Polymer 81 (2015) 1e11

Contents lists available at ScienceDirect

Polymer
journal homepage: www.elsevier.com/locate/polymer

Linking the morphology of a high hard segment content polyurethane

to its thermal behaviour and mechanical properties
Y. Swolfs a, *, E. Bertels b, I. Verpoest a, c, B. Goderis b, c
a
Department of Materials Engineering, KU Leuven, Kasteelpark Arenberg 44 bus 2450, 3001, Leuven, Belgium
b
Department of Chemistry, KU Leuven, Celestijnenlaan 200F, 3001, Leuven, Belgium
c
Leuven Material Research Centre (Leuven-MRC), KU Leuven, Belgium

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

Article history: Understanding and controlling the morphology of thermoplastic polyurethane (TPU) is crucial, as it is
Received 17 July 2015 closely linked with its thermal and mechanical properties. The morphology of a TPU with a high hard
Received in revised form segment content was investigated. When hard segment crystallisation was avoided by fast cooling, a
1 November 2015
crystallisation-induced phase separation occurred upon reheating. At higher temperatures, a second
Accepted 2 November 2015
Available online 6 November 2015
polymorph was additionally created. Cooling slowly from the melt directly induced the formation of both
polymorphic forms. The complex thermal behaviour could hence be explained by the (cold) crystal-
lisation and melting of two polymorphs. The disordered two-phase nanomorphology, revealed by AFM at
Keywords:
Morphological modelling
room temperature after cooling slowly, was validated for both fast and slowly cooled samples at higher
Thermoplastic polyurethane temperatures by fitting model SAXS patterns to time resolved synchrotron SAXS data. Annealing fast
Storage modulus cooled samples at high temperature induced some ordered, lamellar stacks in addition. Finally, the
morphology was linked to the evolution of storage modulus with increasing temperature.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction the TPU reinforcement increases when also the SS phases crystal-
lise [7].
Thermoplastic polyurethanes (TPU) are widely used in indus- The TPU nanomorphology is commonly visualised by Atomic
trial applications, such as foam and sports equipment, and have Force Microscopy (AFM) [8e10] or Transmission Electron Micro-
attracted considerable research interest. TPUs are linear segmented scopy (TEM) [2,5]. However, as these techniques are rarely opera-
copolymers, consisting of alternating soft and hard segments. The tional at elevated temperatures, (synchrotron) Small Angle X-ray
hard segments (HS) tend to separate from the soft segments (SS) by Scattering (SAXS) is the often the method of choice for in-
their crystallisation or do so in the liquid state due to a thermo- vestigations of the nano-morphology as a function of temperature.
dynamic incompatibility with the soft segments. Conversely, the interpretation of SAXS data relies on the use of a
Phase-separated aromatic diphenylmethane diisocyanate (MDI) suitable morphological model, which is nine times out of ten based
TPUs often adopt a lamellar nano-morphology even though several on microscopic observations. SAXS and microscopy are thus highly
authors reported different morphologies. Spherical or cylindrical complementary. The thermal behaviour of TPUs is commonly
morphologies [1] have been reported in MDI extended with studied by Differential Scanning Calorimetry (DSC), which also
ethylene diamine, while a morphology without a specific shape was yields information on the TPU morphology, albeit indirectly. Re-
reported for MDI extended with 2-methyl-1,3-propane diol [2] or lations between thermal behaviour and morphology are often
with butane diol [3]. The TPU morphology affects the mechanical studied by DSC in conjunction with (synchrotron) Wide Angle X-ray
properties [4,5]. The hard domains act as physical cross-links and Diffraction (WAXD) and SAXS [2,11e16]. The size and position of the
reinforce the TPU in a similar fashion to nano-fillers. The rein- thermal transitions varies with thermal history [12,13], mechanical
forcement efficiency is increased when the HS fraction increases deformation [14], composition ratio [15] and soft segment length
and the hard domains are more interconnected [6]. Alternatively, [16].
An intriguing thermal feature of MDI TPUs is the multiple
melting endotherms. Early literature [17,18] attributed these en-
* Corresponding author. dotherms to different types of hydrogen bonds, where the urethane
E-mail address: yentl.swolfs@mtm.kuleuven.be (Y. Swolfs).

http://dx.doi.org/10.1016/j.polymer.2015.11.007
0032-3861/© 2015 Elsevier Ltd. All rights reserved.
2 Y. Swolfs et al. / Polymer 81 (2015) 1e11

NH bond is the donor, and either the urethane carbonyl or the Some of these samples were annealed for 72 h in an oven, either at
macrodiol is acting as acceptor. This hypothesis was later rejected 100
C or 150
C. The samples were then removed from the oven
by Samuels & Wilkes [19], who investigated a TPU without and cooled to room temperature on a steel surface. The same
hydrogen bonding and proved that the explanation of hydrogen production procedure was also performed at a constant cooling rate
bonding is insufficient. Therefore, alternative explanation were of 0.5
C/min. For this material, no extra annealing steps were
coined, based on various levels of packing order [11,20,21]. These applied. This sample is called the slowly cooled sample.
authors suggested that the low temperature endothermic peak was
associated with disordering of the short range order and the high
temperature one to the disordering of the long range order. The 2.2. Differential scanning calorimetry
final peak was attributed to the melting of the microcrystalline
order. Apart from the multiple melting endotherms, TPUs also For the DSC measurements, a DSC Q2000 (Universal V4.3 TA
exhibit annealing endotherms. They occur at 20e50
C above the Instruments) was used, with aluminium TZero lidded pans from TA
annealing temperature [11,16,22,23]. During heating towards the Instruments. Samples of about 3 mg were cut from pressed plates.
annealing temperature or at the annealing temperature, some of The measuring unit was flushed with 25 ml N2/min during all ex-
the ordered structures will melt. Fluid segments that revert to the periments. Heating and cooling rates were 10
C/min in all cases.
solid state must be stable at the annealing temperature and will The endo- and exothermic peak temperatures correspond to the
melt above the annealing temperature. Higher annealing temper- maxima of the transitions, whereas the glass transitions were
atures result in larger annealing induced endotherms, as more calculated as the inflection point. The enthalpy values were
material is melted and re-arranged. measured relative to the baseline. Examples of these baselines will
This paper explores the morphology of MDI TPU with high HS be shown later in Fig. 4.
content and 1,6-hexane diol as chain extender as a function of the
applied cooling rate and the annealing parameters. The link be- 2.3. X-ray scattering
tween the processing induced morphology and the thermal and
mechanical properties (the storage modulus in particular) is Time-resolved Small Angle X-ray Scattering (SAXS) and Wide
investigated. Angle X-ray Diffraction (WAXD) measurements using synchrotron
radiation were performed on the Dutch-Flemish (DUBBLE) beam
2. Experimental line BM26B at the European Synchrotron Radiation Facility (ESRF)
in Grenoble (France). The experiments were executed at a fixed
2.1. Materials wavelength l of 1.24 Å. A 2D multiwire gas-filled detector, with a 13
by 13 cm area and pixel size of 250 mm, was placed at 2.5 m from the
All research was conducted on Irogran D74 P 4742E by Hunts- sample after an evacuated tube for measuring the SAXS. The known
man. This is a polyether-based TPU with high hard segment con- reflections of silverbehenate were used to calibrate the scattering
tent. The polyether is a polytetrahydrofuran with brand name angles, expressed in terms of q, with q ¼ 4psinq/l and q being half
Terathane and molecular weight of 1000. The isocyanate is 4,40 - the scattering angle. The setup enabled measuring SAXS intensities
methylene diphenyl diisocyanate (MDI) with brand name Suprasec in the range from q ¼ 0.0133 Å1 to q ¼ 0.265 Å1, corresponding
1306. The chain extender is 1,6-hexanediol. The HS content is roughly to structures having a size from 2p/q ¼ 470 Å to 24 Å.
73.5 wt%, if all MDI and 1,6-hexanediol are counted as hard WAXD data were collected simultaneously at a position close to the
segment. sample on a 2D VHR CCD detector from Photonic Science. The
The TPU pellets were dried 24 h at 80
C in a vacuum of less than WAXD intensities were measured in the range q ¼ 0.7 Å1 to
5 mbar. This ensured a thorough drying of the pellets. The pellets q ¼ 2.45 Å1, corresponding to sizes from 2.56 Å to 9 Å. Scattering
were poured in a frame of 250  250  2 mm. This was put in angles were calibrated using a polyethylene sample.
between two aluminium foils, treated with Chemlease PR-90. The The SAXS and WAXD patterns were normalised to the intensity
entire setup was put in between a hot press at 220
C and a pres- of the incoming beam, measured by an ionisation chamber placed
sure of 10 bar was applied during the entire cycle. After 10 min at downstream from the sample. The scattering patterns were cor-
220
C, the samples were cooled down according to the cooling rected for the detector response and in the WAXD case additionally
profile in Fig. 1. The sample was removed when the temperature for the dark current prior to azimuthally averaging the isotropic
dropped below 50
C. This sample is called the reference sample. data using home-made software [24]. Both the SAXS and WAXD
patterns were corrected for the scattering due to the empty setup,
taking into account sample and sample holders transmission dif-
ferences. The SAXS and WAXD intensities are both represented as a
function of q.

2.4. Atomic force microscopy

Atomic force microscopy imaging was carried out in Tapping

Mode™ with a Multimode™ Veeco system, operating with a
Nanoscope IV™ Veeco controller and E-type scanner. An Olympus
silicon Micro Cantilever with a resonance frequency of about
250 kHz was used. The measurements were carried out on TPU
samples prepared in an oven at 220
C. A constant cooling rate of
0.5
C/min was applied. Hence, a slowly cooled sample was created
and imaged. Teflon sheets were put on the top and bottom instead
of release agent. The sample surface was scanned without further
Fig. 1. Temperature profile for the cooling of the reference samples. treatment.
 Y. Swolfs et al. / Polymer 81 (2015) 1e11 3

2.5. Dynamic mechanical analysis densities of the hard and soft segment phases respectively. a is the
volume fraction of two-phase regions. Large mixed regions are
A dynamic mechanical analyser DMA Q800 of TA Instruments assumed to not contribute to the observed SAXS intensity [27].
was used for testing all the samples. The rectangular samples have a The factor D takes into account the background caused by local
size of 50  12.5  2 mm, and were tested in three-point bending electron density fluctuations. The scattered wave function F(q) of
mode. The frequency was set to 1 Hz, while the applied strain was the underlying one dimensional electron density profile corre-
0.05%. The samples were first cooled to 80
C, before starting the sponds to:
heating cycle. The heating and cooling rate was always 2
C/min.
!
The glass transition temperature was taken as the temperature at [ sinðq$LP =2Þ s2LPdis $q2
the middle of the drop in storage modulus. FðqÞ ¼ 2$ S $ $exp 
LP q 2
 
2.6. Interpretation of the SAXS data sinðq$[S =2Þ s2 $q2
 2$ $exp  (2)
q 2
For interpreting the SAXS patterns of TPUs, a lamellar domain
The overall one dimensional electron density profile, producing
structure is often assumed. As explained in the introduction, this is
scattering as described by equation (1), can be understood as the
not always adequate [1,2,25]. Fig. 2 schematically depicts two
average electron density profile, seen from the centre of the SS
morphologies, which e as will be demonstrated in the results
layers with monodisperse thickness, [S. This profile evolves from a
section e are relevant to the present system. In these morphologies,
central SS layer over the next neighbour HS phase to the system
the black areas cover 26.5% of the total area and represent the
average density, <r>. The smoothness of the SS sigmoidal transi-
volume occupied by the SS phases, assuming complete phase
tion is governed by s, that of the HS transition by sLPdis . The total
separation from the white areas, which represent the HS volumes.
thickness of the transition zones equals 3s and 3sLPdis respectively.
The percentage occupancy is in line with the chemical composition,
The parameter LP corresponds to twice the distance from the centre
i.e. 73.5 wt% HS, and disregards differences between weight and
of the amorphous layer to the point at which the electron density,
volume fractions. Within the white areas, the blue inclusions
r(x), reaches <r> for the case sLPdis ¼ 0. With this parameter, the
represent HS crystalline material, which in A occupies 40% of the HS
volume fraction of the SS phase, fS, can be defined as
phase and in B 60%. It is assumed that SAXS is dominated by the
electron density contrast between the HS and SS domains and that
[S
hence the electron density difference between crystalline and fS ¼ (3)
LP
amorphous HS volumes is rather small compared to the difference
between SS and HS regions. Obviously, (1fS) equals the HS phase volume fraction, fH, and
The HS phases in morphology A of Fig. 2 are not lamellar and as a result <r>, is given by
dispersed within the SS matrix. However, as the SS phases are
rather thin, they still can be considered as layer-like, say lamellar. 〈r〉 ¼ fS $rS þ ð1  fS Þ$rH (4)
Morphology B of Fig. 2 is closer to the classical lamellar case where
HS and SS phases alternate in a one dimensional stack. The stacks Fig. 3 visualises LP and [S for sLPdis ¼ 0 and for sLPdis ¼ ðLP  [S Þ=3,
are, however, rather short and the lateral dimensions very limited. the limiting case for which [S remains unaffected by the width of
The SAXS intensity function Idis(q) for disordered morphologies the transition to <r> [26].
such as the ones in Fig. 2A and B and anything in between, can be Equation (1) was suggested earlier to model disordered mor-
calculated as [26]: phologies in semicrystalline starch [26]. Note that <r> does not
appear explicitly in equation (1), whereas this is the case in the
C$a$ðrH  rS Þ2 ½FðqÞ2 equivalent of Equation (1) given in Ref. [26]. This difference origi-
Idis ðqÞ ¼ þD (1) nates from inserting the equations (3) and (4) in the equation given
4$p$q2
in Ref. [26]. This allowed putting the factor (rHrS)2 in front of
The parameter C depends on the irradiated volume and the [F(q)]2. Finally, as a correction to [26], the s-containing factor is
beam intensity. The parameters rH and rS are the electron included within F(q).

Fig. 2. Schematic representation of potential TPU morphologies with the black and white areas corresponding to respectively the soft and hard segment phases. The blue inclusions
within the white areas correspond to crystallites within the hard segment phases. In morphology A, the hard segment phases do not adopt lamellar structures whereas the
morphology in B approaches the classical lamellar case with alternating HS and SS layers in a one dimensional stack. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
4 Y. Swolfs et al. / Polymer 81 (2015) 1e11

describe the ordered and disordered fractions. As a result, the po-

LP
tential scattering due to ordered and disordered volumes is
1
contrast-matched and does not contribute to the SAXS intensity.
The experimental SAXS data for the reference and slowly cooled
0.75
samples were fitted to equation (1). This was performed by a least-
ρ(x) 0.5 S Smooth transi on squares minimisation of the logarithmic errors. To obtain this fit, six
Abrupt transi on parameters were allowed to vary: C$aS$(rHrS)2, D, [S, LP, sLPdis and s.
0.25 The thickness s of the transition zone was set to zero to ensure a
stable fit, reducing the effective number of variables to five. The
0 standard deviation sLPdis of the average stack periodicity was con-
-80 -60 -40 -20 0 20 40 60 80 strained to be smaller than (LP[S)/3. This constraint ensures that
the relative density r(x) reaches rHat its maximum.
x (Å)
Only q-values below q ¼ 0.214 Å1 were used in the optimisa-
Fig. 3. Schematic representation of the electron density functions associated with tion to avoid interference with the WAXD data. All parameters were
equation (1) with s ¼ 0 and sLPdis ¼ 0 (blue curve) or sLPdis ¼ ðLP  [S Þ=3 (green dashed allowed to vary as a function of temperature. The initial guess of the
curve). A zero s value implies a sharp transition between the central soft segment solution at each temperature was equal to the solution at the pre-
phase (r(x) ¼ 0) and the high density hard segment phases (r(x) ¼ 1). The evolution
vious temperature. In the case of both reference and slowly cooled
from the hard segment phase density to the system average value (r(x) ¼ 0.75) is sharp
for the blue curve and limiting smooth for the green dashed curve. (For interpretation samples, a good fit was achieved with equation (1), which is
of the references to colour in this figure legend, the reader is referred to the web indicative for disordered morphologies. Electron density profiles,
version of this article.) such as illustrated in Fig. 3, can then be calculated by inverse
Fourier transformation of the scattered wave function F(q):

The SAXS intensity function Iord(q) for an ordered two-phase Z∞

1
morphology composed of alternating lamellae is given by Ref. [26]: rðxÞ ¼ 〈r〉 þ 2$FðqÞ$cosðq$xÞ$dq (8)
2$p
  0
2 32
2$sin q$[
2
S
  For this calculation rH and rS were set equal to 1 and 0 respec-
C$a$ðrH  rS Þ 4 2
5 $SðqÞ$exp q2 $s2 þ D
Iord ðqÞ ¼ 2 tively, as their absolute values cannot be extracted from the lumped
4$p$q q
fitting parameter C$a$(rHrS)2. For that reason the computed
(5) electron densities will be referred to as apparent electron densities.
Note that only half of the electron density profiles are displayed in
The factor between square brackets is the form factor for the SS
the discussion session. Adding a mirror pattern at the negative x-
layers, which are assumed to have a monodisperse thickness of [S.
axis side would yield the full profiles, similar to the schematic one
The function S(q) is the interference function of the paracrystalline
in Fig. 3. At high x-values r(x) reaches 〈r〉. Given equation (4) and
positioning of the soft segment layers and is calculated as:
the fact that rH and rS are set to 1 and 0, the apparent 〈r〉 equals the
  SAXS-based hard segment volume fraction.
1  exp  q2 $s2LPord
SðqÞ ¼    
1 þ exp  q2 $s2LPord  2$exp  0:5$q2 $s2LPord $cosðq$LP Þ 3. Results

(6)
3.1. Thermal transitions during heating
The standard deviation of the average stacking periodicity LP for
such lamellarly ordered morphologies is given by sLPord . To avoid Fig. 4 presents the heat capacity DSC thermograms. In the
negative LP values within the distribution, the ratio sLPord =LP should reference sample, five peaks can be distinguished, labelled as TI, TII,
be less than 0.4. The average HS phase thickness equals LP[S and is
characterised by a distribution identical to that of LP.
The q value at which equation (5) produces a maximum in Iord(q)
is closely related to LP via Bragg's Law, in particular when sLPord is Slowly cooled
low. The scattered intensity given by equation (1) also produces a
peak at a given q value, but the relation with LP is not trivial [26]. Heat capacity
The TPU morphology can also be intermediate between the 1 J/(g.K)
disordered and lamellarly ordered cases. Such a morphology, with a
scattered intensity Iint(q), can be modelled by a linear combination 72h at 150°C
of equations (1) and (5):
72h at 100°C
2
C$aðrH rS Þ Reference
Iint ðqÞ ¼ $
4$p$q2 TIII
  TI TV
2 2 32 3 TII TIV
2$sin q$[
2
S
 
4ð1bÞ½FðqÞ2 þb$4 5 $SðqÞ$exp q2 $s2 5 þD 0 50 100 150 200
q
T (°C)
(7)
Fig. 4. DSC heat capacity thermograms for the four studied samples. The relevant
temperatures for the reference sample are indicated by arrows. The dashed line in-
where b is the fraction of lamellarly ordered and (1b) the fraction dicates the location of the glass transition for the reference system, which was ob-
of disordered volumes. The two fractions are modelled to have the tained by cooling rather rapidly and which resulted in a mixed system. The thin grey
same hard segment fraction by using identical [S and LP values to line indicates the baseline that was used to calculate the enthalpy values.
 Y. Swolfs et al. / Polymer 81 (2015) 1e11 5

TIII, TIV, T V. On top of that, the glass transition at about 40
C is seems to be related to B type crystal formation. At T V, all WAXD
indicated by the dashed line. The temperature and enthalpy of all peaks and shoulders disappear, which indicates all hard seg-
transitions are summarised in Table 1. ments melt. The long range order completely disappears, which
Fig. 5 displays WAXD and SAXS measurements on the reference corresponds to the absence of the A and B polymorphs.
sample at selected temperatures to evaluate the morphology and Concomitantly, the SAXS intensity collapses as illustrated for
phase behaviour as a function of temperature. At room tempera- 205
C, proving that the hard and soft segments fully remix.
ture, crystalline peaks or shoulders are absent in the WAXD-pattern When analysing the WAXD patterns in more depth, the poly-
at 29
C in Fig. 5b. Crystallisation was hence avoided by the fast morph A remnants seems to disappear before polymorph B and
cooling. Furthermore, it can be deduced from the featureless SAXS already during the TIV exotherm. As a result, the WAXD pattern
pattern at 29
C (see Fig. 5a) that the TPU segments remained fully at 180
C only contains the B type reflection, besides a pro-
mixed. In other words, there are no indications for a liquidesolid or nounced amorphous halo.
a liquideliquid type of microphase separation. This morphology
stays valid even after TI, indicating that the TI endotherm is due to 3.2. Thermal transitions during heating after slow cooling
enthalpic recovery at devitrification. This was also demonstrated by
Wilhelm and Gardette for a hexanediol-MDI TPU, albeit with a When the sample is cooled down slowly, the morphology at
different polyether [28]. room temperature is drastically different. During slow cooling,
The TII exotherm is a crystallisation-induced phase separation. the HS have enough time to crystallise, which causes them to
This is proven by two facts. Firstly, three small shoulders are visible separate from the SS. The SAXS peak in Fig. 6a at 34
C reveals
in the WAXD pattern at 120
C in Fig. 5b, which is after the TII that phase separation has indeed occurred. Furthermore, the A as
exotherm. These crystallites with reflections at d-spacing of 3.8 Å, well as the B polymorph are clearly present at room temperature
4.1 Å and 5.1 Å will be labelled polymorph A for the remainder of (Fig. 6b).
this paper. The fact that the shoulders are difficult to observe, Since crystallisation-induced phase separation has already
proves that the crystallinity is low and/or that the crystals them- occurred, the exothermic peak at TII no longer appears (see Fig. 4).
selves are small and/or of little internal order. Secondly, a SAXS The endothermic transition at TIII remains and is again related to
peak is observed in Fig. 5a at 120
C. This is attributed to a phase the partial melting of A-type crystals. The reduction of A type
separation into HS and SS rich phases. crystals can be inferred from a comparison of the 34
C WAXD
The microphase separation remains after the endothermic pattern with the ones at 155
C or 181
C. The patterns at 155
C
signal at TIII, although different in nature as the SAXS peak has and 181
C were taken above the TIII transition and clearly reveal
shifted to smaller q values (see the 164
C pattern in Fig. 5a). weaker type A reflections. In SAXS, a small decrease in
Furthermore, the type A WAXD shoulders seem to have weak- C$a$(rHrS)2 will be observed at TIII as a result of the creation of
ened at 164
C compared to at 120
C, suggesting A-type crystal mixed regions and hence a reduction of a. Details of the SAXS
melting in the TIII endotherm (see Fig. 5b). Not all A-type crys- shape analysis will be presented and discussed in section “3.4
tals, however, melt at this event, as (very faint) remnants of the Morphology”.
reflections remain up to higher temperatures. At 164
C, the first At the TIV exotherm, further growth of the B type crystals occurs
signs of a second crystalline polymorph start appearing. This as can be deduced from the sharper B type reflection in the 181
C
second polymorph, with its single reflection at a d-spacing of WAXD pattern compared to at 155
C. Concomitantly, the SAXS
4.7 Å (indicated with a dotted line in Fig. 5b) will be labelled peak strongly shifts to lower q values. Again, all crystals melt at the
polymorph B. T V peak as deduced from the collapsing SAXS intensities and WAXD
Blackwell and Lee [14] also detected two polymorphs in an crystalline reflections at 203
C in Fig. 6. Further investigation
hexanediol-MDI TPU and proved that they were both triclinic. revealed that polymorph A and B disappear rather simultaneously,
Similarly, D'Hollander et al. [29] found two polymorphs in their but that ultimately only the B type crystals remain prior to
TPU: a triclinic form having multiple diffraction peaks and a completion of the melting process. The long range order as probed
pseudo-hexagonal one with a single reflection at 4.2 Å. Polymorphs by SAXS disappears completely when the last polymorph B crystals
with a single strong reflection (cfr. polymorph B) are quite often have melted.
due to the pseudo-hexagonal packing of stretched-out polymer
chain with a cylindrical symmetry due to (frozen in) rotational 3.3. Thermal transition during heating after annealing the reference
freedom along the polymer chain axis. Crystal structures with well- samples
defined chain packing readily produce multiple reflections (cfr.
polymorph A). Polymorph A and B thus represent crystals with The influence of annealing the reference sample is revealed in
respectively high and low short range order. Fig. 7. The WAXD patterns prove the development of polymorph A.
The WAXD intensity associated with the polymorph B gains Furthermore, the small shoulders after annealing at 100
C become
importance during the small exotherm at TIV. This exotherm thus more pronounced after annealing at 150
C. Note that the low angle

Table 1
Temperature and enthalpy of the thermal transitions in the four studies samples. Not all transitions were present in all samples, and the TIII transition for the sample annealed
at 100
C could not be determined reliably due to the partial overlap with an annealing endotherm.

Tg TI TII TIII TIV TV

Reference Temperature (
C) 37.3 ± 0.5 45.3 ± 0.4 84.7 ± 1.5 148.2 ± 0.8 166.6 ± 1.0 188.8 ± 0.6
Enthalpy (J/g) None 1.3 ± 0.1 11.1 ± 1.7 9.4 ± 1.7 1.5 ± 0.2 4.6 ± 1.2
72 h at 100
C Temperature (
C) n/a 46.9 ± 4.5 n/a n/a 167.3 ± 1.1 186.8 ± 0.3
Enthalpy (J/g) None 2.5 ± 0.6 n/a n/a 2.9 ± 0.6 5.5 ± 0.3
72 h at 150
C Temperature (
C) n/a 43.2 ± 1.7 n/a 160.9 ± 1.5 174.0 ± 0.2 187.4 ± 0.8
Enthalpy (J/g) None 2.2 ± 0.5 n/a 20.9 ± 2.7 2.2 ± 0.4 4.6 ± 0.8
Slowly cooled Temperature (
C) n/a 47.6 ± 1.5 n/a 147.2 ± 0.6 163.8 ± 1.7 184.9 ± 1.3
Enthalpy (J/g) None 4.1 ± 0.5 n/a 3.2 ± 1.0 3.5 ± 1.8 15.0 ± 4.5
6 Y. Swolfs et al. / Polymer 81 (2015) 1e11

Fig. 5. (a) SAXS and (b) WAXD patterns of the reference sample. For the SAXS data, both the experimental and fitted (using equation (1)) data are plotted. The fits for 29
C, 63
C
and 205
C are not presented as the segments are homogeneously mixed at these temperatures. The WAXD data are shifted over the intensity axis for the sake of clarity.

Fig. 6. (a) SAXS and (b) WAXD patterns of the slowly cooled sample. For SAXS, both the experimental and fitted (using equation (1)) data are plotted. The fit for 203
C is not
presented as the segments are homogeneously mixed at this temperature. The WAXD data are shifted over the intensity axis for the sake of clarity.

WAXD reflection is slightly shifted to higher q values (q ¼ 1.29 Å1) develop polymorph B. The phase separation in the annealed sam-
compared to after cooling slowly (q ¼ 1.23 Å1). This may point at A ples is clear from the SAXS data in Fig. 7a. At higher annealing
type crystals with a higher density after annealing. Polymorph B is temperatures, structures have evolved to larger length scales, as the
not present in the samples. For the reference sample, polymorph B SAXS peak has shifted to smaller angles. Note that fitting to equa-
develops within the TIV exothermic peak, which occurs above tion (1) yields poor reproductions of the experimental SAXS pat-
150
C. Hence, the annealing temperature was not high enough to terns, indicating that the morphology cannot be described by this
 Y. Swolfs et al. / Polymer 81 (2015) 1e11 7

Fig. 7. (a) SAXS and (b) WAXD patterns of the reference sample before and after annealing. These data were collected at room temperature, after cooling down from the annealing
temperature. For SAXS, both the experimental and fitted (using equation (1)) data are plotted. The fit for the reference sample is not presented as the segments are homogeneously
mixed at this temperature. The WAXD data are shifted over the intensity axis for the sake of clarity.

limiting case of high disorder. An alternative fitting will be dis- 3.4. Morphology
cussed in section “3.4 Morphology”.
All samples were stored at room temperature for at least 24 h Before analysing the temperature dependent SAXS data, a
before the DSC measurements were started. This results in an microscopic view of the morphology is needed to verify the sug-
enthalpic recovery endotherm TI around 40
C in all samples (see gested models. Fig. 8 presents the AFM phase image of a slowly
Fig. 4). TII is absent in both annealed samples because the phase cooled sample. The Teflon sheet topography too strongly contrib-
separation has already occurred during the annealing or during uted to the height image, hampering its analysis (data not shown).
heating up to the annealing temperature. The most conspicuous 100 nm features in the phase image relate to
The sample annealed at 100
C developed a new DSC peak at what made the height image uninformative. The features of interest
125
C. This is an annealing endotherm and most likely represents
the melting of very small A-type crystals that are only stable up to
temperatures just above the annealing temperature where they
were created. Unfortunately, no time-resolved X-ray data was
available for the annealed samples to verify this hypothesis. At
higher temperatures, the thermal behaviour of the sample
annealed at 100
C closely resembles that of the reference sample
and should be understood in similar terms.
The annealing endotherm is also present in the sample annealed
at 150
C. In that case, however, the peak joins the TIII peak, which is
shifted to higher temperatures and larger compared to in the other
samples. Annealing a fast cooled sample at 150
C seems to produce
a higher amount of more perfect A type crystals compared to when
no annealing was allowed. The perfection of these A type crystals is
however lower compared to the majority of the A type crystals that
were formed during slow cooling. This can be deduced from the
sharper type A reflection after slow cooling (Fig. 6b) as well as from
the fact that the most stable A type crystals after cooling slowly,
survive up to very high temperatures and melt within the broader
T V endotherm where also the B type crystals melt. Clearly, after
cooling fast (reference sample) and irrespective of the annealing
protocols, B type crystals are only generated in the TIV exotherm
and melt in the T V endotherm. Once the majority of A-type crystals
melted in the TIII endotherm, the three systems arrive at a com-
parable state prior to enrolling into the TIV and T V DSC signals (see
Fig. 4 and Table 1). Accordingly, these signals for the sample Fig. 8. AFM phase image at room temperature of a slowly cooled sample. The phase
angle runs from 0 (black) to 4
(white). The inset is artificially coloured for better
annealed at 150
C are comparable with those of the reference
visualisation. (For interpretation of the references to colour in this figure legend, the
samples or the sample annealed at 100
C. reader is referred to the web version of this article.)
8 Y. Swolfs et al. / Polymer 81 (2015) 1e11

are highlighted by artificial colouring in the 500 nm box within

Fig. 8. A disordered morphology can be observed similar to the one
in Fig. 2B. The layer-to-layer distances in the brighter patches fall
between 150 and 300 Å. The SAXS data of the reference and slowly
cooled samples were hence fitted to equation (1). Representative
fits are displayed in Figs. 5 and 6, illustrating that the data was fitted
well. The data with larger deviations between experiment and fit
will not be shown or discussed. At those temperatures, the
morphology was not sufficiently developed yet.
Fig. 9 presents one wing of the related apparent electron density
profiles of the reference and slowly cooled TPU as a function of
temperature. The corresponding parameters, [S, LP, sLPdis , fH and
C$a$(rCrA)2 are presented in Fig. 10. The parameter D is of little
interest. For the reference sample, the functions and parameters
below 114
C and above 191
C are left out. Section “3.1 Thermal
transitions during heating” proved that no phase separation ex-
ists below 60
C. Even though phase separation developed between
60
C and 114
C, the patterns below 114
C were too ill-defined to
ensure a good fit. Above 191
C, the SAXS patterns collapsed due to
complete melting and segment remixing. For the slowly cooled
sample, data are presented up to 191
C for the same reason.
It is convenient to first discuss the SAXS based morphological
development during heating after cooling slowly because this
represents the simplest case. In the temperature range before the
TIII endotherm, except for devitrification, nothing seems to happen
in DSC (Fig. 4). In Fig. 10, the SAXS parameters in this temperature
range remain rather constant, except that [S, LP and C$a$(rHrS)2

Fig. 10. Evolution of (a) [S, (b) LP, (c) sLPdis , (d) fH, and (e) C$a$(rHrS)2 as a function of
temperature for the reference (black curves) and slowly cooled (grey curves) sample
and based on a fitting of the SAXS patterns to equation (1). The vertical transparent
bands demarcate the temperatures ranges of the TIII, TIV, T V thermal transitions. The
horizontal dashed lines are helpful to the discussion (see text).

slightly increase. This is most likely due to thermal expansion with

the SS phases expanding more rapidly than the HS phases such that
the factor (rHrS)2 increases. C and a presumably remain constant
in this temperature range. Furthermore, the HS phase fraction, fH,
probed by SAXS in this range amounts to 80%, which exceeds the
mass based HS chemical share, i.e. 73.5%. Assuming a higher density
for the HS phase, one would expect a lower, rather than a higher HS
SAXS based volume fraction if the phase separation were complete.
Fig. 9. The apparent electron density profiles r(x) for (a) the reference and (b) slowly
It thus seems that part of the SS moieties are included within the HS
cooled TPU in temperature ranges where the fits were reliable. volumes and hence that phase separation is incomplete even after
 Y. Swolfs et al. / Polymer 81 (2015) 1e11 9

slow cooling. For sure, the WAXD data in Fig. 6 do not suggest a value of 78% (Fig. 10d) for the reference sample at 114
C. This
crystallinity of 80%, implying that the HS volumes are only partially value is slightly below the 80% of the slowly cooled sample, but
crystalline and that morphologies similar to Fig. 2 apply with the still above what can be expected from the chemical composition,
hard and soft phase morphology dominating over the morphology suggesting incomplete phase separation or the existence of ho-
established by the crystalline regions in SAXS. mogeneous SS enriched regions that escape the SAXS observa-
In the region associated with the TIII endotherm, [S and LP tion. Large scale segregation more readily happens at higher
continue to increase, albeit at a somewhat higher rate. Concomi- temperatures and when the crystallisation rate is rather low such
tantly, C$a$(rHrS)2 drops to 80% of the original level (80% line in as during slow cooling. During cold crystallisation at TII, very
Fig. 10e). Given that the temperature dependence of (rHrS)2 is little time for segregation is allowed. One can therefore expected
rather small (as deduced from the behaviour in the low tempera- less large scale segregation, accordingly less regions that do not
ture region) and assuming that C remains constant over the entire contribute to the SAXS and hence a fH value closer to what is
experiment, it seems that a decreases. This is associated with the chemically expected. At 114
C, the morphology is finer
melting of A type crystals (cfr. the WAXD data discussed above) and compared to after slow cooling as both [S and LP are smaller for
a homogenisation of about 20% of the volume into volume elements the reference sample. Furthermore, the slightly lower
that are large enough to escape the SAXS observation [27]. This C$a$(rHrS)2 of the slowly cooled compared to for the reference
reasoning builds on the idea that originally, the entire material sample at 114
C (and assuming comparable rH and rS values for
volume was nanoscale phase separated (100% line in Fig. 10e). The the two cases) indicates that a for the reference is slightly larger.
increase of [S, LP and fH might be due to the preferential melting of In other words: there are less SS enriched regions that do not
regions with the lowest HS fraction and the smallest crystals, by contribute to the SAXS for the reference sample. Although C is
which the average values as probed by SAXS increase. This is in not known in absolute terms, it can be defended that the C
agreement with thermodynamic principles, stating that larger values for both samples are identical as normalisations were
crystals are more stable. performed to the intensity of the direct beam measured by an
According to the same reasoning, the observed value for 〈r〉 ionisation chamber downstream from the sample.
corresponds to the average density of the remaining phase sepa- Recall that the crystallinity at this temperature for the reference
rated regions rather than of the entire system. Alternatively, the sample is much lower than for the slowly cooled material (see the
upward drift of fH further away from the chemically expected WAXD patterns in Figs. 5b and 6b). The SAXS parameters thus again
fraction, might be a result of a partial (re)mixing of the SS with HS suggest that the SAXS signal is dominated by phase separated
at the TIII endotherm within the remaining phase separated regions. morphology, rather than by the crystallinity of the material.
A remixing of the SS with the HS should lead to approaching rH and Following this reasoning, the morphologies in Fig. 2A and B would
rS values. Indeed, a decrease in Ref. rH can be calculated by using scatter equally strong in SAXS although their crystallinities are
equation (4) and assuming that 〈r〉 equals the chemical composi- different. Fig. 2A most likely is a fairly rather realistic representa-
tion and rS zero at all time. Calculating (rHrS)2 based on the so tion of the finer morphology of the reference sample, whereas
calculated rH value with rS set to zero, learns that the decrease in Fig. 2B inclines to the case after slow cooling. The very broad WAXD
(rHrS)2 by this effect in the TIII region is smaller than the increase reflections for the reference sample (see Fig. 5) point at small,
due to differential thermal expansion as estimated from the disconnected crystallites as in Fig. 2A whereas the narrower peaks
behaviour of C$a$(rHrS)2 at low temperature. The drops in for the slowly cooled sample are indicative for larger, more con-
C$a$(rHrS)2 at TIII is therefore realised by a decrease of a rather nected crystallites as in Fig. 2B.
than by changes in (rHrS)2. Accordingly, the scenario in which [S, During the endotherm at TIII, C$a$(rHrS)2 of the reference
LP and fH increase by the preferential melting of regions with the strongly decreases due to the melting of A type crystals and the
lowest HS fraction and the smallest crystals is most realistic. ensuing remixing of segments. A higher fraction of A type crystals
Following this line of thought, the initial fH value of 80% (instead of melts compared to after cooling slowly. Only 36% of the sample
the chemically expected lower value) may indicate the presence of volume remains phase separated whereas this was 80% for the
homogeneous SS enriched regions that do not contribute to the slowly cooled sample (see Fig. 10d). Concomitantly, [S, LP and fH
observed SAXS, rather than a partial mixing of the segments in all increase, suggesting that HS-rich regions with larger crystallites
regions. In that case, a would be slightly lower than 1 also at the survive. In this case, the change of these parameters might also
lowest temperatures. reflect remixing processes within the remaining separated regions.
Within the TIV exothermic region, the increase of [S, LP and fH Part of this internal remixing process is indeed reversed at the
continues whereas C$a$(rHrS)2 remains rather constant. Changes, creation of B type crystals in the temperature region towards and at
such as the growth of B type crystals (which could contribute to the the exotherm at TIV. In this region, [S and LP decrease again whereas
increase of LP and fH) thus happen within the remaining semi- fH decreases. C$a$(rHrS)2 remains unaffected, suggesting that the
crystalline regions. In the T V endothermic zone, C$a$(rHrS)2 recrystallisation into B type crystals predominantly happens in A
evolves similarly as in the TIII endothermic region, reflecting the crystalline regions that melted the latest. This is conceivable, as
melting of A and B type crystals together with a remixing of the larger A type crystallites leave larger HS enriched regions, which
different segments. The increase of fH in this temperature range more readily nucleate B type crystallites. At temperatures inter-
again suggests that the regions with the highest HS fraction and the mediate between TIII and TIV, sLPdis increases steeply. This may
largest crystals (mirrored in LP being large) melt the latest. reflect a wide distribution in HS phase sizes and shapes. A similar
Fig. 9b presents the apparent electron density functions of the increase is seen for the slowly cooled sample in the T V endotherm
slowly cooled TPU. The transition from the crystalline density to the where most of the material melts. For the reference sample at T V,
average density is smoothened out as the temperature increases. features are visible that remind of what happened during the TIII
This effect is mainly caused by the increasing average electron endotherm. This is expected for a continued melting and remixing
density of the (remaining) two phase regions by the local HS excess process.
therein. The growth of B crystals during the heating run also con- The apparent electron density functions of the reference TPU in
tributes as mentioned above. Fig. 9b are comparable to for the slowly cooled material. The flat-
Once the crystallisation induced phase separation has been tening of the profiles, however, happens at lower temperatures
established during the exothermic signal at TII, SAXS probes a fH where the A type crystal melt at TIII.
10 Y. Swolfs et al. / Polymer 81 (2015) 1e11

Equation (1) was not able to fully describe the morphology of

the annealed TPUs as the experimental SAXS peaks are too sharp.
Therefore, a fitting was made using equation (7). For this fitting,
two new parameters were added to the five optimisation param-
eters used earlier: sLPord and b. The results of these fits are illustrated
in Fig. 11. For the sample annealed at 100
C, a reasonable fit was
obtained by including 10% of a lamellarly ordered morphology (see
Fig. 11a). This increased to 15% for the sample annealed at 150
C
(see Fig. 11b), indicating that more ordered structures can form at
higher annealing temperatures. For the sample annealed at 100
C,
fH equals 78%, just like for the reference sample right after the TII
exotherm. Also the [S and LP values are comparable, i.e. 21 and 100 Å
respectively. The originally formed dispersed small crystallites and
domains thus (partially) merge laterally during annealing into
lamellae while preserving the separation length scale in the stack
direction. After annealing at 150
C, however, fH decreased down to
76% and [S and LP rose to 32 and 127 Å respectively. The morpho- Fig. 12. Storage modulus as a function of temperature for the four TPUs with different
logical coarsening thus seems to have progressed further after thermal history.
annealing at 150
C compared to after annealing at 100
C since the
distances in the stack direction have also grown. Secondly, the
lower fH value suggests that regions that remained mixed after with a domain structure. These crystalline domains act as physical
cold crystallisation of the reference also phase separated at 72 h cross-links between the liquid chains and increase the modulus by
150
C. All in all, annealing increases the crystallinity and improves this crosslinking as well as by the stiffness increase associated with
the HS phase connectivity. The DSC peak at T V (see Fig. 4) seems to crystallisation.
be somewhat larger after annealing compared to for the reference In the annealed and slowly cooled samples, the crystallisation-
sample, indicating a more efficient conversion to B type crystals induced phase separation has already occurred. This gives a lower
after annealing. This can be due to the larger and more connected modulus value just below the glass transition of the mixed refer-
HS phases by which more readily (larger) B-type crystals can be ence system, because the phases enriched in SS devitrify earlier.
generated at high temperature. This effect is not so obvious in the DSC data in Fig. 4, likely because
Note that sLPdis was found to be zero in both annealed samples. devitrification happens over a rather large temperature range.
This suggests that the morphology is truly intermediate between Phase separation for the slowly cooled and annealed samples also
disordered and lamellarly ordered. yields a higher modulus at temperatures above the glass transition
temperature of the mixed phase. The HS enriched region may
devitrify later because of their own nature, but also because of
3.5. Influence on storage modulus
being connected to HS crystallites. These HS keep the modulus
relatively high after the main glass transition temperature.
Fig. 12 illustrates the storage modulus of the four sample types.
The modulus of the reference only increases once crystallisation
The reference sample is not phase separated yet at room temper-
sets in at the TII exotherm. However, the reference modulus never
ature. Therefore, the reference sample stays reasonably stiff below
reaches the value of the slowly cooled or annealed samples. This
the glass transition temperature of the mixed phase around 40
C.
might be due to the lower crystallinity as well as the presence of
Next, the sample displays elastomeric behaviour. The
dispersed, rather than connected HS phases. A large fraction of the
crystallisation-induced phase separation starts around 60
, corre-
crystallites formed by annealing at 100
C melt during the
sponding to the onset of the TII peak in Fig. 4. This transition
annealing endotherm at about 130
C by which the modulus
changes the TPU morphology from a fairly soft amorphous polymer
tumbles down to the reference value. The modulus up to the onset
above its glass transition temperature to a phase separated polymer

Fig. 11. Fit with equation (7) (green line) to the experimental SAXS patterns (dots) for the reference TPU annealed for 72 h at (a) 100 and (b) 150
C. The disordered and (lamellarly)
ordered shares are represented in blue and grey lines respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
 Y. Swolfs et al. / Polymer 81 (2015) 1e11 11

of the TIII endotherm for the sample annealed at 150
C or after Science project (G.0C12.13). I. Verpoest holds the Toray Chair in
cooling slowly remains rather high. Interestingly, a higher modulus Composite Materials at KU Leuven.
is seen for the sample annealed at 150
C compared to after cooling
slowly. The slowly cooled sample contains a large fraction of B-type
crystals (see Fig. 6b) whereas the sample annealed at 150
C only References
contains A type crystals (see Fig. 7b). The large endotherm close to
TIII for the sample annealed at 150
C suggests that its crystallinity [1] J.T. Garrett, C.A. Siedlecki, J. Runt, Macromolecules 34 (20) (2001) 7066e7070.
[2] A. Saiani, C. Rochas, G. Eeckhaut, W.A. Daunch, J.W. Leenslag, J.S. Higgins,
is relatively high, and likely higher than of the slowly cooled Macromolecules 37 (4) (2004) 1411e1421.
sample. The sharpness of the WAXD peaks suggests more perfect [3] C. Li, J. Liu, J. Li, F. Shen, Q. Huang, H. Xu, Polymer 53 (23) (2012) 5423e5435.
crystallites for the slowly cooled sample, but it seems that the [4] C. Li, J. Han, Q. Huang, H. Xu, J. Tao, X. Li, Polymer 53 (5) (2012) 1138e1147.
[5] M.A. Hood, B. Wang, J.M. Sands, J.J. La Scala, F.L. Beyer, C.Y. Li, Polymer 51 (10)
crystalline fraction is more important to the modulus. (2010) 2191e2198.
[6] H.J. Qi, M.C. Boyce, Mech. Mater. 37 (8) (2005) 817e839.
4. Conclusion [7] G.A. Skarja, K.A. Woodhouse, J. Appl. Polym. Sci. 75 (12) (2000) 1522e1534.
[8] K. Aou, S. Ge, D.M. Mowery, R.C. Zeigler, R.R. Gamboa, Polymer 56 (0) (2015)
37e45.
The investigated TPU with a high hard segment content crys- [9] Y. He, D.L. Xie, X.Y. Zhang, J. Mater. Sci. 49 (21) (2014) 7339e7352.
tallises rather slowly. Crystallinity can be avoided by cooling rapidly [10] Y. He, X. Zhang, J. Runt, Polymer 55 (3) (2014) 906e913.
[11] A. Saiani, W.A. Daunch, H. Verbeke, J.W. Leenslag, J.S. Higgins, Macromolecules
down to below the glass transition temperature. However, crys- 34 (26) (2001) 9059e9068.
tallisation is induced during reheating. This type of crystallisation [12] W.C. Hu, J.T. Koberstein, J. Polym. Sci. Part B Polym. Phys. 32 (3) (1994)
induces a disordered domain morphology as suggested by model- 437e446.
[13] J.W.C. Vanbogart, D.A. Bluemke, S.L. Cooper, Polymer 22 (10) (1981)
ling of the SAXS patterns. Semicrystalline species at room tem-
1428e1438.
perature are obtained by cooling rather slowly or by annealing [14] J. Blackwell, C.D. Lee, J. Polym. Sci. Part B Polym. Phys. 22 (4) (1984) 759e772.
rapidly cooled materials at a sufficiently high temperature. The [15] K. Bagdi, K. Molnar, A. Wacha, A. Bota, B. Pukanszky, Polym. Int. 60 (4) (2011)
level of disorder depends on the thermal history. During slow 529e536.
[16] D.J. Martin, G.F. Meijs, P.A. Gunatillake, S.J. McCarthy, G.M. Renwick, J. Appl.
cooling, a disordered morphology is favoured, while during Polym. Sci. 64 (4) (1997) 803e817.
annealing 12e15% of the material volume adopts a lamellarly or- [17] R.W. Seymour, S.L. Cooper, J. Polym. Sci. Part C Polym. Lett. 9 (9) (1971), 689-
dered morphology. The differences in morphology as a function of &.
[18] S.B. Clough, N.S. Schneider, A.O. King, J. Macromol. Sci. Part B Phys. 2 (4)
thermal history and temperature correlated well with changes in (1968) 641e648.
the storage modulus. [19] S.L. Samuels, G.L. Wilkes, J. Polym. Sci. Part C Polym. Symp. (43) (1973)
Based on SAXS and WAXD measurements, the multiple endo- 149e178.
[20] R.M. Briber, E.L. Thomas, J. Macromol. Sci. Part B Phys. B22 (4) (1983)
therms in the investigated TPU were attributed to the presence of 509e528.
two polymorphs. The complexity in thermal behaviour is further [21] R.W. Seymour, S.L. Cooper, Macromolecules 6 (1) (1973) 48e53.
increased by annealing effects and recrystallisation phenomena. [22] J.W.C. Vanbogart, P.E. Gibson, S.L. Cooper, J. Polym. Sci. Part B Polym. Phys. 21
(1) (1983) 65e95.
Stiffer TPUs can be obtained by tweaking the morphology towards [23] T.K. Chen, T.S. Shieh, J.Y. Chui, Macromolecules 31 (4) (1998) 1312e1320.
higher crystallinities and more ordered phase separated morphol- [24] C.J. Gommes, B. Goderis, J. Appl. Crystallogr. 43 (2010) 352e355.
ogies. This can be exploited by applying suitable thermal histories. [25] C.H.Y. Chentsai, E.L. Thomas, W.J. Macknight, N.S. Schneider, Polymer 27 (5)
(1986) 659e666.
[26] S.V. Gomand, L. Lamberts, C.J. Gommes, R.G.F. Visser, J.A. Delcour, B. Goderis,
Acknowledgements Biomacromolecules 13 (5) (2012) 1361e1370.
[27] B. Goderis, H. Reynaers, M.H.J. Koch, V.B.F. Mathot, J. Polym. Sci. Part B Polym.
The authors thank the staff of DUBBLE at the ESRF in France for Phys. 37 (14) (1999) 1715e1738.
[28] C. Wilhelm, J.L. Gardette, Polymer 39 (24) (1998) 5973e5980.
assisting in the SAXS and WAXD measurements. FWO-Vlaanderen [29] S. D'Hollander, C.J. Gommes, R. Mens, P. Adriaensens, B. Goderis, F. Du Prez,
is gratefully acknowledged for supporting the ESRF/DUBBLE Big J. Mater. Chem. 20 (17) (2010) 3475e3486.