The International Journal of Advanced Manufacturing Technology (2021) 115:2449–2464

https://doi.org/10.1007/s00170-021-07043-0

ORIGINAL ARTICLE

Optimization of metallic powder filaments for additive

manufacturing extrusion (MEX)
Fábio Cerejo 1,2 & Daniel Gatões 2 & M. T. Vieira 2

Received: 26 January 2021 / Accepted: 1 April 2021 / Published online: 25 May 2021
# The Author(s) 2021

Abstract
Additive manufacturing (AM) of metallic powder particles has been establishing itself as sustainable, whatever the technology
selected. Material extrusion (MEX) integrates the ongoing effort to improve AM sustainability, in which low-cost equipment is
associated with a decrease of powder waste during manufacturing. MEX has been gaining increasing interest for building 3D
functional/structural metallic parts because it incorporates the consolidated knowledge from powder injection moulding/
extrusion feedstocks into the AM scope—filament extrusion layer-by-layer. Moreover, MEX as an indirect process can over-
come some of the technical limitations of direct AM processes (laser/electron-beam-based) regarding energy-matter interactions.
The present study reveals an optimal methodology to produce MEX filament feedstocks (metallic powder, binder, and additives),
having in mind to attain the highest metallic powder content. Nevertheless, the main challenges are also to achieve high
extrudability and a suitable ratio between stiffness and flexibility. The metallic powder volume content (vol.%) in the feedstocks
was evaluated by the critical powder volume concentration (CPVC). Subsequently, the rheology of the feedstocks was
established by means of the mixing torque value, which is related to the filament extrudability performance.

Keywords MEX . Filament . Additive manufacturing . Mixing torque . Austenitic stainless steel (316L)

1 Introduction impossible to obtain using traditional manufacturing technol-

ogies is the largest benefit of AM. This manufacturing para-
Additive manufacturing (AM) of powder metals and metal digm, concerning powder metal AM, has attracted significant
alloys is an unavoidable area for Industry 4.0 owing to its interest over the past few years, where AM direct methods
potential to address some of the most significant industrial (e.g. Selective Laser Melting, Electron Beam Melting, etc.)
challenges in the twenty-first century concerning parts/sys- established themselves as technologies for functional/
tem/devices processing [1]. The rising trend to select AM structural metallic parts, with several components approved
processes is based on new design approaches, the ability to by ISO and ASTM standards in industries such as aerospace
create near net shape 3D objects, cloud access to manufactur- and automotive [3]. However, these processes have several
ing, shorter time-to-market, product customization, and circu- drawbacks, the high-power source that can promote micro-
lar economy [2]. Among other factors, the possibility of merg- and macroscopic defects in metallic parts due to the multiple
ing cost savings and new part properties and features that are thermal treatments of the deposited layers; material range, due
to the energy-matter interaction that limits the powder charac-
teristics, like reflectivity and conductivity; and significant
* Fábio Cerejo
[email protected] powder wastes during manufacturing and handling hazard.
Besides, the high cost of direct AM equipment is one of the
* Daniel Gatões
[email protected] major obstacles [4–10].
Material extrusion (MEX) and binder jetting (BJ) [11] are
1
already well-established technologies with market acceptance
IPN - LED&MAT - Instituto Pedro Nunes, Laboratório de Ensaios,
Desgaste e Materiais, Rua Pedro Nunes, 3030-199 Coimbra, Portugal
for the AM. Based on this successful background, MEX and BJ
2
have been investigated with the aim to produce metallic and
University of Coimbra, CEMMPRE - Centre for Mechanical
Engineering, Materials and Processes, Department of Mechanical
ceramic functional/structural components through shaping,
Engineering, Rua Luís Reis Santos, 3030-788 Coimbra, Portugal debinding, and sintering (SDS). In these indirect AM processes,
2450 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 1 MEX manufacturing route through the shaping, debinding, and sintering (SDS) process

unlike direct methods, the material processability is indepen- technologies do not have high pressures that promote the
dent of the power source, which makes the production of func- highest final part densification [12, 14].
tionally graded materials feasible [12]. Regarding shaping, BJ The present study is focused on the MEX technology,
and MEX differ from each other and are often complementary which was initially referred to as the fused deposition of
technologies, since resolution, the optimal number of parts per metals (FDMet), and then as Fused Filament Fabrication
production, the necessity of supports, material, and equipment (FFF) or as Metallic Fused Filament Fabrication (MF3) [15].
costs are unique for each process. In BJ, the binder droplets are MEX is based on the fused deposition modelling (FDMTM)
selectively deposited to interact with powder particles, which technology commercialized by Stratasys Inc. for polymers
present new challenges not associated with MEX, such as and waxes, where the filament is composed of a mixture of a
powder/binder wettability; binder vertical migration, since layer high volume content (vol.%) of metallic powders with organic
height affects the penetration depth of the binder, through time constituents [16]. MEX is suitable for manufacturing geomet-
and capillary force; and binder saturation that must be fine- rical complex metal parts in conjunction with post-shaping
tuned, as well as droplet size and dispensing frequency. steps, such as debinding and sintering [17].
However, the metallic powder is fed independently of the bind- The processing by MEX technology consists of 5 stages
er, which makes the rheological properties not as hard to control (Fig. 1), as follows: Stage 1, materials selection; Stage 2,
as it happens with MEX, which is beneficial to achieve the mixing (1+2=powder material extrusion feedstocks fabrica-
highest vol.% of metallic powder and less binder content in tion); Stage 3, feedstock extrusion (filament); Stage 4, 3D
the shaped 3D object [12, 13]. Even though both technologies part/device built layer-by-layer (green) from extruded fila-
differ in what concerns the powder-binder processability, they ment; and Stage 5, binder removal (debinding) and subse-
also face the same challenges, such as optimizing debinding quent consolidation of metallic powder particles (sintering).
heating profiles to degrade the polymer is time-consuming; The first two stages are fundamental to attain an appropriate
possible undesired reactions from residual polymer ash that viscosity and an excellent balance between flexibility and
could affect the final properties; and porosity, since both stiffness of the filament associated with the highest content

Fig. 2 a Particle size distribution (Malvern Mastersizer 2000) and b powder shape (SEM analysis, SE) of the SS 316L powder
Int J Adv Manuf Technol (2021) 115:2449–2464 2451

(%vol) of metallic powders and a subsequent homogeneity 2 Materials and methods

(chemical and dimensional) needed to attain low porosity
and excellent quality of the parts/devices. 2.1 Characterization techniques
High-quality structural/functional parts/systems/devices
through an SDS process must include efforts to achieve the The characterizations of the powders, feedstocks, and fila-
highest content (vol.%) of the metallic powder particles ments were performed through the following techniques: laser
possible in the feedstock within the steady-state regime diffraction to measure the particle size (Malvern Mastersizer
but always taking into account the final mixing torque val- 2000, Malvern Instruments Ltd, Worcestershire, UK) accord-
ue. This becomes a significant challenge in MEX because ing to ISO 13320:2009(E); helium pycnometry to measure the
the mixture must be manufactured in a filament form. Based density, based on the mean of five runs for each specimen
on these assumptions, there are a few available studies in (Micromeritics AccuPyc 1330, Micromeritics Instrument
the bibliography. Agarwala et al. obtained stiff and straight Corporation, Norcross, GA, USA); scanning electron micros-
filaments from feedstocks with 60 vol.% 17-4PH grade copy (SEM) to analyse the morphology and shape factor (FEI
stainless steel (SS) [18]. Kukla and Gutierrez et al. [19, Quanta 400FEG, FEI Europe BV, Eindhoven,
20] used 55 vol.% SS 316L and 17-4PH; Godec et al. [10] The Netherlands) x-ray diffraction (XRD) according to EN
also selected 55 vol.% SS 316L, as well as Burkhardt et al. 13925:2003, to identify the phases (Philips X'Pert, cobalt ra-
[21]. Anderson et al. [22] produced SS 316L grade fila- diation λkα1= 0.1789 nm and λkα2= 0.1793 nm, Bragg-
ments with 55 vol.%; Kurose et al. [23] used SS 316L grade Brentano geometry, Philips, Eindhoven, Netherlands); and
filaments with 60 vol.%. Gloeckle et al. [24] performed an TGA to analyse the weight variation kinetics (TGA Q500
extensive study on the printability of Ti-6Al-4V filaments V20.13, TA Instruments, DE, USA). The filament was
with up to 60 vol.% of inorganic material, and Singh et al. micro-CT (X-ray micro computed tomography) scanned
[15] used the same Ti-6Al-4V with 59 vol.%. BASF SE® using a Bruker SkyScan 1275 (Bruker, Kontich, Belgium).
has a commercially available SS 316L filament with a me- An acceleration voltage of 80 kV and a beam current of
tallic powder content lower than 60 vol.% [25]. 125 μA was set while using a 1-mm aluminium filter and
The mechanical performance of MEX metallic filaments step-and-shoot mode. Pixel size was set to 6 μm, and random
has been a major limitation for increasing the ratio between mode was used. In total, 1056 projection images were ac-
inorganic and organic constituents since high powder concen- quired at 0.2° angular step with 5 frames average per step
tration can lead to poor extrudability, where the filament be- using an exposure time of 46 ms. The micro-CT images were
comes too brittle to be handled. A balance between stiffness reconstructed with the dedicated manufacturer software.
and flexibility must be guaranteed to promote filament print- A three-point flexural test of each filament was performed
ability [15, 19]. five times using SHIMADZU-EZ-LX (Shimadzu Corp.,
The focus of the present study is to develop filaments with Kyoto, Japan) equipment with a load capacity of 500 N. The
the highest metallic powder volume content (>50 vol.%) that load was applied to the specimen at a rate of 0.5 mm/min with
link the primary MEX filament requirements to a suitable a span length of 20 mm.
viscosity and mechanical behaviour. The selection of the 3D prototype geometries were measured with Focus
highest content of metallic powder was evaluated by critical Variation Microscopy (FVM) using Alicona InfiniteFocusG4
powder volume concentration (CPVC) methodology [26–28]. (Alicona Imaging; Graz, Austria). Hardness measurements, ac-
This procedure aims to promote the highest part green density, cording to ISO 6507-2:2018, were performed with a Shimadzu
which is essential for maintaining the part shape integrity after
debinding and sintering.
The present study aims to contribute, whatever the powder
selected, to high-quality filaments for MEX technology, that is
in conjunction with binder jetting technology, the future of
AM of functional/structural 3D metallic objects.

Table 1 SS 316L powder characteristics

d10 [μm] d50 [μm] d90 [μm] SSA* [kg/m2] ρ [kg/m3] **

4.63 9.43 16.60 786 7896 ± 30.2

*Specific surface area

**Density Fig. 3 SS 316L XRD diffractogram; λ Co = 0.1789 nm (Philips X'Pert)
2452 Int J Adv Manuf Technol (2021) 115:2449–2464

Table 2 Densities of the binder

and additives measured with a M1 TPE ULD-PE S† PӾ
helium pycnometer
(Micromeritics AccuPyc 1330) Density [kg/m3] 970 ± 1 1028 ± 12.2 9144 ± 1.5 983 ± 1 965 ± 0.6

S† Surfactant
PӾ Plasticizer

HMV microhardness tester (Shimadzu Corp., Kyoto, Japan). lower stress in the particles than those from water atomizing,
For each measurement, a load of 9.8 N was applied for 15 s where martensite is more prevalent for a similar powder [30].
by a Vickers indenter. In the present study, the organic constituents of the feed-
stocks were divided into two primary groups: master binder
2.2 Metallic powder, binder, and additives and additives; the last one includes the backbone and surfac-
tant/plasticiser. The selected master binder (M1) was a
The set material for this study was austenitic stainless steel commercial-grade (Atect Corp., Shiga, Japan) that is a mixture
316L (SS 316L) since it is one of the most studied materials in of polyolefin waxes and > 60 wt. % of polyoxymethylene
AM. This material can be a good standard for the methodol- (POM). Although POM is included, this binder is commer-
ogy to be established by the present work and extrapolated to cialized as a thermal-only debinding and previous studies with
other metallic alloys [29]. this binder shown that no carbonaceous residues remained on
The characteristics of the selected powders were studied the final parts. The thermoplastic elastomer (TPE), as well as
using the 4S’s methodology (size and size distribution, shape, an ultra-low density polyethylene (ULD-PE), was used for the
and structure). The particle size is d50 = 9.43 μm. Figure 2 backbone, and a surfactant (stearic acid (S †)) and an external
shows the particle size distribution (a) and particle shape fac- plasticizer (P Ӿ) were used as additives. Density of master
tor (b). Shape factor was close to 1 for the nitrogen-atomized binder and additives (Table 2) was performed to support the
SS 316L powders (Sandvik Osprey Ltd., Neath, UK). theoretical calculations of the volume of these constituents in
Table 1 summarizes the powder characteristics. the feedstock, which is further compared with the practical
The X-ray diffractogram (Fig. 3) of the SS 316L powder results (TGA analysis).
exhibited a biphasic character, where the major phase was
austenite (ICDD 33-0397). However, other peaks with low 2.3 Filament production
I/I0 were present, which can be indexed as (100), (200) that
are typical of ferrite/martensite (ICDD 87-0722). The cooling The vol.% of each organic component in the feedstock was
stress could contribute to the evolution of the austenite phase tailored to achieve the proper filament properties for MEX
into martensite. Nevertheless, nitrogen atomizing results in (rheology and flexibility/stiffness balance). The CPVC and

Table 3 Filament feedstocks

Feedstock Master binder Additives Powder

Backbone Backbone content (vol.%)* S† or PӾ S† or PӾ (vol.%) SS 316 L content (vol.%) Particle size d50 (μm)

F01 M1 - - - - 60 6.85
F02 M1 TPE Y - - 60 6.85
F03 M1 TPE X S† 5 60 6.85
F04 M1 TPE X+5 S† 5 60 6.85
F05 M1 TPE X + 10 S† 5 60 6.85
F06 M1 TPE X PӾ 5 60 6.85
F07 M1 TPE W PӾ 10 60 6.85
F08 M1 ULD-PE X PӾ 5 60 6.85

*The exact ratio of master binder/backbone in the feedstock will be kept confidential. Y, X, and W represent the different amounts of backbone (vol.%).
X = Y – 2.5 vol.%; (½ of S/P content: 5 vol.%).
W = Y – 5 vol.%; (½ of F07 plasticiser (P) content: 10 vol.%).
S† , Surfactant
PӾ , Plasticiser
Int J Adv Manuf Technol (2021) 115:2449–2464 2453

Fig. 4 TGA curves of the SS

316L, M1, TPE, and M1 + TPE
under an N2 atmosphere

feedstock optimization was performed with a torque rheome- measured at multiple points to guarantee its dimensional ac-
ter (Plastograph® Brabender W 50, Brabender GmbH & Co. curacy throughout the filament fabrication process.
KG, Duisburg, Germany) that evaluates the torque variation Instead of using catalytic debinding, thermal debinding
as a function of the powder composition. The temperature was selected, which shows no carbonaceous residues present
inside the 38.5 cm3 mixing chamber of 180 °C was selected during sintering, consistent with previous studies with low-
together with 30 rpm blades rotation speed, taking into ac- pressure injection moulding [31], thus voiding the nitric acid
count previous work with the same master binder [31]. For used for catalytic debinding. The thermogravimetric analysis
the CPVC evaluation, an increment of 1 vol.% powder content highlights that all of the organic constituents of the feedstock
was made approximately every 10 min or after reaching a fully degrade, during the debinding stage (Fig. 4).
steady state. The backbone percentage was not higher than The thermal cycles in this work were selected based on
30% of the total organic portion. Table 3 summarizes the thermogravimetric analysis (TGA). The primary events in
different feedstock compositions (F01–F08, F03A, and the master binder and backbone weight loss curves up to
F03B). Each formulation mixing torque value was evaluated 600 °C (Fig. 4 and Table 4) were the isothermal plateaus
from six independent measurements and done until a steady- during the debinding stage. The beginning and ending values
state regime was achieved. were evaluated from the first derivative (DTG) of the respec-
Small pellets of feedstocks were extruded into a filament tive curve. At 495 °C, the carbonaceous residue was close to 0
form. Filament fabrication was performed in a single screw wt.%.
extruder (Brabender GmbH & Co., Duisburg, Germany) with Concerning the thermal oxidation of the as-received SS
5 heating zones. The temperature of the zones, from feeder to 316L powder, TGA showed that it was quite stable up to
nozzle (ø 1.75 mm), was set at 160, 165, 170, 175, and 180 °C. 600 °C in an N2 atmosphere. An insignificant increase in the
The screw rotation speed was set at 5 rpm. The filament was weight of the powder was noticeable above 500 °C. This is not

Table 4 Weight loss and

degradation temperatures of the Binder component Degradation stage Weight loss [%] Onset [°C] End [°C]
M1, TPE, and M1 + TPE
M1 1st 41 232 314
2nd 13 378 437
3rd 46 437 472
TPE 1st 2 300 327
2nd 98 408 454
M1 + TPE 1st 35 238 320
2nd 11 375 398
3rd 23 427 441
4th 31 450 475
Plasticizer 1st 100 242 276
2454 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 5 TGA curves of the

filaments F03, F06, and F08 with
60 vol.% of SS 316L powder
under an N2 atmosphere

exclusively attributed to the TGA protective atmosphere type 210°C, extrusion speed was limited to 10 mm/s, and extrusion
(N2) because other studies show the same behaviour under an multiplier was set at 1.4. Layer height was set at 0.20 mm, and
Ar + H2(5%) atmosphere [30, 32]. the selected nozzle diameter was 0.40 mm. A glass platform
A significant difference was not detected in the debinding with a layer of glue was used, to promote part adhesion, since
kinetics of the M1 and TPE as raw materials (Fig. 4) and as this 3D printer has no heated build plate feature.
feedstocks constituents (Fig. 5) when mixed with 60 vol.% of
metallic powder. This is an indication that there are no unde- 2.5 Debinding and sintering
sired reactions among the feedstock constituents that could
interfere in the debinding cycle. The heating rates for debinding and sintering were 1 °C/min
Table 5 shows a comparison of the theoretical values and 10 °C/min up to the maximum temperatures of 600 °C
against the final experimental values (wt.%) of the feedstock and 1250 °C, respectively, including several debinding iso-
filaments at 600 °C in order to illustrate the expected SS 316L thermal holding times, based on the TGA results. The fila-
weight (%) after binder degradation based on the SS 316L ments and parts were debinded and sintered under H2 atmo-
vol.% in the feedstock. The small deviation between theoret- sphere at 4×10−2 MPa.
ical and experimental values can be attributed to the experi-
mental evaluation of the densities and can be assumed that the
binder degradation was total.
3 Results and discussion

2.4 Printability 3.1 Feedstock optimization

Green specimens were built on a BQ Prusa I3 Hephestos 3D 3.1.1 Evaluation of the critical powder volume concentration
printer. Extruder nozzle temperature was maintained at
The evaluation of the CPVC in each feedstock was performed
by recording the mixing torque to maximize the metal powder
Table 5 TGA experimental vs theoretical weight reduction of filaments content (vol.%), to promote higher green densities.
F03, F06, and F08
Nevertheless, MEX feedstocks require overcoming new chal-
Feedstock Metallic powder [wt.%] lenges (rheology and flexibility/stiffness balance) because the
filament must be spooled, handled, and extruded through a
Experimental Theoretical* small-diameter nozzle. Therefore, PIM feedstocks must be
F03 92.5±0.1 92.4 modified to be suitable for MEX, and the rheological behav-
F06 92.5±0.1 92.4
iour of the new feedstock must be studied.
F08 92.4±0.1 92.4
Torque values were recorded for the initial mixture of
the master binder and additives F06 (M1 + TPE + P)
*Conversion from SS 316L 60 vol.% with 50 vol.% SS 316L powder. The effect of
Int J Adv Manuf Technol (2021) 115:2449–2464 2455

Fig. 6 Torque as a function of

time of the F06 feedstock at 180
°C with the incremental addition
of 1 vol.% of SS 316L

subsequent additions of metallic powder (1 vol.%) on The CPVC should correspond to the interface torque value
the torque value, shown at every peak from the 10- between the second and third regimes (point of intersection),
min mark, was measured after attaining a steady-state which in the present study was 62 vol.%. However, a great
regime for each percentage (Fig. 6). Based on this eval- number of studies performed concerning the optimization of
uation, the defined ratio of inorganic/organic vol.% powder and binder feedstocks show that the torque value
among all studied feedstocks was maintained. should not exceed 5 N.m for this specific torque rheometer
Figure 7 shows the torque values for the incremental addi- (Plastograph® W 50), to attain the best rheological properties,
tions of 1 vol.% of SS 316L powder (50–65 vol.%). Three in order to guarantee optimal processability [28, 33, 34].
linear regimes can be observed: Based on the CPVC evaluations (Figs. 6 and 7) and keep-
ing in mind that the feedstock flowability to build the green
– The first regime includes up to 58 vol.% of SS 316L part, through a 3D printer with a 0.4-mm nozzle, is promoted
powder, and the torque variation between each addition by the filament (it acts like a piston through the 3D printer pull
is between 1.9 (50 vol.%) and 3.5 N∙m (58 vol.%). system) and not by a screw (high pressures), 60 vol.% (torque
– In the second regime, at up to 62 vol.% of SS 316L value of 4.3 N.m) was selected as the metallic load for all
powder, the mixture rheology changes, which corre- studied feedstocks.
sponds to a slightly higher slope than the first regime,
and the maximum torque is 5.4 N∙m.
– The third regime, where the most significant variation 3.1.2 Effect of additives in the feedstock
occurs (62–65 vol.%), reaches the highest mixing torque
value (8.1 N∙m). In this range, the mixing torque behav- In PIM, the binder generally promotes the best compromise
iour clearly becomes more unstable (cf. Fig. 7). between green integrity and flowability.

Fig. 7 Three different regimes of

torque as a function of powder
volume concentration (feedstock
F06)
2456 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 8 Mixing torques (F01-F05)

as a function of time (Y < X
vol.%.); Stage 1, only binder ad-
dition; Stage 2, feedstock with
additive addition. F01 is the stan-
dard feedstock without additives

However, as previously mentioned, flexibility is one of the approaches were considered: the addition of TPE + surfactant
major characteristics of MEX filaments. For this reason, the (S) and the addition of TPE + plasticizer (P).
backbone, surfactant, and plasticizer content were optimized. Figure 8 shows the impact of the additive composition on
To select the best feedstock composition, two different the final torque value at the end of 30 min. F01 is a feedstock

Fig. 9 F01 and F06–F08 mixing

torques as a function of time (W <
X vol.%); Stage 1, only binder
addition; Stage 2, feedstock with
additive addition. F01 is the stan-
dard feedstock without additives
Int J Adv Manuf Technol (2021) 115:2449–2464 2457

Fig. 10 Torque of the F03A and

F03 mixtures as a function of
time; Stage 1, only binder
addition; Stage 2, feedstock with
additive addition

used in PIM, where there are no concerns about flexibility. In The feedstock F08, which had the composition as F06 ex-
addition to promoting filament flexibility, TPE has a negative cept the TPE was replaced by ULD-PE, had a lower torque
impact on the rheological behaviour of the feedstock. To over- value, but it was still higher than the feedstock with SA (F03).
come this issue, a surfactant (S) was added. Among the feed- Based on torque values of the feedstocks, F06–F08 seem to be
stocks in Fig. 8, F03 (blue curve) showed the most promising suitable for use in MEX.
behaviour (close to 4 N.m).
Although the selected surfactant, stearic acid (SA), clearly
reduced the feedstock torque, other mixtures were studied to 3.1.3 Influence of metal powder particle size
possibly replace it owing to the difficulty of fully removing it
during debinding. Other work reported that SA requires rais- To evaluate the influence of the SS 316L powder par-
ing the debinding temperature from 600 to 700 °C [30]. Thus, ticle size on the rheological behaviour of the feedstock,
SA was replaced with a plasticizer (P), which also acts as a two mixtures with the same vol.% of SS 316 and addi-
rheological modifier with the advantage that it promotes fila- tives but different particle sizes were compared: F03A
ment flexibility. and F03 with d50 = 3.76 μm and d50 = 6.85 μm, re-
Figure 9 shows the different torque values for the remain- spectively (Fig. 10). The finer particle size led to a
ing studied feedstocks (F06–F08). F06 had a final torque (4.6 higher final mixing torque compared with that of F03.
N∙m) that was higher than that of F07 (3.6 N∙m) due to its high For this reason, powder with d50 = 3.76 μm was not
TPE content. Comparing the feedstocks with the same vol.% selected for the studied feedstocks. This is an expected
of all constituents (F03, F06, and F08), the addition of the behaviour as small particles have a high specific surface
surfactant (F03) had a larger influence on the final torque area of contact, which promotes high interparticle fric-
value than a plasticizer (F06 and F08), as expected. tion [35].
However, the plasticizer boosts filament flexibility, which is Table 6 summarizes the average of ten torque values for
a very important requirement. each of the different feedstocks.

Table 6 Torque values of different feedstocks

F01 F02 F03 F04 F05 F06 F07 F08 F03A

Torque [N.m] 2.0±0.04 5.5±0.07 3.7±0.03 4.3±0.04 5.3±0.06 4.5±0.07 3.6±0.05 4.2±0.07 4.6±0.07
Backbone content - Y X X+5 X+10 X W X X
S† or PӾ (vol.%) - - 5 5 5 5 10 5 5
2458 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 11 Spooled filament from

feedstock F06

3.2 Filaments action, a constant height between the extruder nozzle and the
table was preserved for all formulations.
3.2.1 Green filament production Standard filament (powder and binder, F01) was too brittle
to spool. TPE addition resulted in the highest final torque
In this study, the filament was not spooled by an automatic value (5.5 N.m) in mixture F02 curve. Also, the torque profile
system. To avoid filament diameter deviations due to gravity is the most unstable, because the dispersion of powder

Fig. 12 Cross-section fracture

surfaces of filaments F01, F03,
F06, and F08
Int J Adv Manuf Technol (2021) 115:2449–2464 2459

Fig. 13 Micro-CT of the a green filament F06 and b sintered filament F06 – green filament with the highest homogeneity

particles is more problematic in this binder. Thus, F02 was not results (cf. 3.2.2.3). The observed residual pores seem to fol-
extruded in a filament form. The other feedstocks (powder, low the extrusion direction (Fig. 13a, X-Z and Z-Y section),
binder, and additives) could be spooled (F03, F04, F05, and suggesting that the defects may occur in the extrusion process,
F08) but with a higher curvature than those of F06 and F07 resulting in elongated pore geometry. Nevertheless, the
(Fig. 11). It must be emphasized that the best filaments were sintered filament shows that a sintered part with consistent
those resulting from feedstocks with plasticizer and torque density, with no persistent porosity caused by debinding, is
close to 5.0 N∙m. still achievable, indicating that resulting porosity in final parts
may be connected to printing parameters.
3.2.2.2 Structure The X-ray diffractograms (Fig. 14) show the
3.2.2 Characterization
evolution from SS 316L powder to the green and sintered
filament. As referred, the austenitic powder, owing to its prep-
3.2.2.1 Homogeneity SEM observations of the filament cross-
aration technique, besides austenite (ICDD 33-0397), presents
section shown that filament feedstocks without TPE (F01 and
a ferrite/martensite phase (ICDD 87-0722—Fig. 3). This one
F08) appeared to be more homogeneous than the others (F03
increases, as expected, due to the deformation of powders
and F06, Fig. 12), but their flexibility was poor. F03 and F06
(= M1 + TPE + S/P vol.%) were quite similar regarding the
homogeneity of the distribution of metallic powder. A large
powder particle distance in the feedstocks improves flexibility
but decreases density. Thus, considering these two features
(filament flexibly and interparticle distance), a homogenous
powder distribution is crucial, and a suitable balance between
these features is required for the success of MEX.
Micro-CT analysis can be an effective solution for
assessing filament homogeneity without fracture, which can
modify the defect distribution on the observed volumes.
Micro-CT was performed in a representative cross-section of
filament F06 in the green and sintered states (Fig. 13 a and b,
respectively). It can be noted that the green filament (extruded
feedstock in filament form, not subject to any subsequent pro-
cessing step) presents a high density and consistent diameter
throughout its section, reiterating the mechanical behaviour Fig. 14 X-ray diffraction of the sintered filament F06
2460 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 15 Deflection at break (%)

and Eflex of filaments F01 and
F03–F08

during its extrusion (Fig. 14) [36]. However, after sintering, most promising filament feedstocks regarding green
the X-ray diffractogram shows only constituted by the austen- processability.
itic phase, avoiding a post heat treatment, required in other The filament from feedstock F01, without additives, was
additive processes. the most brittle (13%). With the incremental addition of TPE
3.2.2.3 Mechanical behaviour The deflection at rupture (%) to feedstocks F03–F05, the maximum deflection of the fila-
and flexural modulus of elasticity (Eflex) was measured by ments increased. Comparing filaments F03 and F06 (equal
three-point bending tests (Fig. 15). Each value in the figure vol.% of organic constituents), the replacement of the surfac-
is the average of five tests. These results, together with the tant with a plasticizer improved the flexibility. A reduction in
previous torque rheometry study, allowed for selecting the deflection from F06 (91%) to F07 (71%) was noticeable. F07

Fig. 16 Morphological analysis of the F06 green part: a, b, and c top views; d, e, and f side views
Int J Adv Manuf Technol (2021) 115:2449–2464 2461

Table 7 FVM part dimensions (before and after sintering) Table 8 FVM part infill evaluation (before and after sintering)

Height (mm) Diameter (mm) Measurement Distance (mm)

Green Final part Green Final part 1 2 3 4 5 6

Measurement 1 7.2 6.2 19.7 16.5 Green Direction 1 1.6 1.6 1.6 1.6 1.6 1.6
Measurement 2 7.2 6.2 19.7 16.5 Direction 2 1.6 1.6 1.6 1.6 1.5 1.7
Mean 7.2 6.2 19.7 16.5 Final Direction 1 1.3 1.3 1.3 1.3 1.3 1.3
Shrinkage (%) 13 16 Direction 2 1.3 1.3 1.3 1.3 1.2 1.4
Shrinkage (%) 19 18 19 19 19 19

had the highest additive content and, consequently, less TPE

in the feedstock. F08 had ULD-PE instead of TPE. The reduc- The filament with the best mechanical characteristics (F06)
tion in the F08 filament flexibility (44%) is more obvious was used to produce a part consisting of an infill pattern ex-
compared to that of F06, which had the same additive, a plas- trusion lines (0.4 mm) in a grid form and a two-perimeter
ticizer. In fact, filament from feedstock F08 was not used to exterior. Figure 16 shows different details of the green part
produce the green parts because it was brittle, despite its prom- (top and side views) built from the selected filament.
ising mixing torque (4.2 N∙m). Figure 16b shows that a single extrusion line is geometrically
The Eflex results of F07 and F08 were not expected, taking reliable (not considering the flat zones due to superposition of
into account the relationship between the maximum deflection lines), indicating that a filament with consistent diameter and
at break and the Eflex of the remaining filaments. This can be no relevant ovality was achieved, since this would otherwise
attributed to several factors, including porosity and the non- affect the extrusion behaviour. Figure 16d shows that the post-
homogeneous feedstock mixture. Similar behaviour was also deposited layer thickness was almost the same as the 3D de-
reported by another study [19]. sign (0.2 mm).
For the highest magnifications (Fig. 16e and f), it is obvi-
ous that a critical zone exists: interface layers. These bonding
3.3 Proof of concept zones could have a significant role in the occurrence of defects
in the final parts. For this reason, further detailed studies must
The printability of the spoolable filaments (F03-F08) was be performed to understand their influence on the properties of
studied. Filaments with the same surfactant (SA) but dif- the final metallic part.
ferent TPE content (F03 and F04) were successfully The green shrinkage and warpage were evaluated by FVM.
printed. F05 was not printable owing to its high torque The measurements of the external diameter and height were
value, which was promoted by the high amount of TPE performed in two opposite locations. From the results
(vol.%). F08 was too brittle to be printable, although (Table 7), a slight variation in the green dimensions was no-
within optimal mixture torque range. F06 and F07 shown ticeable when compared with the 3D model (ø 20 mm and a
the best printing behaviour. height of 7.0 mm).

Fig. 17 FVM part B infill

evaluation: a green and b sintered
part
2462 Int J Adv Manuf Technol (2021) 115:2449–2464

Fig. 18 Microhardness of the

proof of concept (PoC) part in
comparison with other data [30,
37, 38]

There was a slight variation in the shrinkage between production (metallic powder, binder, and additives), evalua-
the XY plane and the Z-axis. This deviation was also tion of extrudability (torque <4.0–4.5 N∙m, supported by other
reported in other studies and can be associated with the rheological studies), debinding, and sintering. In the filament
build strategy (orientation of the layers) [4, 17]. The feedstocks, the main challenges are to reach the highest me-
diameter shrinkage was, on average, 2.5% higher than tallic powder content with good extrudability and a suitable
in the build direction. Because the amount of shrinkage ratio between stiffness and flexibility. Some filaments, with-
is affected by processing parameters and feedstock char- out surfactant/plasticizer or with different additive contents,
acteristics, a detailed comparison must be avoided with- are unsuitable due to the difficulty to be extruded, owing to
out further investigation. their high mixing torque.
The geometry of the green infill (40%) and shrinkage was From the extruded feedstocks, the selected case study
also evaluated by FVM (Fig. 17). Green part infill was printed shows that, for the filament constituted by SS 316L + M1 +
with high accuracy because the interlayer distance (measure- TPE + P, the best green mechanical characteristics are attained
ment 1 to 6) was constant in both measured directions (0° and due to the excellent homogenization of the mixing, demon-
90°) (Fig. 17a and Table 8). After sintering, accuracy was not strated by micro-CT. Moreover, the best filament, after
affected (Fig. 18b and Table 8). debinding and sintering, continues to show excellent perfor-
In comparison with the previous FVM measurements, mance, concerning defects (porosity) and consequently, the
the part infill (width = nozzle ø, 0.4 mm) shrinkage was best flexural modulus, and deflection at break, assuring an
higher (~19%, Table 8) than in the other directions (13– excellent MEX processability. In the case of austenitic steel
17%, Table 7). The 40% infill configuration can largely powders, the austenite phase is the only phase present,
affect the shrinkage of the part relative to the XY direc- avoiding other costly post heat treatments. In conclusion, this
tions. This is similar to what occurs in PIM parts for study promotes a supported methodology for producing fila-
different thicknesses. ments for MEX and promotes the possibility to make 3D
The microhardness value of the part developed as a proof parts/systems or devices whatever the metallic powders select-
of concept (PoC, Fig. 18) was 1.1 GPa, which is higher than ed, without commercial market offer dependence.
those of other studies for the same material on MEX [37].
Nevertheless, the microhardness values are lower than those
measured in a PIM made from SS 316L and a bulk steel part Author contribution Fábio Cerejo: investigation, conceptualization,
methodology, visualization, and writing—original draft preparation.
[30, 38]. Daniel Gatões: Micro-CT investigation and methodology, writing, and
review. Teresa Vieira: supervision, conceptualization, and review and
editing.
4 Conclusions
Funding This work was supported by the European Regional
Development Fund (ERDF) under the Portuguese program, Programa
High-quality filaments for metallic MEX can be attained by Operacional Factores de Competitividade (COMPETE) [grant agreement
optimization of different manufacturing steps: feedstocks No. POCI-01-0247-FEDER-024533]; and this research is sponsored by
Int J Adv Manuf Technol (2021) 115:2449–2464 2463

FEDER funds through the program COMPETE—Programa Operacional 9. Ikeshoji T-T, Nakamura K, Yonehara M, Imai K, Kyogoku H
Factores de Competitividade—and by national funds through FCT, (2018) Selective laser melting of pure copper. JOM 70:396–400.
Fundação para a Ciência e a Tecnologia, under the project https://doi.org/10.1007/s11837-017-2695-x
UIDB/00285/2020. 10. Godec D, Cano S, Holzer C, Gonzalez-Gutierrez J (2020)
Optimization of the 3D printing parameters for tensile properties
Data availability Data will be available upon request. of specimens produced by fused filament fabrication of 17-4PH
Stainless Steel. Materials 13:774. https://doi.org/10.3390/
ma13030774
Declarations 11. ISO/ASTM Additive manufacturing - General principles - termi-
nology (ISO/ASTM DIS 52900:2018)
Ethical approval Not applicable. 12. Ziaee M, Crane NB (2019) Binder jetting: a review of process,
materials, and methods. Addit Manufact 28:781–801. https://doi.
Consent to participate Not applicable. org/10.1016/j.addma.2019.05.031
13. Bai Y, Williams CB (2018) Binder jetting additive manufacturing
with a particle-free metal ink as a binder precursor. Mater Des 11
Consent for publication Not applicable.
14. Bai Y, Wagner G, Williams CB (2017) Effect of particle size dis-
tribution on powder packing and sintering in binder jetting additive
Conflict of interest The authors declare no competing interests. manufacturing of metals. J Manuf Sci Eng 139:081019. https://doi.
org/10.1115/1.4036640
Open Access This article is licensed under a Creative Commons 15. Singh P, Balla VK, Tofangchi A, Atre SV, Kate KH (2020)
Attribution 4.0 International License, which permits use, sharing, adap- Printability studies of Ti-6Al-4V by metal fused filament fabrica-
tation, distribution and reproduction in any medium or format, as long as tion (MF3). Int J Refract Met Hard Mater 91:105249. https://doi.
you give appropriate credit to the original author(s) and the source, pro- org/10.1016/j.ijrmhm.2020.105249
vide a link to the Creative Commons licence, and indicate if changes were 16. Danforth SC, Agarwala M, Bandyopadghyay A et al (1998) Solid
made. The images or other third party material in this article are included freeform fabrication methods. 17
in the article's Creative Commons licence, unless indicated otherwise in a 17. Gonzalez-Gutierrez J, Cano S, Schuschnigg S, Kukla C, Sapkota J,
credit line to the material. If material is not included in the article's Holzer C (2018) Additive manufacturing of metallic and ceramic
Creative Commons licence and your intended use is not permitted by components by the material extrusion of highly-filled polymers: a
statutory regulation or exceeds the permitted use, you will need to obtain review and future perspectives. Materials 11:840. https://doi.org/
permission directly from the copyright holder. To view a copy of this 10.3390/ma11050840
licence, visit http://creativecommons.org/licenses/by/4.0/. 18. M K Agarwala, Weeren, R. van, Bandyopadhyay, A., et al (1996)
Fused deposition of ceramics and metals: an overview. In:
International Solid Freeform Fabrication Symposium. Texas,
USA, p 8
19. Kukla C, Duretek I, Schuschnigg S, Gonzalez-Gutierrez J (2016)
References Properties for PIM feedstocks used in fused filament fabrication. In:
Proceedings of the World PM2016 Congress & Exhibition.
1. Alcácer V, Cruz-Machado V (2019) Scanning the Industry 4.0: a Hamburg, Germany, 9–13 October 2016, p 5
literature review on technologies for manufacturing systems. Eng 20. Gonzalez-Gutierrez J, Godec D, Kukla C et al (2017) Shaping,
Sci Technol Int J 22:899. https://doi.org/10.1016/j.jestch.2019.01. debinding and sintering of steel components via fused filament
006 fabrication. Croatian Association of Production Engineering,
2. Dilberoglu UM, Gharehpapagh B, Yaman U, Dolen M (2017) The Zagreb, pp 99–104
role of additive manufacturing in the era of Industry 4.0. Proc 21. Burkhardt C, Freigassner P, Weber O et al (2016) Fused filament
Manufact 11:545. https://doi.org/10.1016/j.promfg.2017.07.148 fabrication (FFF) of 316L Green parts for the MIM process. In:
3. Additive Manufacturing Wohlers Report 2018 shows rise in metal Proceedings of the World PM2016 Congress & Exhibition.
additive manufacturing. https://www.additivemanufacturing. EPMA, Hamburg, pp 9–13 October 2016
media/news/wohlers-report-2018-shows-rise-in-metal-additive- 22. Andersen O, Riecker S, Studnitzky T et al (2018) Manufacturing
manufacturing. and properties of metal parts made by fused filament fabrication. In:
4. Lieberwirth C, Harder A, Seitz H (2017) Extrusion based additive Euro PM2018 Proceedings. EPMA, Bilbao, Spain, 14 – 18 October
manufacturing of metal parts. JMEA 7. https://doi.org/10.17265/ 2018, pp 1–6
2159-5275/2017.02.004 23. Kurose T, Abe Y, Santos MVA, Kanaya Y, Ishigami A, Tanaka S,
5. Gong H, Crater C, Ordonez A, Ward C, Waller M, Ginn C (2018) Ito H (2020) Influence of the layer directions on the properties of
Material properties and shrinkage of 3D printing parts using 316L stainless steel parts fabricated through fused deposition of
Ultrafuse Stainless Steel 316LX Filament. MATEC Web Conf metals. Materials 13:2493. https://doi.org/10.3390/ma13112493
249:01001. https://doi.org/10.1051/matecconf/201824901001 24. Gloeckle C, Konkol T, Jacobs O, Limberg W, Ebel T, Handge UA
6. Thompson Y, Gonzalez-Gutierrez J, Kukla C, Felfer P (2019) (2020) Processing of highly filled polymer–metal feedstocks for
Fused filament fabrication, debinding and sintering as a low cost fused filament fabrication and the production of metallic implants.
additive manufacturing method of 316L stainless steel. Addit Materials 13:4413. https://doi.org/10.3390/ma13194413
Manufact 30:100861. https://doi.org/10.1016/j.addma.2019. 25. Nestle, Nikolaus, Hermant, Marie-Claire, Schmidt, Kris (2016)
100861 Mixture for use in a fused filament fabrication process. 40
7. (2020) Making additive manufacturing sustainable: ask the right 26. German RM, Bose A (1997) Injection molding of metals and ce-
question. tct MAG 28:24–25 ramics. Metal Powder Industry, Princeton
8. Lingqin X, Guang C, Luyu Z, Pan L (2020) Explore the feasibility 27. Vieira MT, Martins AG, Barreiros FM, Matos M, Castanho JM
of fabricating pure copper parts with low-laser energy by selective (2008) Surface modification of stainless steel powders for
laser melting. Mater Res Express 7:106509. https://doi.org/10. microfabrication. J Mater Process Technol 201:651–656. https://
1088/2053-1591/abbd08 doi.org/10.1016/j.jmatprotec.2007.11.162
2464 Int J Adv Manuf Technol (2021) 115:2449–2464

28. Ferreira TJ, Vieira MT (2017) Optimization of MWCNT – metal properties of highly concentrated Inconel feedstock alloy 718.
matrix composites feedstocks. Ciência Tecnol Mater 29:87–91. Powder Technol 322:273–289. https://doi.org/10.1016/j.powtec.
https://doi.org/10.1016/j.ctmat.2016.07.010 2017.08.049
29. Carreira P, Cerejo F, Alves N, Vieira MT (2020) In search of the 35. Ibrahim MHI, Muhamad N, Sulong AB (2011) Rheological char-
optimal conditions to process shape memory alloys (NiTi) using acterization of water atomised stainless steel SS316L for Micro
fused filament fabrication (FFF). Materials 13:4718. https://doi. MIM. Adv Mater Res 264–265:129–134
org/10.3390/ma13214718 36. Kaleta J, Wiewiórski P, Wiśniewski W (2017) Investigation of
30. Ferreira TJJ (2018) Microinjection moulding of austenitic stainless martensitic transformation induced by cyclic plastic deformation
steel reinforced with carbon nanotubes, PhD Thesis. University of in austenitic steels. In: Borek W, Tanski T, Brytan Z (eds)
Coimbra Austenitic Stainless Steels - New Aspects. InTech. https://doi.org/
31. Tafti AA, Demers V, Majdi SM, Vachon G, Brailovski V (2021) 10.3390/ijerph14101126
Effect of thermal debinding conditions on the sintered density of 37. Li JB, Xie ZG, Zhang XH, Zeng QG, Liu HJ (2010) Study of metal
low-pressure powder injection molded iron parts. Metals 11:264. powder extrusion and accumulating rapid prototyping. Key Eng
https://doi.org/10.3390/met11020264 Mater 443:81–86. https://doi.org/10.4028/www.scientific.net/
32. Cerejo F, Raimundo AM, Vieira MT, Barreiros FM (2019) Fused KEM.443.81
deposition of WC+Co powder. EPMA, Maastricht Exhibition & 38. Matweb (2019) SS 316L. http://www.matweb.com/search/
Congress Centre, Maastricht DataSheet.aspx?MatGUID=530144e2752b4770
33. Ferreira TJ, Vieira MT (2014) Behaviour of feedstocks reinforced 9a58ca8fe0849969. Accessed 15/02/2021
with nanotubes for micromanufacturing. In: Euro PM 2014 Congr.
Exhib. Proc. EPMA. The Messezentrum Salzburg, Austria
34. Dimitri C, Mohamed S, Thierry B, Jean-Claude G (2017) Influence Publisher’s note Springer Nature remains neutral with regard to jurisdic-
of particle-size distribution and temperature on the rheological tional claims in published maps and institutional affiliations.