3 'MODEL-FREE' RIS

STATISTICAL WEIGHT
PARAMETERS FROM
13
C NMR DATA
O. W. HOWARTH, R. N. IBBETT, L R. HERBERT
and M. A. WHISKENS
Centre for Nuclear Magnetic Resonance, Department of Chemistry, University of
Warwick, Coventry CV4 IAL, UK

3.1 INTRODUCTION
This chapter describes some new attempts to exploit the rich source of conforma-
tional data that is potentially available from the dependence of 13C NMR shifts
upon the local tactic sequence in a substituted carbon-based polymer. The main
link between shift and conformation is the gamma-gauc/ie effect. This is widely
used in the interpretation of 13C chemical shifts in many types of compound. It
was initially identified from the classic work by Dalling and Grant [1], and by
others [2], on the 13C shifts of hydrocarbons, and especially of methylcyc-
lohexanes, in which specific local conformations could be frozen out by either
chemical or physical means. If a molecule contains any partial carbon chain
C1-C2-C3-C4, and if Cl and C4 are mutually y-gauche, thus making the
C1,2-C3,4 dihedral angle approximately ± 60°, then both C1 and C4 experience
an additional shielding, i.e. lowering of <5C, of %5ppm (see Figure 3.1). Similar
shifts are observed when CA is replaced by a wide range of groups [3]. The shift
effects of different groups do not vary in any obvious way with polarity. Indeed,
the shifts that arise from, e.g., hydroxyl substituents may be somewhat smaller
than average. However, they do seem to bear some relationship to the bulk of the
substituent, and indeed they are often interpreted in terms of steric compression.
Various more serious attempts have been made to explain the y-gauche effect
by MO computations. The most recent are those of Barfield and Yamamura [4],
using an ab initio method with a double-C basis set. Although this level of
computation is a little less than that required to give the best available absolute
shieldings, it reproduces very adequately a wide range of steric shifts in hydrocar-
bons (i.e. those dependent solely upon variations in stereochemistry) and also
Polymer Spectroscopy. Edited by Allan H. Fawcett
© 1996 John Wiley & Sons Ltd
 Y - gauche interaction (two of the four present)

syn - axial interaction

"P - gaucheT shifts possible at these atoms,

because of the A C y - gauche interaction

Figure 3.1 Illustration of stereochemical relationships. In this rotational state, A and

B are both y-gauche to the next-nearest non-methylene carbon, and the two A groups are
syn-axial to each other, as are the two B groups

supports the intuitive assumption that y-gauche contributions to the shift of

a given carbon from more than one source are additive. It also draws attention to
several other potential steric contributions to carbon shifts, including those that
arise from quite modest deviations from dihedral angles of ± 60°. These will be
discussed in a later section.
Shifts that arise from the y-gauche relationship may be readily computed for
polymers, provided that the appropriate RIS (rotational isomeric state) probabil-
ity parameters [5] are known from calculations of the potential energy surface of
the macromolecule. The basic procedure is to start with an estimate, e.g. from
molecular modelling calculations, of the free energies and bond angles of the
various energy minima found upon rotation of the two (usually) C—C bonds in
both the meso (m) and the racemic (r) dyad units of the polymer chain. Neigh-
bouring dyads are then permitted to alter these free energies through the steric
restrictions, such as the 1,5 pentane effect, that must apply to the dyad joins.
Although these effects extend in theory to an infinite succession of next-neigh-
bours, the calculated sequence dependence of shifts is usually small beyond the
heptad/hexad level, at most. Thus one obtains an estimate of the true probabili-
ties for occupation of any dyad in any steric sequence. It is then simple to explore
the effects of conformation upon the observed chemical shifts, using, e.g., the
y-gauche shift model.
Such calculations are described in detail by Tonelli in Chapter 2. They work
spectacularly well for polypropylene, with a model having five rotameric states
per chain bond. However, Tonelli has pointed out additional factors that must be
included when other polymers are considered. One is that the methylene shifts in,
e.g., monohalo vinyl polymers, unlike the methine shifts, show some dependence
upon solvent [6]. Another is that no conceivable RIS parameters can explain the
rather large steric shifts that distinguish isotactic from syndiotactic poly(methyl
methacrylate), either in the solution state or, even more so, in the solid state [T].
These can be up to lOppm in magnitude.
A further conceptual challenge arises intuitively. It is reasonable that the shifts
at a given sidechain carbon should depend not only upon the proximity of
y-gauche atoms, but also upon the identity of the atom or group which will
necessarily be syn-axial to it, i.e. its even closer neighbour, in any vinyl polymer.
This relationship is illustrated in Figure 3.1.
There is also a practical consideration. RIS parameters are not at all easy to
calculate in many cases, particularly for disubstituted vinyl polymers. In vinyl
polymers with main chain methine groups, a few of the dyad rotamers will often
have lower energies than the others, because they place both of the methine
protons approximately syn-axial to some other group or chain portion, and thus
minimise steric repulsions. However, this is not possible in disubstituted poly-
mers, and indeed even the gauche and trans RIS states themselves are likely to
offer poor approximations to the real bond torsional angles. There is also some
concern that the contributions of minor conformers may be underestimated if the
model assumes unrealistically rigid elements. Furthermore, some polymers have
shift patterns for every carbon that depend noticeably upon solvent. Such
variations lie beyond the reach of conventional RIS calculations, although they
must obviously bear witness to interesting conformational changes caused by the
influence of the solvent on the potential energy surface of the macromolecule.
The present work is an attempt to address the above challenges by turning the
methodology around. Might it prove possible to obtain reasonable RIS statistical
weight parameters, of use to other modellers, by treating the carbon shifts as
given data and the RIS parameters as variables? Complete sets of such sequence-
dependent shifts have been fairly reliably assigned at the tetrad/pentad level for
many mono- and disubstituted vinyl polymers, and in some cases the spin-spin
coupling constants are also available as further constraints on any model. The
obvious attraction of this approach is that the RIS variables may be overdeter-
mined. Suppose that one allows only three rotameric states per chain bond, and
treats any possible sidechain rotamers as having a constant average contribution.
Even if the probability of each dyad rotamer is allowed to vary independently,
after due regard for symmetry, there are only 10 undetermined probability ratios
in a vinyl polymer. (Note that the ratio of the m- to the r-dyad probabilities is not
significant in the present context). To these 10 must be added a single variable,
such as the g*g*/g~g~ elements of Flory's Un, matrix [8], which controls the
extent by which 1,5 syn-axial chain segments are forbidden. This extent may not
be 100% after allowance for minor distortions of the intervening torsional angles,
especially when the substituent groups are small, and even more so if they are
capable of favourable interactions such as hydrogen bonding. To determine these
11 variables we typically have some 16-30 different steric shifts, or even more if
the assignment can be made at the hexad/heptad level. Thus the quality as well as
the existence of any shift fit might be used to assess the reliability of the
calculation.

3.2 METHODS
The first problem that arises with this approach is that the y-gauche contributions
from many groups are not known accurately. Some may be estimated from the
shifts of molecules is known conformations, e.g. substituted cyclohexanes, but
others remain to be established. An underlying principle of the present study is
that all such parameters be fixed for a given group, even if it appears in several
different polymers. Another principle is that all the carbon resonance patterns of
a given polymer must be fitted simultaneously. Indeed, as a further simplification,
we specify that the shift effects depend only on the substituent, and not on the type
of carbon (e.g. C, CH, CH 2 , CH 3 , CO, CN) under observation. Fortunately, the
numerical value of most shift parameters does not in practice greatly affect the
resulting RIS weightings, within a reasonable range. It is the shift patterns which
dominate the calculation.
The second problem is to meet the challenge of polymers such as poly(methyl
methacrylate), which fit poorly to any scheme based simply on y-gauche contribu-
tions to the methyl shift. Our proposal is that, for such groups, the y-gauche
contributions should be subdivided into three separate syn-axial contributions
from the appropriate groups S to the carbon under observation. Thus, an
a-methyl group in poly(methyl methacrylate) may be y-gauche to the next-nearest
unprotonated carbon, but then it must also be sjw-axial either to another
a-methyl or to a carboxymethyl group or to the chain methylene in the next dyad.
These three will not necessarily cause the same shielding at the original methyl,
although their average shielding effect should not very greatly from a typical
value of, say, 5 ppm. One should note in this context that a hydrogen atom may be
an active substituent, because the shifting unit is in fact both the syn-axial H and
also the gauche C, even though we refer to the shielding parameter for brevity as
being simply 'syn-axiaF. Indeed, one may not even assume that H necessarily
imposes a smaller 'syn-axiaF shielding than any other group. We show below that
this model is capable of allowing surprisingly large methyl shifts.
The next problem arises from Tonelli's observation that some methylene shifts
depend noticeably upon solvent. If these changes are not matched by correspond-
ing changes in other shift patterns, then they cannot be explained by RIS
weightings alone. Tonelli and Schilling have suggested [6] that they may arise
from differential solvation of m- and r-dyads, which should affect methylene shifts
much more that others. We accept this suggestion by creating one more adjust-
able parameter, a variable, solvent-induced shift difference between m- and
r-methylenes. In most cases this turns out to be small, but even so it turns out in
general to give noticeably improved fits.
Finally, one must decide on the size of the RIS dyad matrices. In this respect,
the present work is not altogether 'model-free'. The main problem is that each
gauche state, and perhaps also the trans state, probably exists with two variants
[5], so as to ease the concomitant syw-axial repulsions. In cases where detailed
calculations have been made [9], these variants typically shift the chain dihedral
angle from ± 60° to ± 50° and to ± 70°, with unequal weightings and with a very
low or non-existent intervening energy barrier. It would probably be ideal to use
a generalised 6 x 6 state model in order to allow for this. However, this would
generate an unacceptable number of variables. Instead, we have chosen a simple
3 x 3 state model. This implies the assumption that the probabilities of each
nearby pair of rotational states may reasonably be represented by their sum, and
thus that the shift contributions from each state are comparable. Barfield and
Yamamura [4] have indeed shown that the y-gauche effect varies monotonically
with dihedral angle in an approximately sinusoidal manner, so that the shift
contributions from the ±50° and from the ±70° angles should be not too
different. Although the resulting, simplified RIS parameters will be more artifi-
cial, they should retain their usefulness for most calculations simply because of
their derivation from experimental data.

3.3 SOMECALCULATIONDETAILS
A spreadsheet calculation (using Microsoft Excel version 3.0) was written to
convert the 3 x 3 m- and dyad probability matrices into predicted steric shifts, by
the usual methods of matrix multiplication, to find overall conformational
probabilities. Although the final calculations were only carried out at the
tetrad/pentad level, the calculations assumed further, flanking m-dyads in each
case in order to give more realistic weightings. The linking U i r matrix was
assumed to have one freely variable component, g*g+/g~g~, with the other
elements being fixed at unity. This has the effect of permitting all dyad joins
equally, except for those which result in syw-axial chain segments. The latter are
suppressed in proportion to the variable component. In fact, none of the fitted
values of g*g+/g~g~ exceeds the necessary bounds of zero to approximately
unity.
As described above, a further shift variable was then introduced, to give the
same shift increment to each r-methylene resonance. These increments were never
large, and typically were less than ±0.4ppm. The resulting shift patterns were
compared with the experimental data, which are described in detail below, and
the squared shift differences were summed. In some cases one resonance of
a polymer shows little spread of shifts, and so cannot be reliably assigned, and
here the calculated shifts were constrained within limits set by the overall
linewidth of the data, without regard to their detailed assignment. Such data are
represented in the appropriate figures by single, broad Lorentzian peaks.
The sums of squared differences were then minimized with respect to the above
variables. Minimisation was carried out by the Newton-Raphson method using
Microsoft Solver version 3.0. It was almost invariably well-behaved, and for each
polymer it converged to give essentially the same results from a wide range of
starting parameters. Various calculational refinements were also tested in the
hope of improved fits, but, interestingly, they had little extra success. These
included attempts to reduce the g~g+/g*g~ elements in U n , to values below
unity. This gave good results only with polystyrene. The calculations also permit
an estimate of the reliability of individual parameters. In some cases, the data
closely defined all the parameters. In others, a few parameters, noted below, rose
to unreasonably high values in the final fits, such as y-gauche shifts > lOppm.
However, these could be reduced in all cases to reasonable values without a loss
of fitting accuracy greater than the experimental errors of shift measurement. The
RIS elements were not permitted to become negative. Apart from this, the
calculations were deliberately kept as simple as possible, because of the aim of
practical usefulness to modellers.

3.4 INDIVIDUALPOLYMERS
The method was first tested on polypropylene, because the shifts for this polymer
are reliably known as is its accessibility to shift calculations. Furthermore, both
the methyl and the chain methylene groups have been shown to give (and receive)
y-gauche shifts of close to —5.0 ppm [9]. This usefully reduced the number of
variables in the calculation, though in fact little change was observed even if they
were allowed to float. The resulting fits, using our simplified 3 x 3 RIS approxi-
mation, were not quite as good as those obtained by Schilling and Tonelli [9],
especially in that they slightly underestimated the shift contributions from the
more distant chain groups. However, the fit was still very good, as is shown in
Table 3.1. The syn-axial shift from H was initially permitted to float, but in
a variety of calculations it always came close to —4.0 ppm, and so it was fixed
thereafter at this value.
The next polymer to be tested was poly(acrylonitrile), which is readily available
with minor known variations of tacticity within the approximately atactic
range. Full shift data for this have not been published and are therefore pre-
sented in Table 3.2, using both dimethyl sulphoxide-d6 (DMSO) and water-
d 2 /Na + [SCN] " solvents. Our assignment, based on integration rather than peak
height and using several minor variations in tacticity, reverses a previous
assignment [10] for the CN mrrm and rrrr resonances. The differential effects of
distant groups are small and so the assignments are straightforward, except for
the unresolvable methylene resonances. Each set of RIS calculations led to a CN
 13
Table 3.1 Relative C shifts in polypropylene
CH^ CH^ CH
0
obs. calc. obs. calc. obs. calc.
mmmm 1.99 1.94 mmm 0.79 0.81 mmmm 0.50 0.51
mmmr 1.73 1.72 mmr 1.35 1.37 mmmr 0.45 0.51
rmmr 1.46 1.50 rmr 2.07 2.03 rmmr 0.40 0.51
mmrm 1.10 1.10 mrm 0 0 mmrm 0.17 0.22
rmrm 0.81 0.84 mrr 0.88 0.93 rmrm 0.14 0.23
mmrr 1.27 1.25 rr 1.81 1.81 mrnrr 0.28 0.25
rmrr 0.96 1.00 rmrr 0.23 0.26
mrrm 0 0 mrrrn 0 0
mrrr 0.27 0.23 mrrr 0.04 0.03
rrrr 0.49 0.43 rrrr 0.08 0.05
'y-gauche/d sjw-axial parameters: CH 2 , CH 3 - 5.0ppm, (C)H -4.0ppm.
RMS error 0.039 ppm.

13
Table 3.2 Observed C shifts/ppm for poly(acrylonitrile)
CN cii
DMSO water/NaSCN DMSO water/NaSCN
mmmm 119.73 122.8 mm 26.50 29.40
mmmr 119.66 122.7 mr 27.04 29.77
nnmr 119.59 122.5 rr 27.47 29.85
mmrm 119.36° 122.4
rmrm 119.29 122.3 CH 2
mmrr 119.45 122.5 «32.6 «34.3
rmrr 119.36 122.4 both almost unresolved
mrrm 119.00° 122.0
mrrr 119.09° 122.1
rrrr 119.18 122.2
"Some further fine structure is visible.

and syn-axial S shift parameter of —2.84 ppm. The fits are fairly good, although
they underrepresent the fine structure in the CN region. The favoured conforma-
tions are in reasonable agreement with those deduced by Ganster and Lochmann
[11] for 2,4-dicyanopentane in vacuo.
Similar calculations were then performed for poly(vinyl alcohol). They are
shown as sticks, below the aqueous spectrum, in Figure 3.2. Here the sample is
complicated by incomplete removal of acetate groups. However, the removal was
>90%, and so the dominant peaks in the spectrum are those of fully hydrolysed
regions. The effects of distant groups were once again small. The assignments in
DMSO are published [12]: those in water were assigned largely by analogy,
although the methylene data were sufficiently different to require some confirma-
tion via homo- and heteronuclear shift correlation experiments. The calculated
 CH CH2

Figure 3.2 100.6MHz 13C(1H) NMR spectrum of poly(vinyl alcohol); CH and CH2
regions. Calculated shifts (with atactic weightings) are shown as sticks below, in the same
order of assignment

RIS parameters (Table 3.3) show that H-bonding, even in water, is sufficient to
overcome some of the steric repulsions, including those associated with the 1,5
pentane effect, between 5yw-axial chain segments.
One possible flaw in the poly(vinyl alcohol) calculations is that the shifts might
also be directly and differentially affected by H-bonding. However, an indepen-
dent study of saccharides in water and in DMSO indicates that this is probably
not so. Another interesting feature is that, when the rotamers that were calculated
to be dominant were reset to be totally dominant, then the CH shift pattern was
 Table 3.3 Calculated RIS statistical weight parameters"
T tt tg'lg't tg+fg+t 9 9 g g+/g+g *v <5(CH2)/ppm
polypropylene 391K
m-dyad 0 0 1 0 0 0 0.22 + 0.07
r-dyad 0 0 0.08 0 0.94
poly(acrylonitrile) in DMSO 373 K
m-dyad 0.08 0 1 0 1.79 0 0.41 + 0.20
r-dyad 0 0 0.41 0.51 1.04
poly(acrylonitrile) in water 300K
m-dyad/NaSCN 0 0 1 0 0.65 0 0.40 -0.01
r-dyad 1 0 0 0.12 0.51 0.96
poly(vinyl alcohol) in water 295 K
m-dyad 1 0 0 0.40 0 7.51 0.78 + 0.32
r-dyad 0.22 1 0 1.55 0 0.93
poly(vinyl alcohol) in DMSO 295 K
m-dyad 0 0 0.69 0 0.23 0.75 -0.06
r-dyad 0.18 1 0 2.71 0 0
polystyrene 363 K
m-dyad 0 0 1 0.01 0 0 0* -0.77
r-dyad 0 0 0 0 2.91
poly(methylacrylonitrile) 363 K
m-dyad 0.01 1 0.28 0.39 0 0 0.07 -0.22
r-dyad 0 0.36 1.26 0 0
poly(methyl methacrylate) 300K
m-dyad 1.01 0.31 0 0 0.00 -0.19
r-dyad 0.02 0 0.35 0 0
poly(vinyl chloride) 393 K
m-dyad 0.04 0.41 1 0.37 0 0 0.00 + 1.12
r-dyad 1 0 0 0.16 0 0.87
a
g* and g defined as in Flory [5].
b
g~g+/g+g~ links across the chiral carbons were also completely forbidden in this case.
essentially magnified by a factor of « 3 . It then fitted quite well with the
solid-state data of Terao et al. [13], who obtain a CH shift spread of « 1 0 ppm.
Tonelli (Chapter 2) has interpreted solid-state polypropylene shifts in a similar
way, based on structures deduced by Bunn et al. [14].

Poly(acrylonrtrile) in DMSO

Poly(acrylonitrile) in water/NaSCN

Poly(vinyl alcohol) in water

Poly(vlnyl alcohol) In DMSO

Figure 3 3 {Continued)
 Polypropylene

Polystyrene

Poly(methyiacrylonrtrlle)

Poly(methyl methacrylate)

Figure 3.3 (a),(b) Upper traces: 13 C( 1 H) NMR spectra for eight different polymer
solutions, calculated at the tetrad/pentad level. Lower traces: observed shifts, simplified to
the same level, and further, to broad Lorentzian peaks of representative width, if overlaps
prevent assignment. The numbers are shifts in ppm
 The shift parameter for OH did not converge well, however, perhaps because of
the unusual number of accessible conformations and the presence of only two
types of carbon atom. Values ranging from O to — 5ppm gave acceptable
predictions of the observed shifts. To resolve this problem, we instead used an
independent and strongly convergent value of — 3.5 ppm, obtained from an
extensive study of hexopyranose sugars of known conformation in both water
and DMSO [15]. This value is lower than that for H, which may help to explain
the otherwise curious observations by Stothers and coworkers [16, 17] on
s>w-axial shifts in sterols and related compounds.
The other monosubstituted vinyl polymer to be studied in detail was polysty-
rene. Its spectrum is made complex in the methylene region by large shift
influences at the hexad/heptad level. These probably arise from the anisotropic
bulk of the phenyl rings. Our calculations are certainly naive in ignoring this
anisotropy. The methylene region may speculatively be divided into four broad
subsections, approximate assignments of which were checked by the classical
area method. Our assignments confirm those of Sato etal. [18]. The un-
protonated phenyl carbon shifts are more easily assigned by the same method.
The shift calculations were only successful if all gg links across the chiral carbons
were fully disallowed. They then gave a broadly reasonable but less than perfect
fit. Although this restrictive possibility was also investigated for other polymers,
it only gave useful results in this case. The phenyl shift constant converged well, to
—4.58 ppm.
The next polymer, poly(methylacrylonitrile), presented a more severe test in
that there were no remaining variable shift parameters, and also one further
resonance region to be fitted. An initial assignment was made following Inoue
et al. [19], but with a spectrum taken at higher frequency, and with the methylene
sub-areas rechecked. The fit is encouraging although not perfect.
Finally, it proved possible to obtain an acceptable fit for poly(methyl methac-
rylate) using a carbomethoxy shift parameter of —9.50 ppm. This shift was not
very well defined by the data, although it must certainly be large. (When it was
permitted to float freely in the minimisation, it drifted upward to ^ — 14 ppm,
but with only a very minor resulting improvement in the least-squares fit.) If the
conformations that we calculate to have highest probability are instead set to be
dominant, then the syndiotactic chain becomes all-trans, as observed in the solid
state by Speracek et al. [20], and a 4-8.14 ppm shift increment is predicted on
going from the syndiotactic to the isotactic polymer. The experimental figure is
between + 6.0 and + 8.45 ppm.
The fits are all presented in diagrammatic form in Figure 3.3(a) and (b).
Calculated relative shifts are plotted upright, with a standard Lorentzian
lineshape and with intensities that assume a strictly atactic sample. The vertically
inverted plots are idealisations of the experimental spectra. Broad Lorentzian
bands are used to depict the extent of any overlapped and hence unassigned
resonances. The order of assignment of the resolved peaks, calculated and
observed, is the same in every case expect for the fine structure in the COOMe
and CH 3 (mr) resonances of poly(methyl methacrylate). The same fits could not
be approximated for any of the polymers by any alternative and significantly
different set of minimised RIS parameters. Fits (not shown) have also been
successfully generated for poly(vinyl chloride). In this case the experimental steric
shifts for both carbon types varied by % 2 ppm. The worst methylene fitting error
was 0.11 ppm. The Cl shift parameter minimised to —4.01 ppm, and the m-
methylene shift factor was unusually large, at 1.12 ppm.

3-5 THE CALCULATED RIS PARAMETERS

The complete set of calculated RIS parameters is presented in Table 3.3, and
some of the conformations that are calculated to be dominant are depicted in
Figure 3.4. Those for polypropylene confirm the simple predictions that the
dominant conformations are the four where no carbon is syn-axial to any other
carbon, and that the methylene groups have a very similar bulk to the methyls.
However, those for poly(acrylonitrile) deviate markedly from this. Although it
remains true that the CN groups, like the methyls in polypropylene, resist being
mutually syw-axial, this may be seen from simple modelling to result from
unfavourable dipolar alignment rather than from steric interactions. Also, the
possibility exists that some other mutual orientations of CN groups may be
sufficiently favoured, for the same reason, to override concomitant but less
favourable steric interactions, e.g. between CN and the chain methylenes. Such
dipolar interactions are very hard to model independently because of the
difficulty of defining the local dielectric medium. It is interesting to note that, in
the unusually polar solvent D 2 O/Na + [SCN]", the RIS term tt(m) is decreased
yet further with respect to that in DMSO solvent. This term involves syw-axial
CN groups. Evidently their polar repulsion is increased by the very polar solvent,
perhaps because the cations polarise the CN groups further while being sterically
unable to interpose themselves.
The opposite behaviour is evident in poly(vinyl alcohol). The m-dyad shows
a marked preference for the tt and g~g~ conformations (see Figure 3.4) over the
more normal tg* and g+t. The tt and g~g" conformations both involve closeness
of the OH groups, created presumably by H-bonding, and sufficient to overcome
steric repulsions in the latter case. The same pattern holds for the r-dyad. The tg ~
and g~ t conformations are now the ones with syn-axial hydroxyls, and these are
again favoured along with the unexpected g~g~ conformer, possibly in a rather
distorted form so as to permit another H-bond. The only big effect of changing
the solvent from DMSO to water is a large increase in the g+g+(m) term. Perhaps
water is able to enhance the intramolecular H-bonding that is sterically possible
from this conformation.
 poly(vinyl alcohol)
poly(acrylonitrile) m-diad. tt rotamer
r-diad, tt rotamer

poly(vinyl alcohol)
poly(acrylonitrile)
r-diad, tg rotamer
m-diad, tg* rotamer

poly(acrylonitrile) poly(methyl methacrylate)

m-diad. gg* rotamer m-diad. tt rotamer

Figure 3.4 Some typical dominant dyad conformations

The calculations for polystyrene are discussed above, and should be regarded
with caution. They do, however, yield sensible RIS parameters, which are like
those for polypropylene, except for larger repulsions between phenyl and H than
between methyl and H. In contrast, poly(vinyl chloride) is intermediate between
polystyrene and polypropylene, although in this case a tendency is also evident
for neighbouring C-Cl bonds to lie approximately antiparallel, perhaps for
dipolar reasons.
 The parameters for poly(methylacrylonitrile) are less easy to anticipate, be-
cause most conformations will involve either steric or dipolar repulsions. The
most likely conformations will be g+f(m), tg+(m), tt(r) and g+g+(r), because these
alone allow the CN groups to lie syw-axial to other groups, and not to themselves.
The CN group has less steric bulk than, say, methyl. The calculations bear out
this prediction, although they do to some extent also permit one conformer,
g ~g ~(m), in which neighbouring CN groups are syn-axial.
Not surprisingly, for the reasons discussed above, the fit for poly(methyl
methacrylate) has the biggest error sum of the present series. Nevertheless, the fit
is acceptable given the approximations, which in this case include ignoring the
carboxymethyl orientations. Once again, there is a preference for the four
conformers as above, because the carboxymethyl group also has a low steric bulk
in some orientations. However, the fit also shows a marginal preference for the tt
state in the m dyad, consistently with the solid-state data noted above [20]. This
will no doubt be made possible by a favourable mutual orientation of the
carbonyl dipoles.

3.6 'P-GAUCHE9EFFECTS

The experimental evidence for y-gauche shifts [3] may also point to concomitant
shifts of the two carbons that connect the y-gauche groups, as marked in
Figure 3.1. These 'fi-gauche' shifts seem to be of the same sign as the y-gauche
ones, but of approximately half the magnitude. Unfortunately, because they are
smaller, they are less easy to quantify. We have experimented with the inclusion of
ft-gauche shifts in our model, with values half those for the corresponding
y-gauche shifts. They do not affect the qualitative conclusions above for any
polymer, and the fits are generally a little less satisfactory in each case. Thus,
although such shifts probably do occur, it is not necessary to include them at the
present level of approximation.

3.7 COUPLING CONSTANTS

In principle, it is also possible to use interpf oton coupling data in monovinyl
polymers to assess the conformational equilibria. For such calculations one needs
not only the weightings for the conformations but also their torsional angles, and
a reasonable estimate of the appropriate Karplus relationship [21] between the
HCCH dihedral angle and the corresponding coupling. In practice, it is not easy
to separate the peak splittings due to couplings from those arising from tacticity.
In some cases, and for some components, the separation can be achieved by a 2D
./-resolved spectrum [22]. In favourable case this procedure may also provide an
assignment, because the methylene protons in an r-dyad will be at least approxi-
 CH CH2

ppm
1
Figure 3.5 Partial 2D ./-resolved H NMR spectrum of poly(acrylonitrile) in DMSOd6. The multiplets are effectively rotated into the
vertical dimension. The large peak at 3.06 ppm arises from water, and the uneven ridge marked x is partly an artefact
 Table 3.4 Coupling constants
3
polypropylene J(obs)/Hz 7.0 and 6.5 (m) 7.0 (r)
3
J(calc)/Hz 7.34 and 7.25 (m) 7.13 (r)
3
poly(acrylonitrile) in DMSO J(obs)/Hz 8.4 and 6.0 (m) 7.5 (r)
V(calc)/Hz 8.16 and 5.85 (m) 7.32 (r)

mately isochronic. Figure 3.5 shows an example of such data for poly(acrylonit-
rile) in DMSO. The couplings are compared with calculated values in Table 3.4,
along with those for polypropylene, which have been deduced from the spectra of
oligomers. However, the calculated values are not very sensitive to conformation,
and their absolute values depend on what allowance one makes for libration
within each RI state when setting the t and g couplings for individual vicinal
proton pairs. The present study uses 12.0 Hz (t) and 2.0 Hz (#, polypropylene) or
4.0Hz (g, poly(acrylonitrile)).

3.8 CHARACTERISTICRATIOS
One may in principle check RIS calculations [5] by the measurement of
characteristic ratios <r 2 > 0 /n/ 2 , particularly as a function of tacticity. However,
such correlations present several problems. The published experimental data are
limited in extent and show significant scatter. Also, the calculations require
a good estimate of the chain torsion angles in each conformer. Nevertheless, some
comparisons are possible, because in vacuo modelling calculations are likely to
give reasonably accurate torsion angles even when their predictions of energy are
suspect. We have used our RIS data together with published torsion angles to
calculate the characteristic ratios for polypropylene at various temperatures in
the solution state. Table 3.5 shows that they agree well with the experimental data
[23]. Earlier calculations tended to fit poorly at the extremes of tacticity, perhaps
because they underestimated the contribution of minor conformations. The
calculations were performed using the Biosym RIS package, with a 200-unit
chain and randomised tacticity.

Table 3.5 Characteristic ratios for polypropylene

Calculated ratio Observed ratio Earlier estimates
syndiotactic 7.6(320K) 6.7(318K) 11.0(413K)
atactic 5.4(435K) 5.4(426K) 5.5(415K)
6.3(305K) 6.5(307K)
isotactic 5.3 (420 K) 5.0(456 K) 4.2 (413 K)
5.4(400K) 5.9(398K)
 syndiotactic atactic isotactic

Figure 3.6 Observed (squares) and calculated (lines) characteristic ratios for poly(methyl
methacrylate) at 333 K, plotted as a function of tacticity. The lines are based on RI states
with chain dihedral angles of (upper line) 10,20,100,115, - 1 0 0 and -115°, (central,
broken line) 10,20,100,125, - 1 0 0 and -125°, (lower, hatched line) 15,20,100,110, - 1 0 0
and-110°

Calculations were also performed for poly(methyl methacrylate). These were

more complex, because modelling shows each RI state to be markedly doubled
because of syw-axial interactions. Also, the calculated torsion angles were less
reliable than above. We approached this problem by dividing each RI state into
torsionally close pairs in the calculation, and assuming equal probabilities for
each member of a pair. In particular, this permitted some twist in a chain
dominated by tt states. Calculations are presented in Figure 3.6 for three
reasonable choices of torsional angle sets. The comparison with experiment [24]
is encouraging. However, characteristic ratios alone are not a sensitive test of RIS
weightings, for many combinations of weightings will give similar results.

3,9 CONCLUSIONS
Although there is no guarantee that our simple model will be capable of
predicting the shifts observed for any given polymer, the present fits are en-
couraging. They were undertaken, in part, to assess the validity of the y-gauche
shift model. It is likely that ab initio calculations will soon be able to refine this
model. Indeed, the work of Barfield and Yamamura [4] suggests that unusual
shifts may arise in rather specific conformations over a narrow range of torsion
angles. These may help to explain the rather unexpected shifts sometimes
observed in the solid state, but they are less likely to be a problem in the more
widely averaged local environments of the fluid state. One may therefore
reasonably propose that the present model generates relative RIS parameters for
the fluid state that are of genuine predictive value.

3.10 ACKNOWLEDGEMENT
We thank Dr. CJ. Samuel for many helpful discussions.

3.11 REFERENCES
[1] D. Dalling and D.M. Grant, J. Am. Chem. Soc, 1967,89, 6612.
[2] D.M. Grant and B.V. Cheyney, J. Am. Chem. Soc, 1967,89,5315.
[3] H.-J. Schneider and V. Hoppen, J. Org. Chem., 1978,43, 3866.
[4] M. Barfield and S.H. Yamamura, J. Am. Chem. Soc, 1990,112,4747.
[5] PJ. Flory, Macromolecules, 1974, 7, 381.
[6] A.E. Tonelli and F.C. Schilling, Ace Chem. Res., 1981,14, 235.
[7] A.E. Tonelli, Macromolecules, 1991, 24, 3065.
[8] PJ. Flory, P.R. Sundararajan and L.C. DeBoIt, J. Am. Chem. Soc, 1974,96, 5015.
[9] F.C. Schilling and A.E. Tonelli, Macromolecules, 1980,13, 270.
[10] J. Schaefer, Macromolecules, 1971,4,105.
[11] J. Ganster and J.R. Lochmann, Polymer, 1990,31,1159.
[12] T.K. Wu and M.L. Sheer, Macromolecules, 1977,10, 529.
[13] T. Terao, S. Maeda and A. Saika, Macromolecules, 1983,16,1535.
[14] A. Bunn, E.A. Cudby, R.K. Harris, KJ. Packer and BJ. Say, J. Chem. Soc,Chem.
Commun., 1981,15.
[15] P. Hobley and O.W. Howarth, to be published.
[16] W.A. Ayer, L.M. Browne, S. Fung and J.B. Stothers, Can. J. Chem., 1976,54, 3272.
[17] S.H. Grover, J.P. Guthrie, J.B. Stothers and CT. Tan, J. Magn. Reson., 1973,10,227.
[18] H. Sato, Y. Tanaka and K. Hatada, Makromol. Chem. Rapid. Commun., 1982,3,19.
[19] Y. Inoue, K. Koyama, R. Chujo and A. Nishioka, Makromol. Chem., 1974,175,277.
[20] J. Spevacek, B. Schneider and J. Straka, Macromolecules, 1990, 23, 3042.
[21] J. Kowalewski, Prog. NMR Spectrosc, 1977,11,1.
[22] W.P. Aue, E. Bartholdi and R.R. Ernst, J. Chem. Phys., 1976,65,4226.
[23] W.W. Suter and PJ. Flory, Macromolecules, 1975,8, 765.
[24] P.R. Sundararajan, Macromolecules, 1986,19, 415.