Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Materials & Design 226 (2023) 111687

Contents lists available at ScienceDirect

Materials & Design


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

The effects of thermal annealing on the performance of material


extrusion 3D printed polymer parts
Wangwang Yu a, Xinzhou Wang b,⇑, Xinshun Yin b, Eleonora Ferraris c, Jie Zhang c,⇑
a
School of Mechanical Engineering, Nanjing Vocational University of Industry Technology, Nanjing 210023, China
b
Jiangsu Co-Innovation Center of Efficient Processing and Utilization of Forest Resources, Nanjing Forestry University, Nanjing 210037, China
c
Department of Mechanical Engineering, KU Leuven, Leuven 3000, Belgium

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

 Thermal annealing shrinks the strand FFF Thermal history Cross section

As-printed
length and expands its thickness with
a maximum linear strain of 20%.
 Thermal annealing reverses the
strain–hardening in printed parts and
leads to inferior performance against

Annealed
local tension defromation. Oven
 Crystallite develops after annealing
and locates at the strand-strand
interface. • Thickness swell 20%
• Width shrinkage 5%
• (Length shrinkage 10%)

a r t i c l e i n f o a b s t r a c t

Article history: This article presents the effects of thermal annealing at elevated temperatures (> glass transition temper-
Received 3 October 2022 ature T g ) on the performance of polymer parts via material extrusion 3D printing. Both semi-crystalline
Revised 31 January 2023 and amorphous filaments were used. As-printed parts were designed to be amorphous and then annealed
Accepted 31 January 2023
at 60, 110 and 150 °C for different durations ranging from 50 to 6400 s. The flexural strength and Young’s
Available online 3 February 2023
modulus increased by a maximum of approximately 10%. The increase was ascribed to crystallisation
development during annealing, as confirmed by thermal and morphology characterisations. Hence, this
Keywords:
effect was only observed with semi-crystalline materials. On the other hand, all the annealed parts
Annealing
Crystallinity
expanded in the thickness direction and shrank in the perpendicular plane. The maximum linear strain
Flexural properties reached 20%, while the volume strain was negligible. These morphology changes after annealing reversed
Material extrusion the strain-hardening of the strand and led to inferior strand performance against local tensile deforma-
Relaxation tion. The degradation can outweigh the benefits of crystallinity development. For amorphous parts, the
Strain-hardening degradation reached approximately 25% in both the flexural strength and the modulus.
Ó 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction

Material extrusion is one branch of additive manufacturing


(AM) techniques that selectively dispenses material through a noz-
zle or orifice to make 3D parts. The most common techniques
⇑ Corresponding authors at: No. 159 Longpan Road, Nanjing Forestry University, within this branch rely on physical processes, rather than chemical
Nanjing 210037, China (X. Wang). KU Leuven Campus De Nayer, Jan De Nayerlaan 5, [1] or biological [2] reactions, to achieve material consolidation.
Sint-Katelijne-Waver 2860, Belgium (J. Zhang).
The corresponding consumable materials are thermoplastic poly-
E-mail addresses: xzwang@njfu.edu.cn (X. Wang), jie.zhang2@kuleuven.be (J.
Zhang).
mers and their composites, in the form of either filaments or

https://doi.org/10.1016/j.matdes.2023.111687
0264-1275/Ó 2023 The Author(s). Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

pellets. Fused filament fabrication (FFF) is the most accessible tech- similar amorphous (poly(ethylene terephthalate glycol), PETG)
nique due to its ease of use, low cost, etc. filaments.
In FFF, a filament is fed into a hot-end at a material-specific pro-
cessing temperature well above the glass transition temperature
T g . The material becomes liquid-like and flows under the pressure 2. Material and methods
of the filament itself. The pressure transfer is enabled by a recircu-
lation area that seals the gap between the filament and hot-end 2.1. Material, machine and printing
barrel [3,4]. The polymer melt then flows through a nozzle and is
dispensed on the top of a previously deposited material layer (ex- A white filament of semi-crystalline PLA was used as received in
cept the first layer). During the deposition, the hot strand is typi- the printing. The filament was purchased from Beijing Tiertime
cally stretched by the dragging force of the nozzle while being Technology Co. Ltd. Flexural bars measuring 80104 mm3 were
laid down. Bond formation develops through a series of intimate printed with a UP BOX Plus FFF machine (Beijing Tiertime Technol-
contact, neck formation/wetting, diffusion and randomisation ogy Co. Ltd.). In each build cycle, five specimens were simultane-
steps [5]. The bond development occurs in a highly non- ously printed (Fig. 1). The printing time lasted approximately 2 h.
isothermal process, characterised by a transient cooling rate of The G-code was generated in Slic3rPE 1.42.0. Particularly, the
up to 100 °C/s upon deposition and periodic reheating because of default G-code M221 for the flowrate turning on the hardware side
hot neighbour depositions nearby [6,7]. was restored to 100%, i.e., the command M221 S100 was used.
Due to their additive building and fast-cooling nature, the Table 1 summarises the slicing and printing details. Parameters
strand-strand interfaces are considerably weaker than the strand not listed there took the default values from the slicer.
itself [8]. This drawback largely limits the applications of FFF
printed parts if they are load-bearing. Post-thermal annealing, or 2.2. Temperature simulation and thermal annealing
heat treatment, is commonly used to alleviate the weak bonds
and to improve the overall mechanical integrity. The role of Part cooling was simulated with the T4F3 model (version 1.02)
annealing has been widely investigated recently [9–20].1 In these [6], considering thermal conduction within the part, subjected to
studies, annealing temperatures at or above T g are dominant. Of par- convective and radiative boundary conditions on free surfaces
ticular interest are the cold crystallisation temperature T cc and melt- and the Dirichlet boundary condition on the bottom side. The heat
ing temperature T m for semi-crystalline polymers. The annealing generation due to crystallisation was ignored, assuming a low sig-
duration ranged from seconds to hundreds of hours, based on the
sintering time scale [9] or the crystallisation kinetics [10].
Thermal annealing takes effect by allowing polymer chains to
diffuse across the strand boundaries at elevated temperatures,
hence promoting the strand-strand interfacial performance. The
benefits can be revealed by interfacial tension or shear but may
be insensitive to other forms of deformation. For semi-crystalline
polymers, annealing-induced crystallisation is widely observed
and can be correlated to improved mechanical performance.
Nevertheless, some controversies also occurred. For example,
Jayanth et al. [13] attributed the mechanical performance improve-
ment partly to the porosity decrease. However, Basgul et al. [20]
argued that annealing cannot reduce porosity. In addition, many
researchers observed that printed parts expanded in the thickness Fig. 1. Flexural bars visualised from the G-code in the slicer. Left: five parts in a
direction and shrank in the perpendicular plane after annealing. single build cycle. Right: a demonstration of the G-code characteristics.
These linear strains were small [17,18] or more often neglected.
However, Zhang and Moon [9] reported a maximum strain of
12%. Furthermore, prior proof-of-concept experiments by the Table 1
authors showed that those strains are universal, repeatable and Slicing, printing and simulation details.
can be of higher magnitude. They could lead to a falsified impression Item Detail
of improvement if judging the load-bearing capacity rather than Filament diameter 1.75 (0:05) mm
the normalised strength. These doubts and controversies motivate Nominal nozzle diameter 0.4 mm
the current work, particularly when more and more 3D printed Strand width 442.9 lm
polymer parts are used in medical applications after thermal ster- Spacing [23] 400 lm
Layer thickness 200 lm
ilisation [21].
Vertical shell number 2
This paper investigates the effects of thermal annealing on the Raft 2 layers
physical, thermal and mechanical performance of FFF printed Top/bottom solid layer 0
parts. Printed poly(lactic acid) (PLA) parts were designed to be Infill rectilinear, 100%
amorphous and then annealed at T g ; T cc or (T m  15 °C) for differ- Raster angle 45
Travelling speed 40 mm/s
ent duration from 0.5 to 64 times the crystallisation half time (tak- Volume flow rate 11.5 cm3/h
ing the lower bound of 100 s [22]). The flexural mechanical Part cooling fan on, max
properties, crystallinity development and dimensional changes Nozzle temperature 210 °C
were investigated. The results were compared with parts printed Plate temperature 55 °C
Room temperature 20 (±4) °C
from similar semi-crystalline (poly(vinyl alcohol), PVA) and dis-
Density 1.25 g/cm3
Specific heat capacity 1.314 J/g/K
Thermal conductivity 0.195 W/m/K
Emissivity 0.78
1 Convection coefficient 60 W/m2/K
The annealing details, loading conditions, failure modes, etc., are summarised in
Mesh size optimised for stability
the author note after each reference entry.

2
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

nificance (to be verified). The assumption was supported by the fol- Specimen physical dimensions before and after annealing were
lowing strategies ensuring fast cooling of the printed parts: measured with a digital calliper (accuracy 0.01 mm). The length
was measured with a soft paper ruler (accuracy 0.5 mm) when
 batch printing enabling 5-fold longer inter-layer time [24,25] the part was severely deformed. The weight was measured with
 activating and maximising the part cooling fan for forced con- an analytical balance and then used to calculate the bulk density
vection [25,26] of each specimen. The filament density was measured by Archi-
 using a low layer thickness to facilitate conduction at a high medes method with ethanol of a known density. The bulk porosity
temperature gradient [7,25] was calculated as the relative difference between the two
densities.
The part cooling is characterised by the cooling at the geometric
centre of any single part, which has a primary intra-layer reheating 2.3. Mechanical properties test
time ds of 0.30 s (Fig. 1) and an inter-layer time Ds of 300 s. For
practical reasons, the temperature simulation was performed for As-printed and annealed parts were tested for their quasi-static
a single cuboid of 12 mm long, 6 mm wide and 3 mm high with flexural properties based on the standard ASTM D790-17. For parts
an output data size of 2.2 GB. The original dimension would having a nominal thickness of 4 mm, the span-to-thickness ratio
instead require 450 GB of data storage. The width (6 mm) was cho- was 16:1 (L ¼ 64 mm, D ¼ 4 mm, Fig. 3). The test was performed
sen based on the heat penetration depth concept [6] — a local ther- on an Electronic Universal Testing Machine (model UTM4304,
mal disturbance has a marginal influence on locations 3 mm Shenzhen SUNS Technology Stock Co., Ltd., China) with a 30 kN
distant in printing PLA. In the simulation, the polymer strand at load cell. The bottom side of all the parts faced downwards, bearing
the geometric centre of the 10th layer (z = 2 mm) has a primary tension during the test. The loading speed was 5 mm/min. The flex-
intra-layer time ds of 0.30 s and an inter-layer time Ds of 4.5 s. A ural strain was calculated based on the load cell displacement and
matching ds with a much shorter Ds will suffice to claim the fast measured specimen thickness. During the test, the software failed
cooling nature of the physical process if the cooling in the simula- to return consistent chord moduli. Hence, the Young’s modulus (of
tion is fast. Table 1 presents pertinent material and process param- deepest slope) was obtained afterwards by piece-wise curve fitting
eters. The simulation can be reproduced based on instructions in (details in Section S1 in the supporting online information). For
Mendeley Data [27]. each sample, 5 specimens were tested.
Printed PLA parts were then annealed in an electric drying oven
(model DHG-9146A, Shanghai Jing Hong Laboratory Instrument 2.4. Statistical analysis
Co., Ltd., China). Three treatment temperatures were used: 60,
110 and 150 °C. They are close to the glass transition temperature A two-sample t-test assuming equal variance was performed to
(T g ), peak temperature of cold crystallisation (T cc ) and melting compare any two means (Excel, Microsoft 365). If the statistic was
temperature (T m ) of PLA, respectively. Samples were directly greater than or equal to the two-tailed critical values, the null
placed on metal grids. Different annealing durations from 50 to hypothesis of no difference between the two groups would be
6400 s were chosen based on the crystallisation kinetics (Table 2). rejected. Corresponding results were labelled by one (*), two (**)
The samples were air-cooled on a wooden table after annealing. and three (***) stars, indicative of the probabilities p < 0.05, 0.01
Fig. 2 shows the holistic thermal history of printed parts. and 0.001, respectively.
Table 2
Annealing temperature, duration and sample label. 2.5. Crystallinity model
Temperature [°C] Duration [s] Label
The Avrami equation is commonly used to describe the isother-
60 ( T g ) 800,6400 T60-t800, etc.
110 ( T cc ) 50,100,200,400,800,1600,3200,6400 T110-t50, etc.
mal crystallisation kinetics [22,28]. The relative degree of crys-
150 ( T m ) 50,100,200,400,800,1600,3200,6400 T150-t50, etc. tallinity aðt Þ, defined as a ratio of the absolute crystallinity xðtÞ to
its maximum attainable value x1 , is expressed as
x  n
 aðtÞ ¼ 1  exp kt ; ð1Þ
x1
where t is the crystallisation time, and n is the Avrami index indica-
tive of the nucleation and crystal growth mechanism, typically from

Fig. 2. A holistic view of the thermal history in printed parts since deposition. This
diagram does not reflect the real cooling/heating profile/rate or the time duration. Fig. 3. A diagram of the flexural mechanical test and the failure mode in this work.
For demonstration purposes only. Drawing not to scale.

3
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

1 to 3 for athermic nucleation and 2 to 4 for thermic nucleation 2.6. Differential scanning calorimetry
[29]. The parameter k denotes the crystallisation rate. It has a unit
of time to the (n) power. Furthermore, Nakamura et al. [30] trea- As-printed and annealed parts were tested with differential
ted the non-isothermal process as an infinite sum of short iso- scanning calorimetry (DSC) to probe the crystallinity development.
thermal processes and derived For each sample, a specimen weighing 2–3 mg was carefully cut off
" Z n # from the end side of mechanically tested parts away from the frac-
t
1=n ture. Care was taken not to heat the specimen during the prepara-
aðtÞ ¼ 1  exp  k ds : ð2Þ
0 tions. Tests were performed on a DSC6100 machine (Netzsch,
Germany) in a nitrogen atmosphere. The specimen was ramped
In both equations, the rate constant k strongly depends on the tem- from 20 to 200 °C at a rate of 10 °C/min, followed by a 5-min
perature. The crystallisation half time t 1=2 contains the same infor- isothermal stage, then cooling down to 20 °C at 10 °C/min, after-
mation as k (given n) wards another 5-min isothermal stage, and lastly a second heating
cycle at the same rate.
1=n
t 1=2 ðTÞ ¼ ½ln 2=kðTÞ : ð3Þ The test results were analysed by Netzsch Proteus 8.0.2. The
degree of crystallinity, x, was determined as the ratio between
For PLA, t 1=2 usually attains its minimum at a temperature close to the net specific energy absorbed during cold crystallisation and
T cc and increases in both directions by one order of magnitude for melting (DHcc þ DHm ) to the melting enthalpy of pure PLA crystals
25 °C. t 1=2 should approach infinity at temperatures below T g or (DH*, 93.0 J/g [35]). The determination of DHcc or DHm strongly
above T m , indicating that crystallisation cannot occur under those depended on the baseline choice. Hence, the x determined from
conditions. Fig. 4 provides a survey of literature results the second heating cycle was considered a typical error in DSC,
[22,28,29,31–33]. given that the cooling stage showed no melt-crystallisation.
These data must be extrapolated to the whole temperature win-
dow ½T g ; T m to account for in-process crystallisation. For pure cal- 2.7. X-ray diffraction
culation purposes, the lower bounds of the literature data were
fitted to two models in Table 3, anticipating that the crystallinity Wide angle X-ray diffraction (XRD) was used to double-check the
of the as-printed part would be low (curve fitting details in Sec- crystallinity development. Samples measuring 10  10  4 [mm]
tion S3). Other data sources may report t1=2 magnitudes higher were cut from failed parts after the mechanical test, away from the
than these lower bounds [34]. Thus, the calculation was only used fracture or clamped area. Their bottom sides were diffracted upon
to estimate the upper bound for the crystallinity. X-rays of Cu-Ka1 radiation (k = 0.1541 nm) from an Ultima IV
multi-functional diffractometer (Rigaku, Japan) at 40 kV/30 mA.
The scanning angle 2h spanned from 5 to 90° at increments of
0.02°. The degree of crystallinity was determined as the ratio of the
sum of the peak area to the total area in the diffractogram [36].
Reproducible analyses were performed with Jade 6 (Release
6.5.26), with step-by-step procedures documented in Section S2.

2.8. Morphology investigations

The morphology of the printed parts, particularly of the frac-


tured surface in the mechanical tests, was investigated with an
optical microscope (Keyence VHX 6000, Belgium). The characteris-
tic material plane with an outward normal along the direction n1 =
(1,-1,0) (Fig. 5) was also investigated with mechanical polishing

Fig. 4. Crystallisation half-time t1=2 of PLA at different temperatures T.

Table 3
Numerical models of crystallisation half time t1=2 .

Model 1 Model 2
Model t1=2 ¼ a expðpTÞ t1=2 ¼   T g þT m
a b
cos p T 2
=ðT m T g þDT Þ
þ b expðqTÞ
Parameters a = 1.236106; a = 43.80, b = 43.05.
p = 0.1337,
b = 4.895106; Fig. 5. (a) A demonstration of the polished/sliced location. Drawing not to scale. (b)
q = 0.1097. Anticipated view of the polished/sliced cross-section along the direction n1 = (1,-
2 0.99 0.87
R 1,0). The circled numbers indicate the relative sequence of strand deposition. The
t1=2 (60°C) 405 min 1449 min up voids are more often filled than the down voids and help to identify the positive
t1=2 (165°C) 355 min 1449 min z direction. (c) Alkaline treatment on PLA slices and the chemical reaction
t1=2 (80°C) 28 min 33 min mechanism: a water molecule breaks an ester bond into a carboxylic acid and an
alcohol. The former reacts with hydroxide (OH) and forms a molecule of water. A
Note: T g = 60 °C, T m = 165 °C, DT = 2 °C were used in Model 2. Correspondingly, t1=2 long PLA molecular chain thus breaks into two parts. The existence of the base shifts
is infinity at (T g  1 °C) and (T m + 1 °C) and sufficiently large at T g and T m . the equilibrium toward the products.

4
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

and slicing. This material surface has an expected strand morphol-


ogy in different view angles as indicated in Fig. 5b. The even layers
show the cross-sectional view of the strands; the odd layers show
their side view.
Either side of the thin box in Fig. 5a was sequentially polished
away with silicon carbide papers of P80 and P800 FEPA grades
(Struers), followed by diamond suspensions of grain sizes of 6, 3
and 1 lm. The polished surfaces were then observed under micro-
scopes. Slices of the thin box (thickness 20 lm) were also pro-
duced with a microtome. They were observed under regular and
polarised optical microscopes (Olympus BX41 and BX51, Olympus
Optical Inc., Japan) to investigate crystallinity distributions within
the strand.
To this end, alkali treatment was also performed. Proof-of-
concept experiments showed that PLA reacts with sodium hydrox-
ide (NaOH) aqueous solution through ester hydrolysis (Fig. 5c).
This reaction is slow at room temperature but can be accelerated
by raising the molar concentration of NaOH(aq.) and particularly
the temperature to one above the T g . It is hypothesised this reac-
tion occurs at different rates in the amorphous and crystalline
regions due to different amounts of accessible ester groups. Hence,
it is possible to exploit the kinetic difference to dissolve the amor-
phous PLA region, leaving only the crystallised region.
During the investigations, PLA slices were submerged with
NaOH (aq., 6 mol/L) droplets on a glass slide2, which was placed
on top of a beaker immersed in a water bath at 80 °C (Fig. 5c). After
10 min, the slices were washed with distilled water and then
observed under the microscope. The crystallinity development dur-
ing the alkali treatment was assumed to be marginal since the crys-
tallisation half time at 80 °C was at least 28 min (Table 3).
Fig. 6. Simulated spatial and temporal temperature variations in printed parts. (a)
Temporal profiles at three representative locations. Two solid horizontal lines
3. Results indicate isothermal processes at 85 or 110 °C for 1 min. (b) Temperature
distribution on the 10-th layer (z ¼ 2 mm) at various times. t12 ¼ 54:0 s,
t13 ¼ 58:5 s, the moments when finishing the 12-th and 13-th layers, respectively.
3.1. Reference state of as-printed parts tN = 67.5 s, the total printing time. At t ¼ 1:3tN , the whole layer is colder than T g .

3.1.1. Temperature variations


Since the material deposition at the geometric centre
r2 ¼ ð6; 3; 2Þ [mm] (Fig. 6a), it rapidly cools from the nozzle tem- ally resemble those at r2 . The cumulative time above T g at all three
perature towards a steady state (between the room and plate tem- locations is less than 1 min (39.4, 44.8, 38.6 s for r1 , r2 and r3 ,
perature). After 0.30 s, the primary intra-layer reheating occurs (as respectively), with an average cooling rate of 200 °C/s. According
indicated by the red arrow). However, it only creates a tiny distur- to [39], the semi-crystalline nature of PLA will be depressed when
bance in the profile, as the strand at r2 is still rather hot (96 °C) the steady cooling rate exceeds 10 °C/min. Fig. 6b further demon-
before being thermally contacted. This phenomenon is typical at strates that these temperature histories are representative of the
short intra-layer times [37] or high Biot numbers [6]. After another bulk geometry at various moments: the 10-th layer became rather
0.30 s, a secondary intra-layer reheating occurs, but it does not cold only after a few layers of deposition time, and it only took 20 s
produce any more visible reheating disturbance/peak. The thermal for the whole layer to be frozen (TðrÞ < T g ; 8r) after the deposition.
effects of this and the following intra-layer reheatings are buried in Since the real printed part is bigger, the duration of temperature
the cooling, because the directional conductive heating flux in y- above T g will be lower than the simulations. The following crys-
direction cannot compensate for other forms of energy loss3. tallinity analysis will employ the worse scenario of the cooling pro-
The first major reheating peak appears after 4.5 s due to the file at r2 .
inter-layer reheating of hot material deposition. It raises the tem-
perature at r2 by approximately 1/5 of the temperature difference
3.1.2. Crystallinity estimation
ðT n  TÞ within 0.2 s. The cooling after this peak is relatively slow
With the isothermal kinetic model (Eq. 1), the in-process crys-
due to gradual accumulations of hot neighbours nearby. It is note-
tallinity is bounded above by a value that would develop over an
worthy that this reheating peak does not arise from a single ther-
isothermal process at 110 °C for 1 min (Fig. 6a). Using the lowest
mal contact as in thin-wall structures [6,38] but from the collective
t1=2 of 1.9 min in the calculation, Eq. 1 returns a maximum a of
efforts of all reheating phenomena within the inter-layer time. The
31% for all Avrami index n in the range of ½1; 4 (n = 1, a = 31%; n
following reheating peaks are lower in height but wider in width
= 4, a = 5.2%). More strictly, the crystallinity should be bounded
because they are caused by indirect thermal contacts.
above by that developed after an isothermal process at 85 °C for
Fig. 6a also presents temporal profiles at two boundary loca-
1 min. Taking the experimental data of t1=2 = 14.6 min, the maxi-
tions: r1 and r3 . Their cooling and reheating characteristics gener-
mum a was 4.6%, 8n 2 ½1; 4 .
2
With the non-isothermal kinetic model (Eq. 2), t1=2 is extrapo-
Glass slide is not recommended for future investigations, as it reacts with NaOH
(aq.).
lated to the whole crystallisation window ½T g ; T m and assigned
3
The strand spacing doubles the layer thickness, and the gradient component j @T 1449 min outside this range. The rate constant k is calculated using
@z j
doubles j @T
@y j. Both factors favour thermal conduction in the negative z-direction. Eq. 3. Then, Eq. 2 gives an upper bound
5
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

 Z n
1=n
dae ¼ maxn2½1;4 1  exp  k ds ¼ 0:78%; ð4Þ
t

based on t 1=2 model 1 and the temporal profile at r2 for


t 2 ½42:8; 102:3 s (Fig. 6). With model 2, the upper bound is
0.74%. These calculations are insensitive to t 1=2 choice outside
½T g ; T m if the integration is to be evaluated for a limited duration
below T g or above T m , i.e., dae < 0.8%, 8n 2 ½1; 4 and
t 1=2 >1449 min when T < T g or T > T m .
Regardless of the kinetic model, a relative crystallinity of less
than 5% suggests that the semi-crystalline nature of PLA is
depressed during printing. This amorphous state was verified with
DSC and XRD later.

3.2. Effects of thermal annealing

3.2.1. Mechanical properties


All printed PLA parts showed good dimensional quality. Their
bulk density was between 1.17 and 1.19 g/cm3, corresponding to
a bulk porosity of 4%–6%. Without annealing, these parts showed
an average flexural strength and Young’s modulus of 83.19 and
2966 MPa (Table 4, Fig. 7) in three-point bending. All as-printed
parts failed under cracks initiated from the tensile yielding on
the bottom layer. The crack then slowly propagated upwards.
However, none of the specimens broke apart.
Annealing printed PLA parts at 60 °C (T g ) within 6400 s showed
no significant effects on the dimensional quality, mechanical
response or failure mode. Fig. 7. Flexural properties of as-printed and annealed parts. (a) Strength. (b)
Young’s modulus. The error bars indicate standard deviations. The blue and red
Annealing printed parts at 110 °C (T cc ) induced slight deforma-
stars indicate the statistical significance of test results compared with the as-
tions (Fig. S2). The flexural properties started to deviate from as- printed parts and the maxima within the group, respectively.
printed parts when the duration exceeded 800 s (Fig. 7). The
annealed parts improved in the strength and modulus by a maxi-
mum of 9% and 13%, respectively. For sample T110-t6400, the
specimens also failed from cracks due to local tensile yielding from during the test. The modulus increased by 3% (p = 0.12, one tail)
the bottom side. However, they showed some levels of yield defor- compared with that of as-printed parts; the strength decreased
mations. Some specimens broke apart during the test. by 9% (p = 0.011, one tail). However, the force-time curves always
Annealing at 150 °C (T m  15 °C) did not significantly improve indicated improvement in the test, as shown in the video at
the strength. Instead, significant decreases were observed when https://www.bilibili.com/video/BV1FB4y1z7pR/ or
the duration exceeded 1600 s. Additionally, the Young’s modulus https://youtu.be/LdIvg1q1TPI. This is a falsified impression
only showed marginal improvement at intermediate duration due to the sample thickness expansion after annealing.
(400–3200 s). These promotional effects later vanished after longer
annealing times. All specimens of sample T150-t6400 broke apart
3.2.2. Between-lab reproducibility of the mechanical performance
Considering that numerous factors affect the part performance,
Table 4 the authors reproduced the mechanical experiments on as-printed,
Flexural properties and crystallinity of as-printed and annealed PLA parts.
T110-t6400 and T150-t6400 samples in a different lab, using two
Label Strength [MPa] Modulus [MPa] xDSC [%] xXRD [%] different PLA filaments (Prusa PLA-transparent, ColorFabb PLA-
As-printed 83.19(4.91) 2966(155) 2.6(1.2) 1.9 silver). All the printing, annealing and testing procedures remained
T60-t800 84.34(2.90) 3126(134) - 1.5 the same, as outlined in Section S7.
T60-t6400 84.64(4.31) 2957(67) 2.9(1.0) 1.3 As-printed PLA parts showed a bulk porosity of 5.5% (Table 5),
T110-t50 84.15(3.72) 2906(55) 4.2(2.4) 1.4 consistent with the previous campaign. After annealing, transpar-
T110-t100 82.23(1.64) 3061(80) 7.9(1.5) 1.5 ent PLA parts became less opaque, indicative of crystallinity devel-
T110-t200 84.89(3.45) 3091(158) 12.0(2.5) 1.3 opment. All samples showed apparent shrinkage in length, some
T110-t400 82.28(4.08) 2941(122) 16.6(1.8) 1.7
T110-t800 88.54(2.98) 3284(94) 31.8(2.7) 19.8
distortions (Fig. S2) and a slight weight reduction by 0.3%–1.0%
T110-t1600 89.81(2.20) 3338(60) - - (typical value 0.5%).
T110-t3200 89.24(1.90) 3325(56) 32.9(2.4) 22.2 Stress-strain curves of different Prusa PLA samples are shown in
T110-t6400 90.40(1.94) 3287(75) 33.7(2.3) 23.0 Fig. 8. As-printed parts showed yield deformations before failure
T150-t50 82.90(2.50) 2922(61) - 1.1 under multiple cracks. These cracks independently developed from
T150-t100 82.23(5.81) 2920(60) - 1.1 shear yield under tension on the bottom side (strand necking in
T150-t200 82.64(5.17) 3072(71) 15.4(2.2) 1.0
Fig. 9a) and propagated inwards. The propagation path could
T150-t400 86.83(2.00) 3398(135) 35.3(3.2) 22.7
T150-t800 85.29(2.81) 3319(179) 40.1(1.5) 21.7 locally follow the strand direction, indicating inferior local intra-
T150-t1600 79.26(4.17) 3223(77) - - layer fusions. On the bottom and lateral sides, multiple sites of
T150-t3200 77.32(1.92) 3250(80) - 25.4 crazing/stress whitening were visible due to the local stress con-
T150-t6400 76.04(2.85) 3059(58) 42.3(1.7) 27.2 centration on the gaps between the infill and perimeter (Fig. 1).
‘-’ not tested. Standard deviations between parentheses, except those of xDSC are On the fracture side, a large area of stress-whitening was observed
typical measurement errors. (Fig. 9a).
6
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

Table 5
Physical and flexural properties of as-printed and annealed parts.

Material Sample Bulk porosity eL [%] eW [%] eH [%] eV [%] Strength [MPa] Modulus [MPa]
Prusa As-printed 5.3%(0.1%) 0 0 0 0 90.66(0.60) 3171(28)
PLA T110-t6400 6.8%(0.6%) -6.7(0.4) -4.4(0.3) 13.1(1.0) 1.0(0.6) 98.76(1.60) 3316(27)
T150-t6400 7.1%(3.1%) -10.3(1.3) -5.2(0.8) 21.3(2.3) 3.1(1.3) 88.15(3.27) 3253(144)
ColorFabb As-printed 5.9%(0.8%) 0 0 0 0 82.21(1.45) 2957(94)
PLA T110-t6400 4.8%(0.4%) -4.6(0.2) -2.8(0.3) 6.8(0.5) -0.9(0.4) 91.27(1.01) 3243(36)
T150-t6400 4.5%(1.0%) -4.7(0.2) -3.4(0.4) 8.9(0.8) 0.3(0.9) 86.36(1.44) 3314(33)
PVA As-printed 7.1%(0.4%) 0 0 0 0 65.04(1.06) 3196(64)
T110-t6400 6.1%(0.4%) -0.9(0.1) -0.3(0.1) 1.2(0.1) 0.0(0.2) 74.51(1.60) 3445(53)
T150-t6400 8.1%(0.2%) -3.0(0.2) -1.6(0.1) 4.3(0.4) -0.5(0.3) 77.23(1.34) 3319(76)
PETG As-printed 5.8%(0.4%) 0 0 0 0 70.47(0.70) 2078(18)
T110-t6400 7.0%(0.4%) -8.9(0.8) -6.3(0.6) 18.4(1.9) 1.0(1.1) 60.18(3.84) 1780(58)
T150-t6400 * -7.9(0.4) * 13.1(2.4) * 55.08(2.62) 1492(91)

eL ; eW and eH indicate linear strains in the length, width and height directions after annealing, respectively; eV the volume strain, as calculated by
eV ¼ ð1 þ eL Þð1 þ eW Þð1 þ eH Þ  1. Negative values correspond to shrinkage.
*: measurement not performed due to too much morphological change (Fig. S2).

Fig. 9. Fracture morphology of as-printed and annealed PLA parts. (a1, b1) Bottom
view. (a2, b2) Fracture view. (c1, c2) Polished cross-sectional view. All the
specimens considerably expanded in thickness after annealing (a2 v.s.b2). The
expansion was also significant at the strand level and showed temperature
dependence (c1, c2). Some parts also showed geometric distortions as indicated
by the curvature in b1.

Fig. 8. (a) Force-time curves in the flexural tests; (b) stress-strain curves. During The results of a different PLA brand (ColorFabb, Table 5) gener-
the test, the loading force suggested a falsified impression of improved performance ally confirmed that thermal annealing was beneficial to flexural
after the annealing. However, the stress indicated otherwise.
performance. These stress-strain curves and morphology are pro-
vided in Section S8. However, negative effects of annealing at
150 °C were not observed with this filament. Even so, it is reason-
able to claim that T cc is a better annealing temperature since its
When annealed at 110 °C, both the flexural strength and promotional effect on the strength is more profound.
Young’s modulus significantly increased (Table 5). Most specimens
failed under a single crack without apparent yielding. The strains at 3.2.3. Thermal properties and degree of crystallinity
break considerably deceased. The brittleness resulted from Fig. 10 presents the DSC of different samples. The characteristic
increased free volume contents after annealing [40]. transitional temperatures and enthalpies can be found in Sec-
When annealed at 150 °C, the previous promotion effects van- tion S5. Notably, the cooling and second heating stages for all 13
ished. The flexural strength and Young’s modulus were comparable test samples showed an extremely high degree of consistency,
to those of as-printed parts. All the specimens experienced brittle indicative of consistent material compositions (e.g., no thermal
failures with little yielding. A single crack was also initiated from degradation during annealing) and reliable test results.
the specimen bottom side but led to catastrophe failure. Local The degree of crystallinity x of as-printed parts was estimated
strand necking was no longer observed, but multiple sites of craz- as marginal in Section 3.1.2. The DSC confirmed this reference
ing appeared, indicating limited yielding before failure. state: x was 2.6%, with a typical measurement error of 1.2%
7
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

Fig. 10. (a) DSC heat flow curves of as-printed and annealing PLA parts in the first heating stage. The results are vertically shifted for visual effects. (b1, b2) XRD curves of
annealed samples. The black arrows indicate the annealing time increasing from 50, 100, 200, 400, 800, 3200 to 6400 s.

Fig. 11. Optical images of as-printed (a), annealed (b, c), and alkaline-treated PLA slices (d, e). The first row of images was observed under a regular optical microscope; the
second row was the counterpart under a polarised microscope. Bright areas indicate the crystallites.

(Table 4). This confirmation justified the null volumetric heat gen- Annealing at 110 °C induced crystallisation. The crystallinity
eration assumption in the temperature modelling. increased from 4.2% to 31.8% when the duration rose from 50 to
Annealing at 60 °C for 6400 s made no difference in the crys- 800 s. When further increased to 3200 and 6400 s, only a limited
tallinity of printed parts. The x was 2.9%. The XRD analyses con- incremental crystallinity increase was observed from DSC. Anneal-
firmed the amorphous state with a wide and broad diffraction ing at 150 °C also induced crystallisation. The printed part devel-
peak and a negligible crystallinity of 1.3% (Fig. 10, Table 4). oped an x of 15.4% after annealing for 200 s. A longer annealing

8
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

duration further benefited the crystallisation development with a The alkaline treatment was designed to dissolve the amorphous
maximum x of 42.3% observed at 6400 s. XRD confirmed the same region. Fig. 11d reveals that the skeleton of the strand bottom side
trend of crystallinity development with the annealing time. How- survived the treatment, supporting the crystallinity distributions.
ever, it was not sensitive to intermediate crystallinity states in this In addition, long threads of PLA crystalline aggregates were
study. observed in Fig. 11e (high-resolution figures available in Sec-
Generally, the crystallinity obtained by XRD was lower than tion S11). They reached 1 mm in length and thus could only come
that by DSC. The discrepancy is ascribed to the difference between off from horizontal (odd) layers in the 20-lm-thick slice. These
the two methods (e.g. their interpretations of crystallinity being a observations show that the crystalline structure is one-
mass or a volume ratio) and factors such as the specimen location, dimensional, echoing the Avrami analysis.
the temperature gradients, etc. Additionally, the DSC analysis sug- Although the crystalline structures had preferential distribu-
gested that annealing at 110 or 150 °C might yield different crystal tions and orientations, they cannot yet explain the mechanical
structures, because the melting of sample T150-t6400 showed degradation unless there were large interfacial deformations. This
three peaks, but that of T110-t6400 showed only one. However, aspect is investigated with flexural tests at a lower span-to-
this speculation was not supported in XRD: for the exception of a thickness ratio.
0.22° shift in the 2h angle for the whole pattern, the two samples
behaved identically (Fig. 10-b2). 4.1.2. Interfacial shear deformation
Assuming the annealing was isothermal, an Avrami analysis During the three-point bending of a cuboid bar, the maximum
was performed (Section S6). The crystallisation half times t 1=2 at shear stress is expected on the neutral plane
110 and 150 °C were roughly 300 and 200 s. The corresponding
Avrami indexes n were 0.95 and 0.92, indicative of one- 3P
rzx ¼ ;
dimensional crystal growth and athermic nucleation. This indica- 4WD
tion may result from strand plastic deformation upon deposition.
where P denotes the load force, W the specimen width and D the
Thus, the molecular chains were highly orientated in printed parts.
thickness. In addition, the maximum tensile/compressive stress is
When crystallised, the crystal orientation coincided with the pri-
expected on the bottom/top side with a magnitude of
mary direction of deposited strands.
3PL
jrxx j ¼ ;
2WD2
4. Discussions
where L is the support span distance.
4.1. Explaining the decrease
rxx 2L
¼ ¼ 2  spanthickness ratio:
Previous investigations revealed that annealing at elevated
rzx D
temperatures benefited the part performance. The promotion j rrxxzx j was 32 during the test. Hence, only a limited shear stress/strain
effects had two aspects. On the one hand, the polymer chains
was present compared with the tensile stress/strain.
gained mobility at elevated temperatures above T g and diffused
However, when the span-thickness ratio decreases, shear stress
across the boundaries. The weak boundary healed into a bonded
gains momentum. At a span-thickness ratio of 5:1 with PLA parts,
bulk. This finding was supported by the neck formation in Fig. 9-
the promotional effects of annealing completely vanished. Corre-
a2, b2, in which the sharp edges of voids became rounded after
sponding results are presented in Fig. 12. The exact values can be
annealing. On the other hand, crystallisation occurred during
found in Section S9, along with test details.
annealing. The crystallites benefited the stress transfer and defor-
Similar to the regular span tests, as-printed parts showed some
mation resistance due to their strong and rigid nature. However,
degree of yield and multiple independent crack propagations. The
it was also observed that an excessively high annealing tempera-
annealing rendered the parts brittle. Only adverse effects were
ture or long duration was detrimental to mechanical performance.
observed. However, none of the specimens failed under inter-
Why was that the case?
layer shear. These short-span tests confirmed previous observa-
tions that thermal annealing did have strong adverse effects on
the mechanical performance of printed parts.
4.1.1. Crystallinity distribution
Polymer crystallisation not only competes with diffusion bond-
ing but also compromises the bond quality if it occurs on the inter- 4.1.3. Strand length shrinkage and thickness expansion
face [10]. Morphological observations with PLA slices in Fig. 11 Adverse effects of annealing were primarily observed at higher
reveal the distributions of crystallinity within the strand. temperatures or longer duration. These annealing conditions were
As-printed PLA parts were predicted and verified to be amor- closely related to the dimensional changes (Fig. 9, Fig. S2). The part
phous. Accordingly, the polarised optical image did not reveal length evidently shrank after annealing. A less obvious but more
any crystallinity. significant change was the thickness expansion. Careful dimension
When annealed at either 110 or 150 °C for 6400 s, considerable measurements were taken before and after annealing. Table 5 pre-
degrees of crystallinity developed. They appear bright in the sents the linear and volumetric strains.
polarised images (Fig. 11-b2,c2). Recalling the strand orientations At 110 °C for 6400 s, the annealed parts shrank in length by 5%–
in Fig. 5b, the crystallised region is concentrated at the bottom side 7%, in width by 3%–4%, but expanded in thickness direction by 7%–
of each layer. Additionally, it has a preferential orientation along 13%. Their volume was conserved if applying the three-sigma crite-
the strand. During printing, the extrudate experiences elongational rion to the volume strain (Table 5). The linear strains further devel-
stretches during the deposition, which profoundly aligns the poly- oped at a higher annealing temperature of 150 °C. For Prusa PLA,
mer chains. It is easier for these highly stretched chains to trans- the length shrank by 10.3% and width by 5.2%, but the thickness
form into highly ordered crystal structures. From the expanded by 21.3%. This thickness expansion is indeed responsible
deformation analysis by McIlroy and Olmsted [41], the strand for the falsified impression of the flexural property increase (Fig. 8),
experiences more stretches on the bottom side; thus, the crys- because both the stress and modulus strongly depend on it (stress
talline region is located thereby. rf / D2 , modulus Ef / D3 ).
9
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

Fig. 13. (a) Demonstration of a hot cylinder element of length dl1 at the nozzle
outlet deforming into a strand of length dl2 during the deposition. Figure reproduced
based on the velocity field in [41], with permission from The Society of Rheology.
(b) A thermography view of extruding 200 mm filament at a linear filament feed
rate of 80 mm/min (G1 E200 F80). (c) Printing at a nozzle travelling speed of
40 mm/s (F2400). This infra-red image does not correspond to the actual printing
process. For demonstration purposes only.

Fig. 12. Flexural tests at a span-thickness ratio of 5:1 and a nominal strain rate of 4.1.4. Other influential factors
103 s1. (a) Stress-strain curves. (b) Strength. (c) Young’s modulus. The red stars
Additionally, the strain rate was also observed to be able to
indicate the statistical significance of test results compared with the as-printed
parts.
affect the flexural performance steadily. A higher strain rate in
the range of 5  105–1.6  102 s1 gave rise to an increase in
the flexural strength from 67 to 114 MPa (Fig. S7). The same trend
It is proposed here that all deposited strands in FFF are highly was observed in tensile tests of multiple polymers [42]. However,
stretched/drawn since the targeted strand cross-sectional area is annealing-induced thickness expansion decreased the strain rate
typically smaller than the nozzle capillary. When a hot volume ele- by a maximum of 20%, leading to a maximum of 1.6 MPa degrada-
tion in the strength. Hence, the strain rate was not the dominating
ment dV of length dl1 and cross-sectional area A1 at the nozzle exit
factor in the mechanical performance degradation. However, it is
deforms into a deposited strand of length dl2 and cross-sectional
noteworthy to keep this rate consistent in different datasets to
area A2 (< A1 ), the strand is stretched by the dragging force of
make comparisons.
the nozzle at a ratio of dl2 =dl1 (as demonstrated by the springs M
and W in Fig. 13a). Since the volume is conserved (neglecting den-
sity changes), this stretch ratio of plastic deformation could be 4.2. Is PLA special?
quantified by using the velocity ratio v 2 =v 0 . v 2 is the nozzle trav-
elling speed relative to the build plate. v 0 is the extrudate speed in Does annealing have identical effects on printed parts from
the nozzle capillary, and it can be calculated based on the targeted other filaments as well? Or is PLA a special material in FFF? This
volume flow rate and the nominal nozzle diameter (Fig. 13). Alter- section investigates the effects of annealing on similar and dissim-
natively, a (more) natural state after the die-swell could be used as ilar filaments: semi-crystalline poly(vinyl alcohol) (PVA) and
the reference, thus defining the stretch ratio as v 2 =v 1 . v 1 is the amorphous poly(ethylene terephthalate glycol) (PETG). Details on
extrudate moving speed after die-swell. It can be measured based this material and printing can be found in Section S7.
on the die-swell ratio or the extrudate length (within a given time). Due to similar plate and nozzle temperatures, annealing tem-
The extrudate cooling is considerably slower than that of the peratures of 110 and 150 °C were selected for both materials.
printed parts (Fig. S8); thus, the polymer chains are considered The duration was 6400 s. The experimental details were consistent
to have a more natural status. with those of previous experiments. For PETG sample T150-t6400,
In this study, the stretch ratio of the plastic deformation was the specimen width used undeformed values in the stress calcula-
163% (v 2 =v 0 ) or 276% (v 2 =v 1 ). Either number suggested that the tion. Fig. 14 presents the mechanical properties. The exact values
strand was highly stretched during the deposition. Similarly, the are available in Table 5. Stress-strain curves are provided in Fig. S4.
polymer chains in the layer thickness direction were highly com- Similarly, annealing cannot significantly reduce the bulk poros-
pressed. These deformations were frozen as a consequence of rapid ity (Table 5) without external pressure [20]. However, it can
cooling. Hence, they relaxed back into a (more) natural state when deform and relocate the voids, as observed in the PETG samples
annealed, most notably via strand thickness expansion and length annealed at 150 °C in Fig. S2.
shrinkage. When the annealing temperature was higher, this relax- For PVA parts, annealing at either temperature promoted their
ation mode was more thorough. Hence, the magnitude of the linear mechanical performance. This change can, likewise, be ascribed
strains increased. to increased crystallinity. Despite that increase, the statistical sig-
According to this narrative, as-printed parts consist of highly nificance level indicated that a higher temperature was not bene-
stretched strands, comparable to the status of strain-hardened/ ficial for further increase in the Young’s modulus. By contrast,
drawn polymer parts. Thermal annealing relaxes polymer chains only adverse effects were observed in annealed amorphous PETG
and reverses the strain-hardening process, thus leading to inferior parts. The higher the annealing temperature was, the further the
mechanical performance against (local) tensile deformations. mechanical performance degraded. The maximum degradation in
10
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

Acknowledgements

This work was supported by The Natural Science Foundation of


the Jiangsu Higher Education Institutions of China [22KJA430012]
and the Qing Lan Project (2021).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in


the online version, at https://doi.org/10.1016/j.matdes.2023.
Fig. 14. Flexural mechanical properties of as-printed and annealed PVA and PETG
parts. (a) Strength, (b) Young’s modulus. The stars indicate the statistical signif- 111687.
icance of the test results compared with the as-printed parts within each material
group.
References

[1] S.K. Romberg, A.I. Abir, C.J. Hershey, V. Kunc, B.G. Compton, Structural stability
of thin overhanging walls during material extrusion additive manufacturing of
strength reached 22%, and that of the modulus reached 28%. thermoset-based ink, Add. Manuf. 53 (2022) 102677, https://doi.org/10.1016/j.
However, linear strains of annealed PVA parts were lower than in addma.2022.102677.
the previous case of PLA (Table 5), due possibly to the material’s [2] S. Barnes, L. Kirssin, E. Needham, E. Baharlou, D. Carr, J. Ma, 3D printing of
ecologically active soil structures, Add. Manuf. 52 (2022) 102670, https://doi.
chemical nature. Hence, the adverse effects were marginal. org/10.1016/j.addma.2022.102670.
But for annealed PETG, the same strain magnitude as that of [3] M.P. Serdeczny, R. Comminal, M.T. Mollah, D.B. Pedersen, J. Spangenberg,
annealed PLA was observed, and the degradation was even more Numerical modeling of the polymer flow through the hot-end in filament-
based material extrusion additive manufacturing, Add. Manuf. 36 (2020)
significant. 101454, https://doi.org/10.1016/j.addma.2020.101454.
[4] J. Zhang, E. Vasiliauskaite, A. De Kuyper, C. De Schryver, F. Vogeler, F.
Desplentere, E. Ferraris, Temperature analyses in fused filament fabrication:
5. Conclusions from filament entering the hot-end to the printed parts, 3D Print. Add. Manuf.
9 (2) (2022) 132–142, https://doi.org/10.1089/3dp.2020.0339.
Polymer strands are highly stretched in material extrusion 3D [5] C. Bellehumeur, L. Li, Q. Sun, P. Gu, Modeling of bond formation between
polymer filaments in the fused deposition modeling process, J. Manuf. Process.
printed parts due to the extensional dragging upon deposition. 6 (2) (2004) 170–178, https://doi.org/10.1016/S1526-6125(04)70071-7.
Thermal annealing at elevated temperatures (> T g ) relaxes the [6] J. Zhang, B. Van Hooreweder, E. Ferraris, T4F3: temperature for fused filament
polymer chains and reverses the strain-hardening effects, leading fabrication, Prog. Add. Manuf. 7 (2022) 971–991, https://doi.org/10.1007/
s40964-022-00271-0.
to shrunken length, expanded thickness and inferior strand perfor- [7] J. Zhang, B. Van Hooreweder, E. Ferraris, Fused filament fabrication on the
mance again local tensile deformation. The maximum linear strain moon, JOM 74 (2022) 1111–1119, https://doi.org/10.1007/s11837-021-
reached 20%, while the volume strain was negligible in this study. 05031-z.
[8] T. Yao, K. Zhang, Z. Deng, J. Ye, A novel generalized stress invariant-based
The mechanical degradation reached 25% in strength and modu- strength model for inter-layer failure of FFF 3D printing PLA material, Mater.
lus for amorphous parts. However, promotional effects were still Des. 193 (2020) 108799, https://doi.org/10.1016/j.matdes.2020.108799.
observed with semi-crystalline polymers when crystallisation [9] Y. Zhang, S.K. Moon, The effect of annealing on additive manufactured ULTEM
9085 mechanical properties, Materials 14 (11) (2021) 2907, https://doi.org/
developed after annealing. For printed poly(lactic acid) (PLA) parts, 10.3390/ma14112907. Author note: ULTEM 9085 parts were annealed at 170,
annealing at the peak temperature of cold crystallisation T cc (110 180°C (slightly below and above the Tg) for 24 or 96 h. Annealed parts
°C) was exemplary. The crystallinity increased from null to 34% expanded in the layering direction and shrank in other directions. Most linear
strain < 5%; the maximum was 12%. The tensile strength of the bond line
after 6400 s, leading to 10% increase in flexural strength and
slightly increased; tensile strength along the strand direction slightly
Young’s modulus. The crystalline structures had preferential distri- decreased..
butions at the strand bottom side and orientations along the depo- [10] D.W. Collinson, N. von Windheim, K. Gall, L.C. Brinson, Direct evidence of
interfacial crystallization preventing weld formation during fused filament
sition path. However, a higher annealing temperature or longer
fabrication of poly(ether ether ketone), Add. Manuf. 51 (2022), https://doi.org/
duration benefiting the relaxation tends to erase these promotional 10.1016/j.addma.2022.102604. Author note: PEEK parts were annealed at 148
effects. and 180°C for 6 or 12h, through one or two steps. The heating/cooling rate was
Future work will evaluate the role of thermal annealing in the 1°C/min. Weld line tear and interfacial tensile test. Interfacial failures. In-
process crystallisation can occur on the strand surface if the polymer has fast
interfacial bond quality via pure shear, with possible physical crystallisation kinetics and it competes with inter-layer diffusion. Two-step
constraints in the strand thickness or width directions. In annealing (below, then above Tcc) produced stronger parts. 102604.
addition, the strain rates during annealing will be studied in [11] B. Akhoundi, M. Nabipour, F. Hajami, D. Shakoori, An experimental study of
nozzle temperature and heat treatment (annealing) effects on mechanical
comparison with the crystallisation kinetics for semi-crystalline properties of high-temperature polylactic acid in fused deposition modeling,
polymers. Polym. Eng. Sci. 60 (2020) 979–987, https://doi.org/10.1002/pen.25353.
Author note: PLA parts were annealed at 110°C for 1h. The tensile loading
was aligned with the strand direction. Annealing benefited tensile modulus
6. Data availability statement and strength, resulting from the crystallinity development.
[12] W. Jo, O.C. Kwon, M.W. Moon, Investigation of influence of heat treatment on
mechanical strength of FDM printed 3D objects, Rapid Prototyp. J. 24 (3)
Raw data of DSC, XRD, thermal simulation and modulus deter- (2018) 637–644, https://doi.org/10.1108/RPJ-06-2017-0131. Author note: PLA
mination method are openly available in Mendeley Data [27]. parts were annealed at 160°C for 0, 30, 60, 120 s with external pressure in a
mould. Failure mode: inter-layer debonding. A higher annealing duration and
Other processed or raw data are available from the authors upon
external pressure promoted tensile properties by reducing the part porosity..
a reasonable request. [13] N. Jayanth, K. Jaswanthraj, S. Sandeep, N.H. Mallaya, S.R. Siddharth, Effect of
heat treatment on mechanical properties of 3D printed PLA, J. Mech. Behav.
Biomed. Mater. 123 (2021), https://doi.org/10.1016/j.jmbbm.2021.104764.
Declaration of Competing Interest Author note: PLA parts were annealed at 90, 100, 120°C for 1, 2, 4 h,
followed by slow cooling in the oven to room temperature. Complexed failure
The authors declare that they have no known competing finan- mode in tension observed. Annealing improved the tensile strength and heat
deflection capability. The improvement was ascribed to porosity reduction and
cial interests or personal relationships that could have appeared stress relaxation. Annealing temperature at 100°C outperformed 120°C.
to influence the work reported in this paper. 104764.

11
W. Yu, X. Wang, X. Yin et al. Materials & Design 226 (2023) 111687

[14] A. Szust, G. Adamski, Using thermal annealing and salt remelting to increase [24] M. Faes, E. Ferraris, D. Moens, Influence of inter-layer cooling time on the
tensile properties of 3D FDM prints, Eng. Fail. Anal. 132 (2022) 105932, quasi-static properties of ABS components produced via fused deposition
https://doi.org/10.1016/j.engfailanal.2021.105932. Author note: PLA parts modelling, Procedia CIRP, vol. 42, Elsevier B.V, 2016, pp. 748–753, https://doi.
were annealed at 60, 80°C for 1 h. Failure mode: inter-layer debonding. org/10.1016/j.procir.2016.02.313.
Tensile strength increased only after annealing at 60°C. Tensile modulus [25] J. Zhang, J. Neeckx, J. Troukens, E. Ferraris, A reheating temperature criterion
decreased. Sample deformed during annealing. for adaptive strategy in fused filament fabrication, CIRP Ann. 71 (1) (2022)
[15] K.R. Hart, R.M. Dunn, E.D. Wetzel, Increased fracture toughness of additively 197–200, https://doi.org/10.1016/j.cirp.2022.03.046.
manufactured semi-crystalline thermoplastics via thermal annealing, Polymer [26] S.F. Costa, F.M. Duarte, J.A. Covas, Thermal conditions affecting heat transfer in
211 (2020) 123091, https://doi.org/10.1016/j.polymer.2020.123091. Author FDM/FFE: a contribution towards the numerical modelling of the process: this
note: PLA parts were annealed at 160°C for 2 h in an aluminium fixture. Single paper investigates convection, conduction and radiation phenomena in the
edge notched bending test performed. Failure mode: interlaminar fracture. filament deposition process, Virt. Phys. Prototyp. 10 (1) (2015) 35–46, https://
Annealing improved the fracture roughness through increased interfacial doi.org/10.1080/17452759.2014.984042.
wetting only if the printed part was amorphous before the annealing. [27] W. Yu, X. Wang, X. Yin, E. Ferraris, J. Zhang, Data for: The effects of thermal
[16] J. Dong, C. Mei, J. Han, S. Lee, Q. Wu, 3D printed poly(lactic acid) composites annealing on the performance of material extrusion 3D printed polymer parts,
with grafted cellulose nanofibers: effect of nanofiber and post-fabrication Mendeley Data V4 (2022). https://doi.org/10.17632/tmt5x98p8h.4.
annealing treatment on composite flexural properties, Add. Manuf. 28 (2019) [28] Y. Xu, Y. Wang, T. Xu, J. Zhang, C. Liu, C. Shen, Crystallization kinetics and
621–628, https://doi.org/10.1016/j.addma.2019.06.004. Author note: PLA and morphology of partially melted poly(lactic acid), Polym. Test. 37 (2014) 179–
PLA based composite parts were annealed at 120°C for 12 h. Both static and 185, https://doi.org/10.1016/j.polymertesting.2014.05.012.
dynamic flexural mechanical properties increased as a result of increased [29] A. Balazs, Crystallization kinetics of commercial PLA filament, Commun. – Sci.
crystallinity.. Lett. Univ. Zilina 19 (4) (2017) 15–19, https://doi.org/10.26552/com.
[17] J. Butt, R. Bhaskar, Investigating the effects of annealing on the mechanical c.2017.4.15-19.
properties of FFF-printed thermoplastics, J. Manuf. Mater. Process. 4 (2) (2020) [30] K. Nakamura, K. Katayama, T. Amano, Some aspects of nonisothermal
1–20, https://doi.org/10.3390/jmmp4020038. Author note: PLA and ABS crystallization of polymers. II. Consideration of the isokinetic condition, J.
printed parts were annealed at temperatures 10-30°C higher than their Appl. Polym. Sci. 17 (4) (1973) 1031–1041, https://doi.org/10.1002/
respective T g . The duration was 1 h + 2 h slow cooling inside oven. Complexed app.1973.070170404.
failure mode in tension observed. Annealed parts expanded in z-direction [31] M. Safandowska, A. Rozanski, Ring-banded spherulites in polylactide and its
(strand thickness) but shrank in x-y plane. The load-bearing capacity of blends, Polym. Test. 100 (2021) 107230, https://doi.org/10.1016/j.
annealed parts increased. The increase for ABS was marginal.. polymertesting.2021.107230.
[18] R.M. Dunn, K.R. Hart, E.D. Wetzel, Improving fracture strength of fused [32] H. Xiao, P. Li, X. Ren, T. Jiang, J.-T. Yeh, Isothermal crystallization kinetics and
filament fabrication parts via thermal annealing in a printed support shell, crystal structure of poly(lactic acid): effect of triphenyl phosphate and talc, J.
Prog. Add. Manuf. 4 (3) (2019) 233–243, https://doi.org/10.1007/s40964- Appl. Polym. Sci. 118 (2010) 3558–3569, https://doi.org/10.1002/app.32728.
019-00081-x. Author note: The printed parts had an ABS core with PC [33] H. Fang, Q. Xie, H. Wei, P. Xu, Y. Ding, Physical gelation and macromolecular
shells. The annealing was at 105°C (between T g of the two polymers) for mobility of sustainable polylactide during isothermal crystallization, J. Polym.
168 h. Single edge notched bending test performed. Failure mode: Sci., Part B: Polym. Phys. 55 (16) (2017) 1235–1244, https://doi.org/10.1002/
interlaminar fracture. Limited dimension change compared with un- polb.24381.
shelled parts (volume strain <1.3%, linear strain <2%). Interlaminar [34] S. Liparoti, D. Sofia, A. Romano, F. Marra, R. Pantani, Fused filament deposition
fracture toughness increased by more than 18 owning to the fracture- of PLA: the role of interlayer adhesion in the mechanical performances,
to-ductile transition of ABS. Polymers 13 (3) (2021) 399, https://doi.org/10.3390/polym13030399.
[19] I. Ferreira, C. Melo, R. Neto, M. Machado, J.L. Alves, S. Mould, Study of the [35] E.W. Fischer, H.J. Sterzel, G. Wegner, Investigation of the structure of solution
annealing influence on the mechanical performance of PA12 and PA12 fibre grown crystals of lactide copolymers by means of chemical reactions, Kolloid-
reinforced FFF printed specimens, Rapid Prototyping Journal 26 (10) (2020) Z. Z. Polym. 251 (1973) 980–990, https://doi.org/10.1007/BF01498927.
1761–1770, https://doi.org/10.1108/RPJ-10-2019-0278. Author note: Printed [36] Y. Srithep, P. Nealey, L.-S. Turng, Effects of annealing time and temperature on
parts of PA12 and PA12-carbon fibre annealed at 135, 150, 165°C for 3, 6, 12, the crystallinity and heat resistance behavior of injection-molded poly(lactic
18 h. Heating rate 7°C/min; cooling rate 1°C/min. Complexed failure mode in acid), Polym. Eng. Sci. 53 (3) (2013) 580–588, https://doi.org/10.1002/
tension and bending observed. Both tensile and flexural properties pen.23304.
significantly improved after annealing at 135 and 150°C, accompanied by [37] E. Ferraris, J. Zhang, B. Van Hooreweder, Thermography based in-process
crystallinity increase. Mechanical property decreased when annealed at 165°C monitoring of Fused Filament Fabrication of polymeric parts, CIRP Ann. 68 (1)
due to PA oxidation.. (2019) 213–216, https://doi.org/10.1016/j.cirp.2019.04.123.
[20] C. Basgul, T. Yu, D.W. MacDonald, R. Siskey, M. Marcolongo, S.M. Kurtz, Does [38] H.R. Vanaei, K. Raissi, M. Deligant, M. Shirinbayan, J. Fitoussi, S. Khelladi, A.
annealing improve the interlayer adhesion and structural integrity of FFF 3D Tcharkhtchi, Toward the understanding of temperature effect on bonding
printed PEEK lumbar spinal cages?, J Mech. Behav. Biomed. Mater. 102 (2020), strength, dimensions and geometry of 3D-printed parts, J. Mater. Sci. 55 (29)
https://doi.org/10.1016/j.jmbbm.2019.103455. Author note: PEEK parts were (2020) 14677–14689, https://doi.org/10.1007/s10853-020-05057-9.
annealed at 200, 300°C for 4h, with controlled heating/cooling rates of 12°C/h [39] W. Yu, X. Wang, E. Ferraris, J. Zhang, Melt crystallization of PLA/Talc in fused
above Tg. Parts loaded in compression, compression-shear and torsion. Failure filament fabrication, Mater. Des. 182 (2019) 108013, https://doi.org/10.1016/
mode: inter-layer debonding. Annealing did not markedly improve j.matdes.2019.108013.
mechanical properties, and it could not decrease the part porosity. 103455. [40] L. Malekmotiei, G.Z. Voyiadjis, A. Samadi-Dooki, F. Lu, J. Zhou, Effect of
[21] R. Told, Z. Ujfalusi, A. Pentek, M. Kerenyi, A. Vizi, P. Szabo, S. Melegh, J. Bovari- annealing temperature on interrelation between the microstructural evolution
biri, E. Judit, P. Maroti, A state-of-the-art guide to the sterilization of and plastic deformation in polymers, J. Polym. Sci., Part B: Polym. Phys. 55 (17)
thermoplastic polymers and resin materials used in the additive (2017) 1286–1297, https://doi.org/10.1002/polb.24379.
manufacturing of medical devices, Mater. Des. (2022) 111119, https://doi. [41] C. McIlroy, P.D. Olmsted, Deformation of an amorphous polymer during the
org/10.1016/j.matdes.2022.111119. fused-filament-fabrication method for additive manufacturing, J. Rheol. 61 (2)
[22] S. Saeidlou, M.A. Huneault, H. Li, C.B. Park, Poly(lactic acid) crystallization, (2017) 379–397, https://doi.org/10.1122/1.4976839.
Prog. Polym. Sci. 37 (12) (2012) 1657–1677, https://doi.org/10.1016/j. [42] N. Vidakis, M. Petousis, E. Velidakis, M. Liebscher, V. Mechtcherine, L. Tzounis,
progpolymsci.2012.07.005. On the strain rate sensitivity of fused filament fabrication (FFF) processed PLA,
[23] G. Hodgson, A. Ranellucci, M. Jeff, Slic3r Manual - Flow math, https://manual. ABS, PETG, PA6, and PP thermoplastic polymers, Polymers 12 (12) (2020) 1–15,
slic3r.org/advanced/flow-math Access date: 2022-09-18. https://doi.org/10.3390/polym12122924.

12

You might also like