Effects of Latent Heat in Additive Manufacturing

ed
A. R. Umantsev
Fayetteville State University, Fayetteville, NC, USA

iew
September 29, 2022
ABSTRACT
We analyzed the effects of latent heat in conditions related to the additive manufacturing
technology by solving the free boundary problem of a moving heat source. The most important

ev
finding of our analysis is a possibility of a genuine (not constitutional) supercooling at the fusion
edge of the melt pool. It is measured by the structure forming number of the system, which
changes significantly compared to the latent-heat-free system. Other findings of the present
treatment are increase of the longitudinal size and decrease of the maximum temperature of

r
the melt pool due to the latent heat and the role of the surface heat losses in the process.
There are several important physical properties, which were not accounted for by the free-

er
boundary model developed in this treatment. They can be accounted for by a phase-field
model, which we plan to develop in the future. The latter may also be useful for other
technologies, like laser melting, electron (ion)-beam irradiation, and fusion welding.
pe
KEYWORDS: Additive manufacturing; Crystallization and melting of alloys; Phase transformation
kinetics; Laser treatment; Welding.
DATA STATEMENT: All the research data are provided in the article.
1. BACKGROUND
ot

Additive manufacturing (AM) is making great strides and finding applications in various
industries where there is a need to produce parts with complex geometries and functionalities,
tn

which typically would not be feasible using conventional production techniques [1]. AM
processes (laser- or electron-beam-based) use a high-power heat source to melt the material,
which resolidifies after passage of the beam to a certain distance. Quality and performance of
the parts depend critically on the microstructure of the material after the processing, which
rin

depends critically on the temperature distribution inside the melt pool and beyond.
Jaeger [2,3] and Rosenthal [4] were first to use the analytical methods to address the problem
of temperature distribution in presence of a stationary moving plane, line, or point heat source.
ep

Their solutions show that the temperature distribution exponentially decreases in front of the
moving source but drops slowly behind it. Rosenthal’s paper was particularly influential in the
area of crystal growth where it motivated researchers [5,6,7] to develop the theory of
isenthalpic crystallization when the solid/liquid front moves at constant velocity proportional to
Pr

the initial supercooling of the melt—the so-called kinetic regime of solidification.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 Rosenthal [4] applied his results to welding as an important metal treatment technology. Eagar

ed
and Tsai [8] attempted to resolve a problem of the temperature distribution and geometry of
the melt pool produced by a traveling, Gaussian-distributed heat source. Unfortunately,
presence of the latent heat was ignored in those works, which did not allow the authors to
adequately describe the melting/freezing during the processing.

iew
There have been several attempts, mostly through numerical simulations, to include the latent
heat into the analysis of the temperature distribution in conditions relevant to AM. Goldak et
al. [9] attempted to include the latent heat into the double ellipsoid model of a welding heat
source but ended up not noticing any significant effects. Criales et al. [10] studied the influence
of the material properties on the size and shape of the melt pool in conditions of laser melting

ev
and concluded that “the effect of latent heat [on the temperature distribution and melt pool
shape—A.U.] is negligible”. Peyre et al. [11], Mirkoohi et al. [12], and Proell et al. [13] included
melting, solidification, and solid-state phase change into their models of AM related processes

r
of metallic materials by using the modified specific heat, which relates it to the latent heat of
fusion. The thermal energy deposited into a control volume was absorbed in their models by

er
the latent heat and conducted through the contacting solid boundaries. The equivalent specific
heat or apparent heat-capacity (enthalpy) method models melting/freezing by a mushy phase-
change region where a liquid-phase fraction variable changes from zero to one. The authors of
these works found no significant effects due to the latent heat. Failure to find a significant
pe
effect of the latent heat is due to the equivalent specific-heat formulation, which models a
phase transformation as a homogenization of the material properties in a mushy zone. Such
formulation does not allow for adequate account of the intricate, nonlinear effects of the latent
heat.
ot

Recently, Stump et al. [14] used a computational finite volume method to consider effects of
fluid flow, dynamic material properties, surface heat losses due to radiation and vaporization,
and the latent heat on the size, shape, and solidification characteristics in the melt pool in
tn

conditions of AM. Presence of the latent heat was accounted for in their model by including
into the advective heat equation a source term proportional to the time derivative of the liquid
volume fraction. To relate the latter to the local temperature the authors used, for numerical
convenience, a sigmoidal function of obscure physical meaning. To estimate the effects of the
rin

processes and properties they analyzed the log-log relations between the thermal gradient and
interface velocity at the solidification front. Despite some limitations of the method, they found
that “most of the change in solidification conditions comes from the inclusion of latent heat.”
ep

In this publication, to account for the release and absorption of the latent heat, we use the
method of the free boundary problem, which provides a better description of the prosses than
the homogenization method. We do not consider innate rapid solidification mechanism, which
will be addressed in the future publications. We consider only continuous plane and linear heat
Pr

sources with zero spot thickness and plan to consider a problem of the point heat source in the

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 future publication. The results of this work may be useful for other technologies, like laser

ed
melting, electron (ion)-beam irradiation, and fusion welding.
2. FORMULATION
In the material (laboratory) reference frame, the temperature field T(x,y,z,t) in the solid (s) and

iew
liquid (l) phases obeys the heat equation:
𝜕𝑇𝑠⁄
𝐶𝑠⁄ 𝑙
= 𝜆𝑠⁄ ∇2 𝑇𝑠⁄ − 𝐻 (𝑇𝑠⁄ − 𝑇0 ) (1)
𝑙 𝜕𝑡 𝑙 𝑙 𝑙
where 𝐶𝑠⁄𝑙 is the specific heat at constant pressure and 𝜆𝑠⁄𝑙 is the thermal conductivity of the

ev
respective phase, T0 is the preheating temperature such that initially:

𝑇𝑠⁄ (𝑥, 𝑦, 𝑧, 0) = 𝑇0 (2)

𝑙

r
and H is the heat exchange coefficient from Newton’s law of cooling, which states that “for a
body cooling by forced convection and evaporation the rate of loss of heat is given by

er
𝐻(𝑇 − 𝑇0 )” [3]. Far from the melt pool, the material retains the preheating temperature—the
far-field boundary condition (BC):
pe
𝑇𝑠 ((𝑥, 𝑦, 𝑧) → ∞, 𝑡) = 𝑇0 (3)

If a continuous point, line, or surface source of heat of strength Q(x,y,z,t) (energy density per
unit time) operates in the system, it can be accounted for by adding the heat-source term to
the right-hand side of Eq.(1). Another way to account for the source is by the heat-balance BC:
ot

∫∂Σ 𝜆𝑠⁄𝑙 (𝐧 ∙ 𝛁𝑇)𝑑𝜎 + ∫Σ 𝑄𝑑𝜂𝑑𝜎 = 0 (4)

where  is the closed surface surrounding the domain  of infinitesimal volume, inside which
tn

the sources operate, n is the outer normal of , d is its differential surface area, and d is the
differential thickness of , see Figure 1. Equivalence between the two approaches is established
via the Divergence Theorem.
rin
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 ed
iew
r ev
er
pe
ot
tn

Figure 1. Schematics of the fusion surface and external heat source. Blue line—intersections of
the external and latent sources with the (x,y) plane; ±—the heat source or sink; red line—the
closed surfaces  of infinitesimal volume surrounding the sources; n—the outer normal of ;
rin

Ͱ—the turning points; v—stationary speed of motion of the source.

Under the influence of the heat source, the solid material will melt and resolidify after the
source passes sufficiently far. The melt pool surrounding the source will be bounded by a
ep

closed fusion surface:

̃ (𝑥, 𝑦, 𝑧, 𝑡) = 0
𝛷 (5)

on which the processes of melting and fusion (resolidification) take place, see Fig. 1.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 As known [3,5,7,15], these processes are accompanied respectively by the laten heat release or

ed
absorption with the strength of the surface source:

̃)
𝑄 = −𝐿𝑉𝑛 𝛿(𝛷 (6)

iew
where L>0 is the latent heat per unit volume at the melting temperature of the material, Vn is
the normal component of the local velocity of the fusion surface, and 𝛿(𝛷 ̃ ) is the surface delta
function. Notice that the tangential component of the local velocity of the fusion surface is not
a physically defined quantity.
There are two significant differences between the applications of the boundary conditions to

ev
the external and latent sources of heat. First, the external heat source strength is independent
of its motion while the strength of the latent heat source is proportional to its motion. Second,
the closed surface  surrounding the external heat source is singly connected while that of the
latent heat is doubly connected consisting of two singly connected surfaces surrounding the

r
source and sink, see Fig. 1. The latter can be reconnected such that each of the two singly
connected surfaces belongs to the solid or liquid phase only and that the directions of the

er
normals on the two singly connected surfaces are opposite, see Fig. 1. Then, we obtain the
celebrated heat-balance (Stefan) boundary condition on the fusion surface [3,15]:
pe
𝜆𝑠 ∇n 𝑇𝑠 − 𝜆𝑙 ∇n 𝑇𝑙 = 𝐿𝑉n (7)

where ∇n = 𝐧 ∙ 𝛁 and the outer normal n points into the solid phase everywhere. Notice that it
has opposite directions on the opposite sides of the fusion surface.
Another BC, which applies to the fusion surface, is the condition of equilibrium:
ot

𝑇𝑠 = 𝑇𝑙 = 𝑇𝑀 (8)

where TM is the melting point of the material. The first equality in (8) reflects the zeroth law of
tn

thermodynamics while the second one is manifestation of the condition of phase equilibrium—
equality of the chemical potentials—across the boundary. In this paper we do not take into
account more complicated physics of the interface, which is important for rapid solidification.
rin

The problem described by Eqs.(1-8) is known as a free-boundary problem because, unlike the
predefined position of the external heat source, position and form of the fusion surface (5) is
not known a priory and must be found in a self-consistent solution of the entire problem.
ep

Now suppose that the continuous external heat source Q(x,y,z,t) moves at constant speed v,
called the scanning speed in AM, along the straight line, which, without any loss of generality,
may be chosen as the x-axis of the laboratory reference frame. If a stationary state is
established in the system, then
Pr

Vn = 𝑣 cos 𝜑 (9)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 where  is the direction angle between the x-axis and the normal n to the fusion surface.

ed
An important part of the problem is morphological stability (that is, smoothness) of the moving
fusion surface. As known, failure to maintain it leads to microstructural heterogeneity of the
resolidified material. If the material is an alloy, the condition of the morphological stability is

iew
absence of the constitutional supercooling in the liquid phase ahead of the advancing front of
crystallization [16,17]. This is the case if the temperature gradient into the liquid phase at the
freezing portion of the fusion surface is greater than the gradient of the equilibrium
temperature of the alloy at the same point. The latter is equal to 𝑣∆𝑇𝑠⁄𝑙 ⁄𝐷 where ∆𝑇𝑠⁄𝑙 is the
equilibrium freezing range of the alloy and D is the diffusion coefficient [16,17]. For pure

ev
materials ∆𝑇𝑠⁄𝑙 = 0. Then, the condition of morphological stability of the stationary moving
fusion line may be expressed for the quantity, which we call the structure forming number:

𝜆𝑙 ∇n 𝑇𝑙 𝜆𝑙 ∆𝑇𝑠⁄
𝑙
𝐺≡− > ≥0 (10)

r
𝑣𝐶𝑙 ∆𝑇 𝐷𝐶𝑙 ∆𝑇

Here

Δ𝑇 = 𝑇𝑀 − 𝑇0 er (11)

is the initial undercooling of the material and we take into account that n is the outer normal of
pe
the fusion surface, see Fig. 1. Notice that the absolute value of the so-defined structure
forming number G is the reciprocal of the Péclet number Pe, which is defined as the ratio of the
advective and diffusive transport rates.
As known [2,3,4], if a stationary state is established it is advantageous to transform to the
ot

reference frame moving with the speed v in the direction of motion and the following
transformation of the temperature field is warranted:
tn

−𝑘𝑠⁄ 𝑢
𝑇𝑠⁄ (𝑥, 𝑦, 𝑧, 𝑡) = 𝑇0 + Δ𝑇𝑒 𝑙 Ψ𝑠⁄ (𝑢, 𝑦, 𝑧) (12)
𝑙 𝑙
where
rin

𝐶𝑠⁄ 𝑣
𝑙
𝑘𝑠⁄ = (13)
𝑙 2𝜆𝑠⁄
𝑙

is the thermal wavenumber and

ep

𝑢 = 𝑥 − 𝑣𝑡 (14)

is the transformed coordinate so that in the moving reference frame the coordinates of the
sources Q do not change with time. Then
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 [𝜕𝑢2 + 𝜕𝑦2 + 𝜕𝑧2 − 𝑔𝑠2⁄ ] Ψ𝑠⁄ = 0

ed
(15)
𝑙 𝑙
where we introduced the thermal impedance:
𝐻
𝑔𝑠2⁄ = 𝑘𝑠2⁄ +

iew
(16)
𝑙 𝑙 𝜆𝑠⁄
𝑙

which accounts for the Newtonian surface losses. Then, the fusion surface (5) turns into:

𝛷(𝑢, 𝑦, 𝑧) = 0 (17)

ev
where

∇n 𝑢 = cos 𝜑. (18)

r
The equilibrium BC (8) transforms as follows:

𝑒 −𝑘𝑠𝑢 Ψ𝑠 = 𝑒 −𝑘𝑙𝑢 Ψ𝑙 = 1 (19)

and, taking into account that

∇n 𝑇𝑠⁄ = Δ𝑇𝑒
−𝑘𝑠⁄ 𝑢
er
(∇n Ψ𝑠⁄ − 𝑘𝑠⁄ Ψ𝑠⁄ ∇n 𝑢),
pe
𝑙 (20)
𝑙 𝑙 𝑙 𝑙
the Stefan BC (7) transforms into:
𝐿𝑣
𝜆𝑠 𝑒 −𝑘𝑠𝑢 ∇n Ψ𝑠 − 𝜆𝑙 𝑒 −𝑘𝑙𝑢 ∇n Ψ𝑙 = cos 𝜑. (21)
Δ𝑇
ot

The heat supply BC (4) takes the form of the integral relation:

Δ𝑇 ∫∂Σ 𝜆𝑙 𝑒 −𝑘𝑙𝑢 ∇n Ψ𝑑𝜎 + ∫Σ 𝑄𝑑𝑢𝑑𝜎 = 0 (22)

tn

where we took into account that the directional cosine of the closed surface  surrounding
the external source takes opposite values on the opposite sides, see Fig. 1. Specific features of
the temperature distribution depend on the geometry of the external heat supply.
rin

Below we consider the cases of application of the plane and line heat sources to the materials
with equal thermal properties in the solid and liquid phases so that 𝐶𝑠 = 𝐶𝑙 = 𝐶 and 𝜆𝑠 = 𝜆𝑙 =
𝜆. Then, the equilibrium (19) and Stefan (21) BC’s take the form:
ep

Ψ𝑠 = Ψ𝑙 = 𝑒 𝑘𝑢 (23)
𝐿𝑣
∇n Ψ𝑠 − ∇n Ψ𝑙 = 𝑒 𝑘𝑢 cos 𝜑 (24)
𝜆Δ𝑇
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 3. PLANE HEAT SOURCE

ed
If a plane source of linear strength Q1 is applied, we may take the plane of its application in the
moving reference frame as u=0 and consider the temperature distribution as independent of y
and z. Then, the fusion surface turns into two unconnected planes at u=u+>0 and u=u−<0 on

iew
which the melting and fusion respectively take place. These planes and the plane of application
of the source divide the space into four contiguous domains: sf=(−,u−), lf=(u−,0), li=(0,u+), and
si=(u+,+). In each domain, the heat equation (15) turns into:

𝜕𝑢2 Ψ − 𝑔2 Ψ = 0 (25)

ev
which admits a solution in the form:

Ψ𝑁 = 𝐴𝑁 𝑒 −𝑔𝑢 + 𝐵𝑁 𝑒 𝑔𝑢 (26)

where N=sf, or lf, or li, or si.

r
The trailing front of the melt pool u=u−, where fusion is taking place, is a source of the latent

−
𝑑Ψ𝑙
=
𝐿𝑣
𝑒 𝑘𝑢
er
heat while the leading front u=u+, where melting is taking place, is a sink of it. At both fronts
the equilibrium BC’s (23) are accompanied by the Stefan BC’s (24), which have the same form:
𝑑Ψ𝑠
pe
(27)
𝑑𝑢 𝑑𝑢 𝜆Δ𝑇
because between the leading and trailing fronts the outer normal n and directional cosine cos 𝜑
both change signs, see Fig. 1.
On the external heat-source plane u=0, the thermal equilibrium BC takes the form:
ot

Ψ𝑙𝑖 = Ψ𝑙𝑓 = Ψ𝑚𝑎𝑥 (28)

tn

where Ψ𝑚𝑎𝑥 is the maximum reduced temperature, and the heat-balance BC (22) takes the
form similar to that of the Stefan BC:
𝑑Ψ𝑙𝑖 𝑑Ψ𝑙𝑓 𝑄1
− + =0 (29)
𝑑𝑢 𝑑𝑢 𝜆Δ𝑇
rin

Substituting solution (26) into BC’s (23, 27-29) and taking into account the far-field BC’s (3),
which lead to 𝐴𝑠𝑓 = 𝐵𝑠𝑖 = 0, we obtain a linear system of eight equations for eight quantities,
AN’s, BN’s u−, and u+:
ep

𝐵𝑠𝑓 = 𝐴𝑙𝑓 𝑒 −2𝑔𝑢− + 𝐵𝑙𝑓 = 𝑒 −(𝑔−𝑘)𝑢− (30a)

𝐵𝑠𝑓 + 𝐴𝑙𝑓 𝑒 −2𝑔𝑢− − 𝐵𝑙𝑓 = 2𝑝𝑒 −(𝑔−𝑘)𝑢− (30b)

Pr

𝐴𝑠𝑖 = 𝐴𝑙𝑖 + 𝐵𝑙𝑖 𝑒 2𝑔𝑢+ = 𝑒 (𝑔+𝑘)𝑢+ (30c)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 −𝐴𝑠𝑖 + 𝐴𝑙𝑖 − 𝐵𝑙𝑖 𝑒 2𝑔𝑢+ = 2𝑝𝑒 (𝑔+𝑘)𝑢+ (30d)

ed
𝐴𝑙𝑓 + 𝐵𝑙𝑓 = 𝐴𝑙𝑖 + 𝐵𝑙𝑖 (30e)

𝐴𝑙𝑓 − 𝐵𝑙𝑓 − 𝐴𝑙𝑖 + 𝐵𝑙𝑖 + 2𝑞1 = 0 (30f)

iew
where
𝑄1
𝑞1 = (31)
2𝑔𝜆Δ𝑇

ev
and
𝐿𝑣
𝑝= . (32)
2𝑔𝜆Δ𝑇

r
The triples of Eqs.(30a,b) and (30c,d) can be immediately resolved in terms of u− and u+:

𝐵𝑠𝑓 = 𝑒 (𝑔−𝑘)|𝑢−|
𝐴𝑙𝑓 = 𝑝𝑒 −(𝑔+𝑘)|𝑢−| ; 𝐵𝑙𝑓 = (1 − 𝑝)𝑒 (𝑔−𝑘)|𝑢−|
er (33a)

(33b)
pe
𝐴𝑙𝑖 = (1 + 𝑝)𝑒 (𝑔+𝑘)𝑢+ ; 𝐵𝑙𝑖 = −𝑝𝑒 −(𝑔−𝑘)𝑢+ (33c)

𝐴𝑠𝑖 = 𝑒 (𝑔+𝑘)𝑢+ (33d)

where we have factored in that u+>0 and u−<0. Then Eqs.(30e,f) turn into a system of two
ot

equations for u− and u+:

𝑞1 = (1 + 𝑝)𝑒 (𝑔+𝑘)𝑢+ − 𝑝𝑒 −(𝑔+𝑘)|𝑢−| (34a)

tn

𝑞1 = (1 − 𝑝)𝑒 (𝑔−𝑘)|𝑢−| + 𝑝𝑒 −(𝑔−𝑘)𝑢+ (34b)

which cannot be resolved in terms of the elementary functions. However, comparing (34a)
rin

with (34b), one can find a condition of solvability in the form:

(𝑔 − 𝑘)|𝑢− | ≥ (𝑔 + 𝑘)𝑢+ (35)

If L=0 (p=0, the Rosenthal case) then

ep

0| ln 𝑞1 0 ln 𝑞1
|𝑢− = > 𝑢+ = (36)
𝑔−𝑘 𝑔+𝑘

If L0, but H=0 (𝑔 = 𝑘), see Eq.(34b), the stationary state is possible only if q1=1 (𝑄1 = 𝐶Δ𝑇𝑣)
Pr

that is, without surface losses the stationary motion of a plane heat source is possible at the
specific speed only, and we recover from Eq.(34a) the Rosenthal solution with 𝑢− = 𝑢+ = 0.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 Furthermore, analysis of Eq. (34a) yields another solvability condition in the form of a relation

ed
between the material and control parameters of the stationary state:

𝑞1 ≥ 1 + 𝑝 (37)

Solution (33,34) also allows us to see that at u=0 the reduced temperature reaches its finite

iew
maximum value of:

Ψ𝑚𝑎𝑥 = 𝑞1 − 𝑝[𝑒 −(𝑔−𝑘)𝑢+ − 𝑒 −(𝑔+𝑘)|𝑢− | ] (38)

Notice that the solid-phase temperature field extended into the melt pool would reach its

ev
maximum value at
(𝑔+𝑘)𝑢+ −(𝑔−𝑘)|𝑢− |
𝑢𝑠 = ≤0 (39)
2𝑔

r
The solvability condition (35) prompts us to consider a high-speed approximation where

(𝑔 + 𝑘)|𝑢− | ≫ 1 𝑎𝑛𝑑 (𝑔 − 𝑘)𝑢+ ≪ 1.

In this approximation:
1 𝑞1 −𝑝 1
er
𝑞1
(40)
pe
|𝑢− | ≈ ln > 𝑢+ ≈ ln (41a)
𝑔−𝑘 1−𝑝 𝑔+𝑘 1+𝑝

and
1 𝑞1 (1−𝑝)
𝑢𝑠 ≈ ln (𝑞 <0 (41b)
2𝑔 1 −𝑝)(1+𝑝)
ot

Compared to the latent-heat-free system, the melting front approaches the source while the
fusion front recedes from it so that the melt pool gains net expansion of
tn

0| 0) 1 1−𝑝⁄𝑞1 1
∆𝑢 ≡ (|𝑢− | + 𝑢+ ) − (|𝑢− + 𝑢+ ≈ ln − ln(1 + 𝑝) ≈
𝑔−𝑘 1−𝑝 𝑔+𝑘
(𝑔+𝑘)⁄
2𝑘− 𝑞1
rin

𝑝 2 2 (42)
𝑔 −𝑘

where the last expression is for a system with small p.

In the dimensional terms, the solvability condition (37) and the maximum temperature of the
ep

stationary state (12,38) take the form:

𝑄1 ≥ Δ𝑇√(𝐶𝑣)2 + 4𝜆𝐻 + 𝐿𝑣 (43)

Pr

𝑄1 −𝐿𝑣[𝑒 −(𝑔−𝑘)𝑢+ −𝑒 −(𝑔+𝑘)|𝑢− | ] 𝑄1 −𝐿𝑣

𝑇𝑚𝑎𝑥 = 𝑇0 + ≈ 𝑇0 + ≥ 𝑇𝑀 (44)
√(𝐶𝑣)2 +4𝜆𝐻 √(𝐶𝑣)2 +4𝜆𝐻

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 where the last equality is given in the high-speed approximation. The solvability condition (43)

ed
shows that the minimum strength of the heat-source in the stationary state increases with the
undercooling, scanning speed, surface losses, and latent heat. The maximum temperature (44),
on the contrary, decreases in presence of the latent heat. These effects of the latent heat are
consequences of the proximity of the melting front to the external source plane while the

iew
fusion front is removed from it. In other words, the source-sink dipole of the latent heat is not
symmetric with respect to the external source of heat.
To assess the morphological stability of the stationary moving fusion front in presence of a
plane heat source, we calculate the structure forming number:

ev
𝑔−𝑘−2𝑝𝑔
𝐺= (45)
2𝑘
According to the condition (10), the fusion front is unstable if 𝐺 < 𝜆𝑙 ∆𝑇𝑠⁄𝑙 ⁄𝐷𝐶∆𝑇. However,

r
formula (45) hints at a possibility of genuine, not constitutional, supercooling at the trailing
edge of the melt pool (G<0) if H is small and/or L is large. In other words, if the heat exchange

er
is weak and/or latent heat is large, there will be a region of “cool melt” in front of the trailing
edge of the pool and the microstructure of the recrystallized material will be highly
inhomogeneous, even if the material is one-component.
pe
To find the conditions of stable stationary motion of the melt pool we plot relations (16,37,45)
in the plane of the thermal wavenumber and impedance (k, g), see Figure 2, and find that there
is a domain of material and control parameters where the genuine supercooling is absent, and
the stationary plane front of fusion is stable. The criterion of stationary stability has a critical
point where Eqs.(16,37) and G=0 in Eq.(45) are satisfied simultaneously. This point is
ot

characterized by the maximum strength and speed of motion of the heat source:

𝑄1∗ = (𝐶∆𝑇 + 3𝐿)𝑣 ∗ (46a)

tn

𝜆𝐻 𝜆𝐻∆𝑇
𝑣 ∗ = ∆𝑇√ ≈√ (46b)
𝐿(𝐶∆𝑇+𝐿) 𝐶𝐿
rin

where the last equality is for a material with small latent heat.
ep
Pr

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 ed
iew
r ev
er
pe
ot
tn
rin

Figure 2. The melt-pool stability diagram in the plane (k, g). Curve 1—relation (16) for a given
value of H/; line 2—boundary of stationary motion of the melt pool (37) for q1=2, p=0.5; line
3—boundary of stability of the trailing edge of the melt pool (45); pink polygon—stationary,
stable states of motion of the pool; black circle—operational point with (k=1, g=1.5). Curve 4—
ep

relation (16) for increased value of H/; red star—the critical operational point.
In Figure 3 is depicted the temperature distribution (12,26,33,41a) due to a stationary moving
plane heat source in the approximation of high speed of its motion (40). The operational point
of the process is chosen inside the stationary but outside of the stable domain of motion (black
Pr

circle in Fig. 2). The plot shows that in front of the trailing edge of the melt pool there is a
region of supercooled melt.

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 ed
iew
r ev
er
pe
ot
tn

Figure 3. Temperature distribution due to a stationary moving plane heat source in the high-
rin

speed approximation with q1=2, p=0.5, g=1.5, k=1 (black circle in Fig. 2).
4. LINE HEAT SOURCE
If a line heat source of point strength Q2 is applied, the fusion surface (17) turns into a right
ep

cylinder on which the melting and resolidification take place, see Fig. 1. The point of application
of the source may be taken as u=0 and the temperature field may be considered as
independent of z. Then, it is advantageous to use the polar (to be exact, the z-independent
cylindrical) moving coordinate system
Pr

𝑢 = 𝑟 cos 𝜃 ; 𝑦 = 𝑟 sin 𝜃 (47)

13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 where r and  are respectively the radial distance and polar angle, and the point of application

ed
of the source is the origin. The fusion line:

𝛷(𝑟, 𝜃 ) = 0 (48)

is the intersection of the fusion surface with the plane z=0. The differential geometry provides

iew
an expression for the normal vector:
𝑟Φ 𝐣r −𝑟̇ Φ 𝐣θ
𝐧= (49a)
2 +𝑟̇ 2
√𝑟Φ Φ

ev
and a relationship between the directional  and polar  angles:
𝑟̇ Φ
tan(𝜃 − 𝜑) = (49b)
𝑟Φ

r
where 𝑟Φ (𝜃) is the solution of the fusion line Eq.(48), the dot is differentiation with respect to
the polar angle, and 𝐣r , 𝐣θ are the unit vectors of the polar coordinates. Then,

∇n = 𝐧 ∙ 𝛁2 =
2
𝑟Φ 𝜕𝑟 −𝑟̇ Φ 𝜕θ
2 +𝑟̇ 2
𝑟Φ √𝑟Φ
er (50)
pe
Φ

Notice from (49,50) that the condition (18) is satisfied.

In the polar coordinates, the temperature fields of the liquid and solid phases obey:

𝜕𝑟2 Ψ𝑠⁄ + 𝑟 −1 𝜕𝑟 Ψ𝑠⁄ + 𝑟 −2 𝜕𝜃2 Ψ𝑠⁄ − 𝑔2 Ψ𝑠⁄ = 0

ot

(51)
𝑙 𝑙 𝑙 𝑙
In the liquid phase that is, inside the fusion line, Eq. (51) is satisfied by the following linear
tn

combination of the modified Bessel functions [18]:

Ψ𝑙 = A𝑙 K 0 (𝑔𝑟) + B𝑙 I1 (𝑔𝑟) cos 𝜃 (52)

In the solid phase that is, outside the fusion line, taking into account the far-field BC (3) and
rin

that I1(z→)→ [18], we conclude that Bs=0. However, based on the solution of the plane
heat-source problem, see Eqs.(26,33,39), we may expect the temperature field to be
independent of the polar angel if viewed from a point Os shifted by a distance s behind u=0 as
the origin, see Fig.1:
ep

𝑢 + 𝑠 = 𝑟𝑠 cos 𝜃𝑠 ; 𝑦 = 𝑟𝑠 sin 𝜃𝑠 (53a)

The reverse transformation of (53a) is:

Pr

𝑟 sin 𝜃
𝑟𝑠2 = 𝑟 2 + 2𝑟𝑠 cos 𝜃 + 𝑠 2 ; tan 𝜃𝑠 = (53b)
𝑟 cos 𝜃+𝑠

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 Then:

ed
Ψ𝑠 = A𝑠 K 0 (𝑔𝑟𝑠 ) (54)

Substitution of the solutions (52-54) into BC’s (23,24) yields a system of equations for the fusion
line 𝑟Φ (𝜃) and coefficients of the temperature field As, Al, Bl:

iew
A𝑠 K 0 [𝑔𝑟𝑠 (𝑟Φ , 𝜃 )] = A𝑙 K 0 (𝑔𝑟Φ ) + B𝑙 I1 (𝑔𝑟Φ ) cos 𝜃 = 𝑒 𝑘𝑟Φ cos 𝜃 (55a)

A𝑠 𝑓𝑠 (𝑟Φ , 𝜃 )K1 [𝑔𝑟𝑠 (𝑟Φ , 𝜃 )] − A𝑙 K1 (𝑔𝑟Φ ) + B𝑙 cos 𝜃 [I0 (𝑔𝑟Φ ) −

ev
1 𝑟̇ Φ 𝑘𝑟Φ cos 𝜃 √ 𝑟̇ Φ 2
I (𝑔𝑟Φ ) + 2 I1 (𝑔𝑟Φ ) tan 𝜃] + 2𝑝 cos 𝜑 𝑒 1+( ) =0
𝑔𝑟Φ 1 𝑔𝑟Φ 𝑟 Φ
(55b)

r
where

𝑟̇ 2
𝑓𝑠 (𝑟Φ , 𝜃 ) ≡ √1 + ( Φ) ∇n 𝑟𝑠 (𝑟Φ , 𝜃 ) = Φ Φ (𝑟 Φ
𝑟Φ er
𝑟 +𝑟 𝑠 cos 𝜃+𝑟̇ 𝑠 sin 𝜃 2

𝑟Φ 𝑟𝑠 Φ ,𝜃)

Coefficient Al may be independently found from the heat-balance BC (22) on the external
(55c)
pe
source Q2. Indeed, taking the closed contour  surrounding the source as a small circle of
radius r→0, see Fig.1, and using properties of the modified Bessel functions, see [18], we obtain
that
𝑄2 𝑞2
A𝑙 = ≡ .
ot

(56)
2𝜋𝜆Δ𝑇 √𝜋
Then, excluding the coefficients of As, Al, and Bl from Eqs. (55), we obtain a differential equation
tn

for the fusion line function 𝑟Φ (𝜃):

K1 (𝑔𝑟Φ ) K1 [𝑔𝑟𝑠 (𝑟Φ ,𝜃)] I (𝑔𝑟 )
R(𝑟Φ , 𝜃 ) − 𝑓 (𝑟 , 𝜃 ) = [1 − R(𝑟Φ , 𝜃 )] [ 0 (𝑔𝑟Φ) −
K0 (𝑔𝑟Φ ) K0 [𝑔𝑟𝑠 (𝑟Φ ,𝜃)] 𝑠 Φ I1 Φ
rin

1 𝑟̇ Φ 𝑟̇ 2
(1 − 𝑟 tan 𝜃)] + 2𝑝 cos 𝜑 √1 + ( Φ) (57)
𝑔𝑟Φ Φ 𝑟Φ

where we introduced the Rosenthal function:

ep

R(𝑟, 𝜃 ) ≡ A𝑙 K 0 (𝑔𝑟)𝑒 −𝑘𝑟 cos 𝜃 (58)

If L=0 (p=0), s=0 and we recover from (57) the Rosenthal solution 𝑟R (𝜃) in its implicit form:
R(𝑟R , 𝜃) = 1. For L0 (s0), let us look at the high-speed approximation with 𝑔𝑟Φ ≫ 1. Using
Pr

the asymptotic properties of the modified Bessel functions [18], we find from (57) an
asymptotic expansion of the fusion line equation in the form:

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 𝑠 𝑟̇ Φ 𝑟̇ 2 1

ed
R(𝑟Φ , 𝜃 ) = 1 + 2 sin 𝜃 + 𝑝 cos 𝜑 √1 + ( Φ) + 𝑂 ( ) (59)
2 𝑟Φ 𝑟 𝑔𝑟 Φ Φ

This differential equation may be either integrated numerically or solved perturbatively for 𝑝 ≪
1. Indeed, using the relations:

iew
𝜕𝜃 R(𝑟R ,𝜃) 𝑘 sin 𝜃
𝑟̇R = − ≅ 𝑟 (60a)
𝜕𝑟 R(𝑟R ,𝜃) 𝑔+𝑘 cos 𝜃 R

𝑔 cos 𝜃+𝑘
cos 𝜑R ≅ (60b)
√𝑔2 +𝑘 2 +2𝑔𝑘 cos 𝜃

ev
we obtain the perturbative solution in the form:

𝑘 sin2 𝜃
(𝑔 + 𝑘 cos 𝜃 )2 (𝑟Φ − 𝑟R ) + 𝑠 + 𝑝(𝑔 cos 𝜃 + 𝑘) = 0 (61)

r
2𝑟R

The fusion line may be characterized by the longitudinal {𝑟+ = 𝑟Φ (0), 𝑟− = 𝑟Φ (𝜋)} and lateral
𝜋
er
{𝑟0 = 𝑟Φ (𝜑 = ± )} turning points, see Fig. 1. Then, solution (61) can be used to find the melt
2
pool deformation ∆𝑟 ≡ 𝑟Φ − 𝑟R due to the latent heat:
pe
𝑝
∆𝑟± ≅ ∓ (62a)
𝑔±𝑘
𝑘𝑠 𝐶𝑣𝑠
∆𝑟0 ≅ − =− (62b)
2𝑟R0 (𝑔2 −𝑘 2 ) 4𝑟R0 𝐻
ot

These formulae show contraction of the leading (freezing) edge, expansion of the trailing
(melting) edge, and small contraction of the lateral size. The net expansion of the melt pool of
the line source due to the latent heat effect, defined as ∆𝑢 ≡ ∆𝑟− + ∆𝑟+ , is practically identical
tn

with that of the plane source, Eq. (42).

The relationship between s and p can be found from Eqs.(55a, 59) applied to the longitudinal
0
turning points where 𝜃 ± = 𝜑 ± = , 𝑟̇± = 0, 𝑟s± = 𝑟± ± 𝑠, 𝑓𝑠± = 1. Indeed, in the high-speed
rin

𝜋
approximation, using the properties of the modified Bessel functions [18], we obtain the
following relations:

(𝑟+ + 𝑠)𝑒 2(𝑔+𝑘)𝑟++2𝑔𝑠 = (𝑟− − 𝑠)𝑒 2(𝑔−𝑘)𝑟−−2𝑔𝑠 (63)

ep

𝑞22
)2
(1 + 𝑝 𝑟+ 𝑒 2(𝑔+𝑘)𝑟+
= (1 − 𝑝 𝑟− 𝑒 )2 2(𝑔−𝑘)𝑟−
= (64)
2𝑔
Pr

The latter relation shows that the large-𝑔𝑟Φ approximation corresponds to the case of a strong
heat source: 𝑞2 ≫ 1. If H0 (𝑔 ≠ 𝑘), (64) has an asymptotic solution:

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 1 𝑔±𝑘
𝑟± = (ln 𝑎± − ln ln 𝑎± + ⋯ ); 𝑎± = 𝑞22

ed
(65)
2(𝑔±𝑘) 𝑔(1±𝑝)2

which is an extension of Eq.(62a). Analysis of (64) reveals a line-source solvability condition,

which is practically identical with that of the plane source, Eq.(37), with the replacement of q 1

iew
with q2 that is, Q1 with gQ2. Furthermore, dividing (63) by (64) we obtain the following equation
for the shift s:

4𝑔𝑠 𝑟+ (𝑟− −𝑠) (1+𝑝)2

𝑒 = (66)
𝑟− (𝑟+ +𝑠) (1−𝑝)2

which admits a perturbative solution:

ev
4𝑝 𝑝
𝑠≈ 1 1 ≈ (67)
4𝑔+ + 𝑔
𝑟− 𝑟+

r
Finally, for the coefficients of the temperature field in terms of 𝑟± and s Eqs.(55a) yield in the
large-𝑔𝑟Φ approximation:

2
A𝑠 = √ 𝑔(𝑟± ± 𝑠)𝑒 (𝑔±𝑘)𝑟±±𝑔𝑠
𝜋
er (68a)
pe
B𝑙 = −𝑝√2𝜋𝑔𝑟± 𝑒 −(𝑔∓𝑘)𝑟± (68b)

Then, using (12,52,54,68) we can plot the temperature distribution as a function of u along the
y=0 direction, see Figure 4. Again, the plot shows that in front of the trailing edge of the melt
ot

pool there is a region of supercooled melt. However, the maximum temperature for a line
source is infinite. To the first order of the large-𝑔𝑟Φ approximation, the structure forming
number G of the line source, Eq.(10), is equal to that of the plane source, Eq.(45), and the line-
tn

source stability diagram is similar to that of Fig. 2.

rin
ep
Pr

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 ed
iew
r ev
er
pe
ot
tn

Figure 4. Temperature distribution due to a stationary moving line heat source in the strong
rin

heat-source approximation with q2=2, p=0.5, g=1.5, k=1 (black circle in Fig. 2).
5. COMPARISON WITH OTHER WORKS
A direct comparison of the results of this work with the experimental data is not immediately
ep

possible because of several reasons. First, most of the experiments were conducted in the
conditions of the point source. Although the line-source heating resembles heating of a thin
plate by the point source, the effective thickness of the plate is not apparent. Second, the
Newtonian heat exchange coefficient H was not readily available from the literature. Third, the
Pr

present treatment disregards radiative cooling. Fourth, another source of a discrepancy is that
most of the AM-related experiments were done on alloys characterized by the freezing range

18

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 ∆𝑇𝑠⁄ while we consider a one-component material characterized by the melting point TM. That

ed
𝑙
is why we do not expect a quantitative matching with the experiments. Instead, we are looking
for a qualitative understanding of the effects of various material properties and control
parameters of the process on its outcome. To correct for the missing information the following
adjustments have been made. The strength of the line heat source Q2 was estimated as the

iew
power density (P/R2) multiplied by the ratio of the depth of beam penetration and thickness of
the heat affected zone. H was estimated based on the cooling time of a layer of the thickness of
the heat affected zone, and the melting point was estimated as the average of the liquidus and
solidus temperatures. The results of the calculations of the melt pool dimensions by
Eqs.(65,67) are summarized in the Table 1 and compared with the numerical simulations of the

ev
corresponding processes, which, in due course, were compared to the experimental results.
Table 1. Estimated melt pool lengths in the simulated AM processes of selected alloys. Physical
properties and control parameters of the alloys were obtained from the indicated references.

r
Physical Properties Control Parameters
 
Quantity
Unit
Inconel
625 [14]
TM
K

1531
L
MJ/kg

0.218
C
J/kg*K

725
kg/m3

7846
W/K*m er
30.8
T0
K

298
v
m/s

0.2
P
W

60
R
m

42.9
pe
Inconel
625 [10] 1593 0.227 412 8440 9.83 298 0.13 50 35.0
Ti-6Al-
4V [12] 1903 0.200 677.5 4420 31.05 298 0.006 600 700

Adjustable Dimensional Dimensionless

ot

Parameters Parameters Numbers Melt Pool Length

ratio H T k g q2 p G r+ r- s
𝑀𝑊
tn

𝐾𝑚3 K 1/mm 1/mm mm mm mm

10 −3
500 1233 18.5 18.9 77.3 0.24 −0.23 0.090 4.34 0.011
510−4 500 1295 23.0 24.1 144 0.40 −0.40 0.081 2.69 0.015
810−3 10 1605 .289 .637 17.7 0.084 0.42 2.25 5.24 0.105
Comparing the estimates of the melt pool length from Table 1 with the reported values—0.4
rin

mm [14], 0.1 mm [10], and 2.8 mm [12]—we find that our results are 310 times higher but
retain the qualitative similarity to the simulated ones. The results in Table 1 show that the Ti
alloy has the longest melt pool, which can be attributed to the greatest source power used.
Also, one can see that in the simulated experiments the Inconel alloys were genuinely
ep

supercooled (outside the stationary stable zone of Fig. 2) while the Ti alloy was not (inside the
zone). However, the morphological stability of the Ti alloy was still broken because ∆𝑇𝑠⁄𝑙 =
50𝐾 and D<10−10 m2/s [19] and, hence, the criterion (10) was not satisfied.
Pr

In Figure 5 are depicted the fusion lines of Inconel 625 [10] drawn by the parameters from
Table 1. The blue curve is the Rosenthal solution of R(𝑟R , 𝜃) = 1 and the red curve is the

19

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 trajectory (61), which accounts for the effect of the latent heat. The major deviations between

ed
the two are near the longitudinal turning points where the fusion boundary approaches the
source while the melting boundary recedes from it. The lateral turning points practically do not
deviate from the latent-heat-free curve because there are no latent heat emissions there.
Overall, the difference between the fusion lines is not great because of the large surface heat-

iew
loss coefficient H, see Table 1.

r ev
Figure 5. The fusion line of Inconel 625 [10] with the parameters from Table 1. Blue line—the

direction.
6. DISCUSSION
er
Rosenthal solution; red line—the current result. Dot—external heat source; arrow—scanning
pe
In this paper we analyze the effects of latent heat in conditions related to the additive
manufacturing technology. The free boundary problem is the most effective theoretical
framework to address the problem of five material properties: , H, C, TM, L, and three control
parameters: T0, Q, v, where the heat conduction () and Newtonian heat exchange (H) are not
ot

the only means of cooling. The scanning speed (v) also acts as a means of cooling by exposing
the material (C(TM−T0)v) to the heat source Q. Several effects of the latent heat L depend on the
geometry of the heat source while others are independent of it. Among the latter ones is the
tn

most important finding of our analysis—a possibility of the genuine (not constitutional)
supercooling at the trailing (freezing) edge of the melt pool. In the theory, it is measured by the
structure forming number G of the system, Eq. (10), which changes significantly compared to
the latent-heat-free system. In the experiment, the “cool spot” in front of the trailing edge of
rin

the melt pool will be overgrown by dendrites and turn into a mushy zone. The condition of the
genuine supercooling is characterized by the critical speed and strength of the heat source, see
Fig. 2, which may be used to measure the material properties, e.g., the surface heat losses. The
latter is essential for the establishing of the stationary state in presence of the plane source and
ep

important but not critical for the line one.

One of the findings of the present treatment is contraction at the freezing edge, expansion at
the melting edge, and net longitudinal expansion of the melt pool of the line source due to the
Pr

latent heat. The latter is similar to that due to the plane source, Eq. (42), which can be
explained by the possibility to present a plane source as a linear combination of the line sources

20

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 with the strength of the plane source enhanced by the thermal impedance of the system:

ed
Q1=gQ2. The lateral size of the melt pool formed by the line source does not change significantly
in presence of the latent heat because at the points of maximum separation from the axis of
motion of the external source (the lateral turning points) where =± the strength of the latent
source is zero.

iew
Another finding of the present work is increase of the minimum strength of the heat source for
the stationary motion of the melt pool due to the latent heat, Eq. (43). Moreover, the latent
heat reduces the maximum temperature in the melt pool of the plane source, Eq. (44). These
effects come from the asymmetry of the latent-heat dipole with respect to the external heat
source that is, proximity of the melting boundary (heat sink) to the external source position

ev
while the fusion boundary (heat source) is removed from it.
However, there are several important physical properties and phenomena, which were not
accounted for by the free-boundary model developed in this paper. Among those are the

r
species diffusion, energy, thickness, mobility, and anisotropy of the boundary, energy
dissipation and entropy increase as a result of the latent heat accretion and release. These

er
phenomena, together with a more realistic geometry of the heat source, can be accounted for
by a phase-field model of AM [20], which we plan to develop in the future. The phase-field
model of AM may also be useful for other technologies, like laser melting, electron (ion)-beam
pe
irradiation, and fusion welding.
ACKNOWLEDGEMENTS
This work was supported by the funding provided by the Department of Chemistry, Physics, &
Materials Science at Fayetteville State University.
ot

REFERENCES
1. M. Bayat, W. Dong, J. Thorborg, A.C. To, J.H. Hattel, A review of multi-scale and
tn

multi-physics simulations of metal additive manufacturing processes with focus on

modeling strategies, Additive Manufacturing 47 (2021) 102278
2. J.C. Jaeger; Some problems involving line sources in conduction of heat. Phil. Mag.
35 (7) (1944) 169-179.
rin

3. H.S. Carslaw and J.C. Jaeger, Conduction of Heat in Solids. Clarendon Press, Oxford,
1992, p. 266.
4. D. Rosenthal, The Theory of Moving Source of Heat and Its Application to Metal
Treatment. Trans Amer. Soc. Mech. Engrs. 68 (1946) 849–866.
ep

5. M.E. Glicksman and R.J. Schaefer, Investigation of Solid/Liquid Interface

Temperatures via Isenthalpic Solidification; J. Cryst. Growth. 1 (1967) 297-310
6. R.J. Schaefer, The Validity of Steady-State Dendrite Growth Models; J. Cryst. Growth.
43 (1978) 17-20
Pr

7. A.R. Umantsev, Motion of a Plane Front During Crystallization; Sov. Phys.

Crystallography 30 (1) (1985) 87-91.

21

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818
 8. T.W. Eagar, N.S. Tsai, Temperature fields produced by traveling distributed heat

ed
sources. Weld J 62(12) (1983) 346–355
9. J. Goldak, A. Chakravarti, and M. Bibby, A New Finite Element Model for Welding
Heat Sources; Metallurgical Transactions B, 15B, (1984) 299-305.
10. L.E. Criales, Y.M. Arısoy, and T. Özel, Sensitivity analysis of material and process

iew
parameters in finite element modeling of selective laser melting of Inconel 625; Int J
Adv Manuf Technol 86 (2016) 2653–2666.
11. P. Peyre, P. Aubry, R. Fabbro, R. Neveu and A. Longuet, Analytical and numerical
modelling of the direct metal deposition laser process; J. Phys. D: Appl. Phys. 41
(2008) 025403
12. E. Mirkoohi, J. Ning, P. Bocchini, O. Fergani, K.-N. Chiang and S.Y. Liang, Thermal

ev
Modeling of Temperature Distribution in Metal Additive Manufacturing Considering
Effects of Build Layers, Latent Heat, and Temperature-Sensitivity of Material
Properties, J. Manuf. Mater. Process, 2(3), (2018) 63

r
13. S.D. Proell, W.A. Wall and C. Meier, On phase change and latent heat models in
metal additive manufacturing process simulation, Adv. Model. and Simul. in Eng. Sci.

14.
(2020) 7:24
er
B. Stump, A. Plotkowski and J. Coleman, Solidification dynamics in metal additive
manufacturing: analysis of model assumptions, Modelling Simul. Mater. Sci. Eng. 29
(2021) 035001
pe
15. L.I. Rubinstein, The Stefan Problem. American Mathematical Society, Providence,
Rhode Island, 1971
16. D.A. Porter and K.E. Easterling, Phase Transformations in Metals and Alloys.
Chapman & Hall, GB 1991, p. 215.
W. Kurz and D.J. Fisher, Fundamentals of Solidification, 3rd Ed. Trans Tech Pub,
ot

17.
Switzerland, 1989, p. 50.
18. M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions, NBS Applied
tn

Math Series, Washington, DC, Sec. 9.6-9.8.

19. G. Lindwall, KW. Moon, Z. Chen, et al. Diffusion in the Ti-Al-V System. J. Phase
Equilib. Diffus. 39, 731–746 (2018).
20. A. Umantsev, Field Theoretic Method in Phase Transformations; Springer, Series:
rin

‘Lecture Notes in Physics’, vol 840, April 27, 2012

ep
Pr

22

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4250818