Materials and Design 146 (2018) 249–259

Contents lists available at ScienceDirect

Materials and Design

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

Performance of biocompatible PEEK processed by fused depositionadditive manufacturing

M.F. Arif a, S. Kumar a,⁎, K.M. Varadarajan b,c, W.J. Cantwell d

a Department of Mechanical and Materials Engineering, Khalifa University of Science and Technology, Masdar Institute, Masdar City, P.O. Box 54224, Abu Dhabi, United Arab Emiratesb Department of Orthopaedic Surgery, Harris Orthopaedics Laboratory, Massachusetts General Hospital, 55 Fruit St, Boston, USAc Department of Orthopaedic Surgery, Harvard Medical School, A-111, 25 Shattuck Street, Boston, USAd Department of Aerospace Engineering, Khalifa University of Science and Technology, P.O. Box 127788, Abu Dhabi, United Arab Emirates

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

• Horizontally 3D printed PEEK with ras-ter angles of 0o and 90o exhibits tensileflexural and fracture properties compa-rable to those of molded PEEK

• Specimens built vertically are prone todelamination, exhibiting poorer me-chanical performance due to high ther-mal gradient in the build-direction

• Stick-slip fracture and lower Poisson’sratio are observed for specimens builtvertically, due to the presence of interfa-cial voids.

• Minimizing thermal gradients acrossbeads is the key to producing partswith excellent macroscopic properties

⁎ Corresponding author.E-mail address: s.kumar@eng.oxon.org (S. Kumar).

https://doi.org/10.1016/j.matdes.2018.03.0150264-1275/© 2018 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 13 December 2017Received in revised form 4 March 2018Accepted 5 March 2018Available online 7 March 2018

In this study, the tensile, flexural and fracture behavior of PEEK processed by fused filament fabrication (FFF) isreported. Three different configurations, viz., specimens built horizontally with a raster angle of 0° (H-0°) and90° (H-90°), and vertically with a raster angle of 90° (V-90°) are examined. The best performing specimen interms of its tensile, flexural and fracture toughness properties is H-0°, followed by H-90° and V-90°. The H-0°and H-90° specimens exhibit 85% and 75% of tensile and flexural strengths of molded bulk PEEK, respectively.The fracture toughnesses of the H-0° and H-90° specimens are 78% and 70% of molded bulk PEEK, respectively.However, V-90° specimens show lower tensile, flexural and fracture toughness properties compared to thoseof H-0°, H-90° and molded PEEK. The fracture surface and microtomography analyses indicate that the degreeof interfacial bonding between beads during layer-by-layer buildup, is affected by the thermal gradient acrossthe beads. The PEEK specimen configurations examined here have different thermal gradient in the build direc-tions and such variations manifest themselves in their macroscopic mechanical behavior. The findings of thisstudy provide guidelines for FFF of PEEK to enable its realization in applications such as orthopedic implants.

© 2018 Elsevier Ltd. All rights reserved.

Keywords:Additive manufacturingFused filament fabrication (FFF)Biocompatible PEEKFracture toughnessInterfacial bondingFused deposition modeling (FDM)

250 M.F. Arif et al. / Materials and Design 146 (2018) 249–259

1. Introduction

Compared to established technologies, additive manufacturing(AM) offers the possibility of cost-effective automation of the fabrica-tion process, it enables a digital inventory, and provides greater flexibil-ity to locally design thematerial architecture in three-dimensions [1–4].Fusedfilament fabrication (FFF), also known as fused depositionmodel-ing (FDM) is the most commonly used AM technique for thermoplasticpolymers. In the FFF process, a thermoplastic filament is fed into aheated nozzle, melted or liquefied, and subsequently extruded and de-posited onto a build plate. A gantry moves the nozzle in the horizontalx − y plane as the molten material is deposited. The build plate thenmoves vertically (in the z axis) after completing the deposition in thex− y plane. The deposited layers solidify and bond/weld with adjacentlayers, forming the desired 3D geometry. FFF has been used in the AMofnumerous thermoplastics. The most widely used polymer for FFF is ac-rylonitrile butadiene styrene (ABS) [5,6] and polylactic acid (PLA) [7].Moreover, high impact polystyrene [8], Nylon [8], Ultem® [9], polycar-bonate [8], poly(vinyl alcohol), polypropylene [10] and the mixturesof any two thermoplastics [11] or reinforced thermoplastic composites[12,13] are suitable for FFF. Nanocomposite feedstocks, such asgraphene [14,15] or CNT [16] reinforced filaments, are currentlyattracting interest, due to their superior thermal, electrical andmechan-ical properties.

Processing parameters, material type and geometrical features werefound to influence the final properties of the products made by FFF. Foroptimal properties of components or devicesmanufactured by FFF, tem-perature control is critical. The distribution and variation of temperatureand residual stresses in structures or parts produced by FFFmust be op-timally controlled so as to ensure sufficient internal stress relaxationand a good bonding quality between the adjacent polymer beads. Forexample, warping and interlayer delamination are frequently observeddue to the residual stresses accumulated during the layer-by-layer buildup. These stresses originate from the thermal gradient generated duringthe solidification process that influence the layer's volume contraction,warping and bonding/diffusion quality between beads [17,18]. Thishas shown to pose important challenges particularly for processing ofsemicrystalline polymers as a high degree of crystallinity and highshrinkage coefficient can negatively affect the dimensional stability ofthe part [19]. By using a heated build plate and by controlling the tem-perature of the 3D printer chamber, thermal gradients and their effects,such as warping can be minimized [20,21].

In recent years, there has been a growing interest in the fabricationof high-performance polymers such as polyphenylsulfone, polyetherether ketone (PEEK), polyether ketone ketone (PEKK), etc. via AM dueto their outstanding mechanical properties and chemical stability.PEEK is an expensive thermoplastic having excellent mechanical andchemical resistance properties even at temperatures up to 240 °C.PEEK also exhibits excellent hydrolysis resistance and provides fire,smoke, and toxicity performance. PEEK is one of the few polymers con-sidered for metal replacement in high temperature applications. PEEKhas been used in automotive, aerospace, oil & gas and space applica-tions. This study focuses on FFF of PEEK for bioimplants in order to effec-tively utilize its properties such as biocompatibility, excellent fatigueand wear resistance for such applications [22]. PEEK is typically fabri-cated by an injection molding or extrusion process. AM of high perfor-mance PEEK in particular, holds tremendous promise for biomedicalapplications, such as for design of novel knee replacement implantsthat more closely mimic native physiology. Effort has been made to ad-ditivelymanufacture PEEK via the selective laser sintering (SLS)method[23]. Oxford PerformanceMaterials, Inc. has developedOsteoFab® tech-nology for 3D printing of facial and cranial implants for bone replace-ment using PEKK powders via SLS process [24]. SLS uses a high powerlaser to selectively sinter powders, which causes them to coalesce to-gether due to interfacial and free surface diffusion. SLS can be used tofabricate very complex geometries because it uses the same powder to

support the overhang portions, unlike FFF which requires supportingparts. However, the inherent cost of SLS is high, owing to relatively ex-pensive capital costs and high power laser source.

Recently, FFF has beenpromoted as a cost-effective alternative to SLSfor the fabrication of high-performance polymers, such as PEEK.Valentan et al. [25] developed FFF systems to fabricate two grades ofPEEK-Optima®. The strength of the FFF samples in their study was ap-proximately half of the tensile strength of molded PEEK (~100 MPa).Vaezi and Yang [26] developed the FFF system to fabricate PEEK-Optima® samples. A nozzle temperature of 400–430 °C, an ambienttemperature of 80 °C and a build plate temperature up to 130 °C wereidentified as optimal parameters. The FFF process with 100% infill den-sity results in a sample porosity of 14% and a tensile strength of75 MPa. Wu et al. [27] showed that the warping deformation of FFF-PEEK samples reduces with increasing chamber and nozzle tempera-ture. Wu et al. [28] evaluated the influence of layer thickness (200,300 and 400 μm) and raster angle (0°/90°, 30°/−60° and 45°/−45°)on the mechanical properties of FFF–PEEK samples. A maximum tensilestrength of 56.6 MPa was achieved for samples with a 300 μm layerthickness and a 0°/90° raster angle. Cicala et al. evaluated the tensileproperties of FFF-PEEK samples with raster angle of 0°/90°, 30°/60°and 45°/−45° and found an average tensile strength of 69.04 MPa anda tensile modulus of 3.53 GPa for all tested samples [29]. Yang et al.[30] studied the effect of ambient temperature, nozzle temperatureand post heat-treatments on crystallinity and tensile properties ofFFF–PEEK with a 0° raster angle. In general, a higher ambient tempera-ture, higher nozzle temperature and a higher post treatment tempera-ture and time resulted in higher crystallinity, strength and stiffness.The maximum tensile strength and modulus were found to be 84 MPaand 4 GPa, respectively.

The aforementioned studies mostly focused on the development ofnew FFF machines to accommodate the high processing temperatureof PEEK, and to evaluate the effect of process parameters on the finalproperties of the samples. However, for the latter, the analysis of themechanical performance of the fabricated parts is presented with littledetail, especially on how the spatial thermal gradients during the layerby layer build-up of the sample, can influence the properties. In the cur-rent study, the tensile and flexural properties of samples built horizon-tally (x− yplane)with raster or bead angles of 0° and 90°, and verticallywith a bead angle of 90° with respect to loading direction are evaluated.These three configurations are associated with different thermal histo-ries during FFF-PEEK part build up. Digital image correlation (DIC) anal-ysis is used to measure the axial and lateral strains in the gauge lengthzone of dogbone specimens. Moreover, the fracture behavior of FFF–PEEK samples is also investigated. The fracture study is important,since formation of micro-cracks, particularly in the diffusion/bondingzones between beads, could lead to premature failure of fabricatedparts during their service life. To date, the fracture behavior and inter-layer cohesion studies of FFF fabricated samples are limited to ABS[31,32] and PLA [33,34] through single edge notch bending (SENB) ordouble cantilever beam (DCB) testing configurations. In this work, thefracture behavior of FFF–PEEK specimens is evaluated for the first timeusing compact tension tests for the same fabrication configurations asthose used in tensile and flexural testing.

2. Experimental methods

The configurations used to fabricate the specimens by FFF arediscussed, followed by tensile, flexural and fracture toughness testing,including optical strain mapping with DIC.

2.1. FFF configurations

FFF specimens were fabricated using an Indmatec HPP 155 device(Apium Additive Technologies GmbH). This FFF device is able to fabri-cate samples with a maximum build volume of 135 × 145 × 148 mm3.

251M.F. Arif et al. / Materials and Design 146 (2018) 249–259

Filaments with 1.75 mm diameter made of Victrex® PEEK 450G wereused. The filament exhibited glass transition and melting temperaturesof 153 °C and 340 °C, respectively, based on the 2nd heating cycle datafrom a differential scanning calorimetry (DSC) investigation with aheating rate of 10 °C/min. These transition temperatures closely corre-spond to those found in the Victrex® PEEK 450G database [35]. The fil-ament was fed into a 0.4 mm diameter nozzle by a feeding pressuremechanism via a driver motor and a counter-rotating set of groovedgears. The Simplify3D software package was used to determine the slic-ing sequence and define the FFF process. The FFF process parametersused in this study were as follows:

• Nozzle movement speed: 800 mm/min; first layer: 300 mm/min• Nozzle temperature: 410 °C; first layer: 390 °C• Bed temperature: 100 °C.• Layer height: 0.1 mm; first layer: 0.18 mm• Extrusion width: 0.48 mm• Infill pattern: Rectilinear• Infill density: 100%.

The nozzle movement speed, nozzle temperature and layer heightfor the first layer were set differently in order to enhance adhesion be-tween the sample and the glass build plate. Adhesion was further en-hanced by applying a glue on the plate surface and by the addition of

p

q

p

q

Horizontally fabricated in

plane, 0o beads orientation w.r.t

loading direction

(H-0o)

Horizontally fabric

plane, 90o beads or

loading dir

(H-90

a)

c) d)

Fig. 1. AM of PEEK by FFF: FFF configurations of tensile and compact tension specimens. The sspecimens fabricated horizontally with either c) 0° beads orientation or d) 90° beads orientatiin c) and d) illustrate the magnified view of the corresponding regions p and q in a). SimilFlexural specimen configuration is similar to that of tensile specimen.

brim. Brim is an outline to create a fabricated large ring that surroundsthe part and is attached to it. This holds the edges of the FFF fabricatedpart and thus helps to improve bed adhesion and prevent warping.After fabrication, the chamber was cooled to room temperature by nat-ural convection and the parts were then removed from the build plate.The brim structure surrounding the specimen was detached by hand.Three FFF configurations for the tensile, flexural and compact tensionspecimens used in this work are depicted in Fig. 1.

One specimen for each of the configurations was produced per fab-rication cycle, except for V-90°, tensile and flexural specimens. Fourand three V-90° tensile and flexural specimens were fabricated respec-tively in a single fabrication sequence. Themolten polymer in the V-90°dogbone and flexural samples is deposited onto a small cross-sectionalarea, as seen in Fig. 1b and thus the nozzle travel time in the x− y planeis very short prior to moving in the z direction. This can induce a fabri-cation failure, since the molten polymer is deposited onto a layer thatis still in amolten state. To avoid this, we increased the number of spec-imens so that the fabrication cycle could be done successfully, while en-suring acceptable degree of adhesion between the beads normal to theloading direction (z-direction). A reduced number of specimens re-sulted in fabrication failure while increasing the number of specimensmight increase the thermal gradient between beads in the vertical di-rection, ΔTΔz , due to the longer travel time in x − y plane, leading to aweak bonding. The length of the clamping zone of V-90° samplewas re-duced to reduce the fabrication time, as seen in Fig. 2. H-90° samples

p

q

r

s

rs

ated in

ientation w.r.t

ectiono)

Vertically fabricated in direction,

90o beads orientation w.r.t loading

direction

(V-90o)

b)

e)

pecimens were fabricated either a) horizontally or b) vertically. The deposition pattern ofon, and e) specimens fabricated vertically with 90° beads orientation. Sub-figures p and qarly, r and s in e) illustrate the magnified zone of corresponding regions r and s in b).

H-0o

V-90o

H-90o

Fig. 2. FFF-PEEK tensile specimens.

252 M.F. Arif et al. / Materials and Design 146 (2018) 249–259

were fabricated without outside perimeter/outline in order to avoidbeads with opposing orientations. In H-0° sample, the polymer meltwas initially deposited in the x− y plane following the outline contourof the dogbone sample andwas continued until completely printing theentire gauge length zone, leaving the gap only at the central areas of theclamping zone. These clamping zones were then fabricated with beadsof 45°/−45° orientation to ensure sufficient strength in the transitionzone and thus avoiding sample failure in this zone.

2.2. Tensile testing

Tensile tests were performed according to ISO 527 at a constantcrosshead speed of 1 mm/min at ambient temperature (~20 °C) on aZwick-Roell Z005 universal testing machine (UTM) with a 2.5 kN loadcell. The load cell accuracy was b±0.25% for a load N 10 N and b±1%for a load range of 2.5 to 10 N, and crosshead travel resolution was0.041 μm. DIC technique was used for full-field measurement of strainon the specimens. Random black–white speckle patterns of acrylicpaint were sprayed on the specimen using an airbrush prior to testing.Consecutive speckle images were acquired as a function of load usinga monochrome 5.0 MP camera for strain evaluation. The average engi-neering strain in the axial and lateral directions over the gauge lengthzone was evaluated using Vic-2D software. These strain values werethen used to construct the engineering stress–engineering straincurve. The tensile properties of the samples were extracted from thestress-strain curve. Each of the specimen configurations was tested atleast three times to assess repeatability. The FFF–PEEK tensilespecimens are shown in Fig. 2.

a) b)

3‒15 mm

Fig. 3. a) Geometry of the compact tension specimen, b) FFF-PEEK compact tension specime

2.3. Flexural testing

Three point flexural tests were performed according to ISO 178 at aconstant crosshead speed of 2 mm/min at ambient temperature (~20°C) on a Zwick-Roell Z005 universal testing machine (UTM) with a2.5 kN load cell. The H-0° specimens slipped from the supports afterthe maximum load had been reached but didn't exhibit final failure.Therefore, testing for H-0° was stopped when the load had reduced by20% from the maximum. Each of the specimen configurations wastested three times to ensure repeatability.

2.4. Fracture tests

Fracture testswere conducted to evaluate theMode I fracture tough-ness of the 3D printed PEEK. The compact tension fracture properties ofFFF-PEEK specimens were evaluated following ASTM D5045-14 stan-dard for measuring the plane strain fracture toughness of polymers.Compact tension samples were fabricated in the configuration shownin Fig. 3a, where W = 13–15 mm. After fabricating rectangular blocks,holes for loading pins (∅ 3 mm) were drilled using a benchtop drillingmachine. A notch was introduced by a high precision circular saw of152.4 mmdiameter and 0.508 mm thickness. A sharp notch was subse-quently generated by a slow tapping motion with a fresh razor blade.These specimens were manufactured without either holes or a notch,to ensure a consistent thermal history. The FFF of the holes and the ini-tial cracks would clearly require a change in the deposition patterns ofthe filament beads, leading to a change in specimen's thermal history.Prior to testing, the samples were kept in oven for 12 h at 70 °C to re-move moisture. A crosshead speed of 1 mm/min in a Zwick-Roell

c)

n and c) magnified view of the sharp crack tip region of the specimen prior to testing.

A

B

C

DH-0o

H-90o

V-90o

a) b)A B C D

(%)

4.50

4.05

3.60

3.15

2.70

2.25

1.80

1.35

0.90

0.45

0.00

3.00

2.70

2.40

2.10

1.80

1.50

1.20

0.90

0.60

0.30

0.00

0.40

0.36

0.32

0.28

0.24

0.20

0.16

0.12

0.08

0.04

0.00

25.50

22.95

20.40

17.85

15.30

12.75

10.20

7.65

5.10

2.55

0.00

(%) (%) (%)

Fig. 4. Tensile properties of the FFF-PEEK tested in different directions corresponding to the different build directions: a) Representative stress-strain curve of H-0°, H-90° and V-90°, insertshows extended ×-axis of H-0° specimen, b) axial strain contour at point A, B, C and D in a).

253M.F. Arif et al. / Materials and Design 146 (2018) 249–259

Z005 UTM was used. Critical stress intensity factor KIC was determinedfollowing the method given in ASTM D5045. The adopted scheme as-sumes a linear elastic fracture behavior of compact tension specimenand therefore restriction on linearity of load-displacement responsewas imposed. The 5% secant line criterion of the load-displacementcurve was employed to ensure validity of the KIC, as described inASTM D5045 Section 9.1.1. Moreover, the specimen size was appropri-ately chosen to ensure a plane strain-state at the crack tip. The size cri-terion stipulates that the value of 2.5(KIC/σy)2 must be less than thespecimen thickness, crack and ligament lengths. In the current study,the yield strength (σy) of the molded PEEK was used as a referencedue to different thermal histories associated with 3D printed tensileand compact tension specimens. Moreover, introducing a sharp crackexactly terminating at the bead-to-bead interface is practicallyimpossible.

3. Results and discussion

3.1. Tensile and flexural properties

The tensile performance of specimens fabricated in three differentconfigurations was evaluated. According to ISO 527-1:2012, tensilestrength is the stress at which the first local maximum is observed dur-ing a tensile test. Therefore, the tensile strength of the H-0° sample wasevaluated at the yield point while those of the H-90° and V-90° speci-mens were evaluated at the failure point, as seen in Fig. 4. The H-0°specimen exhibited the highest Young's modulus and tensile strength,followed by the H-90° and V-90° (Fig. 4 and Table 1). In the H-0° spec-imen, the tensile loading direction is parallel to the orientation of the fil-aments or beads and thus this sample showed higher Young's modulus

Table 1Mean values and standard errors for the tensile and flexural properties of H-0°, H-90° and V-9PEEK 450G [35] are presented for comparison.

H-0° H-90°

Tensile propertiesYoung's modulus, Et (GPa) 3.80 ± 0.03 3.54 ±Tensile strength, σm (MPa) 82.58 ± 1.03 (at yield) 72.88 ±Tensile elongation, εb (%) 110.97 ± 5.31 2.91 ±Poisson's ratio, ν 0.44 ± 0.02 0.43 ±

Flexural propertiesFlexural modulus, Ef (GPa) 3.08 ± 0.22 3.06 ±Flexural strength, σf (MPa) 142.0 ± 5.56 124.3 ±

and tensile strength. The H-90° offered a Young's modulus and tensilestrength values that were 7% and 12% lower than those of H-0°, respec-tively. The H-90° specimens exhibited properties approaching those ofthe H-0° specimen, due to the excellent interfacial bonding betweenbeads. The H-90° had a short bead deposition path in the horizontalplane as the nozzle travelled across the width of the specimen, as seenin Fig. 5a. Consequently, the molten polymer diffused and bonded wellwith adjacent beads, since the deposited beads were still in a moltenstate. The H-0° samples had a longer deposition path in the horizontalplane as the nozzle travelled across the length of the specimen (see,Fig. 5b). It is likely that the bead was deposited when the neighboringbead had been solidifying, resulting in poorer interfacial bonding be-tween beads compared to that of the H-90°. Even though the nozzletravel speed was constant, the build-rates were different for H-0° andH-90° configurations due the different deposition paths. It can be seenfrom Fig. 5 that the global in-plane (x − y) build direction for H-90°sample was in the direction of specimen's length, while the build direc-tions for H-0° and V-90° samples were in the direction of specimen'swidth and thickness, respectively. Therefore, the build-rate in x direc-tion for H-90° samples, Vx ¼ Δx

Δt and the build-rate in y direction for H-

0° and V-90° samples, Vy ¼ ΔyΔt . The build-rate in z direction for H-0°,

H-90° and V-90° samples, Vz ¼ ΔzΔt . As seen in Table 2, the global in-

plane build-rate for H-90° sample (62.5 mm/min) was higher thanthat of H-0° sample (4.74 mm/min) and hence lower thermal gradientleading to superior interfacial bonding between beads in x − y plane.It should benoted thatVy for H-0° specimenwas evaluatedwithout con-sidering the build-speed in the middle portion of the clamping zonesthat had alternating 45°/−45° bead orientations.

The difference in thermal history between H-0° and H-90° speci-mens can be seen qualitatively from the specimen color (see, Fig. 2).

0° specimens. Tensile elongation is evaluated at break. The properties of molded Victrex®

V-90° PEEK 450G

0.05 3.03 ± 0.01 4.01.92 (at break) 9.99 ± 0.94 (at break) 98 (at yield)

0.14 0.33 ± 0.03 450.01 0.37 ± 0.005 –

0.21 2.54 ± 0.07 3.87.98 16.40 ± 2.18 165

Table 2Build-rate (Vi) in x, y and z directions, and the deposition time (txy) in a single horizontalplane during the FFF-PEEK process.

Vx

(mm/min)Vy

(mm/min)Vz

(mm/min)txy(min)

Tensile specimens H-0° – 4.74 0.07 1.52H-90° 62.5 – 0.07 1.52V-90°a – 12.64 0.36 0.28

Flexural specimens H-0° – 5.68 0.05 1.9H-90° 42.11 – 0.05 1.9V-90°b – 10.96 0.27 0.33

Compact tensionspecimens

H-0° – 27.63 0.16 0.58H-90° 24.81 – 0.16 0.58V-90° – 26.17 0.4 0.25

a Four specimens in one fabrication batch.b Three specimens in one fabrication batch.

deposition path

deposition patha)

b)

c)

Fig. 5. Schematic of the thermal gradient and build-rate during FFF-PEEK process in a) H-90°, b) H-0°, and c) V-90° samples. Vx ¼ ΔxΔt is the build-rate in x for H-90° samples. Vy ¼ Δy

Δt is the

build-rate in y for H-0° and V-90° samples. Vz ¼ ΔzΔt is the build-rate in z for H-0°, H-90° and V-90° samples.

254 M.F. Arif et al. / Materials and Design 146 (2018) 249–259

The gauge length zone of the H-0° specimen with a higher thermal gra-dient between beads in the x− y plane, and thus weak interfacial bond-ing, had a brownish color. While the H-90° specimen that had a lowerthermal gradient between beads in the x− y plane, and thus better in-terfacial bonding between beads, had a more pale or whiter color. Thiscolor was relatively similar to that of the clamping zone for the H-0°specimen which was purposely fabricated with alternating 45°/−45°bead orientations to have a shorter deposition path, and to avoid failurein the transition or clamping zone during tensile testing. Moreover, thesurface finish can be used as an indicator of the degree of diffusion be-tween beads. As seen from Fig. 2, the surface finish of the H-90° speci-men was relatively rough. Since the molten polymer was depositednext to a bead that was still in its molten state (due to the short deposi-tion path), the beads flowed more easily, forming an overlap region be-tween beads. In contrast, the H-0° specimen had a relatively smoothsurface, since the molten polymer was deposited next to a bead thathad started to solidify. However, even though the interfacial bondingin H-90° was better than that in H-0° specimen, the H-0° specimen ex-hibited superior properties, since themacroscopic performancewaspri-marily due to the strength and strain tolerance of the filament, asdiscussed earlier.

The interfacial bonding between beads is critical, especially betweenthe layers in the vertical direction, particularly for V-90° specimenswhere the loading direction is normal to the interface between beadsin the z direction (see, Fig. 5c). A lower thermal gradient betweenbeads in the z direction required for promoting interfacial adhesion invertically-fabricated specimens was restricted by the fact that the pre-ceding layer should be fully or partially solidified so as to lay thesucceeding layer properly. Therefore, four tensile specimens of V-90°configurations were fabricated simultaneously to increase the deposi-tion time in the horizontal plane (x − y), and thus reducing the build-rate in the vertical (z) direction and avoiding the fabrication failure. Re-sults show that the tensile strength andmodulus of V-90° samples weresignificantly lower compared to those of H-0° and H-90° samples, asseen in Fig. 4 and Table 1. These results indicate incomplete sinteringbetween layers in the vertical direction due to unavoidable large ther-mal gradient in the vertical fabrication direction (z). This can be seenfrom the SEM image of V-90° sample where brittle-like fracture surface

is observed, as seen in Fig. 6c. It can also be seen clearly from the SEMimages of the H-0° and H-90° specimens where the weak interfacecan be observed between the beads in vertical fabrication direction(Fig. 6a and b). It should be noted that, the time to build a single layerin the horizontal plane (x − y) of H-0° and H-90° samples was1.52 min, which was approximately five times higher than that of theV-90° samples (0.28min), due to larger x− y deposition area comparedto that of V-90° samples. Therefore, the H-0° and H-90° samples hadlower build-rate (0.07 mm/min) in vertical direction than that of V-90° samples (0.36 mm/min), as seen in Table 2, and thus larger thermalgradient in z direction. Therefore, we expect poorer interfacial bondingin the vertical direction for H-0° and H-90° samples. However, the effectof thermal gradient in the z direction has less influence on themechan-ical properties of the H-0° and H-90° samples since the weak interfacesbetween beads in the z direction do not experience normal tractionunder tensile loading. A heating zone at the top of the printer's chamberhas been installed to maintain the top surface temperature of the sam-ple and to regulate the chamber temperature, with temperature controlbeing a proprietary technology of the manufacturer. Though it ensuresthat the top surface of the sample remains heated, it cannot fully elimi-nate the low interfacial bonding in the z direction as theminimum ther-mal gradient between beads is restricted by the fact that the preceding

a) b) c)

Fig. 6. SEM fracture surfacemorphology of a) H-0°, b) H-90° and c) V-90° specimens. The H-0° specimenwas evaluated at 110% strain to failure and thus lower in cross sectional area andlarger macroscopic void. Red, yellow and white (arrows and circles) indicate the z-fabrication direction, the raster angle and the load direction, respectively. The circle indicates thedirection normal to the surface.

255M.F. Arif et al. / Materials and Design 146 (2018) 249–259

layer should be fully or partially solidified so as to lay the succeedinglayer.

To further confirm the interfacial defects andmorphology in the z di-rection, micro-computed tomography (μCT) images of the H-0°, H-90°and V-90° tensile samples under no-load have been acquired via Phoe-nix Nanotom®m system. μCT analysis is important to complement theSEM observations since it provides details of 3Dmicrostructure and en-ables visualization of cross-sectional 2D image from the 3D volume,though it has lower resolution than that of the SEM images. As seen inFig. 7a and b, the interfacial defects between the beads in the vertical di-rection in H-0° and H-90° samples cannot be seen as the void dimen-sions were mostly below the μCT voxel size or resolution (3 μm).However, the presence of interfacial defects between beads in the verti-cal fabrication direction was confirmed earlier from the SEM images of

a)

b)

voids

Fig. 7. μCT images of FFF-PEEK samples at no-load; a) 2D image of H-0° sample, the insert showstensile sample, the insert shows 3D reconstructed image of voids, d) 2D image of V-90° compac(arrows or circles) respectively indicate z-fabrication direction, load direction and raster angle

tensile fracture surface, as seen in Fig. 6a and b, and also from the SEMimages of fractured surface of compact tension H-0° and H-90° samples(see, Fig. 11a and b). In fact, some interfacial defects were found in theμCT H-0° and H-90° samples. However, the number of defects observedwas small and the defectswere randomly scattered in a few zones of thesample volume. It is interesting to note that our H-0° andH-90° samplespossess qualitatively a very low void fraction, although FFF processtends to produce large gaps or voids between deposited beads. A lowvoid fraction was achieved in this study by choosing a 100% infill den-sity, a small layer height (0.1 mm) and high nozzle temperature (410°C) that allowed an optimumviscousflowof themolten polymer.More-over, the vertical build-rate for the H-0° and H-90° samples was lower(0.07 mm/min) due to the large fabrication area in the x − y planeand thus the preceding layer had sufficient time to solidify. Therefore,

c)

1 mm

1 mm

d)

voids at cross junctions between beads, b) 2D image ofH-90° sample, c) 2D image of V-90°t tension sample, the insert shows 3D reconstructed image of voids. Red, white and yellow. The circle indicates direction normal to the surface.

Table 3Mean values and standard errors of the fracture toughness (KIC) of the FFF-PEEKsamples.

KIC (MPa·m1/2)

H-0° 5.32 ± 0.17H-90° 4.76 ± 0.39V-90° 0.97 ± 0.03Molded PEEK

Karger-Kocsis and Friedrich [37] 7Beguelin and Kausch [38] 6.75Gensler et al. [39] 6.8Rae et al. [40] 6.87

H-0o

H-90o

V-90o

Fig. 8. Flexural stress-strain curves for the H-0°, H-90° and V-90° FFF-PEEK specimens.

256 M.F. Arif et al. / Materials and Design 146 (2018) 249–259

the succeeding horizontal layer was deposited onto a surface that hadalready solidified, inducing a stable deposition of the polymer meltand thus preventing a void formation at the interface. Concomitantly,this led to weak interfacial bonding due to high thermal gradient be-tween beads along the vertical direction. Different interfacial defectsand morphologies were observed for V-90° tensile samples under no-load, and those results were further confirmed from the V-90° compacttension samples. Voids at the interfaces between beads were observedin both tensile and compact tension V-90° samples, as seen in Fig. 7cand d. As discussed earlier, the deposition of molten polymer in V-90°configuration occurred on the surface of the preceding horizontallayer that was not fully solidified in order to promote an acceptable de-gree of interfacial adhesion between the beads in the vertical fabricationdirection,without inducing a fabrication failure. Consequently, good ad-hesion between beads was achieved over a certain area of the interface,but voids were created primarily due to undulations of themolten layerbeing laid on a partially solidified preceding layer, and shrinkage of thepolymer. Inserts in Fig. 7c and d show the 3D reconstructed image of V-90° tensile and compact tension samples, respectively, which describemore clearly the voids formed at the interface between the beads inthe vertical fabrication direction. It should be noted that, these voidscannot be seen clearly from the SEM images of tensile and compact ten-sion samples in Figs. 6c and 11c. These images were taken on theinterfacially debonded surfaces between beads in the vertical fabrica-tion direction. Voids were mostly developed between the undulatingbeads normal to surface view and thus they cannot be seen clearly.Some voids were also observed in H-0° samples particularly at thecross-junction between beads, as seen in insert of Fig. 7a. It shows an av-erage bead dimension of 0.5 × 0.1 mm2, which corresponds to a layerheight of 0.1 mm and extrusion width of 0.48 mm.

a) b)

Fig. 9. Load-displacement curve of FFF PEEK with

Samples loaded below the yield point (for H-0°) and below failurepoint (for H-90° and V-90°) showed uniform axial strain along thegauge length zone, as can be seen in Fig. 4b, with the maximum stan-dard deviation in longitudinal strains of 0.15%, 0.12% and 0.019%, re-spectively. This indicates that the samples were macroscopicallyuniform, even though the samples were built layer-by-layer. The sam-ple loaded beyond the yield point for H-0° didn't exhibit a uniformaxial strain along the gauge length, with the maximum strain occurringin the middle portion due to its excessive straining. The Poisson's ratioof the H-0° and H-90° samples were approximately the same (see themean values and their corresponding standard error, in Table 1).Much lower Poisson's ratio was observed in V-90° sample and thiscould be attributed to two factors, namely, the bead orientation andthe proliferation of voids during loading. The tensile loading in V-90°specimen is normal to numerousweak bead-to-bead interfaces. This re-sults in significant void growth during loading and thus exhibits lowerPoisson's ratio.

The flexural properties of the FFF-PEEK were also evaluated. It wasfound that the flexural properties followed the same trend as those oftensile properties, as the H-0° specimen had the highest flexuralstrength and modulus, followed by the H-90° and V-90°, as seen inTable 1 and Fig. 8. This is due to nearly the same build-rate and deposi-tion time in x − y plane for tensile and flexural samples, as seen inTable 2. The flexural properties reported here confirms the observationof the effect of printing configurations on tensile properties discussedearlier. Both tensile and flexural strength of H-0°, H-90° and V-90° spec-imens were 15%, 25% and 90% lower than those of molded PEEK 450G,respectively (Table 1). However, the tensile and flexural moduli didn'texhibit similar levels of reduction compared to the PEEK 450G. The ten-sile and (flexural) moduli of the H-0°, H-90° and V-90° specimens were5% (19%), 11% (19%) and 24% (33%) lower than those of molded PEEK450G, respectively. This is due to the fact that the interface zones ofthe layered structure of the FFF-PEEK samples provide strain-toleranceto the system (due to their lower compliance) under bending loadand thus lower the flexural modulus of the samples. The Young's mod-ulus and tensile strength of the H-0° and H-90° specimens were similarto those reported elsewhere [30] or indeed higher [25,26,28,29] thanthe previously reported FFF-PEEK samples. The flexural modulus andstrength of the H-0° and H-90° specimens were higher than the

c)

a) H-0°, b) H-90° and c) V-90° specimens.

Fig. 10. Photograph of fractured compact tension samples of a)H-0°, b) H-90° and c) V-90°specimens.

257M.F. Arif et al. / Materials and Design 146 (2018) 249–259

previously-reported FFF-PEEK samples [26,28]. For theV-90° sample, al-though it had a relatively high Young's modulus, the tensile strengthwas significantly lower than those of H-0° and H-90° specimens andmolded PEEK. The tensile strength of V-90° falls within the range ofmolded LDPE [36].

3.2. Fracture toughness

The fracture test results showed the influence of the FFF fabricationconfigurations and bead orientation on the fracture toughness. The H-0° specimen showed the best performance, followed by H-90° and V-90°, as can be seen in the load-displacement curves (Fig. 9) and KIC

(Table 3). Based on build-rate data presented in Table 2, the interfacialbonding between beads in H-0° compact tension sample was betterthan that in H-0° tensile and flexural samples. On the other hand, H-90° compact tension sample exhibited lower interfacial bonding be-tween beads compared to those of tensile and flexural samples. Thebuild-rates in z direction (Vz) were almost the same for compact ten-sion, tensile and flexural samples and thus exhibited poor interfacialbonding between beads, as discussed earlier. Interestingly, the build-rate in x − y plane for H-0° and H-90° compact tension samples was

a) b)In-plane crack propagation

Out-of-plane crack propagation

In-plane crack propagation

Fig. 11. SEM images of fracture surfaces of a) H-0°, b) H-90° and c) V-90° specimens. Red, yellowload direction. Circle indicates direction normal to the surface.

almost the same and thus we expect similar degree of interfacial bond-ing between beads in x− y plane for both samples. For the ASTMD5045standard adopted here, it is most likely that the sharp tip of the intro-duced crack in H-0° sample is terminating well within the bead (butnot exactly at the interface between beads) and thus the fracture tough-ness value estimated doesn't fully represent the fracture behavior of lay-ered structure of FFF-PEEK. Such limitation was also encountered for H-90° and V-90° samples as it was difficult to generate a sharp crack ex-actly terminating at the interface between beads. Regardless of theselimitations, it can be seen that the trend of KIC was almost the same asthose observed for the tensile and flexural tests. Results indicate thatthe bead orientation with respect to the loading direction contributesthemost to themechanical performance, followed by the interfacial ad-hesion between beads. Interfacial adhesion is affected by the thermalgradient dictated by the in-plane and out-of-plane build-rates (seeTable 2).

The crack tip in H-90° and V-90° compact tension specimens exhib-ited a higher probability of advancement through the bead-bead inter-face where the interfacial zone between beads acted as the weakestlink, unlike the H-0° specimen wherein the crack tip advanced normalto the bead orientation (see, Fig. 10). The crack tip in H-0° specimen re-quired more energy to break the beads transversely, resulting in higherfracture toughness. This effect can also be seen on the fracture surface,where the crack path in H-0° specimen sometimes deflected into theweakest link, leading to out-of-plane crack propagation, while thecrack propagation in H-90° and V-90° specimens consistentlydelaminated along the bead-bead interface, as shown in Figs. 10 and11. The difference in performance between H-90° and V-90° specimenswas related to the different degree of interfacial bonding between beadsassociated with the different thermal gradients, as discussed in the pre-vious section. After the maximum load had reached, a softening re-sponse was observed as seen in Fig. 9b due to the deflection of crackinto neighboringweak bead-bead interface. The SEM images of the frac-ture surfaces of compact tension samples shown in Fig. 11a, b and chelped to draw a similar conclusion to those of tensile samples asweak interfacial bonding can be found between beads in the verticalfabrication direction. The stick-slip crack growth (after the maximumload) was observed in V-90° sample under constant applied extensionrate loading conditions due to the presence of voids at the interface be-tween beads. These voids act as a crack arrester. Stick-slip fracture inthis case was characterized by oscillatory crack tip velocities and crackgrowth jumps resulting in periodic load fluctuations (see, Fig. 9c) [41].The strain contour atmaximum load of the three sample configurationswas evaluated. It showed that the strain was a maximum at the crackfront, as seen in Fig. 12. The highest value of maximum strain wasfound in H-0°, followed by H-90° and V-90° specimens. The H-0° speci-men had the highest strain tolerance as the crack tip advanced normal

c)

andwhite (arrows or circles) respectively indicate z-fabrication direction, raster angle and

a) b) c)6.10

5.23

4.36

3.49

2.62

1.75

0.88

0.01

-0.86

-1.73

-2.6

(%) 4.30

3.71

3.11

2.51

1.92

1.33

0.73

0.14

-0.46

-1.05

-1,65

0.47

0.41

0.34

0.28

0.21

0.15

0.09

0.02

-0.04

-0.11

-0.17

(%) (%)

Fig. 12. DIC contour of strain at maximum load of a) H-0°, b) H-90° and c) V-90° specimens.

258 M.F. Arif et al. / Materials and Design 146 (2018) 249–259

to the bead orientation, and thus the highest fracture toughness of thesample. While comparing the current results with those from tests onbulk, molded PEEK using compact tension geometry at room tempera-ture, we noticed that the fracture toughnesses of H-0° and H-90° sam-ples, respectively were 78% and 70% of the molded PEEK fracturetoughness (see, Table 3). The low fracture toughness in the V-90° con-figuration reduced the full potential of PEEK to be applied in highload-bearing applications but the value was still within the range ofbrittle polymers such as acrylic, epoxy and polystyrene [42].

4. Summary

The tensile and flexural properties of PEEK fabricated by FFF havebeen studied. The fracture behavior of FFF-PEEK has been reported forthe first time. Three specimen configurations; H-0°, H-90° and V-90°have been tested to failure in conjunction with optical strain mapping.The configuration that exhibited a superior performance, in terms oftensile strength and modulus, flexural strength and modulus, and frac-ture toughness, was H-0°, followed by H-90° and V-90°. The H-0° andH-90° showed 85% and 75% attainable tensile and flexural strengths ofthe bulk molded PEEK respectively. The fracture toughnesses of the H-0° and H-90° specimenswere 78% and 70% of themolded PEEK fracturetoughness, respectively. However, the V-90° sample exhibited low ten-sile, flexural and fracture toughness properties compared to those of H-0°, H-90° and bulk PEEK. The performance of these three configurationsrepresents different thermal gradients between beads during layer bylayer buildup in FFF.

It was found that different fabrication path can promote a differentdegree of interfacial adhesion between beads which in turn can have asignificant influence on the tensile and fracture toughness properties.This information is directly relevant to the fabrication of biomedical im-plants with sufficient mechanical strength for long-term in-vivo appli-cations. Short polymer deposition path in x − y plane promoted thebest bead-to-bead interfacial adhesion since the diffusion is fasterwhen the beads were in molten condition. However, it had drawbackin the final surface roughness due to the formation of bead-to-beadoverlap zone. Therefore, short deposition path could be considered forthe inner part of the sample while the long deposition path could beemployed for the outer surface to obtain better surface quality. Theshort deposition path need not necessarily be implemented with a 90°bead orientation as alternate bead orientations such as 45°/−45° and60°/−60° also exhibit almost the same thermal gradient in the horizon-tal layer as that of the 90° bead orientation.

Diffusion lines (interfaces atwhich the bonding is weak)were foundat the interface between the beads in the vertical fabrication direction inall specimen configurations due to high thermal gradient across beads.This had a dramatic effect when the load was applied normal to thisweakest zone, as exemplified by the tensile, flexural and fracture tough-ness properties of V-90° samples. Results suggest that minimizing ther-mal gradients across beads during FFF of thermoplastics with a high

melting temperature is the key to producing parts with excellent mac-roscopic material properties and dimensional stability. Thermal gradi-ent effects remain an important issue in FFF, leading to limitations inmechanical performance relative to conventionally molded samples.Ways such as precise control of the chamber temperature, post heat-treatment of the FFF fabricated samples [30] and microwave heatingof FFF samples fabricated with coated-CNT filament [43], would helpto minimize these limitations. The Poisson's ratio calculation can beused as a comparative indicator of the specimen's quality. For examplethe V-90° specimen had much lower Poisson's ratio compared to theH-0° and H-90° specimens which was mostly due to the opening ofthe voids during loading. The current study serves as an importantfirst step in the development of comprehensive guideline for the FFFof high temperature and high performance thermoplastics (e.g. PEEK)to enable high-end engineering applications such as for biomedicalimplants.

Acknowledgements

Authors would like to thank to Abu Dhabi Education Council (ADEC)for providing the research grant (EX2016-000006) through “the ADECAward for Research Excellence (A2RE) 2015”. SK would like to thankProfessor Brian Wardle, Massachusetts Institute of Technology for hiscomments on the MS.

References

[1] S. Kumar, B.L. Wardle, M.F. Arif, Strength and performance enhancement of bondedjoints by spatial tailoring of adhesive compliance via 3D printing, ACS Appl. Mater.Interfaces 9 (2017) 884–891, https://doi.org/10.1021/acsami.6b13038.

[2] S. Kumar, B.L. Wardle, M.F. Arif, J. Ubaid, Stress reduction of 3D printed compliance-tailored multilayers, Adv. Eng. Mater. (2017)https://doi.org/10.1002/adem.201700883 (1700883).

[3] J. Liljenhjerte, S. Kumar, Pull-out performance of 3D printed composites with em-bedded fins on the fiber, MRS Proc. 1800 (2015)https://doi.org/10.1557/opl.2015.645 (mrss15-2135597).

[4] J. Liljenhjerte, P. Upadhyaya, S. Kumar, Hyperelastic strain measurements and con-stitutive parameters identification of 3D printed soft polymers by image processing,Addit. Manuf. 11 (2016) 40–48, https://doi.org/10.1016/j.addma.2016.03.005.

[5] D.P. Cole, J.C. Riddick, H.M. Iftekhar Jaim, K.E. Strawhecker, N.E. Zander, Interfacialmechanical behavior of 3D printed ABS, J. Appl. Polym. Sci. 133 (2016) 1–12,https://doi.org/10.1002/app.43671.

[6] R. Gautam, S. Idapalapati, S. Feih, Printing and characterisation of Kagome latticestructures by fused deposition modelling, Mater. Des. 137 (2018) 266–275,https://doi.org/10.1016/j.matdes.2017.10.022.

[7] A. Tsouknidas, M. Pantazopoulos, I. Katsoulis, D. Fasnakis, S. Maropoulos, N.Michailidis, Impact absorption capacity of 3D-printed components fabricated byfused deposition modelling, Mater. Des. 102 (2016) 41–44, https://doi.org/10.1016/j.matdes.2016.03.154.

[8] N.G. Tanikella, B. Wittbrodt, J.M. Pearce, Tensile strength of commercial polymermaterials for fused filament fabrication 3D printing, Addit. Manuf. 15 (2017)https://doi.org/10.1016/j.addma.2017.03.005.

[9] R.J. Zaldivar, D.B. Witkin, T. McLouth, D.N. Patel, K. Schmitt, J.P. Nokes, Influence ofprocessing and orientation print effects on the mechanical and thermal behaviorof 3D-Printed ULTEM® 9085 Material, Addit. Manuf. 13 (2017) 71–80, https://doi.org/10.1016/j.addma.2016.11.007.

259M.F. Arif et al. / Materials and Design 146 (2018) 249–259

[10] S. Hertle, M. Drexler, D. Drummer, Additive manufacturing of poly(propylene) bymeans of melt extrusion, Macromol. Mater. Eng. 301 (2016) 1482–1493, https://doi.org/10.1002/mame.201600259.

[11] A.R. Torrado, C.M. Shemelya, J.D. English, Y. Lin, R.B. Wicker, D.A. Roberson, Charac-terizing the effect of additives to ABS on the mechanical property anisotropy ofspecimens fabricated by material extrusion 3D printing, Addit. Manuf. 6 (2015)16–29, https://doi.org/10.1016/j.addma.2015.02.001.

[12] Z. Weng, J.Wang, T. Senthil, L. Wu, Mechanical and thermal properties of ABS/mont-morillonite nanocomposites for fused deposition modeling 3D printing, Mater. Des.102 (2016) 276–283, https://doi.org/10.1016/j.matdes.2016.04.045.

[13] F. Ning, W. Cong, Z. Hu, K. Huang, Additive manufacturing of thermoplastic matrixcomposites using fused deposition modeling: a comparison of two reinforcements,J. Compos. Mater. (2017)https://doi.org/10.1177/0021998317692659(2199831769265).

[14] K. Prashantha, F. Roger, Multifunctional properties of 3D printed poly(lactic acid)/graphene nanocomposites by fused deposition modeling, J. Macromol. Sci. A 54(2017) 24–29, https://doi.org/10.1080/10601325.2017.1250311.

[15] S. Dul, L. Fambri, A. Pegoretti, Fused deposition modelling with ABS–graphene nano-composites, Compos. A: Appl. Sci. Manuf. 85 (2016) 181–191, https://doi.org/10.1016/j.compositesa.2016.03.013.

[16] W.W. Yu, J. Zhang, J.R. Wu, X.Z. Wang, Y.H. Deng, Incorporation of graphitic nano-filler and poly(lactic acid) in fused deposition modeling, J. Appl. Polym. Sci. 134(2017) 1–11, https://doi.org/10.1002/app.44703.

[17] A. Kantaros, D. Karalekas, Fiber Bragg grating based investigation of residual strainsin ABS parts fabricated by fused deposition modeling process, Mater. Des. 50 (2013)44–50, https://doi.org/10.1016/j.matdes.2013.02.067.

[18] C. Kousiatza, D. Karalekas, In-situmonitoring of strain and temperature distributionsduring fused deposition modeling process, Mater. Des. 97 (2016) 400–406, https://doi.org/10.1016/j.matdes.2016.02.099.

[19] M. Spoerk, J. Sapkota, G. Weingrill, T. Fischinger, F. Arbeiter, C. Holzer, Shrinkage andwarpage optimization of expanded-perlite-filled polypropylene composites inextrusion-based additive manufacturing, Macromol. Mater. Eng. (2017)https://doi.org/10.1002/mame.201700143 (1700143).

[20] B.N. Turner, R. Strong, S.A. Gold, A review of melt extrusion additive manufacturingprocesses: I. Process design and modeling, Rapid Prototyp. J. 20 (2014) 192–204,https://doi.org/10.1108/RPJ-01-2013-0012.

[21] S.F. Costa, F.M. Duarte, J.A. Covas, Thermal conditions affecting heat transfer in FDM/FFE: a contribution towards the numerical modelling of the process, Virtual Phys.Prototyp. 10 (2015) 35–46, https://doi.org/10.1080/17452759.2014.984042.

[22] S.M. Kurtz, J.N. Devine, PEEK biomaterials in trauma, orthopedic, and spinal im-plants, Biomaterials 28 (2007) 4845–4869, https://doi.org/10.1016/j.biomaterials.2007.07.013.

[23] S. Berretta, K.E. Evans, O.R. Ghita, Predicting processing parameters in high temper-ature laser sintering (HT-LS) from powder properties, Mater. Des. 105 (2016)301–314, https://doi.org/10.1016/j.matdes.2016.04.097.

[24] OsteoFab® Implants, http://oxfordpm.com/cmf-orthopedics/osteofab-implants,Accessed date: 1 January 2018.

[25] B. Valentan, Ž. Kadivnik, T. Brajlih, A. Anderson, I. Drstvenšek, Processing poly(etheretherketone) on a 3d printer for thermoplastic modelling,Mater. Technol. 47 (2013)715–721.

[26] M. Vaezi, S. Yang, Extrusion-based additive manufacturing of PEEK for biomedicalapplications, Virtual Phys. Prototyp. 10 (2015) 123–135, https://doi.org/10.1080/17452759.2015.1097053.

[27] W.Z. Wu, P. Geng, J. Zhao, Y. Zhang, D.W. Rosen, H.B. Zhang, Manufacture and ther-mal deformation analysis of semicrystalline polymer polyether ether ketone by 3Dprinting, Mater. Res. Innov. 18 (2014) S5-12–16, https://doi.org/10.1179/1432891714Z.000000000898.

[28] W.Wu, P. Geng, G. Li, D. Zhao, H. Zhang, J. Zhao, Influence of layer thickness and ras-ter angle on the mechanical properties of 3D-printed PEEK and a comparative me-chanical study between PEEK and ABS, Materials (Basel) 8 (2015) 5834–5846,https://doi.org/10.3390/ma8095271.

[29] G. Cicala, A. Latteri, B. Del Curto, A. Lo Russo, G. Recca, S. Farè, Engineering thermo-plastics for additive manufacturing: a critical perspective with experimental evi-dence to support functional applications, J. Appl. Biomater. Funct. Mater. 15(2017) 0, https://doi.org/10.5301/jabfm.5000343.

[30] C. Yang, X. Tian, D. Li, Y. Cao, F. Zhao, C. Shi, Influence of thermal processing condi-tions in 3D printing on the crystallinity and mechanical properties of PEEK material,J. Mater. Process. Technol. 248 (2017) 1–7, https://doi.org/10.1016/j.jmatprotec.2017.04.027.

[31] K.R. Hart, E.D.Wetzel, Fracture behavior of additivelymanufactured acrylonitrile bu-tadiene styrene (ABS) materials, Eng. Fract. Mech. 177 (2017) 1–13, https://doi.org/10.1016/j.engfracmech.2017.03.028.

[32] N. Aliheidari, R. Tripuraneni, A. Ameli, S. Nadimpalli, Fracture resistance measure-ment of fused deposition modeling 3D printed polymers, Polym. Test. 60 (2017)94–101, https://doi.org/10.1016/j.polymertesting.2017.03.016.

[33] Y. Song, Y. Li, W. Song, K. Yee, K.Y. Lee, V.L. Tagarielli, Measurements of the mechan-ical response of unidirectional 3D-printed PLA, Mater. Des. 123 (2017) 154–164,https://doi.org/10.1016/j.matdes.2017.03.051.

[34] M. Spoerk, F. Arbeiter, H. Cajner, J. Sapkota, C. Holzer, Parametric optimization ofintra- and inter-layer strengths in parts produced by extrusion-based additivemanufacturing of poly(lactic acid), J. Appl. Polym. Sci. (2017), 45401. https://doi.org/10.1002/app.45401.

[35] https://www.victrex.com/en/datasheets (accessed November 20, 2017).[36] J.E. Mark (Ed.), Polymer Data Handbook, Oxford University Press, 1999.[37] J. Karger-Kocsis, K. Friedrich, Temperature and strain-rate effects on the fracture

toughness of poly(ether ether ketone) and its short glass-fibre reinforced compos-ite, Polymer (Guildf.) 27 (1986) 1753–1760, https://doi.org/10.1016/0032-3861(86)90272-7.

[38] P. Beguelin, H.H. Kausch, The effect of the loading rate on the fracture toughness ofPoly(methyl methacrylate), Polyacetal, Polyetheretherketone and modified PVC, J.Mater. Sci. 29 (1994) 91–98, https://doi.org/10.1007/BF00356577.

[39] R. Gensler, P. Béguelin, C.J.G. Plummer, H.-H. Kausch, H. Münstedt, Tensile behaviourand fracture toughness of poly(ether ether ketone)/poly(ether imide) blends,Polym. Bull. 37 (1996) 111–118.

[40] P.J. Rae, E.N. Brown, E.B. Orler, The mechanical properties of poly(ether-ether-ke-tone) (PEEK) with emphasis on the large compressive strain response, Polymer(Guildf.) 48 (2007) 598–615, https://doi.org/10.1016/j.polymer.2006.11.032.

[41] T.W. Webb, E.C. Aifantis, Stick-slip instabilities in fracture, Comput. Mech. '95,Springer Berlin Heidelberg, Berlin, Heidelberg 1995, pp. 1353–1358, https://doi.org/10.1007/978-3-642-79654-8_219.

[42] M.F. Ashby, Materials Selection in Mechanical Design, 4th ed. Elsevier, 2011.[43] C.B. Sweeney, B.A. Lackey, M.J. Pospisil, T.C. Achee, V.K. Hicks, A.G. Moran, et al.,

Welding of 3D-printed carbon nanotube–polymer composites by locally inducedmicrowave heating, Sci. Adv. 3 (2017), e1700262. https://doi.org/10.1126/sciadv.1700262.