**1. Introduction**

Additive manufacturing (AM), also popularly known as 3D printing, is at the frontier of development for manufacturing diverse parts and has also been referred to as the third industrial revolution [1–3]. AM has an advantage over other conventional manufacturing techniques, making it possible to print complex shapes without the need for several conventional processing steps, such as expensive tooling, dies, or casting molds [3–5]. There are several reviews on AM covering different aspects from process dynamics to post-processing [6–10]. This paper focuses on one such type of AM process, popularly known as directed energy deposition (DED), (also more specifically as laser engineered net shaping (LENS™), blown powder additive manufacturing, laser metal deposition system, and directed laser deposition), which has attracted significant attention due to its ability to print metals and potentially any metal-alloy system, notably functionally gradient materials [11,12]. Another important application utilizing DED is the remanufacturing or repairing of a component to increase its lifespan and hence reduce environmental impact [12,13].

DED systems have a concentrated energy source and a stream of raw material, both intersecting at a common focal point, generally in the presence of an inert shield gas. The energy density generated at a particular point melts the raw materials in and around that spot, giving rise to melt pool formation. There are various types of DED systems, which include (but are not limited to): Powder-feed and wire-feed based DED (on the basis of the type of feedstock), melt based DED, and kinetic energy based DED (on the basis of the type of energy source). Melt based DED could be further subclassified as laser based DED, electron-beam based DED, plasma based DED, and electric arc based DED. The powder based DED system has been studied extensively in the literature and is the most commonly used metal DED technique. It predominantly uses a laser beam as the heat source. Wire based DED processes provide a lower resolution as compared to laser-beam powder based processes, but have a higher deposition rate and the ability to build larger structures [14,15]. They generally use an electron-beam, plasma, or electric arc as the heat source. Electron-beam based AM (EBAM), which has a high energy focused electron beam in vacuum, can fuse almost any metal. EBAM is commercialized by Sciaky, Inc. and it is mainly used for manufacturing near net shape parts [16]. Plasma based AM uses a controlled plasma source to melt the metal particles. This is a relatively new AM technology and is commercially being used by Norsk Titanium to build mainly titanium parts [17]. Electric arc based DED melts the wire feed to deposit the layers. Emerging technology, like metal big area additive manufacturing (mBAAM) [14], takes advantage of the principle of electric arc welding to print big parts. Kinetic energy based DED systems, often referred to as Cold Spray, use a converging-diverging nozzle to accelerate micron sized particles to supersonic velocities [18]. Beyond a critical impact velocity, micron sized particles adhesively bond to the substrate and build up material in the form of a coating as well free standing bulk components [19]. Figure 1 summarizes the di fferent DED categories in the form of a flowchart.

**Figure 1.** Classification of Directed Energy Deposition (DED) systems.

This paper will focus on the powder based DED systems in detail. It covers sections on metal and alloy systems, emphasizing the potential of DED, modelling e fforts and process variables (surface tension, Marangoni e ffect, dimensionless numbers, energy distribution in a DED process, process-microstructure relationships, dilution), common defects (porosity, solute segregation and changes in chemical compositions, printability of alloys), mechanical properties (tensile strength, hardness, fatigue, residual stress), DED process control and monitoring, determination of optimal processing parameters by establishing process maps and the regions where high amount of defects are expected, application, and emerging technologies (DED of metal parts in biomedical applications, welding and cladding, repair, bulk combinatorial alloy design, construction materials, and hybrid AM). The paper will conclude with an overview of possible future perspectives of the field.

#### *Metal and Alloy Systems*

DED has been used to print mainly functionally graded materials, metal-matrix composites, and coatings. Each system was developed for a targeted application (e.g., enhancing biocompatibility, improving oxidation resistance, mechanical, and tribological properties, interfacial strength, etc.). Table 1 summarizes selected material systems with an emphasis on the potential applications.


**Table 1.** Selected studies showing DED as an emerging method to print complex metal and alloy systems.


#### **Table 1.** *Cont.*

\* This table shows potential for printing diverse combinations of materials with DED, rather than looking at all applications or functionalities in grea<sup>t</sup> detail.

#### **2. DED Process Variables and Modelling E** ff**orts**

#### *2.1. Overview of Powder-Fed DED Process Physics and Thermal History*

DED is a non-equilibrium processing technique, which has very fast cooling rates, often on the order of 10<sup>3</sup> to 10<sup>5</sup> K/s [55,56] for laser and electron beam energy sources. Major process parameters for laser based DED include: Laser power, laser beam spot size, powder or wire feed rate, scanning speed, carrier gas flow rate, clad angle, feedstock properties, and layer dimensions. Therefore, a diverse set of processing parameters coupled with the complex transport phenomena, including conduction of heat into the substrate, convection due to Marangoni e ffects, and radiation accompanied by the shield gas, lead to a di fficulty in understanding the e ffect of these individual process parameters on the overall DED process. Figure 2 schematically shows the complex thermal history during the multi-layer DED process, and the trend of an increasing peak temperature with an increasing layer number due to the accumulation of heat in the system [57].

Modelling e fforts are beneficial to complement experimental data. The model should be close to the real DED system, taking into account the transient temperature and heat flow, complex transport phenomena, heating and cooling cycles, solidification rate, etc. These in turn give important information about the microstructure, defects, texture, and mechanical properties [57]. Table 2 summarizes some important modelling efforts of the DED process in the literature.

**Figure 2.** Schematic showing the thermal cycles for three consecutive layers during DED, and thecorrespondingpeaktemperaturesforeachlayer.

**Table 2.** Selected studies on the modelling efforts of various DED processes.


#### *2.2. Surface Tension and Marangoni E*ff*ect*

The Marangoni effect is a convective heat transfer phenomenon, which affects the melt pool flow dynamics and indirectly contributes towards porosity, which is a major concern in DED processed materials. In 1982, Heiple and Roper [71] postulated the theory that Marangoni forces generated due to the differences in surface tension and temperatures along the melt pool lead to more spattering and circulation of the liquid melt pool. The movement is mainly from regions of high surface tension to low surface tension (γ), finally leading to variable melt pool penetration. The strength of the Marangoni flow for any DED process can be determined through the dimensionless Marangoni number (*M*a) [72], as provided in Equation (1):

$$M\_{\mathbf{a}} = \frac{\mathbf{d}\mathbf{y}}{\mathbf{d}T} \frac{\mathbf{d}T}{\mathbf{d}x} \frac{L^2}{\mathbf{\eta}\alpha} \tag{1}$$

where γ is the surface tension, d*T*/d*x* is the temperature gradient, α is the thermal diffusivity, *L* is the characteristic length, and η is the viscosity of the melt pool. The surface tension gradient (the slope of the graph) qualitatively governs the melt pool movement. Figure 3 schematically shows how the variation of the surface tension with temperature affects the melt pool geometry. Figure 3a shows how the melt pool length is small due to a negative surface tension gradient and signifies bulk turbulence flow in the melt pool. Figure 3b shows how the melt pool depth increases with a positive gradient of surface tension and surface turbulence occurs in the melt pool, which could also potentially trap undesired oxides in the bulk. Figure 3c shows the transition from a positive to negative surface tension gradient at a certain temperature, *T*o. This transition also indicates a melt pool flow transition from surface turbulence to bulk turbulence. The surface tension and therefore the internal melt pool flow could be controlled, to a certain extent, using surface active elements. For example, in an Fe system, changes in the concentration of the surface-active elements, like sulphur and oxygen, were shown to modify the internal melt pool flow [72,73].

**Figure 3.** Schematic of Marangoni effect using the Heiple–Roper theory of weld pool geometry, depending on the surface tension and temperature of the melt pool, as applicable for DED systems. (**a**) Melt pool geometry when the surface tension gradient is negative, (**b**) melt pool geometry when the surface tension gradient is positive, and (**c**) melt pool geometry when the surface tension gradient shifts from positive to negative.

Besides the Marangoni flow force, other forces, like aerodynamic drag (outward drag forces caused by the plume formed above the melt pool), buoyancy (upward movements of the melt pool due to density changes caused by thermal gradients inside the melt pool), electromagnetic, and Lorentz forces (forces due to electric and magnetic fields generated by the source), may also be present during the DED process [72].

#### *2.3. Dimensionless Numbers*

Other than the Marangoni number, there are several other dimensionless numbers which enable capturing of the accumulative effect of various process parameters [74]. While these dimensionless numbers are not specifically developed for DED, they can capture the DED process variable relationships very well. Table 3 summarizes three such dimensionless numbers for laser based DED systems (these can be extended to other heat sources as well).


**Table 3.** Dimensionless numbers and their definitions.

\* Characteristic length refers to either the thickness of the layers, the melt pool width or depth, and the laser beam spot size; characteristic velocity is considered to be the maximum velocity of the molten metal; characteristic time is defined as the length of the melt pool divided by the scanning speed.

#### *2.4. Energy Distribution in a DED System*

During DED, there should be conservation of mass, momentum, and energy, like any other physical system. Therefore, keeping track of how the initial heat source energy is distributed during the process will be beneficial in further improving the DED process to reduce energy loss and maximize energy for melting powders. Calorimetric measurements of the energy absorbed by the substrate (*Q*ABS), energy absorbed by the powder (*Q*DEP), energy reflected by the substrate (*Q*REF), and energy lost by the powder (*Q*LOST) due to evaporation and lack of fusion for DED of Ti-6Al-4V and Inconel were performed in the literature [77]. The important parameter is the bulk absorption coefficient (β), which gives the ratio of the energy for a particular component of energy (*Q*ABS, *Q*DEP, *Q*REF, or *Q*LOST) with respect to the total energy supplied to the system (*Q*IN). The energy balance equation is shown in Equation (2) and the corresponding bulk absorption coefficient, β, is presented in Equation (3) [77]:

$$Q\_{\rm IN} = Q\_{\rm ABS} + Q\_{\rm DEP} + Q\_{\rm REF} + Q\_{\rm LOST} \tag{2}$$

$$
\beta\_{\rm ABS} + \beta\_{\rmDEP} + \beta\_{\rm REF} + \beta\_{\rm LOSST} = 1 \tag{3}
$$

where βABS refers to the bulk absorption coefficient due to *Q*ABS, βDEP refers to the bulk absorption coefficient due to *Q*DEP, βREF refers to the bulk absorption coefficient due to *Q*REF, and βLOST refers to the bulk absorption coefficient due to *Q*LOST. From this study, it was experimentally proven that about 60% of the total initial energy was lost when using a laser as heat source, due to reflection by powders and also by powders not absorbing enough energy to melt.

#### *2.5. Process–Microstructure Relationship*

The energy source used during DED can be approximated as the Rosenthal solution of a moving heat source (laser, electron beam, plasma, or arc) on an infinite substrate. The microstructures obtained through DED can be predicted using the two important parameters: Thermal gradient, *G* (K/cm), and solidification front velocity (or interfacial velocity), *R* (cm/s). The relationship between *G* and *R* gives the thermal process maps [78,79]. Figure 4 shows graphs describing relationships between parameters, like *G*, *R*, arc length, undercooling, and supercooling. The mathematical relationships for the cooling rate, thermal gradient, and solidification front velocity are as follows (the reader could refer to [56] for a derivation of these equations):

$$\text{Cooding rate} : \begin{array}{l} \partial T\\ \frac{\partial}{\partial t} \end{array} \tag{4}$$

$$\text{Thermal Gradient}:\ G = |\nabla T|\tag{5}$$

$$\text{Solidification front velocity}: R = \frac{1}{G} \frac{\partial T}{\partial t} \tag{6}$$

where *T* is the temperature and *t* is the time. The melt pool circumference (MPC) length used in Figure 4 is the length of the circumference of the melt pool, and measurement starts from the bottom to the top of the melt pool in this study.

**Figure 4.** The relationships explaining the process maps for *G* vs. *R*, derived from [80]; (**a**) graph of undercooling/constitutional supercooling versus melt pool circumference (MPC) length showing the transition from low nucleation to high nucleation (columnar to equiaxed transition), (**b**) trend of the columnar to equiaxed transition from the graph of *G* versus *R*, (**c**) variation of the *G*/*R* ratio with respect to the MPC length, and (**d**) the MPC length as defined from the bottom to the top.

Some of the important trends derived from solidification front velocity and thermal gradient studies are:


To conclude, as the MPC length increases, there is an increasing trend of undercooling and constitutional supercooling, leading to more columnar-to-equiaxed transition (CET), till the *G*/*R* ratio reaches 1. Therefore, CET is dictated by the thermal history and can be engineered according to requirements [80].
