Journal of Quantitative Spectroscopy & Radiative Transfer: A.V. Gusarov
Journal of Quantitative Spectroscopy & Radiative Transfer: A.V. Gusarov
Journal of Quantitative Spectroscopy & Radiative Transfer: A.V. Gusarov
a r t i c l e i n f o a b s t r a c t
Article history: Laser powder-bed processing is widely employed in 3D printing. The technological processes strongly de-
Received 22 June 2020 pend on such effective radiative-transfer properties of powder beds as the absorptance and the radiation
Revised 2 October 2020
penetration depth. However, it is still not completely clear what are the principal factors influencing these
Accepted 3 October 2020
properties and what are the possible ranges of their variation. For theoretical investigation of the radiative
Available online 6 October 2020
transfer, a ray-tracing method is developed where the effective properties of powder beds are expanded
Keywords: in series of the reflectance of the solid-phase surface. Once the coefficients of the power series are cal-
Absorptance culated, they can be used for any combination wavelength/material thus facilitating parametric analysis.
Cubic structure Powder beds are modeled by equal spheres packed in regular cubic-symmetry structures of various den-
Normal-directional reflectance sity. The calculated radiation penetration depth can vary from few particle diameters for close-packed
Normal-hemispherical reflectance structures with moderate solid-phase absorptance to few tens of particle diameters for less dense power
Packed spheres
beds of a highly reflecting material. The effective absorptance of the powder bed is revealed to decrease
Powder bed
Radiative transfer
significantly with the packing density. The calculated dependence of the powder-bed absorptance versus
Ray tracing the solid-phase absorptance and the density is validated by the known experimental data. The calculated
3D printing angular distribution of the radiation reflected by a powder bed slightly deviates from the cosine-like dis-
tribution corresponding to the diffuse reflection law. It is validated by the known experimental data. A
modified model of equivalent medium is proposed to approximate the obtained ray tracing results by
analytic expressions.
© 2020 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.jqsrt.2020.107366
0022-4073/© 2020 Elsevier Ltd. All rights reserved.
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
ter/wavelength ratio allows one to use the ray optics for explaining
light absorption in the powder beds. The most rigorous theoretical
results were obtained by MCRT.
Wang et al. [14,15] developed an MCRT model to calculate ab-
sorptance of an array of randomly distributed non-intersecting
polydisperse spheres. For the specular reflection law, this model
satisfactorily agrees with experimental data. Boley et al. [16] cal-
culated the absorptance of three model arrays of spheres: a bed of
two close-packed layers of equal spheres and two beds of polydis-
perse spheres with a Gaussian and a bimodal distributions. They
revealed a dependence of the effective absorptance of the powder
bed on its structure model accepted for calculations, which is es-
pecially important for highly reflective materials. In view of this Fig. 1. Reflection and transmission of collimated incident radiation by an array of
fact, it is not surprising that the values of the effective absorptance opaque specularly-reflecting spheres. The numbers near the exiting rays designate
reported by Boley et al. are considerably less than the correspond- the number of reflections from spheres.
ing values reported by Wang et al. Zhang et al. [10] compared the
close-packed structure of equal spheres with denser structures ob-
tained by filling the pores between the spheres with additional recently Kovalev et al. [21] reported the ray tracing results indi-
smaller spheres. Their MCRT calculations did not found a clear de- cating that incident radiation penetrates into packed beds to the
pendence of the effective absorptance on the density. depth of about few particle sizes. Gusarov and Kruth [17] explained
Another theoretical approach is the model of effective absorb- the penetration in the framework of the RTE model and found that
ing scattering medium that reduces the problem to the radiative the principal parameter responsible for the penetration depth is
transfer equation (RTE). The parameters of this model can be esti- the specific surface. These results suggest that the arrangement of
mated from the parameters of the powder bed. Gusarov and Kruth particles on the surface of the packed bed is also important for
[17] solved a half-space problem for RTE and found the effective reflection and penetration. Zarrouati et al. [22] reported consid-
absorptance of the powder bed as function of the absorptance of erable periodic variations of packing density near walls of a con-
the flat surface. In this approximation, the effective absorptance is tainer with a packed bed. Similar irregularities are possible near
independent of powder bed density, which contradicts the recent the top surface too.
MCRT results [16]. Surprisingly, the RTE model agrees well with The gravity force increases as the particle size cubed while the
experimental data, see [18]. adhesion Van der Waals force between particles is proportional to
Not only the effective absorptance but also the penetration pro- the particle size [23]. That is why the adhesion force is negligible
file and volumetric absorption affect laser processing of powder in packed beds of macroscopic particles and becomes comparable
beds. Transmittance of powder beds was measured [19] to estimate with or greater than the gravity force in powder beds. This fact
volumetric light absorption. Both the MCRT and RTE models indi- essentially determines the lower density of powders [12,20]. Ex-
cate that the powder bed structure significantly influence the pen- perimentally measured packing density of powders [9] varies de-
etration depth. Normal-directional reflectance measured in [9] can pending on the material, particle size and shape. Generally, the
provide additional information about light interaction with pow- density is less for finer powders and powders of irregular parti-
der beds. However, the analysis of these data using an RTE model cles [9]. Prediction of powder bed properties in a wide range of
[9] did not result in clear conclusions. It was probably due to re- packing density is a challenge. The measured reflectance of powder
strictions of the model. beds gradually increases with density [9]. This tendency has not
The weak point of the existing ray-tracing studies of radiative been confirmed theoretically. Thus, the RTE model proposes that
transfer in powder beds is modeling the realistic structure of the the effective reflectance is independent of the density [17]. How-
beds. There are known works aiming to simulate mechanical de- ever, comparison with ray tracing [16,21] shows that this model
position of powders in the conditions typical for industrial laser systematically underestimates the reflectance.
powder-bed machines, for example [12], but the known theoretical In summary, ray tracing appears to be the most reliable theo-
investigations of radiation transfer do not use such realistic mod- retical approach to radiative transfer in the powder beds but prin-
els. cipal difficulties arise in modeling the packing structure. There is
Thus, Wang and Kruth [14] generated their packed-bed mod- no yet a systematic theoretical study of the packing density in-
els by sequential random placing of particles until there is a free fluence in the range important for laser powder-bed applications.
space for a particle. They did not communicate the resulted pack- The present work proposes to analyze a set of regular structures
ing density. Boley et al. [16] and Kovalev and Gusarov [20] used of spheres within a wide range of packing density. The principal
the sequential rain model for random packing. Boley et al. did not questions to answer concern the difference in radiative transfer be-
communicate the packing density too. Kovalev and Gusarov con- tween high-symmetry lattices of packed spheres and the influence
sidered the adhesion Van der Waals force in addition to the gravity of the structure orientation relative to the packed bed surface.
force. Therefore, the packing density varied with the particle size
because of varying the adhesion/gravity force ratio [20]. However, 2. Ray tracing
one cannot be sure if such variation is realistic because the rain
model considerably differs from the blade spreading typically used The great particle size/wavelength ratio justifies using ray trac-
in laser powder-bed processes. Boley et al. [16] and Zhang et al. ing. In a packed bed, the ray tracing is not applicable in thin gaps
[10] studied close-packed regular structures of spheres but their between particles around contact points. The thickness of such a
density was significantly greater than typical powder densities ex- gap is of the order of the wavelength. Therefore, the cross section
perimentally measured by Gusarov et al. in the range from 0.33 to area of the gap is much less than the projected area of a particle.
0.67 [9]. Thus, the relative error introduced is expected to be low.
Kovalev and Gusarov [20] found a considerable anisotropy in The example of Fig. 1 shows that one can distinguish inci-
their model packed beds. It is not clear if one can expect a signifi- dent rays subjected to one, two etc. reflections from a non-convex
cant anisotropy in realistic powder beds. Wang and Kruth [14] and opaque object. A novel ray tracing technique is proposed. It is ap-
2
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Table 1
Structure models of powder beds.
plicable in assumption that reflectivity ρ is independent of the in- dense. However, close packing is believed to be favorable to im-
cidence angle. In this case, every reflection of a ray attenuates ray’s prove the laser powder bed fusion process and is considered in
energy by a factor of ρ . Then, the energy becomes ρ n of the initial the ray tracing simulation of Boley et al. [16] and Zhang et al. [10].
energy after n reflections. Every ray is traced until it is reflected The present study includes the FCC and HCP structures to compare
back or passes forward through the object analyzed. Then, the fol- the results with the previous works. Besides, one cannot exclude
lowing series give the effective reflectance R and transmittance T another application domain where the theoretical results on FCC
of the object: and HCP would be demanded.
∞
For each structure, surfaces parallel to several low-index planes
R rn
= ρn, (1) are tested as shown in Figs. 2 and 3. These figures also show the
T tn calculation domains for ray tracing chosen for each combination
n=0
structure/reflecting surface. The optimal calculation domain is a
where rn is the fraction of incident rays turned back after n re-
prism bounded by vertical planes in 3D (see Fig. 2) and a rect-
flections and tn the fraction of incident rays passed forward after
angle in 2D (see Fig. 3). Everywhere excluding DI(110), the nearest
n reflections from the object. Here r0 = 0 by definition because no
vertical mirror planes of the given structure are chosen as the lat-
ray can turn back without a reflection while t0 can be positive.
eral boundaries of the calculation domain. In the case of DI(110)
Of course, all the incident rays are directed on the object. How-
shown in Fig. 2(l), one cannot find the necessary number of mir-
ever, the object may have through holes. Therefore, transmittance
ror planes perpendicular to the (110) plane. Therefore, two paral-
through the object is possible without reflections.
lel mirror planes and two parallel planes perpendicular to them
The assumption that ρ is independent of the incidence angle, is
bound the calculation domain from the lateral sides. The distance
acceptable for metals [24]. The advantage of such an approach is
between the no-mirror planes equals the lattice period in the di-
that the ray tracing computation is independent of the value of
rection normal to them.
reflectivity ρ . Once factors rn and tn are calculated by ray trac-
Vertical incident rays are generated on the top of the calcula-
ing, they can be applied to any combination material/wavelength
tion domain. The mirror conditions are imposed on the mirror lat-
defining the value of ρ . This offers the generality of results nec-
eral boundaries. The periodic conditions are imposed on the non-
essary for the present study. Numerical ray tracing computation is
mirror lateral boundaries. The spheres reflect the rays specularly. A
stopped if a ray is not reflected back or transmitted forward after
ray is traced until the number of reflections from the spheres at-
a predefined great number of reflections N. Therefore, the series
tains a predefined number N or it exits through the top or bottom
in Eq. (1) are truncated in numerical calculation. Partial sums of N
boundary of the calculation domain. Fractions rn and tn are calcu-
terms approach R and T with the confidence interval estimated as
lated as the ratio of rays exited from the top and bottom bound-
aries, respectively, after n reflections from the spheres to the to-
R
N
rn
0< − ρ n ≤ EN ρ N+1 , (2) tal number of incident rays. The number of incident rays for each
T tn calculation is of the order of 1010 . Appendix A presents analytic
n=0
calculation of factors r1 and r2 for the close-packed row of circles
where
in two dimensions (see Fig. 3(a) or (c)) that were used to test the
N
N algorithm for numerical calculations.
EN = 1 − rn − t n, (3)
n=1 n=0
3. Normal-hemispherical reflectance and absorptance
is the fraction of rays neither turned back nor passed forward af-
ter N reflections. The left inequality of Eq. (2) follows from the fact Fig. 4 shows the first 12 reflection factors rn (a) numerically
that all terms of the series in Eq. (1) are positive. In the right in- calculated by ray tracing and the effective absorptance (b) of the
equality of Eq. (2), fraction EN given by Eq. (3) is the upper bound 2D structures. It has to be noted that there are no through gaps
for both the factors rN+1 and tN+1 . in such structures. Therefore, transmittance T = 0 and absorptance
Table 1 lists the studied periodic structures with their princi- A = 1 − R, where R is calculated by Eq. (1). The effective absorp-
pal characteristics, the coordination number and the relative den- tance is plotted versus the absorptance of a flat surface a = 1 − ρ
sity. Typical powder beds used in laser processing has the density in the range from 0.01 to 1. For the close-packed row of circles
in the range from 0.3 to 0.67 [9]. The density depends on the in- corresponding to Sqr(10) or Hex structures, the first two reflec-
terplay of the gravity and the Van der Waals adhesive forces and tion factors calculated by ray-tracing (black circles in Fig. 4(a))
varies with the powder material and particle size [21]. The regimes coincide with the corresponding values calculated analytically in
of mechanical spreading of powder also influence the resulting Appendix A (crosses). Analytical calculations become very cumber-
powder bed density [12]. The mentioned range of density is ap- some for three and more reflections. Therefore, only the cases of
proximately covered by the DI, SC, and BCC structures, see Table 1. one and two reflections are considered. The value of the reflec-
The close packed FCC and HCP structures are considerably more tion factor rn depends on the kind of 2D structure, especially in
3
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 2. Beds of equal spheres packed into regular structures with the top horizontal surface parallel to the crystallographic plane indicated by the Miller indices: (a)–(c) FCC;
(d) HCP; (e)–(g); (h)–(j) SC; (k)–(m) DI. Three views are shown for each powder bed model: Top view on the reflecting surface (right); Axonometric view with the reflecting
surface on the top (middle); Calculation domain for ray tracing (left).
the range of n from 3 to 10 (see. Fig. 4(a)). Nevertheless, the ef- structure and ABCABC ... for the FCC one (see Fig. 2(c) and (d)). The
fective absorptance is not sensitive to either the structure or the first two layers are common for FCC(111) and HCP(001). Two lay-
orientation (see. Fig. 4(b)). Probably, it is because r1 and r2 are ers reflect noticeably more than one layer (see Fig. 5(b)), especially
nearly identical for all the structures, which have the most weight for highly reflective materials with low a. It is in line with the
in Eq. (1). reflection factors rn (see Fig. 5(a)): the difference between them
Fig. 5 compares the effective reflectance R for beds consisting increases with n while the terms with great n are important for
of close-packed monolayers. The stacking is ABAB ... for the HCP highly reflective materials only, according to Eq. (1). The difference
4
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 3. Beds of equal circles packed into regular 2D structures: (a) Hexagonal lattice
with a close-packed row of circles on the top; (b) Hexagonal honeycomb lattice; (c)
Square lattice with a close-packed (10) row of circles on the top; (d) Square lattice
with a (11) top plane; Two views are shown for each powder bed model: The lattice
with an elementary cell (dached lines) and two nearest vertical mirror lines (right);
Calculation domain for ray tracing (left).
5
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
6
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 6. Effective absorptance of a deep bed of equal spheres depending on the structure and the top surface indicated by the Miller indices.
7
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 8. Average effective absorptance of a deep bed of equal spheres compared with the absorptance of Sqr(10) structure in 2D, other models, and experiments. The insert
estimates the relative error of the modified model of equivalent medium.
points are near the full red curve calculated by ray tracing and the are easily oxidized too. The oxidation can explain the discrepancy
dashed red curve calculated by the modified equivalent-medium between the experiments and the models for these three powders.
model, Eq. (11), for the SC structure. However, the powders of Al, Gusarov et al. [9] measured the effective reflectance of an oxidized
Ti, and W are significantly more absorptive and correspond bet- copper powder and the same powder after annealing in hydrogen
ter to the equivalent medium model, Eq. (5). It would not vali- atmosphere. They found that the annealed powder is considerably
date the equivalent medium model because it contradicts the re- less absorbing. The red point for copper in Fig. 8 corresponds to
cent ray-tracing results. Boley et al. [11] supposed that aluminum the annealed powder. It perfectly agrees with the ray tracing (see
is oxidized and, therefore, absorbs more. Titanium and tungsten the red full-line curve).
8
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 10. Powder bed absorptance versus the packing density: Comparison of exper-
iments of Gusarov et al. (closed circles) with the ray tracing (open circles) and the
modified equivalent-medium model (full lines).
Fig. 9. (a) Back reflection by a sphere on the top of a powder bed. (b) First reflec-
tion factor r1 calculated by ray tracing (points) and theoretically estimated (line).
9
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
diffuse surface,
(θ )
R (θ ) = . (16)
2 cos(θ )
The powders of iron and copper consisted of spherical particles
with the diameter approximately from 125 to 160 μm. Their pack-
ing density d was 0.55 and 0.58, respectively [9]. It is around the
packing density of the SC structure, see Table 1. That is why this
structure is chosen for ray tracing. The values of the solid-phase
absorptance a indicated in Fig. 14 are taken as reported in Ref.
[9]. The experiment and the modeling excellently agree at reflec-
tion angles θ < 70◦ for Cu and θ < 80◦ for Fe. The modeling
confirms the experimentally observed tendency of increasing the
normal-hemispherical reflectance with θ at θ > 60◦ . A discrepancy
at oblique angles can be explained by the experimental errors due
Fig. 12. Total angular distributions of reflected radiation (θ ) for the models of a to the influence of the borders of the cuvette with the powder and
deep powder bed with the SC structure at selected values of the solid-phase ab- a non-ideal alignment of the sensor relative to the cuvette [9].
sorptance a.
Fig. 13. Angular distribution of radiation reflected by a deep bed of equal spheres packed in the FCC, BCC, SC, and DI structures at selected values of the solid-phase
absorptance a. Packing density d is indicated on the top of each diagram.
Fig. 15. Partial depth profiles fn (z) for 20 layers of SC(100) structure.
Fig. 14. Normal-directional reflectance R(θ ): Comparison between the experiments
of Gusarov et al. at 0.79 and 0.633 μm wavelengths (points) and the present ray
tracing for the SC structure at the indicated values of the solid-phase absorptance That is why averaging over a great number of particles does not
a (lines).
eliminate the angular points. Such details cannot be generalized to
random packing.
Full lines in Fig. 17 present the profiles of radiative energy flux
where indices designate orientation. averaged over orientation for the FCC, BCC, SC, and DI structures
The curves plotted in Figs. 12, 13, and 16 exhibit angular points, at various values of the flat-surface reflectance of the solid phase
i. e. points where the derivative is not continuous. Probably, the an- a. It has to be noted that the value of the normalized energy flux
gular points are due to shadowing of a particle by another one. It q at z ≤ −D/2 is equal to the absorptance. The flux decays with z
occurs at equivalent positions and directions in regular structures. but generally does not tend to zero because a fraction of the in-
11
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Fig. 16. Normalized net flux of radiative energy q(z) for the models of a deep pow-
der bed with the SC structure at selected values of the solid-phase absorptance a.
12
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
13
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
CRediT authorship contribution statement shows the minimum angle α = α1 when the ray is directed on
another circle after the first reflection. We draw the ray reflected
A.V. Gusarov: Conceptualization, Methodology, Software, Formal by a symmetry line in point F to make the figure more compact.
analysis, Writing - original draft. Fig. A1(b) shows the minimum angle α = α2 when the ray turns
back after three reflections by the spheres. Thus, the incident ray
Acknowledgement turns back after one reflection if 0 ≤ α < α 1 and after two reflec-
tions if α 1 ≤ α < α 2 . Therefore,
This work was supported by the Ministry of Science and Higher
r1 = sin α1 , r2 = sin α2 − sin α1 . (A.1)
Education of the Russian Federation under project 0707-2020-
0034. To find α 1 , consider similar triangles ACO and BCD in Fig. A1(a),
14
A.V. Gusarov Journal of Quantitative Spectroscopy & Radiative Transfer 257 (2020) 107366
Table A1 [4] Wang C-A, Ma L-X, Tan J-Y, Liu L-H. Study of radiative transfer in 1d densely
Parameters of rays to calculate r2 . packed bed layer containing absorbing’scattering spherical particles. Int J Heat
Mass Transf 2016;102:669–78.
sin α2 = 0.82296097 α2 = 55.382307◦ δ = 27.304904◦ [5] Johnson E, Tar I, Baker D. A Monte Carlo method to solve for radiative effective
thermal conductivity for particle beds of various solid fractions and emissivi-
ties. J Quant Spectrosc Radiat Transf 2020;250:107014.
[6] Kang HH, Kaya M, Hajimirza S. A data driven artificial neural network model
Table A2 for predicting radiative properties of metallic packed beds. J Quant Spectrosc
First two reflection factors for normal incidence on Radiat Transf 2019;226:66–72.
a close-packed row of circles. [7] Taine J, Enguehard F. Statistical modelling of radiative transfer within non-Bee-
√ rian effective phases of macroporous media. Int J Thermal Sci 2019;139:61–78.
17 + 1 [8] Tolochko NK, Laoui T, Khlopkov YV, Mozzharov SE, Titov VI, Ignatiev MB. Ab-
r1 = = 0.64038820 r2 = 0.18257277
8 sorptance of powder materials suitable for laser sintering. Rapid Prototyping J
20 0 0;6(3):155–60.
[9] Gusarov AV, Bentefour EH, Rombouts M, Froyen L, Glorieux C, Kruth J-P. Nor-
mal-directional and normal-hemispherical reflectances of micron- and submi-
Exclude |DG| and |EF| from Eqs. (A.6), (A.7), (A.19): cron-sized powder beds at 633 and 790 nm. J Appl Phys 2006;99:113528.
[10] Zhang D, Wang W, Guo Y, Hu S, Dong D, Poprawe R, et al. Numerical simu-
|BD| = R(2 − sin α − cos δ ) tan β − R(cos α − sin δ ). (A.20) lation in the absorption behavior of Ti6Al4V powder materials to laser energy
during SLM. J Mater Process Technol 2019;268:25–36. doi:10.1016/j.jmatprotec.
Exclude |AO| and |BD| from Eqs. (A.2), (A.4), (A.20): 2019.01.002.
[11] Boley CD, Mitchell SC, Rubenchik AM, Wu SSQ. Metal powder absorptivity:
|CD| = [(2 − sin α − cos δ ) sin β − (cos α − sin δ ) cos β ]|CO|. modeling and experiment. Appl Opt 2016;55(23):6496–500. doi:10.1364/AO.
55.006496.
(A.21) [12] Meier C, Weissbach R, Weinberg J, Wall WA, John Hart A. Critical influences
of particle size and adhesion on the powder layer uniformity in metal addi-
Exclude |CO| and |CD| from Eqs. (A.3), (A.5), (A.21): tive manufacturing. J Mater Process Technol 2019;266:484–501. doi:10.1016/j.
jmatprotec.2018.10.037.
(2 − sin α − cos δ ) sin β − (cos α − sin δ ) cos β + cos γ = 1. [13] Meier C, Penny R, Zou Y, Gibbs J, John Hart A. Thermophysical phenomena in
metal additive manufacturing by selective laser melting: fundamentals, model-
(A.22) ing, simulation and experimentation. Annu Rev Heat Transf 2018;20:241–316.
[14] Wang XC, Kruth J-P. A simulation model for direct selective laser sintering of
From triangle DEI, metal powders. In: BHV Topping, editor. Computational techniques for materi-
als, composites and composite structures. Edinburgh: Civil-Comp Press; 20 0 0.
|DE|
= tan ε . (A.23) p. 57–71.
|EI| [15] Wang XC, Laoui T, Bonse J, Kruth J-P, Lauwers B, Froyen L. Direct selective laser
sintering of hard metal powders: experimental study and simulation. Int J Adv
Exclude |DE|, |DG|, |EF|, |EI| and |FI| from Eqs. (A.6), (A.7), (A.16)– Manuf Technol 2002;19(5):351–7.
(A.18), and (A.23): [16] Boley CD, Khairallah SA, Rubenchik AM. Calculation of laser absorption by
metal powders in additive manufacturing. Appl Opt 2015;54(9):2477–82.
(2 − sin α − cos δ ) sin ε = (cos α − sin δ ) cos ε . (A.24) doi:10.1364/AO.54.002477.
[17] Gusarov AV, Kruth J-P. Modelling of radiation transfer in metallic powders
Substitution of at laser treatment. Int J Heat Mass Transf 2005;48(16):3423–34. doi:10.1016/
j.ijheatmasstransfer.2005.01.044.
π π [18] Gusarov AV, Smurov I. Radiation transfer in metallic powder beds used in laser
ε = 2α − , β = 2δ − ε , and γ = 3α − 2δ − processing. J Quant Spectrosc Radiat Transf 2010;111(17):2517–27. doi:10.1016/
2 2
j.jqsrt.2010.07.009.
(A.25) [19] Rombouts M, Froyen L, Gusarov AV, Bentefour EH, Glorieux C. Light extinc-
tion in metallic powder beds: correlation with powder structure. J Appl Phys
reduces Eqs. (A.22) and (A.24) to 2005;98:013533.
[20] Kovalev OB, Gusarov AV. Modeling of granular packed beds, their statistical
(2 cos 2δ − cos δ ) cos 2α + (2 sin 2δ − sin δ ) sin 2α = 1, (A.26) analyses and evaluation of effective thermal conductivity. Int J Thermal Sci
2017;114:327–41. doi:10.1016/j.ijthermalsci.2017.01.003.
[21] Kovalev OB, Gusarov AV, Belyaev VV. Morphology of random packing of micro-
(2 − sin α − cos δ ) cos 2α + (cos α − sin δ ) sin 2α = 0. (A.27) particles and its effect on the absorption of laser radiation during selective
melting of powders. Int J Eng Sci 2020;157:103378. doi:10.1016/j.ijengsci.2020.
Numerical solution of the system of Eqs. (A.26), (A.27) gives the re- 103378.
[22] Zarrouati M, Enguehard F, Taine J. Statistical characterization of near-
sults shown inTable A1. Table A2 summarizes the reflection factors
wall radiative properties of a statistically non-homogeneous and anisotropic
calculated by Eq. (A.1). porous medium. Int J Heat Mass Transf 2013;67:776–83. doi:10.1016/j.
ijheatmasstransfer.2013.08.021.
References [23] Khmyrov RS, Ableeva RR, Gusarov AV. Metallographic study of denudation in
laser powder-bed fusion. Proc CIRP 2020;94:194–9. doi:10.1016/j.procir.2020.
[1] Wu H, Gui N, Yang X, Tu J, Jiang S. A matrix model of particle-scale radiative 09.037.
heat transfer in structured and randomly packed pebble bed. Int J Thermal Sci [24] Howell JR, Meng MP, Siegel R. Thermal radiation heat transfer. Taylor & Fran-
2020;153:106334. doi:10.1016/j.ijthermalsci.2020.106334. cis; 2015.
[2] Ruiz G, Ripoll N, Fedorova N, Zbogar-Rasic A, Jovicic V, Delgado A, et al. Exper- [25] Gusarov AV. Statistical approach to radiative transfer in the heterogeneous me-
imental and numerical analysis of the heat transfer in a packed bed exposed dia of thin-wall morphology-II: applications. J Heat Transf 2019;141:012701.
to the high thermal radiation flux. Int J Heat Mass Transf 2019;136:383–92. doi:10.1115/1.4040958.
[3] Asakuma Y, Honda I, Yamamoto T. Numerical approach to predicting the effec-
tive thermal conductivity of a packed bed of binary particles. Powder Technol
2019;354:886–92.
15