Materials and Design 133 (2017) 549–558

Contents lists available at ScienceDirect

Materials and Design

journal homepage: www.elsevier.com/locate/matdes

Numerical prediction of the printable density range of lattice structures

for additive manufacturing
Niyazi Tanlak a, b, * , Dirk Frederik De Lange a, c , Wim Van Paepegem a
a
Mechanics of Materials and Structures, Department of Materials Science and Engineering, Ghent University, Technologiepark-Zwijnaarde 903, Zwijnaarde 9052, Belgium
b
SIM M3 Program, Technologiepark 935, Zwijnaarde B-9052, Belgium
c
Universidad Autónoma de San Luis Potosí, Facultad de Ingeniería, Zona Universitaria, San Luis Potosí, Mexico

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Even if ideal conditions are assumed,

there is a maximum printable density
limit for lattices being fabricated with
powder bed fusion.
• When lattice size increases, the print-
able range widens and approaches
to ideal conditions while it gets nar-
rower when the powder lump size
increase and the machine precision
exacerbates.
• The powder adhesion and the prob-
lems due to strut inclination can also
be modeled with powder size.
• The printable density range in work-
ing conditions will be narrower than
the ones found in this work.
• Finding optimized process parame-
ters or the advancement of technol-
ogy will approach the real limits to
the ones found in this paper.

A R T I C L E I N F O A B S T R A C T

Article history: Structured cellular structures are nowadays printed using additive manufacturing methods like powder
Received 21 April 2017 bed fusion. The relative density of the cellular structures has a big role in the suitability of a lattice for
Received in revised form 23 July 2017 printing due to the minimum printable radius constraint and powder being trapped inside an inclusion. In
Accepted 4 August 2017 this work, the theoretical limits of the printable range of relative density of different lattice types are found
Available online 7 August 2017
based on the cell size using computer methods by leaving other process parameters for further research as
the current parameters are the most basic ones. The results are approximated using simple polynomials to
Keywords: enable practical usage.
Additive manufacturing
© 2017 Elsevier Ltd. All rights reserved.
3D printing
Powder bed fusion
Structured cellular structures
Lattice
Printable relative density

* Corresponding author at: Mechanics of Materials and Structures, Department of Materials Science and Engineering, Ghent University, Technologiepark-Zwijnaarde 903,
Zwijnaarde 9052, Belgium.
E-mail address: niyazi.tanlak@ugent.be (N.Tanlak).

http://dx.doi.org/10.1016/j.matdes.2017.08.007
0264-1275/© 2017 Elsevier Ltd. All rights reserved.
550 N. Tanlak et al. / Materials and Design 133 (2017) 549–558

1. Introduction and loose powder adhesion to struts. Wang et al. [47] marked the
powder adhesion to struts as unavoidable in selective laser melt-
Cellular structures (a.k.a. foams) are known to have a good strength ing (SLM). The overhanging struts increase the problem. Optimizing
to weight ratio. They are used in many engineering applications rang- design and process control can minimize powder adhesion. Powder
ing from passive safety measures to space applications. By introducing adhesion is linked to balling phenomenon giving rise to beads on
structured foams, engineers have better control in tailoring the over- laser melted surfaces on building direction front by Yan et al. [42].
all material properties according to their needs. However, structured Abele et al. [48] claimed that building orientation has not any mean-
cellular materials are hard to produce with conventional manufactur- ingful effect on the printability of scaffolds in SLM processes. They
ing methods. Nonetheless, there are studies about the fabrication of also claimed that powder adhesions have not any meaningful effect.
periodic cellular materials with conventional methods, such as sheet On the other hand, Sallica-Leva et al. [49] concluded that powder
forming,perforated/slottedsheetfolding/drawing[1–4],extrusion[5], adhesion can act as stress concentrators, which in turn affects fatigue
wire assembly [6,7], and investment casting methods [8–12], how- strength of the porous structures. Hazlehurst et al. [50] concluded
ever, these techniques have limited flexibility with respect to structure that surface waviness variation and heterogeneities have a big effect
type and dimensions. On the other hand, additive manufacturing (AM) on the stiffness of lattice structures fabricated by SLM by using
allows an improved flexibility in geometry and size of lattice types. CoCrMo powder. Kim et al. [51] claimed that post-SLM sintering or
That opens up new possibilities for engineering applications. jet blasting can result in localized removal of these powder off the
Based on engineers’ needs, the unit cell of a structured foam is struts without changing the macro-properties of the scaffolds.
tailored such that the global properties of the edifice are adjusted to Leary et al. [52] found that downward facing surfaces of inclined
fulfill the specifications. Type, size and volume fraction of the unit struts had higher surface roughness than upward facing surfaces.
cell (lattice) are subjected to adjustments to reach the desired prop- They worked on overhanging struts having inclinations of 0, 35.26,
erties. Although there are studies about the mechanical properties of 45 and 90◦ being made up of AlSi12Mg by using SLM. They reported
foams [8,9,12-38], the printable ranges of the volume fraction of dif- that horizontal struts (0◦ inclined) were not printable for any of the
ferent lattice types with various cell sizes are not known thoroughly. diameters they tested. But they reported that reducing the spanned
Intuitively, it is tempting to think that such a cellular structure is distance can improve manufacturability for low inclination struts.
always printable. However, the printable density limits may change Similar surface roughness results of Leary et al. were found by Fox
dramatically with respect to the cell size and the type of lattice. et al. [53], Triantaphyllou et al. [54], Drescher et al. [55], and Suard
Either the lower limit or the higher limit of printable density may be et al. [56]. For Ti6Al4V, cantilever strut elements were studied by
paramount according to the needs at hand. The lower printable den- Mazur et al. [57]. They varied inclination angle in the range of
sity limit is based on the minimal feature size that can be printed 0–60◦ with 10◦ increments, and strut diameter of 0.3–1.0 mm with
by a specific process. The maximum printable density limit is prin- 0.1 mm increments. They found out that the minimum manufac-
cipally based on the density up to which all raw material can be turable strut angle is 20◦ and the minimum manufacturable strut
evacuated after manufacture. Although the entrapment of raw mate- diameter is 0.3 mm. On the other hand, they also found out that
rial formally allows the manufacture, the entrapped material will add struts as big as 3 mm are manufacturable while supported at both
mass in comparison with the designed density, which is considered ends. But horizontal struts ended up with oversized diameters due to
unacceptable. For better customization, the limits of the feasibility of high powder adhesion. Weißmann et al. [58] studied vertical and 45◦
printing should be known beforehand. In this work, we will focus on inclined struts fabricated by SLM and Electron Beam Melting (EBM).
selective powder bed fusion or sintering, but can equally be applied The struts from the SLM process were closer to the target specifica-
to several other AM processes. tion than the EBM-manufactured struts. The 45◦ inclined struts had
The printability of lattice structures using powder bed fusion bigger deviations from the desired.
depends on unit cell shape, relative density (i.e. strut radius) of unit Rehme and Emmelmann [59] defined an aspect ratio as cell
cell, strut inclination, powder particle size, powder material, machine edge length to strut diameter to assess the manufacturability by
precision, laser power, laser spot size, laser scanning speed, layer changing strut radius, cell size, and cell type. For 316-L stainless
thickness, and hatch spacing. The effect of spacing between parallel steel, the printable range was in between 3.36 and 22.9 from rela-
laser scans (a.k.a. hatch spacing) on manufacturability was studied by tively low to high, respectively. They also reported that printability
Zhang et al. [39]. They concluded that hatch spacing must be greater increases when strut length gets shorter. About minimum print-
than the laser spot size. They also advised that three times bigger pore able radius, researchers reported several values. The successful pro-
diameter than the maximum powder size must be present because duction of strut diameter of 0.05 mm were achieved by Van Bael
partially melted powder particles stick on strut surfaces. et al. [43] and Hao et al. [60] stated that struts with 0.025 mm radius
Sing et al. [40] reported that laser scan speed and laser power were possible. The studies investigating manufacturability of lattice
did not affect Young’s modulus. But strut diameter or lattice type structures, to authors’ knowledge, concentrated on specific param-
affected the elastic constant significantly while they did not affect the eters for small variations and only for a limited number of lattice
dimensional accuracy. They also reported that with the increase of types. Only one study [59] mentions explicit limits for one machine
laser scan speed or laser power, the powder adhesion on the struts precision and powder type. However, after its publication, other
decreased for Ti6Al4V. Loh et al. [41] found similar results. Other researchers [43,60] reported results beating those values. Through
researchers like Yan et al. [42], Van Bael et al. [43], Pattanayak et al. [44] the advancements in technology, the printable density limits are
also reported strut diameter increase due to powder adhesion. Qiu et changing. On the other hand, the printable density is dependent on
al. [45] studied the effect of laser scanning speed and laser power on the specifications of each process, and in fact, each machine and the
strut size, strut morphology, and surface structures. They compared raw material used. Therefore, there is a need for a tool estimating
two different laser power (150 and 400 W). 400 W resulted in thicker printable density limits of lattice structures based on given condi-
powder adhesion to struts with bigger deviation with respect to the tions. In this work, the authors try to estimate those limits only
laser power of 150 W for AlSi10Mg. Besides, they also found out that by using the most fundamental parameters affecting those limits:
the scanning speeds below 3000 mm/s affected the strut diameter. the machine precision and powder size for different lattice types by
Tsopanos et al. [46] reached similar results. They showed that high leaving the study of other parameters to later studies. Since these are
laser power led to thicker struts of lattices. the most basic factors in terms of printability and some other param-
Van Bael et al. [43] claimed three reasons for the deviations from eters can be modeled with these, the study can be used as the basis
the CAD model: inadequately chosen beam offset, staircase effect, point for further research.
 N. Tanlak et al. / Materials and Design 133 (2017) 549–558 551

Cubic Bcc Fcc Octahedron CubicBcc CubicFcc BccFcc Octet CubicBccFcc

Fig. 1. Unit cell types used.

2. Problem formulation

To create the CAD models of the lattices, a preprocessor of a commercial CAE software [61] is used together with Python scripting language
to enable the algorithms presented in this study to be embodied. The unit cells with cubic outer boundary space are considered. All struts in a
lattice are assumed to have the same circular cross-sectional area. Fig. 1 shows the unit cell types used schematically.
The numerical prediction of the lower and higher printable density limits of lattice structures for AM problem can be stated mathematically
as follows:

Find : r = rl r = rh

Minimizes qlow = q(r, t, L, rmin ) Maximizes qhigh = q(r, t, L, Vp )

For : 1) ∀t ∈ T | T = {Cubic, Bcc, Fcc, Octahedron, CubicBcc, CubicFcc, BccFcc, Octet, CubicBccFcc}
2) ∀L | L ∈ IR +
3) ∀rmin ∈ r̄min | r̄min = {rmin | rmin ∈ IR+ & rmin ≤ L}
4) ∀Vp ∈ V̄p | V̄p = {Vp | Vp ∈ IR3 & 0 < Vp < L3 } (1)

Such That :1) rmin = r ≤ L 1) 0 < r ≤ L

2) nvoid = 1
2
3) prinsmax
on ∂ Y ≥ Amin
4) ∃Ys | Ys = {P ∈ Yv ∧ P, c ≤ rlump }
c : I → IR3 & c ⊂ Yv & I = Iv & min( j(c)
1
) ≥ rlump

where r is the strut radius; rl is the critical strut radius value related to qlow ; qlow is the lowest printable relative density; rh is the critical strut
radius value related to qhigh ; qhigh is the highest printable relative density; L is the cell size of a lattice; nvoid is the number of independent void
volumes inside the cell; rmin is the minimum printable radius which is a machine specific value; t is the lattice type; Vp is the volume of the
sphere circumscribing the biggest powder particle used; rinsmax is the radius of the biggest sphere that can fit into every chunk of void volume;
Amin is the corresponding powder cross-sectional area; Yv is the void volume of a lattice; ∂ Y is the surface of the cube enclosing a lattice; Ys is
a sub-volume of Yv ; P is a point; c is a space curve; is Euclidean distance operator; rlump is the corresponding radius  of Vp ; I is the interval
1
over which c defined; Iv is the interval over which Yv defined; j is the curvature of a curve at every point; min j(c) is the minimum radius
of curvature of the curve c.
More specifically, the main objective of this study is to find the minimum and the maximum printable relative density for a given lattice
type, lattice size, the minimum printable radius, and the powder lump size. The strut radius values should not be smaller than zero and should
not be bigger than the lattice cell size. While finding the minimum and the maximum printable relative density, the void volume should be
monolithic (see the constraint 2 on the right-hand side of Eq. (1)). One may think that a single lattice having disconnected voids can be printable
as long as the voids are connected to one of the faces of the cubic envelope. However, when considering periodic lattices, and considering that
the faces are symmetry planes, these disconnected voids will turn into closed cavities which entrap powder. By constraining the sub-volume
formation, it is ensured that the powder can go out as long as at least one face is open. Hence, the number of void volumes inside a lattice
should be equal to unity.
The powder used in the AM process has a size. Moreover, the unused powder may form lumps during printing, which in turn makes their
effective radius even bigger, e.g. by partial sintering of loose particles, or by spatter from the process [62], landing on the unused powder zones.
Thus, having a monolithic void volume is not enough on its own in order to take the powder out. In Fig. 2, a 2D schematic shows how a powder
particle can go through the void between two struts. In order to ensure that the powder in the void spaces can be evacuated after printing, the
void volume has to have a minimum hole area which is bigger than the corresponding cross-sectional area of the powder lumps on faces of
the lattices (See the constraint 3 on the right-hand side of Eq. (1).).
Besides, there has to be enough space for powder inside the void volume to let the powder go through. To ensure so, a void sub-volume
is required, whose branches’ cross-sectional area has to be equal to the cross-sectional area of the powder lump, at least. A space curve, c, is
employed to define the sub-volume over the same interval of the void volume. Since the powder is a sphere, checking the cross-sectional area
is not enough to prevent powder entrapment. Hence, the curvature of the space curve is also constrained as defined in the constraint 4 on the
right-hand side of Eq. (1). In other words, there should be, at least, a sub-volume, Ys , inside the void volume whose interval is the same as
the interval the void volume defined over so that it touches the faces of the enclosing cube. By adding the distance to the formulation of Ys , a
pipe-like volume whose cross-section is a circle with rlump is created throughout the interval. By checking the curvature of the curve, one will
be sure that the powder sphere will go through the pipe-like sub-volume; therefore the void volume.
On the other hand, machines are subjected to precision constraints, as well. Hence, they have a limit of the smallest printable radius. For
this reason, the smallest printable radius should be taken into account for estimating limits of being printable for a lattice. Therefore, in order
to find qlow for each lattice type, this constraint has to be taken into account (See the constraint-1 of Eq. (1)).
552 N. Tanlak et al. / Materials and Design 133 (2017) 549–558

problem is to find the maximum qhigh using only one constraint at a

time. Then, it is to choose the smaller qhigh among the two solutions
found.
Af is a continuous function that shows the hole area on a face of
lattice, which is a function of relative density, q, that is, in turn, a
function of strut radius, r. To find q(rh ), one needs to find Af (rh ) =
pe2 . By changing the problem to Af (r) − pe2 = 0, one can solve it by
using the bisection method, which does not need to have a derivative
of the function. e is an infinitesimal positive quantity, which is taken
as 10−5 by considering that the precision of the software to define a
geometric entity is 10−6 .
Fig. 2. A 2D schematic for powder passage through the void between two struts. nvoid is a piece-wise function that shows the number of indepen-
dent void volumes inside of the lattice, which is a function of q, which
is, in turn, a function of strut radius, r. Then, one needs to find maxi-
mum r that satisfies nvoid (r = rh ) = 1. A typical change of nvoid along
3. Methodology relative radius, r̄ = Lr , is shown in Fig. 3.
While having a function like above, nvoid , at hand, finding r̄h is
The problem defined in Eq. (1) is, initially, attacked by assum- tricky. First, the function is piecewise discontinuous. Second, it is
ing the powder as infinitesimally thin, because of the fact that the not closed-form function of the free variable, r or r̄. At this juncture,
infinitesimal assumption of the powder gives the limit cases. Besides, a zeroth order method looms as the most convenient alternative,
it gives the location of the critical points which are inside the lattice which does not require calculation of the derivatives. Therefore, the
at which the formation of the disconnected void volumes starts. In bisection method is chosen for the search of r̄h for the case of nvoid =
other words, these critical points are the separation points (bottle- 1 as well. However, the algorithm needs a modification as described
necks) that keep the void volume monolithic. Then using the results in the next section.
of the infinitesimal problem, the problem with finite size powder is
attacked. 3.1.1. Modified bisection method
The bisection method is originally intended to find roots for con-
3.1. Infinitesimally thin powder problem tinuous functions like f(x) = 0 on the interval [xl , xu ]. Nonetheless,
the problem at hand is to find the discontinuity. Turning nvoid as
In this scenario, it is assumed that the powder is infinitesimally nvoid := nvoid − 1 will not help finding r̄h because of the fact that
thin and the machine used has infinite precision. As a result, it can be any values between 0 and r̄h will be the solution of the bisection
considered that any shape is printable, and the powder can be taken method. Therefore, the method is slightly modified to narrow the
out from a hole as long as its area is bigger than zero regardless of search space systematically. In Fig. 4, the flow chart of its modified
however small it is. Through this problem, the maximum theoretical form is illustrated.
printable density can be investigated. In Fig. 4 a and b designates the lower and upper bound of the
Needless to say, the lower limit, only qlow (r, t, L), will be zero in updated interval; c designates the midpoint of the interval. The mod-
this case. Hence, in this case, the qhigh (r, t, L) will be calculated. All in ified bisection method first finds the midpoint of the interval and
all, the problem reduces to the form as follows: then evaluates it. If f(c) = 1, then it updates the lower bound of
the interval by setting the midpoint as a = c; if not, it updates the
Maximize qhigh = q(r, t, L, Amin = pe2 ) upper bound of the interval by setting the midpoint as b = c. This
Subjected to : Af on ∂ Y ≥ pe2 loop continues until the difference between b and a is smaller than
a specified positive small value, e. As can be seen from Fig. 4, the
nvoid = 1 (2)
original algorithm is modified by changing the interval update from
multiplication of function values, f(a)f(c) < 0, to f(c) = 1.
Since the powder is infinitesimal, there is no need to check hole
area in the void volume as long as the void volume is monolithic and
3.2. Finite size powder problem
the void volume can reach the faces. The solution approach to this

3.2.1. Calculation of qlow

The calculation of qlow is independent from qhigh . Therefore, the
problem is reduced to

Minimize q(r, t, L, rmin )

Subjected to : r = rmin (3)

To find qlow , it is only needed to calculate the relative density, q,

based on the specified the minimum printable radius, rmin .
To make the calculations easier, the relative density, q, is calcu-
lated based on the relative radius, r̄. Then a polynomial is fit to the
r
data. By plugging the dimensionless minimum printable radius, min L ,
to the polynomial fitted, g, one can find the lower relative density
limit, qlow .

 
rmin
qlow = g r̄ = , t, L (4)
Fig. 3. Typical number of void volumes versus the relative strut radius in unit cell. L
 N. Tanlak et al. / Materials and Design 133 (2017) 549–558 553

up, in this case, the printable interval of the relative density of lattice
structures will be narrower than the infinitesimal powder problem.
As can be viewed from the results of the infinitesimally thin pow-
der problem (see Table 3 in Section 4.1), there are bottlenecks in the
void volume. To calculate the higher limit, qhigh , one has to find the
locations and the orientations of these bottlenecks, which are the
smallest interconnecting sections that keep the void volume mono-
lithic and determine whether the cross section lets the powder go
through. From the infinitesimally thin powder problem, one already
knows the positions of the bottlenecks. On both sides of these sep-
aration points (bottlenecks), the cross section increases. Therefore,
checking the void volume’s cross-sectional area at these points and
normal to the bottleneck axis will be enough to check the powder
entrapment. The simplified problem can then be stated as follows:

Maximize q(r, t, L, Vp )
Subjected to : nvoid = 1
2
prins max
on ∂ Y ≥ Amin
2
prins max
on C .P . ≥ Amin (5)

where C.P. stands for the cross plane perpendicular through the
bottleneck points.
The removal of the powder is based on the fact that the mono-
lithic void volume in each cell leads to an interconnected monolithic
void in the multi-cell structure. Therefore the removal of the pow-
der in the voids is possible with help of gravity, vibration and air
pressure, as long as the powder particles can pass through the
bottlenecks in the interconnected cell voids (Fig. 5).
In the infinitesimally thin powder case, the void volume is main-
tained together only with point contact. Slightly bigger values than
the critical radius, rh , will result in multiple void volumes and create
twin points on each side of the newly separated void volumes. Using
the CAD model of a lattice for rh (Vp = 43 pe3 ), one may pinpoint the
Fig. 4. Flow chart of modified bisection method for function at hand which is depicted
in Fig. 3. bottlenecks and their orientations. One may partition the void vol-
ume at the desired location and at the desired orientation such that
one can find the smallest cross-section using CAD representation for
any given strut radius, r (see Fig. 3). Finding the biggest radius of a

3.2.2. Calculation of qhigh

Since the smallest printable radius is introduced, the lower limit
of the relative density will be higher. Likewise, since the constraint of
the smallest area in the void volume is imposed for taking the pow-
der out, the higher limit of the relative density will be lower. To sum

Fig. 5. A few hypothetical scenarios for powder and hole cross-sectional area, which Fig. 6. Typical void cross-sectional shape change in terms of strut radius around
powder can be taken out. separation point.
554 N. Tanlak et al. / Materials and Design 133 (2017) 549–558

circle that can be inscribed on the smallest section of the void vol- Table 1
rh
Critical relative radius, , values corresponding to upper relative density limits, qhigh .
ume, rinsmax , is yet another optimization problem which can be stated L
rh
as follows: Lattice type L qhigh

Cubic 0.5 0.9420

Bcc 0.3553 0.9886
Maximize rins
Fcc 0.2887 0.9073
Such that : Y = {(x, y) | x, y ∈ IR ∧ x2 + y2 ≤ f (x, y)} Octahedron 0.2024 0.5849
Cubic Bcc 0.2588 0.8902
X = {(x, y) | x, y ∈ IR ∧ (x − xG )2 + (y − yG )2 ≤ rins
2
∧ (x, y) ∈ Y} Cubic Fcc 0.2071 0.7365
(6) Bcc Fcc 0.2142 0.9065
Octet 0.2041 0.9073
Cubic Bcc Fcc 0.2071 0.9660
where rins is the radius of an inscribed circle on a arbitrary cross
section through the void volume of a lattice, Y; IR is real numbers;
f(x, y) is the defining function of boundary of the critical cross- depicted in Fig. 7 for the lattice types given. As can be seen from the
sectional hole shape of a lattice; (xG , yG ) is the central location of the figure, qhigh does not depend on the lattice size, L.
inscribed circle. In Table 1, upper density limits, qhigh , and the corresponding crit-
As depicted in Fig. 6, the cross-sectional shape of the void around ical relative radius values, rh , are given. The Octahedron has a very
the separation point (S.P.) changes like nested polygons depending low upper limit of printable density in comparison to the others. The
on the strut radius, r, since the struts of a lattice are oriented always reason of getting low relative printable density limit for Octahedron
the same and they have the same radius. Hence, to find the rinsmax is that Octahedron has struts surrounding an inner cavity. When the
reduces to lower rh (Vp = 43 pe3 ) in Section 3.1 as much as the mini- struts radius increases, the central cavity easily gets disconnected
mum allowed circular hole area specified by rlump . Mathematically, it from outer cavities, which are still quite large, resulting in a low rela-
can be stated as the following equation: tive density limit. On the other hand, Bcc has the highest limit. Here,
    for Bcc lattice, the struts stem from the center and for a growing strut
4 3 4 radius, the cavities remain connected almost up to the disappearance
rh Vp = prlump = rh Vp = pe3 − rlump (7)
3 3 of the cavities.
In Table 3, the corresponding solid and void volumes are graphi-
In dimensionless form, qhigh can be found using the following cally shown from the different point of views for all lattice types. As
equation: can be seen from Table 3, for Cubic, CubicFcc, and CubicBccFcc, the
bottlenecks are on the face of the corresponding lattice. However, for
⎛   ⎞
rh Vp = 43 pe3 the rest, the bottleneck is inside of the corresponding lattice.
rlump
qhigh = g ⎝ − ⎠ (8) Since q = q(r̄) relationship will be used in all limit calcula-
L L
tions (not only for infinitesimal powder assumption but also for
finite powder assumption as explained in Section 3.2.2), q versus
where g is the polynomial function fitted to q = q(r̄). r̄ is plotted in Fig. 8 for all lattice types. The corresponding data
of the printable portion, under infinitesimal powder assumption, of
q = q(r̄) (see the right plot of Fig. 8) is fitted to cubic polynomials
4. Results
in order to give a practical tool for designers (see Table 2). In other
words, the polynomials are valid until the critical relative radius val-
4.1. Infinitesimally thin powder problem
ues specified for every lattice type in Table 1. The polynomial fit
is confined to the printable zone due to the fact that achieving the
After calculating the corresponding rh values for each constraint
intended density is not possible outside of it.
separately (see Eq. (2)), the smaller one is taken among the two and
by using that value the corresponding relative density is calculated
as explained before. The resultant printable upper-density limits are 4.2. Finite size powder problem
In Fig. 9, both lower and upper limits of density are plotted by
assuming rmin = 0.2 mm and rlump = 0.2 mm. These limits are sen-
sitive to the cell size. As the cell size increases, the limits approach to
the infinitesimal case. When it gets smaller, the two curves approach
to each other, which means the printable density range gets smaller,
until the two curves cross each other, which means only one den-
sity value is printable. Beyond that crossing point, no lattice having
smaller density value is printable.

Table 2
Coefficients of the polynomials, g, fitted to r̄ for q.

Lattice type r̄3 r̄2 R2

Cubic −11.3138 9.4248 0.999

Bcc −39.1919 21.7656 0.999
Fcc −54.6277 26.6574 0.999
Octahedron −61.8194 26.6573 0.999
CubicBcc −69.1653 31.1904 0.999
CubicFcc −91.3146 36.0823 0.999
BccFcc −133.8197 48.4230 0.999
Octet −154.5098 53.3147 0.999
CubicBccFcc −170.5703 57.8478 0.999

Fig. 7. Upper limits for several lattice types based on the two criteria together. R2 is the coefficient of determination.
 N. Tanlak et al. / Materials and Design 133 (2017) 549–558 555

Table 3
Solid and void volumes of upper thresholds using infinitesimal powder assumption.

Fig. 9 shows the curves for only rmin = 0.2 mm and rlump = 0.2 Based on the crossing point of the upper and lower limits, there is
mm. However, the possible combinations of rmin and rlump values are a relationship like the following:
endless. Hence, it is necessary to build a dimensionless graph such
that one can find the corresponding printable range for given rmin  
r rlump
rmin rh Vp = 43 pe3 rlump
and rlump . In Fig. 10, such a graph is given, whose min
L and L are = − (9)
plotted on the same axis. However, completely independent values L L L
can be selected for them. The relative density values on the graph are
limited up until the corresponding the highest relative density, qhigh , Using this equation, one can calculate the other r value if one
of each lattice. In order to clarify how to use the figure, the print- is given. Fig. 11 shows the boundary of printable zone in terms of
r rlump
able range is found using L = 5 mm, rmin = 0.2 mm, and rlump = dimensionless parameters min L and L . The regions under the lines
0.1 mm for Octet lattice. The resultant range is [0.075, 0.843]. The are the printable zones in Fig. 11. If a combination is on the boundary
corresponding values are marked in Fig. 10. then that means only one value can be printable.
556 N. Tanlak et al. / Materials and Design 133 (2017) 549–558

Fig. 8. Relative density, q, versus relative radius, r̄, of the struts for some lattice types: left: full range; right: the printable range for the infinitesimally thin powder.

The minimum printable feature radius based on strut inclination the realized and planned diameter of struts is a function of strut incli-
can be found either in the literature or by experiment. In literature, nation. Therefore, one can expect a lattice having different diameter
there are several minimum printable radius values in terms of strut struts after printing even if a lattice having equal diameter struts are
inclination. This constraint can be treated via the machine precision intended. To guess the printable limits at such conditions, one needs
(the minimum printable radius) terms in this study. That value can be to estimate radius difference based on inclination angle. After taking
plugged into the equation (Eq. (4)). Moreover, the manufacturability the biggest difference expected, one can hypothetically assume a big-
of horizontal struts was discussed in the literature. Mazur et al. [57] ger powder lump as much as the biggest radius difference foreseen
reported that by using SLM, they achieved to print horizontal struts to find the highest printable limit.
up to 3 mm if they are supported at both ends. In the current state- Despite having useful estimations of the printable density range
of-the-art, the lattices having horizontal struts (e.g. Cubic, CubicBcc, based on the assumptions made here, there is a need for a further
CubicFcc, CubicBccFcc) are not printable unless they are smaller than study which includes other parameters for manufacturability of lat-
3 mm for Ti6Al4V. Hence, the reader must be aware of this limita- tice structures. One can say that the real printable density range may
tion (can be different for every powder material) when they are using be smaller than the one found here as explained in Eq. (10).
Fig. 10.
Powder adhesion problem can also be taken into the account
by the powder lump size. As an example powder lump size can be
qreal corr
low = qlow + qlow = qlow (r̄ + r̄corr )
taken as a couple of times bigger than the biggest powder size used.
The exact amount has to be investigated for every material. Zhang qreal corr
high = qhigh − qhigh = qhigh (r̄ − r̄corr ) (10)
et al. [39] suggested using three times bigger pore sizes. Further-
more, as reported by several authors [52–58], the difference between

Fig. 9. Thresholds of lattice structures using rmin = 0.2 mm and rlump = 0.2 mm. Fig. 10. Printable thresholds of lattice structures in dimensionless form.
 N. Tanlak et al. / Materials and Design 133 (2017) 549–558 557

parameters for exact limits. It is easily foreseeable that the real print-
able density range will be narrower than the ones found in this
work. Furthermore, finding the optimized process parameters or the
advancement of technology will approach the printable density lim-
its to the ones found in this paper. Therefore, this study can be taken
as a basis point for further investigations.

Acknowledgment

The work leading to this publication has been funded by

“M3AMCAE” project, which fits in the MacroModelMat (M3) research
program funded by SIM (Strategic Initiative Materials in Flanders)
and VLAIO (Flanders Innovation & Entrepreneurship).
Siemens PLM Softwares and Materialise are gratefully acknowl-
edged for being partners in this project.
The second author acknowledges the sabbatical research fellow-
ship awarded by the Mexican Research Council (CONACYT) and the
invitation by UGent and its partners to collaborate in the M3AMCAE
project.

References

[1] T. Bitzer, Honeycom Technology, Cahpmann & Hall. 1997.

[2] D.J. Sypeck, H.N.G. Wadley, Cellular metal truss core sandwich structures,
Advanced Engineering Materials 4 (10). (2002)
[3] F. Cote, V.S. Deshpande, N.A. Fleck, A.G. Evans, The compressive and shear
responses of corrugated and diamond lattice materials, International Journal of
Solids and Structures 43 (20) (2006) 6220–6242.
[4] G.W. Kooistra, H.N. Wadley, Lattice truss structures from expanded metal
rmin rlump sheet, Materials & Design 28 (2) (2007) 507–514.
Fig. 11. Boundary of printable zones in terms of L and L . [5] J. Cochran, K. Lee, D. McDowell, T. Sanders, B. Church, J. Clark, B. Dempsey,
A. Hayes, K. Hurysz, T. McCoy, J. Nadler, R. Oh, W. Seay, B. Shapiro, Low den-
where qreal
low
, qreal
high
are the real lower and upper printable limits, sity monolithic metal honeycombs by thermal chemical processing, Fourth
Conference on Aerospace Materials, Processes and Environmental Technology,
respectively; qcorr are the corrections terms. Huntsville, Alabama, 2000,
In order to fabricate lattice structures with precise dimensions, it [6] D.J. Sypeck, H.N.G. Wadley, Multifunctional microtruss laminates: textile syn-
is important to select appropriate processing conditions. Therefore, thesis and properties, Journal of Materials Research 16 (3) (2001) 890–897.
finding optimized parameters will approach the printable limits to [7] D.T. Queheillalt, V.S. Deshpande, H.G. Wadley, Truss waviness effects in cellular
lattice structures, Journal of Mechanics of Materials and Structures 2 (9) (2007)
the ones in this study. 1657–1675.
[8] V.S. Deshpande, N.A. Fleck, Collapse of truss core sandwich beams in 3-point
bending, International Journal of Solids and Structures 38 (36–37) (2001)
5. Conclusion 6275–6305.
[9] J. Wallach, L. Gibson, Mechanical behavior of a three-dimensional truss
The main and novel conclusion of this study is that even if ideal material, International Journal of Solids and Structures 38 (40–41) (2001)
7181–7196.
conditions are assumed, there is a maximum printable density limit [10] S. Chiras, D. Mumm, A. Evans, N. Wicks, J. Hutchinson, K. Dharmasena,
for each lattice type. Among the lattice types studied, Octahedron H. Wadley, S. Fichter, The structural performance of near-optimized truss
lags far behind the other lattice types on the highest printable limit core panels, International Journal of Solids and Structures 39 (15) (2002)
4093–4115.
while Bcc has the widest range. In the Octahedron lattice, the void [11] J. Wang, A.G. Evans, K. Dharmasena, H.N.G. Wadley, On the performance of truss
enclosed by the central void quickly get separated from the relatively panels with Kagome cores, International Journal of Solids and Structures 40
large remaining voids in the cell corners. (25) (2003) 6981–6988.
[12] W.X. Wang, D.M. Luo, Y. Takao, K. Kakimoto, New solution method for homog-
The printable range is highly dependent on the cell size as enization analysis and its application to the prediction of macroscopic elastic
well. When unit cell size increases, the printable range widens constants of materials with periodic microstructures, Computers & Structures
and approaches to the limit case of infinitesimal powder case. On 84 (15–16) (2006) 991–1001.
[13] A. Askar, A.S. Cakmak, A structural model of a micropolar continuum, Interna-
the other hand, the printable range gets narrower when the pow-
tional Journal of Engineering Science 6 (10). (1968)583–589.
der lump size increase and the machine precision exacerbates (i.e. [14] Z. Bazant, M. Christensen, Analogy between micropolar continuum and grid
increasing rmin ). Dimensionless plots are created to enable finding frameworks under initial stress, International Journal of Solids and Structures
the printable range for a given cell type, cell size, the minimum 8 (3) (1972) 327–346.
[15] J.M. Guedes, N. Kikuchi, Preprocessing and postprocessing for materials based
printable radius and lump size. on the homogenization method with adaptive finite-element methods, Com-
This approach for finding the printable density limits is valid for puter Methods in Applied Mechanics and Engineering 83 (2) (1990) 143–198.
the lattices having a circular cross-section, and the same radius for [16] I.G. Masters, K.E. Evans, Models for the elastic deformation of honeycombs,
Composite Structures 35 (4) (1996) 403–422.
all their struts. For the lattices having cross-sections other than cir- [17] L. Gibson, M. Ashby, Cellular Solids: Structure and Properties, Cambridge
cle and having the same dimension for all their struts, this approach University Press. 1997.
can also be used by assuming the strut radius, r, to be equal to the [18] B. Hassani, E. Hinton, A review of homogenization and topology optimization
I - homogenization theory for media with periodic structure, Computers &
radius of the circumscribing circle of the cross-section. But, in that
Structures 69 (6) (1998) 707–717.
case, this approach can be a conservative estimation of the relative [19] B. Hassani, E. Hinton, A review of homogenization and topology optimization II
density limits. Yet, it needs verification. For better approximation, - analytical and numerical solution of homogenization equations, Computers &
the problem defined in Eq. (6) has to be solved. Structures 69 (6) (1998) 719–738.
[20] J. Chen, Y. Huang, M. Ortiz, Fracture analysis of cellular materials: a strain
Although printable limits based on only two parameters have gradient model, Journal of the Mechanics and Physics of Solids 46 (5) (1998)
been found in this work, further studies are needed on other process 789–828.
558 N. Tanlak et al. / Materials and Design 133 (2017) 549–558

[21] X.L. Wang, W.J. Stronge, Micropolar theory for two-dimensional stresses [43] S.V. Bael, G. Kerckhofs, M. Moesen, G. Pyka, J. Schrooten, J. Kruth, Micro-C-
in elastic honeycomb, Proceedings of the Royal Society of London A: T-based improvement of geometrical and mechanical controllability of selec-
Mathematical, Physical and Engineering Sciences 455 (1986) (1999) tive laser melted Ti6Al4V porous structures, Materials Science and Engineering:
2091–2116. A 528 (24) (2011) 7423–7431.
[22] R. Christensen, Mechanics of cellular and other low-density materials, Interna- [44] D.K. Pattanayak, T. Matsushita, H. Takadama, A. Fukuda, M. Takemoto, S.
tional Journal of Solids and Structures 37 (1–2) (2000) 93–104. Fujibayashi, K. Sasaki, N. Nishida, T. Nakamura, T. Kokubo, Fabrication of
[23] E.W. Andrews, G. Gioux, P. Onck, L.J. Gibson, Size effects in ductile cellu- bioactive porous Ti metal with structure similar to human cancellous bone by
lar solids. Part II: experimental results, International Journal of Mechanical selective laser melting, Bioceramics Development and Applications 1 (2011)
Sciences 43 (3) (2001) 701–713. 1–3.
[24] V. Deshpande, N. Fleck, M. Ashby, Effective properties of the octet-truss lat- [45] C. Qiu, S. Yue, N. Adkins, Influence of processing conditions on strut structure
tice material, Journal of the Mechanics and Physics of Solids 49 (8) (2001) and compressive properties of cellular lattice structures fabricated by selective
1747–1769. laser melting, Mat Sci Eng A 628 (2015) 188–197.
[25] W.E. Warren, E. Byskov, Three-fold symmetry restrictions on two-dimensional [46] S. Tsopanos, R.A.W. Mines, S. McKown, Y. Shen, W.J. Cantwell, W. Brooks, C.J.
micropolar materials, European Journal of Mechanics a-Solids 21 (5) (2002) Sutcliffe, The influence of processing parameters on the mechanical properties
779–792. 1873–7285. of selectively laser melted stainless steel microlattice structures, J Manuf Sci
[26] A.J. Wang, D.L. McDowell, In-plane stiffness and yield strength of periodic metal Eng 132 (041011). (2010)
honeycombs, Journal of Engineering Materials and Technology 126 (2) (2004) [47] D. Wang, Y. Yang, R. Liu, D. Xiao, J. Sun, Study on the designing rules and pro-
137–156. http://dx.doi.org/10.1115/1.1646165. cessability of porous structure based on selective laser melting (SLM), Journal
[27] R.S. Kumar, D.L. McDowell, Generalized continuum modeling of 2-D periodic of Materials Processing Technology 213 (10) (2013) 1734–1742.
cellular solids, International Journal of Solids and Structures 41 (26) (2004) [48] E. Abele, H.A. Stoffregen, K. Klimkeit, H. Hoche, M. Oechsner, Optimisation
7399–7422. of process parameters for lattice structures, Rapid Prototyping Journal 21 (1)
[28] Z. Fang, Asymptotic homogenization and numerical implementation to predict (2015) 117–127.
the effective mechanical properties for electromagnetic composite conductor, [49] E. Sallica-Leva, A. Jardini, J. Fogagnolo, Microstructure and mechanical behavior
Journal of Composite Materials 38 (16) (aug 2004) 1371–1385. of porous Ti-6Al-4V parts obtained by selective laser melting, Journal of the
[29] Z. Fang, B. Starly, W. Sun, Computer-aided characterization for effective Mechanical Behavior of Biomedical Materials 26 (2013) 98–108.
mechanical properties of porous tissue scaffolds, Computer-Aided Design 37 [50] K. Hazlehurst, C. Wang, M. Stanford, Evaluation of the stiffness characteristics
(1) (2005) 65–72. of square pore CoCrMo cellular structures manufactured using laser melting
[30] D. Mohr, Mechanism-based multi-surface plasticity model for ideal truss lat- technology for potential orthopaedic applications, Materials Design 51 (2013)
tice materials, International Journal of Solids and Structures 42 (11–12) (2005) 949–955.
3235–3260. [51] T.B. Kim, S. Yue, Z. Zhang, E. Jones, J.R. Jones, P.D. Lee, Additive manufac-
[31] M.F. Ashby, The properties of foams and lattices, Philosophical Transactions tured porous titanium structures: through-process quantification of pore and
of the Royal Society A-Mathematical Physical and Engineering Sciences 364 strut networks, Journal of Materials Processing Technology 214 (11) (2014)
(1838) (2006) 15–30. 2706–2715.
[32] M. Doyoyo, J.W. Hu, Plastic failure analysis of an auxetic foam or inverted strut [52] M. Leary, MaciejMazur, J. Elambasseril, M. McMillan, T. Chirent, Y. Suna, M.
lattice under longitudinal and shear loads, Journal of the Mechanics and Physics Qian, M. Eastona, M. Brandt, Selective laser melting (SLM) of AlSi12Mg lattice
of Solids 54 (7) (2006) 1479–1492. structures, Materials and Design 98 (2016) 344–357.
[33] C.C. Foo, G.B. Chai, L.K. Seah, Mechanical properties of Nomex material and [53] J.C. Fox, S.P. Moylan, B.M. Lane, Effect of process parameters on the sur-
Nomex honeycomb structure, Composite Structures 80 (4) (2007) 588–594. face roughness of overhanging structures in laser powder bed fusion additive
[34] S. Demiray, W. Becker, J. Hohe, Numerical determination of initial and subse- manufacturing, Procedia CIRP 45 (2016) 131–134.
quent yield surfaces of open-celled model foams, International Journal of Solids [54] A. Triantaphyllou, C.L. Giusca, G.D. Macaulay, F. Roerig, M. Hoebel, R.K. Leach, B.
and Structures 44 (7–8) (2007) 2093–2108. Tomita, K.A. Milne, Surface texture measurement for additive manufacturing,
[35] K. Sab, F. Pradel, Homogenisation of periodic Cosserat media, IJCAT 34 (1) Surface Topography: Metrology and Properties 3 (2) (2015) 024002.
(2009) 60. [55] P. Drescher, T. Reimann, H. Seitz, Investigation of powder removal of net-struc-
[36] A.L. Kalamkarov, I.V. Andrianov, V.V. Danishevs’kyy, Asymptotic homogeniza- tured titanium parts made from electron beam melting, International Journal
tion of composite materials and structures, Applied Mechanics Reviews 62 (3). of Rapid Manufacturing Vol. 4 (2). (2014)
(2009) [56] M. Suard, G. Martin, P. Lhuissier, R. Dendievel, F. Vignat, J.-J. Blandin, F.
[37] C. Yan, L. Hao, A. Hussein, P. Young, D. Raymont, Advanced lightweight 316L Villeneuve, Mechanical equivalent diameter of single struts for the stiffness
stainless steel cellular lattice structures fabricated via selective laser melting, prediction of lattice structures produced by electron beam melting, Additive
Materials & Design 55 (2014) 533–541. 1873–4197. Manufacturing 8 (2015) 124–131.
[38] T. Zhao, C. Chen, Z. Deng, Elastoplastic properties of transversely isotropic [57] M. Mazur, M. Leary, S. Sun, M. Vcelka, D. Shidid, M. Brandt, Deformation and
sintered metal fiber sheets, Materials Science and Engineering: A 662 (2016) failure behaviour of Ti-6Al-4V lattice structures manufactured by selective
308–319. laser melting (SLM), Int J Adv Manuf Technol 84 (2016) 1391–1411.
[39] S. Zhang, Q. Wei, L. Cheng, Effects of scan line spacing on pore characteristics [58] V. Weißmann, P. Drescher, R. Bader, H. Seitz, H. Hansmann, N. Laufer,
and mechanical properties of porous Ti6Al4V implants fabricated by selective Comparison of single Ti6Al4V struts made using selective laser melting and
laser melting, Materials Design 63 (2014) 185–193. electron beam melting subject to part orientation, Metals 7 (3) (2017) 91–113.
[40] S. Sing, W. Yeong, F. Wiria, B. Tay, Characterization of titanium lattice struc- [59] O. Rehme, C. Emmelmann, Rapid manufacturing of lattice structures with selec-
tures fabricated by selective laser melting using an adapted compressive test tive laser melting, Proceedings of SPIE 6107: 61070K. 1-61070K. 12., 2006,
method, Exp Mechan 56 (2015) 735–748. [60] L. Hao, D. Raymont, C. Yan, A. Hussein, P. Young, Design and additive manu-
[41] L.-E. Loh, C.-K. Chua, W.-Y. Yeong, J. Song, M. Mapar, S.-L. Sing, Z.-H. Liu, D.-Q. facturing of cellular lattice structures, Innovative Developments in Virtual and
Zhang, Numerical investigation and an effective modelling on the selective Physical Prototyping (2011) 249–254.
laser melting (SLM) process with aluminium alloy 6061, International Journal [61] Abaqus CAE User’s Guide.
of Heat and Mass Transfer 80 (2015) 288–300. [62] D. Wang, S. Wu, F. Fu, S. Mai, Y. Yang, Y. Liu, C. Song, Mechanisms and character-
[42] C. Yan, L. Hao, A. Hussein, D. Raymont, Evaluations of cellular lattice structures istics of spatter generation in {SLM} processing and its effect on the properties,
manufactured using selective laser melting, International Journal of Machine Materials & Design 117 (2017) 121–130.
Tools & Manufacture 62 (2012) 32–38.