Additive Manufacturing 22 (2018) 508–515

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Interlayer fracture toughness of additively manufactured unreinforced and T

carbon-fiber-reinforced acrylonitrile butadiene styrene
⁎
Devin Young, Nelson Wetmore, Michael Czabaj
Department of Mechanical Engineering, University of Utah, 1495 East 100 South 1550 MEK, Salt Lake City, UT 84112, United States

A R T I C LE I N FO A B S T R A C T

Keywords: This study presents development of a test method for characterization of interlayer, mode-I fracture toughness of
Additive manufacturing fused filament fabrication (FFF) materials using a modified double cantilever beam (DCB) test. This test consists
Interlayer fracture toughness of DCB specimen fabricated from using unidirectional FFF layers, an 8 μm Kapton starter crack inserted in the
Fused filament fabrication midplane during the printing process, and reinforcing glass/epoxy doublers to prevent DCB arm failure during
Fused deposition modeling
loading. DCB specimens are manufactured with a commercially available 3D printer using unreinforced
Acrylonitrile butadiene styrene
Acrylonitrile Butadiene Styrene (ABS) and chopped carbon-fiber-reinforced ABS (CF-ABS) filaments. To examine
the effect of the FFF printing process on fracture toughness, additional ABS and CF-ABS specimens are hot-press
molded using the filament material, and tested with the single end notch bend (SENB) specimen configuration.
The fracture toughness data from DCB and SENB tests reveal that the FFF process significantly lowers the mode-I
fracture toughness of ABS and CF-ABS. For both materials, in situ thermal imaging and post-mortem fractography
shows, respectively, rapid cool-down of the rasters during filament deposition and presence of voids between
adjacent raster roads; both of which serve to reduce fracture toughness. For CF-ABS specimens, fracture
toughness is further reduced by inclusion of poorly wetted chopped carbon fibers. Although this study did not
attempt to optimize the fracture performance of FFF specimens, the results demonstrate that the proposed
methodology is suitable for design and optimization of FFF processes for improved interlayer fracture perfor-
mance.

1. Introduction To date, limited research has been conducted to systematically re-

Additive manufacturing by fused filament fabrication1 (FFF) holds
promise for on-demand manufacturing of tailored objects for a wide
variety of applications, including rapid prototyping, part replacement,
and tooling for composite layup, to name a few. Despite the apparent
advantage over more traditional manufacturing methods (e.g. manu-
facturing by subtraction, injection molding), FFF parts often suffer from
poor mechanical characteristics, limiting their broader adaptation for
end-use, load-bearing components. The poor performance of FFF parts
often results from the use of polymeric base materials that are selected
based on rheological properties to improve the extrusion process, in-
stead of materials that exhibit exceptional mechanical properties. Ad-
ditionally, the quality of final fully-formed FFF parts is affected by part
intricacy and the corresponding print path, and the differential cool-
down and solidification of the individual rasters, among other factors.
The complexity of the FFF process results in parts that are often highly
anisotropic with properties that are difficult to characterize, let alone
predict.
late properties of FFF base materials, the printing process, the resulting
structure, and the mechanical performance. A few studies have found
that optimizing parameters like extrusion and print bed temperatures,
print speed, layer height, and infill percentage can increase tensile
strength and stiffness of FFF parts [1–3]. However, regardless of para-
meter optimization, FFF parts still exhibit lower properties compared to
those obtained by conventional polymer processing methods such as
compression or injection molding [1,4–8]. A few recent studies have
attempted to increase strength and stiffness of FFF parts by combining
polymeric filaments with chopped glass [9] and carbon fibers [10–12],
carbon nanotubes [10,13], zinc oxide nanorods [14], and Jute plant
fibers [14]. In most cases, the addition of reinforcement resulted in
marginal increase in mechanical properties, that are still well below
properties of metallic alloys and continuous-fiber reinforced polymers.
In addition to poor mechanical performance, FFF parts exhibit low
interlayer and intralayer fracture properties. In the context of this work,
interlayer fracture is defined as decohesion of adjacent layers of an FFF
part, and is similar to delamination between plies of a tape-laminate

⁎
Corresponding author.
E-mail address: m.czabaj@utah.edu (M. Czabaj).
1
Fused Filament Fabrication is a material extrusion technology as per ISO/ASTM 52900.

https://doi.org/10.1016/j.addma.2018.02.023
Received 15 November 2017; Received in revised form 7 February 2018; Accepted 21 February 2018
Available online 06 June 2018
2214-8604/ © 2018 Published by Elsevier B.V.
D. Young et al. Additive Manufacturing 22 (2018) 508–515

2. Specimen design and manufacturing

2.1. DCB specimens

The DCB specimens were designed based on the guidelines provided

in the ASTM D5528 standard with one notable exception. Given the low
stiffness of ABS and CF-ABS materials relative to fiber-reinforced
composites, the FFF-manufactured DCB specimens were stiffened with
composite doublers. This was done to reduce the overall specimen de-
flection at the onset of fracture, and to prevent flexure-induced failure
of the arms. Based on previous experience with fracture-toughness
testing of thin adhesives, 3.1 mm thick G10 glass/epoxy laminates were
used as the doubler material. The actual ABS and CF-ABS specimens
were manufactured by printing twenty 0.2 mm thick layers. Similar to
the D5528 standard, an 8 μm Kapton film (SPEX SamplePrep;
Metuchen, NJ, USA) was placed between the 10th and 11th layer
Fig. 1. Definition of interlayer and intralayer fracture in FFF manufactured during the printing process to create a sharp starter crack. The resulting
unidirectional components. DCB specimen geometry is depicted in Fig. 2. As seen in this figure, total
specimen thickness was approximately 10.3 mm, total length was
152 mm, and initial starter crack length was 50.8 mm. After final ma-
composite. Intralayer fracture is defined as decohesion between ad-
chining the average specimen width was 22.5 mm, which corresponds
jacent raster roads, and is analogous to intralaminar ply cracking. The
to approximately 46 raster roads for ABS specimens and 23 raster roads
difference between the two fracture types in unidirectional FFF com-
for CF-ABS specimens.
ponents is shown in Fig. 1. To date, characterization of fracture prop-
The FFF specimens were manufactured in pairs by printing a 48 mm
erties of FFF components has been attempted indirectly using tensile
× 158 mm × 4 mm plaque as shown in Fig. 3. In this figure, the heavy
testing [9,15]. A recent study utilized several double cantilever beam
lines correspond to the outline of the entire plaque, while the dashed
(DCB) and T-peel test configurations to measure adhesion between two
lines show the cut lines for each specimen. The light-orange region
Polyjetted photopolymer layers [16]. Currently, there are no viable
corresponds to location of the Kapton insert. To ensure that the crack
fracture toughness tests that can be applied to a broad range of FFF
faces remain closed during the print and doubler adhesion, a 13 mm
materials.
section was printed between the end of the Kapton film and the right
In light of the above discussion, this study aims to develop a test
edge of the plaque. To reduce the effect of thermal residual stresses, the
methodology for characterization of interlayer fracture toughness
four corners of the plaque were filleted to a 2 mm radius. Each layer
properties of FFF materials. This test has the potential to be used in
was printed using a rectilinear fill pattern with the raster orientation
optimization of FFF processes for improved interlayer fracture tough-
aligned along the direction of the expected crack growth. This was done
ness, evaluation of new FFF materials, or assessment of novel interlayer
to prevent possible crack migration caused by the undulations between
inclusions (e.g. sensors, antennas) and their effect on fracture perfor-
the rasters. To minimize the internal void content, the infill percentage
mance. The test geometry used in this study is based on a DCB test
was set to 100%. In the context of FFF, infill percentage is defined as the
described in ASTM D5528 test standard for unidirectional fiber-re-
degree to which the internal volume is filled with raster roads. Al-
inforced polymer matrix composites [17]. Given the maturity and
though an infill of 100% implies no gap between adjacent rasters, some
straightforward nature of D5528, the specimen geometry, test proce-
voiding is usually observed due to the elliptical cross-section of the
dures, and data reduction are adopted as closely as possible. The suit-
solidified rasters. Thus 100% infill minimizes void content but does not
ability of the test described herein is evaluated by testing specimens
fully eliminate it.
manufactured from Acrylonitrile Butadiene Styrene (ABS), and ABS
The ABS specimens were fabricated using Polylac® PA-747 natural
reinforced with short carbon fibers (CF-ABS). All specimens are man-
ABS filament (Village Plastics; Barberton, OH, USA). The extrusion was
ufactured using a commercially available 3D printer. To examine the
performed using a 0.35 mm dia. circular nozzle at 235 °C. The print bed
effect of the FFF printing process, fracture toughness was also measured
was set to 95 °C throughout the printing process. The CF-ABS specimens
using ABS and CF-ABS filaments that were hot-press molded (HPM).
were fabricated using 3DXTech CarbonX™ carbon-fiber reinforced fi-
The HPM materials were tested using the single end notch bend (SENB)
lament (3DXTech; Byron Center, MI, USA), which has a reported fiber
specimen configuration according to ASTM standard D5045 [18]. In
volume fraction of 15% [19]. Based on resin digestion and optical mi-
what follows, a description of the specimen manufacturing process is
croscopy, the average fiber length in CarbonX™ filament is approxi-
presented, including printing of DCB specimens, adhesion of reinforcing
mately 81 μm, with standard deviation of ± 20 μm. The CF-ABS speci-
doublers, and insertion of interlayer starter cracks during the printing
mens were printed using 0.50 mm dia. circular nozzle, an extrusion
process. The test procedure is presented, followed by results from
temperature of 225 °C, and print bed temperature of 60 °C. The reduced
fracture toughness testing, fractography, and in situ thermal imaging.
print bed temperature was used to reduce warping of CF-ABS plaques

Fig. 2. Geometry of the modified DCB specimen.

509
D. Young et al.
Fig. 3. Schematic of the FFF plaque used for manufacturing of DCB specimens.
during printing. The printing was performed using a Lulzbot TAZ 5 FFF
printer (Aleph Objects, Inc.; Loveland, CO, USA) placed inside of a
Plexiglas® (Arkema; King of Prussia, Pennsylvania, USA) enclosure to
minimize cooling due to ambient air currents. Prior to printing, the
print bed was coated with a thin layer of Avery® Permanent Glue (Avery
Products Corporation; Strongsville, OH, USA) to promote adhesion to
the extruded filament.
To simplify placement of the Kapton insert mid-print, all specimens
were printed at a speed of 35 mm/s. Placement of Kapton between the
10th and 11th layer was done using a custom U-shaped jig as depicted
in Fig. 4. The Kapton film was stretched taut across the arms of the jig
and taped in place using adhesive Kapton tape. Prior to placement, the
Kapton film was coated with Avery® glue to promote adhesion to the
extruded filament. The actual placement was performed without stop-
ping the printing process. This was achieved by opening the enclosure
door and quickly butting the jig against the specimen while the nozzle
was in the final stage of printing the 10th layer. After completion of the
printing process, all specimens were carefully detached from the print
bed and rinsed with water to remove excess Avery® glue.
The next step of specimen manufacturing was the adhesion of the
doubler material. To promote good adhesion between the ABS material
and the doublers, all mating surfaces were textured by surface abrasion.
The top and bottom surfaces of the FFF specimen were abraded in two
steps. First, the surfaces were sand-blasted using 180 grade brown
aluminum oxide grit at 275 KPa. Second, a 240-grit sand paper was
used to manually abrade both surfaces until a consistent texture was
achieved on both sides. The G10 doublers were sand-blasted using the
same grit and pressure, but manual abrasion was omitted. After surface
abrasion, the FFF specimens and doublers were wiped clean with iso-
propyl alcohol. The adhesion was performed using Loctite® 9460TM
Fig. 4. Placement of the Kapton insert in-between the midplane of the FFF
specimen.
510
Additive Manufacturing 22 (2018) 508–515
Hysol® (Henkel; Dusseldorf, Germany) structural adhesive mixed with
0.59 mm dia. glass beads. The cure was performed at room temperature
and 765 KPa pressure for 24 h. The room temperature cure was ne-
cessary to avoid buildup of thermal residual stresses caused by differ-
ential thermal expansion of ABS relative to G10. Note that surface
preparation and adhesive were chosen specifically for ABS-to-G10 ad-
hesion, and may not apply to other FFF and doubler materials. Also,
depending on the flexural stiffness of other FFF materials that may be
studied with this method, alternative doubler thicknesses should be
considered.
After cure, the resulting sandwich plate was cut to exact specimen
dimensions given in Fig. 2, using a 1.6 mm thick diamond saw blade.
Aluminum MW2001-6 loading hinges from Genuine Aircraft Hardware
Co. (Paso Robles, CA, USA) were adhered to the doublers using Loctite®
9460TM Hysol® and a custom hinging jig. Prior to adhesion, the hinges
were sandblasted and cleaned with isopropyl alcohol. After hinging, a
band saw was used to cut a notch through the leading 8 mm of material
allowing the crack faces to freely open. The final configuration of the
DCB specimen modified for FFF materials is depicted in Fig. 5. In total,
6 ABS and 5 CF-ABS specimens were manufactured. DCB specimens
were numbered with the designation “material-Y-X”, where “material”
is ABS or CF-ABS, Y is the plaque number, and X is the specimen
number.
2.2. SENB specimens
The SENB specimens were made in accordance with ASTM D5045-
14 [18] test method for fracture toughness of plastics materials. Resin
plaques with dimensions of 76 mm × 76 mm × 6 mm were made from
both FFF filament feedstock by HPM. Strands of ABS and CF-ABS fila-
ment were chopped into 1–3 mm pieces, arranged in the mold, and
vacuum bagged. Both materials were cured at 230 °C for 105 min, fol-
lowed by a 60-min cool-down to room temperature. To achieve proper
consolidation, 766 kPa and 919 kPa pressures were used for ABS and
CF-ABS plaques, respectively. After cure, both plaques were trimmed
into rectangular beam specimens with dimensions of 44 mm × 10 mm
× 5 mm. Notches were made by first scoring one side of the SENB with
a jeweler’s saw, followed by tapping a razor blade into the notch to
extend the crack. Additional 3-point bend specimens were manu-
factured and left un-notched. These specimens were used to correct the
measured displacement for compliance of the testing apparatus, load-
pin penetration, and specimen compression.
In total, 6 ABS and 3 CF-ABS specimens were hot press molded.
Molded SENB specimens were labeled in the same manner as the DCB
specimens with the letter “M” appended to the end, to represent the
molding process (i.e. “material-Y-XM”).
D. Young et al.
Fig. 5. Final configuration of a DCB specimen modified for FFF materials.
Fig. 6. Typical force-displacement data from ABS and CF-ABS tests.
3. Testing procedure
DCB specimens were tested in an MTS servo-hydraulic load frame
instrumented with a 445 N load cell. Specimen displacement was
measured using the load frame’s internal linear variable displacement
transducer. The ABS and CF-ABS specimens were loaded at a rate of
2 mm/min and 1 mm/min, respectively. Different loading rates for each
material were needed to enable accurate monitoring of crack extension.
The force-displacement data were recorded once every second. Crack
growth was recorded using two DinoXLite AM–414M (AnMo
Electronics Corporation; New Taipei City, Taiwan) digital microscopes
placed on each side of the specimen. Both cameras were interfaced to
an in-house LabVIEW program that synched image acquisition with the
corresponding force and displacement values. During each test, crack
extension was recorded every 1 mm for the first 10 mm beyond the
Kapton insert. Thereafter, measurements were to be taken every 5 mm
until 30 mm of total growth had occurred. For two specimens, ABS-1-2
and ABS-3-2, doubler debonding was observed after 13 mm and 10 mm
of crack growth, respectively. As a result, for two specimens, data re-
duction was performed over a smaller crack growth range. All CF-ABS
specimens exhibited stick/slip fracture, where crack extension alter-
nated between stable (i.e. quasi-static) and unstable (dynamic). As a
result, crack extension and the corresponding force-displacement data
points, were recorded only during stable extension. Following the
D5528 standard [17], a few ABS and CF-ABS specimens were unloaded
after approximately 3 mm of crack extension to assess the extent of
damage (e.g. yielding, microcracking, doubler debonding) in the spe-
cimens’ arms. All specimens unloaded linearly back to the origin; thus,
511
Additive Manufacturing 22 (2018) 508–515
the unloading procedure was omitted for the remaining specimens.
The mode I fracture toughness, GIc, for ABS and CF-ABS specimens
were calculated using the unmodified beam theory expression from
[17]:
3Pc δc
GIc =
2ba (1)
where Pc is the critical load at the onset of fracture, δc is the corre-
sponding critical load-point deflection, b is specimen width, and a is the
crack length. Use of this data reduction method, instead of a more ac-
cepted compliance calibration [17], was based on previous work where
DCB specimens exhibited nonlinear response prior to onset of fracture
[20].
The SENB specimens were tested on an Instron 4303 screw-driven
load frame in a 3-point bend configuration. The support span was
40 mm, and the loading roller diameter was 6.35 mm. Load was re-
corded using a 5 kn load cell, and displacement was recorded using an
encoder in the load frame’s lead screw. Loading rate was 3 mm/min.
Following the D5045-14 standard, GIc values were determined using the
area method [18].
4. Results and discussion
4.1. Fracture toughness and SEM fractography
Representative force-displacement data for ABS and CF-ABS DCB
specimens are shown in Fig. 6. The hollow circular marks in each curve
indicate the force-displacement points where the instantaneous crack
tip positions were measured, and used in calculation of GIc. As seen in
this figure, both specimen types were slightly nonlinear prior to mac-
roscopic crack extension, which was likely caused by the inherent onset
of cracking in the center of DCB specimens and/or some crack-tip
plasticity. The ABS specimens exhibited stable crack growth throughout
the entire test. Conversely, the onset of crack growth in the CF-ABS
specimens was initially unstable, resulting in rapid load drop combined
with dynamic crack advance, Δa, of approximately 4–6 mm. This was
typically followed by 1–2 mm of stable crack growth, and another un-
stable extension. The process of stable and unstable growth was re-
peated throughout the entire test, and was observed in all CF-ABS
specimens tested. To ensure consistency when comparing CF-ABS and
ABS propagation toughness data, all crack-extension points captured
during dynamic advance (not shown in Fig. 6) were omitted from data
reduction.
The non-precracked (NPC) fracture toughness values measured from
the Kapton insert, GIc−NPC, were determined based on the 5%-offset/
maximum-load definition of Pc as defined in [17]. The precracked (PC)
toughness, GIc−PC, values were measured from a sharper, naturally
occurring crack front on the next increment of growth. In addition,
average propagation toughness, Gavg, for each specimen was calculated
by averaging toughness values measured between the second and last
D. Young et al.
Fig. 7. Mode I fracture toughness for ABS and CF-ABS specimens (solid sym-
bols) with benchmark fracture toughness values of hot press molded material
(hollow symbols).
increment of growth. A summary of these data for ABS and CF-ABS
specimens is presented in Table A1 of the Appendix A. For complete-
ness, Table A1 of the Appendix A includes GIc−NPC calculated based on
5%-offset and maximum-load definitions of Pc. The same data are de-
picted in Fig. 7 as a function of crack extension, Δa. In this figure, in-
itiation toughness at Δa = 0 (i.e. GIc−NPC) is based on the 5%-offset
method. In addition to the DCB data, Fig. 7 includes toughness values of
molded ABS and CF-ABS measured using the SENB test. Given the in-
herently unstable nature of crack extension in the SENB tests, only one
value of toughness was recorded per specimen. Also, since all specimens
are initially precracked by tapping a sharp razor blade into a pre-ma-
chined notch, all toughness values from SENB specimens are referred to
as PC (i.e. measured from a sharp, naturally occurring crack front).
Consequently, SENB data in Fig. 7 are presented as single points
(hollow symbols) at Δa ≅ 0, with error bars corresponding to plus/
minus one standard deviation in toughness. The summary of toughness
data from SENB tests is presented in Table A2 of the of the Appendix A.
Following fracture toughness testing, several FFF specimens were
split open and imaged using a scanning electron microscope (SEM). The
SEM imaging was performed with magnification of 50x, 300x, and
1000x. All specimens were imaged within the region of stable crack
growth, which was determined based on the force versus crack-exten-
sion data. Fig. 8 depicts a set of representative SEM images obtained for
each material type, with ABS and CF-ABS shown in the left and right
columns, respectively. In all six micrographs, the direction of macro-
scopic crack growth is from right to left, as indicated. The white square
boxes in the upper four images indicate regions that were examined at a
higher magnification.
Examining Fig. 8, at 50x the ABS fracture surfaces are textured,
clearly showing resin ridgelines and evidence of textured microflow
that is inherent to mode-I crack growth in polymers [21]. Images of ABS
at 300x and 1000x show that the surface roughness is caused by sig-
nificant plastic deformation during crack growth. The CF-ABS fracture
surfaces appear much smoother at 50x, showing little evidence of
ductile deformation during crack growth. At 300x and 1000x, the CF-
ABS fracture surfaces reveal a large number of short carbon fibers,
which are either completely separated from, or only partially bonded
to, the surrounding ABS resin. The CF-ABS specimens show very little
evidence of fiber pull-out or fiber bridging across adjacent rasters,
512
Additive Manufacturing 22 (2018) 508–515
further indicating lack of fiber-to-ABS adhesion. At 1000x, the resin in
CF-ABS specimens shows evidence of plastic deformation, suggesting
some adhesion between adjacent rasters. Finally, for both specimen
types, images taken at 50x show a series of evenly spaced intra-raster
voids (vis. lengthwise channels between adjacent rasters) that are in-
herently formed during filament extrusion.
Examining the data presented in Figs. 7 and 8 and Table A1 and A2
of the Appendix A, several observations can be made about the influ-
ence of the FFF process on the mode I fracture toughness of 3D printed
parts. Comparing the GIc-PC values resulting from the two manu-
facturing methods (FFF vs. HPM), the FFF process resulted in a twofold
reduction in toughness for ABS specimens, and tenfold reduction for CF-
ABS specimens. Comparing the GIc-PC and GIc-avg values for both mate-
rial types, the FFF process resulted in relatively flat R-curves. For a few
ABS specimens, the slight increase in GIc with Δa was related to intra-
raster crack migration and raster bridging (i.e. process that is analogous
to fiber bridging in tape-laminate composites). For CF-ABS specimens,
the flat R-curves are consistent with the lack of fiber bridging and fiber
pull-out seen in Fig. 8. Examining the toughness data in Table I, the
difference in GIc values based on 5%-offset and max-load definitions of
Pc was approximately 30% for ABS, and 6% for CF-ABS. These differ-
ences are consistent with the SEM observations, confirming ductile and
brittle fracture processes for ABS and CF-ABS specimens, respectively.
4.2. Effect of inter-raster voids
The apparent reduction of mode-I fracture toughness in FFF speci-
mens can be linked directly to the manufacturing process. As shown in
Fig. 8, extrusion of filament during FFF results in formation of evenly
spaced inter-raster voids, which decrease the total surface area involved
in the fracture process. However, Eq. (1) assumes that the crack surface
area is equal to the entire specimen width, b, times the crack length, a.
Since the actual width of the fractured material is less than b, Eq. (1)
underestimates the actual fracture toughness (i.e. energy per area)
needed to fracture FFF rasters. To estimate the “true” fracture tough-
ness of ABS and CF-ABS rasters, GIc-raster, the effective specimen width,
b*, involved in the fracture process was measured directly. This was
done by optically imaging fracture surfaces of all tested specimens, and
using grayscale thresholding to subtract from the nominal specimen
width the cumulative width of the inter-raster voids. This analysis
showed that, on average, inter-raster voids corresponded to 38% and
11% of width for ABS and CF-ABS specimens, respectively. Given the
corrected width for all specimens, the fracture toughness values for all
FFF specimens were recalculated using Eq. (1) and plotted in Fig. 9. As
seen in this figure, for ABS specimens, the width correction resulted in
average GIc-raster that was approximately 17% lower than GIc from SENB
specimens. The width correction did, however, result in a large scatter
in the toughness data, which was likely caused by variability in ex-
traction of b* using an optical method. For CF-ABS, GIc-raster increased
marginally, and remained almost 7 times lower than GIc from SENB
specimens.
4.3. Effect of print temperature
The data in Fig. 9 suggests that the presence of inter-raster voids has
a strong effect on the apparent fracture toughness of FFF parts; how-
ever, excessive voiding does not fully explain the significant reduction
in toughness, especially for the CF-ABS specimens. To investigate this
further, a selected number of specimens were imaged using a FLIR 2420
(FLIR® Systems, Inc.; Wilsonville, OR, USA) thermal camera in situ
during the printing process of the midplane layers. An example of this
imaging is presented in Fig. 10, where the left image shows the full-field
temperature of the print, and the right plot shows the temperature
profile along the newly printed raster. Note that the left image is plotted
on a plateau equalization scale to enhance the contract between the
newly deposited raster and the surrounding material. Examining
D. Young et al. Additive Manufacturing 22 (2018) 508–515

Fig. 8. Representative fracture surfaces from ABS (left) and CF-ABS (right) specimens. The scale bars on the left apply to both sets of images. The white boxes indicate
magnified regions in subsequent micrographs. Inter-raster voids are indicated with white arrows.

Fig. 10, the ABS raster temperature drops by nearly 100° C in the first
second after deposition, and reaches the glass transition temperature of
105° C in approximately 4 s. The CF-ABS rasters cool much quicker, and
reach 105° C in approximately 1 s. Within 4.5 s, both rasters reach their
respective bed temperatures of 95° C and 60° C. Similar observations
were reported in Ref. [22], suggesting that rapid raster cooldown
during FFF is not specific to this study. By comparison, specimens
created using HPM remain at temperature of 230° C for the entire mold
time of 105 min, allowing polymer chains sufficient time to fully en-
tangle. The quick cool-down of FFF rasters, especially for CF-ABS parts,
leave little time for proper polymer chain entanglement, effectively
reducing the bond strength between adjacent rasters.

4.4. Effect of chopped carbon fibers

The evidence presented in the above two sections suggest that

Fig. 9. Corrected mode I fracture toughness for ABS and CF-ABS specimens
recalculated by accounting for inter-raster voids.
voiding and processing temperature are the primary reasons for re-
duction in fracture toughness in FFF specimens. However, for CF-ABS
specimens, reduction in fracture toughness can also be attributed to the
chopped carbon fibers. As shown in bottom right image in Fig. 8,
carbon fibers suffer from poor resin wetting, and effectively act as
brittle inclusions. Previous studies have shown that such brittle inclu-
sions can reduce mode I fracture toughness in polymers [23,24]. Ex-
amining the HPM data in Table II, these inclusions appear to reduce the
fracture toughness of CF-ABS specimens by approximately 10% relative

513
D. Young et al. Additive Manufacturing 22 (2018) 508–515

Fig. 10. Results from in situ thermal imaging

of an ABS and CF-ABS DCB specimens during
fabrication. The left image depicts the full-field
temperature profile of the ABS specimen at the
end of raster deposition. The right image shows
the temperature between points A and B as a
function of distance and time from nozzle for
both ABS and CF-ABS.

to pure ABS. This effect, in combination with voiding and processing
temperature, is the likely explanation for a nearly ten-fold reduction in
mode I fracture toughness of FFF CF-ABS compared to HPM CF-ABS.
5. Conclusions
This study presents a test methodology for characterization of in-
terlayer fracture toughness of FFF components using a modified version
of ASTM D5528 standard. DCB specimens were manufactured with a
commercially available 3D printer using pure ABS, and ABS reinforced
with short carbon fibers. Novel contributions of this work include: 1)
design of unidirectional FFF DCB specimens; 2) insertion of Kapton
starter cracks during the printing process; 3) adhesion of reinforcing
glass/epoxy doublers to prevent DCB arm failure during loading; 4) and
data analysis based on fractographic imaging and in situ full-field
monitoring of the print temperature using a thermal camera. The test
data obtained in this study revealed that, if not properly optimized, the
FFF process can significantly reduce mode I fracture toughness relative
to parts manufactured using HPM. Detailed investigation of the FFF
specimens attributed reduction in fracture toughness to the presence of
inter-raster voids which are inherent to the FFF process, and the rapid
cool-down of rasters during filament deposition. For CF-ABS specimens,
toughness was further reduced by inclusion, and poor wetting, of the
chopped carbon fibers.
It must be noted that during the course of this research, no attempts
were made to optimize the FFF processing parameters to improve the
fracture toughness. Rather, processing was optimized to create DCB
specimens with good dimensional quality, which was necessary to
develop a reproducible test method. In order to more fully understand
the differences between interlayer toughness of the two FFF materials
described herein, a more in depth investigation may be required. It is
possible that careful optimization of the print conditions can sig-
nificantly improve the interlayer toughness of the CF-ABS materials.
This, however, was not the main focus of this work, and is currently a
subject of a separate ongoing study. Nevertheless, the above discussion
demonstrates that the proposed DCB test can act as an effective tool for
screening interlayer performance of FFF components and various
manufacturing processes. This method, combined with other material
characterization tests developed specifically for polymeric FFF com-
ponents, may prove critical for their broader acceptance in structural
applications.
Funding
This work was supported by NASA Langley Research Center [grant
number NNX14AQ74A].
Acknowledgements
This work is funded by NASA Langley Research Center (Grant No.
NNX14AQ74A) with Dr. Emilie J. Siochi as the grant monitor. Authors
would like to thank Dr. Siochi and Dr. James Ratcliffe from NASA LaRC
for their invaluable technical support. In addition, the authors would
like to thank Mr. Jim Baughman from NASA LaRC for help with the
SEM analysis, and Mr. Jeff Kessler for the Utah Composites Laboratory
for help with fracture toughness testing.

Appendix A

Table A1
Summary of fracture toughness data from ABS and CF-ABS specimens, along with the corresponding averages (AVG), standard deviation (SD), and coefficients of
variation (CV).
Sample 5% offset MAX GIc-PC (kJ/m2) GIc-avg (kJ/m2)
GIc-NPC (kJ/m2) GIc-NPC (kJ/m2)

ABS-1-1 1.39 1.73 1.76 1.65

ABS-1-2 1.19 1.95 1.95 2.08
ABS-2-1 1.14 1.45 1.57 1.52
ABS-2-2 1.43 1.92 1.86 1.94
ABS-3-1 1.24 1.66 1.71 1.72
(continued on next page)

514
D. Young et al. Additive Manufacturing 22 (2018) 508–515

Table A1 (continued)

Sample 5% offset MAX GIc-PC (kJ/m2) GIc-avg (kJ/m2)

GIc-NPC (kJ/m2) GIc-NPC (kJ/m2)

ABS-3-2 1.34 1.78 1.69 1.90

AVG 1.29 1.75 1.76 1.80

SD 0.12 0.18 0.13 0.21
CV (%) 9.0 10.5 7.6 11.5

CF-ABS-1-1 0.45 0.47 0.44 0.43

CF-ABS-2-1 0.34 0.38 0.38 0.37
CF-ABS-2-2 0.29 0.30 0.30 0.29
CF-ABS-3-1 0.31 0.34 0.34 0.32
CF-ABS-3-2 0.34 0.36 0.39 0.39

AVG 0.35 0.37 0.37 0.36

SD 0.06 0.06 0.05 0.06
CV (%) 17.6 17.1 14.3 15.5

Table A2
Summary of fracture toughness data from SENB tests.
ABS CF-ABS

2
Sample GIc-NPC (kJ/m ) Sample GIc-NPC (kJ/m2)

ABS-1-M 3.28 CF-ABS-1-M 3.38

ABS-2-M 3.52 CF-ABS-2-M 3.23
ABS-3-M 3.35 CF-ABS-3-M 2.67
ABS-4-M 3.67 – –
ABS-5-M 3.38 – –

AVG 3.44 AVG 3.09

SD 0.15 SD 0.38
CV (%) 4.49 CV (%) 12.21

References
[1] Ahn Sung-Hoon, et al., Anisotropic material properties of fused deposition modeling
ABS, Rapid Prototyp. J. 8 (4) (2002) 248–257.
[2] B.M. Tymrak, M. Kreiger, M. Joshua, Pearce Mechanical properties of components
fabricated with open-source 3-D printers under realistic environmental conditions,
Mater. Des. 58 (2014) 242–246.
[3] Anoop Kumar Sood, Raj K. Ohdar, Siba S. Mahapatra, Parametric appraisal of
mechanical property of fused deposition modelling processed parts, Mater. Des. 31
(1) (2010) 287–295.
[4] Aaron M. Forster. Materials Testing Standards for Additive Manufacturing of
Polymer Materials: State of the Art and Standards Applicability. NIST Interagency/
Internal Report (NISTIR)-8059, 2015.
[5] W.C. Smith, R.W. Dean, Structural characteristics of fused deposition modeling
polycarbonate material, Polym. Test. 32 (8) (2013) 1306–1312.
[6] A. Bagsik, V. Schöppner, E. Klemp, FDM part quality manufactured with Ultem*
9085, 14th International Scientific Conference on Polymeric Materials vol. 15,
(2010).
[7] R.J. Zaldivar, et al., Influence of processing and orientation print effects on the
mechanical and thermal behavior of 3D-printed ULTEM® 9085 material, Addit.
Manuf. 13 (2017) 71–80.
[8] Carsten Koch, Luke Van Hulle, Natalie Rudolph, Investigation of mechanical ani-
sotropy of the fused filament fabrication process via customized tool path genera-
tion, Addit. Manuf. 16 (2017) 138–145.
[9] W. Zhong, et al., Short fiber reinforced composites for fused deposition modeling,
Mater. Sci. Eng. A 301 (2) (2001) 125–130.
[10] George Tsiakatouras, Eirini Tsellou, Constantinos Stergiou, Comparative study on
nanotubes reinforced with carbon filaments for the 3D printing of mechanical parts,
World Trans. Eng. Technol. Educ. 12 (3) (2014) 392–396.
[11] Halil L. Tekinalp, et al., Highly oriented carbon fiber–polymer composites via ad-
ditive manufacturing, Compos. Sci. Technol. 105 (2014) 144–150.
[12] Fuda Ning, et al., Additive manufacturing of carbon fiber reinforced thermoplastic
composites using fused deposition modeling, Compos. B-Eng. 80 (2015) 369–378.
[13] M. Gardner John, et al., 3-D printing of multifunctional carbon nanotube yarn re-
inforced components, Addit. Manuf. 12 (2016) 38–44.
[14] R. Torrado Angel, et al., Characterizing the effect of additives to ABS on the me-
chanical property anisotropy of specimens fabricated by material extrusion 3D
printing, Addit. Manuf. 6 (2015) 16–29.
[15] J.F. Rodriguez, J.P.J.E. Thomas Renaud, Maximizing the strength of fused-deposi-
tion ABS plastic parts, Solid Freeform Fabrication Proceedings, Notre Dame, IN,
1999.
[16] Ivan Vu, et al., Characterization of multi-material interfaces in PolyJet additive
manufacturing, Solid Freeform Fabrication Symposium (2014).
[17] ASTM International, ASTM D5528-13, Standard test method for mode I inter-
laminar fracture toughness of unidirectional fiber-reinforced polymer matrix com-
posites, West Conshohocken, PA, 2013.
[18] ASTM International, ASTM D5045-14, Standard test methods for plane-strain
toughness and strain energy release rate of plastic materials, West Conshohocken,
PA, 2014.
[19] 3DXTECH. http://www.3dxtech.com/carbonx-carbon-fiber-abs-3d-printing-
filament/, 2017 (Accessed 19.06.17).
[20] Michael W. Czabaj, Barry D. Davidson, Determination of the mode I, mode II, and
mixed-mode I–II delamination toughness of a graphite/polyimide composite at
room and elevated temperatures, J. Compos. Mater. 50 (16) (2016) 2235–2253.
[21] E. Greenhalgh, Failure Analysis and Fractography of Polymer Composites, CRC
Press, LLC, Boca Raton FL, 2009 p. 169.
[22] Jonathan E. Seppala, D. Kalman, Migler, Infrared thermography of welding zones
produced by polymer extrusion additive manufacturing, Addit. Manuf. 12 (2016)
71–76.
[23] K. Friedrich, U.A. Karsch, Fatigue crack growth and fracture in glass sphere filled
nylon 6, Polym. Compos. 3 (2) (1982) 65–74.
[24] K. Friedrich, Microstructural efficiency and fracture toughness of short fiber/ther-
moplastic matrix composites, Compos. Sci. Technol. 22 (1) (1985) 43–74.

515