Powder Technology 192 (2009) 178–183

Contents lists available at ScienceDirect

Powder Technology

j ourna l homepage: www.e lsev ie r.com/ locate /powtec

Effect of particle size on three dimensional printed mesh structures

Kathy Lu ⁎, Matthew Hiser, William WuVirginia Polytechnic Institute and State University, Materials Science and Engineering Department, 213 Holden Hall-M/C 0237, Blacksburg, VA 24061, USA

⁎ Corresponding author. Tel.: +1 540 231 3225; fax: +E-mail address: klu@vt.edu (K. Lu).

0032-5910/$ – see front matter © 2008 Elsevier B.V. Adoi:10.1016/j.powtec.2008.12.011

a b s t r a c t

a r t i c l e i n f o

Article history:

Three dimensional printing Received 3 June 2008Received in revised form 1 October 2008Accepted 13 December 2008Available online 24 December 2008

Keywords:Three dimensional printingParticle sizeMesh structureGreen strengthSurface smoothnessBinder spreading

is a unique technique that can print complex 3D structures that cannot beproduced by other means, especially for rapid prototyping purpose. In this study, 3D mesh structures arecreated by three dimensional printing with four different TiNiHf powder sizes. Mesh structure green strengthand surface smoothness are characterized in order to produce high quality 3D structures. Binder spreadingtime and spreading rate in the TiNiHf powders are measured in order to understand binder penetrationdifference in the mesh structures. The study shows that smaller TiNiHf particle size produces higher meshstructure green strength and surface smoothness, consistent with the observation that binder spreading isslower for smaller TiNiHf particles.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Three dimensional printing (3DP) is a unique technique that canprint complex 3D structures that cannot be produced by other means,especially for rapid prototyping purpose [1–4]. During 3DP, a 3Dstructure model, formatted into a [.stl] file (standard trianglelanguage), is converted by a slicing routine into a compilation of two-dimensional slices representing the 3D part. The slice file is furtherformulated into instructions that control the movement of the 3Dprinting components. The powder is spread by a counter-rotatingroller onto a build platform inside a build box. By means of ink-jetprinting technology, a printhead, containing an array of binder fluidjets, rasters across the layer of the powder and deposits binder dropletsin those locations defined by the current 2D slice of the 3D structuremodel. Subsequently, the build platform advances downward by onelayer thickness and a new layer of powder is spread, which is thenprinted by the printhead. This procedure is repeated layer after layeruntil the 3D part is completed. After the designed 3D part is printed,the particles are held together by the binder used. The printed part canbe removed from the surrounding unbound powders. However, theprinted structures are not strong enough to be used directly and needto be sintered to densify the matrix.

The 3DP technique has demonstrated the capability of fabricatingparts of a variety of materials, including ceramics, metals, and poly-mers with an array of unique geometries [3,5–9]. However, substantialwork is still needed to explore and improve the ability of formingintricate fine structures. This is because the quality of the printedstructures is affected by many factors such as binder saturation level,

1 540 231 8919.

ll rights reserved.

printing layer thickness, and particle size. The first two factors havebeen studied in our prior work [10]. In order to obtain printed 3Dstructures with high integrity, small feature sizes, and accuratedimensions, particle size effects need to be studied.

In order to advance the understanding of particle size effect on theproperties of printed 3D structures, a gradient 3D mesh structure hasbeen designed. The green strength and surface smoothness of theprinted mesh structures are compared for four different size TiNiHfpowders. Since the interaction between the particles and binder playan important role in the 3D mesh structure creation, binder spreadingtime and rate for different size TiNiHf particles are studied with a highspeed digital optical microscope. Mesh structure integrity and surfacesmoothness are closely related to the particle size and binderspreading.

2. Experimental procedure

Four TiNiHf powders with different particle sizes were speciallymade (Crucible Research, Pittsburgh, PA) in order to evaluate particlesize effect on 3D mesh structure properties. The choice of TiNiHfpowder was based on our on-going research needs and the availabilityof the powder types [11]. The approach and knowledge gained shouldbe applicable to many other powder systems. The particle size rangeswere: less than 20 μm(−635mesh), 20–45 μm(+635 to−325mesh),45–75 μm (+325 to −200 mesh), and 75–150 μm (+200 to −100mesh). The particle size distributions of the four TiNiHf powders fromlaser light scattering analysis (Horiba, LA-950, Irvine, CA) are shown inFig. 1.

A gradient 3D mesh structure was designed as shown in Fig. 2.Therewere three layers in themesh structure and each layer had threecircular wires with diameters of 5, 10, and 15 mm, respectively. Each

Fig. 1. Particle size distributions of four different TiNiHf powders used in the study.

179K. Lu et al. / Powder Technology 192 (2009) 178–183

layer had different numbers of radial wires: top layer, 16 wires; middlelayer, 8 wires; and bottom layer, 4 wires. Wires on each layer hadsquare cross-section shape with 400 μm edge length. Vertical wiresbetween the adjacent layers had round cross-section shape with400 μm diameter. This round shape was used in order to achieve goodjunction nodes in the mesh structure.

A 3D printer (RX-1, ProMetal, Irwin, PA) was employed to print thedesigned 3D mesh structure. An acrylic-based proprietary binder wasused during the printing process. Based on the previous study withless than 20 µm TiNiHf powders [10], binder saturation level of 170%was used for each powder (the 3D printer allowed the bindersaturation level to vary from 40–200%). Printing layer thickness wasset to be twice of the largest particle size for a given powder. For thefour powders used in this study (less than 20 μm, 20–45 μm, 45–75 μm, and 75–150 μm), the printing layer thicknesses were 40, 90,150, and 300 μm, respectively. Each printed layer was cured by a built-in heat lamp at 90% power for 60 s before the next layer was printed.After the 3Dmesh structure was printed, it was further cured at 170 °Cfor an hour in an oven to strengthen the green structure. The loosepowder surrounding the printedmesh structurewas removedwith anair blower.

A Texture Analyzer test console equipped with a 5 kN load cell(Stable Micro Systems, Surrey, UK) was used for green strengthevaluation [12]. The console was set to record compressive load andthe crosshead was lowered monotonically at a speed of 0.1 mm/min.The shape of the load vs. displacement curve was recorded and thepeak load was determined from the curve as the 3D mesh structurebreaking force. The green strength was taken as the breaking strengthand was calculated as the ratio of the breaking force to the averagecross section area in the mesh structure horizontal direction.

A syringe (1 cm3 25G 5/8 Tuberculin, Becton Dickinson & Company,Franklin Lakes, NJ) was used to simulate the binder spray from theprinthead. The binder drop size used was 0.1 mL. TiNiHf powder of agiven size was put into a shallow container and then rolled even usingthe 3D printer. This process produced the TiNiHf powder packing thatresembled what existed during the 3D printing. Because the binderspreading process was too fast to be quantified by routine timing and

Fig. 2. 3D mesh structure de

spreading distance measurement, a high speed digital opticalmicroscope (KH-7700, Hirox Company, River Edge, NJ) was used tofilm the binder spreading process. A software (SolveigMM AVITrimmer, Solveig Multimedia Company, Tomsk, Russia) was used totrim the video files in order to measure the binder spreading time onthe scale of microseconds and binder spreading distance on the scaleof micrometers. The binder spreading experiment was repeated threetimes for each TiNiHf powder size.

3. Results and discussion

3.1. Particle packing and green strength

In order to study particle size effect on the 3D mesh structureproperties, the TiNiHf particles should be saturated with the acrylicbinder to the same extent during the 3DP process. One of the criticalvariables is binder drop volume, which can be varied based on threeparameters: printing layer thickness, binder saturation level, andpowder packing rate. The 3D printer surveys these parameters anddetermines the corresponding binder drop volume before anyprinting process is started. Because of this pre-existing correlation,printing layer thickness and binder saturation level need to be pre-determined; and powder packing rate needs to be measured. Asstated, the binder saturation level is 170% and the printing layerthickness is twice of the largest particle size for a given particle sizedistribution. These selections are based on our prior study as well asthe need to spread at least one layer powder each time [10]. In thisstudy, the printing layer thickness used is 40, 90, 150, and 300 μm,respectively for the four TiNiHf powders. The most challenging yetnecessary parameter to determine is powder packing rate. Thisparameter determination is more involved because powder packing inthe printbed is different from that of bulk loose powder packing; theroller spreading process by the 3D printer creates a different powderpacking rate for each powder. In this study, powder packing rate hasbeen measured by printing a rectangular TiNiHf powder specimen of5×5×2 mm3 for each TiNiHf particle size. After the 3D printing, theprinted sample with volume Vprint has been completely dried and thendissolved in a graduated beaker containing distilled water. The solidvolume of the loose powder, Vpowder, is thenmeasured in the graduatedbeaker. The powder packing rate ρ has been calculated using:

ρ=Vpowder

Vprint×100k: ð1Þ

Based on Eq. (1), the powder packing rates of the four TiNiHfpowders (less than 20 μm, 20–45 μm, 45–75 μm, and 75–150 μm) are35, 42, 54, and 62%, respectively. Clearly, the 3DP roller spreadingprocess has a larger impact on the packing of smaller size TiNiHfpowders. The powder packing rate has been substantially reducedduring the 3DP process for powders smaller than 75 µm. When the

sign used in this study.

Fig. 3. Green strength of 3D mesh structures from four different size TiNiHf particles.

180 K. Lu et al. / Powder Technology 192 (2009) 178–183

particle size is greater than 75 µm, the powder packing rate by the 3Droller spreading is almost the same as loose random powder packingrate, 64% [13,14]. This is an observation that has not been reported, yetunique to the 3DP process. The low powder packing rate is a directresult of powder spreading. Smaller particles have poorer flowabilityand cannot move and pack as well as large particles. Such drasticpowder packing rate change with particle size during the 3DP processwill have significant impact on binder spreading and printed structurestrength and should be carefully evaluated. Based on the powderpacking rates measured for the four TiNiHf powders, and the bindersaturation level and the printing layer thickness chosen, the binderdrop volume is determined to be 180 pL by the 3D printer for all themesh structures produced in this study.

The printed 3D mesh structures are loosely bound by the binderbut strong enough to be handled. The green strength of the printed 3Dmesh structures from different size TiNiHf powders is shown in Fig. 3.Overall, the green strength decreases drastically with increasingTiNiHf particle size. The mesh structure from the less than 20 µmpowder has the highest green strength, around 315 kPa. The meshstructure from the 20–45 μm powder has 102 kPa green strength. Themesh structure from the 45–75 µm powder has even lower greenstrength, around 35 kPa. The mesh structure from the 75–150 µmpowder is too fragile for measurable green strength to be obtained.This result is opposite to the powder packing rate trend measured,which shows that larger size TiNiHf powder has higher powderpacking rate. Put it differently, higher powder packing rate results inlower green strength, which is different from other studies [15–17].This is another unique characteristic of the 3DP process not observedfor randomly packed powder materials. The result can be understoodfrom several aspects. First, the 3D mesh structure strength is affectedby the number of the contact points between TiNiHf particles. Eventhough smaller TiNiHf powder has lower packing rate, there are morecontact points for a given cross section area. Second, the smaller radiiof the smaller TiNiHf particles in contact have a higher tendency toattract binder flow and offer higher bonding strength at the contactpoints. The end result is that the contact points in a given area and thebonding strength at the contact points play more significant role andoffset the adverse impact from the lower powder packing rate. Third,binder spreading process is different for different size powders as tobe discussed in Section 3.3. This can also contribute to the 3D meshstructure strength difference.

3.2. Mesh structure surface evaluation

The 3D mesh structures printed with the four different size TiNiHfpowders are shown in Fig. 4. Similar to what was previously reported[10], the printed wire meshes have larger dimensions than designed,650–770 µm vs. 400 µm. This discrepancy is mainly caused by binderspreading. During the 3DP process, the binder is jetted and the wiremeshes are printed based on the 400 µm width design. However, thebinder is not confined along the designed width line and lateralspreading is instantaneous. This inevitably increases the actual wirewidth. The specific wire width increase is a function of powder size,

powder packing rate, and binder drop volume. This is another newaspect that should be considered for the 3DP technique. From 3D partdesign point of view, this dimensional increase due to binderspreading can lead to dimensional tolerance loss if not addressed,especially if fine dimensions are required.

Fig. 4 also shows that except for the mesh structure printed with75–150 µm powder, the other three mesh structures have good shaperetention of the designed 3D structure. This means that even thoughthe largest particle size (75–150 µm) powder has the best flowabilityand the highest powder packing rate, the shape retention ability isundesirable. Good powder flowability is not necessarily beneficial formesh structure creation, another aspect unique to the 3DP process. Ifthe powder flowability is too high, the roller spreading rate, whichdictates how fast each layer of powder spreads, needs to be adjustedfor each powder size. Otherwise, a high roller spreading rate willcreate a high shear force that can cause shifting of the printed layersfrom large particles, as seen in Fig. 4(d).

In addition to the 3D mesh structure shape retention ability,printed wire mesh surface smoothness can be examined forqualitative comparison of the TiNiHf particle size effect on the 3DPprocess. The surface difference can be seen from the right side imagesof Fig. 4(a) to (d). Small TiNiHf particles produce more uniform surfacethan large TiNiHf particles. To provide more direct comparison, themesh surface has been measured by 3D profiling as shown in Fig. 5. Inthe images, the valley indicates the joining point/boundary betweenthe particles. The peak indicates the top of individual particles. Since allthe experiments have been conducted under the same condition, thecolors in Fig. 5 can be used to compare mesh structure surfacesmoothness differences. The mesh structures printed with the fourdifferent size TiNiHf powders have different surface smoothness. FromFig. 5(a) to (d), the peak to peak and the peak to valley differences (bothin spacing and height) increase as particle size increases. The meshsurface from smaller TiNiHf particles shows less surface profilevariation than the mesh surface from larger TiNiHf particles. This isconsistent with the images shown in Fig. 4, meaning smaller particlesproduce smoother surface.

3.3. Binder spreading

For the four powders studied, binder spreading diameter sizes areas follows based on the average of three measurements at eachcondition: 3426.6 µm for the b20 µm size powder, 4006.9 µm for the20–45 µm size powder, 3974.0 µm for the 45–75 µm size powder, and3781.9 µm for the 75–150 µm size powder. Since the binder spreadingdiameter is particle size, particle packing, and binder spreading timedependent, the binder spreading time and rate are used for moredirect comparison.

Horizontal direction binder spreading time and rate in the TiNiHfpowder bed are shown in Fig. 6. For the same amount of binder used(0.1 mL), the binder spreading time drastically decreases when theTiNiHf particle size increases from less than 20 µm to greater than45 µm. As the particle size further increases to 75 µm, the binderspreading time becomes almost constant. At less than 20 µm particlesize, the binder spreading time is 2903 ms. As the particle sizeincreases to 20–45 μm, the binder spreading time decreases to1426 ms. At 45–75 μm particle size, the binder spreading time is786ms. As the particle size further increases to 75–150 µm, the binderspreading time stays almost the same. This indicates that smallerTiNiHf powder size causes slower binder flow in the powder bed.However, there is a lower limit for the binder spreading time, whichbecomes constant when the TiNiHf particle size is larger than 75 µm.

The binder spreading rate in the horizontal direction shows thesame process but in a slightly different format from the binderspreading time. At less than 20 µm particle size, the binder spreadingrate is the slowest, 0.07 µm/ms. As the particle size increases to 20–45 μm, the binder spreading rate increases to 0.71 µm/ms. At 45–

Fig. 4.Mesh structure images for four different size TiNiHf powders: (a) less than 20 µm, (b) 20–45 μm, (c) 45–75 μm, and (d) 75–150 μm. The left side image is 3D structure. The rightside image is 2D mesh wire structure.

181K. Lu et al. / Powder Technology 192 (2009) 178–183

75 μm particle size, the binder spreading rate is 1.03 µm/ms. As theparticle size further increases to 75–150 µm, the binder spreading ratestays almost the same, at 0.98 µm/ms. Fig. 6 also shows that binder

spreading rate deviation becomesmuch larger for large TiNiHf particlesizes. This large spreading rate variation could have also contributedto the poorer surface smoothness observed in Fig. 4(d).

182 K. Lu et al. / Powder Technology 192 (2009) 178–183

Several factors lead to the binder spreading time and ratedifferences. On the one hand, smaller TiNiHf particles produce highercapillary force for the binder to flow and spread. Also, smaller TiNiHf

Fig. 5. Wire mesh surface 3D profiles for different TiNiHf powders: (a) less than 20 µm,(b) 20–45 μm, (c) 45–75 μm, and (d) 75–150 μm.

Fig. 6. Binder spreading time and rate for different size TiNiHf powders.

particles have lower powder packing rate, such as 35%. There is higherporosity in the powder bed for the binder to spread. Higher capillaryforce and more available porous space are beneficial for the binder topenetrate into the powder bed in a shorter time. On the other hand,smaller TiNiHf particles produce smaller and more tortuous porousstructure in the powder bed. Also, there are more TiNiHf particlesurface area and frictional force between the binder and the particles.These two factors hinder binder spreading. Overall, small pores andincreased total surface area in a TiNiHf powder bed outweigh thebenefits from higher capillary force and porosity, create higher binderflow hindrance, and thus result in longer binder spreading time andlower binder spreading rate.

Binder spreading rate plays a dominant role for the binder distrib-ution in the printed powder. The distance of binder lateral spreading isalways larger than that of binder vertical spreading. This is because thebinder lateral spreading is in the 360° direction while the verticalspreading is in one direction only. When the binder lateral spreadingrate is high, the spreading time is shorter; and more lateral spreadingoccurs. At the same time, binder spreading in the vertical direction iscompromised because there is less binder available for vertical spread-ing. This is especially the casewhen the printing layer thickness is large.For the large TiNiHf particles, there are faster binder lateral spreadingand larger printing layer thickness. Since the binder is the dominantfactor to impart green strength, some TiNiHf particles may notbe bonded in certain layers for the larger TiNiHf powder bed. This sub-sequently results in low green strength for the corresponding 3D wiremesh structures.

While the above analysis can successfully explain the experimentalobservations, binder spreading is a complicated process [18,19]. Theresults obtained from individual droplet spreading should be carefullyconsidered before being applied to the 3DP process that involves anarray of inkjets with pico-liter droplet sizes. Binder spreading from thebinder jet array instead of individual binder drop needs to be analyzedwhen the 3D printed part dimensional tolerance is critical. This aspectneeds to be further studied in future work.

4. Conclusions

In this study, 3D mesh structures are created by three dimensionalprinting with four different size TiNiHf powders: (a) less than 20 µm,(b) 20–45 μm, (c) 45–75 μm, and (d) 75–150 μm. The study shows thatsmaller TiNiHf particle size offers higher mesh structure greenstrength and surface smoothness. For the less than 20 µm powder,the mesh structure green strength is highest, at 315 kPa. Also, smallersize TiNiHf powder has slower binder spreading rate. For the less than20 µm powder, the binder spreading rate is 0.07 µm/ms. Binderspreading time and rate differences for different size TiNiHf particlesare explained based on particle packing and the resultant porousstructure difference. Themesh structure strength and surface smooth-ness are closely related to the particle size and binder spreading.

183K. Lu et al. / Powder Technology 192 (2009) 178–183

Acknowledgments

This material is based upon work supported by the Departmentof Energy under Award Numbers DE-FC07-06ID14739 and DE-FG26-06NT42741.

References

[1] E. Sachs, M. Cima, P. Williams, D. Brancazio, J. Cornie, J. Eng. Ind. 114 (1992) 481.[2] X.W. Yin, N. Travitzky, P. Greil, Int. J. Appl. Ceram. Technol. 4 (2007) 184.[3] H. Seitz, W. Rieder, S. Irsen, B. Leukers, C. Tille, J. Biomed. Mater. Res. 74B (2005)

782.[4] T.R. Jackson, H. Liu, N.M. Patrikalakis, E.M. Sachs, M.J. Cima, Mater. Des. 20 (1999)

63.[5] S.A. Uhland, R.K. Holman, M.J. Cima, E. Sachs, Y. Enokido, Mater. Res. Soc. Symp.

Proc. 542 (1999) 153.

[6] J. Moon, J.E. Grau, M.J. Cima, E.M. Sachs, J. Am. Ceram. Soc. 83 (2000) 2401.[7] B.M. Wu, M.J. Cima, Polym. Eng. Sci. 39 (1999) 249.[8] D. Dimitrov,W. vanWijck, K. Schreve, N. de Beer, Rapid Prototyping J. 12 (2006) 42.[9] J. Sicre-Artalejo, F. Petzoldt, M. Campos, J.M. Torralba, Int. J. Powder Metall. 44

(2008) 35.[10] K. Lu, W.T. Reynolds, Powder Technol. 187 (2008) 11.[11] C. Story, K. Lu, W.T. Reynolds, D. Brown, Int. J. Hydrogen Energy 33 (2008) 3970.[12] K. Lu, J. Mater. Sci. 43 (2008) 652.[13] R.M. German, Particle Packing Characteristics, Metal Powder Industries Federation,

Princeton, NJ, 1989, pp. 89–119.[14] M.N. Rahaman, Ceramic Processing and Sintering, Taylor and Francis Group, New

York, NY, 2003, pp. 328–420.[15] I.H. Moon, J.S. Choi, Powder Metall. 28 (1985) 21.[16] B. Sustarsic, L. Kosec, T. Span, M. Jelenko, M. Torkar, Mater. Technol. 35 (2001) 351.[17] Z. Zhang, B. Liu, F.F. Lange, J. Am. Ceram. Soc. 88 (2005) 1411.[18] X.L. Zhao, J.R.G. Evans, M.J. Edirisinghe, J.H. Song, J. Am. Ceram. Soc. 85 (2002) 2113.[19] R.K. Holman, S.A. Uhland, M.J. Cima, E. Sachs, J. Colloid Interface Sci. 247 (2002)

266.