Soft Matter

View Article Online

PAPER View Journal | View Issue
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

Origin of the low-frequency plateau and the light-

scattering slow mode in semidilute poly(ethylene
Cite this: Soft Matter, 2019,
15, 2666 glycol) solutions
Martha Koziol, Karl Fischer and Sebastian Seiffert *

A low-frequency plateau is often found in the rheological spectra of various kinds of semidilute
solutions of polymers and other colloids; also, many such solutions have been reported to show slow-
modes in their dynamic light scattering autocorrelation functions. Both these observations may lead to
the hypothesis of weak associative network structures built by the dissolved polymer chains or colloidal
building blocks. To challenge this hypothesis, we conduct a series of comparative studies on semidilute
solutions of poly(ethylene glycol) by using classical rheology as well as passive microrheology based on
dynamic light scattering, along with structural studies using static light scattering. Although we indeed
find a low-frequency plateau using classical shear rheology, even at elevated temperatures where
potential polymer aggregates should be broken, no such plateau is observed in any of our microrheology
experiments. Also, dynamic and static light scattering studies on the polymer solutions do not confirm the
presence of larger structural entities: no slow mode can be detected in the autocorrelation function of the
scattering intensity signal, and this signal is angle independent if the samples are purified by a thorough
Received 5th November 2018, procedure of filtration. Based on these findings, we conclude that the low-frequency plateau in classical
Accepted 23rd February 2019 rheology results is an instrument effect caused by erroneous recording of the phase angle, although the
DOI: 10.1039/c8sm02263a magnitude of the torque lies well within the resolution of the rheometer. We also conclude that slow
modes in dynamic light scattering on solutions of poly(ethylene glycol) are impurity-based artifacts rather
rsc.li/soft-matter-journal than due to actual associated structures.

Introduction (iv) the energy-elastic glassy region at frequencies higher

than the inverse shortest relaxation time.
The usual viscoelastic spectrum of homopolymer solutions However, in numerous publications, a bend-over of G 0 to a
in the semidilute concentration range consists of up to four second weak plateau in the low-frequency region has been
regimes, as shown in Fig. 1A: reported, as shown in Fig. 1B;1–8 furthermore, this finding is
(i) the terminal relaxation region at low frequencies (which not limited to polymer solutions, but has also been reported for
is phenomenologically described by the Maxwell model) other colloidal samples such as micellar systems.9,10 This low-
(ii) potentially an entropy-elastic plateau where the storage frequency plateau of the storage modulus can indicate the
modulus G 0 is frequency-independent and dominates over the presence of a weak network structure that is able to store elastic
loss modulus G00 at medium frequencies (provided that the energy. Such a network structure may be presumed to form as a
chains are long, overlapping, and entangled) consequence of a weak non-covalent association and aggrega-
(iii) a transition into the glassy region in the range of tion of the polymer chains in a sample at rest, whereas at a
frequencies between the inverse entanglement time (in the higher shear rate, as given at a higher frequency if the shear
case of entangled chains and presence of regime ii) or the amplitude is fixed, these structures might be destroyed such
inverse Rouse or Zimm time (in the case of unentangled chains that the normal flow and Maxwell-type scaling of the frequency-
and absence of regime ii) and the inverse shortest relaxation dependent storage and loss moduli are observed. Indeed, for
time (which is phenomenologically described by the Kelvin– aqueous solutions of the popular polymer poly(ethylene glycol)
Voigt model) (PEG), formation of helical aggregates due to hydrogen bonding
and due to the amphiphilic character of the polymer molecules
Johannes Gutenberg-Universität Mainz, Institute of Physical Chemistry,
has been reported.11 Alternatively, a weak network structure
Duesbergweg 10–14, D-55128 Mainz, Germany. may also be caused by small air bubbles embedded in the
E-mail: sebastian.seiffert@uni-mainz.de viscous polymer solution. Such air bubbles may even be

2666 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Paper Soft Matter

ability of the PEG chains to enclose hydrophobic impurities can

be considered as an origin for aggregate formation.13–15 All
these structures can be presumed to create a weak network that
may weakly store elastic energy, but that may also be destroy-
able by sufficiently strong shear forces, which will be present in
rheology at shear oscillation frequencies higher than a certain
lower limit at a given and fixed shear amplitude, thereby
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

leading to a transition from the low-frequency weak plateau

to a common Maxwellian flow regime, as shown in Fig. 1B.
The presumption of weak network structures and polymer
aggregates may further be seen to be in agreement with the
findings of various dynamic light scattering studies on polymer
solutions. Whereas the normalized autocorrelation function
obtained from these experiments should actually just show a
single-exponential decay that is addressable to the cooperative
diffusive motion of overlapping PEG chain segments, as shown
in Fig. 1C, many measurements show an additional slow mode,
as shown in Fig. 1D.16,17 In the light-scattering community, the
reasons for the possible existence of such slow modes have
been and still are a topic of debate for more than two decades.18
In recent publications, the slow mode has been attributed to
the appearance of large aggregates,14,19 air bubbles12 or
impurities,15 just as also addressed above as possible explana-
tions for a low-frequency plateau in the rheological spectra.
Alternatively, viscoelastic relaxation20 and relaxation of corre-
lated blobs of overlapping chain segments17 have been dis-
cussed as causes for a slow mode. However, a direct connection
between the weak low-frequency plateau in the storage modulus
G 0 in rheology and the appearance of a slow mode in dynamic
light scattering has not been reported so far.
In this paper, we challenge the hypothesis of the presence of
aggregates and clusters in semidilute PEG solutions and their
presumed effects on the rheological and light scattering proper-
ties of these solutions. We do this by systematic investigation
from a microscopic and a macroscopic perspective. For this
purpose, we use microrheology based on dynamic light scatter-
ing (DLS) as well as classical oscillatory shear macrorheology to
Fig. 1 Schematics of ideal and frequent actual oscillatory-shear rheo- examine the frequency-dependent viscous and elastic contribu-
logical spectra (panel A and B) and dynamic-light-scattering autocorrelation
tions of the rheological properties of the polymer solutions. We
curves (panel C and D) of semidilute polymer solutions. (A) Ideal viscoelastic
spectrum of a homopolymer solution in the semidilute entangled concen- choose three different temperatures (278 K, 293 K, and 323 K)
tration regime, with the frequency-dependent storage modulus G 0 (blue full to achieve variations in the external thermal energy. On top of
line) and loss modulus G00 (blue dashed line). (B) Actual viscoelastic spectrum that, we probe the dynamics and microstructure of the PEG
as often measured for many samples, displaying a bend-over of the elastic solutions by dynamic and static light scattering (DLS and SLS)
modulus G 0 (black full line) to a low-frequency plateau in the flow regime.
to detect potential aggregates or micro-phase separated domains
(C) Ideal monomodal autocorrelation function of a semidilute polymer
solution as obtained by dynamic light scattering (DLS). (D) Actual bimodal larger than 10 nm.
autocorrelation function with a typical slow mode as often measured for
many samples.

Experimental section
stabilized by surface-adsorbed amphiphilic PEG chains, Materials and sample preparation
thereby acting as weak network nodes. In an earlier study, The material basis for our investigations is narrowly dispersed
evidence for such bubbles has been found in dilute PEG PEG samples with weight averaged molar masses of 1860 g mol1
solutions,12 and in the semidilute regime, these may become (Carl Roth, denoted PEG-2k) and 23 540 g mol1 (Sigma Aldrich,
a cause for weak network formation. Furthermore, the forma- denoted PEG-20k); in addition, we partly use a PEG sample with
tion of large PEG chain clusters due to interactions between the a nominal molar mass of 100 000 g mol1 (Alfa Aesar, denoted
hydrophobic CH2–CH2 units has been reported, and also the PEG-100k). These polymers are first purified by dissolution in

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 2666--2676 | 2667
 View Article Online

Soft Matter Paper

dichloromethane and subsequently precipitated in cold diethyl aqueous solution is examined by DLS, yielding a value of
ether. After drying in a high vacuum, the PEG is dissolved in water, RH = 23.7 nm based on an angular-independent self-diffusion
filtered through a Whatman Anotop filter (25 mm, 0.02 mm pore coefficient of Ds = 9.16  108 cm2 s1. DLS-based microrheol-
size) and a Millex-LGs filter (200 nm pore size), and dried again. To ogy is then performed by loading the polymer solutions with
characterize the molar mass of the polymers and their dispersity, the gold–PEG core–shell nanoprobes at a low gold concen-
SEC measurements are carried out with a 1260 Infinity GPC/SEC- tration of 5 mg L1 to avoid multiple scattering and inter-
system (Agilent) equipped with an RI-detector and a Shodex pre- particle interactions. The microrheology samples are prepared
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

column, along with Shodex OHpak SB 804, Shodex OHpak SB by dissolving the polymer in water, adding the gold, homo-
803,and Shodex OHpak SB 802.5 columns, with an eluent mixture genization, freeze drying, and redissolution of the sample in
of water/acetonitrile (70 : 30) at a flow rate of 1 mL min1. SEC 2 mL of water to obtain a homogenous distribution of the
calibration is carried out with poly(ethylene glycol) standards from nanoparticles in the viscous polymer solutions. All light scatter-
PSS, Germany. We determined the polydispersity index (PDI) of ing quartz glass cuvettes are thoroughly pre-cleaned with
1.03 (PEG-2k) and 1.07 (PEG 20-k). Aqueous solutions of the peroxymonosulfuric acid and aqua regia, and dust particles
poly(ethylene glycol)s with different concentrations below, at, and are removed by rinsing with hot acetone. The polymer solutions
above the overlap concentration (100, 200, 300, and 400 g L1 are then poured into clean cuvettes by slow filtration through
for PEG-2k; 35, 70, 150, and 200 g L1 for PEG-20k; 200 g L1 for syringe filters (Millex-LGs, 200 nm pore size and Whatman
PEG-100k) are prepared by dissolving the respective amount of PEG Anotop, 25 mm and 20 nm pore size). This filtration step is
in deionized water. The solutions are gently shaken at 40 1C for performed in a dust-free laminar flow box to obtain dust-free
three days to allow for complete dissolution of the polymer. The polymer solutions. Dynamic light scattering measurements are
overlap concentration c* is calculated as c* = (3Mw)/(4pNARG3) E performed on a Multi goniometer ALV-CGS-8F SLS/DLS 5022F
[Z]1,21 with the polymer molar mass Mw, the Avogadro number NA, with a Uniphase He/Ne Laser (25 mW, l = 632.8 nm) and a
the polymer-coil radius of gyration RG, and the intrinsic viscosity vertically polarized laser beam, with eight simultaneously working
[Z] = 12.5  103 Mw0.78.22 We calculated radii of gyration of ALV-7004 multi-tau correlators at eight fiber optical ALV/High QE
1.62 nm for PEG-2k, 5.13 nm for PEG-20k, and 11.46 nm for APD avalanche photodiodes (all components: ALV-Laser
PEG-100k, using the formula hRG2i1/2 = 61/2  8.88  102 Mw1/2.23 vertriebsgesellschaft mbH, Langen, Germany). A constant tem-
With these values, the overlap concentration c* was determined to perature of 20 1C for all measurements is assured (0.1 1C) by a
be 200 g L1 (PEG-2k), 35 g L1 (PEG-20k), and 11 g L1 (PEG-100k). Lauda ultra thermostat RKS C6 (Lauda Dr Worbser GmbH,
As a further characteristic quantity, the semidilute-solution correla- Königshofen, Germany). A Glan–Thompson polarizer with
tion length x, which can be seen as a mesh size of the transient vertical polarization (VV-mode) is installed in front of the detector
solution-network, is appraised as x = hRG2i1/2(c*/c)0.75 with the to block any light depolarized by the gold nanoparticles.26 The
polymer concentration c.24 For microrheology, probe colloids with measured and normalized intensity autocorrelation function,
a core–shell structure are used. The cores are spherical gold
hI ðq; tÞi  hIðq; t þ tÞi
nanoparticles with a diameter of 15 nm purchased from Nano- gð2Þ ðq; tÞ ¼ (1)
hIðq; tÞi2
partzt, whereas the shell is a PEG (Mw = 5000 g mol1) layer. With
that, the probe colloids are significantly larger than the estimated is converted into the amplitude autocorrelation function Fs(q,t)
mesh size of the semidilute solution-network, so we expect compar- via the Siegert relation
ability between the microrheology and the macrorheology results.25
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Table 1 summarises the measured and calculated molecular and Fs ðq; tÞ ¼ gð1Þ ðq; tÞ ¼ gð2Þ ðq;tÞ  1 (2)
structural properties of the PEG samples.
with the angular- and time-dependent intensity I(q,t), and the lag
Dynamic light scattering (DLS) and microrheology
time t. In these equations, the scattering vector q is defined as
Before the actual microrheology experiments are performed, q = (4pnD/l)sin(y/2), where nD, y, and l, are the refractive index
the hydrodynamic radius of the gold nanoparticles in an of the solvent, the scattering angle, and the laser wavelength.
The amplitude autocorrelation function Fs(q,t) is the Fourier
transform of the van-Hove-autocorrelation function, describing
Table 1 Molecular and structural characteristics of the PEG-2k and
PEG-20k samples used in this work the time- and space-dependent particle motion as an ensemble.
The experimental autocorrelation data are fitted by an exponential
Mw Mn RG c* c xcalcd function g(1)(q,t) = Aexp(t/tR) for ideal diluted solutions of
Sample (g mol1) (g mol1) PDI (nm) (g L1) (g L1) (nm)
dispersed spherical particles, with the amplitude A and the
PEG-2k 1810 1860 1.03 1.62 200 100 2.72 characteristic relaxation time tR, and the diffusion coefficient D
200 1.62
300 1.20 is computed as D = (tRq2)1. The hydrodynamic correlation
400 0.96 length xH for polymer solutions in the semidilute concentration
regime can be calculated by a Stokes–Einstein analog relation
PEG20-k 23 540 22 080 1.07 5.13 35 35 5.13
70 3.05 kB T
150 1.72 xH ¼ (3)
200 1.39 6pZD

2668 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Paper Soft Matter

with the Boltzmann constant kB, temperature T, and the viscosity purified as described above. First, we investigate the linear
Z. For the microrheology conversion process, the time-dependent viscoelastic regime boundaries of every sample by implement-
mean-square displacement hDr(t)2i (MSD) of the probe colloids ing an amplitude sweep at a constant frequency of o = 1 rad s1
in the polymer solutions is extracted from a stretched biexponen- and amplitudes in the range of 0.01–100% deformation. Then,
tial fit function using the Einstein–Smoluchowski identity frequency sweeps are carried out in a frequency range of
hDr(t)2i = 6Dt and o = 0.1–100 rad s1, performed from high to low frequencies,
  
2  with a fixed amplitude within the linear viscoelastic regime
2 DrðtÞ
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

ð1Þ (LVE) of the sample.

g ðq; tÞ ¼ gð0Þ  exp q (4)
6

The MSD is then converted into the time-dependent creep com-

Results and discussion
pliance J(t) according to J(t) = (pa/kBT)hDr(t)2i, with the particle
radius a.27 To calculate the complex shear modulus G*(o), the In this work, like in others before,1–8 a frequency-independent
numerical inversion method introduced by Evans et al.28 based on plateau in the storage modulus G 0 is detected by using oscilla-
io ðJ1  J ð0ÞÞ tory shear rheology at low frequencies, as exemplified for PEG-
¼ io  J ð0Þ þ ð1  expðiot1 ÞÞ solutions with molar masses of 20 kg mol1 and 100 kg mol1,
G ðoÞ t1
both at c = 200 g L1, in Fig. 2. The value of this plateau lies up
N  
expðiotN Þ X Jk  Jk1 (5) to three decades above the calculated low-torque limit of the
þ þ
Z tk  tk1 rheometer, whose torque resolution is appraised as
k¼2
3Tmin
 ðexpðiotk1 Þ  expðiotk ÞÞ Gmin ¼ ; (6)
2pR3 g0
is used with the free software Compliance to Complex Moduli V3.
with the minimal modulus Gmin, the instrument’s minimum
Static light scattering torque Tmin = 0.5 Nm, the cone radius R, and the deformation
amplitude g0.29 In both solutions, the frequency-independent
Static light scattering (SLS) is performed using an ALV-SP86
plateau persists up to frequencies of o = 10 rad s1; until then,
goniometer equipped with an ALV 3000 correlator, an ALV High
the storage modulus curves bend over and approach the
QE APD avalanche photodiode fiber optical detection system,
expected Maxwell-type terminal-flow power-law dependence
and a Uniphase He/Ne laser (22 mW, l = 632.8 nm). All
with a log–log slope of 1.8. Even with different rheometer
measurements are carried out at 293 K and at multiple detec-
geometries (cone–plate of different sizes, concentric cylinder,
tion angles of y ranging from 301 to 1501 in steps of 101. We
and double gap), the low-frequency plateau is obtained in every
use the Rayleigh ratio of toluene RTol = 1.39  105 cm1 at
case, such that we exclude interfacial tension effects as a
l = 632.8 nm to calculate the Rayleigh ratio of the samples. PEG
possible reason for its appearance. In contrast to the unusual
solutions without gold nanoparticles are prepared at different
curve shape of G 0 , the loss modulus G00 is in accordance with
concentrations in the semidilute regime, all of which at
the Maxwell model over the whole frequency range investi-
or above the respective overlap concentration (200, 300, and
gated, with a log–log slope of 1. Besides general errors in the
400 g L1 for PEG-2k, and 35, 70, and 200 g L1 for PEG-20k).
The sample and cuvette preparation is analogous to the dynamic
light scattering procedure described above. For the experimental
data evaluation, we plot Kc/R against the square of the scattering
vector q2 with the contrast factor K = 4p2nD2(dn/dc)2/NAl4, using
the refractive index of the solvent nD, the refractive index
increment (dn/dc), and the refractive index n of the polymer
solutions. The refractive index increments for both PEG samples
were experimentally determined to be 0.135 mL g1 (PEG-2k)
and 0.139 mL g1 (PEG-20k) at l = 632.8 nm.

Rheology
To compare the microrheology results to analogues obtained by
classical macrorheology, we performed classical shear experi-
ments using an Anton Paar modular compact rheometer of type
MCR 302 (Anton Paar, Graz, Austria) equipped with a cone–
plate measuring system, CP50-1/TG, with a cone radius of
Fig. 2 Frequency-dependent storage (squares) and loss moduli (circles)
24.983 mm and a cone angle of 0.9961 at 293 K. For all
of a PEG-20k solution (red) and a PEG-100k solution (blue), each at a
measurements, a solvent trap is used to suppress evaporation concentration of 200 g L1, measured macroscopically with a rheometer
of the solvent. To assure comparability, the same PEG samples at an amplitude well within the linear viscoelastic regime. The dashed line
as in the light scattering studies are used, prepared and represents the calculated low-torque limit of the instrument.

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 2666--2676 | 2669
 View Article Online

Soft Matter Paper

determination of low torque values, the unexpected curve- rates, these network structures may be destroyed, leading to a
shape of the storage modulus G 0 at low frequencies can be transition to the normal terminal flow regime in the viscoelastic
caused by phase-angle uncertainties as an effect of the instru- spectrum.
ment, as comprehensively described by Velankar and Giles.30 To examine whether the low-frequency plateau has its origin
According to their work, the total error in G 0 can be expressed as in inaccurate phase-angle values or appears as a consequence
a combination of the error in the complex modulus |G*| and of a real underlying structural feature in the samples, and to get
the error in the phase angle: a closer insight into this structure, we performed microrheology
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

on the polymer solutions, thereby extending the accessible

dG0 djG j frequency range and obtaining microscopic information about
¼  tanðdÞdd (7)
G0 jG j the viscoelastic response of the system.
As this measurement technique does not depend on any
These errors particularly occur if the values of G 0 and G00 differ phase-angle determination based on the quality of the initial
significantly, so even if their values lie well above the low torque torque and displacement signal, we may be able to identify
limit of the rheometer, its phase-angle resolution has to be whether inaccurate phase angles or microscopic structural
taken into account. Especially, fluid samples with low torque irregularities are the origin of the low-frequency plateau. The
values exhibit a noisy measurement signal, thus leading to second advantage of the microrheology technique for our
uncertainties in the correct phase-angle determination. In the investigation is that the effective shear forces are smaller than
words of Velankar and Giles:30 ‘‘as d approaches 90 degrees, the those in macrorheology, as there is no external load applied to
fractional error in G 0 grows without bound even when the error the sample in microrheology, but instead, just the common
in phase angle dd remains finite. This is just a quantitative way thermal energy drives a diffusive motion of nanoscopic probe
of stating that it is difficult to characterize the elasticity of particles inside the polymer solutions. Therefore, we may
weakly-viscoelastic materials’’. The rheological spectrum of our presume that, if the hypothesis of a structural reason is correct,
PEG-100k solution in Fig. 2 (blue symbols) shows that due to the low-frequency plateau in the storage modulus should be
these errors, a low-frequency plateau can exist even at high even more extended to even higher frequencies when probed by
absolute modulus-values of around 0.3 Pa, despite having such passive microrheology.
nearly the same ratio of G 0 and G00 as the pEG-20k solution
(red symbols).
As an alternative interpretive approach to the latter method- Microrheology
inherent line of argument, the frequency-independent curve Our experiments are based on dynamic light scattering, where
shape of G 0 can be seen to be not an artifact, but to actually we track the Brownian motion of spherical gold nanoparticles
reflect the existence of a weak network structure in the sample with a PEG corona through the polymer solutions. The PEG
that is able to store external deformation energy as long as that corona provides thermodynamic compatibility between the tracer
energy is weak. At low shear frequencies, along with low shear particles and their surrounding solution-network matrices. As an
rates at a given shear amplitude, this might be the case, example, Fig. 3A shows a set of the normalized autocorrelation
whereas at higher shear frequencies and therefore higher shear functions of a PEG solution containing a small amount of the

Fig. 3 (A) Normalized DLS autocorrelation function g(1)(t) of an aqueous solution of PEG-20k (c = 70 g L1), containing PEG-coated gold nanoparticles
(d = 20 nm, Mw,PEG–corona = 5000 g mol1) as probe colloids observed at scattering angles of 301, 601, 901, and 1201 (black, red, green, and blue symbols),
along with fits and corresponding residuals to a bi-Kohlrausch–Williams–Watts (bi-KWW) function. (B) Angular dependence of the inverse relaxation
times, showing two different diffusive processes.

2670 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Paper Soft Matter

colloidal gold tracer particles. The experimental autocorrelation The second observed process is slower than the fast one by
data are all well fitted by a stretched biexponential function of two orders of magnitude and corresponds to the hindered gold
Kohlrausch–Williams–Watts type. tracer diffusion. With an increase in the polymer concentration
 C !  E ! and therefore the solution viscosity, the velocity of the tracer
ð1Þ t t decreases due to greater obstruction by the surrounding med-
g ðtÞ ¼ A þ B  exp  þ D  exp  (8)
t1 t2 ium. Based on this principle, the viscoelastic properties of that
medium can be quantified from the tracer motion. As it is
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

From this fit, two relaxation times, tslow and tfast, of the diffusion driven by the thermal energy kBT only, no external forces are
processes can be computed as applied, and hence, measurements in the linear viscoelastic
    regime are always ensured; also, the risk of measurement-
t1 1 t2 1
tslow ¼  G and tfast ¼  G (9) induced destruction of the relevant microscopic sample struc-
C C E E
tures is excluded. The tracer mean-square displacement (MSD)
with the Gamma function G and the fit-function exponents C and is extracted from the DLS autocorrelation fit function and
E. Plotting the inverse relaxation times tslow1 and tfast1 versus converted into the complex shear modulus by a numerical
the square of the scattering vector q2 yields two straight lines inversion method that works without any need for a Fourier
through the origin, with slopes representing two different diffu- transform of the experimental data and thereby minimizes
sion coefficients, as can be seen in Fig. 3B. Fig. 4 shows the fast transformation errors (see Experimental part).
(filled squares) and the slow (open circles) diffusion coefficients Fig. 5 shows the resulting frequency-dependent data for the
depending on the polymer concentration of the PEG-20k samples. loss and storage modulus of a PEG-20k solution at c = 150 g L1,
We attribute the fast coefficient Dfast to the cooperative in comparison with the corresponding data from classical
motion of the overlapping PEG chain segments. This coopera- macroscopic shear rheology. We achieve good agreement
tive diffusion coefficient increases with the increase of the between the loss modulus data from both, macro- and micro-
polymer concentration in the semidilute region (c 4 35 g L1 rheology, in the overlapping frequency range from 101 to
for PEG-20k and c 4 200 g L1 for PEG-2k), as the freely 60 rad s1. In contrast, the low-frequency plateau in the storage
movable polymer segments between overlap points of the modulus G 0 does not appear in the microrheology data. As the
semidilute solution-network decrease in length if the polymer shear forces in microrheology are as weak as possible (only
concentration increases,24 thereby leading to a faster relaxa- thermal energy kBT drives the tracer particles), according to our
tion, whereas in the dilute regime, the fast diffusion coefficient above hypothesis of weak network structures in the PEG solu-
is nearly independent of the polymer concentration and can be tions, these should be more likely to be destroyed in macro-
described by the following equation:31 rheology than in microrheology, such that the latter method
should show them to be even more extended than the first
Dapp = Ds(1 + kcD)(1 + hRG2izq2) (10) method. The opposite is found, though. A possible reason for
the discrepancies in micro- and macrorheology could be inter-
with kD = 2A2Mw  (RH3/Mw), the mean-square radius of
action effects between the colloidal gold tracers and the polymer
gyration hRG2iz, the second virial coefficient A2, the hydro-
chains, in particular depletion effects, whereby the gold tracers
dynamic radius RH, and the molar mass Mw. Extrapolation of
are able to create their own microenvironment resulting in a
the concentration c and the scattering vector q2 to zero results
in the self-diffusion coefficient Ds of the single polymer coils.

Fig. 4 Concentration dependence of the fast (filled squares) and the slow Fig. 5 Frequency dependence of storage and loss modulus (black and
(open circles) diffusion coefficients determined by DLS on aqueous solu- red) for a PEG-20k solution (c = 150 g L1), obtained by DLS-based
tions of PEG-20k. The grey bar marks the polymer overlap concentration c*. microrheology (lines) and by classical shear rheology (open symbols).

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 2666--2676 | 2671
 View Article Online

Soft Matter Paper

layer of depleted polymer concentration surrounding the probe

particle. In this case, a microscopic origin for the second plateau
found in macrorheology would not be detectable. To check
for the latter supposition, we design and apply a new DLS
auto correlation function from previously known experimental
parameters:
    
t t
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

gð1Þ ðtÞ ¼ B  C  exp  þ D  exp 

t1;polymer t2;polymer
    
t t
þ E  F  exp  þ G  exp 
t1;tracer t2;tracer
(11)

In this function, the first addend represents the matrix polymer

solution; for the amplitude ratios C and D and for the two
relaxation times t1,polymer and t2,polymer, we plug in the estimated Fig. 7 Normalized autocorrelation data of g(1)(t) and their fit functions
values from dynamic light scattering experiments of the polymer obtained from a PEG-2k sample (c = 100 g L1) after several filtration steps
solutions without tracer particles at the same concentrations, with a Millex-AAs filter with 800 nm pore size (black symbols), a Millex-
respectively. This procedure is possible if we assume that the LCRs filter with 450 nm pore size (blue symbols), a Millex-VVs filter with
100 nm pore size (green symbols), and a combination of a Whatman
gold tracers have no influence on the polymer relaxation.
Anotop filter with 0.02 mm pore size and a Millex-LGs filter with 200 nm
The second addend contains contributions from the tracer pore size (violet symbols).
particles; for the amplitude ratios F and G, we use parameters
from the fit function of the gold tracer in aqueous solution
without polymer. Only the overall amplitude ratios B and E and further into this matter, we investigated the dynamics and
the tracer relaxation times t1,tracer and t2,tracer are left open to structure of the polymer solutions by dynamic (DLS) and static
fitting, as the increasing viscosity of the polymer solutions (SLS) light scattering.
compared with water leads to a decelerated tracer diffusion.
Dynamic light scattering
The residuals in Fig. 6 show good agreement between the
designed fit function and the raw data, confirming our first DLS measurements are performed on the semidilute polymer
assumption that the initial polymer relaxation is not influenced solutions without gold nanoparticles to analyze the scattering
by interactions with the gold nanoparticles such as depletion. contributions of the polymers only. Before being able to do
In conclusion, the former suggested weak PEG network that, we need to implement a series of multiple filtration steps
structures could not be discovered by microrheology, leading to remove impurities. To explicitly illustrate the impact of this
us to the supposition that their appearance in macrorheology filtration sequence, Fig. 7 shows the normalized autocorrela-
indeed originates from a phase-angle instrument effect. To go tion functions of a freshly prepared PEG-2k solution with a
concentration of c = 100 g L1, obtained after one to four
filtration steps. Note that this concentration is below the over-
lap threshold of c* = 200 g L1, so in this solution, we should
definitely not expect any slow modes. Anyhow, a diffusive slow
mode still appears, and only after the last filtration step with a
combination of a Millex-LGs filter (200 nm pore size) and a
Whatman Anotop filter (25 mm, 20 nm pore size) is it success-
fully removed and also does not reappear. Via the amplitudes of
the different modes in the autocorrelation functions before that
removal, we can estimate the concentration portion of impu-
rities and their contribution to the total scattering intensity in
the previous filtration steps. As an example, this is done for the
PEG-2k solution after the first filtration step, where a tri-
exponential fit function with the amplitudes B, C, and D has
to be used to adequately fit the experimental data.
     
ð1Þ t t t
Fig. 6 DLS autocorrelation raw data (black filled squares) of an aqueous g ðtÞ ¼ A þ B  exp  þ C  exp  þ D  exp 
solution of PEG-20k (c = 70 g L1) containing PEG-coated gold nano-
t1 t2 t3
particles (d = 20 nm, Mw,PEG-corona = 5000 g mol1) along with a fit of a (12)
designed function consisting of the separate contributions from the PEG
matrix and the gold tracers (red line). The residuals between the fit and the Three hydrodynamic radii are calculated to be RH,1 = 1.4 nm,
raw data are shown as a blue line at the bottom. RH,2 = 185 nm, and RH,3 = 540 nm. In general, the total

2672 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Paper Soft Matter

scattering intensity Itotal consists of the sum of the scattering species (RH,3 = 540 nm), but this 0.79% still has a contribution
contributions of i species Ii, and via the normalized amplitude to the overall scattering intensity of about 62%. This estimate
ratios Ai, the intensity proportion of each species can be shows how crucial it is to thoroughly filter samples in DLS
estimated separately. analytics. Such filtration, however, removes material from
the sample.
X
N
Itotal ¼ Ii (13) To assure that a desired polymer concentration is used in
i¼1 DLS experiments along with assuring good sample purity, we
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

therefore recommend filtering stock solutions, then freeze-

with
drying them, and finally preparing the samples from the
X
N
filtered and freeze-dried material. Using such purified solu-
Ii ¼ Ai  Itotal (14)
i¼1
tions in the semidilute concentration regime, it is now possible
to gain structural insights by fitting the amplitude autocorrela-
The amplitude ratio Ai of each relaxation process is proportional tion function with a biexponential fit function to evaluate
to the concentration ci and the molar mass Mi of the species i. the cooperative diffusion coefficient, as shown in Fig. 8A. The
Ai B ciMi (15) second addend of the fit function is chosen to take into account
polymer polydispersity, even though it is small. Using the
To estimate the molar mass M of each species i, we use the classical cooperative diffusion coefficient Dcoop, we can calculate the
scaling of the radius of gyration of monodisperse Gaussian coils hydrodynamic correlation length xH of the samples according
Rg B M0.6 (16) to eqn (3). Both these characteristic data are summarized in
Table 2 for the two classes of PEG samples investigated in our
and convert that radius into the hydrodynamic radius RH through study. An increase in the polymer concentration leads to a
use of a relation reported by Burchard and colleagues32 reduction of the hydrodynamic correlation length of the chain-
overlap network,24 and as a result of this, an increase in the
Rg/RH = 1.78 (17)
cooperative diffusion coefficient due to the faster wobbling of
to obtain

1.78RH B M0.6 (18) Table 2 Cooperative diffusion coefficients and correlation lengths of
differently concentrated semidilute solutions of the polymers PEG-2k
The concentration portion xi of the i th species is then and PEG-20k, as obtained by DLS
obtained by
Polymer c (g L1) Dcoop (cm2 s1) xH (nm)
Ai ci
ci  0:6
and xi ¼ N (19) PEG-2k 200 2.09  10 6
1.03
1:78  RH;i P
ci 300 2.21  106 0.97
i¼1 400 2.34  106 0.92

From eqn (13)–(19) and the three hydrodynamic radii, we PEG-20k 35 7.57  107 2.84
estimate that only 0.79% of the sample consists of the largest 70 9.94  107 2.16

Fig. 8 (A) Experimental normalized amplitude autocorrelation function g(1)(t) of a PEG-20k solution (c = 70 g L1), recorded at eight different scattering
angles (301, 561, 641, 811, 981, 1151, 1241, and 1581), and residuals between a biexponential fit function and the raw data at the bottom. (B) The cooperative
diffusion coefficient Dcoop versus the square of the scattering vector q2 shows no angular dependence and yields a hydrodynamic correlation length of
xH = 2.16 nm.

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 2666--2676 | 2673
 View Article Online

Soft Matter Paper

Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

Fig. 9 Static light scattering results for PEG-2k (A) and PEG-20k (B) at three different concentrations, measured at 293 K with the experimental
parameters l = 632.8 nm, dn/dc = 0.135 mL g1 (PEG-2k) and 0.139 mL g1 (PEG-20k), and n = 1.3288.

the chain strands between overlap points can be detected. After

thorough purification of the polymer samples and the solutions
by means of several filtration steps, we obtain no slow mode in
the light scattering autocorrelation functions. From these find-
ings, we conclude that colloidal-scale impurities, even in trace
amounts only, are the origin of a marked slow mode in PEG
solutions, but if these impurities are removed by tedious
but effective sample purification, no slow mode is observed. In
such purified samples, the diffusion coefficient Dcoop is inde-
pendent of the square of the scattering vector q2 (Fig. 8B),
indicating a solution containing no particles or polymer coils
with a hydrodynamic radius bigger than 10 nm. These findings
are in accordance with an earlier work by Devanand and Selser33
and lead us to the picture of homogeneous semidilute solutions
without PEG aggregates, microphase separation, or crystalline
PEG micro- or nanophases. Furthermore, the appearance of Fig. 10 Micro- and macrorheology results at different temperatures,
stabilized air bubbles and impurities cannot be verified. 278 K (blue), 293 K (green), and 323 K (red), of a PEG-2k solution (c = 200 g L1).

Static light scattering

As a final step, we perform static light scattering experiments to pre-conclusion, the present results intensify our above assump-
detect the microstructure on scales larger than 10 nm, for tion that the low-frequency plateau in rheology appears to be
example, in view of a potential presence of larger aggregates rather an effect of the rheometer and not due to actual
of phase-separated PEG chains. We probe the same samples as structural features of the PEG chains in solution. To tie up this
in the dynamic light scattering experiments described in the hypothesis, we undertake further investigations on the tem-
section above. Plotting the inverse Rayleigh ratio against the perature dependence of the low-frequency plateau in macro-
scattering vector shows no angular dependence of the mea- scopic rheology results.
sured scattering intensity, as shown in Fig. 9. No reliable slopes
can be extracted out of the fitted straight lines, such that we Influence of temperature
conclude all correlation lengths to be smaller than 10 nm. As a To further investigate the influence of potentially weak aggre-
result of the static light scattering experiments, we cannot find gates such as hydrogen bonded or hydrophobic associations on
any evidence of clusters or micro-phase separated domains of the low-frequency plateau in rheology results, we probe our
PEG in the aqueous solutions. Similar results have also been samples at different temperatures. Such interactions can be
found by Borchard et al.34 These findings confirm our previous broken at higher temperatures, which should lead to a distinct
DLS results. After thorough purification, the polymer solutions change in the rheological behavior of the polymer solutions.
contain completely and homogeneously dissolved PEG chains, Fig. 10 shows the frequency-dependent storage and loss
and we can also exclude the hypothesis of air bubbles with sizes moduli of the PEG-2k sample in solution at c = 200 g L1,
in the range of 0.1–1 mm in the solutions.12 As a first brief measured at three different temperatures using macrorheology

2674 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Paper Soft Matter

and at two different temperatures using microrheology. We see References

that the low-frequency plateau in the macrorheology data exists
at every measured temperature, wherein the storage modulus of 1 A. W. Bosnian, L. Brunsveld, B. J. B. Folmer, R. P. Sijbesma
that plateau shows no systematic dependence on the chosen and E. W. Meijer, Macromol. Symp., 2003, 201, 143–154.
temperature. In the microrheology data, by contrast, again, no 2 T. Vermonden, M. J. van Steenbergen, N. A. M. Besseling,
such plateau is seen in any case. This finding further rules A. T. M. Marcelis, W. E. Hennink, E. J. R. Sudhölter and
out the hypothesis of weak attractive interactions between the M. A. Cohen Stuart, J. Am. Chem. Soc., 2004, 126, 15802–15808.
3 F. van de Manakker, T. Vermonden, N. El Morabit, C. F. van
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

PEG chains causing a weak network structure in semidilute

solutions at rest. Nostrum and W. E. Hennink, Langmuir, 2008, 24, 12559–12567.
4 S. Hackelbusch, T. Rossow, P. van Assenbergh and S. Seiffert,
Macromolecules, 2013, 46, 6273–6286.
5 T. Rossow, A. Habicht and S. Seiffert, Macromolecules, 2014,
Conclusions 47, 6473–6482.
When light scattering is performed on thoroughly purified 6 H. Goldansaz, Q. Voleppe, S. Piogé, C. A. Fustin, J. F. Gohy,
semidilute aqueous poly(ethylene glycol) solutions, no evidence J. Brassinne, D. Auhl and E. van Ruymbeke, Soft Matter,
for any sort of complex structures is observed, neither slow 2015, 11, 762–774.
modes in DLS nor angular dependencies in SLS. These results, 7 M. J. Sánchez-Miranda, E. Sarmiento-Gómez and J. L. Arauz-Lara,
though, require multiple careful purification steps of the Eur. Phys. J. E: Soft Matter Biol. Phys., 2015, 38, 3.
samples before and after the preparation of the solutions. We 8 A. Habicht, S. Czarnecki, T. Rossow and S. Seiffert, J. Polym.
speculate that improper purification might have been the cause Sci., Part B: Polym. Phys., 2017, 55, 19–29.
of many contrary findings reported in the literature. In addi- 9 D. Gaudino, R. Pasquino and N. Grizzuti, J. Rheol., 2015, 59,
tion, in light-scattering-based microrheology, no low-frequency 1363–1375.
plateau of the storage modulus is observed in any case, in 10 J. E. Moore, T. M. McCoy, L. de Campo, A. V. Sokolova,
contrast to results of macrorheology on the same polymer C. J. Garvey, G. Pearson, B. L. Wilkinson and R. F. Tabor,
solutions, in which such a low-frequency plateau is observed J. Colloid Interface Sci., 2018, 529, 464–475.
even at high temperatures. Hence, we consider this plateau to 11 (a) K. Tasaki, J. Am. Chem. Soc., 1996, 118, 8459–8469; (b) A. Azri,
be an instrument effect of the classical shear rheometer, even P. Giamarchi, Y. Grohens, R. Olier and M. Privat, J. Colloid
though the data lie two decades above the low-torque limit of Interface Sci., 2012, 379, 14–19; (c) A. Azri, M. Privat, Y. Grohens
the instrument. In line with the work of Velankar and Giles,30 and T. Aubry, J. Colloid Interface Sci., 2013, 393, 104–108.
erroneous phase-angle values appear to be the origin of this 12 J. Wang, Macromolecules, 2015, 48, 1614–1620.
artifact. The method of microrheology overcomes this limita- 13 (a) W. F. Polik and W. Burchard, Macromolecules, 1983, 16,
tion and accurately probes both the high- and low-frequency 978–982; (b) M. Polverari and T. G. M. van de Ven, J. Phys.
regimes, indicating no presence of any sort of transient or weak Chem., 1996, 100, 13687–13695; (c) M. Duval and D. Sarazin,
network structures in PEG semidilute solutions, in line with Polymer, 2000, 41, 2711–2716; (d) B. Hammouda, Polymer,
similar conclusions made by Devanand and Selser33 and 2009, 50, 5293–5297.
Faraone et al.35 These findings also appear transferrable to 14 P. Zhou and W. Brown, Macromolecules, 1990, 23, 1131–1139.
other types of soft-matter systems, as a macrorheology-probed 15 B. Porsch and L.-O. Sundeloef, Macromolecules, 1995, 28, 7165–7170.
low-frequency plateau was also found, for example, in supra- 16 (a) M. Adam and M. Delsanti, Macromolecules, 1977, 10, 1229–1237;
molecular polymer gels based on PNIPAAm36 or even in systems (b) W. Brown and T. Nicolai, Colloid Polym. Sci., 1990, 268, 977–990;
with micellar components,9,10 which may all be well explain- (c) D. L. Ho, B. Hammouda and S. R. Kline, J. Polym. Sci., Part B:
able by our current conclusions of methodological limits in Polym. Phys., 2003, 41, 135–138; (d) J. Li, W. Li, H. Huo, S. Luo and
classical rheology and sample impurities in light scattering. C. Wu, Macromolecules, 2008, 41, 901–911; (e) M. Duval and
E. Gross, Macromolecules, 2013, 46, 4972–4977.
17 J. Wang and C. Wu, Macromolecules, 2016, 49, 3184–3191.
18 S. Seiffert, Prog. Polym. Sci., 2017, 66, 1–21.
Conflicts of interest
19 M. Duval, Macromolecules, 2000, 33, 7862–7867.
There are no conflicts to declare. 20 T. Jian, D. Vlassopoulos, G. Fytas, T. Pakula and W. Brown,
Colloid Polym. Sci., 1996, 274, 1033–1043.
21 Q. Ying and B. Chu, Macromolecules, 1987, 20, 362–366.
22 F. E. Bailey, J. L. Kucera and L. G. Imhof, J. Polym. Sci., 1958,
Acknowledgements
32, 517–518.
We sincerely thank Dimitris Vlassopoulos for stimulating 23 P. Gregory and M. B. Huglin, Makromol. Chem., 1986, 187,
discussions and valuable suggestions on the interpretation of 1745–1755.
macrorheology data in the low-frequency limit. This work was 24 M. Daoud, J. P. Cotton, B. Farnoux, G. Jannink, G. Sarma,
supported by the German Research Foundation (DFG) under H. Benoit, C. Duplessix, C. Picot and P. G. de Gennes,
grant no. SE 1888/7-1. Macromolecules, 1975, 8, 804–818.

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 2666--2676 | 2675
 View Article Online

Soft Matter Paper

25 E. C. Cooper, P. Johnson and A. M. Donald, Polymer, 1991, 31 Light Scattering from Polymer Solutions and Nanoparticle Dis-
32, 2815–2822. persions, ed. W. Schärtl, Springer-Verlag, Berlin Heidelberg,
26 S. Balog, L. Rodriguez-Lorenzo, C. A. Monnier, B. Michen, 1st edn, 2007.
M. Obiols-Rabasa, L. Casal-Dujat, B. Rothen-Rutishauser, 32 (a) W. Burchard, M. Schmidt and W. H. Stockmayer, Macro-
A. Petri-Fink and P. Schurtenberger, J. Phys. Chem. C, 2014, molecules, 1980, 13, 1265–1272; (b) W. Burchard and
118, 17968–17974. W. Richtering, Prog. Colloid Polym. Sci., 1989, 80, 151–163.
27 J. J. Duffy, C. A. Rega, R. Jack and S. Amin, Appl. Rheol., 2016, 33 K. Devanand and J. C. Selser, Nature, 1990, 343, 739–741.
Published on 26 February 2019. Downloaded by North Carolina State University on 11/6/2023 4:03:42 PM.

26, 15130. 34 H. Venohr, V. Fraaije, H. Strunk and W. Borchard, Eur.

28 R. M. L. Evans, M. Tassieri, D. Auhl and T. A. Waigh, Phys. Polym. J., 1998, 34, 723–732.
Rev. E: Stat., Nonlinear, Soft Matter Phys., 2009, 80, 12501. 35 A. Faraone, S. Magazù, G. Maisano, P. Migliardo, E. Tettamanti
29 Complex Fluids in Biological Systems. Experiment, Theory, and and V. Villari, J. Chem. Phys., 1999, 110, 1801–1806.
Computation, ed. S. Spagnolie, Springer, New York, 1st edn, 2015. 36 T. Rossow, S. Hackelbusch, P. van Assenbergh and
30 S. Velankar and D. Giles, Rheol. Bull., 2007, 76, 8–20. S. Seiffert, Polym. Chem., 2013, 4, 2515.

2676 | Soft Matter, 2019, 15, 2666--2676 This journal is © The Royal Society of Chemistry 2019