walls should experience cooling and heating rates from 1K/s to 105K/s, which is typical for
an LMD process22.
Now, we perform a thermodynamic analysis to understand why non-oxide precipitates are present in our
LMD 316LSS walls. e rst step involves analyzing the most stable (i.e., lowest Gibbs free energy) equilibrium
phases that could exist in the liquid and solid states of 316LSS. is analysis is performed using the ermoCalc
soware and the results are presented in sectionS4 of the supplementary material for several dierent tempera-
tures in both the solid and liquid states. It reveals that the main phase in the solid 316LSS is γ-austenite and the
main phase in the liquid is molten 316LSS, henceforth called Liquid A (LA). Amongst all the possible oxides,
MnSiO3 is the most stable equilibrium phase in both the liquid and solid states; in the liquid state it exists in the
form of Liquid B (LB), which is an Mn–Si–O rich liquid with composition close to MnSiO3 having trace amounts
of Fe and S. Non-oxide (carbide, sulde and phosphide) phases do not exist above the equilibrium liquidus
temperature i.e., above 1703K (1430°C). Below the equilibrium solidus i.e., 1673K (1400°C), M–S (M = Mn,
Cr and Fe in descending wt.%) is the most stable equilibrium non-oxide phase.
Figure6. Comparison of equivalent diameters, also called size, of oxide precipitates (Oxides), non-oxide
precipitates (Non-oxides), and the oxide (Mixed oxides) and non-oxide (Mixed non-oxides) inclusions in
mixed precipitates in L1–L5. Equivalent diameter is the diameter of a circle whose area is equivalent to the
corresponding area of the precipitate/inclusion in its HAADF image. Values in black font represent maximum
and minimum equivalent diameters. Values in colored font represent average equivalent diameters, which
are the diameters of a circle whose area is equivalent to the average of the areas of all precipitate/inclusions
belonging to a category. e colored dashes represent the individual sizes of a precipitate/inclusion. Gnuplot
v5.4 (http:// www. gnupl ot. info) was used to prepare this gure.
Figure7. Temperature v/s time curves obtained from the heat transfer FE simulation of LMD of the 3-layer
wall of Fig.1b at the locations X, Y and Z. e temperature rates shown inside the plot have units K/s.
Superscript ‘c’ implies cooling and ‘h’ implies heating. e rates in red, blue and green font correspond to
the rates encountered by points X, Y and Z, respectively, during LMD. e subscripts ‘min’ and ‘max’ to the
temperature rates correspond to minimum and maximum, respectively, rates for a given curve. EQ solidus
stands for the equilibrium solidus temperature.
is the minimum temperature below which M–S (M = Mn,
Cr and Fe) does not precipitate for any temperature rate in the range 1–107K/s (see discussion related to Figs.9
and 10). While the simulation begins at 0s, however, X and Y are deposited at 5.7s and Z is deposited at 9.9s
aer the start. Gnuplot v5.4 (http:// www. gnupl ot. info) was used to prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Next, element solubility analysis is performed in order to understand the temperature dependent equilib-
rium composition and volume fraction of the most stable phases that coexist in the solid (γ-austenite, M-S and
MnSiO3) and liquid (LA and LB) 316LSS.
Figure8a shows the equilibrium atomic composition and stoichiometry of M–S as a function of temperature.
Starting as purely MnS below 473K, the composition of M-S evolves between 473K and equilibrium solidus as
Mn is increasingly substituted mainly by Cr and in much lesser proportion by Fe from the γ-austenite. Mean-
while, the equilibrium atomic composition and stoichiometry of MnSiO3 (not shown) does not change with
temperature until equilibrium solidus is reached.
Figure8b shows the equilibrium volume fraction as a function of temperature for γ-austenite, M-S, MnSiO3,
LA and LB. In the solid-state, γ-austenite is the dominant phase with some non-zero volume fraction of MnS
and MnSiO3. Between equilibrium solidus (1673K) and liquidus (1703K), the phase fraction of γ-austenite
decreases and that of LA correspondingly increases. At liquidus and beyond, all γ-austenite is replaced by LA.
e volume fraction of MnSiO3 also decreases above 1673K; however, up to 1703K, the decrease in MnSiO3 is a
combination of dissolution as well as the fact that the density of LA is lower than that of γ-austenite. At ~ 1743K,
MnSiO3 begins to transform into LB and the transformation is complete at ~ 1783K. Meanwhile, just above
equilibrium solidus, M-S dissolves into LA faster than γ-austenite transforms into LA; at 1683K, all M-S has
dissolved into LA.
e data used to plot the elemental solubility curves is then used to perform the precipitation kinetics simula-
tions for nucleation and growth of M–S inclusions on MnSiO3. Recalling that TEM observations show 72.22% of
the 36 mixed precipitates have oxide inclusions at their core and non-oxide inclusions surrounding the oxides,
the most important assumption made during the precipitation kinetics simulations is that M-S inclusions can
nucleate only on existing MnSiO3 inclusions (see Methods). e precipitation kinetics simulations are performed
during solidication/cooling from the liquid phase and heating from the room temperature.
Figure9a,b show the M–S diameter and volume fraction, respectively, as a function of temperature for dif-
ferent cooling rates, including 2 × 104K/s, which is approximately the average of the cooling rates encountered
between 1273 K and 1673K (Fig.7). e non-equilibrium solidus 1643K (1370°C) reported by Dépinoy etal.30
is considered as the reference solidus temperature for these simulations. e model parameters have been chosen
such that the predicted average diameter of M-S inclusions matches the measured average size of the non-oxide
inclusions in the mixed precipitates in Fig.6 for 2 × 104K/s at the non-equilibrium solidus. is tting results in
an excellent match between the predicted average diameters at cooling rates 1K/s and 8.33K/s and the experi-
mental measurements of Kim etal.16; thus, validating not only the precipitation kinetics predictions but also FE
predicted temperature rates.
For cooling rates 1K/s, 10K/s, 102K/s and 103K/s, precipitate growth saturates prior to reaching equilib-
rium solidus. For 104K/s, it saturates at the non-equilibrium solidus. For cooling rates 105–107K/s, precipitates
continue to grow below the non-equilibrium solidus, however, the increase in their diameter is overestimated
below this temperature because the simulations have been performed using the diusion coecient of M–S in
liquid (Methods). When the simulations are reperformed (not shown) using the diusion coecient of M–S in
γ-austenite, which is approximately 5 orders of magnitude lower than that in liquid, the changes in diameters
below the non-equilibrium solidus for all temperature rates are much smaller. Furthermore, at any cooling rate,
M–S does not nucleate/grow below 1273K i.e., below 0.76Tm (Tm = equilibrium melting point), which is also
the lowest annealing temperature for 316LSS. e predicted average diameters of M–S at the non-equilibrium
solidus are (in nm) 3395.1, 1131, 375.5, 144.5, 107.8, 29.1, 8.5 and 3.6, respectively, for cooling rates (in K/s) 1,
10, 102, 103, 104, 105, 106 and 107.
Figure8. Equilibrium atomic composition and volume fraction curves. (a) Equilibrium atomic composition of
only M–S (M = Mn, Cr and Fe), and (b) equilibrium volume fraction of γ-austenite, LA, LB, MnSiO3 and M–S,
as a function of temperature. ‘EQ solidus’ stands for equilibrium solidus. Gnuplot v5.4 (http:// www. gnupl ot.
info) was used to prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Figure9b shows that the M–S volume fraction for cooling rates from 1 to 103K/s is ~ 0.021%. At cooling rates
higher than 103K/s, the volume fraction begins to decrease faster than the increase in cooling rate. At 2 × 104K/s,
the volume fraction is ~ 0.0145% at non-equilibrium solidus, which is still comparable to the one obtained at
1K/s. At 105K/s, the M–S volume fraction is less than 0.002%, and it is negligibly small at 106 and 107K/s. A
signicant change in M–S volume fraction is not expected for temperatures lower than non-equilibrium solidus
at any cooling rate.
Figure10 shows the M–S volume fraction as a function of temperature for dierent heating rates, includ-
ing 2.08 × 103K/s, which is the lowest predicted temperature rate encountered by point Y between 1273 K and
1673K. For these simulations, the diusion coecient of M–S in γ-austenite has been used and, for simplicity,
equilibrium solidus is considered as the reference solidus. Note that during rapid heating, the non-equilibrium
solidus should be higher than the equilibrium solidus, however, this is not considered here because the precipi-
tation kinetics simulations use the equilibrium solubility curve of M–S (Fig.8) and this curve shows that M–S
has already completely dissolved at only 10K above the equilibrium solidus.
Similar to the solidication/cooling case, the volume fraction of M–S does not change below 1273K at any
heating rate. e volume fraction of M–S for heating rates 1K/s and 10K/s rapidly increases to 0.025% but only
above 1273K, and then decreases with increasing temperature until equilibrium solidus is reached. A similar
trend occurs at 102K/s but starting from 1373K and reaches a slightly lower maximum volume fraction than
at 1K/s. At 103K/s, the volume fraction starts to increase from 1473K and reaches to ~ 0.008% at equilibrium
solidus. At 2.08 × 103K/s, the volume fraction starts to increase from ~ 1573K and reaches a maximum of
0.0025%, which is an order of magnitude lower than that occurring during cooling at 2 × 104K/s. For heating
rates ≥ 104K/s, the change in M–S volume fraction is negligible at any temperature.
e most important results of the precipitation kinetics simulations are as follows: (i) there is no change to
the M–S average diameter or volume fraction below 1273K for any temperature rate, (ii) during solidication/
cooling, the M–S volume fraction only increases at cooling rates lower than 105K/s in magnitude, and (iii) dur-
ing heating, the maximum M–S volume fraction increases at rates lower than 104K/s.
Combining the results obtained from the FE simulations (Fig.7) and the precipitation kinetics simulations,
it can be seen that the M–S volume fraction increase during solidication/cooling at 1.15 × 104 – 2.91 × 104K/s
is higher by an order of magnitude than that during heating at the lowest rate 2.08 × 103K/s. erefore, the
nucleation and growth of M–S on oxide inclusions should be signicant only during solidication/cooling at
temperatures above 1273K; the heating should play a negligible role. is explains why M-S inclusions have
been observed in TEM lamellae L1–L5.
Since the inert gas atomization of 316LSS is expected to have similar maximum solidication/cooling rates as
the LMD process, the precipitation kinetics simulations also explain why M–S inclusions in LMD 316LSS have
a similar size to those observed in TEM lamella LP taken from the inert gas atomized 316LSS powder (Fig.5).
Furthermore, the analysis also explains why M–S inclusions have not been observed in TEM studies of SLM
316LSS24,7. During SLM, the maximum temperature rates of 106–107K/s can be encountered at temperatures
above 1273 K23. At these rates, the predicted average diameters of M–S precipitates are very small and their
volume fraction is negligible (Fig.9), which makes it very dicult to detect them even with high-resolution
STEM imaging.
Figure9. M–S precipitation kinetics simulation predicted (a) average diameter (log scale) and (b) volume
fraction as a function of temperature during solidication/cooling at dierent temperature rates. ‘EQ solidus’
and ‘NEQ solidus’ respectively stand for the equilibrium and non-equilibrium solidus temperatures; the latter
(1643K) is obtained from Dépinoy etal.30 e black curves in (a,b) correspond to approximately the average
cooling rate obtained from Fig.7 between 1273 and 1673K. e red and green empty circles in (a) represent the
M–S precipitate diameters reported by Kim etal.16 for cooling rates 1K/s and 8.33K/s, respectively. e black
empty circle represents the non-oxide inclusion diameter in mixed precipitates obtained from Fig.6. Gnuplot
v5.4 (http:// www. gnupl ot. info) was used to prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Finally, although precipitation kinetics of non-oxide inclusions other than M–S have not been performed,
nevertheless, similar thermodynamic arguments can also be used to explain their presence in LMD 316LSS
and inert gas atomized 316LSS studied in this work, as well as their absence in SLM 316LSS in other works24,7.
In conclusion, the TEM studies performed in this work have revealed a plethora of dierent kinds of non-
oxide inclusions (carbides, suldes, phosphides and intermetallics) in the LMD 316LSS samples studied in this
work. Furthermore, 70.9% of all the precipitates found in the TEM lamellae extracted from the LMD 316LSS
samples are either purely non-oxide inclusions or mixed precipitates containing non-oxide inclusions together
with oxide inclusions. ese results prove that non-oxide inclusions can indeed form during DED-type AM
processes such as LMD.
e presence/absence of non-oxide inclusions is dependent on the temperature rates experienced during
processing as well as a minimum temperature below which they do not form. e precipitation kinetics simula-
tions performed in this work show that new non-oxide inclusions form only during solidication/cooling (and
not heating) at temperatures above 1273K (1000°C) i.e., 0.76Tm, and at temperatures rates less than or equal to
105K/s. e FE simulations performed in this work, supported by existing literature22, reveal that the highest
temperature rate magnitudes encountered during the LMD of 316LSS samples are less than 105K/s at tempera-
tures above 1273K; these thermal conditions occur only during the initial stages of the heating–cooling cycles in
a material just aer its deposition. Furthermore, similar rates must be experienced during inert gas atomization
because the size of inclusions in gas atomized 316LSS are similar to the ones obtained in LMD 316LSS. ese
predictions explain the presence of non-oxide inclusions in both inert gas atomized 316LSS powder and LMD
316LSS samples. Meanwhile, the highest temperature rates typically encountered during powder-bed based tech-
niques such as SLM at temperatures above 1273K are in the range 106–107K/s23. erefore, non-oxide inclusions
should not form during SLM 316LSS, which is fully supported by existing TEM studies24,7.
316LSS powder properties and LMD process parameters. e 316LSS powder has been produced
via the inert gas atomization process. is process involves slowly melting a wrought alloy and pouring it into
an atomization chamber from the top. While being poured, the molten alloy is acted upon by high-speed inert
gas (Ar) jets that disperse it into smaller droplets. ese droplets rapidly solidify and cool during their descent
and they are collected into a crucible at the bottom of the chamber. e solidication/cooling temperature rates
attained during this process depend on several factors such as desired powder particle sizes, gas used, gas jet
speed, etc. A sieve analysis has revealed the particle sizes range from 45 to 90μm. Following the gas atomization
process, the powder is typically stored in a sealed container, transported to the location where it is used for AM,
and stored in a powder feeding chamber for the AM process.
e powder-based LMD process, also known as Direct Metal Deposition (DMD) or Laser Engineered Net
Shaping (LENS), had been carried out inside the “Mobile” machine from BeAM. More details on the material,
powder characteristics and specications of the LMD machine can be found in 31,32.
A single-track bidirectionally-printed 3-layer wall had been built on a hot-rolled 316LSS substrate using this
machine (Fig.1a). A 3-layer wall ensures that (i) precipitates formed in the top layer are not aected by chemical
heterogeneities on the substrate surface, and (ii) solidication/cooling rates are similar to those experienced by
the rst deposited layer during its deposition and comparable to those occurring during gas atomization. e
LMD process parameters used to print the samples are summarized in Table2. e nal dimensions of the as-
built wall are 100mm × 0.8mm × 0.6mm.
During the LMD process, the 316LSS powder particles are mixed with an inert gas (Ar in our case) and then
transported to the focusing head, which contains laser ber optic cable, gas and powder inlet, focusing lenses
and a nozzle to direct the powder enveloped in inert gas. e laser melts the powder and the molten powder gets
Figure10. M–S precipitation kinetics simulation predicted volume fraction as a function of temperature
during heating at dierent temperature rates. ‘EQ solidus’ stands for the equilibrium solidus temperature.
e black curve corresponds to the minimum heating rate obtained from Fig.7 between 1273 K and 1673K.
Gnuplot v5.4 (http:// www. gnupl ot. info) was used to prepare this gure.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
deposited on the substrate or the part being built. While the focusing head continues its trajectory to deposit the
layer, the molten powder solidies, typically in a few milliseconds, and then cools down, until the subsequent
laser passage either on the same layer but adjacent to the material that is just deposited or on the layer above the
deposited material. Subsequent deposition results in thermal cycling of the material in the solid-state and the
number of cycles, their amplitudes and temperature rates depend on the number of laser passages and layers
that remain to be built.
Scanning Electron Microscopy (SEM): equipment and TEM lamellae preparation tech‑
nique. e as-built wall and substrate had been mechanically cut near the mid-section and along the direc-
tion normal to the build and print directions (section B-B in Fig.1a). is cross-section had been polished rst
using SiC papers with dierent grits (from 800-grit to 4000-grit), followed by diamond paste polishing with grit
sizes 3μm and 1μm, and nally ion polishing. en, this cross-section has been analyzed via SEM using a FEI
Quanta 650 FEG Environmental-SEM microscope equipped with the symmetry detector (Oxford Instruments)
for EBSD measurements.
e EBSD map shown in Fig.1b for the 3-layer wall is a single map with a resolution of 2000 (height) × 1750
(width) pixels (1 × 0.88 mm2) of step size 0.5μm. e acceleration voltage of the electron beam was 20kV. e
indexing success rate was 99%. e EBSD map shown in Fig.S1 for the 60-layer wall was stitched from 80 images
of 1005 (width) × 692 (height) pixels (0.51 × 0.35 mm2) each with step size 0.5μm. e acceleration voltage of
the electron beam was 30kV. e indexing success rate was 97.4%.
e 316LSS powder has also been analyzed via SEM. Some powder particles had been embedded inside a
diluted conductive carbon cement and polished using the same approach as above.
Following the preliminary EBSD analysis, the sample had been transported to a FEI Helios Nanolab 660
dualbeam SEM microscope, which is equipped with a dual beam Focused Ion Beam (FIB). Using the standard
li-out process, the FIB-SEM is used to extract 4 TEM lamellae: 3 lamellae (L1, L2 and L3) from the 3-layer wall,
2 lamellae (L4 and L5) from the 60-layer wall, and 1 lamella (LP) from the powder sample.
TEM: equipment and analysis
e six lamellae had been studied in an aberration-corrected FEI Titan3 G2 60–300 TEM microscope operating
at 300kV. is TEM is equipped with a Cs probe corrector, and a series of detectors including a High-Angle
Annular Dark Field (HAADF) detector. It is also equipped with 4 EDS detectors, which allow generating chemi-
cal maps.
EDS takes advantage of the X-ray uorescence to analyze the composition of the sample.
Working in STEM mode, it is possible to scan the sample and store the local X-ray spectrum at each pixel
in the scanned zone. However, characteristic X-ray peaks from each element are convoluted with background
contributions as well as any overlap with peaks of other elements; for example, the
peak of S has an overlap
with the
peak of Mo. It is necessary to deconvolute these peaks in order to generate images separating the
composition of each element in the scanned zone. e deconvolution procedure ts the experimental peaks using
a linear combination of theoretical peaks of all the alloying elements in 316LSS. is method also removes the
background contributions (e.g. Cu from the TEM grid). Following deconvolution, images are quantied using
the standardless Cli-Lorimer method33 to link experimental intensities to the relative amount of each element.
All EDS images shown in this work are deconvoluted and quantied such that they are devoid of any thickness,
density or peak overlap. EDS images of elements with lower concentration require brightness adjustment to
facilitate visualization.
EDS line prole plots have been generated from the quantied EDS spectra for multiple precipitates in order
to study the variation in elemental composition across dierent line paths through the precipitates. e plots
have been generated by averaging over a width of 10 pixels.
Precipitate size (equivalent diameter) determination has been performed from the TEM images using the
Fiji soware.
Heat transfer FE simulation setup
Solid-state heat transfer FE simulations have been performed to simulate LMD of the 3-layer wall in Fig.1 in
order to generate the temperature v/s time curves shown in Fig.7.
Governing equations and variational formulation. e heat transfer model used in this work is a
continuum-based initial boundary value problem that accounts for heat conduction, convection and radiation
based on the work of Weisz-Patrault34. Let
be a temporally evolving domain with boundary
. Let
T=[0, tmax]
be the time interval of interest, where
is known. e governing equations for the heat transfer
problem at a material point
in (t)×T
Table 2. LMD process parameters used to build the 316LSS wall in Fig.1.
Laser power Powder ow rate Deposition speed Initial vertical position / vertical displacement of focusing head
225W 0.1083g/s 33.33mm/s 3.5 mm above substrate / 0.2mm aer deposition of one layer
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
e initial conditions are:
e boundary conditions are:
is the absolute temperature,
is the second order thermal conductivity tensor,
is the specic heat at
constant volume,
is the dierential operator vector,
is the heat ux vector,
is the temperature gradient,
is the initial temperature
is the bottom surface of the baseplate that does not evolve with time and it
is applied a Dirichlet boundary condition.
are dened as those sur-
faces that have been respectively subjected to surface heat ux from the laser beam
, forced convective heat
due to the Ar-gas owing from the nozzle of the focusing head, natural air convective heat transfer
and radiative heat ux
, respectively; note that these surfaces evolve with time.
are the con-
vective heat transfer coecients for Ar-gas and air, respectively.
are the Stefan-Boltzmann constant and
emissivity of the deposited material, respectively.
is the innitive temperature. e laser beam interaction
with the material is modeled as a surface heat ux condition having a 2D Gaussian distribution with maximum
and absorptivity
acting on a circular area with radius
centered at
. e Ar-gas heat
ux is also modeled as a Gaussian with its peak at the center of the laser beam
and a radius
RAr >Rlaser
that is large enough to cover some area around the build.
At a given time step, the following variational (weak) form of the governing equations with the boundary
conditions is solved:
where a Euler backward time integration scheme has been used,
is the temperature test function,
is the
approximate solution sought,
· ∇
and the subscript
is the current time step.
e weak form has been implemented in the FEniCS35 open-source (LGPLv3) FE computational platform
for solving partial dierential equations and the non-linear Newton iterative solver is used.
Mesh. e simulated geometry shown is created based on the experimental setup (Fig.1). It consists of a
baseplate (not shown in Fig.1 but present during the experiment), the hot-rolled 316LSS substrate, and the
3-layer LMD 316LSS wall. e baseplate, substrate and wall are created as independent volumes and merged to
form a single material volume for the simulations. e geometry is then decomposed into subdomains as shown
in Fig.S5 and meshed using the open-source gmsh36 soware.
e structured regions of the substrate and baseplate were meshed using 6mm seeds and the structured region
of wall was meshed using 0.1mm seeds. An unstructured region was necessary to transition from the 0.1mm
seeds of the wall to the 6mm seeds of the substrate and baseplate. is unstructured region starts around the wall
and goes to a distance of 15mm in all directions to ensure a uniform increase in mesh element size. All elements
are chosen to be tetrahedral with quadratic interpolation (P2). A conformal mesh is obtained aer recombining
using “transnite” mesh options in gmsh (Fig.S6).
For simplicity, mass addition is simulated via a mesh element addition procedure. e initial mesh only
contains the substrate and the baseplate. When the laser passes over the region where the rst mass is added, a
new element is generated at the location where the laser passes. However, this element is not added to the exist-
ing mesh. Instead, a new geometry and mesh is generated. is new mesh is an augmentation of the geometry
and mesh of the previous time step with one additional geometric element to mimic the deposition of some
material on a layer. At the beginning of the deposition of each layer, a cuboid element of size 0.6mm (along
) × 0.6mm (along
) × 0.2667mm (along
) is generated. Each of the subsequent elements added to that layer
are of size 0.1mm × 0.6mm × 0.2667mm. Finally, for the 3-layer simulation, the initial and nal FE meshes are
composed of 170056 and 457948 elements, respectively, with approximately 100 mesh elements added for the
additional geometry element.
· ∇
in (t)×T
)=Tini x(0)
T(x,t)=T0on ∂�base ×T
laser (x,t)=2ηlaser
on ∂�laser(t
Ar (x,t)=hAr
on ∂�Ar(t
hair T
on ∂�
on ∂�(t
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Simulation setup. e FE simulation parameters are presented in Table3. e baseplate has the same
length and breadth as the substrate but its thickness is 40mm (5 times the thickness of the substrate in Fig.1).
At the rst timestep, a cuboidal element of 0.6mm × 0.6mm × 0.2667mm is generated and initialized to 1885K.
e laser beam and the Ar-gas jet are centered on the top surface of this element and although the element
is initialized to 1885K, the action of the Ar-gas almost instantaneously reduces the temperature to ~ 1773K
(1500°C), which is similar to the melt temperature at deposition. All the remaining surfaces are provided with
natural air ow
boundary condition. In addition, radiation heat loss condition
is imposed over all sur-
(∂�rad =∂�
e simulation starts with the rst deposition at
s. e simulation time step is 3ms. Deposition of
each layer takes 994 steps i.e., 2.982s; this value corresponds to a deposition speed of 33.53mm/s, which is very
close to LMD deposition speed of 33.33mm/s. Between successive layer depositions, a dwell time of 1.221s
(407 steps) is provided to account for focusing head deceleration and reverse acceleration. Aer the deposition
of the nal layer, a dwell time of 0.609s (204 steps) is provided. erefore, the total simulation steps are 4000,
which correspond to a total simulated time of 12s. e simulation was performed on a single CPU thread and
it took 22h 26min 12s to complete.
Thermodynamic calculations
e phase stability and elemental solubility analysis is performed using the ermo-Calc soware version 2019b
with the TCFE9 database.
A mean eld precipitate kinetics model based on the classical nucleation and growth theories3741 is used
to predict the nucleation, growth and coarsening kinetics of M–S precipitates on MnSiO3 precipitates already
present in γ-austenite/LA. M–S precipitates are assumed to be spherical and their growth is assumed to be con-
trolled by the diusion in γ-austenite/LA. eir growth rate is calculated following Zener’s approximate solution
for spherical precipitates42. e Gibbs–omson eect43 is also taken into account. γ-austenite/LA is assumed
to be an ideal supersaturated solid/liquid solution containing both interstitial and substitutional elements. e
solubility of each element in γ-austenite, MnSiO3, M-S, LA and LB is taken from the ermoCalc soware
generated data used to plot Fig.8.
An important step in the precipitation kinetics simulations is to estimate the number of potential nucleation
sites for M-S. In order to understand how these sites are chosen, the TEM results for L1 – L5 are reconsulted
(Fig.3, S3 and S6). ese results show that 72.22% of all the mixed precipitates contain oxide inclusions at
their core and non-oxide inclusions surround them. ese numbers strongly suggest that non-oxide inclusions
mainly nucleate on oxide inclusions, a phenomenon that has already been observed in other studies performed
at slower cooling rates15,16. Meanwhile, Fig.6 shows that (i) the average size of oxide precipitates is larger than the
average size of oxide inclusions in mixed precipitates and (ii) with the exception of 2 oxide inclusions in mixed
precipitates, the remaining oxide inclusions are either in the size range of oxide precipitates or smaller than the
smallest oxide precipitate. ese results show that the growth of non-oxide inclusions could stunt further growth
of oxide inclusions. erefore, when computing the number of potential nucleation sites for M–S, the following
assumptions have been made: (i) only one M–S inclusion grows on one oxide inclusion, and (ii) the diameter
of an oxide inclusion is the average size of the oxide inclusions in mixed precipitates from Fig.6 i.e., 116.19nm.
Oxide inclusions are assumed to be only MnSiO3 and to have a spherical shape. We also assume that the
maximum possible equilibrium phase fraction of MnSiO3 (Fig.8) is present in γ-austenite/LA. e composition
of γ-austenite/LA is assumed to be such that all oxygen is used up to form MnSiO3; it is as follows (in at. %):
Cr—18.063, Ni—12.027, Mo—1.448, Mn—1.464, Si—1.331, P—0.025, S—0.009, C—0.05 and Fe—balance. In
the following, a maximum volume fraction of
max =0.3%
(rounded up from the equilibrium calculation below
solidus in Fig.8) is considered for MnSiO3.
e diameter of MnSiO3 precipitates is taken as the average diameter of oxide inclusions in mixed precipitates
from Fig.6. From the average mixed oxide volume (
r=58.095 ×109
m from Fig.6) and
the MnSiO3 number density (
m-3) can be estimated as:
Table 3. Parameters for the FE simulations appearing in Eqs.(1)–(9).
(W m−2 K−4)5.67 × 10–8
(J kg−1 K−1) 500
(W m−2 K) 12,500 × 106
(W m−2 K) 15 × 106
(K) 300
(mm) 20
(mm) 0.6
(W mm−1 K−1) 16.3
(kg m−3) 8000
(W) 225
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
We have assumed that only one M–S precipitate nucleates on one MnSiO3. erefore,
is the potential
number of nucleation sites for M–S nucleation on MnSiO3. Finally, the diusion coecient is given as:
(m2/s) is the pre-exponent factor,
(J/mol) is the activation energy,
is the universal gas constant,
is the temperature, and
is a tting parameter.
e parameters used for the precipitate kinetics simulations are reported in Table4; some were either obtained
from the literature41,44 or tted to reproduce the measured non-oxide inclusion size reported in Fig.6. For further
information about this well documented model see38,40,45. e model calculates a particle size distribution fol-
lowing the Euler-type multi-class approach38 and only the average diameter of the particle is plotted in Fig.9a.
Received: 6 February 2021; Accepted: 4 May 2021
1. DebRoy, T. et al. Additive manufacturing of metallic components—Process, structure and properties. Prog. Mater Sci. 92, 112–224
2. Chao, Q. et al. On the enhanced corrosion resistance of a selective laser melted austenitic stainless steel. Scripta Mater. 141, 94–98
3. Saeidi, K., Kvetková, L., Lofaj, F. & Shen, Z. Austenitic stainless steel strengthened by the insitu formation of oxide nanoinclusions.
RSC Adv. 5, 20747–20750 (2015).
4. Wang, Y. M. et al. Additively manufactured hierarchical stainless steels with high strength and ductility. Nat. Mater. 17, 63–71
5. Smith, T. R., Sugar, J. D., San Marchi, C. & Schoenung, J. M. Strengthening mechanisms in directed energy deposited austenitic
stainless steel. Acta Mater. 164, 728–740 (2019).
6. Tan, Z. E., Pang, J. H. L., Kaminski, J. & Pepin, H. Characterisation of porosity, density, and microstructure of directed energy
deposited stainless steel AISI 316L. Addit. Manuf. 25, 286–296 (2019).
7. Deng, P. et al. e origin and formation of oxygen inclusions in austenitic stainless steels manufactured by laser powder bed fusion.
Addit. Manuf. 101334. https:// doi. org/ 10. 1016/j. addma. 2020. 101334 (2020).
8. Zheng, B., Zhou, Y., Smugeresky, J. E., Schoenung, J. M. & Lavernia, E. J. ermal behavior and microstructure evolution during
laser deposition with laser-engineered net shaping: Part II. Experimental investigation and discussion. Metall. Matter Trans. A 39,
2237–2245 (2008).
9. Gladman, T. Precipitation hardening in metals. Mater. Sci. Technol. 15, 30–36 (1999).
10. Wasnik, D. N., Dey, G. K., Kain, V. & Samajdar, I. Precipitation stages in a 316L austenitic stainless steel. Scripta Mater. 49, 135–141
11. Han, G. et al. Properties of oxide lms formed on 316L SS and model alloys with modied Ni, Cr and Si contents in high tempera-
ture water. Corros. Sci. 106, 157–171 (2016).
12. Wonneberger, R. et al. Oxidation of stainless steel 316L—Oxide grains with pronounced inhomogeneous composition. Corros.
Sci. 149, 178–184 (2019).
13. Huang, X. et al. Oxidation behavior of 316L austenitic stainless steel in high temperature air with long-term exposure. Mater. Res.
Express 7, 066517 (2020).
14. Wakoh, M., Sawai, T. & Mizoguchi, S. Eect of S content on the MnS precipitation in steel with oxide nuclei. ISIJ Int. 36, 1014–1021
15. Kim, H. S., Lee, H.-G. & OH, K.-S. Precipitation behavior of MnS on oxide inclusions in Si/Mn deoxidized steel. Metals Mater. 6,
305–310 (2000).
16. Kim, H. S., Lee, H.-G. & Oh, K.-S. MnS precipitation in association with manganese silicate inclusions in Si/Mn deoxidized steel.
Metall. Mater Trans. A 32, 1519 (2001).
17. Tanaka, Y., Pahlevani, F., Moon, S.-C., Dippenaar, R. & Sahajwalla, V. Insitu characterisation of MnS precipitation in high carbon
steel. Sci. Rep. 9, 10096 (2019).
22 =3.65 ×1018/m
RT ,
Table 4. Precipitation kinetics model parameters41,44.
Molar volume of γ-austenite (m3/mol) 7 × 10–6
Molar volume of M-S (m3/mol) 1.1 × 10–5
Interfacial energy (J/m2) 0.25
Heterogeneous nucleation factor 0.5
Mn in γ-austenite D0 = 0.178 × 10–4, Q0 = 2.64 × 105, α = 1
Mn in LA D0 = 1.8 × 10–7, Q0 = 1.3 × 104, α = 1
Cr in γ-austenite D0 = 0.169 × 10–4, Q0 = 2.639 × 105, α = 1
Cr in LA: D (m2/s) D = 4.9 × 10–9
S in γ-austenite D0 = 7.52 × 10–4, Q0 = 2.364 × 105, α = 1
S in LA D0 = 4.33 × 10–8, Q0 = 3.5 × 104, α = 1.45
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
18. Mandiang, Y. & Cizeron, G. Phosphorus eects on mechanical properties changes in type 316 Ti stainless steels. Mater. Sci. Eng.
A 206, 241–248 (1996).
19. Hsieh, C.-C. & Wu, W. Overview of intermetallic sigma ( σ ) phase precipitation in stainless steels. ISRN Metall. 2012, e732471.
https:// www. hinda wi. com/ journ als/ isrn/ 2012/ 732471/ (2012).
20. Warren, A. D., Griths, I. J. & Flewitt, P. E. J. Precipitation within localised chromium-enriched regions in a type 316H austenitic
stainless steel. J. Mater. Sci. 53, 6183–6197 (2018).
21. Ellingham, H. J. T. Reducibility of oxides and sulphides in metallurgical processes. J. Soc. Chem. Ind. 63, 125–160 (1944).
22. Hofmeister, W. & Grith, M. Solidication in direct metal deposition by LENS processing. JOM 53, 30–34 (2001).
23. Scipioni Bertoli, U., Guss, G., Wu, S., Matthews, M. J. & Schoenung, J. M. In-situ characterization of laser-powder interaction and
cooling rates through high-speed imaging of powder bed fusion additive manufacturing. Mater. Des. 135, 385–396 (2017).
24. Riabov, D., Hryha, E., Rashidi, M., Bengtsson, S. & Nyborg, L. Eect of atomization on surface oxide composition in 316L stainless
steel powders for additive manufacturing. Surf. Interface Anal. 52, 694–706 (2020).
25. Mullis, A. M., Farrell, L., Cochrane, R. F. & Adkins, N. J. Estimation of cooling rates during close-coupled gas atomization using
secondary dendrite arm spacing measurement. Metall. Mater. Trans. B 44, 992–999 (2013).
26. Shukla, P., Mandal, R. K. & Ojha, S. N. Non-equilibrium solidication of undercooled droplets during atomization process. Bull.
Mater. Sci. 24, 547–554 (2001).
27. Zeoli, N. & Gu, S. Computational simulation of metal droplet break-up, cooling and solidication during gas atomisation. Comput.
Mater. Sci. 43, 268–278 (2008).
28. Barkia, B. et al. On the origin of the high tensile strength and ductility of additively manufactured 316L stainless steel: Multiscale
investigation. J. Mater. Sci. Technol. 41, 209–218 (2020).
29. Höganäs portfolio for additive manufacturing. https:// www. hogan as. com/ globa lasse ts/ downl oad- media/ share point/ broch ures-
and- datas heets--- all- docum ents/ addit ive- manuf actor ing_ hogan as- addit ive- stand ard- portf olio_ 2853h og. pdf.
30. Dépinoy, S., Sennour, M., Ferhat, L. & Colin, C. Experimental determination of solute redistribution behavior during solidication
of additively manufactured 316L. Scr. Mater. 194, 113663 (2021).
31. Balit, Y., Charkaluk, E. & Constantinescu, A. Digital image correlation for microstructural analysis of deformation pattern in
additively manufactured 316L thin walls. Addit. Manuf. 31, 100862 (2020).
32. Balit, Y. et al. High resolution digital image correlation for microstructural strain analysis of a stainless steel repaired by directed
energy deposition. Mater. Lett. 270, 127632 (2020).
33. Cli, G. & Lorimer, G. W. e quantitative analysis of thin specimens. J. Microsc. 103, 203–207 (1975).
34. Weisz-Patrault, D. Fast simulation of temperature and phase transitions in directed energy deposition additive manufacturing.
Addit. Manuf. 31, 100990 (2020).
35. Alnæs, M. et al. e FEniCS Project Version 1.5. ans 3 (2015).
36. Geuzaine, C. & Remacle, J.-F. Gmsh: A 3-D nite element mesh generator with built-in pre- and post-processing facilities. Int. J.
Numer. Methods Eng. 79, 1309–1331 (2009).
37. Kozeschnik, E., Svoboda, J. & Fischer, F. D. Modied evolution equations for the precipitation kinetics of complex phases in multi-
component systems. Calphad 28, 379–382 (2004).
38. Perez, M., Dumont, M. & Acevedo-Reyes, D. Implementation of classical nucleation and growth theories for precipitation. Acta
Mater. 56, 2119–2132 (2008).
39. Rougier, L. et al. Numerical simulation of precipitation in multicomponent Ni-base alloys. Acta Mater. 61, 6396–6405 (2013).
40. Hou, Z. et al. Quantitative modeling and experimental verication of carbide precipitation in a martensitic Fe-0.16wt% C-4.0wt%
Cr alloy. Calphad 53, 39–48 (2016).
41. You, D., Michelic, S. K., Wieser, G. & Bernhard, C. Modeling of manganese sulde formation during the solidication of steel. J.
Mater. Sci. 52, 1797–1812 (2017).
42. Zener, C. eory of growth of spherical precipitates from solid solution. J. Appl. Phys. 20, 950–953 (1949).
43. Perez, M. Gibbs-omson eects in phase transformations. Scr. Mater. 52, 709–712 (2005).
44. Oikawa, H. Lattice diusion in iron-A review. Tetsu-to-Hagane 68, 1489–1497 (1982).
45. Gaudez, S. Kinetics and Microstructural Evolutions During the Tempering of Martensitic and Nano-Bainitic Low Alloyed Steel : InSitu
Experimental Study and Modelling. (Université de Lorraine, 2021).
M.V.U., M.B.H.S. and E.H. are grateful to the Fédération Francielliene de Mécanique (F2M) for their support
through the Coup de Pouce 2019 grant. M.V.U. and S.G. are grateful to the European Research Council (ERC) for
their support through the European Unions Horizon 2020 research and innovation programme (Grant agreement
No. 946959). N. M. is grateful to the doctoral school of Ecole Polytechnique for their support through the PhD
scholarship program. FIB-SEM and TEM work was carried out using the facilities available at the MSSMat labo-
ratory within the MATMECA consortium, which is supported by the Agence National de la Recherche (ANR)
under the contract number ANR-10-EQPX-37. Finally, the authors would like to thank Y. Balit for performing
the 3D printing of the LMD sample used in this work.
Author contributions
M.V.U. conceived the main idea, designed and led this study. M.V.U. and M.B.H.S. are the main contributors
to the writing of this manuscript, designing of the gures and tables. S.G. developed the precipitation kinetics
model and performed the ermoCalc and precipitation kinetics simulations. N.M. developed the FE imple-
mentation of the heat transfer model and performed the FE simulations of the LMD process. L.Y. performed the
TEM studies while being assisted by M.V.U and M.B.H.S. M.B.H.S. and L.Y. performed the post-processing of
the TEM data. E.H. prepared the TEM lamellae. S.H. performed the EBSD studies. A.T. helped with the sample
preparation for EBSD analysis and assisted with the EBSD studies. All authors contributed to the discussions
related to the results and their analysis.
Competing interests
e authors declare no competing interests.
Additional information
Supplementary Information e online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 021- 89873-2.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Scientic Reports | (2021) 11:10393 | 
Correspondence and requests for materials should be addressed to M.V.U.
Reprints and permissions information is available at
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access is article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. e images or other third party material in this
article are included in the articles Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© e Author(s) 2021
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
... Until recently, precipitates occurring in as-built LMD 316LSS or SLM 316LSS were reported to be only Mn-Si-rich oxide precipitates [22,[24][25][26][27][28][29]. However, in a recent TEM study on five lamellae extracted from an LMD 316LSS thin-wall, Upadhyay et al. [30] reported the presence of a large volume fraction of non-oxide precipitates including sulfides, carbides, phosphides and intermetallics, along with Mn-Si-rich oxide precipitates. The non-oxide precipitates were found to be smaller in their average size in comparison to the oxide precipitates. ...
... The non-oxide precipitates were found to be smaller in their average size in comparison to the oxide precipitates. An investigation into the origin of these non-oxide precipitates [30], with the help of finite element and precipitation kinetics simulations, revealed that non-oxide precipitates can indeed occur in LMD 316LSS but not in SLM 316LSS. The simulation results also suggested that precipitates could form/disappear and evolve in the solid-state. ...
... The material and most of the experimental setup used in this work has been described in detail in Upadhyay et al. [30]. In the following, only the essential details are recalled for the sake of completeness and self-sufficiency of this paper. ...
Full-text available
During additive manufacturing of alloys, just after local heat-matter interactions, a molten material undergoes rapid solidification. Then, for the rest of the building time, it is subjected to cooling/heating cycles in the solid-state i.e., solid-state thermal cycling. The thermo-mechanical forces generated during solid-state thermal cycling can trigger a plethora of micro-mechanisms that can bring about significant microstructural changes that determine the eventual mechanical properties of as-built parts. In this work, the aim is to gain insight on solid-state thermal cycling-driven evolution of submicron-sized precipitates in an austenitic stainless steel using transmission electron microscopy. To that end, thin-film lamellae are extracted from a pre-built sample and subjected to different in-situ solid-state thermal cycles inside a transmission electron microscope. The solid-state thermal cycles are designed to understand the role of temperature amplitude and rates, number and type of thermal cycles, and post-process annealing on precipitate evolution. High angle annular dark field imaging and energy dispersive X-ray spectroscopy before and after each thermal cycle provide a deep insight on the contribution of different thermal cycling factors on the evolution of precipitate composition, size and morphology. Common trends include diffusion of Mn and Si from Mn-Si-rich oxides into the surrounding matrix, formation of Cr rings around oxide precipitates and S redistribution in non-oxide precipitates. Similar Cr rings and S distributions were also found in precipitates in as-built samples studied in (Upadhyay et al., Sci. Rep. 11 (2021) 10393), which strongly supports the representativeness of these results with respect to what occurs during additive manufacturing.
... Vacuum melted argon gas atomised stainless steel AISI 316L powder (35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50) μm) supplied by Mimete S.r.l. and nanometre sized (30-100 nm) hexagonal WC powder supplied by US Research Nanomaterials Inc. were used in this study. The morphology of these powders is shown in Fig. 1. ...
... The precipitated hard Fe 2 Si phase plays a crucial role in determining the material response. However, while improving hardness and wear, corrosion, fatigue and fracture resistance of the specimens could be compromised by the brittle intermetallic Fe 2 Si precipitates [46][47][48]. An evident diffraction peak of SiC was observed in the specimens' spectrum and confirmed by the JCPDS card 89-1396. ...
Functionally grading material composition in laser-powder bed fusion grants the potential for manufacturing complex components with tailored properties. The challenge in achieving this is that the current laser-powder bed fusion machine technology is designed to process only powdered feedstock materials. This study presents a multi-feedstock material printing methodology for laser-powder bed fusion. Utilising colloid nebulisation, tungsten carbide nanoparticles were successfully deposited onto powder beds of stainless steel 316L during the laser-powder bed fusion process. By this means, a controlled volume of tungsten carbide nanoparticles was uniformly dispersed onto powder beds under the inert processing chamber atmosphere. As a result, specimens printed with this methodology showed an increase in strength. Similarly, the colloid medium played an important role in the resulting microstructures. It led to the formation of consistent and stable meltpools and a strong crystallographic texture. Recommendations are given for the successful dispersion of higher volumes of nanoparticles. Additionally, insights into application prospects for material nebulisation in laser-powder bed fusion are presented and discussed.
... The influence of other oxide inclusions (summarized in Table 2) on the corrosion performance of LPBF-316L is not discussed in the literature. Besides LPBF, researchers [225] had reported nano-inclusions with different compositions, including sulphur-rich inclusion in 316L manufactured by laser metal deposition (LMD). However, the influence of these inclusions on the corrosion performance of the alloy is not reported in the literature. ...
Full-text available
The applications of laser-powder bed fusion (LPBF), an emerging additive manufacturing (AM) technique, are rapidly growing in various industries. The superior and consistent mechanical and corrosion properties of LPBF-printed components are essential for engineering applications. The 316L stainless steel (SS) is an essential alloy with widespread applications from household items to nuclear and aerospace industries. Extensive research is conducted to understand and improve the mechanical properties of LPBF printed 316L SS. Studying the corrosion behavior of LPBF printed 316L has attracted only limited attention. Additionally, a discrepancy in the corrosion performance of LPBF printed 316L has been reported due to the complex microstructure and defects introduced during LPBF. Therefore, understanding the influence of processing parameters and feedstock on defects and microstructure becomes critical in understanding the processing-corrosion relationships and producing LPBF printed 316L components with reproducible properties. This review presents the influence of feedstock, processing parameters, and post-processing techniques on manufacturing defects, microstructure, and corrosion performance of LPBF printed 316L. Strategies and hypotheses to improve the corrosion resistance of LPBF printed 316L are also presented.
... The addition of manganese also proves suitable only to a limited extent. When active gas is used as a shielding gas, the alloying elements of silicon and manganese have an affinity to react with oxygen, act as getters and form detrimental silicon and manganese oxides [63][64][65][66][67][68]. Non-metallic inclusions and oxides form especially on a large scale in AM-manufactured components and have a detrimental effect on the mechanical properties of these components [69]. ...
Full-text available
In the present study, the application and tailoring of the alloy composition of chromium martensitic hot-work steels using metal cored wires (MCW) for wire arc additive manufacturing (WAAM) in a modified short-circuit metal transfer process is demonstrated. The nickel content was varied and the alloys were fabricated as tubular-cored wires with various powder fillings. By recording the material transfer at high speed during processing, evidence was gathered indicating the suitability of the fabricated cored wires for WAAM. Optimized process parameters were identified by taking a Design of Experiment (DoE) approach and additive manufacturing (AM) structures were fabricated from the chromium martensitic hot-work tool steel alloys. The microstructure and mechanical properties of the parts were subsequently characterized. The phase fraction of the polygonally shaped delta ferrite could be reduced and microstructural refinement could be achieved by adding nickel to the investigated hot-work tool steel. In addition to molybdenum-enriched precipitates that covered the grain boundaries, randomly scattered non-metallic inclusions and oxides were observed. Modifying the microstructure by adding nickel also affects the mechanical properties of the product: an increase in hardness, impact toughness and yield strength as the nickel content increased in the AM structures fabricated by WAAM was observed.
Full-text available
New developments at synchrotron beamlines and the ongoing upgrades of synchrotron facilities allow the possibility to study complex structures with a much better spatial and temporal resolution than ever before. However, the downside is that the data collected are also significantly larger (more than several terabytes) than ever before, and post-processing and analyzing these data is very challenging to perform manually. This issue can be solved by employing automated methods such as machine learning, which show significantly improved performance in data processing and image segmentation than manual methods. In this work, a 3D U-net deep convolutional neural network (DCNN) model with four layers and base-8 characteristic features has been developed to segment precipitates and porosities in synchrotron transmission X-ray micrograms. Transmission X-ray microscopy experiments were conducted on micropillars prepared from additively manufactured 316L steel to evaluate precipitate information. After training the 3D U-net DCNN model, it was used on unseen data and the prediction was compared with manual segmentation. A good agreement was found between both segmentations. An ablation study was performed and revealed that the proposed model showed better statistics than other models with lower numbers of layers and/or characteristic features. The proposed model is able to segment several hundreds of gigabytes of data in a few minutes and could be applied to other materials and tomography techniques. The code and the fitted weights are made available with this paper for any interested researcher to use for their needs (
Additive manufacturing exhibits great potentials for the fabrication of novel materials due to its unique non-equilibrium solidification and heating process. In this work, a novel nano-oxides dispersion strengthened Co28Cr9W1.5Si (wt.%) alloy, fabricated by laser powder bed fusion (LPBF), was comprehensively investigated. During the layer-by-layer featured process, in-situ formation of Si rich, amorphous, nano-oxide inclusions was observed, whose formation is ascribed to the high affinity of Si to oxygen. Meanwhile, distinctive body-centered cubic (BCC) Co5Cr3Si2 nano-precipitates with an 8-fold symmetry were also confirmed to appear. The precipitates, rarely reported in previous studied Co-Cr alloys, were found to tightly bond with the in-situ oxidization. Furthermore, the morphologies, the size distributions as well as the microstructure of the interface between the matrix and the inclusions were investigated in detail and their influence on the tensile deformation was also clarified. The existence of transition boundaries between these inclusions and the matrix strongly blocked the movement of dislocations, thereby increasing the strength of the alloy. It was understood that when the plastic deformation proceeds, the fracture occurs in the vicinity of the oxide inclusions where dislocations accumulate. A quantitative analysis of the strengthening mechanism was also established, in which an additional important contribution to strength (∼ 169 MPa) caused by the effects of in-situ formed oxide inclusions was calculated.
Full-text available
Chemical measurement points obtained with energy dispersive X-ray spectroscopy (EDS) in scanning transmission electron microscopy (STEM) were processed by a modified weighted interval rank sort (WIRS) method in order to study the solute redistribution during solidification of an additively manufactured 316L stainless steel. Scheil-Gulliver calculations give a good first approximation, although the extent of segregation is lower experimentally than theoretically. These discrepancies are believed to be due to the fast solidification rate involved. STEM-EDS data processed by the modified WIRS method is a strong and reliable technique for characterizing the segregation profile for additively manufactured metals, and the proposed method is shown to reduce the experimental uncertainty associated with the quantification of minor elements.
Full-text available
The initial oxide state of powder is essential to the robust additive manufacturing of metal components using powder bed fusion processes. However, the variation of the powder surface oxide composition as a function of the atomizing medium is not clear. This work summarizes a detailed surface characterization of three 316L powders, produced using water atomization (WA), vacuum melting inert gas atomization (VIGA), and nitrogen atomization (GA). X‐ray photoelectron spectroscopy (XPS) and scanning electron microscopy analyses were combined to characterize the surface state of the powders. The results showed that the surface oxides consisted of a thin (~4 nm) iron oxide (Fe2O3) layer with particulate oxide phases rich in Cr, Mn, and Si, with a varying composition. XPS analysis combined with depth‐profiling showed that the VIGA powder had the lowest surface coverage of particulate compounds, followed by the GA powder, whereas the WA powder had the largest fraction of particulate surface oxides. The composition of the oxides was evaluated based on the XPS analysis of the oxide standards. Effects of Ar sputtering on the peak positions of the oxide standards were evaluated with the aim of providing an accurate analysis of the oxide characteristics at different etch depths.
Full-text available
The oxidation behavior of 316L stainless steel exposed at 400, 600 and 800 °C air for 100, 500 and 1000 h was investigated using different characterization techniques. Weight gain obeys a parabolic law, but the degree of deviation of n index is increasingly larger with the increase of temperature. A double oxide film, including Cr 2 O 3 and Fe 2 O 3 oxide particles in outer and FeCr 2 O 4 oxides in inner, is observed at 400 °C. As regards to samples at 600 °C, a critical exposure period around 100 h exists in the oxidation process, at which a compact oxide film decorated with oxide particles transforms to a loose oxide layer with a pore-structure. In addition, an oxide film containing Fe-rich outer oxide layer and Cr-rich inner oxide layer is observed at 600 °C for 500 and 1000 h. Spallation of oxide scale is observed for all samples at 800 °C regardless of exposure periods, resulting in different oxidation morphologies, and the degree of spallation behavior is getting worse. A double oxide film with the same chemical composition as 600 °C is observed, and the thickness increases over exposure periods.
Full-text available
Deformations within a microstructural gradient zone of stainless steel repaired specimens are investigated. The repair, added material by Directed Energy Deposition over a hot rolled sheet substrate, was tested in monotonic tensile experiments. In situ tests, scanning electron microscope images combined with high resolution digital image correlation and electron backscatter diffraction maps, permitted to monitor the local strain distribution. The strain distribution is homogeneous in the substrate and exhibits a heterogeneous pattern in the printed part with localization correlating spatially with the position of interlayers. The vicinity of the interface has smaller strains and exhibits larger hardness.
Full-text available
In this contribution, a simplified macroscopic and semi-analytical thermal analysis of directed energy deposition (DED) is presented to obtain computationally efficient simulations of the entire process. Solidification and solid-state phase transitions are taken into account. The model is derived for laser metal powder directed energy deposition, although it can be simply adapted for other focused thermal energy (e.g., electron beam, or plasma arc). The gas flow used for carrying the powder significantly influences cooling conditions, which is included in the model. The proposed simulation strategy applies to multilayer composites with a wide range of shapes in the horizontal plane and arbitrary laser scanning strategies (continuous way, back and forth, etc.). The proposed work provides a simple tool to study the influence of most process parameters, design in-situ experiments and in turn develop optimization loops to reach material requirements and specific microstructures. In-situ pyrometer measurements have been compared to the model, and good agreement has been observed with 2.6% error in average. The model is used to demonstrate the effect of various process parameters for a simple cylindrical geometry and a more complex auxetic cell.
Full-text available
We report that 316L austenitic stainless steel fabricated by direct laser deposition (DLD), an additive manufacturing (AM) process, have a higher yield strength than that of conventional 316L while keeping high ductility. More interestingly, no clear anisotropy in tensile properties was observed between the building and the scanning direction of the 3D printed steel. Metallographic examination of the as-built parts shows a heterogeneous solidification cellular microstructure. Transmission electron microscopy observations coupled with Energy Dispersive X-ray Spectrometry (EDS) reveal the presence of chemical micro-segregation correlated with high dislocation density at cell boundaries as well as the in-situ formation of well-dispersed oxides and transition-metal-rich precipitates. The hierarchical heterogeneous microstructure in the AM parts induces excellent strength of the 316L stainless steel while the low staking fault energy of the as-built 316L promotes the occurrence of abundant deformation twinning, in the origin of the high ductility of the AM steel. Without additional post-process treatments, the AM 316L proves that it can be used as a structural material or component for repair in mechanical construction.
Full-text available
Manganese sulphide (MnS) is one of the major non-metallic inclusions in steel with huge impact on steel property. In the case of high carbon steel, due to higher sulphur content and its brittleness, controlling MnS formation is one of the main issues. MnS has a complicated precipitation mechanism during solidification in liquid and solid steel and at the interface with oxide inclusions. Higher sulphur content, lower melting point and different oxide inclusions in high carbon steel will cause MnS precipitation at different stages. In this study, different stages of MnS precipitation from liquid and/or solid in high carbon steel and at the interface with oxide inclusion were investigated comprehensively via two different types of High Temperature Confocal Scanning Laser Microscope (HTCSLM). Samples were analysed further using SEM-EDS for better understanding the pertaining mechanisms. MnS precipitation on the surface of liquid steel was observed in situ in a HTCSLM by the use of a concentric solidification technique. Additionally, formation of MnS following solidification and at the interfaces of oxide inclusions, was investigated in situ in a HTCSLM, which has a uniform temperature profile across the specimen. These comprehensive descriptions about different stages of MnS precipitation in high carbon steel have been conducted for the first time and provide crucial information for controlling MnS morphology in high carbon steel.
Nano-bainitic steels represent a new class of alloys, whose microstructure consists of nanostructured bainitic ferrite formed at low temperature with a high amount of retained austenite leading to a high ductility and high tensile strength of the steel. Formation of nano-bainite has been studied thoroughly in literature as well as tempering of nano-bainitic steels. More recently it has been shown that adding carbide forming elements such as V and Mo increases the resistance to softening and hence the mechanical properties of nano-bainite at moderate temperature. Investigating the secondary carbide precipitation inside a nano-bainitic microstructure is thus necessary to optimize the thermal treatments for this promising new class of steels. Three initial microstructures of the same steel composition are investigated: martensite, martensite + retained austenite and nano-bainite. Studying the more conventional case of martensite has served as a basis to better understand the microstructure evolutions inside the nano-bainitic steel. The microstructural evolutions during the tempering were followed by complementary experimental techniques including dilatometry, in situ high energy synchrotron X-ray diffraction (HEXRD), conventional and high resolution TEM. The sequence of carbides precipitation and dissolution (transition-iron-carbides, cementite, and alloyed carbides) both during heating and holding is shown similar for the three initial microstructures. The kinetics are similar as well as cementite chemical composition and size distributions of cementite and alloyed carbides. It has been shown too that the three microstructures present a high retained austenite stability. Moreover, the analyses of the lattice parameters evolutions all along the tempering treatment associated with carbon mass balances have allowed to better understand the carbon distributions between carbides and matrix phases. The nucleation and growth model from a previous work was upgraded to take into account secondary precipitation and different new features (e.g. para-equilibrium interface condition for first stage of cementite growth, etc.). This model predicts the kinetics of precipitation, the particle densities and size distributions as well as matrix and carbides mean composition for different tempering conditions. Apart from the comparison with the experimental results that is discussed, it allowed to interpret the similar tempering behaviour for the three initial microstructures.
The origins of nano-scale oxide inclusions in 316 L austenitic stainless steel (SS) manufactured by laser powder bed fusion (L-PBF) was investigated by quantifying the possible intrusion pathways of oxygen contained in the precursor powder, extraneous oxygen from the process environment during laser processing, and moisture contamination during powder handling and storage. When processing the fresh, as-received powder in a well-controlled environment, the oxide inclusions contained in the precursor powder were the primary contributors to the formation of nano-scale oxides in the final additive manufactured (AM) product. These oxide inclusions were found to be enriched with oxygen getter elements like Si and Mn. By controlling the extraneous oxygen level in the process environment, the oxygen level in AM produced parts was found to increase with the extraneous oxygen level. The intrusion pathway of this extra oxygen was found to be dominated by the incorporation of spatter particles into the build during processing. Moisture induced oxidation during powder storage was also found to result in a higher oxide density in the AM produced parts. SS 316 L powder free of Si and Mn oxygen getters was processed in a well-controlled environment and resulted in a similar level of oxygen intrusion. Microhardness testing indicated that the oxide volume fraction increase from extraneous oxygen did not influence hardness values. However, a marked decrease in hardness was found for the humidified and Si-Mn free AM processed parts.
In additive manufacturing, the process parameters have a direct impact on the microstructure of the material and consequently on the mechanical properties of the manufactured parts. The purpose of this paper is to explore this relation by characterizing the local microstructural response via in situ tensile test under a scanning electron microscope (SEM) combined with high resolution digital image correlation (HR-DIC) and Electron Backscatter Diffraction (EBSD) maps. The specimens under scrutiny were extracted from bidirectionally-printed single-track thickness 316L stainless steel walls built by directed energy deposition. The morphologic and crystallographic textures of the grains were characterized by statistical analysis and associated with the particular heat flow pattern of the process. Grains were sorted according to their size into large columnar grains located within the printed layer and small equiaxed grains located at the interfaces between successive layers. In situ tensile experiments were performed with a loading direction either perpendicular or along the printing direction and exhibit different mechanisms of deformation. A statistical analysis of the average deformation per grain indicates that for a tensile loading along the building direction, small grains deform less than the large ones. In addition, HR-DIC combined with EBSD maps showed strain localization situated at the interface between layers in the absence of small grains either individual or in clusters. For tensile loads along the printing direction, the strain localization was present in several particular large grains. These observations permit to justify the differences in yield and ultimate strength noticed during macroscopic tensile tests for both configurations. Moreover, they indicate that an optimization of the process parameters could trigger the control of microstructure and consequently the macroscopic mechanical behavior.