FOURTH INTERNATIONAL SYMPOSIUM ON WOOD AND PULPING CHEMISTRY

April, 27-30, 1987 Paris - Palais des Congres

VOLUME 1
ORAL PRESENTATIONS
COMPARISON OF CELLULOSE MOLECULAR WEIGHTS DETERMINED BY HIGH
PERFORMANCE SIZE EXCLUSION CHROMATOGHAgKY AND VISCOMETRY
ROBERT EVANS and ADRIAN F.ft. WALLIS
CSIRO DIVISION OF CHEMICAL AND WOOD TECHNOLOGY,
PRIVATE BAG 10,
CLAYTON, VICTORIA 3160,
AUSTRALIA
ABSTRACT
The molecular weight distribution (MWD) of cellulose samples was determined by high
performance size exclusion chrotnatography (HPSEC) of their tricarbanilate derivatives. The
HPSEC columns were calibrated with a series of tri-carbanilates whose weight average molecular
weights had been obtained by low angle laser light scattering (LALLS) photometry. Improved
procedures for the preparation of tricarbanilate derivatives are discussed. The equation
recommended in the standard method SCAN C15:G2 for the estimation of cellulose DP from
intrinsic viscosity in cupriethylenediamine, DP0.905 = 0.75 [] , is incorrect.
tentative replacement, DP0.85 = 1.1 [], based upon our results and those of others, is
suggested.
INTRODUCTION
The degree of polymerisation (DP) of cellulose samples is usually estimated from intrinsic
viscosities of solutions of cellulose in 0.5M cupriethylenediamine hydroxide (cuene)
(1,2,3) by application of the Mark-Houwink-Sakurada (MHS) equation [] = KMv , where
[] is the intrinsic viscosity, Mv is the viscosity average molecular weight and K and are
empirical constants. Although viscometry provides an indication of the DP of cellulose,
knowledge of the molecular weight distribution (MWD) of cellulose, conveniently determined
by high performance size exclusion chromatography (HPSEC) of solutions of cellulose
derivatives, is more useful, cellulose tricarbanilate (CTC) is a preferred derivative for this
purpose (4), as it may be prepared without degradation of the cellulose and gives a stable,
fully substituted product. However, some cellulose samples, e.g. regenerated cellulose,
require activation before they may be derivatised, and care must be taken to ensure that the CTC
samples reflect the MWD of the parent celluloses (5).
It is apparently not always appreciated that, for general applicability, the viscosity average
molecular weight (Mv ) must be used in the MHS correlation (6) . If the polymer standards
are almost monodisperse then Mn (number average) or preferably Mw (weight average) can be
uaed as an approximation to Mv If, however, only poly-disperse samples are available (as is
usually the case), Mv should be estimated, for example by means of polydis-persity correction
factors (7,8), before the MHS correlation is made. In some systems, for example cellulose
trinitrate (CTN)/acetone, - 1 and thus Mv - Mw (8)

For log-normal distributions with polydispersity about 2 and = 0.85, Mv is about 5% smaller
than Mn or about 90% greater than M Once the values of K and a have been correctly
determined the MHS equation can be used within the limitations of the concept to estimate the
viscosity average molecular weight for samples of arbitrary MWD.
During our work on the characterisation of cellulose samples we noted that the MHS parameters
suggested in the standard methods and elsewhere (1,3) for the cellulose/ cuene system had
been incorrectly determined. The recommended constants for this system were originally
derived by Immergut et al. (9) who used osmometry to measure the Dpn of CTN. The DP
calculated from this correlation, on the assumption that the result is a viscosity average, would
be too low by a factor of ca Mw/Mn (simple distributions). The polydispersity indices
(Mw /Mn ) of their samples were not reported but they may have been in the range 1.5-3.
To obtain a valid equation, the viscosity average molecular weight for each sample should have
been estimated using a measured polydispersity index and a reasonable approximation for a before
the correlation with intrinsic viscosity is made. Of the many published formulae, it is unfortunate
that the one chosen as an example in the standard methods is probably the most misleading.
Although others have commented on this matter (10), there is still evidence of misunderstanding
(11). In this paper we present the results of some of our work on the determination of cellulose
MWD.
EXPERIMENTAL
Cellulose samples used were bleached cotton linters with an a-cellulose content of 99.3%, a
selection of hydrolysed cotton lintera, Avicel PH101 and some regenerated celluloses. Cellulose
viscosities (Tappi pipette viscosities) were measured by the Tappi method T230 om-82.
Preparation of cellulose tricarbanilates
(a) Activation with ammonia/pyridine
some samples, e.g. regenerated celluloses, were
activated before derivati-sation. Dry cellulose (100 mg) was placed in a 50 mL reaction bottle
with liquid ammonia (10 mL) and pyridine (1 mL) under a blanket of dry nitrogen. The
ammonia was allowed to evaporate and, after ca 2 1, the residual ammonia was driven off by
gentle heating.
(b) Derivatisation
Dry cellulose (100 mg) , or cellulose activated as above, was treated in a
50 mL stoppered reaction bottle with phenylisocyanate (2 mL) in dry pyridine (20 mL) at
80C. After dissolution of the cellulose, the CTC was recovered by precipitation of the pyridine
solution in 200 mL saline aqueous methanol, prepared by mixing one part 0.02M aqueous sodium
sulphate solution with methanol to give 2 parts final mixture.
The derivative was collected by centrifugation, washed with the saline solution and dried. Nitrogen
contents were measured as 8.10 0.19 % (calc. for CTC : 8.09%).
HPSEC of CTC solutions
HPSEC was carried out on a system made up of a Waters WISP 710B autosampler and 481 UV
detector, and a Spectra-Physics SP8700 pump and SP4100 computing integrator using the
Spectra-Physics proprietary programs MWDCAL and MWDADP. Additional numerical analysis
was performed on an Olivetti M24 computer using programs written in DAOS, a data aquisition
and analysis operating system from Laboratory Software Associates, Melbourne. The columr.
system was a series of three Showdex crosslinked polystyrene columns KP-8Q6, -B05 and -804
with polystyrene exclusion limits 4 x 107 , 4 x 106 and 4 x 10 daltons respectively.
Stabilised tetrahydrofuran (THF) was delivered to the columns at a rate of 1,0 mL/min. The

injection volume was 40 PL of a 0.02% solution of CTC in THF containing 0.05% toluene as a
reference marker. Detection was by UV at 235 nm.
Photometry
CTC samples chosen for use as calibration standards in HPSEC were characterised by low angle
laser light scattering (LALLS) photometry using a Chrdmatix KMX6 photometer in conjunction
with a Chromatix KMX16 differential refractometer. The solvent was stabilised, chromatography
grade THF. The 4.5-5.5 degree annulus and 1-5 field stop were used, dn/dc was determined
from three samples to be 0.165 0.002 cm3 g -1 . The second virial coefficient was ca 4.6 x 104
cm3 mol g -1 and appeared to decrease with increasing DP .
RESULTS AND DISCUSSION
Activation of cellulose samples
A significant part of our research requires the determination of the MWD of regenerated cellulose
samples. The published methods for direct preparation of the tri-carbanilate derivative have been
found to be ineffective with regenerated cellulose samples. Activation methods have been described
which can be used to overcome this difficulty: Sundquist and Rantanen (12) found that regenerated
cellulose could he activated by soaking in water for 24 hours followed by solvent exchanging with
dry pyridine. Any residual water present in the reaction mixture would produce almost 12 times its
mass of di-phenylurea. Cael et al. (11) prepared CTC from regenerated cellulose by conversion to
the methylol derivative (DMSO/paraformaldehyde), regeneration, washing, drying, and finally,
derivatisation in pyridine with phenylisocyanate.
These methods are, however, inconvenient for routine application. We therefore developed a very
simple ammonia/pyridins activation method which could be used for both cellulose I and cellulose
II. The reaction/ dissolution rate for both high and low molecular weight cellulose samples was
found to be substantially increased by the above activation procedure. No significant effect of
the activation treatment on the MWD (HPSEC) of CTC from cotton linters was observed.
Quantities of ammonia and pyridine are not critical; we found it convenient to use the volumes
stated above. Treatment time, including evaporation, was approximately two hours,
although a lev minutes exposure to the activation solution, followed by forced evaporation,
also gave satisfactory resulta. Residual ammonia, which would have reacted with phenylisocyanate to give phenylurea, was driven off by gentle heating. Liquid ammonia alone was not
effective as an activating agent for regenerated cellulose if driven off before the addition of
pyridine.
other activating agents such as DMSO in ammonia can be used but more care must be taken with
the DMSO system as we found that considerable molecular weight degradation of dissolved CTC
from cotton linters ocourred over 18 h at 80C in DMSO. DMSO has been used as a
derivatising medium for the tricarbanilation of cotton celluloses (13).
Isolation of tricarbanilates
Methanol alone has been used as a precipitation medium, but Sutter and Burohard (14) and, more
recently, Wood et al. (5) have pointed out that low DP CTC is soluble in methanol. The latter
advocate the use of 50/50 v/v aqueous methanol or, preferably, evaporation of the reaction mixture
to obtain material suitable for size exclusion chromatography.
We had also found methanol to be inadequate as a precipitation medium when attempting to recover
the tricarbanilate of an extensively hydrolysed sample of cellulose (Dpw ~ 20) . Although 50%
aqueoUB methanol, in large excess, precipitated all the derivative and dissolved the
methylcarbanilate (formed by reaction of excess phenylisocyanate with methanol) , diphenylurea

was co-precipitated. Pyridine increases the solvent power of the precipitation medium; therefore
care must be taken to control the pyridine/precipitant ratio. We are currently determining the best
water/methanol/pyridine combination for the recovery of low molecular weight CTC. A volume
ratio of 10:1 precipitation medium : reaction medium yields moat of the low molecular weight CTC
but vary little diphenylurea, as indicated in Figure 1 where the diphenylurea peak retention time is
close, to 34 min. The optimum ratio appears to be between 10:1 and 20:1. Of course, many other
systems would be of use in separating CTC from the other reaction products (e.g., acetone/water,
dioxane/ water, THF/water) . A more rigorous purification procedure is being developed using an
aqueous washing medium containing sufficient methanol to give similar differential solvency to that
of the precipitation medium. Temperature also affects the solubility of CTC oligomera in the
precipitation medium. The temperature should not be allowed to rise very far above ambient,
particularly during

Figure 1. Effect of precipitation/reaction medium volume ratio on MWD of low DP CTC.

Solid line: ratio 20/1; broken line: ratio 10/1.
While we agree with Wood et al. (5) , that evaporation of the reaction mixture is the only way
to recover all of the CTC oligomer, and is sufficient for HPSEC analysis, other procedures such as
viscometry, light scattering photometry and oamometry require physical isolation from
impurities.
A further problem encountered was that of the colloidal stability of precipitated low DP CTC.
Some colloidal low molecular weight samples were freeze/thaw stable in the precipitation
medium. This was overcome by the addition of a small amount of sodium sulphate to the
water (before mixing with methanol) such that the final salt concentration in the mixture was
about 0.01M. The most stable systems required several hours to coagulate.
HPSEC calibration
CTC samples covering a range of DP were chosen as calibration standards for HPSEC. These
comprised a cotton linters sample, ten hydrolysed cotton linters samples and an Avicel PH101
sample. The DPw range covered was approximately 350 to 5000 as determined by light
scattering photometry. The calibration curve was found to be adequately represented by a fifth
order polynomial which was approximated by a point-to-point calibration curve on the, SpectraPhysics computing integrator. Two general criteria were used to generate the calibration curve.
First, the DPw values obtained using the calibration had to agree with those determined by
LALLS photometry and, second, the number distribution (obtained by division of the mass
distribution by the calibration function) and the weighted mass distribution (obtained by

multiplication by the calibration, function) had to be of reasonable and consistent form. The
low DP end of the calibration curve was additionally checked using the retention time of
cellobiose carbanilate. The above calibration criteria were satisfied simultaneously by
computer-assisted iteration. The second criterion was found to be necessary because of the
lack of molecular weight information outside
the range of our standards. Although this procedure may seem to be too subjective, we found
that the final curve could be altered very little without generating obvious inconsistencies in the
shapes of the weighted differential distribution functions. Although an on-line LALLS detector
should make this calibration procedure unnecessary at high molecular weights, we recommend
that the number and weighted mass distributions always be inspected for anomalous behaviour at
the extremes. Errors and noise levels are relatively large in these areas and arc greatly
amplified by application of the calibration function. Although we have yet to include corrections
for axial dispersion, such corrections should have little effect on the calculated DPv , which is
used in the MHS correlation. Figure 2 compares DPw values obtained from LALLS and HPSEC.

Figure 2.
samples.

Comparison of_HPSEC and LALLS methods for determining DPw values of CTC

Universal calibration using published MHS parameters for polystyrene and CTC was found to be
unsatisfactory, particularly at the extremes of the calibration curve.
Cael et al. (11) recently used the MHS correlation to compare the viscosity/molecular weight
relationships of the tricarbanilates and trinitrates of cellulose. They found an apparent
discrepancy between the two correlations: at a given cuene viscosity, the DPw (their notation)
of the nitrate appeared to be very much lower than that of the tricarbanilate. Random chain
degradation during the nitration process was considered an unlikely explanation in view of the
consistency of the data with those from several investigations in other laboratories.
The MHS formula used by Cael et al. for CTN in acetone was that published by Sihtola et al. (3).
However, the latter authors had used the formula published by Immergut et al. (9). As
mentioned above, this formula was based on osmotic pressure data and is therefore an approximate
relationship between intrinsic viscosity and DP, not DPw (as assumed by Cael et al.). This
misinterpretation accounts for most of the apparent discrepancy.
Cael et al. compared their MHS correlation for CTC in THF (based on a universal calibration
procedure rather than on their measured intrinsic viscosities) with those obtained by Danhelka et_al.
(15), El Ashmawy et al. (16) and Valtasaari and Saarala (17). Their Figure 5 (11) depicts the various
data in relation to their predicted MHS plot. Although it appears that the predicted line is roughly
compatible with the published data, we make the following comments:
1. One set of data, that of E. Ashmawy et al. (16), we believe to be invalid because very impure
cellulose was used (-cellulose content 78%-80%) and because the degree of substitution (DS) of

their carbanilate was only 2.75. The authors had erroneously assumed that the DS of their
derivatives could be obtained from the relationship:
DS = 3N/8.09 where N is the percent nitrogen content of the derivative. The correct relationship is:
DS = 162N/ (1400-119N) . A low DS would, for example, tend to decrease chain extension and
thereby decrease intrinsic viscosity at a given DP. The average DS of their five samples was 2.61
0.23, giving an average repeating unit molecular mass of 473 instead of 519 (fully substituted
derivative) .
2. One of the two sets of data taken from Danhelka et al. (15) is not for CTC in THF but for CTC in
acetone.
Danhelka et al. did publish two sets of CTC/THF data but Cael et al. inadvertently plotted the
CTC/acetone data instead of the first set of CTC/THF data.
3. Valtasaari and Saarela (17) reported a set of results obtained privately from Burchard and Keilich
and used these data to calibrate their HPSEC by the 'universal' method. These results were not
included in the plot even though they were probably obtained with highly fractionated CTCS (14)
and, in any case, must be more comprehensive and accurate than the indirectly measured HPSEC
data. It is assumed that the intrinsic viscosity of CTC in THF at 20C is similar to that at 25C.

Figure 3. Mark-Houwink-Sakurada plot for CTC in THF.

Diamonds: DPv data from this work transformed using intrinsic viscosity data from ref. 11;
squares: Burchard and Keilich DPw data (17); circles: Danhelka et al. DPw data (15).
We have redrawn Figure 5 from reference (11) (see our Figure 3), including our own data
transformed by the relationship (11) between intrinsic viscosity of CTC in THF and Tappi pipette
viscosity (Figure 4) . More extensive results will be reported when the intrinsic viscosities of
our CTC samples have been directly determined.

Figure 4. Intrinsic viscosity of CTN in acetone (filled circles) and 1.75 x intrinsic viscosity
of CTC in the THF (open circles) vs Tappi pipette viscosity. Data from reference 11.
All the DP values shown in our Figure 3 approximate the viscosity average. The resulting MHS
parameters for CTC in THF are: = 0.S5 and K = 0.0047. The parameters suggested by Cael et al
. (11), i.e., = 0.84 and K = 0.0043, give considerably higher DP, even allowing for
possible differences between Dpw and DPv . This observation
is in agreement with that of Korner et al. (18) who used the parameters from reference (11) for a
univeral calibration procedure. They found that universal calibration gave much higher DPw values
than those obtained by LALLS photometry and by viscometry.
Cael et al. (11) have published intrinsic viscosity results for both CTC in THF and CTN in acetone.
Those data are shown in Figure 4 and indicate that the intrinsic viscosity of CTN in acetone is 1.75
times greater than that of CTC in THF. The ratio of the intrinsic viscosity of CTN in acetone to that
of cellulose in cuene has been given (19) as 1.83. From these two ratios and the above MHS
parameters for CTC in THF we can determine an approximate relationship between cuene intrinsic
viscosity and DPv :
DPv0.85 = 1.1[].
Corrections to this equation will be made when we have determined the cuene intrinsic viscoaities
of our samples.
CONCLUSION
Suppliers and consumers of cellulosic products have been using for many years a relationship
between intrinsic viscosity in cuene and DP which gives very misleading results. Of the many
published relationships, the one referred to in the standards is also the most incorrect and
should be replaced. A more accurate equation is suggested which is based on data from this
work and several other sources. Further adjustment of this equation requires direct correlation
of cuene intrinsic viscosity with DPv over a wide range.
REFERENCES
1. SCAN C15:62. Viscosity of cellulose in cupriethylene-diamine solution,
scandanavian Pulp, Paper and Board Testing Committee, Stockholm, Sweden (1962)
2. ASTM D1795-62 Standard test method for intrinsic viscosity of cellulose. American
Society for the Testing of Materials, Philadelphia, PA. (1984).
3. SIHTOLA, H., KYRKXUND, B., LAAMANSN, L. and PALBMIUS, I.
Comparison and conversion of viscosity and DP values determined by different methods. Pan.
Puu, 45, 225-232 (1963).
4. SCHROBDER, L.R. and HAIGH, F.C.
Cellulose and wood pulp polysaccharides.
Gel permeation ehramatographic analysis. Tappi, 62, 103-105 (1979).
5. WOOD, B.F., CONNER, A.H. and HILL, C.G. The effect of precipitation on the
molecular weight distribution of cellulose trlcarhanilate. J. Appl. polym.. Sci. , 31, 37033712 (1986).
6. FLORY, P.J. Principles of Polymer Chemistry, Cornell University Press, N.Y., (1953)
Chapt. VII, p.313.
7. BAREISS, R.B. Polymolecularity correction factors, in Polymer Handbook, (ed. J.
Brandrup and E.H. Immergut) . Part IV, p. 115 (1975) .
8. PEEBLES, L.H. Rates of polymerization and depolymeri-zation, average molecular
weights and molecular weight distribution of polymers, in Polymer Handbook, (ed.
J. Brandrup and E.H. Immergut). Part II, p.405 [1975).

9. IMKERGOT, C.H., SCHURZ, J. and MARK, H. viskoeitats-zahl-MolekulargewichtsBeziehung fur Cellulose und Untersuchungen von Nitrocellulose in verschiedenen Losungsmitteln.
Monatsh. chem., 84, 219-249 (1953).
10. HOLTZER, A.M., BENOIT, H. and EOTY, P. The molecular configuration and
hydrodynamic behavior of cellulose trinitrate. J. Phys. Chem., 58, 624-634 (1954).
11. CAEL, J.J., CIETEK, D.J. and KOLPAK, F.J. Application of GPC/LALLS to
cellulose research. J. Appl. Folym. Sci., Appl. Polyp. Symp., 37, 509-529 (1983).
12. SUNDQUIST, J. and RANTANEN, T. characterisation of technical pulps by
carbanilation of cellulose. Pap. Puu, 65, 733-737 (1983).
13. LAURIOL., J.M., FROMENT, P., PLA, F. and ROBERT, A. E"tud.e de la
polymolecularite de celluloses degradees - cas des hydrocellulasee. Collogue Sciences et
Industries du Bois, Grenoble, Group de Travail No. 3, 71-82 (1962).
14. SDTTER, W. and BURCHARD, W. Comparative 3tudy of the hydrodynamic properties
of cellulose and amylose tricarbanilateB in dilute solutions. MaJtromol. Chem., 179, 19611980 (1978).
15. DANHELKA, J., KOSSLER, I. and B0HACK0VA, V. Determination of molecular
weight distribution of cellulose by conversion into tricarbanilate and fractionation. J. Polym.
Sci., Polym. Chem. Ed., 14, 287-298 (1976).
IS. El ASHMAWY, A.E., DANHELKA, J. ar.d KOSSLER, I.
Determination of molecular weight distribution of callulosic pulps by conversion into
tricarbanilate, elution fractionation and GPC. Sven. Papperstidn. , 77, 603-^608 (1974) .
17. VALTASAAKI, L. and SAARELA, K. Determination of chain length distribution of
cellulose by gel permeation chromatography using the tricarbanilate derivative. Pap. Puu, 57, 510 (1975).
18. KORNER, H.U., GOTTSCHALK, D. and PULS, J. Anwendbar-keit der GelPermeations - chromatographie zur BaBtimmung der Molekulargewichtsverteilung und des
Polymerisatlongrades von cellulosen. Papier (Darmstadt), 38, 255-261 (1984).
19. MARX-FIGIKI, M. and SCHDLZ, G.V. Die viskosimetrische
Molekulargewichtsbestimmung von Cellulosen und Cellulusenitraton unter Standardbedingungen .
Hakromol. Chem., 54, 102-118 (1962).