Polymer 54 (2013) 2632e2638

Contents lists available at SciVerse ScienceDirect

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

Separation of cellulose acetates by degree of substitution

Hewa Othman Ghareeb, Wolfgang Radke*
Fraunhofer-Institut für Betriebsfestigkeit und Systemzuverlässigkeit LBF, Bereich Kunststoffe, Schlossgartenstraße 6, D-64289 Darmstadt, Germany

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

Article history: A liquid chromatographic method was developed allowing separating cellulose acetates (CA) with
Received 1 December 2012 respect to degree of substitution (DS). The normal-phase gradient separation, which is based on an
Received in revised form adsorptionedesorption mechanism, uses a multi-step gradient of dichloromethane (DCM) and methanol
20 March 2013
(MeOH) on a bare silica column as stationary phase. Applicability of the developed multi-step gradient
Accepted 21 March 2013
allows separating CAs within the DS-range DS ¼ 1.5e2.9, which is the target DS-range for most CA
Available online 28 March 2013
applications. To the best of our knowledge the developed method for the first time allows determination
of the DS-distribution of intact CA chains over a wide DS-range.
Keywords:
Cellulose acetate
Ó 2013 Elsevier Ltd. All rights reserved.
Degree of substitution
Degree of substitution distribution

1. Introduction differently substituted AGUs can be applied to characterize the

Due to the architectural complexity of cellulose derivatives,
analytical methods for comprehensive analysis of the molecular
heterogeneities have become increasingly important in cellulose
research field. Cellulose derivatives are complex copolymers being
heterogeneous at least in molar mass and chemical composition.
These heterogeneities need to be characterized as they critically
affect many properties of these derivatives such as adhesion
strength, solubility, viscoelasticity and drug release from hydro-
philic tablets just to name a few [1e7]. The heterogeneity in chem-
ical composition of cellulose derivatives includes the distribution of
the substituents within the individual anhydroglucose units (AGUs),
i.e. the partial degree of substitution (DS) of O-2, O-3, and O-6 atoms,
as well as the distribution of the substituents among the polymer
chains (heterogeneity of 1st order) and along the polymer chains
(heterogeneity of 2nd order). DS refers to the average number of
substituted hydroxyl groups per AGU. Therefore, DS can take values
between 0 and 3 for unbranched cellulose derivatives.
Today cellulose derivatives are mainly characterized in terms of their
molar mass and molar mass distribution (MMD), their average DS and
of the distribution of the substituents on both the monomer and olig-
omer levels. NMR techniques are applied to obtain a first insight in the
partial DS within the monomer units [8e16]. Alternatively complete
acidic chain degradation and subsequent determination of the resulting
* Corresponding author.
E-mail address: wolfgang.radke@lbf.fraunhofer.de (W. Radke).
composition of cellulose derivatives [17e20]. In another characteriza-
tion pathway, the polymer chains are completely or partially degraded
using acids or enzymes. The partial degradation of the chains results in a
mixture of monomers and/or oligomers present in different molar ra-
tios. These mixtures are separated and characterized subsequently in
detail using various analytical techniques, alone or in combination, such
as size exclusion chromatography (SEC) [21e24], anion-exchange
chromatography (AEC) [23e28], gaseliquid chromatography (GLC)
[21,27,29,30] and mass spectrometry (MS) [26e35]. This gives infor-
mation on the monomer composition as well as the substituent dis-
tribution on the oligomeric levels. Such data can be compared with the
product distribution expected for a specified monomer distribution
along the chain at a certain DP, e.g. a completely random monomer
distribution. This procedure therefore yields information on the dis-
tribution of the differently substituted AGUs along the polymer chain
(2nd order heterogeneity).
In general, any or at least a large amount of information on
whether the different monomeric or oligomeric units result from
the same or from different chains is lost, if the sample is fully or
partially degraded. Therefore, establishing methods to characterize
the chemical heterogeneity of cellulose derivatives on the level of
intact polymeric chains is still a highly challenging task.
Gradient HPLC has been shown to be a powerful tool for sepa-
rating (co)polymer molecules according to chemical composition
[36,37]. Such separations allow determining the chemical compo-
sition distribution (CCD) of the copolymers, i.e. the distribution of
the monomer units among the different polymer chains. Strictly,
cellulose derivatives have to be regarded as complex polymers

0032-3861/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.polymer.2013.03.041
 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638 2633

composed of at least eight different monomer units, which are the 2. Experimental section
differently substituted AGUs. Today, it is hardly possible to separate
intact cellulose chains of such complexity by the fractions of 2.1. Materials
the differently substituted AGUs. In our somehow more naïve pic-
ture we regard cellulose derivatives as being composed of only CA samples with average DS ranging from DS ¼ 1.5 to DS ¼ 2.9
two kind of structural units, the AGUs and the substituents. A were used. Five of the samples (sample 4, DS ¼ 1.72; sample 13,
separation with regard to DS is therefore similar to a separation of DS ¼ 2.42; sample 14, DS ¼ 2.45; sample 15, DS ¼ 2.45; sample 17,
binary copolymers according to chemical composition. Thus, the DS ¼ 2.92) were kindly provided by Rhodia Acetow GmbH (Freiburg
DS-distribution is the analogue to the chemical composition dis- im Breisgau, Germany). The others were prepared in our laboratory
tribution of binary copolymers. The efforts on separation and by partial saponification of a commercial high DS sample (sample
characterization of cellulose derivatives, particularly CAs, with 16, DS ¼ 2.60, Acetati, Italy). Information on the average DS, weight
respect to the DS-distribution are very limited and some of the average molar masses (Mw), weight average degrees of polymeri-
attempts described in literature are briefly discussed here: zation (DPw) and molar mass dispersity (Ðm) of the samples are
SEC with MALLS/RI/UV detections has been used to gain insight summarized in Table 1. The details on the synthesis and charac-
on the substitution distribution along the chains of CA [38]. The terization of the samples are given elsewhere [48]. Dichloro-
procedure requires the presence of substituents that can be methane (DCM), dimethyl acetamide (DMAc), dimethyl sulphoxide
detected by an UV-detector in relation to the molar mass deter- (DMSO) and methanol (MeOH) (VWR, Darmstadt, Germany) were
mined by MALLS/RI. It was found that DS varies with molar mass, of HPLC grade and used as received.
with the lower molar mass fractions being less substituted than the
higher molar mass fractions. However, the DS gradient over the 2.2. Chromatographic system and conditions
MMD curve does not represent a true separation in terms of DS,
because SEC separation is based on size, not on chemical compo- For the chromatographic separations a Shimadzu HPLC system
sition. Furthermore, since CAs do not contain UV-active groups, consisting of a DGU-14A degasser, an FCV-10ALvp solvent mixing
they had to be modified to the respective amide derivatives. Thus, chamber, an LC-10ADvp pump and an SIL 10ADvp auto sampler
the modification procedures have to be quantitatively performed in were used. For detection an evaporative light scattering detector
order to achieve reliable results. (ELSD, model PL-ELS 1000, Polymer Laboratories, UK) was added.
Other techniques used to obtain insight into the CCD of CAs are The detector was operated at a nebulization temperature of 50  C,
fractionated precipitation and thin layer chromatography (TLC). an evaporation temperature of 90  C and a gas flow of 1.0 SLM. The
Fractionated precipitation involves dissolving a solid polymer in a flow rate of the mobile phase was 1.0 mL/min unless mentioned
good solvent and precipitating the desired fractions stepwise by otherwise. Data collection and processing were performed using
decreasing the solubility as a result of the controlled addition of a ‘WinGPC Software version 7.0’ (Polymer Standards Service GmbH,
non-solvent. However, this method is laborious and requires large Mainz, Germany).
amounts of solvents and samples. In addition it is not very selective The experiments were performed on a pure silica stationary phase
and difficult to automate. In most cases, the isolated fractions of CAs Nucleosil, 5 mm particle size, 100 
A pore diameter, 250 mm  4.0 mm
varied in DP but were of the same DS [39e43]. Thus, the separa- I.D. (Macherey & Nagel GmbH, Düren, Germany). The column tem-
tions were based on molar mass rather than on chemical compo- perature was kept constant at 35  C using a column oven K4 (Techlab
sition. Kamide et al. evaluated the use of TLC for the separation of GmbH, Erkerode, Germany). The injected sample volume was 15e
CAs according to DS and molar mass [44,45]. By stepwise changing 20 mL with concentrations of 1.5e2.0 g/L, if not stated otherwise.
the eluent composition, they were able to identify experimental
conditions where clear dependences of retardation factor (Rf) on
either DS or molar mass were observed. The resulting separations Table 1
Characterization data of samples used (samples from the industrial source are
according to DS and molar mass were found to be independent of
marked in grey).
molar mass and DS, respectively. The methods allowed calculating
both the DS-distribution and the MMD. However, TLC is difficult to Sample name DSa LS-Mwb,e [g/mol] DPwc Ðmd
quantify and has a poor reproducibility. Sample 1 1.53 57,600 254 3.0
Regarding the application of gradient chromatography, two sep- Sample 2 1.59 72,700 318 3.4
Sample 3 1.66 73,500 317 3.5
aration systems for different DS-ranges are reported in literature. Sample 4 1.72 44,600 190 2.6
Both systems followed reversed-phase liquid chromatography under Sample 5 1.81 68,100 286 3.2
different conditions. The first system reported by Floyd et al. allowed Sample 6 1.87 64,900 269 3.4
the separation of cellulose diacetate in the range of DS ¼ 2.3e2.7 Sample 7 1.92 63,200 260 3.2
Sample 8 1.95 76,400 313 3.5
[46]. The separation was carried out on a poly(styrene-co-divinyl
Sample 9 2.09 70,700 283 3.3
benzene) based column in a linear gradient from acetone/water/ Sample 10 2.16 71,900 284 3.2
MeOH (4:3:1) to acetone in 15 min at a flow rate of 0.8 mL/min. The Sample 11 2.19 71,000 279 3.5
samples eluted in the order of increasing average DS. The second Sample 12 2.27 74,400 289 3.3
system reported by Asai et al. was applied for the separation of cel- Sample 13 2.42 64,400 244 3.2
Sample 14 2.45 64,700 244 3.3
lulose triacetate in the range of DS ¼ 2.7e2.9 [47]. The separation was Sample 15 2.45 93,300 352 3.4
performed on a Waters Novapak-phenyl column in a gradient from Sample 16 2.60 69,800 257 3.5
chloroformeMeOH (9/1):MeOHewater (8/1) [2:8] to 100% in 28 min Sample 17 2.92 175,000 613 4.4
at a flow rate of 0.7 mL/min. a
Determined by 1H NMR [48].
Both systems allow calculating DS-distributions from a linear b
Determined by SECeMALLS in N,N-dimethyl acetamide/lithium chloride [48].
c
correlation between DS and retention volume. However, the appli- Calculated from Mw and DS [48].
d
cability of both methods is confined to the DS-ranges investigated. Based on PMMA-equivalent molar mass [48].
e
Note: The molar masses of chemically heterogeneous copolymers have to be
Therefore, the aim of the present paper was to develop a chro- regarded as apparent molar masses. For the present samples, the error due to
matographic method capable of separating CAs according to DS chemical heterogeneity was estimated to be of approximately the same size as the
over a wide DS-range. experimental error in LS measurement.
2634 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638
DMSO was used as a sample solvent. The eluents were DCM as a weak
eluent and MeOH as a displacer.
The void volume (V0 ¼ 2.8 mL) of the column was determined
from the elution volume of a low molar mass polystyrene standard
(Mw of PS ¼ 410 g/mol) using pure tetrahydrofuran (THF) as eluent.
The dwell volume (Vd) was determined to be 2.3 mL by subtracting
the void volume from the onset of the increasing UV-signal due to a
linear gradient starting from pure THF and running to THF con-
taining 0.3% acetone.
3. Results and discussion
For developing the separation method, a bare silica stationary
phase was used.
In a previous study only DMSO and DMAc/LiCl were identified to
dissolve all our samples, irrespective of DS [48]. Since the samples
were only partially soluble in DMAc without LiCl, DMSO was
preferred as solvent due to the incompatibility of non-volatile LiCl
with evaporative light scattering detection (ELSD). However, DMSO
was not selected as the initial eluent since it creates a high back
pressure and detector noise. Most importantly, however, DMSO acts
as desorption promoting liquid (desorli) and thus the use of DMSO as
eluent did not result in polymer adsorption to the stationary phase.
Seeking for conditions resulting in polymer adsorption, 0.1 mL of
1.0 mg/mL solutions of the lowest and highest DS CA (sample 1,
DS ¼ 1.53 and sample 17, DS ¼ 2.92, respectively) in DMSO were taken
and a second solvent (in this case, DCM) was added dropwise to identify
the composition at which precipitation occurs. It turned out that the
DMSO solutions could be infinitely diluted with DCM without visible
precipitation, despite the fact that the samples could not be dissolved
directly in pure DCM. Therefore, dissolution of the CAs in DMSO and
subsequent dilution with DCM appeared to be a suitable way to inject
the samples at adsorbing conditions without precipitation.
Methanol is known to be a strong displacer in normal phase
chromatography, but is a non-solvent for CA. To identify the
maximum amount of methanol in DCM rendering solubility of CAs,
Fig. 1. Overlay of normalized chromatograms of CAs having different DS (15 mL injection volume, 1.5 g/L of each sample, linear gradient change from 100% DCM to 100% MeOH (a),
100% DCM to 50% MeOH (b), 100% DCM to 35% MeOH (c) and 100% DCM to 13% MeOH (d) in 10 min). The dotted lines represent the eluent composition at the detector.
0.1 mg/mL of the lowest and highest DS CA (sample 1, DS ¼ 1.53 and
sample 17, DS ¼ 2.92, respectively) were prepared in DMSO/DCM
(10/90, v/v) and MeOH was added dropwise. It was found that the
samples were soluble until the MeOH content reaches 62% for the
sample of highest and 80% for the lowest DS.
A simple linear 10 min gradient running from 100% DCM to 100%
MeOH was applied to confirm that CAs (dissolved in DMSO) adsorb
completely from DCM as initial eluent on the silica stationary
phase, while the addition of MeOH causes desorption. The resulting
normalized chromatograms along with the eluent composition at
the detector are shown for selected samples (sample 1, DS ¼ 1.53;
sample 5, DS ¼ 1.81; sample 6, DS ¼ 1.87 and sample 17, DS ¼ 2.92)
in Fig. 1a.
As can be seen from Fig. 1a, all the peaks elute within the gradient
(i.e. at times larger than 5.1 min) but at different elution times, i.e. at
different eluent compositions. This means that all the samples are
initially adsorbed onto the stationary phase despite the use of the
DMSO as sample solvent. Desorption of the low DS samples in the
DS-range DS ¼ 1.5e1.8 occurs at significantly higher MeOH contents
than required to elute samples of DS > 1.8. Samples of the DS > 1.8
are not separated but coelute. For all samples except for sample 17
(DS ¼ 2.92) a sharp peak appears at a low elution volume within the
gradient prior to the elution of the main broad sample peak.
To increase the resolution for the high DS samples the gradient
slope was systematically reduced by running linear 10 min gradi-
ents from 100% DCM to 50%, 35% and 13% MeOH, respectively. The
normalized chromatograms of selected samples are shown in
Fig. 1bed. As the gradients becomes more shallow, the peaks of the
higher DS samples shift to higher elution volumes and a better
resolution for samples of higher DS is obtained. However, bimodal
chromatograms were also observed for the shallow gradients. The
origin of these bimodalities will be investigated in further experi-
ments below.
As can be seen from Fig. 1a, samples of low DS require up to
approximately 50% MeOH for complete elution of the peak. Thus,
when the final MeOH content decreases from 100% MeOH to 50%,
 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638 2635

35%, and 13% MeOH, the lower DS samples might be partially or

completely retained in the column. Thus any linear 10 min gradient
ending at a specific MeOH content is capable of separating only a
specific DS-range. The information about the influence of the
gradient on the DS-range possible to separate is summarized in
Table 2.
From Table 2 it becomes clear that within each gradient a
particular DS-range can be separated while the samples of lower DS
may not be desorbed and those possessing a higher DS may coelute.
This indicates that none of these linear gradients is capable of
separating the samples for the whole DS-range.
In order to determine the recoveries, isocratic experiments were
performed for selected samples at MeOH contents higher than the
MeOH amount required to desorb the sample in the gradient. Well
known concentrations of the samples were injected and well
defined volumes of the effluents were collected covering the
elution range of the sample peak. Afterwards the collected effluent
was injected without column. The sample concentration in the
effluent was determined by comparing the peak areas of the
effluent with the ones obtained from a detector calibration curve
established without column under the same experimental condi-
tions. By this procedure recoveries of 98% for sample 5, 96% for
sample 10 and 99% for sample 13 were obtained. Thus, the re-
coveries can be regarded as being quantitative.
Beside the limitations of the DS-range, another problem arose
from the early eluting sharp peaks at low retention times seen in
Fig. 1aed. From a variety of experiments it was concluded that the
use of DMSO was the main cause of this peak. It should be
mentioned, however, that the injection of pure DMSO did not result
in a peak, showing that appropriate detector settings were used.
Only the combination of DMSO together with injected sample
resulted in the appearance of the peak. Breakthrough peaks, i.e.
sample peaks from non-retained sample components due to the Fig. 2. Effect of additional isocratic or gradient steps on occurrence of prepeaks
use of a solvent being a strong eluent, can also be ruled out, as such (sample 15 (DS ¼ 2.45) dissolved in DMSO, 20 mL injection volume, 1.5 g/L of the
peaks are expected to elute with the solvent peak around 3.2 mL, sample). The dotted lines represent the eluent composition at the detector.
while the retention volumes of the prepeaks clearly indicate
adsorption onto the stationary phase. Therefore the early peaks
might be due to adsorption of DMSO to the polar silica surface linear gradient from 0 to 1% MeOH completely removes the pre-
similar to the sample molecules. To minimize or to avoid the pre- peak, resulting in a monomodal elution of the sample (Fig. 2d).
peaks, the influence of additional isocratic and gradient elution Since it was not possible to separate all samples over the com-
steps was investigated. Within these additional steps, adsorbed plete DS-range in a single 10 min linear gradient, a multi-step
DMSO may be desorbed while the polymer molecules should gradient was applied. The resulting chromatograms for some
remain adsorbed on the column. In the subsiding gradient DMSO- selected samples obtained in the final optimized gradient are shown
free sample molecules may start eluting. To examine the applica- in Fig. 3. The multi-step gradient program is included in the same
bility of the modified gradients, isocratic steps of 100% DCM for figure.
5 min and 10 min, respectively as well as a 10 min linear gradient As can be seen, CAs of different DS elute at different retention
step of 0e1% MeOH, respectively, were added before the gradient times without any prepeaks. The less polar the CA i.e. the higher the
step to 13% MeOH. The effect on the chromatograms of sample 15 DS, the lower is the MeOH content required for desorption. The
(DS ¼ 2.45) is shown in Fig. 2aed along with the eluent composi- peak maxima are observed at MeOH contents of less than 20%, but
tions at the detector. some peaks of low DSs extend to MeOH contents of up to 40%.
As can be seen, the prepeaks were not eliminated by the addi- However, the amount of MeOH required for desorption is much
tion of isocratic steps in pure DCM. The peaks are merely shifted to lower than required for the precipitation (MeOH content of 62% for
higher elution volume by 5 mL and 10 mL, which correspond to the the sample of highest DS and 80% for the lowest DS). It can there-
duration of the respective additional isocratic step (b and c, fore be concluded that the separation in the developed gradient is
respectively, relative to a in Fig. 2). In contrast, the addition of a based on an adsorptionedesorption mechanism rather than on
precipitationeredissolution.
Fig. 4 shows the dependence of chemical composition (DS) on
Table 2
Relation between the gradient and the DS-range possible to separate.
elution volume at peak maximum (Vapex) for all samples. The in-
dustrial samples are indicated by a filled rectangle (-) and those
Linear gradient Resolved approx. Coeluted approx. synthesized in this work including the precursor (sample 16,
DS-range DS-range
t ¼ 0 min t ¼ 10 min DS ¼ 2.60) by an open one (,). The composition obtained by NMR
100% DCM 100% MeOH DS ¼ 1.5 to DS ¼ 1.8 DS > 1.8 should be close to the weight average composition. In case of co-
100% DCM 50% MeOH DS ¼ 1.8 to DS ¼ 2.2 DS > 2.2 monomers of identical molar mass or for a homogeneous copolymer
100% DCM 35% MeOH DS ¼ 2.2 to DS ¼ 2.4 DS > 2.4 the composition by NMR is a true weight average. A correct assign-
100% DCM 13% MeOH DS ¼ 2.4 to DS ¼ 2.9 e
ment of an elution volume to DS from NMR is thus complicated. The
2636 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638
Fig. 3. Superimposed chromatograms of CAs having different DS in a multi-step
gradient (15 mL injection volume, 1.5 g/L of each sample). The dotted line represents
the eluent composition at the detector.
assignment of the peak maximum to DS obtained by NMR serves
therefore only as a first approximation.
Despite some scattering a clear decrease of DS with elution
volume is observed. Since the DPw and Ðm of the samples are very
similar (see Table 1) it can be concluded that a separation with
regard to DS is realized within the investigated DS-range for sam-
ples of identical DP.
To elucidate the molar mass contribution to the separation process
under these conditions, samples of similar DS but varying DP are
required. However, such samples were not available. In order to get at
least some indication on the influence of molar mass on the separa-
tion, the elution volumes for two data sets 1 (sample 3, sample 4) and
2 (sample 13, sample 14, sample 15) having nearly similar DS but
distinct difference in molar mass are inspected more closely. The
elution volumes at peak maximum (Vapex) for these samples deter-
mined from Fig. 4 are as follows: Vapex ¼ 36.2 and 35.6 mL for sample 3
(Mw ¼ 73,500 g/mol) and sample 4 (Mw ¼ 44,600 g/mol), respectively,
and Vapex ¼ 27.8, 27.0 and 26.8 mL for sample 13 (Mw ¼ 64,400 g/mol),
sample 14 (Mw ¼ 64,700 g/mol) and sample 15 (Mw ¼ 93,300 g/mol),
respectively. For samples of identical composition (DS) but varying
molar mass, an increase in elution volume with increasing molar
Fig. 4. Chromatographic retention as a function of DS for the optimized gradient (-:
industrial samples, ,: samples synthesized and the precursor). The ellipses indicate
the samples having similar average DS but different molar masses.
mass, which vanishes at sufficient high molar masses, would be ex-
pected. However, this cannot be conclusively observed for the two
data sets. Therefore we cannot finally conclude whether the molar
mass effect has already vanished, or not.
In order to calculate the DS-distribution from the chromato-
grams, the dependence of DS on elution volume from Fig. 4 was
fitted to the following equation:
 
DSi ¼ exp A þ B  Ve þ C  Ve2 (1)
where DSi is the DS at any elution volume of Ve and A, B and C the
adjustable parameters. Using the dependence of DS on elution
volume, the normalized DS-distribution, w(DS) was calculated from
the ELSD signal via:
S  DV
wðDSÞ ¼ (2)
DDS  SðS  DVÞ
where S is the ELSD signal which as a first approximation is
assumed to be proportional to concentration. DV is the volume
difference of two adjacent data points and DDS the difference in DS
of two adjacent data points.
However, the ELSD’s response was reported to be non-linear in
concentration and might depend on the nature of the mobile phase
and of the solute [49e51]. To investigate the effect of DS on ELSD-
response, identical amounts of the samples were injected without
column using pure DCM as eluent. The resulting peak areas are
plotted as a function of average DS in Fig. 5.
As can be seen, the peak area decreases by approximately 15%
over the DS-range from DS ¼ 1.5 to DS ¼ 2.9. When this procedure
was repeated for different concentrations, the same dependence of
peak area on DS was observed. However, the dependence of peak
area on concentration was linear until an injected mass of approx.
0.03 mg, corresponding to the highest injected mass used in the
experiments. Therefore, the dependence of peak area on concen-
tration can be assumed to be linear for the injected mass range used
in this investigation.
Since the overall change of detector response with DS is rather
low, the ELSD signals for the calculation of DS-distribution were
used without correction.
From equation (2) the DS-distribution for all of the CA samples
was determined and the results for the selected samples are rep-
resented in Fig. 6.
Fig. 5. Dependence of ELS peak area on DS (15 mL injection volume, 2.0 g/L of each
sample, DCM as eluent, no column used).
 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638 2637

Table 3
Weight average DS (DSw) calculated from DS-distribution and variance of the DS-
distribution peaks (the samples marked in grey are industrial samples).

Sample name 1
H NMR DS DSw s2
Sample 1 1.53 1.54 0.05
Sample 2 1.59 1.65 0.06
Sample 3 1.66 1.70 0.12
Sample 4 1.72 1.78 0.01
Sample 5 1.81 1.73 0.11
Sample 6 1.87 1.82 0.12
Sample 7 1.92 1.75 0.09
Sample 8 1.95 1.83 0.12
Sample 9 2.09 1.87 0.05
Sample 10 2.16 2.04 0.14
Sample 11 2.19 2.19 0.11
Sample 12 2.27 2.20 0.13
Sample 13 2.42 2.39 0.07
Sample 14 2.45 2.42 0.06
Sample 15 2.45 2.46 0.08
Sample 16 2.60 2.44 0.05
Sample 17 2.92 2.88 0.0007

Fig. 6. DS-distribution of CAs of different DS calculated from equation (2).

therefore be concluded that the presented separation method can

Samples of different average DS possess different width, i.e. a
different degree of chemical heterogeneity, with the highest DS
sample (sample 17, DS ¼ 2.92) being fairly uniform in DS while the
sample of lowest DS (sample 1, DS ¼ 1.53) contains polymer chains
ranging from DS ¼ 0.8 to DS ¼ 1.9. It is easily understood that for
very high and very low average DS narrow DS-distributions are
expected, as DS ¼ 0 and DS ¼ 3 correspond to chemically homo-
geneous products. Thus, for a completely random substitution
process the dependence of peak widths on DS should run through a
maximum. However, from Fig. 6 no systematic increase in degree of
heterogeneity in going from the highest to the lowest DS can be
observed, indicating that for the samples in our investigation the
heterogeneities are influenced by other factors as well. This is also
reflected by pronounced tailing towards low DS for sample 1 and 3
and the bimodal DS-distribution of sample 9.
It should be noted though that the assignment of DS to elution
volume is reliable only for DS values larger than DS ¼ 1.53, i.e. at
elution volumes lower than 38 mL. For samples components
eluting outside this calibrated region the assigned DS values should
be regarded with caution.
An advantage of the chromatographic separation method as
compared to NMR is that chromatography does not allow only
calculating the weight average degree of substitution (DSw)
provide new insights into the heterogeneity of CA allowing corre-
lating process parameters with chemical heterogeneity and finally
application properties.
Although the present manuscript addresses the characterization of
CAs, gradient chromatography should also be applicable to the sepa-
ration of other cellulose derivatives or polysaccharides in general. It
should be mentioned however that other types of heterogeneity, e.g.
side chain length in hydroxyethyl celluloses or -starches or branching
in starch derivatives will add additional levels of complexity.
4. Conclusions
CAs of different DS in the DS-range DS ¼ 1.5e2.9 were suc-
cessfully separated using a normal-phase multi-step gradient from
pure DCM to MeOH. The separation is based on adsorptione
desorption mechanism rather than on precipitationeredissolution.
The developed method was applied to determine the 1st order
chemical heterogeneity (DS-distribution) and the variance of the
DS-distribution of CA samples. Significant differences in the
chemical heterogeneity were observed even for samples of com-
parable average DS. These data can further be used to distinguish
samples of different origins or to study the effect of synthesis
conditions on the chemical heterogeneity.

SwðDSÞ  DDS  DS Acknowledgement

DSw ¼ (3)
SwðDSÞ  DDS
Hewa Othman Ghareeb would like to acknowledge the Deutscher
but the DS variance (s2) as well Akademischer Austauschdienst (DAAD) for providing a PhD schol-
arship. The allocation of samples by Rhodia Acetow GmbH is grate-
SwðDSÞ  DDS  ðDS  DSw Þ2 fully acknowledged.
s2 ¼ (4)
SwðDSÞ  DDS
References
The variance gives a quantitative measure of the width of the
DS-distribution. The calculated values for both the DSw and the [1] Scandola M, Ceccorulli G. Polymer 1985;26(13):1953e7.
variances of the CAs are given in Table 3. [2] Buchanan CM, Edgar KJ, Wilson AK. Macromolecules 1991;24(11):3060e4.
[3] Clasen C, Kulicke W-M. Progress in Polymer Science 2001;26(9):1839e919.
As expected, the DS values determined by NMR and DSw values [4] Rinaudo M. Biomacromolecules 2004;5(4):1155e65.
calculated from DS-distribution are quite close to each other. The [5] Bonet M, Quijada C, Muñoz S, Cases F. Journal of Adhesion Science and
large difference in the value of variance for the highest and lowest Technology 2005;19(2):95e108.
[6] Viridén A, Wittgren B, Andersson T, Larsson A. European Journal of Pharma-
DS confirms the observation of the difference in degree of hetero-
ceutical Sciences 2009;36(4e5):392e400.
geneity between these two samples. As can be seen the majority of [7] Viridén A, Larsson A, Schagerlöf H, Wittgren B. International Journal of
our synthesized samples have larger variances compared with the Pharmaceutics 2010;401(1e2):60e7.
industrial ones, indicating that our samples are more heteroge- [8] Tezuka Y, Tsuchiya Y. Carbohydrate Research 1995;273(1):83e91.
[9] Kamide K, Okajima K. Polymer Journal 1981;13(2):127e33.
neous in DS than the industrial ones. This might be due to a dif- [10] Sei T, Ishitani K, Suzuki R, Ikematsu K. Polymer Journal 1985;17(9):1065e9.
ference in the synthetic route or the process parameters. It can [11] Kowsaka K, Okajima K, Kamide K. Polymer Journal 1986;18(11):843e9.
2638 H.O. Ghareeb, W. Radke / Polymer 54 (2013) 2632e2638
[12] Takahashi S-I, Fujimoto T, Barua BM, Miyamoto T, Inagaki H. Journal of
Polymer Science Part A: Polymer Chemistry Edition 1986;24(11):2981e93.
[13] Tezuka Y, Imai K, Oshima M, Chiba T. Macromolecules 1987;20(10):2413e8.
[14] Kowsaka K, Okajima K, Kamide K. Polymer Journal 1988;20(10):827e36.
[15] Nehls I, Philipp B, Wagenknecht W, Klemm D, Schnabelrauch M, Stein A, et al.
Papier 1990;44(12):633e40.
[16] Tezuka Y. Carbohydrate Research 1993;241:285e90.
[17] Ho FF-L, Klosiewicz DW. Analytical Chemistry 1980;52(6):913e6.
[18] Reuben J, Conner HT. Carbohydrate Research 1983;115:1e13.
[19] Reuben J. Carbohydrate Research 1986;157:201e13.
[20] Verraest DL, Peters JA, Kuzee HC, Raaijmakers HWC, van Bekkum H. Carbo-
hydrate Research 1997;302(3e4):203e12.
[21] Erler U, Mischnick P, Stein A, Klemm D. Polymer Bulletin 1992;29(3e4):349e56.
[22] Heinze T, Erler U, Nehls I, Klemm D. Die Angewandte Makromolekulare
Chemie 1994;215(1):93e106.
[23] Richardson S, Lundqvist J, Wittgren B, Tjerneld F, Gorton L. Bio-
macromolecules 2002;3(6):1359e63.
[24] Horner S, Puls J, Saake B, Klohr E-A, Thielking H. Carbohydrate Polymers
1999;40(1):1e7.
[25] Saake B, Lebioda S, Puls J. Holzforschung 2004;58(1):97e104.
[26] Richardson S, Andersson T, Brinkmalm G, Wittgren B. Analytical Chemistry
2003;75(22):6077e83.
[27] Kragten EA, Kamerling JP, Vliegenthart JFG. Journal of Chromatography
1992;623(1):49e53.
[28] Cohen A, Schagerlöf H, Nilsson C, Melander C, Tjerneld F, Gorton L. Journal of
Chromatography A 2004;1029(1e2):87e95.
[29] Adden R, Müller R, Mischnick P. Cellulose 2006;13(4):459e76.
[30] Mischnick P. Journal of Carbohydrate Chemistry 1991;10(4):711e22.
[31] Cuers J, Unterieser I, Burchard W, Adden R, Rinken M, Mischnick P. Carbo-
hydrate Research 2012;348:55e63.
[32] Adden R, Niedner W, Müller R, Mischnick P. Analytical Chemistry 2006;78(4):
1146e57.
[33] Schagerlöf H, Richardson S, Momcilovic D, Brinkmalm G, Wittgren B,
Tjerneld F. Biomacromolecules 2006;7(1):80e5.
[34] Bashir S, Critchley P, Derrick PJ. Cellulose 2001;8(1):81e9.
[35] Arisz PW, Kauw HJJ, Boon JJ. Carbohydrate Research 1995;271(1):1e14.
[36] Glöckner G. Gradient HPLC of copolymers and chromatographic cross-frac-
tionation. Berlin: Springer-Verlag; 1991.
[37] Pasch H, Trathnigg B. HPLC of polymers. Berlin: Springer; 1997.
[38] Fischer S, Thümmler K, Volkert B, Hettrich K, Schmidt I, Fischer K. Macro-
molecular Symposia 2008;262:89e96.
[39] Mardles EWJ. Journal of the Chemical Society, Transactions 1923;123:1951e7.
[40] McNally JG, Godbout AP. Journal of the American Chemical Society
1929;51(10):3095e101.
[41] Herzog RO, Deripasko A. Cellulosechemie 1932;13:25e31.
[42] Sookne AM, Rutherford HA, Mark H, Harris M. Journal of Research of the
National Bureau of Standards 1942;29:123e30.
[43] Rosenthal AJ, White BB. Journal of Industrial and Engineering Chemistry
1952;44:2693e6.
[44] Kamide K, Manabe S-I, Osafune E. Die Makromolekulare Chemie 1973;168:
173e93.
[45] Kamide K, Matsui T, Okajima K, Manabe S-I. Cellulose Chemistry and Tech-
nology 1982;16(6):601e14.
[46] Floyd TR. Journal of Chromatography 1993;629(2):243e54.
[47] Asai T, Shimamoto S, Shibata T, Kawai T, Teramachi S. Macromolecular
Symposia 2006;242(1):5e12.
[48] Ghareeb HO, Malz F, Kilz P, Radke W. Carbohydrate Polymers 2012;88(1):
96e102.
[49] Righezza M, Guiochon G. Journal of Liquid Chromatography 1988;11(9e10):
1967e2004.
[50] Guiochon G, Moysan A, Holley C. Journal of Liquid Chromatography
1988;11(12):2547e70.
[51] Righezza M, Guiochon G. Journal of Liquid Chromatography 1988;11(13):
2709e29.