Materials and Design 187 (2020) 108387

Contents lists available at ScienceDirect

Materials and Design

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

Prediction of plasticization in a real biopolymer system (starch) using

molecular dynamics simulations
Hüsamettin D. Özeren, Richard T. Olsson, Fritjof Nilsson, Mikael S. Hedenqvist ⁎
School of Engineering Sciences in Chemistry, Biotechnology and Health, Fibre and Polymer Technology, KTH Royal Institute of Technology, SE-100 44 Stockholm, Sweden

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

• A molecular simulations methodology

was developed to be able to predict
plasticization of polymers (tested here
on a complex system; thermoplastic
starch).
• The methodology, and the simulated
systems, were evaluated against experi-
mental density, glass-transition data
and mechanical properties.
• In the simulations, it was possible to ob-
serve the transition region where plasti-
cization occurs experimentally (20–30
wt.% plasticizer).
• Plasticizer percolation was suggested as
a possible cause for the occurrence of
the transition region.

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

Article history: Virgin biopolymers are often brittle, which means that they need efficient, sustainable, non-toxic plasticizers for
Received 3 November 2019 most practical applications. Although the mechanical properties of biopolymers plasticized with e.g. sugars have
Received in revised form 20 November 2019 been extensively investigated, the explanation why efficient plasticization normally only occurs above 20 wt%
Accepted 22 November 2019
plasticizer is still lacking. In this work, starch/glycerol was used as a model system to show that all-atom molec-
Available online 27 November 2019
ular dynamics (MD) simulations can be used to capture the transition region at 20–30 wt% plasticizer, where
Keywords:
plasticization becomes pronounced. Tensile properties and PVT data (densities and glass transition tempera-
Plasticization tures) were obtained both from MD simulations and from measurements on real starch/glycerol materials,
Prediction confirming that MD could capture the experimentally observed transition region. Also, the simulated glycerol dif-
Simulation fusivity correlated well with the trends in the mechanical properties. Percolation theory was used to derive a
Biopolymer probable explanation of the observed transition. The results indicate that the MD methodology can be used
Starch also for other polymer/plasticizer systems and has the potential to be a valuable tool for optimizing the type
Glycerol and amount of plasticizer in a given polymer, as well as being a tool for the design of new efficient plasticizers.
© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

⁎ Corresponding author. For environmental reasons, it is very important to find sustainable

E-mail address: [email protected] (M.S. Hedenqvist). alternatives to fossil oil-based plastics. Polymers from plants, such as

https://doi.org/10.1016/j.matdes.2019.108387
0264-1275/© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
2 H.D. Özeren et al. / Materials and Design 187 (2020) 108387

polysaccharides, are viable replacements. Starch-based biopolymers are in this case glycerol plasticized starch. The motivation for studying
among the most promising substitutes, because of their good film- this particular system is that glycerol is the most commonly used plas-
forming properties and their wide application range, comprising e.g. ticizer for biobased polymers (including starch), and starch is likely to
food packaging, mulch films and bags as well as non-film applications become a more important alternative to fossil-based polymers in the fu-
(e.g., disposable kitchen items) [1–4]. Owing to its high moisture sensi- ture sustainable society. It was especially interesting to see if it was pos-
tivity, starch is typically blended with less moisture-sensitive materials sible to predict and explain the “transition” between a 20 and 30 wt%
such as polylactic acid (PLA) [5], polycaprolactone (PCL) [5] or polyeth- plasticizer where effective plasticization sets in. Our hypothesis is that
ylene (PE) [6] to obtain fully biobased or semi-biobased products, being the transition is due to the formation of percolated paths of plasticizer
fully or partially biodegradable. Fillers are also used to limit the effects of molecules in this region. Glycerol plasticization of starch and the corre-
moisture on mechanical properties [7–9]. Because of the rigid mono- lation between hydrogen bonding and mechanical properties has been
meric units and a high number of hydrogen bonds, pure starch is brittle investigated using MD in a previous article [35]. Even though that
and has a high glass transition temperature and a melting point above study had merits, it contained two important issues: the atomistic sys-
its degradation temperature [10,11]. To obtain a tougher and more duc- tems were not validated against experimental densities and the proper-
tile material that can be processed like conventional thermoplastics, ties of the amorphous computer-generated systems were compared
starch is plasticized (refer to thermoplastic starch (TPS)) [12–16]. Glyc- with experimentally determined properties of semi-crystalline starch.
erol is a good plasticizer for starch and is also the most widely used plas- In contrast, the computer-generated amorphous starch systems here
ticizer for biobased polymers in research. The choice of glycerol content were validated by comparing the simulated densities with the corre-
depends on the targeted application; the minimum amount of glycerol sponding experimental values of amorphous starch. The degree of plas-
commonly used for plasticization is N25% [17,18]. Co-plasticization has ticization was assessed by examining the glass transition temperature,
also been considered for improving the mechanical and physical prop- the mechanical properties and the glycerol diffusivity. Although care
erties of starch [19]. Although the properties, e.g., the modulus, are usu- was taken to perform all experiments on fully dry starch, a limited
ally reported as a function of plasticizer content, the mechanisms amount of moisture was inevitable sorbed by the samples during all
leading to plasticization are generally not discussed, nor the reason measurements, except with DSC. Hence, to consider any effects of the
why sufficient plasticization often occurs within the region of presence of a small amount of water in the samples, the simulated
20–30 wt% plasticizer/glycerol [5]. Migration of plasticizers from the glycerol-starch systems were modeled both in dry state and with a
material is a problem with hydrophilic plasticizers such as glycerol, es- water content of 3 wt%.
pecially at high relative humidity, and it is important to assess the risk The starch/glycerol system, which is a bio-based commercial mate-
for this in different applications [20,21]. rial with complex and strong interactions between the polar plasticizer
There are four different theories on plasticizer mechanisms [22–33]; and the polymer, was chosen as a first test of our new methodology,
the lubricity, gel, free volume and mechanistic theories. In the lubricity which combines MD with experimental validation measurements. If
theory [27–29], the plasticizer is considered to glide along with the the methodology proves successful for this system, it can also be tested
polymer chains during deformation providing new gliding planes be- on other complex systems (e.g. proteins) as well as on less complex sys-
tween the chains. The gel theory [30,31] considers the polymer as a tems (e.g. semi-polar and non-polar plasticizer-polymer systems). The
gel with non-covalent points of attraction (e.g. through van der Waals development of a predictive tool for determining plasticizer efficiency
forces, hydrogen bonds) along adjacent chains. The action of the in bio-based polymers will enable the optimization of more environ-
added plasticizer is to push the points of attractions further apart, mentally friendly polymers, which (for instance) can be used as renew-
which increases the chain mobility. In the free volume theory [31–33], able packaging materials, sustainable insulating building materials,
it is considered that the addition of a plasticizer contributes with more ecological clothing and furniture, insulation for electrical cables, seals
free volume to the polymer. It results in higher free volume of the poly- for electric power technology and covers for white goods.
mer/plasticizer system and increases the chain mobility. The mechanis-
tic theory (solvation-desolvation equilibrium theory) [22,31] is similar
to the gel theory. The difference is that in the gel theory the plasticizer 2. Experimental
molecules are fixed at the points of attraction whereas in the mechanis-
tic theory they can move along the chains. In addition to the classical 2.1. Materials and film preparation
theories, mathematical models for plasticization have also been devel-
oped. For instance, Mauritz et al. [34] proposed a model, based on the Starch (amylopectin (AP) from waxy maize [CAS: 9037-22-3]) was
classical theories, that could predict the glass transition of plasticized purchased from Sigma-Aldrich and glycerol [CAS: 56-61-5], purity:
polyvinyl chloride (PVC). This plasticizer model accounted for the struc- 99.5%, was purchased from Alfa Aesar. The waxy maize had two maxima
tural features (chain length, branching etc.) of different plasticizer in the degree of polymerization (DP) at ca. 30 and ca. 61 (Fig. S1).
types. Starch films were cast from an aqueous dispersion with 1 and 3 wt%
A full understanding of the factors causing the plasticization and of amylopectin (with respect to total weight of film-forming solution)
why a certain content of plasticizer is needed would enable the de- and various amounts of glycerol (0, 10, 20, 30, 40, 50 wt%, with respect
velopment of tailor-made, even more powerful plasticizers, which to total mass of amylopectin and glycerol). The dispersion was heated at
also show low migration. It is also of considerable importance to be 5 °C/min to 90 °C, where amylopectin was gelatinized. It was kept for
able to assess and predict the plasticizer efficiency of various plasti- 10–15 min at 90 °C, until the dispersion became transparent. The dis-
cizer candidates for a certain polymer system. Using accurate predic- persion with 3 wt% amylopectin was subsequently poured into petri
tion tools, the number of experiments to determine the optimal dishes and left to dry slowly at 50% relative humidity (RH) and 23 °C
plasticizer can be reduced and new effective plasticizers can be de- for 7 days to yield films with a thickness of 0.5 ± 0.05 mm. The
signed and rapidly tested in the computer, before making extensive gelatinized dispersion with 1 wt% amylopectin was dried faster in an
experimental work. oven at 50 °C and ambient relative humidity to yield films with a thick-
Reliable predictions, however, require that the plasticizer/polymer ness of 0.1 ± 0.01 mm. The petri dishes were coated with a layer of
mixture is correctly described, which is particularly challenging for sys- polytetrafluoroethylene supported by an aluminum foil (Bytac Z-27).
tems with strong intermolecular interactions, such as polar (e.g. hydro- The films are referred to as E1 and E3, where 1 and 3 refers to the amy-
gen bonding) plasticizer/polymer systems. This inspired us to lopectin content in the film forming dispersion. A starch system with
investigate if it was possible, using molecular dynamics (MD) simula- glycerol is referred to as either E1-GX or E3-GX, where X is the wt%
tions, to accurately describe and capture the features of such a system, glycerol.
 H.D. Özeren et al. / Materials and Design 187 (2020) 108387 3

2.2. X-ray diffraction (XRD) NPT production runs (2 ns) were performed to compute glass transition
temperatures (Tg). This was followed by NVT diffusion simulations
X-ray diffraction measurements were carried out in a Panalytical (2 ns) and NPT tensile-test simulations. The convergence of the MD sim-
X'pert Pro operating at 40 kV, and 40 mA, with Cu Ka radiation. Diffrac- ulations was carefully controlled by monitoring the density and the en-
tion patterns were recorded using a radiation wavelength of 1.54 Å. The ergy of the systems. After equilibration, the standard deviation for the
crystallinity ratio was determined according to the Rietveld method density was b10 (kg/m3) also for the presumably most scattered sys-
[36]; the crystalline peaks were resolved from the amorphous halo tems. Since the drift was also negligible, it could be concluded that the
and the crystallinity was obtained as the area of the crystalline peaks di- systems had converged. Typical energy profiles for 2 ns Tg simulations
vided by the total intensity within a 2θ of 5–40°. [37] and for 2 ns each of NPT and NVT for diffusion simulations are plotted
in Fig. S5 and Figs. S6-S7 (Supporting Information), respectively.
2.3. Differential scanning calorimetry (DSC)
3.1. Modelling of starch
The measurements were performed with a Mettler Toledo DSC1. The
samples, with a ca. 8 mg weight, were sealed in an aluminum pan (PE In the simulations of the amorphous starch, only linear chains were
No. 0219-0041) with a hole in the lid. After an initial heating at 150 °C used. The double helix amylose model, containing 36 glucose monomer
for 2 h, to ensure that the sample was dry, it was cooled to −80 °C or units in each chain (Fig. 1a), was derived using the molecule databank of
−20 °C at a rate of 10 °C/min and then heated to 250 °C at the same Bioptics [45]. To have a precise polar force description, a fully atomistic
rate. The sample was always exposed to inert atmosphere (nitrogen, model of the starch was used, instead of using a united atom approach.
with a purge rate of 50 ml/min). The atoms were modeled using the CHARMM force field, [46,47] which
uses the Lennard–Jones 12–6 potential to represent the intermolecular
2.4. Dynamic mechanical thermal analysis (DMTA) dispersion–repulsion interactions between atom pairs, the Lorentz–
Berthelot rule was applied for pairwise interactions. [48], and the cou-
The dynamic mechanical analysis (DMA) was conducted in tensile lomb interactions were added to the intermolecular potential. The con-
mode using a DMA Q800. Samples with a size of 25 × 4 mm2 were stant charges were taken from CHARMM. The 1–4 interactions,
used. All runs were performed in a “multi-frequency, strain” mode at considering atoms separated by three covalent bonds, included
1 Hz, 0.08% strain, from −80 to 120 °C at a heating rate of 3 °C/min (unscaled) nonbonded interactions and dihedral/torsional contribu-
with a 15-mm gauge length. The starch films were stored in a degassed tions, whereas the intramolecular interactions included bond and
desiccator containing silica gel for at least 48 h before analysis. bend angle harmonic stretching. Values of the pairwise, bond, bend
angle and dihedral parameters are provided in Tables S1–S11.
2.5. Tensile testing
3.2. Modelling of glycerol and water
The tensile testing was carried out in an Instron Testing Instrument
model 5944 at 23 °C with a 500 N load cell. The films were kept in a A number of recent publications cover simulations of glycerol
degassed desiccator with silica gel for at least 48 h before the test. The (Fig. 1b) with all-atom MD [49], using different force fields. Since the
crosshead speed and gauge length were 25 mm/min and 25 mm, re- studies focused on different material properties, it was difficult to di-
spectively. Straight specimens (30–35-mm long and 4-mm wide) rectly compare the reliability of the used force fields. Fortunately, Jahn
were cut from the film and tensile tested according to ASTM D882. et al. [49] have compared glycerol force fields to examine how well
The film thickness was measured with a Mitutoyo 10C-1128 μm they can reproduce experimental data for various material properties.
(Mitutoyo Scandinavia AB, Sweden). Five specimens were tensile tested In this study on glycerol plasticization, the force field that most satisfac-
for each sample. Specimens fractured at the grips were not included in torily reproduced dynamic properties was chosen [50], because our pri-
the analysis. mary interest was dynamic rather than thermodynamic glycerol
properties. This force field, which was initially developed by Chelli
2.6. Density measurements et al. [51] and later modified by Blieck et al. [50] was based on the
AMBER-FF force field, although with the 1–4 interactions scaled by
The density was measured (three replicates) using the Archimedes 0.833 (Lennard–Jones) or 0.5 (coulomb interactions). Akumni et al.
principle. For each glycerol filler fraction, three specimens were [52] continued the work of Jahn et al. [49] by examining glycerol–
weighed in air and in n-hexane at room temperature using a Mettler To- water mixture systems and found that the TIP4P model gave more accu-
ledo AL104 balance. rate results than the TIP3P model. Therefore the four-point TIP4P water
model for biomolecules [53] was used with glycerol in this work. It ex-
3. Molecular dynamics (MD) simulation tends the traditional three-point TIP3P model by adding an additional
site, usually massless, where the charge associated with the oxygen
MD simulations were performed using the open source LAMMPS atom is placed. This site M is located at a fixed distance away from the
molecular dynamics simulation package [38]. The polymer chains and oxygen along the bisector of the H\\O\\H bond angle. The O-M distance
glycerol were built and placed into the cell/box using the Amorphous was 0.125 Å in the model.
Cell module in Biova Material Studio 2016. The temperature and pres-
sure were controlled by a Nose–Hoover barostat [39,40] and thermostat 3.3. Simulation details
[41,42]. The long range contribution to the electrostatic energy was cal-
culated via Ewald summation [43] with an accuracy value of 10−3. For MD simulations were performed on starch/glycerol systems (0, 10,
the long range pairwise energy calculations, a tail correction was 20, 30, 40, 50 wt% glycerol) at 1 atm, with and without 3 wt% water.
added. To equilibrate the cells, a well-established 21-step slow decom- Three chains (molar mass of each polymer chain: 9098 g/mol) and the
pression method was applied to each simulation cell, using the chosen required amount of glycerol and water molecules were placed in a
force fields [44]. In this method, the temperature and pressure are cubic box/cell with periodic boundary conditions, referred to as S3-
changed systematically up and down using NPT and NVT ensembles in W0 (without water) or S3-W3 (3 wt% water). The box side length var-
order to relax the system to obtain the most energy-mimimized state. ied with temperature and was in the range of ∼30–45 Å (total number of
The 21-step equilibration procedure (Supporting Information SI, atoms: 3537–8255, B Tables, SI). In the simulation of the mechanical
Table S13) had a total simulation time of ≈15 ns. After equilibration, properties a greater system with 10 chains (number of atoms:
4 H.D. Özeren et al. / Materials and Design 187 (2020) 108387

a b

Fig. 1. a) Starch monomer and b) glycerol.

11790–27450, box size: 60–65 Å) were used to improve the accuracy of function of time t and fitting the linear part of the curve with the linear
the obtained modulus and stress values. The thermostat used a time (3D) Einstein equation ref.:
constant for temperature coupling of 0.1 ps, and the barostat used a
pressure relaxation time of 1 ps. The cutoff distance for starch was 12 MSDðt Þ ¼ 6Dt ð1Þ
Å. For glycerol, however, the pairwise interactions were brought
smoothly to zero from r1 = 10 Å to r2 = 12 Å using a polynomial
switching function [54]. The results were independent of the starting More recently, models that include the initial non-linear, confined
configuration as observed by replicating the described procedure. diffusion part of the MSD curve have been suggested [56,57]. An expo-
nential term is added, yielding:
3.3.1. Glass transition temperature (Tg)
0 1
Tg was calculated via isotropic-isobaric cooling of filled cells systems t
−
from 427 °C (or 277 °C) to −123 °C in intervals of 25 °C and at a pressure MSDðt Þ ¼ A@1−e τ A þ 6Dmacro t ð2Þ
of 1 atm. Every temperature was simulated for 2 ns, leading to a total of
34 ns MD simulations with a 1 fs timestep in the isotropic NPT ensem-
ble. The cell dimensions were fully coupled in the x-, y- and z-axes where Dmacro is the long-term random walk diffusion coefficient, and A
and the specific volumes were averaged from the last 200 ps at each and τ are parameters characterizing the confined diffusion process. A
temperature. A pairwise regression method, known as broken-stick re- is the characteristic dimension of the confinement:
gression, was applied to determine the Tg values. Two linear slopes, each
in the glassy and rubbery phases, were fitted to the volume–      
temperature curve. The exact breakpoint of these two lines was defined A ¼ r x 2 eq þ r y 2 eq þ r z 2 eq ¼ 3L2 ð3Þ
as the Tg value.
and L is the characteristic length of the confined domain. In this domain,
3.3.2. Mechanical properties the diffusivity (Dmicro) can be obtained as (in 3D):
The Young's modulus and the yield strength were determined from
the stress–strain curve (SSC). Hossain et al. [55] showed that using dif-
Dmicro ¼ A=6 τ ð4Þ
ferent chain lengths of the simulated polymer (polyethylene) changed
the yield strength significantly. However, the trends in the simulated
SSC were the same as for experimental data. In addition, using very The MD systems for the NVT diffusion simulations (2 ns) were first
high strain rates, as in the case of nanosize MD simulations, results in equilibrated with a 21-step decompression method (15 ns, Table S13,
a higher Young's modulus and yield strength than would be obtained final temperature 300 K) and then relaxed to desired temperatures
experimentally. In order to improve the accuracy and to avoid scattered using NPT MD simulations (2 ns). The mean square displacement data
pressure values, larger systems were therefore needed for the stress– were generated from the trajectories with Diffusion Coefficient Tool in
strain simulations than for the other simulations. A total of 10 simulated VMD software [58].
chains (abbreviated S10) was used, together with the required propor-
tions of glycerol and water. The deformation simulation was performed 4. Results & discussion
uniaxially with a 1-fs timestep. A constant stress (strain rate of 108s-1)
was applied along each principal axis of the simulation cell in separate 4.1. Starch structure
simulations, allowing the cell dimensions in the other two orthogonal
directions to change without constraints at a pressure of 1 atm and X-ray diffraction spectra of starch with various glycerol contents are
the cells were stretched until 50% strain. The tensile test was performed shown in Fig. 2. The material with no plasticizer was fully amorphous
in the NPT ensemble at 300 K (27 °C) and the Young's modulus was ob- for both the slowly (E3) and more rapidly (E1) formed films. However,
tained from the stress–strain slope between 0.05% and 3% strain. The in the presence of glycerol, the slowly cast system crystallized into B-
error bars referring to simulated modulus originate from the uncer- type crystals (Fig. 2a, peaks at 2θ ≈ 15, 17, 22°) [59–61]. Hence, the
tainty of determining the slope, and those referring to the simulated plasticizer-induced increase in chain mobility led to crystallization dur-
maximum stress originate from the noise in the region of maximum ing the drying/casting process, as also shown elsewhere [62], and the
stress (observed over a 0.2% strain interval) (Fig. 5). crystallinity increased with increasing glycerol content (Table 1). For
the more rapidly dried samples (E1 samples), all samples were amor-
3.3.3. Diffusion simulations phous (Fig. 2b). It should be noted that cast amylose films get a high
Traditionally, the diffusion coefficient D from an MD simulation is crystallinity, ca. 35% (B-type crystals), which is nearly independent of
determined by plotting the mean square displacement (MSD) as plasticizer content [63].
 H.D. Özeren et al. / Materials and Design 187 (2020) 108387 5

Fig. 2. X-ray diffractograms representing, from top to bottom, starch samples with 50, 40, 30, 20, 10 and zero % glycerol. a) and b) refer to the E3 and E1 samples, respectively. {Vermeylen,
2004 #175}.

4.2. Pressure-volume-temperature (PVT) data and glass transition similar simulated water-free Tg and the DSC-determined Tg of the amor-
temperature phous material with glycerol contents of ≤20 wt%. The Tg values for the
30–50 wt% glycerol samples, obtained from the simulation, were, how-
The simulated volume versus temperature data are shown in Fig. 3 ever, significantly higher than those obtained from DSC. This trend has
for the pure starch system and those of the other systems are provided been observed also for other polymer systems. Han et al. [68] obtained
in Figs. S8–S9. In Table 1, the simulated and experimental densities at close agreement between experimental and simulated Tg for stiff sys-
room temperature are given, and these are in good agreement. A devia- tems (Tg well above room temperature, e.g. for polystyrene), while sim-
tion from the rule of mixture was noticed for the density of the starch/ ulated Tg's tended to be somewhat higher than experimental ones for
glycerol system, which is expected when mixing a small polar molecule more molecularly mobile systems (Tg close to or lower than room tem-
with a polar polymer matrix. The increase in the degree of molecular perature, e.g. polyisobutylene and polybutadiene). The trend of larger
packing that occurs, especially at lower contents of glycerol, has also differences in glass transitions in more flexible systems was observed
been reported for maltodextrin matrices [64]. also for epoxy thermosets [69]. A comparison between simulated and
In Fig. 4a, experimental Tg values (from DSC data) and simulated Tg experimental Tg, at similar cooling rates, can be performed by shifting
values (from PVT data) are plotted as a function of glycerol content. simulated high-cooling-rate Tg-values to experimental conditions.
The experimental and simulated Tg values of pure starch were similar, Soldera and Metatla [70], used the well-known Williams-Landel-Ferry
slightly above 200 °C [66,67]. For all glycerol-containing samples, the (WLF) equation [33]:
experimental Tg was higher for the semi-crystalline material than for  
the amorphous one, although the differences were in some cases rela- τ −C 1 T−T g
log10 at ¼ log10 ¼ ð5Þ
tively small. Due to the plasticization effect of water, the simulated Tg τg C 2 þ T−T g
was always lower for the water-containing system; 3 wt% water re-
duced the Tg by 15–25 °C, depending on the glycerol content. Again, where at is the shift factor, τ and τg are in this case the time scales of the
the reason why we chose to simulate systems with and without 3 wt% measurement (the inverse of the cooling rates) at T (simulated Tg) and
water was that, although efforts were made to test the samples experi- Tg (experimental Tg), respectively. C1 and C2 are constants, whose “uni-
mentally in the fully water-free condition, it is expected that a minor versal” values are C1 = 17.44 and C2 = 51.6 K [33]. Using the experi-
amount of water is taken up by the sample during the set up and/or run- mental cooling rate (10 °C/min) and the simulated cooling rate
ning of the experiments. The exception is the DSC measurements, which (estimated from the 25 °C cooling step over 2 ns; 7.5·1011 °C/min)
represented fully water-free systems. Hence, by simulating with both 0
and 3 wt% water, it is more straightforward to compare simulations and
experiments.
The error bars in the simulated Tg data reflect the sensitivity on
where to place exactly the broken sticks along the V-T data. Notice the

Table 1
Measured and simulated densities (ρ) at 20 °C.a

Glycerol ρ (kg/m3) E1 ρ S3-W0 Differenceb ρ Differencec

wt% samples (kg/m3) % S3-W3 %
(kg/m3)

0 1455 ± 2 1420 2 1393 4

10 1490 ± 2 1416 5 1424 4
20 1456 ± 10 1394 4 1405 4
30 1427 ± 7 1399 2 1393 2
40 1397 ± 4 1380 2 1352 3
50 1347 ± 14 1350 b1 1324 2
100 1261 – – – –
a
The value for anhydrous glycerol is taken from ref. [65].
b
The difference between experimental and simulated (without water) density.
c
The difference between experimental and simulated (with 3 wt% water) density. Fig. 3. Specific volume as a function of temperature at 1 atm for pure starch.
6 H.D. Özeren et al. / Materials and Design 187 (2020) 108387

Fig. 4. Experimental (E1, E3) and simulated (S3-W0, S3-W3) glass transition temperatures (left). S3-W0 is the dry system and S3-W3 contains 3 wt% water. DMA results of the E3 samples
(right).

[71] the difference in Tg from simulation and experiment was estimated molecules, which will be elaborated further in a section below. Before
to be ~80 °C. This is on the same order of magnitude as observed be- the transition (at lower glycerol content), the samples were brittle to
tween the simulated and experimental glass transition temperatures the extent that it was impossible to run DMA on these, whereas above
for the more flexible systems (30–50 wt% glycerol) (Fig. 4a). The reason the transition, samples were ductile and tough and could be tested
for the observed similar Tg's, despite the very different cooling rates, for with DMA. Fig. 5b shows that the DMA-determined glass transitions
the stiffer systems (glycerol content ≤20 wt%), is due to the actual molar (considered here to correspond to the peak in tan (δ)) were similar to
mass of the simulated polymer. The glass transition temperature de- those obtained by DSC, i.e., with values at −50 to −60 °C.
creases non-linearly with decreasing molar mass (effects stronger at
lower molar mass) [72]. Based on experimental data on starch and 4.3. Mechanical properties
maltodextrins [73], it is expected that the molar mass used here
(9098 g/mol) will reduce significantly the simulated Tg. Noteworthy, In Fig. 5, the stiffness (Young's modulus) and strength (maximum
along with our findings here, is that the effects of molar mass (concen- stress) are shown as a function of glycerol content. The samples with a
tration of chain ends) decreases with increasing plasticization, as ob- glycerol content below 20 wt% were too brittle to be tensile tested
served experimentally on water plasticized starch and maltodextrins whereas the thin amorphous sample (E1) with 50 wt% glycerol was
[73]. Hence, the effect of chain ends decreases with increasing overall too tacky to be tensile tested. However, as with the glass transition tem-
mobility of the system (increasing plasticizer content). The molar perature, the largest change in the experimentally determined stiffness
mass in the different simulated systems referenced above (polystyrene, occurred in the 20–30 wt% glycerol region (Fig. 5a). The semi-crystalline
polybutadiene and polyisobutylene) were also low material (E3) was stiffer than the amorphous material (E1) and the
(8000–10,000 g/mol) [68]. presence of water reduced the simulated modulus. In the water free-
Of importance, however, is that the simulations (Fig. 4a) can capture system, the modulus simulation captured an experimentally observed
both the experimentally observed stronger decrease in Tg between 20 modulus transition occurring between 20 and 30 wt%. The simulated
and 30 wt% plasticizer and the leveling off in Tg beyond 30 wt% plasti- moduli were higher than the experimental values due to the unavoid-
cizer. This behavior has also been reported before, even with films con- able significantly higher tensile strain rates in the simulations, which
taining water [17,18]. The simulated decrease in Tg between 20 and is in agreement with the results from other MD studies on plasticized
30 wt% plasticizer was however more moderate. We hypothesize that polymers [74]. Similarly, the simulated strength was significantly higher
the “transition” in the degree of plasticization (decrease in Tg) in this re- than the corresponding experimental strength (Fig. 5b). The more
gion is associated with the onset of percolation of plasticizer/glycerol glassy behavior at high strain rates, even for plasticized systems, has

Fig. 5. Young's modulus (left) and maximum stress (right) as a function of glycerol content from experiments and simulations.
 H.D. Özeren et al. / Materials and Design 187 (2020) 108387 7

also been observed experimentally [75]. The strength decreased in the The percolation threshold of a well dispersed, isotropic mixture with
presence of water (compare S10-W0 and S10-W3) and the semi- isotropically oriented (inert) ellipsoidal fillers can be determined as a
crystalline material (E3) was generally stronger than the amorphous function of aspect ratio (=length/width) directly from tabulated data
material (E1). The transition in properties between 20 and 30 wt% glyc- [79,80]. The average glycerol aspect ratio was approximately 2, both
erol was less apparent in the strength data than in the Tg data. Notewor- for energy minimized glycerol (Fig. 1b) and for glycerol in MD cells
thy is that while the Tg values level off and decrease only little beyond (Fig. S12a). This predicts a percolation threshold at 25–26 vol% [79,80]
30 wt% plasticizer, the stiffness and strength show a large further de- when the glycerol molecule is represented as a prolate ellipsoid. Since
crease, both in the experiments and in the simulations. the mechanical MD simulations showed no anisotropy and since a
good dispersion was observed (Figs. 6 and S12b), it can be concluded
4.4. The transition region that the percolation threshold most probably coincided with the exper-
imentally observed transition region. Even though this correlation does
The glycerol plasticization efficiency increased distinctly in a transi- not guarantee a definite link, the observation still significantly
tion region between a glycerol content of 20 and 30 wt% (weight frac- strengthens our hypothesis that efficient plasticization occurs when a
tion: ϕwt = 0.2–0.3), corresponding to a glycerol volume content of percolated path of plasticizer molecules develops in the polymer matrix.
22–33 vol% (volume fraction: ϕvol = 0.22–0.33), neglecting the pres- Accordingly, the results indicate that if the plasticizer dispersion is good
ence of water. The calculation was done using the rule of mixture ϕvol and the plasticizer molecular geometry is properly defined, the transi-
= ρstarch/(ρstarch + (1/ϕwt-1) ρglycerol), where ρstarch and ρglycerol are tion region should be possible to identify also for other plasticizer-
the densities of starch and glycerol, respectively (Table 1). There is a polymer systems. The formation of percolated glycerol paths is visual-
wealth of studies in literature showing that a content above 20 wt% ized in Fig. 6, showing simulation snapshots with 3 starch molecules,
plasticizer is needed for effective plasticization of biopolymers, and we 3 wt% amylopectin, 0% water and 20–50 wt% glycerol. Percolated
refer here, as an example, to a study on similar starch-glycerol systems paths of glycerol can be observed from 30 wt% and above. It should be
as ours, where a similar transition was observed experimentally [76]. noted that the 50 wt% glycerol system shown in Fig. 6, according to
We hypothesize that this transition occurs when a percolated path of Fig. S12a, tended to be somewhat less well dispersed as the other sys-
plasticizer molecules develops throughout the structure, in analogy tems, indicating some phase separation towards very large plasticizer
with the theories for electrical percolation in composites with conduc- contents.
tive fillers [77,78]. This hypothesis does not contradict the existing lu-
bricity, gel, free volume and mechanistic theories of plasticization; it 4.5. Glycerol diffusion
rather complements them by predicting the required amount of plasti-
cizer needed for effective plasticization [20]. Percolation may here refer Migration of plasticizers should generally be minimized, because it
to as the onset of cooperative motions of local plasticized regions will deteriorate the long-term properties of the material and eventually
throughout the structure. cause leakage of unhealthy chemicals. Therefore, it is important to

a b

c d

Fig. 6. MD simulation cell for simulated glycerol-starch systems with 0% water and 20–50 wt% glycerol. Glycerol is in red and starch is in gray. (a) 20, (b) 30, (c) 40 and (d) 50 wt% glycerol.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
8 H.D. Özeren et al. / Materials and Design 187 (2020) 108387

predict the migration of plasticizers in order to minimize these unde-

sired effects.
The glycerol diffusivity was determined by fitting the MD generated
MSD curves with Eq. 2. For the 2 ns diffusion simulations, only the first
1.7 ns were used in the fitting, to avoid the poor statistics at the end of
the MSD curves. Representative MSD curves with corresponding fits
are presented in Fig. 7 and the other MSD curves are given in Fig. S13.
As observed in Figs. 8 and S13 and also in Table 2, the long-term glyc-
erol random-walk diffusivity (Dmacro) increased overall with increasing
glycerol content. Note that also in the diffusivity data a threshold/tran-
sition was observed in the 20–30 wt% glycerol region, as observed in e.g.
the PVT data. The same trend was observed both for completely dry sys-
tems and for systems containing 3 wt% water, although the absolute dif-
fusivity values were, as expected, higher for the latter. The initial clear
non-linear part of the MSD curves indicated that the glycerol (on a
short time-scale) moved in a confined way in the starch matrix
(Figs. 8 and S13) [68,81]. As comparison, the MSD for a 100% glycerol
system was a fully straight line (not shown). The short-term confined
Fig. 8. MD-simulated Dmacro as a function of glycerol content with (triangles) and without
glycerol diffusivity (Dmicro) was higher than the random-walk diffusiv-
(squares) 3 wt% water.
ity (Dmacro). Both the confined diffusivity and the confined domain
size (obtained by monitoring the motion of the central glycerol carbon),
considering the full range in glycerol content, increased with both in- assessing the overall plasticizer efficiency and how it depends on the
creasing glycerol and water content. Hence, also the confined glycerol plasticizer loading, even in complex systems with strong inter- and in-
diffusivity was affected by the overall molecular mobility in the system, tramolecular interactions, such as the present starch-glycerol system.
but did not seem to correlate with the specific changes in the system at The methodology used here can probably be applied to other complex
the transition region; possibly due to the local character of the confined and less complex polymer-plasticizer systems (biobased (e.g. polyesters
process (percolation occurs over larger distances). like PLA) and also petroleum-based (e.g. polyamide) plasticized poly-
As demonstrated here with simulations, it is relatively easy and mers). In addition, from the same simulations, it is possible to estimate
straightforward to determine the risk of substantial migration. It can the plasticizer diffusion properties as a function of plasticizer content,
be assessed rapidly for essentially any environment, temperature, rela- which is important information when determining the risk of plasticizer
tive humidity etc. This enables a “safer” choice of plasticizer for a given migration. The simulation of dry and slightly moisturized systems
polymer in any current application, and also enables new applications seemed to be a good strategy to account for the fact that even dried
where the risk of migration has to be determined beforehand and min- polar materials may take up some moisture during certain activities,
imized (e.g. in children toys [82] and medical products [83]). such as mechanical measurements. The future work will include inves-
tigating whether molecular dynamics simulations can be used to also
5. Conclusions rank different molecules for their plasticization efficiency. Polar
polymer–plasticizer systems will also be investigated at higher mois-
It was shown in this study that with molecular dynamics simulations ture contents (higher relative humidity) than those investigated here.
it was possible to capture the transition in the degree of plasticization It has been suggested that part of the plasticization efficiency of a
that occurred between a 20 to 30 wt% plasticizer, as observed experi- sugar like glycerol is to attract moisture. The present work showed
mentally by a large drop in glass transition temperature and associated that the plasticization with glycerol occurred readily when there was
changes in mechanical properties. This indicates that molecular dynam- low or no water present, even though the simulations showed a consis-
ics simulations have the potential to be used as a prediction tool in tently lower glass transition temperature, modulus and strength in the
presence of 3 wt% water.
To conclude, the approach used here in assessing the details of the
plasticizer action and the prediction of plasticizer efficiency has the po-
tential to be a valuable instrument for industry to choose the best plas-
ticizer candidate for a given polymer and product and also for rapid
screening of new designed potential plasticizers before they are actually
synthesized. It also allows to assess the risk of migration and to optimize
plasticizer content in plastic products for current and new applications.
This is especially important nowadays when there is an urgent need to

Table 2
Obtained diffusivities and L using eqs. 2 and 4.

System Dmicroa Dmacroa L (Å)

S3-GLY10-W0 7.70 ± 0.32 0.62 ± 0.00 1.14 ± 0.03

S3-GLY20-W0 6.72 ± 1.00 0.63 ± 0.01 1.34 ± 0.08
S3-GLY30-W0 6.34 ± 0.12 1.00 ± 0.02 1.19 ± 0.00
S3-GLY40-W0 7.99 ± 0.94 1.24 ± 0.06 1.43 ± 0.10
S3-GLY50-W0 9.79 ± 0.18 1.70 ± 0.05 1.94 ± 0.13
S3-GLY10-W3 9.11 ± 0.04 0.71 ± 0.01 1.24 ± 0.02
S3-GLY20-W3 9.32 ± 0.59 0.70 ± 0.01 1.17 ± 0.05
S3-GLY30-W3 8.31 ± 0.38 1.45 ± 0.16 1.26 ± 0.02
S3-GLY40-W3 8.14 ± 0.26 2.59 ± 0.11 1.92 ± 0.01
S3-GLY50-W3 12.11 ± 0.01 3.19 ± 0.04 1.84 ± 0.07
Fig. 7. MD simulated mean square displacement (MSD) of glycerol in the starch matrix
with 3 wt% water and 20 and 50 wt% glycerol. The data was fitted with Eq. 2. a
Unit of Dmicro and Dmacro is 10−8 cm2 s−1.
 H.D. Özeren et al. / Materials and Design 187 (2020) 108387 9

replace oil-based polymers with biobased polymers, which often shows [8] A.J.F. De Carvalho, A.A.S. Curvelo, J.A.M. Agnelli, Wood pulp reinforced thermoplastic
starch composites, Int. J. Polym. Mater. Polym. Biomater. 51 (7) (2002) 647–660,
more brittle behavior. https://doi.org/10.1080/714975803.
It should be mentioned that plasticization of starch, even with an op- [9] B. Chen, J.R.G. Evans, Thermoplastic starch–clay nanocomposites and their charac-
timized plasticizer system (low degree of migration and a ductility to- teristics, Carbohydr. Polym. 61 (4) (2005) 455–463, https://doi.org/10.1016/j.
carbpol.2005.06.020.
wards that of polyethylene), does not solve its high water sensitivity
[10] S. Mali, L.S. Sakanaka, F. Yamashita, M.V.E. Grossmann, Water sorption and mechan-
[16–19] and limits the applications to low-humidity environments or ical properties of cassava starch films and their relation to plasticizing effect,
single-use/edible packaging. However, the decrease in strength associ- Carbohydr. Polym. 60 (3) (2005) 283–289, https://doi.org/10.1016/j.carbpol.2005.
ated with plasticization can, at least partly, be compensated for by the 01.003.
[11] Y. Nakamura, S. Miyachi, Effect of temperature on starch degradation in Chlorella
addition of fillers (e.g. nanosized clay [17]).
vulgaris 11h cells, Plant Cell. Phys. 23 (2) (1982) 333–341, https://doi.org/10.
1093/oxfordjournals.pcp.a076354.
Data availability [12] Y.C. Zhang, C. Rempel, Q. Liu, Thermoplastic starch processing and characteristics-a
review, Crit. Rev. Food Sci. Nutr. 54 (10) (2014) 1353–1370, https://doi.org/10.
1080/10408398.2011.636156.
The data that support the findings of this study are available from [13] R. Mohamed, N.M.N. Nurazzi, M.I.S. Aisyah, F.M. Fauzi, Swelling and tensile proper-
the corresponding author upon reasonable request. ties of starch glycerol system with various crosslinking agents, Innov. Polym. Sci.
Tech. 223 (2017) https://doi.org/10.1088/1757-899x/223/1/012059.
[14] S. Mali, M.V.E. Grossmann, M.A. Garcia, M.N. Martino, N.E. Zaritzky, Effects of con-
Declaration of competing interests
trolled storage on thermal, mechanical and barrier properties of plasticized films
from different starch sources, J. Food Eng. 75 (4) (2006) 453–460, https://doi.org/
The authors declare that they have no known competing financial 10.1016/j.jfoodeng.2005.04.031.
[15] J. Viguie, S. Molina-Boisseau, A. Dufresne, Processing and characterization of waxy
interests or personal relationships that could have appeared to influ-
maize starch films plasticized by sorbitol and reinforced with starch nanocrystals,
ence the work reported in this paper. Macromol. Biosci. 7 (11) (2007) 1206–1216, https://doi.org/10.1002/mabi.
CRediT Author Statement 200700136.
Hüsamettin D. Özeren: Conceptualization, Methodology, Software, [16] M.I.J. Ibrahim, S.M. Sapuan, E.S. Zainudin, M.Y.M. Zuhri, Physical, thermal, morpho-
logical, and tensile properties of cornstarch-based films as affected by different plas-
Investigation, Writing - original draft. ticizers, Int. J. Food Prop. 22 (1) (2019) 925–941, https://doi.org/10.1080/10942912.
Richard T. Olsson: Formal analysis, Resources. 2019.1618324.
Fritjof Nilsson: Validation, Formal analysis, Writing - review & [17] A. Heydari, I. Alemzadeh, M. Vossoughi, Functional properties of biodegradable corn
starch nanocomposites for food packaging applications, Mater Design 50 (2013)
editing. 954–961, https://doi.org/10.1016/j.matdes.2013.03.084.
Mikael S. Hedenqvist: Supervision, Funding acquisition, Writing - [18] E. Basiak, A. Lenart, F. Debeaufort, How glycerol and water contents affect the struc-
review & editing. tural and functional properties of starch-based edible films, Polymers-Basel 10 (4)
(2018). https://www.doi.org/ARTN412. https://doi.org/10.3390/polym10040412.
[19] M. Esmaeili, G. Pircheraghi, R. Bagheri, Optimizing the mechanical and physical
CRediT authorship contribution statement properties of thermoplastic starch via tuning the molecular microstructure through
co-plasticization by sorbitol and glycerol, Polym. Int. 66 (6) (2017) 809–819,
https://doi.org/10.1002/pi.5319.
Hüsamettin D. Özeren:Conceptualization, Methodology, Software, [20] M.L. Sanyang, S.M. Sapuan, M. Jawaid, M.R. Ishak, J. Sahari, Effect of plasticizer type
Investigation, Writing - original draft.Richard T. Olsson:Formal analy- and concentration on tensile, thermal and barrier properties of biodegradable films
sis, Resources.Fritjof Nilsson:Validation, Formal analysis, Writing - re- based on sugar palm (Arenga pinnata) starch, Polymers-Basel 7 (6) (2015)
1106–1124, https://doi.org/10.3390/polym7061106.
view & editing.Mikael S. Hedenqvist:Supervision, Funding acquisition, [21] B. Adhikari, D.S. Chaudhary, E. Clerfeuille, Effect of plasticizers on the moisture mi-
Writing - review & editing. gration behavior of low-amylose starch films during drying, Dry. Technol. 28 (4)
(2010) 468–480. https://www.doi.org/Pii921711782. https://doi.org/10.1080/
07373931003613593.
Ackowledgements [22] A.K. Doolittle, The Technology of Solvents and Plasticizers, Wiley, New York, 1955.
[23] A.K. Doolittle, Mechanism of plasticization, Plasticizer Technology, Reinhold, New
The work was financed by the Swedish Research Council (VR), pro- York, 1965.
[24] K. Sears, Salvation and plasticization, in: C.A.H. Leonard, I. Nass (Eds.), Encyclopedia
ject number 2016-04453. Carolin Menzel is thanked for providing the
of PVC: Resin Manufacture and Properties, CRC Press 1986, pp. 440–448.
molecular size distribution data. [25] H.G. Moneypenny, General compounding, in: B. Rodgers (Ed.), Rubber
Compounding: Chemistry and Applications, CRC Press 2004, pp. 394–403.
Appendix A. Supplementary data [26] A.D. Godwin, Plasticizers, in: M. Kutz (Ed.), Applied Plastics Engineering Handbook,
2ndWilliam Andrew 2016, pp. 487–503.
[27] A. Kirkpatrick, Some relations between molecular structure and plasticizing effect, J.
Supplementary data to this article can be found online at https://doi. Appl. Phys. 11 (4) (1940) 255–261, https://doi.org/10.1063/1.1712768.
org/10.1016/j.matdes.2019.108387. [28] F.W. Clark, Chem. Ind. 60 (1941) 225.
[29] R. Houwink, Proc. XI Int. Cong. Pure Appl. Chem, London, 1947 575–583.
[30] W. Aiken, T. Alfrey Jr., A. Janssen, H. Mark, Creep behavior of plasticized vinylite
References VYNW, J. Polym. Sci. 2 (2) (1947) 178–198, https://doi.org/10.1002/pol.1947.
120020206.
[1] A. Jiménez, M.J. Fabra, P. Talens, A. Chiralt, Edible and biodegradable starch films: a [31] D.F. Cadogan, C.J. Howick, Plasticizers, Ullmann's Encyclopedia of Industrial Chemis-
review, Food Bioprocess Technol. 5 (6) (2012) 2058–2076, https://doi.org/10.1007/ try 2000, p. 605.
s11947-012-0835-4. [32] A. Marcilla, M. Beltran, Mechanism and plasticizer action, in: G. Wypych (Ed.),
[2] S.A. Attaran, A. Hassan, M.U. Wahit, Materials for food packaging applications based Handbook of Plasticizers, Third editionChemTec Publishing 2017, pp. 119–134.
on bio-based polymer nanocomposites: a review, J. Thermoplast. Compos. Mater. 30 [33] M.L. Williams, R.F. Landel, J.D. Ferry, The temperature dependence of relaxation
(2) (2017) 143–173, https://doi.org/10.1177/0892705715588801. mechanisms in amorphous polymers and other glass-forming liquids, J. Am.
[3] T.N. Prabhu, K. Prashantha, A review on present status and future challenges of Chem. Soc. 77 (14) (1955) 3701–3707, https://doi.org/10.1021/ja01619a008.
starch based polymer films and their composites in food packaging applications, [34] K.A. Mauritz, R.F. Storey, B.S. Wilson, Efficiency of plasticization of PVC by higher-
Polym. Compos. 39 (7) (2018) 2499–2522, https://doi.org/10.1002/pc.24236. order di-alkyl phthalates and survey of mathematical models for prediction of poly-
[4] R.P. Babu, K. O'Connor, R. Seeram, Current progress on bio-based polymers and their mer/diluent blend Tg's, Journal of Vinyl Technology 12 (3) (1990) 165–173, https://
future trends, Prog. Biomater. 2 (1) (2013) 8, https://doi.org/10.1186/2194-0517-2- doi.org/10.1002/vnl.730120309.
8. [35] J.H. Yang, K.K. Tang, G.Q. Qin, Y.X. Chen, L. Peng, X. Wang, H.N. Xiao, Q.Y. Xia, Hydro-
[5] J. Muller, C. Gonzalez-Martinez, A. Chiralt, Combination of poly(lactic) acid and gen bonding energy determined by molecular dynamics simulation and correlation
starch for biodegradable food packaging, Materials 10 (8) (2017) https://doi.org/ to properties of thermoplastic starch films, Carbohydr. Polym. 166 (2017) 256–263,
10.3390/ma10080952. https://doi.org/10.1016/j.carbpol.2017.03.001.
[6] F.J. Rodriguez-Gonzalez, B.A. Ramsay, B.D. Favis, High performance LDPE/thermo- [36] H.M. Rietveld, A profile refinement method for nuclear and magnetic structures, J.
plastic starch blends: a sustainable alternative to pure polyethylene, Polymer 44 Appl. Crystallogr. 2 (2) (1969) 65–71, https://doi.org/10.1107/
(5) (2003) 1517–1526, https://doi.org/10.1016/S0032-3861(02)00907-2. S0021889869006558.
[7] M. Kaseem, K. Hamad, F. Deri, Preparation and studying properties of polybutene-1/ [37] P.H. Hermans, A. Weidinger, Quantitative X-ray investigations on the crystallinity of
thermoplastic starch blends, J. Appl. Polym. Sci. 124 (4) (2012) 3092–3098, https:// cellulose fibers. A background analysis, J. Appl. Phys. 19 (5) (1948) 491–506, https://
doi.org/10.1002/app.35350. doi.org/10.1063/1.1698162.
10 H.D. Özeren et al. / Materials and Design 187 (2020) 108387

[38] S. Plimpton, LAMMPS-large-scale atomic/molecular massively parallel simulator, [61] D. Lafargue, B. Pontoire, A. Buleon, J.L. Doublier, D. Lourdin, Structure and mechani-
Sandia National Laboratories, 2007. cal properties of hydroxypropylated starch films, Biomacromol 8 (12) (2007)
[39] S. Nosé, A unified formulation of the constant temperature molecular dynamics 3950–3958, https://doi.org/10.1021/bm7009637.
methods, J. Chem. Phys. 81 (1) (1984) 511–519, https://doi.org/10.1063/1.447334. [62] M.T. Kalichevsky, E.M. Jaroszkiewicz, S. Ablett, J.M.V. Blanshard, P.J. Lillford, The glass
[40] W.G. Hoover, Constant-pressure equations of motion, Phys Rev A Gen Phys 34 (3) transition of amylopectin measured by DSC, DMTA and NMR, Carbohydr. Polym. 18
(1986) 2499–2500, https://doi.org/10.1103/PhysRevA.34.2499. (2) (1992) 77–88, https://doi.org/10.1016/0144-8617(92)90129-E.
[41] S. Nosé, A molecular dynamics method for simulations in the canonical ensemble, [63] A. Rindlav-Westling, M. Stading, A.M. Hermansson, P. Gatenholm, Structure, me-
Mol. Phys. 52 (2) (1984) 255–268, https://doi.org/10.1080/00268978400101201. chanical and barrier properties of amylose and amylopectin films, Carbohydr.
[42] W.G. Hoover, Canonical dynamics: equilibrium phase-space distributions, Phys Rev Polym. 36 (2–3) (1998) 217–224, https://doi.org/10.1016/S0144-8617(98)00025-3.
A Gen Phys 31 (3) (1985) 1695–1697, https://doi.org/10.1103/PhysRevA.31.1695. [64] M. Roussenova, M. Murith, A. Alam, J. Ubbink, Plasticization, antiplasticization, and
[43] P.P. Ewald, Die Berechnung optischer und elektrostatischer Gitterpotentiale, Anna. molecular packing in amorphous carbohydrate-glycerol matrices, Biomacromol 11
Phys. 369 (3) (1921) 253–287, https://doi.org/10.1002/andp.19213690304. (12) (2010) 3237–3247, https://doi.org/10.1021/bm1005068.
[44] G.S. Larsen, P. Lin, K.E. Hart, C.M. Colina, Molecular simulations of pim-1-like poly- [65] N.-S. Cheng, Formula for the viscosity of a glycerol−water mixture, Ind. Eng. Chem.
mers of intrinsic microporosity, Macromolecules 44 (17) (2011) 6944–6951, Res. 47 (9) (2008) 3285–3288, https://doi.org/10.1021/ie071349z.
https://doi.org/10.1021/ma200345v. [66] D. Benczedi, I. Tomka, F. Escher, Thermodynamics of amorphous starch-water sys-
[45] R. Steane, http://www.biotopics.co.uk/jsmol/amylose2.html#, Access Date: 10-15- tems. 1. Volume fluctuations, Macromolecules 31 (9) (1998) 3055–3061, https://
2018. doi.org/10.1021/ma970143e.
[46] B.R. Brooks, R.E. Bruccoleri, B.D. Olafson, D.J. States, S. Swaminathan, M. Karplus, [67] D. Benczedi, I. Tomka, C. Panayiotou, Volumetric properties of starch-water mix-
CHARMM: a program for macromolecular energy, minimization, and dynamics cal- tures, Fluid Phase Equilib. 138 (1–2) (1997) 145–158, https://doi.org/10.1016/
culations, J. Comput. Chem. 4 (2) (1983) 187–217, https://doi.org/10.1002/jcc. S0378-3812(97)00165-9.
540040211. [68] J. Han, R.H. Gee, R.H. Boyd, Glass transition temperatures of polymers from molecu-
[47] O. Guvench, S.S. Mallajosyula, E.P. Raman, E. Hatcher, K. Vanommeslaeghe, T.J. lar dynamics simulations, Macromolecules 27 (1994) 7781–7784, https://doi.org/
Foster, F.W. Jamison, A.D. MacKerell, CHARMM additive all-atom force field for car- 10.1021/ma00104a036.
bohydrate derivatives and its utility in polysaccharide and carbohydrate-protein [69] N.J. Soni, P.H. Lin, R. Khare, Effect of cross-linker length on the thermal and volumet-
modeling, J. Chem. Theory Comput. 7 (10) (2011) 3162–3180, https://doi.org/10. ric properties of cross-linked epoxy networks: a molecular simulation study, Poly-
1021/ct200328p. mer 53 (4) (2012) 1015–1019, https://doi.org/10.1016/j.polymer.2011.12.051.
[48] H.A. Lorentz, Ueber die Anwendung des Satzes vom Virial in der kinetischen Theorie [70] A. Soldera, N. Metatla, Glass transition of polymers: atomistic simulation versus ex-
der Gase, Anna. Phys. 248 (1) (1881) 127–136, https://doi.org/10.1002/andp. periments, Phys. Rev. E 74 (6) (2006) https://doi.org/10.1103/PhysRevE.74.061803.
18812480110. [71] Y.X. Zhou, S.T. Milner, Average and local T-g shifts of plasticized PVC from simula-
[49] D.A. Jahn, F.O. Akinkunmi, N. Giovambattista, Effects of temperature on the proper- tions, Macromolecules 51 (10) (2018) 3865–3873, https://doi.org/10.1021/acs.
ties of glycerol: a computer simulation study of five different force fields, J. Phys. macromol.8b00271.
Chem. B 118 (38) (2014) 11284–11294, https://doi.org/10.1021/jp5059098. [72] J.L. Barrat, J. Baschnagel, A. Lyulin, Molecular dynamics simulations of glassy poly-
[50] J. Blieck, F. Affouard, P. Bordat, A. Lerbret, M. Descamps, Molecular dynamics simu- mers, Soft Matter 6 (15) (2010) 3430–3446, https://doi.org/10.1039/b927044b.
lations of glycerol glass-forming liquid, Chem. Phys. 317 (2–3) (2005) 253–257, [73] R.G.M. van der Sman, M.B.J. Meinders, Prediction of the state diagram of starch
https://doi.org/10.1016/j.chemphys.2005.05.045. water mixtures using the Flory-Huggins free volume theory, Soft Matter 7 (2)
[51] R. Chelli, P. Procacci, G. Cardini, R.G. Della Valle, S. Califano, Glycerol condensed (2011) 429–442, https://doi.org/10.1039/c0sm00280a.
phases part I. A molecular dynamics study, PCCP 1 (5) (1999) 871–877, https:// [74] H.D. Özeren, M. Balçık, M.G. Ahunbay, J.R. Elliott, In silico screening of green plasti-
doi.org/10.1039/a808958b. cizers for poly(vinyl chloride), Macromolecules (2019) https://doi.org/10.1021/acs.
[52] F.O. Akinkunmi, D.A. Jahn, N. Giovambattista, Effects of temperature on the thermo- macromol.8b02154.
dynamic and dynamical properties of glycerol-water mixtures: a computer simula- [75] M.J. Kendall, C.R.A. Siviour, Rate dependence of poly (vinyl chloride), the effects of
tion study of three different force fields, J. Phys. Chem. B 119 (20) (2015) plasticizer and time − temperature superposition rate dependence of poly (vinyl
6250–6261, https://doi.org/10.1021/acs.jpcb.5b00439. chloride), the effects of plasticizer and time – temperature superposition, Proceed-
[53] H.W. Horn, W.C. Swope, J.W. Pitera, J.D. Madura, T.J. Dick, G.L. Hura, T. Head-Gordon, ings of the Royal Society A: Mathematical, Physical and Engineering Sciences
Development of an improved four-site water model for biomolecular simulations: (April) (2014).
TIP4P-Ew, J. Chem. Phys. 120 (20) (2004) 9665–9678, https://doi.org/10.1063/1. [76] P. Myllärinen, R. Partanen, J. Seppälä, P. Forssell, Effect of glycerol on behaviour of
1683075. amylose and amylopectin films, Carbohydr. Polym. 50 (4) (2002) 355–361,
[54] P.J. Steinbach, B.R. Brooks, New spherical-cutoff methods for long-range forces in https://doi.org/10.1016/S0144-8617(02)00042-5.
macromolecular simulation, J. Comput. Chem. 15 (7) (1994) 667–683, https://doi. [77] M. Wahlander, F. Nilsson, R.L. Andersson, C.C. Sanchez, N. Taylor, A. Carlmark, H.
org/10.1002/jcc.540150702. Hillborg, E. Malmstrom, Tailoring dielectric properties using designed polymer-
[55] D. Hossain, M.A. Tschopp, D.K. Ward, J.L. Bouvard, P. Wang, M.F. Horstemeyer, Mo- grafted ZnO nanoparticles in silicone rubber, J. Mater. Chem. A 5 (27) (2017)
lecular dynamics simulations of deformation mechanisms of amorphous polyethyl- 14241–14258, https://doi.org/10.1039/c6ta11237d.
ene, Polymer 51 (25) (2010) 6071–6083, https://doi.org/10.1016/j.polymer.2010. [78] F. Nilsson, J. Kruckel, D.W. Schubert, F. Chen, M. Unge, U.W. Gedde, M.S. Hedenqvist,
10.009. Simulating the effective electric conductivity of polymer composites with high as-
[56] F. Daumas, N. Destainville, C. Millot, A. Lopez, D. Dean, L. Salome, Confined diffusion pect ratio fillers, Compos. Sci. Technol. 132 (2016) 16–23, https://doi.org/10.1016/
without fences of a G-protein-coupled receptor as revealed by single particle track- j.compscitech.2016.06.008.
ing, Biophys. J. 84 (1) (2003) 356–366, https://doi.org/10.1016/S0006-3495(03) [79] E.J. Garboczi, K.A. Snyder, J.F. Douglas, M.F. Thorpe, Geometrical percolation-
74856-5. threshold of overlapping ellipsoids, Phys. Rev. E 52 (1) (1995) 819–828, https://
[57] Y.H. Zhang, A.V. Tzingounis, G. Lykotrafitis, Modeling of the axon plasma membrane doi.org/10.1103/PhysRevE.52.819.
structure and its effects on protein diffusion, PLoS Comput. Biol. 15 (5) (2019), [80] Y.B. Yi, A.M. Sastry, Analytical approximation of the percolation threshold for over-
e1007003. https://doi.org/10.1371/journal.pcbi.1007003. lapping ellipsoids of revolution, P Roy Soc a-Math Phy 460 (2048) (2004)
[58] T. Giorgino, Computing diffusion coefficients in macromolecular simulations: the 2353–2380, https://doi.org/10.1098/rspa.2004.1279.
diffusion coefficient tool for VMD, Journal of Open Source Software 4 (41) (2019) [81] G.E. Karlsson, U.W. Gedde, M.S. Hedenqvist, Molecular dynamics simulation of oxy-
1698, https://doi.org/10.21105/joss.01698. gen diffusion in dry and water-containing poly(vinyl alcohol), Polymer 45 (11)
[59] Q.X. Zhang, Z.Z. Yu, X.L. Xie, K. Naito, Y. Kagawa, Preparation and crystalline mor- (2004) 3893–3900, https://doi.org/10.1016/j.polymer.2003.12.082.
phology of biodegradable starch/clay nanocomposites, Polymer 48 (24) (2007) [82] P.J. Halley, R.W. Truss, M.G. Markotsis, C. Chaleat, M. Russo, A.L. Sargent, I. Tan, P.A.
7193–7200, https://doi.org/10.1016/j.polymer.2007.09.051. Sopade, A Review of Biodegradable Thermoplastic Starch Polymers, Polymer Dura-
[60] R. Vermeylen, B. Goderis, H. Reynaers, J.A. Delcour, Amylopectin molecular structure bility and Radiation Effects, American Chemical Society, 2007 287–300.
reflected in macromolecular organization of granular starch, Biomacromol 5 (5) [83] A.J.F. de Carvalho, E. Trovatti, Biomedical applications for thermoplastic starch,
(2004) 1775–1786, https://doi.org/10.1021/bm0499132. Biodeg. Bio. Polym. Env. Biomed. Appl, Wiley 2016, pp. 1–23.