POLITECNICO DI TORINO

Master’s Degree in Mechanical Engineering

Master’s Degree Thesis

Effect of Process Parameters on the Quality of 17-4

PH Samples Produced by Directed Energy Deposition

Tutors: Candidate:
Federica Bondioli Lorenzo Badi
Alberta Aversa
Alessandro Carrozza
Erica Librera

April 2021
Contents
Abstract ............................................................................................................................................................. 2
1 Additive Manufacturing Technologies ....................................................................................................... 3
1.1 Brief History of AM Development ..................................................................................................... 5
1.2 Generic AM process........................................................................................................................... 7
1.3 Classification of AM processes .......................................................................................................... 9
1.3.1 Powder Bed Fusion ...................................................................................................................... 10
1.3.2 Directed Energy Deposition ......................................................................................................... 11
1.4 Metal Powder Manufacturing ......................................................................................................... 15
2 Stainless steel ........................................................................................................................................... 17
2.1 17-4 PH ............................................................................................................................................ 19
2.1.1 AM microstructure ...................................................................................................................... 19
2.1.2 Mechanical properties ................................................................................................................. 22
3 Materials and Methods ............................................................................................................................ 24
3.1 DED Machine in Prima Additive....................................................................................................... 24
3.2 Powder............................................................................................................................................. 26
3.3 Single Scan Tracks ............................................................................................................................ 30
3.4 Massive Cubes ................................................................................................................................. 33
4 Results and Discussion ............................................................................................................................. 38
4.1 Powder............................................................................................................................................. 38
4.2 Single Scan Tracks ............................................................................................................................ 42
4.3 Cubes ............................................................................................................................................... 50
4.3.1 Porosity ........................................................................................................................................ 52
4.3.2 Microstructure ............................................................................................................................. 68
4.3.3 Hardness ...................................................................................................................................... 70
5 Conclusions .............................................................................................................................................. 74
References ....................................................................................................................................................... 75

1
Abstract

This thesis project is the result of the collaboration between Politecnico di Torino and Prima industrie S.p.A.
The group Prima Industrie operates in the field of the mechanical industry, in particular it develops,
manufacture and markets lasers, sheet metal cutting machinery, electronics, CNCs components and additive
manufacturing systems. This last field is the most recent and is managed by the newly constituted Prima
Additive division. It deals with both Directed Energy Deposition and Powder Bed Fusion technologies.
Directed Energy Deposition is an innovative Additive Manufacturing method that represents the evolution of
the cladding technology. However, it is still on evolution, and, for this reason, the process parameters must
be evaluated and optimized case by case. In the framework of a scientific support, Prima Additive assigned
the analysis of single scan tracks and massive cubes samples made of 17-4 PH. The single scan tracks were
produced covering a wide range of process parameters and the aim was providing a correlation between the
process parameters and the meltpools’ characteristics. The massive cubes, instead, were printed using
different sets of parameters selected by the experience with the objective of analysing their quality.
In this dissertation, the state of the art of Additive Manufacturing technologies is described, focusing on the
metal based ones and presenting the metal powder production. Then, the 17-4 PH stainless steel is
presented, providing an outline of its characteristics when used in Additive Manufacturing processes. The
experimental section of the thesis describes the machine of Prima Additive, the powder feedstock used, how
the samples were produced, prepared, and characterized. The feedstock powder was analysed in terms of
particle size distribution, flowability and tap density. It was observed an excellent flowability, crucial for DED
applications. Then, the on-top and the cross-section observations were carried on the single scan tracks,
defining the deposition effectiveness and the meltpools’ geometrical characteristics. Some correlations
between them and process parameters were found. The massive cubes’ quality was analysed in terms of
porosity, microstructure, and hardness. Generally, they had high hardness and good porosity values, even if
few sets of parameters showed more defects than others. In conclusion, the single scans’ analysis allowed to
exclude the most critical sets of process parameters while the overall most promising sets were found
through the massive cubes’ analysis.

2
1 Additive Manufacturing Technologies
The term Additive Manufacturing (AM) denotes the industrial process of fabricating an object, by the addition
of layers, from a 3D CAD model (Fig. 1.1). This innovative technology is particularly suitable for complex
shaped components that would be impossible, or inconvenient, to produce using traditional subtracting
methods [1]. The term 3D-printing is informal and used by the general public or in a domestic DIY context.
The first commercialized machines date back to the mid-80s, when the first patent was filed in the USA. In
this period, this technology was used only to build conceptual prototypes and so was known as Rapid
Prototyping (RP). The main advantage was that the model was built in short time, few hours, without tools
nor process planning required. Afterwards, not only prototypes but also casting patterns (Rapid Casting) and
tools (Rapid Tooling) were built. During the last two decades, the expression Additive Manufacturing has
been spreading, since end-usable parts could be produced, reaching the highest and most valuable
application of the technology [1].

Fig. 1.1 General additive manufacturing process [2]

Nowadays, amongst manufacturing processes, AM technologies are the fastest growing in the global market
and they are becoming faster and cheaper year after year [3]. The advantages of adopting these technologies
are both in process and product: one machine can produce almost any kind of shape in just one
manufacturing step and without needing a mould or special tools, leading to a reduced operator intervention
[4]. So, thanks to the design freedom, complex hollow shapes can be created, obtaining an optimized and
lightweight product. Consequently, in AM time and costs are only related to size, not to complexity, in first
approximation. Traditional manufacturing, instead, is based on subtractive processes and requires a certain
level of human interaction, for example, part must be repositioned or relocated within the machine or even
in a different one. In this case, the cost of a component strongly depends on its complexity, as schematized
in Fig. 1.2.

3
 Fig. 1.2 Comparison between part costs in traditional manufacturing and AM [4]

On the other hand, AM technologies are a fairly new and still evolving reality, thus they have some drawbacks
that restrict their field of application. First of all, large objects cannot be produced due to the working
chamber size limitations and, in some cases, due to lack of strength in the material itself. Building large
objects might also be impractical due to the limited machine speed [5]. Then, even if prices are falling, both
machines and materials are still expensive, so onerous investments from the companies are usually required
[6]. Also, the materials choice is currently limited and these are mostly polymers [4]. At the end, object
produced through AM often have some imperfections in the superficial finish that might result ribbed and
rough [6]. For this reason, a certain amount of post-processing is typically required.
Even if many researches and improvements have been carried on during the last decades, it is unrealistic that
AM techniques will make the traditional ones obsolete. It is expected, instead, that AM will progressively
increase its role as complementary technology in the industrial landscape [6].

4
1.1 Brief History of AM Development
The concept of producing a 3D object layer by layer occurred long before the advent of CAD software or AM
ideas.
One of the first patents dates back to 1902: George J. Peacock filed a patent for laminated horseshoes
“Method of making composition horseshoes” [7]. It consisted of many layers of strong fabric with a thin
rubber in between. The attempt was to create custom horseshoes in a much easier way than in the traditional
methods.
The origins of the real AM technology, however, are intricate to define. There were definitely a lot of
researches in the 1950s and 1960s, but the applications became effective only thanks to remarkable
development of associated technologies, such as computers, lasers, controllers, CNC machines etc. in the
early 1980s [1]. The first material used in AM was a photopolymer resin, invented in the 1950s by the
chemical industry DuPont (Wilmington, Delaware) [8]. Through a process called polymerization, the liquid
resin hardened when exposed to a light source, such as a laser. From the late 60s to the 80s many studies
and experiments on the resin were carried out, mostly in USA, but also in Japan and France [1].
In 1984, Charles Hull of UVP, Inc. (San Gabriel, California) invented and filed the first stereolithography
apparatus to produce three-dimensional objects [9]. He described a process of directing a computer-
controlled UV laser beam in a vat of photopolymer resin: 3D prototypes were formed by curing the resin
layer by layer while submerging the build platform deeper into the vat while between [8] (Fig. 1.3).
In 1986, Hull and Raymond Freed co‐founded 3D Systems Inc. (Valencia, California) [10]. The following year,
SLA-1 system was launched: (Fig. 1.4) this was the first ever commercialized AM system in the world and
showed impressively detailed parts for that time [8].

Fig. 1.4 The SLA-1 system: the first AM technology

Fig. 1.3 Basic working principle of
launched in 1987 by 3D systems [1]
stereolithography apparatus [12]

In the following years, Japanese companies were particularly concentrated on improving stereolithography
processes, while companies in USA and Israel were developing other approaches to additive manufacturing
[8]. The most of AM technologies, indeed, was invented during the 90s.
The first Fused Deposition Modelling (FDM) was commercialized by Stratasys (Eden Prairie, Minnesota) in
1991 [3]. A thermoplastic material was extruded in filament form trough a computer-controlled deposition
head, creating a 3D object [6]. In 1993 the Massachusetts Institute of Technology patented the Binder Jetting
(BJ): an inkjet mechanism added droplets of liquid binder i.e., glue to layers of ceramic powder [3]. In
particular, Soligen, Inc. (Northridge, California) applied this principle and commercialized the Direct Shell

5
Production Casting (DSPC), informally called Rapid Casting, to create shells for investment casting processes
[8]. In 1994 Solidscape (Wilton, New Hampshire) invented the InkJet Printing (IJP) and launched the Model
Maker [13]: The mechanism consisted of an inkjet printing head that deposited wax layer by layer. Using the
same concept, 2 years later, 3D Systems sold the first 3D printer: this was the first mention of the term “3D
printers” among industrial machinery [3].
In 1991 Helisys (Torrance, California) invented the Laminated Object Manufacturing (LOM) [14]. A roll of
plastic material or paper was unrolled, stacked onto the part, and bonded to the previous layers, then, a
computer guided laser cut the sheet, and the excess of material was rolled onto the waste roll [1].
In 1992 DTM (Austin, Texas) launched the Selective Laser Sintering (SLS) system [14]. The heat of a laser
melted powder materials, building layer by layer. A similar machine was patented one year later by Ralf
Larson of Larson Brothers co.: The Electron Beam Melting (EBM), which was, at the time, the first AM process
using a metal powder. In fact, the electron beam was able to weld metallic powder creating a thin layer of
solid metal [11].
In 2000 Precision Optical Manufacturing (Auburn Hills, Michigan) introduced the Direct Metal Deposition
(DMD) technology [8], otherwise called Directed Energy Deposition. Its working principles will be extensively
presented in the following chapters.
Starting from the 2000s, AM has been used for production parts and the last two decades are characterized
by an exponential growth of AM development [4]. Both the existing technologies and materials keep on
evolving while new ones are studied, developed, and commercialized [1].
In 2020 additive manufacturing became very popular to public opinion due to the COVID-19 pandemic
because it allowed to produce quickly and cost-effectively specialized surgical instruments and medical
devices like valves, masks and everything that may prove to be of primary importance in the fight of the first
wave [15]. The idea started in Italy [17] but quickly spread in Europe and USA, so much successfully that
AMable, a consortium supported by the European Commission, published an Open Call for solution ideas to
fight COVID-19 adopting additive manufacturing [18].

6
1.2 Generic AM process
The overall process chain is almost the same, regardless from technology adopted and is summarized in Fig.
1.5

Fig. 1.5 The eight stages of a generic AM process [1]

The process chain consists in the steps required to generate the physical part starting from its concept.
1. Generation of a CAD model
Computer Aided Design (CAD) is basically the first step of every product design process in industry.
The output file fully describes the geometry of the part, either in terms of a 3D solid or a set of
surfaces [14]. The most common CAD software are SolidWorks, Katia, Autodesk Inventors, Solid Edge,
Creo, NX, etc. [19]

2. STL conversion
The CAD model must be converted into acceptable format for the AM machine (STL). Luckily, almost
every CAD software can easily convert it. The term STL was derived from the first commercial AM
technology: STereoLithograhy [1] and it is also the acronym of “Standard Triangle Language" or
"Standard Tessellation Language" [20]. This file represents external closed surfaces with a triangular
mesh, forming the basis for slices calculation. Each triangle is described with its three vertexes and
its normal vector direction, indicating where the material is placed.

3. Transfer to AM Machine and STL File Manipulation

During the conversion process, some errors can occur such as gaps between cells, inverted normals
or intersections of triangles [14]. Therefore, the first task is to verify that everything in the part is
correct. Then, the user can manipulate the part, for example, it is possible to duplicate it to build
7
 more than one part in the same job. It is also possible to linearly scale STL files, to be slightly larger
or slightly smaller than the original to consider shrinkage or post-processing coating, typical of some
AM methods [1]. The user can set the position within the build chamber and the orientation to
optimize the process as well. In fact, additional supports are generally required to hold in place
geometries during fabrication [4] and the correct orientation of the part can significantly decrease
their number and size. They are usually calculated and added to the part by the software and may
be formed of the same material as the part, or a different one. They will be either mechanically
removed or dissolved away in secondary operations [1]. Then, the STL model is then sliced into a
sequence of parallel planes. This operation is called slicing and it can be performed in two ways.
Direct slicing is the simplest method because all the layers have the same thickness, instead, in more
sophisticated applications, adaptive slicing is adopted. In this case, the build layer thickness is
automatically adjusted to accommodate the surface geometry [4].

4. Machine setup
The AM machine must be properly set up prior to the build process. Usually, the build chamber must
be cleaned from previous works, then, the raw material is loaded [14]. The energy source level is
selected, and the inert gas injection dosed. Finally, the head motion is set up, including scanning
speed, hatching distance and layer thickness.
All these parameters are significantly specific to the machine, process, and material. Usually, for
advanced machines, the parameters are chosen to optimize final quality, part cost or building time.
Unlike classic subtractive technologies, a wrong setup may still produce a part without any serious
accidents, although its final quality will probably be unacceptable [1].

5. Build
While the previous steps require some human interaction, the build phase is fully automatic. The
material deposition and the layer creation can be separate or simultaneous. The machines have a
deposition head or a moving platform and they can combine from 3 to up to 5 or even more axis,
including roto-tilting mechanisms for particularly complex shapes. As long as no errors are detected
(overheating, no more raw material) the machine repeats the process until the build is complete [1].

6. Removal
At the end of the building process, the part is removed from the building chamber and a human
interaction may be required. For this reason, safety interlocks ensure that the temperatures are
sufficiently low and that there are no moving parts [6]. Eventual excess of material in the camber is
cleaned away, and additional supports are removed carefully to avoid possible damages to the part.
Also, they can be made of a different material from the part, to facilitate the operation.

7. Post-Processing
Unfortunately, only few AM processes build a ready-for-use part: most of them require a significant
amount of post-processing [5]. This step is strongly application specific too, as it depends on output
accuracy and material properties. Post-processing can be a simple clean up, manual sandpapering,
machine polishing, priming, and painting or a more complex operation, such as infiltration or surface
coating application to strengthen the final product. Heat treatments or post-curing can be required
too [14].

8
 8. Application: when all requirements are satisfied, the part is finally completed. The AM processes
cover a significantly wide field of application, which is constantly increasing [1]: aerospace,
automotive, medical and dental, jewellery, architectural, design and furniture, fashion, food industry,
filters and lattice structures, hinges and assemblies, heat exchanges with optimized dissipation
surface and new opportunities for fluid dynamics.

1.3 Classification of AM processes

Two different methods of classification may be applied to the AM processes: by material class (polymer,
metal etc.) or by raw material form (powder, wire etc.). A certain variety of materials can be processed using
the AM technologies: polymers, metals, ceramic and even organic materials. Up to now, polymers have been
prevalent in the AM landscape due to their easier formability, but metals have been spreading. AM
technologies for production of metallic parts are reported in Table 1.1. Powder Bed Fusion (PBF) and Directed
Energy Deposition (DED) are the predominant ones and will be exposed in the following dedicated
paragraphs.

Sheet Lamination

Selective Laser
Binder Jetting
Sintering

Selective Laser
Powder Bed Fusion
Metal AM Melting

Electron Beam
Melting

Powder DED
Directed Energy
Deposition
Wire DED
Table 1.1 Metal AM processes

9
1.3.1 Powder Bed Fusion

In PBF techniques, the object is built on a lowering platform, layer after layer, by an energy source directed
on a powder bed. The working principle is depicted in Fig. 1.6. A layer of powder is deposited from the
delivery system to the fabrication bed by a roller or a recoating blade, then the energy source selectively
melts (or sinters) a part of the layer, the platform moves downwards, and the sequence is repeated until the
built is completed. The powder bed is usually pre-heated to achieve better compaction and lower residual
internal stresses.

Fig. 1.6 Schematic of Powder Bed Fusion process in Selective Laser Melting
configuration [14]

Depending on the energy source type, powder bed processes are divided into electron beam or laser. Laser
processes can be further divided into Selective Laser Melting (SLM) and Selective Laser Sintering (SLS), they
are essentially the same but SLS gives a lower amount of heat, sintering the metal powder rather than melting
it and can be applied also to plastics, glass, and ceramics.
Laser technologies exploit photons directed by a scanner system through a galvanometer mirror, without
moving the laser source for higher efficiency. They adopt inert atmosphere of nitrogen or argon to minimize
oxidation and porosities, as long as these defects are typically responsible for lower mechanical
performances.
Instead, Electron Beam Melting (EBM) uses a thermionic emission gun [3] which emits electrons from a
tungsten filament. The environment is a vacuum chamber (high vacuum, 10-5 Pa), which makes easier to keep
very low oxygen levels. First, the beam preheats the layer and then melts it using a 70-250 µm layer thickness.
The high energy of electrons permits a high melting capacity. The beam is deflected by electromagnetic coils,
providing a very fast and accurate control resulting in higher productivity then laser processes. EBM also
produces lower thermal stresses, but the surface finish is poorer due to larger particle sizes being used.

10
1.3.2 Directed Energy Deposition

The principle behind DED technique resembles robotic welding and laser cladding (i.e., bonding together two
different metallic parts) technologies. Depending on the manufacturer, DED machines sharing the same
general approach are indicated by an extreme variety of denominations: 3D Laser Cladding, Directed Light
Fabrication, Direct Metal Deposition, Laser Based Metal Deposition, LaserCast, Laser Consolidation, Laser
Direct Casting, Laser Freedom Fabrication, LasForm, Laser Engineered Net Shaping etc. [1].
DED creates part by melting and depositing material at the same time, from thin wire or powder feedstock
and the focused heat source can be an electron beam or a laser (Fig. 1.7).

Fig. 1.7 DED configurations: (a) laser beam and powder feedstock and (b) electron beam and wire [21]

Wire feedstock implies a deposition efficiency close to 100%, as approximately all the material used is melted
and deposited onto the substrate [21]. Powder, instead, is a more versatile feedstock, it allows to build more
complex objects, especially if they have thick and thin regions [1]. Electron beam systems require a vacuum
chamber and do not have significant oxidation issues [21]. Laser systems, instead, require some methods to
keep oxidation under control that will be exposed in the following pages. The most common configuration is
laser (L-DED) using powder feedstock [14] and so the following considerations will refer to this one in
particular.

11
During the process, the laser melts a small point of the substrate, or previous layer, producing the so-called
melt pool, highlighted in Fig. 1.8. Its dimensions depend on many factors such as the material, laser
characteristics and other process parameters. As powder is deposited into the melt pool, it melts and
solidifies as soon as the laser beam moves away, creating a thin track. Given the short duration of the melting
process, the high cooling rates (1000-5000 K/s) typically generated, produce large thermal gradients and non-
equilibrium grain structures impossible to obtain via traditional processes like casting. Neighbour tracks are
overlapped, typically 25% of their width, creating a layer. This results in re-melting of the previously
deposited material.

Fig. 1.8 Melt pool in DED [22]

The deposition head can be mounted on a robotic arm [23],[24] or on a classic CNC machine from 3 up to 5
axes [25]. Typically, the object remains in a fixed position but, in case of particularly complex geometries,
also the substrate holder can translate, rotate, or tilt, increasing the number of degrees of freedom [25].
Moreover, the beam does not need to be strictly perpendicular to the platform. Within certain angles, the
oblique deposition is equally effective thanks to the high velocity of particles [1] (Fig. 1.9).

Fig. 1.9 An example of oblique deposition [26]

12
Powder particles are transported and ejected though the nozzles by an inert gas carrier, such as nitrogen or
argon. Different nozzles configurations exist (Fig. 1.10), it can be a single lateral one, 3 or 4 nozzles coaxial to
laser beam or a continuous ring, creating a conical nozzle.

Fig. 1.10 Different nozzle configurations. a) wire, b) single lateral

nozzle, c) radially symmetric nozzles and d) conical nozzle. [27]

The 4 nozzles configuration is the most common because it is a good compromise between quality outcome
and simplicity of construction [1] (Fig. 1.11).

Fig. 1.11 Deposition head with 4 powder nozzles

configuration adopted by Prima Additive [25]

DED technology has some unique characteristic in the AM landscape: it is used to build new parts onto a flat
metal plate but also to repair, coat and add features to existing components. DED, in fact, does not require
a powder filled chamber, so dealing with much greater objects than in other AM techniques is possible.
Moreover, technological development makes possible to adopt different material within the same object
[21], avoiding cross contamination problems, almost inevitable in PBF. This technique is also known as
Gradient Metal Alloys [28]. Functionally graded materials can be built controlling the composition in each

13
point of the component, allowing to strategically vary its physical and the mechanical properties. However,
this technology is used only in extremely advanced applications.
As previously mentioned, in order to minimize oxidation phenomena and delamination problems, the
atmosphere around the meltpool can be controlled in two different ways with inert gas, typically nitrogen or
argon. The whole building chamber can be filled with the gas, like in the PBF, or the shielding gas solution
can be adopted. In this latter case, a gas flow comes from dedicated nozzles in the deposition head and
creates a local protection around the meltpool, this method can be adopted when very large components
are processed. Aversa et al. [29] observed that 316L samples produced in a controlled build chamber
atmosphere had finer and fewer oxides than the ones produced with the shielding gas method. On the other
hand, the latter method is also suitable in case of particularly large parts that might not suit the building
chamber.

Fig. 1.12 Scheme of the different methods to achieve a controlled atmosphere [29]

Generally, the final part has a very high density, and its mechanical properties are usually better than the
ones produced with traditional methods. On the other hand, surface finish is quite poor due to the presence
of adhered partially molten particles.
DED is a rather new technology and there is a lack of standardization concerning process parameters and
machines. The former can be studied and optimized in order to achieve maximum density, quality,
mechanical properties, but also cost and working time. These can be classified in main groups: powder
feeder, laser, scanning speed and geometrical parameters. The powder feeder controls the amount of
powder feedstock injected into the system and it is measured in g/min. The speed of the particles depends
on the carrier gas flow rate, which is measured in l/m. Laser-related parameters include laser power and spot
diameter. The scanning speed refers to the linear velocity of the head during the deposition of the scans and
it is usually measured in mm/min. The geometrical parameters define the characteristics of the deposition in
terms of overlapping between single beads. The hatching distance is the distance between the centre of two
consecutive scans while the z-step, or Δz, is the increase of head height between two consecutive layers. Δz
must be as close as possible to the actual layer thickness. In the end, the scan pattern, also canned scanning
strategy, identifies the tracks’ directions within one layer and how the subsequent layer changes. A large
variety of possibilities exists, an example is reported in Fig. 1.13.

14
 Fig. 1.13 Scanning strategies: a) Parallel horizontal b) parallel vertical c) Anti-parallel
horizontal d) Anti-parallel vertical e) Out-Spiral and f) In-Spiral [16]

1.4 Metal Powder Manufacturing

Powder Metallurgy (PM) is a set of techniques used to produce a component from metal powder. It was
adopted long before the development of additive manufacturing technologies, in fact, classic PM has existed
since the end of 19th century. In classic PM, the metal powder is compressed into the desired shape by a rigid
die, roll, or injection mould and then sintered i.e., heated at a temperature well below melting. Instead, in
AM technologies many layers of powder are stacked and melted [14].
There are several methods to produce metal powder, which strongly influence powder characteristics and,
consequently, the properties of the final component. For example, the final component density reflects the
presence of the internal porosity of particles and air gaps between them, also chemical composition is
important in order to avoid possible undesired phases or defects like oxide films. Also powder flowability has
an important role in AM because it affects its spread or deposition behaviour. It mainly depends on particles’
size and morphology. The preferred shape is spherical because it guarantees higher flowability and better
results. Some methods accomplish rather spherical particles, while others require further processing.
Different properties can even occur amongst different vendors applying the same technique [30]
The powder production technologies are divided in two main groups: mechanical and physical/chemical
methods. Mechanical ones, also called atomization processes, are the most common and are based on the
dispersion of a molten metal by a high-pressure jet of gas or liquid. Physical/chemical methods, instead,
involve a transformation in chemical composition and structure from raw material.
The gas atomization (Fig. 1.14a, e) is the most widespread process, and it is used for a large variety of alloys
like Fe, Ni, Co but also Ti and Al [14]. A furnace melts, usually under vacuum, the metal, then the molten
metal enters the chamber through nozzles in a thin flow and it is dispersed into small droplets by a high-
pressure inert gas that quickly solidifies into spherical (or near to sphere) shape particles. Their size
distribution strongly depends on the metal and system used. An inert gas may be used to avoid oxidation of
particularly reactive metals like titanium alloys. Water atomization is a similar process, but it employs a high-
pressure water jet instead of a gas and it is recommended for unreactive metals such as steels. It is less
15
expensive but can create oxidation problems and very irregular shape particles due to even higher cooling
rate.
Another atomization technique is the plasma atomization (Fig. 1.14d), it is suitable for reactive and high
melting temperatures metals like Ti. A plasma torch is used to disperse the metal wire feedstock into
droplets. Particles are characterized by highly spherical shape, high flowability and high purity powder with
low oxygen content.
The mechanical milling (Fig. 1.14c), instead, is the simplest method and has been used for a long time to
produce metallic powder, it includes compression, collision, and shear processes but it is not suitable for AM
applications since the final particles appear extremely irregular shaped. [30]
The most common physical-chemical method is Electrolysis (Fig. 1.14b). Due to its intrinsic characteristics, it
is appliable only to pure metals and not alloys and so its adoption is rather limited. [30]

Fig. 1.14 SEM images of different powders: gas atomized In718 (a) and Ti–6Al–4V (e); electrolysed Fe (b); mechanically
milled Fe–18Cr–8Ni 12Mn–N (c); plasma atomized Ti–6Al–4V (d). [30]

16
2 Stainless steel
The steels are the most widespread alloy in the mechanical field. Their peculiar components are iron and
carbon but also many other alloying elements can be added such as Mn, Cr, Mo, Ni, Mn, Si while other
elements are considered harmful even in low percentages like S, P, O, H, N [32]. Alloying elements modify
the Fe-C phase diagram, thus they are generally classified as austenite stabilizer or ferrite stabilizer,
depending on the phase field they enhance.
The stainless steels are adopted in aggressive environments due to their high corrosion resistance. According
to the EN 10088 standard, steel having a minimum chromium content of 10.5% can be classified as stainless
[31] even if, in real applications, a higher percentage is commonly adopted. Chromium has the ability to
create an extremely thin layer of oxide (Cr2O3) on the surface, protecting the material beneath. This layer is
formed spontaneously when the steel is exposed to an oxidizing environment (e.g., air), moreover, it adheres
really well and it immediately restores when scratched [32].
Depending on the composition and on the microstructure at room temperature, stainless steels are classified
using the Schaeffler’s diagram in Fig. 2.1

Fig. 2.1 Schaeffler diagram for stainless steels [33]

Four main groups are identified: martensitic, ferritic, austenitic, and duplex.
Martensitic stainless steels have the lowest contents of alloying elements, their mechanical properties
depend on the heat treatments adopted, in some case can achieve exceptional ultimate tensile strength, up
to 2000 MPa [32]. Due to their high C content, they are the least corrosion resistant among the stainless
steels [31]
Ferritic stainless steels have moderate mechanical properties and, depending on the content of Cr and other
alloying element, the corrosion resistance ranges from moderate to very good. They have a BCC lattice
structure and show ductile-to-brittle transition. [32]
Austenitic stainless steels represent the most widespread group, they have a considerable oxidation and
corrosion resistance, they have an FCC lattice structure and so they are amagnetic and do not show ductile-

17
to-brittle transition. Their mechanical properties are good but cannot be enchanted by heat treatments
because they do not have critical points. [32]
Duplex stainless steels have a biphasic structure of austenite and ferrite thanks to the balanced combination
of Cr and Ni [31]. They conjugate high mechanical properties and high corrosion resistance.
An additional group, precipitation-hardenable, is not strictly based on microstructure but on the type of heat
treatment they usually require [34]. Also called precipitation hardening, or PH, these steels were invented
with the aim of enhancing mechanical properties, particularly the fracture toughness, and increasing the
corrosion resistance of the classic martensitic stainless steels. They are further classified into subgroups,
depending on the microstructure they show, into martensitic, semi-austenitic and austenitic PH stainless
steels [34]. The hardening mechanism is achieved through solution annealing and aging heat treatments:
alloying elements like copper, aluminium, niobium, and titanium form finely dispersed precipitations,
strengthening the crystalline matrix of the steel. The corrosion resistance is ensured by the presence of
chromium, nickel, and molybdenum and a very low carbon content, usually lower than 0.1 % [31].

18
2.1 17-4 PH
The 17-4 PH alloy can assume many different denominations: the aforementioned one is according to ASTM,
the EN 10027 designation is X5CrNiCuNb16-4, while the DIN one is 1.4542 and finally the AISI denomination
is AISI 630. Due to its composition, reported in Table 2.1, it is classified as a martensitic precipitation
hardening stainless steel. An example of its microstructure is reported in Fig. 2.2 The most common heat
treatments are solution annealing at 1040°C for 30-60 min and ageing at 480-630 °C for 1-4 hours [31]. Upon
them, Cu and Ni precipitate into intermetallic compounds.
Table 2.1 Chemical composition of 17-4 PH stainless steel [31]

Name %C %Si %Mn %P %S %Cr %Ni %Mo %Cu %Nb

17-4 ≤0.07 ≤0.70 ≤1.50 ≤0.040 ≤0.015 15.0 ÷ 3.50 ÷ ≤0.06 3.00 ÷ 5 x %C ÷
PH 17.0 5.00 5.00 0.45

Fig. 2.2 Martensitic microstructure in a 17-4 PH sample after aging

treatment at 620°C [31]
17-4 PH shows high mechanical properties and corrosion resistance [35], making it suitable for applications
in aerospace, naval, chemical, nuclear, energy and biomedical industries. After thermal treatments, it is
possible to achieve a tensile strength of 850-1400 MPa with an elongation between 12% and 25% [31].

2.1.1 AM microstructure

According to Yadollahi et al. [36], precipitation hardening steels are one of the most attracting stainless steels
to be processed with AM techniques because of their martensitic/austenitic microstructure and good
weldability. 17-4 PH in particular, besides its high mechanical properties and corrosion resistance, has good
deposition characteristics [37]. For all these reasons, it is well established in AM applications and studies [23],
[24], [35-40].

19
Generally, the production process strongly influences the characteristics of the metal and, in this case, DED
technologies provide cooling rates extremely higher than conventional ones. Thermal history, indeed, defines
microstructural features i.e., grain size and morphology, which reflect on mechanical properties
consequently [36]. For this reason, in order to achieve the desired mechanical properties, it is crucial to study
the influential parameters on microstructure. Studies on different metal alloys used in AM [35], [36] [41-44],
show that the resulting solidified microstructure depends on many factors. The most effective are
temperature gradient (G) and growth rate (R) i.e., ratio between cooling rate and thermal gradient [41].

𝑐𝑜𝑜𝑙𝑖𝑛𝑔 𝑟𝑎𝑡𝑒
𝑅 = 𝑡ℎ𝑒𝑟𝑚𝑎𝑙 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 = 𝑔𝑟𝑜𝑤𝑡 𝑟𝑎𝑡𝑒 (2.1)

𝐺 = 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 𝑎𝑡 𝑙𝑖𝑞𝑢𝑖𝑑/𝑠𝑜𝑙𝑖𝑑 𝑖𝑛𝑡𝑒𝑟𝑓𝑎𝑐𝑒 (2.2)

Their product G x R determines the grain size, a higher G x R results in finer structures while the ratio G/R
determines their morphology, as schematized in Fig. 2.3

Fig. 2.3 Diagram showing how temperature gradient

and growth rate affect morphology and size of the
final microstructure [44]

20
The reason of this phenomenon is that, within the meltpool, different heat transfer mechanisms take place
[42]. As showed in Fig. 2.4, conduction occurs along the meltpool borders, creating a heat affected zone
(HAZ) in the lower layers. In the middle of the meltpool, the metal remains liquid for a slightly longer time,
enhancing convection. Lateral sides, instead, are in contact with the environment and radiation is
predominant.

Fig. 2.4 Gaussian distribution of laser beam intensity

and different heat transfer mechanisms in the
meltpool [43]

These different mechanisms affect G and R parameters and consequently the microstructure: Dendrites with
columnar structures (Fig. 2.5 e) grow in direction of maximum thermal gradient, perpendicularly to meltpool
border towards the centre [35]. In the centre, instead, grains have an equiaxed structure (Fig. 2.5 d). Meltpool

Fig. 2.5 Different structures in meltpools (a), (b), (c): equiaxed (d) and columnar (e) in an
AISI 316L sample processed by DED [42]

21
borders are visible in Fig. 2.5 a), b) and it worth observing that they are curved due to gaussian distribution
of laser intensity.
The microstructure of the as-built 17-4 PH samples produced by DED is depicted in Fig. 2.6 and it is composed
of martensite and retained austenite. Their percentages depend mainly on the gas process [38]: Argon
atmosphere gives a predominantly martensitic microstructure, while in nitrogen atmosphere, due to its
austenite stabilizing effect (it can dissolve in meltpool), the retained austenite can reach up to 50% [39].

Fig. 2.6 Equiaxed and columnar grains in a 17-4 PH sample [38]

2.1.2 Mechanical properties

The most common mechanical properties in studies are divided into hardness, usually Vickers hardness (HV)
and tensile properties i.e., yield strength (YS), ultimate strength (US) and elongation (ε). As-built samples
exhibit good properties, but these can be further enhanced by heat treatments, as previously stated. A
comparison of them is shown in Table 2.2.
Table 2.2 Mechanical properties of 17-4 PH processed by AM before and after heat treatments [35], [36], [40]

Sample type YS [MPa] US [MPa] ε [%] Hardness [HV]

As-built 500-700 950-1050 20-50 250-350
After heat treatments 750-1170 1200-1450 7-20 350-450

The beneficial effects of thermal treatments are manifold: besides the aforementioned precipitation
phenomena, retained austenite suppression occurs. Studies showed that the presence of retained austenite
has deleterious effects on hardness [37], yield and ultimate tensile strengths but increases elongation [36].
Also, Pavlina et. al [45] demonstrated a certain linear correlation between tensile strength and hardness.
Furthermore, AM processed samples are characterized by a strong anisotropy due to their inhomogeneous
microstructure, but the meltpool borders may disappear after heat treatments [38], obtaining a more
homogeneous microstructure.
Although the mechanical properties of the 17-4 PH alloy are mainly affected by thermal treatments, studies
revealed that also the parameters typical of AM processes affect them.

22
The Linear Energy Density (LED) is defined as the ratio between laser power and scanning speed. Studies
showed that a lower LED has a positive effect both on porosity [23] and microhardness [24], [38]. High LEDs,
in fact, promote evaporation and pores formation [37], [40] which have a harmful effect of mechanical
properties. Besides, at equal laser powers, high scanning speed values could promote grain refinement, and
possibly an increase in microhardness, since there is a shorter interaction-time between the laser and the
metal [24].
Not only the LED but also the building direction can affect the tensile properties, as demonstrated in various
studies [36], [37], [46]. The built was defined horizontal (or 0°) when the scans were parallel to the tensile
direction and vertical (or 90°) in the perpendicular case. As showed in Fig. 2.7, scan directions created
different structures within the sample, which was characterized by a certain anisotropy. For this reason,
horizontal samples had higher tensile properties i.e., YS, US, and ε. Yadollahi et al [41] reported that the
interfaces between layers created preferential sites for crack formation and propagation because they
contained more defects. Oh et al. [37], instead, suggested that heat treatments could homogenize the
structure of the samples, reducing their performance discrepancy.

Fig. 2.7 Cross sections and representation of different samples built in the horizontal (a) and in
the vertical (b) direction [46]

Mechanical properties depend also on the powder characteristics and powder feed rate: samples built using
recycled powder showed a slight degradation of tensile strength and much lower elongation: the reasons
could be the more irregular shapes of the particles, that cause an increase in fusion defects, pore number
and size [47]. Higher powder feed rate decreases porosity, this effect is more relevant at higher scanning
speed [40].

23
3 Materials and Methods
This chapter describes the experimental procedures of the thesis. First, the characteristics of the DED
machine are presented, then, the powder properties are illustrated, as well as the additional
characterizations on it. Turning to the main core of the work, the samples are presented and divided into
single scan tracks and massive cubes. For each of them, the production, the preparation, and the analysis are
detailly explained.

3.1 DED Machine in Prima Additive

The samples were produced in Prima Additive by a DED machine used for experimental researches (Fig. 3.1).
The machine is a prototype, and it is composed by a complex apparatus, which includes some different
commercially available parts.

Fig. 3.1 DED machine prototype in Prima Industrie [48]

The main unit is built on the base of a 3-axes gantry CNC machine and it is controlled through a dedicated
software developed by Prima Electro. When the main door is closed, the working chamber is protected from
the external environment. This is crucial when a controlled atmosphere is necessary and the working area is
filled with inert gas to minimize the oxygen internal content. In this case the glovebox system allows the
manipulation of the pieces inside without opening the door. Also, a special dark glass filters the laser light,
allowing the operators to see inside the chamber during the building operations safely.
The control panel allows the operator to manage the machine functions. The head can be controlled manually
to position it in the correct place prior to the deposition phase. The samples are built using dedicated files,
called Part Programs that contain all the process parameters, in particular laser power and head movements.
The powder and the gas flow parameters, instead, are controlled through the powder feeder.

24
The laser source is a commercial Ytterbium fibre laser, with a maximum power of 3 kW (Fig. 3.2). The laser
beam arrives to the deposition head through an optical fibre cable. The wavelength is 1070 nm, and the laser
spot has a diameter of 2mm.

Fig. 3.2 Fibre laser [49]

A chiller is used to cool down the laser, keeping it at a controlled temperature. Another chiller unit is used to
cool down the entire optical chain and nozzle, which is close to the process.
The powder feeder is constituted by two hoppers and a control panel (Fig. 3.3a). The hoppers have separated
circuits so, they can be used to store up to two different materials. The powder feedstock level can be seen
through a round glass window at the base of each hopper Fig. 3.3b). Behind it, there is a gear system that

Fig. 3.3 Powder feeder (a) and detail of gear system (b)

doses the powder feed rate, whose rotational speed is set through the control panel. In fact, there is a precise
correlation between the main gear RPM and the powder feed rate measured in g/min.

25
The control panel also manages the flow rates of the gases used as powder transport and as shielding. The
inert gases used generally for the direct deposition process are nitrogen and argon. The transport gas is used
to push the particles into the pipe system, up to the deposition head.
The deposition head (Fig. 3.4) was designed by Prima Additive and the spacer part is produced using the
Powder Bed Deposition technique. Its innovative design includes complex internal channels in a slim
structure at the same time. The powder is ejected through four holes coaxial to laser beam.

Fig. 3.4 Deposition head [48]

During the deposition, in case of a controlled atmosphere process, an oxygen sensor monitors the oxygen
level inside the chamber. In fact, it must be kept at minimum levels to avoid oxidation phenomena.
At the end of the building process, a certain amount of waste powder is present, and a metallic container
allows an easier collection of it. However, some residues of powder can be present outside the container, all
around the building chamber. For this reason, an industrial vacuum cleaner is used. Different material
powders require different bags for further specific recycling treatments or disposal.

3.2 Powder
The 17-4 PH powder used to produce all the samples studied, was produced by an external partner supplier
and it was indicated for additive manufacturing applications. The supplier provided a test certificate, in
accordance with the standard EN 10204-3.1. The chemical composition is reported Table 3.1 and the sieve
analysis results in Table 3.2.
Table 3.1 Powder chemical composition

Element C Si Mn P S Ni Cr Mo Nb Cu N O
Weight 0.05 0.61 0.27 0.014 0.004 3.02 15.25 0.04 0.26 3.13 0.09 0.0140
composition (%)

26
Table 3.2 Particle size distribution

Grain size [µm] +250 +180 -150 -106 -53 -45

Distribution (%) 0.0 4 84 52 4 1

The powder was further analysed in Politecnico’s laboratories. In particular, it was observed under the
Scanning Electron Microscope (SEM), then, the flowability test was carried out and its density was calculated.
The SEM adopted was the Phenom XL, produced by the Thermo Fisher Scientific (Fig. 3.5). Compared to
optical microscopes, SEMs offer a much higher magnification and image resolution, this model offered a 320x
– 200.000x magnification and a resolution of 10 nm, according to the specifications provided by the supplier.

Fig. 3.5 Phenom XL a) and samples holder b) [51]

The SEM analysis was useful not only to check the size and shape of particles, but also to observe possible
defects like satellites. All the collected images were further processed through the image-processing software
ImageJ to calculate the particle size distribution of the powder in the following way: first, the original image
was opened (Fig. 3.6a), and the scalebar was used to calibrate the measurements. Then, the colours were
inverted, a darkness threshold was set and the particles, i.e., the darkest regions, were highlighted (Fig. 3.6b).
At the end, the area of each particle was measured (Fig. 3.6c).

Fig. 3.6 ImageJ steps for particle size analysis: original image (a), inverted colours image with darkest regions
highlighted (b), particles counted and measured (c)

27
The flowability is an important parameter for DED applications and it expresses the ability of a powder to
flow. High flowability is usually required to achieve a powder flow as steady as possible from the hopper [14],
and hence a stable deposition, but also to avoid system occlusions. The friction between particles depends
on their physical characteristics like dimension, shape, and roughness. Generally, large, round, and smooth
particles flow faster [30]. Also, humidity can negatively affect flowability, sticking particles together, thus the
powder must be kept dry. The flowability test was performed following the ASTM B213 standard [70]. The
apparatus is shown in Fig. 3.7: it consisted of a conical funnel, called Hall, with a calibrated angle, surface
roughness and orifice diameter of 2.54 mm, and a cylindrical brass cup with a capacity of 25 cm3. The bottom
of the funnel was placed approximately 25 mm above the cup using a support. So, 50 grams of powder were
poured into the Hall funnel, using a finger to block them. When the finger was released, the powder began
to flow, if this did not occur, a gentle tap to the funnel was allowed, and the emptying time was measured.
In case the tap was ineffective, a funnel with a bigger hole (5.08mm), called Carney, could be adopted.
Consistently, the unit of measure of flowability is [s/50g].

Fig. 3.7 Flowability test apparatus [52]

According to the ASTM B212 standard [71], the density of a powder is reported using two parameters:
apparent or bulk density and tap or tapped density. The standard defines how to calculate them: the
apparent one was obtained filling up to the edge the aforementioned cup with the nominal capacity of 25
cm3 and measuring the weight. Then, the container was placed onto a shaking plate that compacted the
existing powder, so it was filled up to the edge again and weighted. The procedure was repeated until the
change between the last two measurements was negligible, the last result was used to obtain the tap density.
These two density values were then used to calculate the Hausner ratio and the Compressibility index as
follows:
𝜌𝑡𝑎𝑝
𝐻𝑎𝑢𝑠𝑛𝑒𝑟 𝑅𝑎𝑡𝑖𝑜 = (3.1)
𝜌𝑎𝑝𝑝𝑎𝑟𝑒𝑛𝑡

𝜌𝑡𝑎𝑝 −𝜌𝑎𝑝𝑝𝑎𝑟𝑒𝑛𝑡
𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑖𝑏𝑖𝑙𝑖𝑡𝑦 𝐼𝑛𝑑𝑒𝑥 = 𝜌𝑡𝑎𝑝
(3.2)

28
As well as the previously mentioned flowability expressed in s/50g, these two parameters are commonly
used to judge the quality of a powder in terms of flow. 7 categories of flow exist in the literature and they
are shown in Table 3.3.
Table 3.3 Flowability characterization [38]

29
3.3 Single Scan Tracks
Depending on the studies carried and their relative objectives, many different types of samples were built.
For this thesis project single scan tracks and massive cubes were considered.
Single scan tracks (SST) are the simplest shaped samples: they consist of a linear deposition on the baseplate.
The SST of this thesis work are 40 mm long and are illustrated in Fig. 3.8.

20 mm

Fig. 3.8 Single Scan Tracks deposited on two baseplates

This type of sample is particularly suitable for preliminary studies and characterizations of materials and
processes, allowing to estimate the feasibility window. SST samples do not have geometrical implications
because, of course, the multiple meltpool interactions, such as the hatching distance and the z-step, are not
present. Secondly, a large number of specimens can be printed changing in a wide range their process
parameters. These are summarized in Table 3.4 and consist of 5 different values of laser power (P), 4 of
scanning speed (v) and 2 powder feed rates (F) for a total of 40 different combinations. A nitrogen gas carrier
flow was always constant, and no shielding gas was adopted.
According to the trade secret policy of the company, all the process parameter values are a confidential
information and for this reason they are expressed as a percentage of the maximum value considered.
Table 3.4 Building parameters adopted in the SST production

Laser power P 50%, 61%, 72%, 83%, 100%

Scanning speed v 36%, 57%, 79%, 100%
Powder flow F 70%, 90%

The Linear Energy Density (LED) is defined as the ratio between the laser power and scanning speed, it is
measured in J/mm and it is a commonly adopted as a parameter in single scan tracks studies.
𝑃
𝐿𝐸𝐷 = (3.3)
𝑣

The analysis on the SSTs allowed to study the dimensions and geometry of the single meltpool, with the aim
to carry out a first exclusion on process parameters that did not guarantee satisfactory results and find the
correlations between them and the meltpools’ geometry. The samples required a certain metallographic
preparation, i.e., a series of steps, to obtain a plain, mirror polished and clean surface.

30
The baseplate was cut around the SSTs using a cutting machine: the plate was held between two clamps,
where an abrasive cermet circular blade cut it transversally. During this operation, thermal alterations had
to be carefully avoided, so two nozzles supplied a lubricant flow that prevented overheating, blade excessive
wear and sparks. Also, the machine was equipped with a fume extractor fan. Then, single scans were cut at
around half length, to ensure that a steady state deposition was reached.
Due to the small size of the specimens, the single scan tracks were mounted in resin to facilitate their
handling. They were put into a rectangular rubber mould, as shown in Fig. 3.9, which was manually filled with
an acrylic bicomponent resin. This operation took place under chemical fume hood due to the hazardous
vapours released.

Fig. 3.9 Rubber moulds used to mount samples in resin. Keeping trace of the
sample ID number is crucial.

After the cutting operations, the surfaces of the samples showed many irregularities and blade scratches, so
a grinding process was necessary and the polishing machine Presi Minitech 233 was used (Fig. 3.10).

Fig. 3.10 Polishing machine Presi Minitech 233

This machine consisted of a rotating plate where abrasive paper was fixed by a ring. A series of abrasive
papers was used, with a progressively finer grit: P400, P800, P2000 and P4000, with a respective grain size of
35, 21.8, 10.3 and 5 μm. A water flow was directed onto the plate to avoid excessive heat and to remove
31
waste particles. Subsequently, to achieve a finer mirror finish, velvet cloth disks with diamond paste of grain
size 3 and then 1 μm were used. Those cloths required an alcohol-based lubricant rather than water.
Then, the samples were carefully washed with water and a drop of liquid soap to remove any impurities from
the surface and then gently dried by a compressed air flow. An example of the final result is shown in Fig.
3.11.

Fig. 3.11 SST mounted in resin and polished

The first analysis was an on-top observation of single scans using a stereomicroscope Leica EZ4W (Fig. 3.12a),
this could be useful to detect noticeable deposition irregularities or any other issues. Then, images of the
cross-sections were taken with a Leica DMI 5000 optical microscope (Fig. 3.12b) at 50x magnification.

(a) (b)

Fig. 3.12 Leica EZ4W stereomicroscope (a) and Leica DMI 5000 optical microscope (b) [53]

32
The cross-section images allowed to see the meltpools. The geometrical features considered were track
width (W), growth (G), and depth (D), as highlighted in Fig. 3.13, and they were measured with the image-
processing software ImageJ.

Fig. 3.13 Characteristic features of single scans: Width

(W), Growth (G) and Depth (D)

3.4 Massive Cubes

The massive cubes are a type of sample that allows to study how the meltpool interacts one another both
horizontally and vertically. Their building requires much more time and material feedstock than the SSTs.
Prima Additive chose the process parameters amongst the best ones of the feasibility window found in the
SST analysis, in terms of highest productivity. They also considered their past experience.
The massive cubes had theoretical dimensions of 20x20x15 or 20x25x15 mm, the scanning strategy adopted
was 0-90°, like as SST deposition, Nitrogen gas carrier flow was always constant, and no shielding gas was
adopted.
The parameters adopted for the cubes production are reported in Table 3.5 and consist of 3 different values
of laser power, 4 of scanning speed, 4 powder feeder (powder speed RPM), 3 hatching distance and 3 layer
thickness for a total of 7 different combinations.
Table 3.5 Building parameters adopted in massive cubes productions

Laser power P 50%, 61%, 72%

Scanning speed v 57%, 64%, 71%, 79%
Powder flow F 50%, 70%, 80%, 100%
Hatching distance HD 71%, 93%, 100%
Z step ΔZ 56%, 78%, 100%

The main evaluation parameter adopted for the analyses of the massive cubes was the Volumetric Energy
Density (VED) and it is measured in J/mm3.
𝑃
𝑉𝐸𝐷 = 𝑣⋅𝐻𝐷⋅∆𝑍 (3.4)

Generally, lower VEDs were associated to higher powder flow rates with the aim of obtaining a rather
balanced deposition. In fact, with limited powder flow rates, the amount of the deposited powder was
insufficient, leading to a poor outcome.

33
The cubes, like the SSTs, underwent a certain metallographic preparation process. First of all, the baseplate
was cut around the cubes (Fig. 3.14a), avoiding the separation between them. Samples were then sectioned
using a smaller and more accurate cutting machine Brillant 220 (Fig. 3.14b).

Fig. 3.14 Massive cube ready for baseplate cut (b) and the brillant 220
cutting machine (b)

All the cubes together are shown in Fig. 3.15.

20 mm

Fig. 3.15 Set of all the 8 cubes produced and analysed

After the cut, the massive cubes were consequently sectioned and polished. In Fig. 3.16 the two cross-
sections of the same cube are shown before and after the polishing operation.

10 mm

Fig. 3.16 Massive cube sections before and after grinding and
polishing operations

The characterization on the massive cubes regarded hardness, porosity, optical and SEM microscope analysis.
So, the objective was to identify the best sets of process parameters and also how they influence the cubes’
quality.

34
Hardness
Hardness is a material characteristic expressing its penetration resistance and it is commonly measured by a
machine that forces an indenter onto the specimen [32]. According to the standard procedures, the indenter
is positioned onto the surface and a load is applied, when it is removed, a mark is visible on the surface.
Measuring its dimensions, it is possible to calculate hardness. Different hardness tests exist, such as Vickers,
Brinell, Rockwell, Knoop depending on the machine.
In this study, the Vickers hardness, was considered. The instrument used employed a square base pyramidal

Fig. 3.17 Vickers hardness schematics [54]

indenter made of diamond, with a vertex angle of 136°, as reported in Fig. 3.17. It is worth noticing that the
square shape guarantees less uncertainty in measurement than, for example, a circle. Impressions were seen
through the ocular of the instrument and measured.
Besides the length of impression’s diagonals, the parameters defining the Vickers test are the applied load
and its time. Finally, the hardness value is given by the formula:
𝑃
𝐻𝑉 = 0.1891 (3.5)
𝑑2

Where P is the applied load in N and d is the mean value of the diagonals in mm.
In this experimental activity, a load of 0.3 Kg was applied for 15 s. For each sample, 3 measurements were
performed at the bottom, centre, and top part of the cross section surface, for a total of 9 measurements.
The aim was not only to obtain an average value for each sample, but also to investigate if the hardness
values changed along the height of the sample.

35
Porosity
Cubes were placed on the optical microscope at a 100x magnification and a set of 20 images was taken for
each of them in an unbiased way. The images were further analysed through ImageJ software, through the
passages indicated in Fig. 3.18: a darkness threshold was set with the aim of considering the pores but
excluding possible scratches, still remaining after the polishing. The percentage of black areas represented
the pores percentual porosity.

c
Fig. 3.18 Imagine elaboration: a) original image. b) darkest
region highlighted. c) only black and white image

Furthermore, the software also measured the area and the Feret’s diameter of each black dot. The Feret’s
diameter corresponds to the distance between the two farthest points of a pore, as represented in Fig. 3.19.

Fig. 3.19 Graphical representation of maximum Feret diameter [33]

This data was useful to make a dimensional distribution analysis of the defects.

36
SEM observation
A Scanning Electron Microscope (SEM) Phenom XL was used to analyse the samples, in particular to observe
fusion defects such as pores, partially molten powder particles or oxides. In case of suspected oxides, the
technique Energy Dispersive X-ray Spectrometry (EDS) was adopted: exploiting electron diffraction, it could
provide an estimation of chemical composition of the selected spot.

Microstructure
In order to observe the microstructure of the cubes, an acid etching was performed. The acid solution, in
fact, selectively attacked grain boundaries, highlighting the microstructure. Marble’s reagent was prepared
mixing 4 g of copper sulfate (CuSO4) in 20 mL of hydrochloric acid (HCl) and 20 mL of water, then, always
under fume hood, it was gently swept on the sample using a cotton swab and then dried. This technique is
called swabbing and it is particularly suitable for strong acid solutions like this one.
Therefore, the samples were again observed under the optical microscope. From each sample, some images
were taken at 50x, 100x, 200x, and 500x magnifications to investigate the microstructure and observe defects
or particularities.

37
4 Results and Discussion
4.1 Powder
The scanning electron microscope (SEM) images of a small sample of the powder are collected in Fig. 4.1

Fig. 4.1 SEM images of powder, general view a) and some defects in particular view b), c)

It was observed that the powder had a considerable amount of defects, the most common ones were
satellites of various dimensions. Other defects were elongated shape particles, irregular surfaces, and some
not perfectly molten particles. In some other particles, small black dots were visible, and they might be
oxides. An example of these suspect oxide is shown in Fig. 4.2.

Fig. 4.2 Possible oxides found on powder particles

38
A total of 22 images was analysed using ImageJ software to measure the diameter of each particle detected.
According to the powder supplier’s datasheet, only 1% of the particles had a diameter below 45 μm. So, in
this analysis, the lower threshold for the diameter was set at 20 μm. In order to calculate the particle size
distribution of the powder, 18 diameter classes were defined with an amplitude of 10 μm, up to 200 μm in
diameter, i.e., the maximum value found. Then, the particles were sort and put in the corresponding diameter
class, obtaining the simple numerical distribution. With the purpose of obtaining a more significant outcome,
the area-normalized distribution was calculated. Firstly, every particle was approximated as a circle, and its
area A was calculated from the diameter d.
𝑑2
𝐴=𝜋⋅ (4.1)
4

Each i-th class contained ni elements. The weight of each class wti was calculated summing the areas of all its
particles.
𝑖 𝑛
𝑤𝑡𝑖 = ∑𝑘=1 𝐴𝑘 (4.2)

Finally, the influence of the i-th class (Xi) was calculated as the ratio between the corresponding weight wti
and the sum of the weights of the 18 classes.
𝑋𝑖 = 𝑤𝑡𝑖 /(∑18
ℎ=1 𝑤𝑡ℎ ) (4.3)

As reported in Table 4.1, the highest number of particles had a diameter between 40 and 60 μm. Particles
above 100 μm diameter were less than 10% of the total and no particles larger than 200 μm in diameter were
detected. In the area-normalized distribution, instead, bigger particles had more weight, and in particular,
particles above 100 μm covered almost 25% of the total.
Table 4.1 Calculated particle size distribution of the powder

Diameter Numerical Area-normalized Diameter Numerical Area-normalized

Range [μm] Distribution Distribution Range [μm] Distribution Distribution
20-30 6.7% 1.0% 110-120 1.4% 4.5%
30-40 11.4% 3.5% 120-130 0.7% 2.4%
40-50 20.0% 9.7% 130-140 0.3% 1.4%
50-60 23.0% 16.4% 140-150 0.2% 1.2%
60-70 12.9% 13.0% 150-160 0.3% 1.9%
70-80 8.7% 11.5% 160-170 0.2% 1.4%
80-90 6.1% 10.4% 170-180 0.1% 0.8%
90-100 4.6% 10.0% 180-190 0.3% 2.7%
100-110 2.5% 6.5% 190-200 0.2% 1.9%

39
The numerical distribution is presented in Fig. 4.3 adopting two graphical representations: the standard and
the cumulative cases. The diameters reported on the horizontal axis indicate the upper limit of each class.

Fig. 4.3 Numerical distributions of particle diameters

Analogously, the area-normalized distributions, are provided in Fig. 4.4 and, in particular, the cumulative
distribution allowed to calculate the characteristic parameters of the powder D10, D50 and D90, respectively
46, 75 and 140 μm.

Fig. 4.4 Area-normalized distributions of particle diameters

The area-normalized distribution had a much wider peak, between 50 to 100 μm, and the particles above
120 μm gained much more weight. In fact, in the numerical distribution, they were extremely few.
The area-normalized distribution was considered the most significative and so it was compared to the particle
size distribution provided by the supplier (Table 4.2). The same class diameters of the supplier’s datasheet
were adopted.
Table 4.2 Comparison between the supplier’s datasheet and calculated distribution

Grain size [μm] +250 +180 -150 -106 -53 -45

Supplier distribution [%] 0 4 84 52 4 1
Calculated cumulative area-normalized distribution [%] 0 5 91 80 20 9

40
The calculated distribution did not perfectly reflect the supplier’s one because the content of smallest
particles was overestimated. In particular, this method overestimated the number of particles below 106 μm,
this was even more noticeable below 53 μm. However, particles above 250 μm were not found in neither of
the two cases. In other words, the powder appeared finer than expected. The reasons of this might be
manifold, first of all, the supplier adopted a physical sieve method instead of an image-based one. Then, even
at the lowest SEM magnification i.e., 320x, a single image did not contain many particles and a significative
amount of them was cut at the edges. This issue affected more likely larger particles. The software, of course,
could not consider the cut particles, even if they were cut only for a small portion and this might be the
reason of the underestimation of larger particles.
The other analysis on the powder included flowability and density calculations. The flowability test was
conducted using the Hall funnel, the measurements were repeated three times and the results are shown in
Table 4.3. The average value of 15.7 ± 0.2 s/50g was found.
Table 4.3 Flowability test results

Test Value [s/50g]

1 15.9
2 15.5
3 15.6
Average 15.7
Standard Deviation 0.2

Afterwards, the apparent and the tap densities were measured so, the Hausner Ratio and the Compressibility
Index were calculated (Table 4.4).
Table 4.4 Density results

Apparent Density [g/cm3] 4.45

Tap Density [g/cm3] 4.71
Hausner Ratio 1.06
Compressibility Index 5.5%

According to the previously mentioned Table 3.3, the powder placed in the first category with an excellent
free flow.
Unlike the particle size distribution, the flowability and the density values were not present on the supplier’s
datasheet, hence could not be counterchecked.

41
4.2 Single Scan Tracks
The stereomicroscope pictures for the on-top analysis are sort by building parameters in Fig. 4.5. The SST
were observed at about half of their length to ensure that a steady-state deposition was reached. Most of
SST had in the start and the end points a faded deposition, indeed not representative.
Two general trends were found: the SSTs produced with the same scanning speed showed a width increase
with laser power. Instead, SSTs with the same power exhibited more adhered particles when the scanning
speed increased. These behaviours were observed also by Oh et al. [40] on 17-4 PH samples. In particular,
relatively to the latter trend, they stated that the surface roughness and waviness increased with the
scanning speed. Beside this, some critical issues were found.
50% power samples exhibited a poor quality, especially at 70% powder feed rate and high scanning speed:
depositions had slightly irregular borders and many unmelted particles on them. The same issue, at minor
extent, was observed also at 61% power and 100% scanning speed. Samples with the same LEDs but higher
powers, instead, did not show this issue, suggesting that the ineffective or incomplete melting was caused
by an insufficient laser power rather than an insufficient LED. On the other hand, 100% power samples had
much more regular borders and less unmelted particles but showed a slightly colourful surface. The colours
ranged from brown and gold to blue and purple and were present both on the samples and on their borders.
This is a consequence of a more severe oxidation phenomena and higher thermal shock caused by the higher
temperature reached. However, samples with same LEDs but lower powers did not show as much severe
oxidations. Eo et al. [72] investigated the effects of laser power on AISI 316L samples produced by DED. They
used different laser powers but kept the energy density constant and observed that higher power caused
higher oxidation. Moreover, all the samples of this thesis were produced in free air, without any controlled
atmosphere, local or total. The adoption of a controlled atmosphere might have reduced the oxygen partial
pressure on the meltpool [72] and so the oxidation.
In conclusion, according to this simple qualitative analysis, both scarce power and excessive power should
be avoided because they led to some issues, a middle power might be preferrable, as suggested in different
studies [40], [65] [72], [73].

42
 5 mm

Fig. 4.5 On-top views of single scans, sort by building parameters

43
The images from the optical microscope of the cross-section are reported in Fig. 1.2. Three different colours
were used to assess the quality of the deposition in terms of geometry.

2 mm

Fig. 4.6 Cross-section views of single scans, sort by building parameters

44
The adopted criteria for the preliminary classification were based on a comparison among the samples and
were the following:

• High acceptability (green). Scans that showed all of these characteristics: regular shape, average
growth 0.4 mm < G < 0.6 mm , and sufficient penetration D > 0.15 mm.
• Middle acceptability (orange). Scans that showed a lower growth 0.3 mm < G < 0.4 mm or an
excessive one G > 0.6 mm but a sufficient penetration D > 0.15 mm.
• Low acceptability (red). Scans that showed irregular shape, either insufficient growth G < 0.3 mm, or
insufficient penetration D < 0.15 mm.

At first observation, the most critical samples were the ones built with 50% laser power, for both powder
flows, as anticipated in the on-top analysis.

In particular, for 70% powder flow, the sample 1A (Fig. 4.7) exhibited very large pores on the left side. Defects
like those would surely be unacceptable in a final component.

Fig. 4.7 Large porosities on sample 1A

For 90% powder flow, samples 6C and 6D had highly irregular shapes and almost no penetration (Fig. 4.8).

Fig. 4.8 Irregular profiles on samples 6C (a) and 6D (b)

Furthermore, it was observed that almost every sample produced at 100% scanning speed had an insufficient
material deposition. Their dimensions were in fact rather smaller than the corresponding ones with the same
power.

After a first qualitative analysis of on-top and cross-section views, it was concluded that 50% and 100%
laser power and 100% scanning speed parameters brought to critical single scan tracks.

Deeper investigations were carried on the cross-section images, measuring the geometrical features i.e.,
Width (W), Growth (G) and Depth (D). The results are listed besides the process parameters in Table 4.5.

45
Table 4.5 Single scan tracks results

Sample Scanning Powder Width Growth Depth Powder

Power LED G/D
ID Speed flow [mm] [mm] [mm] Efficiency
1A 50% 36% 50% 70% 1.441 0.283 0.144 2.0 12%
1B 50% 57% 31% 70% 1.537 0.215 0.159 1.4 16%
1C 50% 79% 23% 70% 1.475 0.245 0.075 3.3 23%
1D 50% 100% 18% 70% 1.353 0.157 0.082 1.9 17%
2A 61% 36% 61% 70% 1.892 0.175 0.249 0.7 10%
2B 61% 57% 38% 70% 1.752 0.391 0.175 2.2 32%
2C 61% 79% 28% 70% 1.72 0.309 0.172 1.8 34%
2D 61% 100% 22% 70% 1.61 0.284 0.09 3.2 38%
3A 72% 36% 72% 70% 2.088 0.694 0.142 4.9 43%
3B 72% 57% 45% 70% 1.885 0.468 0.18 2.6 42%
3C 72% 79% 33% 70% 1.824 0.383 0.175 2.2 45%
3D 72% 100% 26% 70% 1.764 0.273 0.159 1.7 40%
4A 83% 36% 83% 70% 2.368 0.331 0.293 1.1 23%
4B 83% 57% 52% 70% 2.07 0.443 0.211 2.1 43%
4C 83% 79% 38% 70% 1.933 0.369 0.216 1.7 46%
4D 83% 100% 30% 70% 1.87 0.303 0.268 1.1 47%
5A 100% 36% 100% 70% 2.633 0.618 0.204 3.0 48%
5B 100% 57% 63% 70% 2.34 0.522 0.191 2.7 57%
5C 100% 79% 45% 70% 2.184 0.355 0.23 1.5 50%
5D 100% 100% 36% 70% 2.016 0.309 0.2 1.5 51%
6A 50% 36% 50% 90% 1.784 0.627 0.126 5.0 26%
6B 50% 57% 31% 90% 1.654 0.442 0.093 4.8 27%
6C 50% 79% 23% 90% 1.471 0.313 0.153 2.0 23%
6D 50% 100% 18% 90% 1.552 0.196 0.077 2.5 20%
7A 61% 36% 61% 90% 1.941 0.391 0.186 2.1 17%
7B 61% 57% 38% 90% 1.854 0.391 0.205 1.9 27%
7C 61% 79% 28% 90% 1.708 0.347 0.161 2.2 30%
7D 61% 100% 22% 90% 1.695 0.246 0.12 2.1 27%
8A 72% 36% 72% 90% 1.809 0.293 0.188 1.6 12%
8B 72% 57% 45% 90% 1.792 0.36 0.198 1.8 24%
8C 72% 79% 33% 90% 1.943 0.494 0.159 3.1 49%
8D 72% 100% 26% 90% 1.774 0.315 0.175 1.8 36%
9A 83% 36% 83% 90% 2.125 0.282 0.26 1.1 14%
9B 83% 57% 52% 90% 1.941 0.517 0.153 3.4 37%
9C 83% 79% 38% 90% 2.015 0.426 0.175 2.4 44%
9D 83% 100% 30% 90% 1.813 0.309 0.164 1.9 36%
10A 100% 36% 100% 90% 2.591 0.831 0.183 4.5 50%
10B 100% 57% 63% 90% 2.347 0.599 0.192 3.1 52%
10C 100% 79% 45% 90% 2.233 0.522 0.192 2.7 59%
10D 100% 100% 36% 90% 2.065 0.353 0.202 1.7 47%

46
For each sample, W, G and D were plotted in function of the LED for each of the 5 laser powers (Fig. 4.9).
W was between 1.3 and 2.7 mm and, in first approximation, it slightly increased with the powder flow rate.
W had the clearest trend: it increased with LED i.e., power increase and scanning speed decrease. However,
the trend was less certain at 90% powder flow and 72% power. Also the sample with 50% power, 50% LED
and 70% powder flow had an irregular behaviour, but it was due to its irregular shape, previously described
in Fig. 4.7. High LEDs, so, created a larger meltpool, increasing the width of the scan track. In other words,
more powder got effectively melted, enhancing the deposition effectiveness.
G lied between 0.15 and 0.85 mm and most of the values were condensed in the interval 0.2-0.45 mm Again,
a slightly increase with powder flow was observed in first approximation. Its trend, instead, showed a quick
increase between 15% and 40% LED and then, in some samples continued to increase, while in others,
decreased inverting the trend and. In both width and growth, 90% powder flow samples had a light increase,
compared to 70% powder flow ones. It might be concluded that LED promoted the growth until a certain
limit value, around 50%-60%. After that, the deposition becomes less effective.
D was the most fluctuating characteristic, it varied between 0.07 and 0.3 and had the least certain trend:
generally, D seemed to increase with LED and slightly decrease with powder flow increase. This might be
reasonable, because, with the same amount of energy supplied, higher powder flow means more powder to
melt, and so, less energy remaining to melt the substrate.

Fig. 4.9 Evaluation of geometrical features for 70% a) and 90% b) powder flow

Some more precise information were searched in the literature. In fact, single scans are a standard type of
sample and are typically studied in DED applications.
Ho et al. [40] focused their study on 17-4 PH SSTs produced by DED and suggested that if a higher energy is
supplied i.e., higher laser power, the melt pool dimensions increase and so W, G and D. When the scanning

47
speed decreases, instead, only W and G increase, while D remains almost constant. They also stated that an
increase in powder feed rate increases W and G but does not significantly affect D.
Returning to the topic of how energy splits to melt the supplied powder and the substrate, the ratio G/D
could be considered. Studies [58], [59] showed that, indeed, G/D is an important parameter to estimate the
deposition effectiveness. If it is too low, namely G/D < 0.5, not enough new material is deposited and an
excess of previous layers material is melted, finally resulting in an irregular shape component with the
collapsed top portion. Conversely, if it is too high i.e., G/D >2 the bonding to previous layer is too weak and
so delamination and cracks may occur. The optimal range is set at 0.5 < G/D < 1.
G/D ratio was plotted, again, in function of the LED for each of the 5 laser powers (Fig. 4.10). All the values
lied in the range 0.7-5, even if the most of them was between 1-2 and 1.5-2.5, respectively for 70% and 90%
powder flow rates. The separated effect of powder flow rates on G and D previously analysed was confirmed
here, with their ratio: G/D slightly increases with powder flow rates.

Fig. 4.10 G/D ratios for 70% a) and 90% b) powder flow. The optimal and the acceptable regions are highlighted in
green and yellow, respectively

Going into details, only one sample (P 61%, v 36%, F 70%) lied in the aforementioned optimal range: all the
others had a higher G/D ratio. In case of 70% powder flow, the majority of the samples lied in the acceptable
interval between 1 and 2 but were fewer for 90% powder flow. So, it could be concluded that, under G/D
point of view, lower powder flow was preferrable.
Another key parameter in single scan study is the powder efficiency [60] which represents the fraction of
delivered powder that has actually solidified into the track. Powder feedstock is rather expensive, so it is
preferrable to obtain the efficiency as high as possible. It is calculated as
𝐴𝑔 ⋅𝑣⋅𝜌
𝑃𝑒 = (4.4)
𝐹

48
Where ρ is the material density, corresponding to 7.8 g/cm3 for 17-4 PH, as found in the literature [31]. F is
the powder flow rate and Ag is the growth area in mm2, calculated approximatively as half ellipse:
𝜋⋅𝑊⋅𝐺
𝐴𝑔 = (4.5)
4

The results were plotted in function of LED and power, like the previous graphs. Pe lied within a wide interval,
from 10% to 60% (Fig. 4.11), precisely, the highest value detected was 59% in sample 10C (100% P, 79% v,
90% F). Generally, a power increase led to a strong increase in Pe. Also, Pe was higher at higher scanning speed
and the most of 36% v samples were characterized by a considerable drop in P e, even if it did not happen in
100% power samples. The increase of powder feed rate increased Pe only in the 50% power samples, while
it did not have significant effect in 100% power samples and decreased in the others.

Fig. 4.11 Powder efficiencies for 70% a) and 90% b) powder feed rate

Summing up, the on-top analysis allowed to exclude the 100% and 50% values of laser power because they
lead to severe oxidation phenomena and insufficient and/or irregular deposition, respectively. The
ineffectiveness of 50% P deposition was confirmed by the cross-section images that highlighted their poor
quality. Then, the meltpools’ dimensions were measured and studied in function of LED and power. W
exhibited a clear trend, it increased with LED and P. G showed a similar behavior but, in some samples, after
a maximum value it decreased. Both W and G slightly increased with F. D, instead, slightly decreased with F,
even if its trend was not clear. The ratio G/D was calculated to judge the deposition effectiveness. All the
samples, except one, had an excessive G and were above the optimal range, even if the majority of them was
acceptable. In this case, G/D increased with F and so lower value was preferred. At the end, the powder
efficiency was calculated in order to understand the amount of powder that actually solidified into the scan
track. Low power and low scanning speed were particularly detrimental, leading to very low P e values. So,
higher power and middle-high scanning speed were preferrable.

49
4.3 Cubes
The analysis on the cubes started with some preliminary observations of their general deposition quality,
regarding their shape, size and surface finish.
All the lateral surfaces were flat and perpendicular to the baseplate. The top surfaces, conversely, were not
as flat as the lateral ones. For this reason, the minimum and the maximum heights were measured for each
cube and compared to the theoretical value of 15 mm. The obtained height ranges were further studied in
function of the VED. The deposition quality was classified through colours and, in general, higher VED
produced better cubes, as depicted in Fig. 4.12.

Fig. 4.12 Evaluation of the heights in function of the VEDs

All the green cubes had a VED above 70% and their height was slightly above the expected value of 15 mm.
The difference between the maximum and minimum heights were very small too, around 1 mm. Thus, they
had a good dimensional stability, and the top surfaces were almost flat. The yellow cubes intersected the
value of 15 mm, and the height spans were larger, resulting in a less flat top surface. Critical issues were
observed in the red cubes. Not only their heights were well below the expected value, but also the central
parts of their upper faces actually collapsed, resulting in an unacceptable outcome. The cube #6, for example,
had both excessive z-step and hatching distance, leading to the lowest VED amongst all the cubes. Despite
its 100% powder feed rate, the final outcome is a very bad deposition in terms of geometrical stability. This
happened because, during the building, after the deposition of each layer, the deposition head moved up
more than the actual growth of the just deposited layer. Also, higher z-step implies a lower number of layers
deposited, further enhancing the sample collapse problem. For these reasons, the z-step of the deposition
head must be set as close as possible to the layer growth. At the opposite, setting a too low z-step would
cause an excessive remelt of the previous layers and an inefficient deposition.

50
Furthermore, some other issues were observed on some samples regarding their low surface finish. Sample
#1 had many adhered round particles on its lateral surfaces, their sizes generally increased with the cube’s
height, up to 1mm in diameter. Analog particles but lower in number and size were present also on the cubes
#4 and #7. The other cubes, instead, showed a much better lateral surface finish, in particular #3 and #6.
The aforementioned samples’ characteristics are compared in Fig. 4.13.

Fig. 4.13 Examples of cubes with the highlighted features. White labels indicate the sample ID number

The finish of the top surfaces, instead, were very similar one another. Scan tracks were clearly visible and
created a ribbed surface. This is typical of the DED deposition and depend on the laser spot diameter.
In conclusion, it was observed that high VEDs produced geometrically stable cubes but with a poor lateral
surface finish. On the other hand, low VEDs had opposite outcomes. Thus, according to this preliminary
observation, the best process parameter set was associated to both the cubes #5 and #8. They combined the
best geometrical stability with a very good lateral surface finish, compared to the other samples.
After this preliminary observations, the cubes’ cross-sections were analysed in terms of porosity and
hardness. The building parameters and the main results of these analysis are provided in Table 4.6

51
Table 4.6 Cubes’ building parameters and general results of the porosity and hardness analysis

Sample ID 1 2 3 4 5 6 7 8
Power 50% 61% 72% 61% 72% 72% 50% 72%
Scanning speed 57% 57% 79% 64% 64% 79% 71% 64%
Powder flow 50% 70% 70% 70% 80% 100% 70% 80%
HD 71% 93% 93% 93% 93% 100% 71% 93%
ΔZ 56% 78% 78% 56% 78% 100% 78% 78%
VED 100% 67% 58% 84% 71% 42% 57% 71%
Build Rate [cm3/h] 24.0 43.7 60.1 35.1 49.1 83.2 42.0 49.1
Porosity [%] 0.092 0.044 0.078 0.031 0.041 0.034 0.033 0.030
Standard Deviation [%] 0.044 0.016 0.042 0.018 0.022 0.052 0.029 0.037
Average Hardness [HV] 406 400 392 418 413 399 415 403
Standard Deviation [HV] 4.9 11.6 7.1 5.5 7.2 11.0 10.8 7.1

The build rate expresses how fast the deposition is, in terms of built volume per unit of time [64]. It is
calculated as the product of scanning speed, hatching distance and layer thickness and it is measured in
cm3/h.
𝐵𝑢𝑖𝑙𝑑 𝑟𝑎𝑡𝑒 = 𝑣 ⋅ 𝐻𝐷 ⋅ 𝛥𝑍 (4.6)

This index represents an additional judgement parameter that was used at the end of the analysis.

4.3.1 Porosity

From each image taken at the optical microscope, a percentual value of porosity was found. The results of
all images collected are reported in Table 4.7. In particular, to facilitate the readability, a scale of colours
was assigned: green for values below 0.030% up to red for values over 0.17%. Porosity values, in fact, lied in
a rather wide range, from 0.007% in samples #6 and #8 to 0.249% in sample #6.

52
Table 4.7 Results of images analysis for porosity

Sample ID
1 2 3 4 5 6 7 8
1 0.119 0.055 0.116 0.024 0.042 0.021 0.030 0.054
2 0.130 0.053 0.217 0.029 0.048 0.010 0.033 0.065
3 0.090 0.049 0.134 0.049 0.040 0.016 0.016 0.017
4 0.072 0.027 0.079 0.034 0.026 0.028 0.012 0.027
5 0.021 0.041 0.086 0.071 0.056 0.014 0.048 0.014
6 0.168 0.048 0.044 0.013 0.028 0.011 0.015 0.041
7 0.178 0.031 0.027 0.026 0.030 0.023 0.019 0.012
8 0.111 0.069 0.074 0.024 0.095 0.027 0.026 0.014
Image porosity [%]

9 0.100 0.049 0.051 0.014 0.023 0.044 0.150 0.050

10 0.122 0.042 0.068 0.015 0.010 0.017 0.024 0.170
11 0.046 0.024 0.049 0.012 0.039 0.020 0.014 0.027
12 0.134 0.052 0.066 0.026 0.077 0.047 0.041 0.008
13 0.031 0.089 0.072 0.024 0.027 0.249 0.025 0.011
14 0.047 0.057 0.058 0.039 0.044 0.023 0.019 0.010
15 0.092 0.035 0.050 0.051 0.024 0.038 0.033 0.016
16 0.069 0.031 0.091 0.075 0.032 0.007 0.028 0.007
17 0.061 0.027 0.070 0.031 0.021 0.019 0.033 0.008
18 0.127 0.030 0.045 0.016 0.041 0.018 0.033 0.010
19 0.050 0.029 0.107 0.021 0.084 0.044 0.018 0.019
20 0.077 0.033 0.065 0.020 0.026 0.008 0.033 0.015

Average Porosity
0.092 0.044 0.078 0.031 0.041 0.034 0.033 0.030
[%]
Standard
0.044 0.016 0.042 0.018 0.022 0.052 0.029 0.037
Deviation [%]

Then, for each sample, standard deviation was calculated. It was observed that, in some of them, due to their
extremely wide range of porosities, standard deviation even exceeded the average value. All the samples
achieved a good average porosity, below 0.1%. #1 and #3 were respectively 0.092% and 0.078%, while the
others were well below these values, between 0.044 and 0.030. The best samples were #4, #6, #7 and #8.

53
For a better visualization of the results, the average porosity values were plotted in function of VED and
pictures with the porosity value close to the average ones were placed beside the corresponding samples
(Fig. 4.14).

Fig. 4.14 Average porosity and relative images

Moreover, the effects of other parameters, such as laser power, scanning speed and powder flow were
singularly investigated in Fig. 4.15. However, due to the low number of samples available and the high
number of parameters varying among them, no plausible trends were found.
However, Mathoho et al. [37] studied the effects of process parameters on 17-4 PH samples produced by
DED and concluded that high scanning speed promoted pore formation, probably because the meltpool
moved too fast to fill some pores. On the other hand, an insufficient scanning speed might cause an excessive
heat concentration, causing evaporation and pores formation High LEDs, in fact, promote evaporation and
pores formation. They also observed that lower powder feed rate resulted in higher porosity samples.

54
 Fig. 4.15 Average cube porosity in general a) and divided by power b), scanning speed c) and powder flow d)

Maximum porosity images were collected in Fig. 4.16. They exhibited a decreasing trend with respect to the
VEDs, even if sample #1 represented an exception. Also, an interesting behaviour was observed: high porosity
percentages could be caused by different factors: the presence of one large pore, such as samples #6 and #8,
many small pores like #1 and #4 or an intermediate situation between them, like #3 and #7.

Fig. 4.16 Collection of images with the highest average porosity

55
Nevertheless, the average value is not the only parameter to judge the quality of the samples: different
studies [55], [62], [63], [74] showed that the size of the pores could strongly influence the mechanical
properties of the AM processed components. For example, few large pores would be more detrimental than
many small ones, even if the average porosities were the same. For this reason, the following analysis
concentrated on the pore size.
Hence, the dimensions of the largest pores were measured, both in terms of area and of maximum Feret
diameter. In Fig. 4.17, the majority of samples had the trends of maximum porosity images and largest pores
almost coincident, but not samples #1 and #3. This means that in those samples, the maximum porosity was
mainly distributed in smaller pores rather than in one large.

Fig. 4.17 Comparison of maximum porosity images and largest pores

After this consideration, all the images were further processed to classify the dimensions of each visible pore,
rather than considering the total porosity. The first parameter measured was the area: pores below the
threshold of 4 μm2 were considered negligible or just image artifacts, thus were excluded from the following
analyses. Then, assuming each pore a perfect circle, the equivalent diameter deq, was calculated from its area
A as

4⋅𝐴
𝑑𝑒𝑞 = √ (4.7)
𝜋

The second parameter measured was the maximum Feret diameter was measured. Its lowest value found
was 2.7 μm, corresponding to an area of 4 μm2 and will be discussed later.
The largest pore of all was found in sample #6 and had an area of 2629 μm2, corresponding to an equivalent
diameter of 57.9 μm. Its maximum Feret diameter was 69.1 μm.
24 diameter classes were defined at a step size of 2.5 μm, up to 70 μm. For each sample, the pores were
divided in these classes. The numerical frequency expressed how many pores belonged to each class. The

56
results were expressed as percentages of the total number of pores detected in the sample, because, trivially,
the total number depends on how many images were taken.
As reported in Table 4.8, for every sample, more than 80% of the detected pores had an equivalent diameter
below 5 μm and up to 95% of the total number had an equivalent diameter below 7.5 μm2. Also, the most of
classes from 20 μm to 70 μm was empty. This means that larger pores were extremely low in number. To
facilitate the readability of the results the empty classes were excluded from the following tables.
Table 4.8 Numerical frequency of equivalent diameters, divided into diameter classes

Sample ID
Class interval [μm]
1 2 3 4 5 6 7 8
0 - 2.5 14.7% 19.2% 19.2% 19.0% 13.2% 20.0% 23.1% 17.4%
2.5 - 5 73.2% 69.2% 73.3% 61.6% 72.2% 66.3% 64.8% 63.9%
5 - 7.5 9.7% 9.0% 5.6% 11.9% 9.6% 8.0% 9.6% 13.9%
7.5 - 10 1.3% 1.8% 1.1% 4.5% 2.9% 4.0% 2.1%
10 - 12.5 0.7% 0.3% 0.7% 3.0% 1.5% 3.5%
12.5 - 15 0.3% 0.3% 0.3% 0.6%
15 - 17.5 0.3% 0.6% 0.7%
17.5 - 20 0.1%
20 - 22.5 0.1% 0.3%
30 - 32.5 0.4%
40 - 42.5 0.7%
57.5 - 60 0.6%

This approach, however, was not enough satisfactory because the importance of the bigger pores, even if
they were only few, was supposed to be much more than few percentage points. For this reason, a diameter-
normalization was carried out in the following way.
Each i-th class contained ni elements. The weight of each class wti was calculated summing the diameters of
all its pores.
𝑖 𝑛
𝑤𝑡𝑖 = ∑𝑘=1 𝑑𝑘 (4.8)

Finally, the influence of the i-th class (Xi) was calculated as the ratio between the corresponding weight wti
and the sum of the weights of the 24 classes.

𝑋𝑖 = 𝑤𝑡𝑖 /(∑24
ℎ=1 𝑤𝑡ℎ ) (4.9)
As reported in Table 4.9, after this calculation, the few large pores could cover up an important percentage
of total porosity area: 59% in sample #6 and 48% in #8. Consequently, these samples showed lower
percentages in the smaller classes than the other samples.

57
Table 4.9 Diameter-normalized frequency of equivalent diameters, divided into diameter classes

Sample ID
Class interval [μm] 1 2 3 4 5 6 7 8
0 - 2.5 9.7% 13.1% 13.5% 11.5% 8.3% 11.9% 15.9% 9.9%
2.5 - 5 67.5% 65.0% 70.9% 53.0% 63.3% 54.9% 59.5% 51.3%
5 - 7.5 16.0% 14.7% 9.7% 17.6% 14.8% 11.8% 16.3% 20.3%
7.5 - 10 2.8% 4.3% 3.1% 9.8% 6.5% 9.0% 5.1%
10 - 12.5 2.2% 0.7% 2.1% 8.1% 4.3% 9.3%
12.5 - 15 1.1% 1.1% 1.1% 1.8%
15 - 17.5 1.1% 2.2% 2.5%
17.5 - 20 0.7%
20 - 22.5 1.7%
22.5 - 25 0.8%
30 - 32.5 3.2%
40 - 42.5 6.7%
57.5 - 60 8.3%

Sample #3 had a good porosity distribution because no excessively big pores were found and so the highest
contribution is given by very small pores (below 5 μm). Sample #7, on the other hand, started to show some
issues because, even if it did not have pores in the range 10-30 μm, only one pore gave a not negligible
contribution of 3.2%. Sample #6 was the worst because more than the 10% of its total porosity was
concentrated in a very few and very large pores, which, under the mechanical properties point of view, might
be the most detrimental [62] because they might reduce the loaded cross-section area [55]. However, as
mentioned in Chapter 1.3.2, samples built via DED technology shows very small pores and low average
porosity values. Kleszczynski et al. [63] stated that only huge defects (> 1mm long) have negative effects on
yield strength of 17-4 PH samples. Such large defects are typically caused by other AM techniques, such as L-
PBF. Thus, the pores found in this thesis were too small to represent an issue.
Moreover, Biswal et al. [55] stated that the shape of voids is more meaningful than their size and the
elongated ones could be particularly detrimental to mechanical properties. For this reason, a deeper
investigation was conducted on the shape of the pores measuring the Feret’s diameters. Then, the results
were compared to the equivalent diameters previously calculated.
So, a similar statistical analysis was conducted on the Feret diameters maintaining the same diameter classes.
The numerical distributions are provided in Table 4.10, while the diameter-normalized ones in Table 4.11.

58
Table 4.10 Numerical frequency of maximum Feret diameters, divided into diameter classes

Sample ID
Class interval [μm]
1 2 3 4 5 6 7 8
2.5 - 5 65.4% 67.7% 67.9% 59.0% 64.9% 73.1% 72.2% 66.0%
5 - 7.5 20.9% 23.1% 21.2% 24.6% 21.6% 18.9% 18.9% 17.4%
7.5 - 10 8.8% 7.4% 6.8% 7.5% 8.2% 3.4% 6.4% 10.4%
10 - 12.5 1.8% 0.5% 1.6% 5.2% 3.2% 2.3% 1.4% 2.8%
12.5 - 15 1.8% 0.3% 0.7% 1.9% 1.2% 1.1% 0.7% 0.7%
15 - 17.5 0.4% 0.8% 0.7% 1.1% 0.3% 0.7%
17.5 - 20 0.1% 0.2% 0.7% 0.3% 1.4%
20 - 22.5 0.3% 0.3% 0.6% 0.6%
22.5 - 25 0.1% 0.3%
25 - 27.5 0.3% 0.2%
37.5 - 40 0.4%
40 - 42.5 0.1% 0.7%
67.5 - 70 0.6%

Table 4.11 Diameter-normalized frequency of maximum Feret diameters, divided into diameter classes

Sample ID
Class interval [μm]
1 2 3 4 5 6 7 8
2.5 - 5 48.0% 52.2% 50.8% 39.8% 48.0% 53.3% 55.5% 45.4%
5 - 7.5 23.7% 29.2% 25.3% 27.4% 25.2% 22.4% 24.2% 19.1%
7.5 - 10 14.3% 13.1% 11.5% 11.6% 13.5% 6.1% 11.9% 15.8%
10 - 12.5 3.8% 1.2% 3.4% 10.7% 7.0% 4.9% 3.3% 5.8%
12.5 - 15 4.7% 0.7% 1.8% 4.6% 3.1% 3.1% 2.2% 1.8%
15 - 17.5 1.2% 2.5% 2.2% 3.3% 0.9% 2.0%
17.5 - 20 0.5% 0.8% 2.5% 1.0% 4.6%
20 - 22.5 1.1% 1.2% 2.5% 2.3%
22.5 - 25 0.5% 1.3%
25 - 27.5 1.4% 1.2%
37.5 - 40 3.0%
40 - 42.5 1.1% 5.4%
67.5 - 70 7.9%

The distributions were similar to the equivalent diameter ones: small diameter pores were much more
numerous, but they had a lighter impact on the total porosity.

59
In order to study the circularity of the pores, the diameters-normalized distributions of the equivalent
diameters and the Feret diameters’ ones were compared in Fig. 4.18.
Sample ID

Fig. 4.18 Diameter-normalized distributions of the Feret (a) and the equivalent (b) diameters

Generally, the peaks below 10 μm were higher and narrower in deq than in Feret. The peaks corresponding
to larger, instead, had the same shapes but were shifted on the left because they usually indicate one single
pore.
Excluding the case limit of a perfectly round pores

𝑑𝑒𝑞 < 𝑀𝑎𝑥𝐹𝑒𝑟𝑒𝑡 (4.10)

Since these two values did not differ much, it was concluded that pores had predominantly a rather regular
shape. Thus, Feret diameters found were consistent with the corresponding equivalent diameters and no
large and very elongated shaped voids were present.
By way of example, a comparison between the two diameters in the cube #1 is shown in Fig. 4.19. It was in
line with the aforementioned considerations.

60
 Fig. 4.19 Comparison between Max Feret and equivalent diameter-
normalized distributions

A deeper inspection on the shapes of the largest pores found, i.e., with an area above 100 μm2, was
performed, since they had a greater importance compared to the others. In this case, both minimum and
maximum Feret diameters were measured (Fig. 4.20) and their ratio, called Feret Ratio, was calculated.

Fig. 4.20 Representation of Min and Max Feret diameters

The ratio equals 1 for perfectly round pores and the more the shape is irregular, the lower the ratio gets. The
results are collected in Table 4.12.

61
Table 4.12 Details of pores above 100 μm2

Sample ID Area [μm2] deq [μm] Feret Max [μm] Feret Min [μm] Feret Ratio

6 2629 58 69 53 0.76
8 1261 40 42 40 0.95
7 793 32 39 30 0.76
3 429 23 27 21 0.77
5 382 22 23 22 0.95
1 280 19 40 19 0.47
2 191 16 17 16 0.94
6 181 15 21 13 0.64
8 178 15 18 15 0.84
2 175 15 22 15 0.67
5 165 15 15 14 0.93
1 159 14 15 14 0.94
1 132 13 26 12 0.46
6 123 13 13 12 0.93
4 121 12 17 10 0.61
1 120 12 13 12 0.94
5 114 12 13 12 0.89
1 110 12 21 12 0.56
5 109 12 14 12 0.88
8 109 12 16 11 0.71
1 107 12 19 13 0.68
8 107 12 18 10 0.58

The highlighted values represented the most irregular shaped pores, and they all were on the sample #1. The
picture of the largest one is reported in Fig. 4.21. The cube #1 had the highest VED, suggesting that excessive
VEDs might promote irregular pores.

62
 Fig. 4.21 The most irregular shaped and largest pore found in sample #1

According to the analysis performed, the pores found did not represent a big issue to the mechanical
properties, neither in terms of size [63], nor in terms of shape [55].
In conclusion, cube #4 had excellent average porosity (0.030%) and also a quite low amount of large pores.
On the other hand, cube #6 had a similar average value (0.034%) but showed some large pores.
Cube #4 might have one of the best process parameters, but a practical consideration was done comparing
the previously mentioned build rates [64]. The values ranged between 24 and 83 cm3/h and were reported
in Fig. 4.22.

Fig. 4.22 Build rates of the samples

Not necessarily high build rates resulted in high porosity, for example cube #1 had the lowest build rate and
the highest porosity. Cube #4 had a rather low build rate and so a compromise between porosity and build
rate was necessary: cubes #5 and #8 had the same process parameters and showed a slightly higher porosity
but a 40% higher build rate, for this reason, this might be considered the best set of parameters. Finally, the
cube #7 was in an intermediate situation between #4 and #5 both in terms of porosity than build rate.

63
After these considerations, the samples were again observed under the optical microscope, focusing on the
research of defects. Some pores were photographed at higher magnification, i.e., 500x and a selection of
them was collected in Fig. 4.23. In some pictures, suspect oxides were present but due to their nanometric
dimensions, they were considered negligible and probably caused by the manipulation of the samples rather
than their building process.

Fig. 4.23 Porosities on cubes #2 (a), #4 (b) and #7 (c) at 500x magnification

Main defectiveness was found on the cube #3 at the interface with the base plate (Fig. 4.24). In particular,
an extremely high number of pores was present on the baseplate and many defects, about 100 μm long,
were spread along the interface line.
Such a high defectiveness is clearly unacceptable and, maybe, was caused by an excessive scanning speed
(100%) combined to an insufficient powder feed rate (70%).

Fig. 4.24 Defects and porosities at the base interface in cube #3 at 100x and 500x magnifications

64
Many other defects were found in cube #6 (Fig. 4.25). Their size and shape were compatible with powder
particles and so it was concluded that those were partially melted particles.

Fig. 4.25 Defects on cube #6 at 200x (a) and 500x (b) magnification

This cube was the only built with 100% powder flow rate. This excessive value, probably, had promoted the
presence of partially molten particles. Also, the scanning speed was 100% and this resulted in a lower
interaction time to effectively melt all particles, despite the 100% laser power.
Some other defects are reported in Fig. 4.26, after the acid etching.
a) b) c)

Fig. 4.26 Fusion defects (a), (b) and pore (c) in the sample #6 at 200x magnification after the etching

65
Samples were further observed under Scanning Electron Microscope (SEM) and few elongated pores were
found (Fig. 4.27). Their lengths were irrelevant i.e., below 10 μm, in most cases. The only exception was
found on the cube #6, with a length of around 20 μm.

a) b) c)

d) e) f)

Fig. 4.27 SEM images of sample #1 (a,b,c), sample #2 (d), sample #3 (e) and sample #6 (f)

66
 During the SEM analysis, the only noteworthy element was observed on sample #6 and it was judged as a
potential oxide Fig. 4.28. All the possible oxides observed so far had nanometric dimensions but this one was
much larger, with a diameter of 4.8 μm.

Fig. 4.28 Suspected oxide on sample #6

Sample 6
Thanks to its larger dimension, a deeper inspection was possible. An Energy Dispersive Spectroscopy (EDS)
was performed selecting two points: the centre of the suspect oxide and the surrounding metal for a
comparison. EDS is a function of SEM microscopes and offered an estimation of the chemical composition of
the selected points, exploiting the diffraction of the electrons. The results are reported in Table 4.13.
Table 4.13 Comparison of chemical composition between metal and suspect oxide

Metal Suspect Oxide

Element Atomic Weight Atomic Weight
Element
Conc. [%] Conc. [%] Conc. [%] Conc. [%]
Fe 76.34 77.68 O 50.85 27.03
Cr 16.79 15.91 Cr 15.40 26.60
Cu 2.44 2.83 Si 18.42 17.19
Ni 2.37 2.54 Fe 9.09 16.87
Si 2.05 1.05 Mn 5.51 10.06

The composition of the metal was in line with the expected values, found in the supplier’s datasheet. The
suspected oxide, instead, was characterized by a high oxygen content and also chromium and silicon. The
corresponding oxides are Cr2O3 or SiO2 but the uncertainty in this case was too high. In conclusion, on the
basis of the analysis carried out, the presence of other relatively large oxides was not detected but, however,
further researches might be necessary.

67
4.3.2 Microstructure

After the etching, the samples were once again observed under the optical microscope to study their
microstructure. At 50x and 100x magnifications, the meltpool borders were visible. All the samples were built
adopting the 0-90° scanning strategy. Thus, the longitudinal and the transverse cross-sections of the scan
tracks were alternated in the z direction. The transverse cross-sections were obviously more significative
because they showed much more information. The grain growth directions were, as expected and discussed
in Chapter 2, preferentially perpendicular to the meltpool borders, as shown in Fig. 4.29.

Fig. 4.29 Examples of meltpool borders at 50x magnification in samples #7 (a) and #6 (b)

Also, it was noticed that the scan tracks of the sample #7 had a higher overlap than the ones of the sample
#6. The reason was that the hatching distance adopted during the building was respectively 71% and 100%.
Lower hatching distance promoted a higher overlapping. Besides that, the dimensions of the scan tracks were
different, too. Consistently to what observed in the SST analysis, they generally increased with laser power,
while they decreased with the scanning speed.
At 500x magnification microstructurecubewas
1 visible: martensite cubeand
7 δ-ferrite dendrites were observed,
consistently to what was suggested in many studies [37], [38], [40], [65]. A higher concentration of dendrites
was detected at the meltpool borders (Fig. 4.30).

a) b) Martensite

ides

Martensite ‐ ferrite ‐ ferrite

Fig. 4.30 Microstructure of sample 1 a) and sample 7 b)

68
In contrast to the previous optical microscope and SEM observations, in some samples, the presence of small
oxides (below 5 μm in diameter) was observed in few different spots. In all probability, they were caused by
the etching itself, as long as they were not present before.

As already mentioned, within the same samples, rather different microstructure appeared in different areas
of the meltpools. In particular the δ-ferrite dendrites number, shape, and size (Fig. 4.31)

a) b) c)

Fig. 4.31 Sample #7 at the base (a), middle (b) and top (c) portions

Considering these internal variations, every sample showed a similar microstructure. Laser power and
scanning speed did not change enough to influence the microstructure. Also, such a great variety did not
allow a characterization based on the building process parameters of the cubes.
Cubo 7
Base middle top

69
4.3.3 Hardness

All the hardness measurements performed are reported in Table 4.14 .

Table 4.14 Hardness measurements

Position Standard Total Standard

Sample ID Position Measured Hardness [HV] Average Deviation Average Deviation
[HV] [HV] [HV] [HV]
Top 405 413 407 408 4
1 Centre 403 398 405 402 4 406 5
Bottom 413 402 406 407 6
Top 394 385 399 393 7
2 Centre 397 398 387 394 6 400 12
Bottom 416 407 418 414 6
Top 380 395 389 388 8
3 Centre 386 389 389 388 2 392 7
Bottom 394 404 398 399 5
Top 419 405 422 415 9
4 Centre 420 415 419 418 3 418 6
Bottom 423 420 422 422 2
Top 417 405 419 414 8
5 Centre 401 408 412 407 6 413 7
Bottom 420 415 422 419 4
Top 389 385 398 391 7
6 Centre 399 410 393 401 9 399 11
Bottom 416 392 412 407 13
Top 407 408 401 405 4
7 Centre 408 416 415 413 4 415 11
Bottom 420 433 430 428 7
Top 390 398 405 398 8
8 Centre 406 400 400 402 4 403 7
Bottom 410 413 409 411 2

The lowest value reported was 388 HV in sample #3 and the highest one was 428 HV in sample #7. In most
of the samples, hardness decreased in the z direction in fact, it was higher at the bottom and lower at the
top. An exception is represented by cube #1, where this trend was not observed (Fig. 4.32).

70
 Fig. 4.32 Hardness values at different z positions

The reason of this this might be caused by the typical characteristics of heat transfer during the DED process.
The bottom layers had slower cooling rate because of the many new layers deposited. These, in fact, helped
the layers below to maintain a rather high temperature, vaguely resembling a precipitation hardening heat
treatment. Top layers, instead, did not benefit of this effect and showed a lower hardness in most of the
samples.
Then, the general average values were studied in function of the Volumetric Energy Density parameter (Fig.
4.33). Even if results were rather dispersed, a slight hardness increase with VED increase seemed to be
present, except for cube #1. The particularly lower powder feed rate of this cube might have influenced it
characteristics.

Fig. 4.33 Average cube hardness in general a) and divided by power b), scanning speed c) and powder flow d)

71
The effects of other parameters, such as laser power, scanning speed and powder flow were further
investigated. Once again, due to the low number of samples available and the high number of parameters
varying among them, no plausible trends were found.
The samples with highest hardness were #4 and #7 while the least hard was #6. A higher hardness value is
not necessarily a positive characteristic because, generally, even if it is correlated to a higher tensile strength
[45], it also involves lower elongation and ductility values [32]. Thus, the hardness goal depends on the final
application of the component.
Additionally, the obtained hardness values were compared with other values found in the literature for the
same material. The considered values referred to samples not only produced via AM techniques such as DED
and L-PBF, but also via other conventional processes like press & sinter and wrought (Fig. 4.34).

Fig. 4.34 Hardness comparison of as-built 17-4 PH samples produced

via different techniques [37], [38], [66-69]

The samples studied in the thesis were in the as-built condition, i.e., without any heat-treatments. In this
case, the obtained hardness values were in line with the ones found in the samples produced by DED, both
around 400 HV. The L-PBF technique, instead, achieved rather low hardness values, around 330 HV. The
conventional methods reached values around 350 HV. So, the samples produced by Prima Additive were
amongst the hardest ones.
However, as extensively discussed in Chapter 2, the 17-4 PH stainless steel lends itself well to heat treatments
that can increase its hardness. For this reason, different heat treatments are commonly performed, so they
are contemplated in the majority of studies and applications. In almost every case, the solution annealing
heat treatment (1100°C for 1 hour) caused a decrease in the hardness, up to 10%. On the contrary, if the
solution annealing was followed by an ageing heat treatment (480°C for 1 hour), a considerable increase in
the hardness was always recorded, up to 50% more than the as-built condition. Thus, if a high hardness is
desired, a solution annealed followed by an ageing is the recommended heat treatment. In fact, the hardness
of the samples might reach the values of 520-530 HV, as suggested in studies [37], [38].

Summing up all the analysis on the cubes, the preliminary observations concluded that an insufficient LED,
when associated to high hatching distance and z-step, caused an unacceptable geometrical instability. The
samples #3 and #6 were, indeed, the most critical. On the contrary, a too high LED (above 80%) caused a poor
surface finish, with many adhered particles (samples #1 and, at lower extent, #4). This might be problematic
72
in case of objects with small geometries because an invasive post-processing treatment would not be
feasible. The most promising set of parameters were #2, #5 (that is the same of #8) and #7. The overall
porosities of the samples were very low, even if the worst samples (#3 and #6) exhibited large and numerous
pores and defects. These two samples also showed the lowest hardness. The most promising samples had a
very similar porosity and hardness, so, another parameter considered: the build rate. The samples #5 and #8
had a build rate higher than the #7 and #2 of almost 17% and 12%, respectively. This, for sure, represents a
great advantage for the manufacturer in terms of both time and costs.

73
5 Conclusions
In this thesis project, different samples in 17-4 PH realized by Directed Energy Deposition were examined,
investigating how process parameters affected their geometrical and structural characteristics. First of all,
the powder was analysed, determining its particle size distribution and the relative D10, D50 and D90
parameters. Morphology of particles was mostly round, even if many defects were present, such as satelites.
Flowability and density were calculated, finding excellent results.
Single Scan Tracks were observed on-top and in cross-section views. The former highlighted some issues in
samples produced at minumum and maximum laser power. Cross-section analysis consisted in the evaluation
of the geometrical characteristics of meltpools (width, growth, depth, G/D) and also in calculation of powder
efficiencies (Pe). According to G/D evaluations, the lower powder feed rate was preferrable and higher P e
values were achieved by high power and medium-high scanning speed samples.
Massive Cubes were e tensively analysed. Vicker’s microhardness was evaluated at different heights within
the cubes, discovering a decrescent trend with the height. The average hardness were between 392 and 418
HV. Porosity was evaluated in terms of average values, obtaining vey good results, between 0.030% and
0.092%. Further researches were carried on single pore dimension and their statistical distribution: small
pores were more numerous but together convered a smaller area than the few large ones. Besides pores,
under the SEM and optical microscope, other defects were spotted, such as fusion defects and partially
melted powder particles. Finally, the martensitic and δ-ferrite microstructure was observe. The low number
of samples available and their high number of changing process parameter did not allow to find a one-to-one
correlation between characteristics and parameters. Despite that, a comparison between them allowed to
find the best set of parameters.
Moreover, future studies and researches might be performed on the best set and on very similar
combinations: for example, new samples might be built and tensile tests performed to investigate the effects
on the mechanical properties.

74
References
[1] I. Gibson, D. W. Rosen, and B. Stucker, Additive manufacturing technologies: Rapid prototyping to
direct digital manufacturing. Springer US, 2010.
[2] “http://www.esa.int/ESA_Multimedia/Images/ 014/04/Additive_Manufacturing_process.” .
[3] M. K. Niaki and F. Nonino, “The Management of Additive Manufacturing Enhancing Business Value.
Springer Series in Advanced Manufacturing,” 018. [ nline]. Available:
http://www.springer.com/series/7113.
[4] A. Salmi, P. Minetola, and A. Saboori, “Additive Manufacturing Systems and Materials,” Course Slides
- Politecnico di Torino, 2019.
[5] T. S. Srivatsan and T. S. Sudarshan, Additive Manufacturing. CRC Press, 2015.
[6] M. Srivastava, S. Rathee, S. Maheshwari, and T. K. Kundra, Additive Manufacturing. Boca Raton, FL :
CRC Press/Taylor &amp; Francis Group, 2019.: CRC Press, 2019.
[7] G. J. Peacock, “Method of making composition horseshoes,” US746143A, 190 .
[8] T. Wohlers and T. Gornet, History of Additive Manufacturing, Wohlers Reports. 2014.
[9] C. Hull, “Apparatus for production of three-dimensional objects by stereolithography,” US Patent
4,575,330, 1984.
[10] “https://www.3dsystems.com/our-story.” .
[11] R. Larson, “Method and device for producing three-dimensional bodies,” US Patent 578656 A, 1993.
[12] G. Kim et al., “Three-Dimensional Printing: Basic Principles and Applications in Medicine and
Radiology,” Korean Journal of Radiology, vol. 17, p. 182, Feb. 2016, doi: 10.3348/kjr.2016.17.2.182.
[13] R. Sanders, L. Forsyth, and K. Philbrook, “3-D Model maker,” US Patent 5506706, 1996.
[14] L. Yang et al., “Additive Manufacturing of Metals: The Technology, Materials, Design and Production
Springer Series in Advanced Manufacturing,” 017. [ nline]. Available:
http://www.springer.com/series/7113.
[15] “https://www.ansa.it/canale_motori/notizie/componentie_tech/ 0 0/03/ 3/coronavirus-
stampanti-3d-strategiche-per-supporti-medicali_c8e6d71c-a02f-4aa4-85a0-fc5fe3547685.html.” .
[16] A. Ahrari, K. Deb, S. Mohanty, and J. H. Hattel, “Multi-Objective Optimization of Cellular Scanning
Strategy in Selective Laser Melting,” 017.
[17] “https://www.lastampa.it/cronaca/ 0 0/03/14/news/coronavirus-mancano-le-valvole-in-
rianimazione-facciamole-con-la-stampante-3d-1.38593309.”
[18] “https://clustercollaboration.eu/open-calls/amable-calls-ideas-additive-manufacturing-covid-19.” .
[19] S. Tornincasa and E. Chirone, Disegno tecnico industriale, vol. 1. Edizioni Il Capitello, 2014.
[20] “https://www.loc.gov/preservation/digital/formats/fdd/fdd000504.shtml.”.

75
[21] S. L. Sing, C. F. Tey, J. H. K. Tan, S. Huang, and W. Y. Yeong, “ - 3D printing of metals in rapid
prototyping of biomaterials: Techniques in additive manufacturing,” in Rapid Prototyping of
Biomaterials (Second Edition), Second Edition., R. Narayan, Ed. Woodhead Publishing, 2020, pp. 17–
40.
[22] F. Wirth, S. Arpagaus, and K. Wegener, “Analysis of melt pool dynamics in laser cladding and direct
metal deposition by automated high-speed camera image evaluation,” Additive Manufacturing, vol.
21, pp. 369–382, May 2018, doi: 10.1016/j.addma.2018.03.025.
[23] Z. Yu et al., “Effect of laser remelting processing on microstructure and mechanical properties of 17-
4 PH stainless steel during laser direct metal deposition,” Journal of Materials Processing Technology,
vol. 284, Oct. 2020, doi: 10.1016/j.jmatprotec.2020.116738.
[24 B. Abiodun, Sisa Pityana, M olisi Brendon Shongwe, and Titilayo Akinlabi Esther, “Effect of Scanning
Speed on Laser Deposited 17-4PH Stainless Steel,” 017.
[25] “https://www.primaadditive.com/the-laser-metal-deposition-process/.” .
[26] “https://www.forgemag.com/articles/84865-the-use-of-direct-metal-deposition-dmd-additive-
manufacturing-on-forging-dies.” .
[27] “https://www.spilasers.com/application-additive-manufacturing/what-is-laser-metal-deposition
[28] D. C. Hofmann et al., “Developing Gradient Metal Alloys through Radial Deposition Additive
Manufacturing,” Scientific Reports, vol. 4, no. 1, May 2015, doi: 10.1038/srep05357.
[29] A. Aversa et al., “The role of Directed Energy Deposition atmosphere mode on the microstructure
and mechanical properties of 316L samples,” Additive Manufacturing, vol. 34, Aug. 2020, doi:
10.1016/j.addma.2020.101274.
[30] A. Popovich and V. Sufiiarov, “Metal Powder Additive Manufacturing,” in New Trends in 3D Printing,
InTech, 2016.
[31] M. Boniardi and A. Casaroli, “Stainless steels Gruppo Lucefin Research & Development,” 014.
[32] S. Barella and A. Gruttadauria, Metallurgia e Materiali Non Metallici, Esculapio. 2016.
[33] “https://www.welderdestiny.com/dissimilar-metals-welding.html.” .
[34] ASM International, “Alloy Digest Sourcebook: Stainless Steels,” 000. [ nline]. Available:
www.asminternational.org.
[35] H. K. Rafi, D. Pal, N. Patil, T. L. Starr, and B. E. Stucker, “Microstructure and Mechanical Behavior of
17-4 Precipitation Hardenable Steel Processed by Selective Laser Melting,” Journal of Materials
Engineering and Performance, vol. 23, no. 12, pp. 4421–4428, Dec. 2014, doi: 10.1007/s11665-014-
1226-y.
[36] A. Yadollahi, N. Shamsaei, S. M. Thompson, A. Elwany, and L. Bian, “Mechanical and microstructural
properties of selective laser melted 17-4 ph stainless steel,” in ASME International Mechanical
Engineering Congress and Exposition, Proceedings (IMECE), 2015, vol. 2A-2015, doi:
10.1115/IMECE2015-52362.
[37] I. Mathoho, E. T. Akinlabi, N. Arthur, M. Tlotleng, and N. W. Makoana, “The effects of LENS Process
parameters on the behaviour of 17-4 PH stainless steel.”
76
[38] I. Mathoho, E. T. Akinlabi, N. Arthur, and M. Tlotleng, “Impact of DED process parameters on the
metallurgical characteristics of 17-4 PH SS deposited using DED,” CIRP Journal of Manufacturing
Science and Technology, vol. 31, pp. 450–458, Nov. 2020, doi: 10.1016/j.cirpj.2020.07.007.
[39] S. Cheruvathur, E. A. Lass, and C. E. Campbell, “Additive Manufacturing of 17-4 PH Stainless Steel:
Post-processing Heat Treatment to Achieve Uniform Reproducible Microstructure,” JOM, vol. 68, no.
3, pp. 930–942, 2016, doi: 10.1007/s11837-015-1754-4.
[40] W. J. h, Y. Son, S. Y. Cho, S. W. Yang, G. Y. Shin, and D. S. Shim, “Solution annealing and precipitation
hardening effect on the mechanical properties of 630 stainless steel fabricated via laser melting
deposition,” Materials Science and Engineering A, vol. 794, Sep. 2020, doi:
10.1016/j.msea.2020.139999.
[41] A. Yadollahi, N. Shamsaei, S. Thompson, and D. Seely, “Effects of process time interval and heat
treatment on the mechanical and microstructural properties of direct laser deposited 316L stainless
steel,” Materials Science and Engineering A, Dec. 2015.
[42] A. Saboori et al., “An investigation on the effect of powder recycling on the microstructure and
mechanical properties of AISI 316L produced by Directed Energy Deposition,” Materials Science and
Engineering A, vol. 766, Oct. 2019, doi: 10.1016/j.msea.2019.138360.
[43] B. de La Batut, . Fergani, V. Brotan, M. Bambach, and M. el Mansouri, “Analytical and Numerical
Temperature Prediction in Direct Metal Deposition of Ti6Al4V,” Journal of Manufacturing and
Materials Processing, vol. 1, no. 1, p. 3, Jul. 2017, doi: 10.3390/jmmp1010003.
[44] T. DebRoy et al., “Additive manufacturing of metallic components – Process, structure and
properties,” Progress in Materials Science, vol. 92. Elsevier Ltd, pp. 112–224, Mar. 01, 2018, doi:
10.1016/j.pmatsci.2017.10.001.
[45] E. J. Pavlina and C. J. van Tyne, “Correlation of Yield Strength and Tensile Strength with Hardness for
Steels,” Journal of Materials Engineering and Performance, vol. 17, no. 6, Dec. 2008, doi:
10.1007/s11665-008-9225-5.
[46] P. Guo, B. Zou, C. Huang, and H. Gao, “Study on microstructure, mechanical properties and
machinability of efficiently additive manufactured AISI 316L stainless steel by high-power direct laser
deposition,” Journal of Materials Processing Technology, vol. 240, pp. 12–22, Feb. 2017, doi:
10.1016/j.jmatprotec.2016.09.005.
[47] F. Ahmed et al., “Study of powder recycling and its effect on printed parts during laser powder-bed
fusion of 17-4 PH stainless steel,” Journal of Materials Processing Technology, vol. 278, Apr. 2020, doi:
10.1016/j.jmatprotec.2019.116522.
[48] D. Pacucci, “ ttimizzazione del processo Direct Energy Deposition dell’acciaio H13,” 019.
[49] “http://www.fsg-international.com/cnc-shop-machine-tool-fiber-laserjan-3-fibermak-momentum-
sm-40003000-x-1500.html.”.
[50] A. Lucenti, “Effetti dei trattamenti termici sulle microstrutture e sulle proprietà meccaniche
dell’acciaio AISI 316L da Deposizione Diretta,” 019.
[51] “www.phenom-world.com.”.

77
[52] “http://www.labulk.com/astmb 1 -free-flowing-metal-powders-apparent-density-hall-flowmeter-
funnel/.”
[53] “https://www.leica-microsystems.com/products/.” .
[54] “https://sites.google.com/site/gobanengineeringnotes/materials-testing/hardness-testing/vickers-
hardness-test?tmpl=% Fsystem% Fapp% Ftemplates% Fprint% F&showPrintDialog=1.”
[55] R. Biswal, A. K. Syed, and X. Zhang, “Assessment of the effect of isolated porosity defects on the
fatigue performance of additive manufactured titanium alloy,” Additive Manufacturing, vol. 23, pp.
433–442, Oct. 2018, doi: 10.1016/j.addma.2018.08.024.
[56] A. Carrozza et al., “Single Scans of Ti-6Al-4V by Directed Energy Deposition: A Cost and Time Effective
Methodology to Assess the Proper Process Window,” Metals and Materials International, Jan. 2021,
doi: 10.1007/s12540-020-00930-3.
[57] D. McGlinchey, “Bulk Solids Handling: Equipment Selection and peration,” 008.
[58] L. Peng et al., “Direct laser fabrication of nickel alloy samples,” International Journal of Machine Tools
and Manufacture, vol. 45, no. 11, pp. 1288–1294, Sep. 2005, doi: 10.1016/j.ijmachtools.2005.01.014.
[59] E. Capello and B. Previtali, “The influence of operator skills, process parameters and materials on clad
shape in repair using laser cladding by wire,” Journal of Materials Processing Technology, vol. 174, no.
1–3, pp. 223–232, May 2006, doi: 10.1016/j.jmatprotec.2006.01.005.
[60] U. de liveira, V. celík, and J. T. M. de Hosson, “Analysis of coa ial laser cladding processing
conditions,” Surface and Coatings Technology, vol. 197, no. 2–3, pp. 127–136, Jul. 2005, doi:
10.1016/j.surfcoat.2004.06.029.
[61] “https://www.olympus-ims.com/en/insight/breaking-down-the-technical-cleanliness-workflow-
part- /.” .
[62] P. Wang, H. Lei, X. Zhu, H. Chen, and D. Fang, “Influence of manufacturing geometric defects on the
mechanical properties of AlSi10Mg alloy fabricated by selective laser melting,” Journal of Alloys and
Compounds, vol. 789, pp. 852–859, Jun. 2019, doi: 10.1016/j.jallcom.2019.03.135.
[63] S. Kleszczynski and A. Elspaß, “Influence of isolated structural defects on the static mechanical
properties of PBF-LB/M components,” in Procedia CIRP, 2020, vol. 94, pp. 188–193, doi:
10.1016/j.procir.2020.09.036.
[64] D. Herzog, K. Bartsch, and B. Bossen, “Productivity optimization of laser powder bed fusion by hot
isostatic pressing,” Additive Manufacturing, vol. 36, Dec. 2020, doi: 10.1016/j.addma.2020.101494.
[65] A. A. Adeyemi, E. Akinlabi, R. M. Mahamood, K. . Sanusi, S. Pityana, and M. Tlotleng, “Influence of
laser power on microstructure of laser metal deposited 17-4 ph stainless steel,” IOP Conference
Series: Materials Science and Engineering, vol. 225, p. 012028, Aug. 2017, doi: 10.1088/1757-
899x/225/1/012028.
[66] LeBrun, T.; Nakamoto, T.; Horikawa, K.; Kobayashi, H. “Effect of retained austenite on subsequent
thermal processing and resultant mechanical properties of selective laser melted 17–4 PH stainless
steel.” Mater. Des. 2015, 81, 44–53.

78
[67] Irrinki, H.; Jangam, J.S.D.; Pasebani, S.; Badwe, S.; Stitzel, J.; Kate, K.; Gulsoy, O.; Atre, S.V. E
ects of particle characteristics on the microstructure and mechanical properties of 17–4 PH stainless
steel fabricated by laser-powder bed fusion. Powder Technol. 2018, 331, 192–203.
[68] P. K. Samal, N. N. North, A. Hoganas, I. Hauer, and H. Ab, “Properties of 17 – 4 PH stainless steel
produced via press and sinter route.”
[69] Grupo Lucefin. (2018). Precipitation Hardening Stainless Steel. http://www.lucefin.com/wp-
content/files_mf/1.4542pha63062.pdf
[70] ASTM B 13−17 Standard test methods for flow rate of metal powders using the hall flowmeter funnel,
ASTM Int.
[71] ASTM B212−17 Standard Test Method for Apparent Density of Free-Flowing Metal Powders Using the
Hall Flowmeter Funnel, ASTM Int.
[72] D. R. Eo, S. H. Park, and J. W. Cho, “Controlling inclusion evolution behavior by adjusting flow rate of
shielding gas during direct energy deposition of AISI 316 L,” Additive Manufacturing, vol. 33, May
2020, doi: 10.1016/j.addma.2020.101119.
[73] H. Gu, H. Gong, D. Pal, K. Rafi, T. Starr, B. Stucker, Influences of energy density on porosity and
microstructure of selective laser melted 17-4PH stainless steel, 2013 Solid Freeform Fabrication
Symposium, 2013 474-89.
[74] Z. E. Tan, J. H. L. Pang, J. Kaminski, and H. Pepin, “Characterisation of porosity, density, and
microstructure of directed energy deposited stainless steel AISI 316L,” Additive Manufacturing, vol.
25, pp. 286–296, Jan. 2019, doi: 10.1016/j.addma.2018.11.014.

79